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 |
|---|---|---|---|---|---|
Synthesis of mRNA in eukaryotes involves the coordinated action of many enzymatic processes , including initiation , elongation , splicing , and cleavage . Kinetic competition between these processes has been proposed to determine RNA fate , yet such coupling has never been observed in vivo on single transcripts . In this study , we use dual-color single-molecule RNA imaging in living human cells to construct a complete kinetic profile of transcription and splicing of the β-globin gene . We find that kinetic competition results in multiple competing pathways for pre-mRNA splicing . Splicing of the terminal intron occurs stochastically both before and after transcript release , indicating there is not a strict quality control checkpoint . The majority of pre-mRNAs are spliced after release , while diffusing away from the site of transcription . A single missense point mutation ( S34F ) in the essential splicing factor U2AF1 which occurs in human cancers perturbs this kinetic balance and defers splicing to occur entirely post-release .
Co-transcriptional processing of nascent pre-mRNA is a central mechanism for gene regulation in eukaryotes and requires temporal coordination between transcription initiation , elongation , splicing , and cleavage . Each of these processes is carried out by megadalton macromolecular complexes acting at a single genetic locus , and kinetic competition between these processes has been proposed to determine RNA fate ( Bentley , 2014 ) . Genome-wide studies across organisms indicate heterogeneous distributions of both RNA polymerase and nascent RNA along the gene , suggesting that kinetic checkpoints exist throughout the gene , including at promoter-proximal sites , translation start sites , intron–exon boundaries , and at the 3′ end of genes ( Core et al . , 2008; Nechaev et al . , 2010; Churchman and Weissman , 2011; Hah et al . , 2011; Larson et al . , 2014 ) . However , population studies reflect the balance of kinetic rates and are unable to resolve the multiple competing processes occurring at a single gene . Moreover , genome-wide measurements lack the time-resolution which might provide mechanistic clues about the underlying enzymatic processes . The hypothesis of kinetic competition is that a fast process will out-compete a process which may in fact be more energetically preferred . Kinetic competition during the transcription cycle has been shown to influence splice site selection during alternative splicing ( de la Mata et al . , 2003 ) , recruitment of factors which release promoter-proximal pausing ( Li et al . , 2013 ) , and even RNAi-mediated genome defense ( Dumesic et al . , 2013 ) . Since these processes occur within the dynamic milieu of the nucleus , the stochastic interactions between macromolecules may result in a range of possible outcomes for the nascent RNA . Stochastic RNA synthesis—the phenomenon whereby the inherently stochastic nature of bio-molecular encounters and reactions leads to a non-deterministic production of transcripts—has been directly visualized in multiple organisms ( Golding et al . , 2005; Chubb et al . , 2006; Yunger et al . , 2010; Larson et al . , 2011 , 2013 ) . Yet , stochastic RNA processing—the possibility that stochastic bio-molecular reactions might lead to non-deterministic pathways/outcomes in the making and maturation of an RNA—has never been directly observed , and the potential consequences for gene regulation are largely unexplored . Alternatively , regulatory checkpoints have been proposed which safeguard against such stochastic RNA processing events , providing a level of quality control . For example , the model of exon definition requires that splicing of the terminal intron relies on synergy between 3′ end formation , nascent RNA cleavage , and intron excision ( Berget et al . , 1977; Niwa et al . , 1990 ) . Similarly , multiple studies indicate an increased density of nascent RNA present at the 3′ end of genes or in the chromatin-bound fraction , suggesting that nascent RNA is retained at the site of transcription to ensure correct processing ( Glover-Cutter et al . , 2008; Brody et al . , 2011; Carrillo Oesterreich et al . , 2010; Bhatt et al . , 2012 ) . In both the competition model and the checkpoint model , kinetics plays a prominent role , but in the latter case , the cell has developed additional safeguard mechanisms . In this article , we use an in vivo single-molecule RNA imaging approach to directly measure kinetic coupling between transcription and splicing of a human β-globin reporter gene . The approach is based on simultaneous dual-color imaging of both the intron and exon of the same pre-mRNA using both PP7 and MS2 stem loops ( Bertrand et al . , 1998; Chao et al . , 2008 ) . We find that kinetic competition results in multiple competing pathways for pre-mRNA splicing . Splicing of the terminal intron occurs stochastically both before and after transcript release , indicating there is not a strict quality control checkpoint . Post-release splicing occurs from freely diffusing transcripts in the nucleus and is an order of magnitude faster than splicing at the site of transcription . A single missense mutation ( Ser34Phe ) in the zinc finger domain of the conserved splicing factor U2AF1 which is recurrent in multiple cancers ( Yoshida et al . , 2011; Graubert et al . , 2012; Waterfall et al . , 2014 ) changes the balance , making all splicing post-release . This same effect can also be observed on the endogenous , un-modified fragile X mental retardation mRNA ( FXR1 ) . Our results show that kinetic competition governs the stochastic balance between multiple competing pathways for RNA synthesis and processing and that this balance is perturbed by oncogenic mutations .
To visualize transcription , splicing , and release of single transcripts in living cells , we used time-lapse fluorescence microscopy of multiply labeled RNA ( Bertrand et al . , 1998 ) . We stably integrated into U2-OS cells , a human β-globin reporter with a DNA cassette that encodes for 24X PP7 RNA hairpins in the second intron ( Chao et al . , 2008 ) and a 24X MS2 hairpin cassette in the 3′ UTR ( Boireau et al . , 2007 ) ( Figure 1A ) . The constitutively expressed PP7-coat protein tagged with mCherry ( PCP-mCherry ) and MS2-coat protein tagged with GFP ( MCP-GFP ) bind with high affinity ( the on rate for MCP is 0 . 54 μM–1s–1 . At a nuclear concentration of 1 μM , the average time for the MCP to bind a completed stem loop is 1 . 85 s ) ( Buenrostro et al . , 2014 ) to the RNA stem loops as homodimers , tagging each cassette with 48 fluorophores of a single color , resulting in orthogonal labeling of two different parts of the nascent transcript ( Hocine et al . , 2013; Martin et al . , 2013; Buenrostro et al . , 2014 ) . Since the PP7 cassette is intronic , unspliced RNAs appear in both colors , while spliced RNAs are only visible in green . Time-lapse imaging of cells in 3D reveals a temporally fluctuating diffraction-limited spot , co-localizing in both colors , that corresponds to the transcription site ( TS ) where nascent transcripts are synthesized ( Figure 1B–C , Videos 1 and Videos 2 ) . We observed mature mRNA ( exon only ) diffusing in both the nucleus and the cytoplasm ( Video 3 ) , and we verified expression of the protein product , indicating the message is spliced and translated correctly ( Figure 1—figure supplement 1B–C ) . 10 . 7554/eLife . 03939 . 003Figure 1 . Real-time measurement of transcription and splicing in living cells . ( A ) Schematic of the human β-globin report gene construct . Reporter splicing efficiency >95% by qRT-PCR ( Figure 1—figure supplement 1C ) . ( B ) 3D images of diffraction-limited spot in both channels corresponding to the transcription site ( TS , arrow ) . Bar: 4 µm . ( C ) Fluorescence fluctuations recorded at the TS reflect stochastic transcriptional events . Dotted lines are background traces recorded in the nucleus , 8 µm away from the TS . ( D and E ) Examples of pre- and post-release splicing observed when the intron ( red signal ) disappears simultaneously with ( D ) or before ( E ) the exon ( green signal ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00310 . 7554/eLife . 03939 . 004Figure 1—figure supplement 1 . Human β-globin reporter gene . ( A ) Detail of the reporter construct ( See ‘Materials and methods’ ) . Top schematic drawn to scale . ( B ) Imaging of gene expression from nascent transcripts ( arrow ) to protein product ( blue channel shows CFP-SKL protein product of the reporter gene accumulating in peroxisomes ) . Images are maximum intensity projections of z-stacks ( Δz = 0 . 25 μm , exposure = 1 s ) . ( C ) Splicing efficiency and ( D ) poly ( A ) tail site/length of the β-globin reporter , measured in three conditions: mock-transfected cells and cells transfected with either wild type ( WT ) or mutant ( S34F ) splicing factor U2AF1 ( see Materials and methods ) . Error bars are SEM calculated over four measurements . ( E ) Expected fluorescence time profiles for a single transcript . When the PP7 cassette is transcribed , the red fluorescence signal increases progressively ( as RNA stem loops are formed ) and plateaus once the polymerase exits the cassette . The same applies to the green fluorescence signal when the PP7 cassette is transcribed . If splicing is post-release , red and green signals drop simultaneously when the unspliced RNA is released and diffuse away . If splicing is pre-release , the red fluorescence drops before the green fluorescence reflecting that intron removal occurred before the release of a spliced transcript . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00410 . 7554/eLife . 03939 . 005Figure 1—figure supplement 2 . Integration site of the β-globin reporter and copy number analysis . ( A ) Example of an integration site of the reporter plasmid as identified by whole genome sequencing . Reads aligning to both genomic and plasmid sequences are shown at the bottom . The alignments identify the genomic position of the insertion and the location of the breaks in the plasmid . The number of repeats of the plasmid at the insertion site cannot be known from sequencing data . ( B ) Three insertion sites were identified in the cell line . ( C ) Semi-quantitative PCR was used to confirm the insertion sites and to estimate the total copy number of the integrated plasmid construct in the cell line . ( D ) Quantification of the PCR products shows , as expected , that amplicons internal to the plasmid were more amplified than the amplicons at the junctions . ( E ) Calibration curves were made by amplifying varying amounts of G-block DNA carrying the same primer pairs as used in ( C ) . ( F ) Correcting the data in ( D ) with the calibration curves in ( E ) and taking the ratio of internal-to-junction PCR products yields a total copy number of 5 . 48 ± 1 . 47 . Error is SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00510 . 7554/eLife . 03939 . 006Video 1 . Time-fluctuating transcription sites . Cells show a diffraction-limited fluorescent spot colocalizing in both colors ( red: intron , green: exon ) , corresponding to the transcription site of the reporter gene . The fluorescence intensity of each site fluctuates over time as nascent transcripts are synthesized , spliced , and released from the transcription site . Large orange shapes in nuclei are nucleoli ( Ferguson and Larson , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00610 . 7554/eLife . 03939 . 007Video 2 . Tracking of a transcription site in 4D . The video shows , for the intron and exon signals ( left and center panels ) , the maximum intensity projected image from the top ( square image ) and from the sides ( rectangle images ) , revealing the transcription site ( TS ) in three dimensions ( 3D ) and over time ( 4D ) . Image analysis is used to track the TS over time in both colors . The blue box and cross indicate the location of the TS as found by the tracking algorithm . The right panel is the merge of both signals . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00710 . 7554/eLife . 03939 . 008Video 3 . Spliced RNAs diffusing in the nucleus and the cytoplasm . Cells are imaged here with a high laser power and a short exposure time so that diffusion of single RNAs can be appreciated . It reveals a population of transcripts diffusing in both the nucleus and the cytoplasm , as evidenced by fast fluctuations observed in the exon signal ( right panel ) . These transcripts are , for the most part , already spliced since the intron signal ( center panel ) does not show the same fluctuations . In these imaging conditions , unspliced transcripts are only visible at the transcription site ( bright spot in the nucleus colocalizing in both color; see merge in left panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 008 By simultaneously observing the fluorescence intensity of the intron and the exon of a single nascent transcript , one can determine when the intron is excised from the pre-mRNA . We find that both pre- and post-release splicing are visible as single events at the same gene over time ( Figure 1C ) . In the case of post-release splicing , the intronic fluorescence appears , followed by the exon fluorescence , followed by a coincidental drop in both colors reflecting the release of an unspliced RNA ( Figure 1D and Figure 1—figure supplement 1E ) . For pre-release splicing , there is a delay between the drop-off of the red and green signals , indicating intron removal before release ( Figure 1E ) . Diffusion of the pre-mRNA away from the TS is rapid , which accounts for the precipitous drop of the signal in the time trace ( Figure 1C ) . Most of the time , multiple nascent RNA is present at the TS , necessitating a general analysis method for extracting kinetic information from the time traces ( Larson et al . , 2011 ) . We developed a dual-color fluctuation correlation analysis approach for analyzing the complete transcription cycle , resulting in two temporal auto-correlation functions and a single temporal cross-correlation function ( Figure 2A; See ‘Materials and methods’ ) . These functions measure , over many traces , how a fluctuation in one fluorescence channel is statistically correlated with a fluctuation in either the same or the other channel after a given time delay . 10 . 7554/eLife . 03939 . 009Figure 2 . Transcription and splicing kinetics are revealed by fluctuation analysis of dual-color fluorescence intensity time traces . ( A ) Auto- and cross-correlation functions quantify statistically correlated fluctuations occurring at different time delays , respectively within the same or between two signals . ( B ) Correlation functions ( G ( τ ) ) of experimental time traces ( N = 21 ) . Auto-correlations ( red and green curves ) are symmetrical by construction . Cross-correlations ( blue and magenta curves ) are two halves of a single continuous curve . Inset: short-delay behavior of the cross-correlation reveals that 13 ± 5% of the RNAs are spliced pre-release ( p-value: pre-release fraction ≠ 0% and 100%; z-test ) . ( C ) Schematic representing stochastic pre- and post-release splicing . Purely pre-release splicing imposes the cross-correlation to have the same rising slope on both sides of the y-axis , while purely post-release makes the intron-to-exon cross-correlation ( blue curve , positive delay ) start as a plateau . The change of slope at τ = 0 delay is indicative of the fraction of splicing events occurring before release . ( D ) Spliceostatin A abolishes pre-release splicing . ( E ) Camptothecin delays the decay of the intron-to-exon cross-correlation and increases the pre-release fraction . All correlation functions are normalized by the value of the cross-correlation at 0 delay ( Grg ( 0 ) ) . Error: SEM ( bootstrap ) . Control correlation functions are shown in Figure 2—figure supplement 1G–H . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 00910 . 7554/eLife . 03939 . 010Figure 2—figure supplement 1 . Fluorescence time traces and correlation functions . ( A ) Three examples of dual-color fluorescence time traces recorded at TS ( left ) and the corresponding correlation functions ( right ) . 21 of such traces were used to compute the average correlation functions shown in ( B ) . ( B–F ) Correlation functions under different experimental conditions . Each panel shows crosscorrelations ( main graph ) , autocorrelations ( right inset ) and crosscorrelations magnified around t = 0 ( left inset ) . ( G ) For each transcription time trace , a background trace was measured in the nucleus at a constant offset from the TS ( 2 . 4 μm on average ) . Correlation functions from these traces reflect technical bias ( cell movement , coat protein diffusion , imaging or tracking artifacts ) and are mostly flat . ( H ) Cross-correlation function obtained after swapping the green channels between pairs of time traces ( shuffled data ) also reveals the absence of technical bias . NB: data in ( G ) is from a different cell line ( same reporter but different genomic insertion site ) . In all cases except ( G and H ) , all four correlation functions are normalized by Grg ( 0 ) . Errorbars: SEM ( bootstrap ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01010 . 7554/eLife . 03939 . 011Figure 2—figure supplement 2 . Geometry of the correlation functions . ( A ) If fluorescence profiles are approximated by step functions ( left ) , correlation functions are piecewise linear ( right ) . Autocorrelations decrease linearly and reach 0 at a delay equal to the persistence time of the red and green signals . Cross-correlation shows four angles , each reflecting the delay between a rise or fall in the red signal and a rise or fall in the green signal . ( B ) This can be generalized to the case where fluorescence signals rise as ramps ( spanning the width of each cassette ) . In this case , the sharp angles described above become smooth when one or two of the fluorescence transitions they involve is a ramp . ( C ) If the red fall precedes the green rise ( i . e . , splicing precedes transcription of MS2 cassette ) , crosscorrelation at τ = 0 is null because the two signals never overlap . If the red signal falls while the green signal is up ( i . e . , splicing occurs after the MS2 cassette but before release ) , crosscorrelation at τ = 0 is non-null and is increasing with the same slope on either side of the y-axis . Finally , if both signal fall at the same time ( i . e . , splicing succeeds or coincides with transcript release ) , the crosscorrelation shows a break of slope at τ = 0 , with a positive slope for τ < 0 and a null slope for τ > 0 . ( D ) The correlation functions originating from a heterogeneous ( stochastic ) population of transcripts is simply the average of the correlation functions for each transcript . Hence , all the pre-release splicing events contribute a positive slope on either site of the y-axis , while all the post-release splicing events only contribute a positive slope on the left side of the axis . The resulting crosscorrelation displays a change of slope that directly reflects the fraction of pre- and post-release splicing events . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01110 . 7554/eLife . 03939 . 012Figure 2—figure supplement 3 . Estimation of the fraction of pre-release splicing from slopes in the crosscorrelations . In panels ( A ) and ( B ) , we estimate the fraction of splicing events that occur pre-release from experimental correlation functions ( i . e . , Figure 2B ) , by applying the principle shown on Figure 2—figure supplement 2D . Straight lines are fit to the crosscorrelation function on either side of the y-axis ( A ) by performing a non-linear least-square fit on the derivatives ( B ) . Standard errors are obtained by bootstrapping the derivatives and only the darkened points are used in the fit . The pre-release fraction is obtained as the ratio of the two slopes . ( C ) Simulated data ( from models II . 4 and II . 6; see ‘Materials and methods’ and Supplementary file 2 ) were used to assess the accuracy of this method . The pre-release fraction effectively observed in the simulations is compared to the one obtained by fitting the slopes of the correlations functions . Each point corresponds to a set of simulations preformed with a given set of parameters . The amount of data used per point in the main graph is similar to the experimental data shown in ( A ) and is higher in the inset , yielding a more precise estimation . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01210 . 7554/eLife . 03939 . 013Figure 2—figure supplement 4 . Mechanistic schemes . This figure presents the five mechanistic schemes that are compared to our data . The central part of each panel depicts the fluorescence time profiles lined up with the reporter construct , indicating the names of the time distributions of the different sections of the profile ( i . e . , time the polymerase spends in each region of the construct ) . The right part of each panel indicates what is assumed in terms of kinetic relationship between splicing ( i . e . , intron removal ) , elongation ( i . e . , progression or pausing ) , and release ( i . e . , including some 3′ end processing time or retention on the chromatin ) . ( Scheme I ) Splicing never happens pre-release . ( Scheme II ) Once the 3′ splice site ( 3′ss ) is reached and as long as the RNA is as the transcription site , splicing is kinetically independent from all other processes such as elongation and 3′ end processing/retention . It means that affecting either process will change the balance of post- vs pre-release splicing . ( Scheme III ) Once the 3′ splice site is reached , an obligatory checkpoint forces the polymerase to pause until the splicing reaction completes . ( Scheme IV ) The transcript can only be spliced once the 3′ end of the gene is reached . Once the intron is removed , the RNA may be released after an additional processing/retention time . ( Scheme V ) Splicing may happen any time after the 3′ss is reached but a checkpoint mechanism ensures the RNA is spliced before its release . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01310 . 7554/eLife . 03939 . 014Figure 2—figure supplement 5 . Model comparison using a Bayesian Information Criterion ( BIC ) . ( A–F ) Examples of best fits ( lines ) onto experimental correlation functions ( circles; untreated , untransfected condition ) with 6 of the 21 competing models . Left and top right graphs: crosscorrelations; bottom right graph: autocorrelations . ( G–K ) All 21 models were compared on five experimental conditions , using the BIC score ( the lower the better , see Supplementary file 1—Appendix 3 ) . The number of parameters for each model is indicated on the right of panel ( G ) . The BIC accounts for the variable number of parameters . * Models II . 4 , IV . 4 , and V . 3 are the three models that we retain from our analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01410 . 7554/eLife . 03939 . 015Figure 2—figure supplement 6 . Counting transcripts at the transcription site . ( A–B ) The average intensity of the single-RNA particle diffusing in the nucleus detected in one channel of a confocal video ( e . g . , Figure 4A or Video 5 ) was used to normalize the intensities of all the spots found in that channel . Panels ( A ) and ( B ) show the distribution of normalized intensities of all the single RNAs ( centered around 1 ) as well as the transcription site ( TS ) for the red and the green channels respectively . This allows estimating the average number of RNAs at the TS that are labeled in red and green . ( C ) Repeating this analysis for multiple cells ( N = 9 ) and averaging shows that there are more red RNAs than green RNAs . ( D ) The average of the ratio between the number of red and green RNAs at the TS is very close to the expected value of 1 . 4 calculated from the fitting parameters shown in Table 1 . Errors: SEM over cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 015 The correlation functions for β-globin ( Figure 2B , Figure 2—figure supplement 1A–B ) reveal the kinetic features hidden in the fluorescence time traces and encapsulate the stochastic kinetics of single-transcript synthesis ( Figure 2—figure supplement 2 , Supplementary file 1—§2 ) . At short time delays ( <40 s ) , the cross-correlation reflects the order of splicing and release events ( Figure 2B–C and Supplementary file 1—§2 ) . If splicing never occurs before transcript release , the intron-to-exon cross-correlation ( blue ) starts horizontally at positive delays . On the other hand , if splicing always occurs before release , this function starts with a positive slope in alignment with its negative counterpart ( magenta ) . In the case where splicing is stochastic and both outcomes may occur , their relative probability of occurrence is given by the change of slope of the cross-correlation at 0 delay ( Figure 2C , Figure 2—figure supplements 2D and 3 ) . The experimental cross-correlation determined from ∼2000 individual β-globin transcripts indicates that splicing occurs before release for a fraction of transcripts ( 13 ± 5% , Figure 2B inset ) . We note that this change in slope at short time delay only denotes the relative order of splicing vs release but says nothing about the kinetics of these two processes . In summary , these data demonstrate that splicing and release are not firmly constrained to occur in a specific order ( p < 0 . 003 ) . At longer delays ( >40 s ) , other features of the transcription cycle are visible . For example , the delay at which the intron-to-exon cross-correlation ( Figure 2B , blue circles ) starts decreasing ( ∼60 s ) corresponds to the elongation time between the two cassettes ( 2573 bases apart ) , resulting in an elongation rate of ∼2 . 6 kb/min . Finally , the decay at time-scales > 100 s relates to the dwell time of transcripts at the TS ( Figure 2—figure supplement 2A , B ) . Specifically , the long decays observed in all correlation functions indicates that RNA is not immediately released after transcription of the poly ( A ) site . Rather , the transcript remains at the TS for a duration which reflects either a pause at/near the poly ( A ) site , transcription past the termination site , or a post-cleavage retention of the transcript within the diffraction-limited spot ( Hofer and Darnell , 1981; Glover-Cutter et al . , 2008; Brody et al . , 2011; Carrillo Oesterreich et al . , 2010; Bhatt et al . , 2012 ) . To confirm our assignment of transcription cycle events to features in the correlation curve , we treated cells with drugs known to affect different aspects of RNA synthesis . The splicing inhibitor spliceostatin A ( SSA ) ( Kaida et al . , 2007 ) abolished splicing at the TS as evidenced by the disappearance of the rise in the intron-to-exon correlation function ( Figure 2D ) . Treatment with camptothecin ( CPT , a topoisomerase I inhibitor known to slow down elongation [Singh and Padgett , 2009] ) resulted in a marked shift of the decreasing part of the intron-to-exon cross-correlation to longer delays ( Figure 2E ) , which is the expected manifestation of slower elongation ( Figure 2—figure supplement 2A–B , Supplementary file 1—§2 ) . As an additional control , when shuffling channels between traces or when using time traces recorded away from the TS , the correlation functions are flat ( Figure 2—figure supplement 1G–H ) , supporting the fact that the correlation functions shown on Figure 2B reflect the molecular events happening at the TS . Finally , we emphasize that an essential advantage of this approach is that correlation functions reveal single-transcript kinetics even from signals where multiple transcripts are present at any given time ( Supplementary file 1—§1 ) . As illustrated in Figure 3 , a single transcription and splicing event results in correlation functions with a peak near zero delay , reflecting the intra-transcript kinetics ( Figure 3A ) . If three transcripts are present , additional peaks appear at non-zero delay , due to inter-transcript kinetics but all the correlations resulting from intra-transcript kinetics accumulate around 0 ( Figure 3B ) . After averaging over many transcription and splicing events , inter-transcript correlations disappear , leaving only a central peak which reflects the kinetics of single transcripts ( Figure 3C ) . See also Video 4 and Figure 3—figure supplement 1 . 10 . 7554/eLife . 03939 . 016Figure 3 . Correlation functions reflect single-transcript kinetics . ( A ) A dual-color time trace with a single transcription event yields correlation functions with features around 0 delay and flat elsewhere . ( B ) When several transcription events are present in a time trace , the correlation coming from each individual RNA accumulates around 0 delay , while all the correlation between pairs of RNAs distributes uniformly on the delay axis . ( C ) When there are many transcription events per time trace and/or many traces are used to produce an average correlation function , the correlation from single transcripts dominates and that from pairs of transcripts averages out . The resulting correlation functions hence reflect single-transcript kinetics . Time traces shown are simulations where the statistics of transcript kinetics are similar to those we measured by live cell imaging . Traces in ( C ) have the same duration and number of transcripts as estimated in experimental data ( e . g . , Figure 1C ) . See Video 4 for an animation of how the correlation functions converge as the number of transcripts increases . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01610 . 7554/eLife . 03939 . 017Figure 3—figure supplement 1 . Correlation functions with several gene copies at the TS reflect single-transcript kinetics . The same principles as shown on Figure 3 can be applied to a case where several independent genes are at the transcription site ( TS ) . The contributions of pairs of transcripts from distinct genes ( inter-gene ) behave like inter-transcript correlations in Figure 3B and hence distribute uniformly on the delay axis and simply average out . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 01710 . 7554/eLife . 03939 . 018Video 4 . Correlation functions reveal single transcript kinetics . This video shows the convergence of the correlation functions for increasing number of transcripts in a time trace . See also Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 018 The preceding conclusions are general and make no reference to a specific model . To gain further insight , we developed mathematical models which relate the shape of the correlation functions to the timing of the underlying molecular processes ( see ‘Materials and methods’ ) . We generated five different mechanistic schemes: ( I ) purely post-release splicing , ( II ) independence between splicing and elongation/release , ( III ) polymerase pausing at the 3′ splice site ( ss ) until splicing is complete , ( IV ) splicing only during 3′ end retention of the transcript , and ( V ) release only after splicing is complete ( Figure 2—figure supplement 4 and ‘Materials and methods’ ) . For each one of these general schemes , different time distributions were tested for elongation , splicing , and release ( Supplementary file 2 ) . Since by construction , the intron-to-exon cross-correlation at 0 delay is necessarily null in scheme III and have a null slope in scheme I , these two schemes can be ruled out ( See Figure 2—figure supplement 5A , D for fits ) . The three other schemes were better at fitting the correlation curves but the best model is one from scheme II , that is where splicing is independent of elongation and transcript release ( See Figure 2—figure supplement 5 and discussion on Model comparison in ‘Materials and methods’ ) . In this 3-parameter model ( Table 1 ) , splicing occurs a fixed amount of time after the 3′ss has been transcribed , and transcript release involves a stochastic delay after the poly ( A ) site is reached . No pause at the 3′ ss was needed to fit the data . This observation does not rule out pausing at these sites but suggests that such a pause would be much shorter than the other timescales observed . Notably , our data are fit better with a model having a fixed time for intron removal rather than a stochastic ( exponential ) one , arguing for several sequential kinetic steps ( Aitken et al . , 2011; Schmidt et al . , 2011 ) . In total , the β-globin-terminal intron splicing time was 267 ± 9 s after the polymerase passes the 3′ ss . This measurement of splicing time is consistent with previous estimates either in vivo on cell populations ( Singh and Padgett , 2009 ) or in vitro at the single-molecule level ( Hoskins et al . , 2011 ) , suggesting that PP7 stem loops do not perturb splicing kinetics of this intron , contrary to MS2 stem loops ( Aitken et al . , 2011; Schmidt et al . , 2011 ) . As an independent validation of our modeling results , we counted the number of red and green RNAs at transcription sites using a normalized ratiometric approach ( Zenklusen et al . , 2008 ) ( See ‘Materials and methods’ and Figure 2—figure supplement 6 ) . The average red-to-green ratio of 1 . 41 is indistinguishable from the expected 1 . 4 value predicted by our modeling analysis of the correlation functions . 10 . 7554/eLife . 03939 . 019Table 1 . Kinetics of transcription and splicing under different experimental conditionsDOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 019Elongation rate ( kb/min ) Mean 3' end dwell time ( sec ) Splicing time ( sec ) Pre-release fraction ( % ) Control2 . 60 ± 0 . 16116 . 1 ± 5 . 8267 ± 915 . 9 ± 3 . 2SSA+2 . 41 ± 0 . 26126 . 7 ± 5 . 7485 ± 62 **3 . 5 ± 2 . 4 **CPT+1 . 44 ± 0 . 09 **111 . 0 ± 10 . 3251 ± 1024 . 9 ± 6 . 9U2AF1 ( wt ) 2 . 24 ± 0 . 27120 . 7 ± 4 . 9280 ± 816 . 6 ± 3 . 3U2AF1-S34F2 . 64 ± 0 . 11166 . 0 ± 7 . 0 **694 ± 176 *2 . 1 ± 2 . 6 **The table shows result of fits with model II . 4 ( ‘Materials and methods’ and Supplementary file 2 ) . Pre-release fraction is deduced from the 3 other parameters . Errors are propagated SEM from correlation functions . * p-value<0 . 05 , ** p-value<0 . 005 ( two-sided z-test vs control ) . Importantly , both SSA and CPT drug treatments only affected a single parameter ( Table 1 ) , arguing that splicing is kinetically independent of elongation/termination . Based on our measurements , splicing occurring at the TS is rarely completed during elongation but rather during a pause at the 3′ end of the gene . Because transcript release is stochastic , the 3′ end dwell time can be shorter or longer than the necessary time to remove the intron , resulting in splicing that can occur either before or after release . Since the majority of β-globin pre-mRNA is released from the TS before splicing of the terminal intron , we addressed the question of where and when this splicing takes place . We observed a mobile population of unspliced pre-mRNA ( co-localized intron/exon ) diffusing in the immediate vicinity of the TS ( Figure 4A–B , Video 5 ) . In contrast , spliced mRNA ( exon only ) could be observed diffusing throughout the nucleus . A small population of red-only particles was also recorded , which could be due either to the false-discovery rate of the segmentation algorithm or the presence of free lariats ( See ‘Materials and methods’ ) . The radial distribution of mRNA and pre-mRNA indicated an enrichment for unspliced transcripts within 2 . 4 ± 0 . 1 µm of the TS ( Figure 4C , Figure 4—figure supplement 1 ) , meaning that splicing occurs faster than diffusion throughout the nucleus . This enrichment disappears upon treatment with splicing inhibitor SSA , in which case most of the transcripts are unspliced and dispersed throughout the nucleus ( Figure 4C , Video 6 ) . From the measured diffusion coefficient ( D = 0 . 12 µm2/s , Figure 4—figure supplement 2; See ‘Materials and methods’ ) , we calculated that post-release splicing takes place on average 13 ± 1 s after departure from chromatin . This time is much shorter than the expected 137 s it would take if pre- and post-release splicing kinetics were identical ( calculated from Table 1 ) . From this observation , it is tempting to speculate that transcripts are released only after they have passed a particular rate-limiting step in spliceosome assembly , explaining why the catalytic step occurs very soon after release . However , this interpretation is inconsistent with the fact that 3′end retention time is not affected by SSA treatment which impairs binding of U2 and hence affects recruitment of all the snRNPs except U1 ( Corrionero et al . , 2011 ) . Note also that we cannot formally exclude the possibility that co-localized intron/exon particles are actually excised introns still in complex with transcripts . In summary , although introns can be retained for over 4 min on chromatin , once the transcript is released , splicing is 10-fold faster for freely diffusing transcripts . In most cases , the nascent intron is retained until transcription reaches the 3′ end of the gene and then removed either before or shortly after release of the transcript ( Figure 5 ) . 10 . 7554/eLife . 03939 . 020Figure 4 . Visualization of splicing occurring after release from chromatin . ( A ) Individual frames from live-cell confocal imaging showing intron ( red dots ) , exon ( green dots ) , and the merged image . White arrow: TS . Bar: 4 µm . ( B ) Fluorescence intensity profile along the line in the inset shows co-localized intron/exon ( unspliced pre-mRNA ) and exon only ( spliced mRNA ) . ( C ) Radial distributions of mRNA ( green ) and pre-mRNA ( orange ) , as well as pre-mRNA under SSA treatment ( black ) are shown as a function of distance from the TS . Density distributions are normalized by the distribution of random ( uniform ) positions within the nucleus ( see ‘Materials and methods’ ) . Error: SEM ( bootstrap over 9 cells ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02010 . 7554/eLife . 03939 . 021Figure 4—figure supplement 1 . Localization of single RNAs diffusing in the nucleus . ( A ) Example of a raw confocal image from Video 5 . Many particles are visible in the exon channel ( center ) . The intron channel ( left ) shows fewer particles that colocalize with green ones ( See merge , right ) . ( B ) The spot-finding algorithm , run independently on the red and green images , generates a list of red and green spot coordinates . Spots colocalized by less than 250 nm were paired , yielding a list of red-only , green-only , and colocalized particles . ( No spots are indicated on the red image since all of the red spots colocalize with a green one in this frame . ) ( C ) Plotting , for all the frames , the position of all the red spots relative to all the green spots in the same frame reveal a population of particles colocalizing in the two channels within a 250-nm radius . ( D ) The radial distribution of the distances between colocalized particles and the transcription site is shown here normalized by the radial distribution of random locations within the nucleus . Three exprimental conditions are shown: for three experimental conditions: untransfected cells and cells transfected with wild-type ( wt ) or mutant ( S34F ) U2AF1 . The depletion at very short distances is a technical artifact of spot detection ( 2 very close particles are detected as a single one ) . ( E ) These distributions are fitted with a Gaussian function with three parameters: standard deviation σ , height h , and baseline y0 . Errors: SEM over cells . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02110 . 7554/eLife . 03939 . 022Figure 4—figure supplement 2 . Measure of RNA diffusion in the nucleus by RICS . ( A ) Spatial RICS autocorrelation function of green channel from a RICS measurement . Color code is the same as the vertical axis of ( B ) . ( B ) A 2D fit to a two component diffusion model gives a good fit with fast and slow moving components ( 60% , 1 . 64 μm2/s and 40% , 0 . 095 μm2/s ) . Bottom graph is the fit and top graph is the residual . ( C ) Resulting measurements of free coat protein ( MS2-GFP ) and mRNA diffusion coefficients . Error bars: SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02210 . 7554/eLife . 03939 . 023Video 5 . Single-RNA imaging reveals a transient population of unspliced transcript diffusing away from the transcription site . Using high-power confocal laser scanning microscopy , we were able to observe single transcripts with a better temporal resolution than with widefield imaging ( Video 3 ) . The video shows a single cell with an active transcription site ( TS , bright spot visible in both signals ) and diffusing RNA particles ( left: intron , center: exon , right: merge ) . Although most of the RNAs diffusing in the nucleus are spliced ( visible only in the exon signal ) , few unspliced RNAs ( visible in both colors ) are detectable in the vicinity of the TS as they diffuse away . Spatial distribution and diffusion analyses revealed that this population is very transient ( Figure 4C and Figure 4—figure supplements 1 and 2 ) . Large shapes in the nucleus are nucleoli ( Ferguson and Larson , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02310 . 7554/eLife . 03939 . 024Video 6 . Single-RNA imaging with splicing inhibitor SSA . Imaging conditions are identical as in Video 5 , but cells are treated with splicing inhibitor spliceostatin A ( SSA ) . RNAs diffusing in the nucleus are now visible in both color , indicating that they are unspliced . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02410 . 7554/eLife . 03939 . 025Figure 5 . Schematic of β-globin transcription cycle kinetics . Transcript synthesis and processing can occur through different pathways , the choice of which is governed by a kinetic competition between transcription and splicing . After transcription of the 3′ splice site , intron removal takes about 260 s and elongation until the end of the gene , about 55 s . Hence , splicing does not occur during elongation . The transcript is retained at the 3′end of the gene for a stochastic amount of time that can be shorter or longer than the remaining time to excise the intron . This results in two possible outcomes: either an unspliced pre-mRNA is released and then spliced very rapidly or splicing occurs while the transcript is retained on chromatin before being released . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 025 Because the balance of kinetic competition determines where and when introns are excised from the pre-mRNA , we then sought to determine whether trans-acting factors regulated splicing by perturbing this balance . Deep sequencing studies in myelodysplastic syndrome , chronic lymphocytic leukemia , acute myeloid leukemia ( AML ) , breast cancer , lung adenocarcinoma , and hairy cell leukemia have all revealed the existence of mutated factors involved in 3′ ss recognition ( Yoshida et al . , 2011; Graubert et al . , 2012; Brooks et al . , 2014; Waterfall et al . , 2014; TCGA , 2012 ) . One recurrent change is a heterozygous point mutation in U2 auxiliary factor 1 ( U2AF1 ) , which is an essential factor for recognition of the AG dinucleotide consensus motif ( Figure 1A ) . The serine-to-phenylalanine ( S34F ) missense mutation in the zinc finger domain results in disparate changes in alternative splicing patterns including exon skipping , exon inclusion , and alternative 3′ ss selection ( Brooks et al . , 2014 ) . Although somatic genetics clearly indicate the importance of this mutation ( Waterfall et al . , 2014 ) , there is no functional or mechanistic understanding of how U2AF1-S34F works at the molecular level . We performed time-lapse imaging on cells expressing moderate levels of either wild-type U2AF1 or U2AF1-S34F , both fused to a cerulean fluorescent protein . We note that this experimental condition recapitulates the in situ case , because the mutant U2AF1 is present against the background of at least one copy of the wild-type allele . Correlation functions revealed that U2AF1-S34F completely abolishes pre-release splicing ( Figure 6A , horizontal slope ) and prolongs transcript 3′ end dwell time ( Table 1 ) . Post-release imaging of transcripts showed a local enrichment for unspliced pre-mRNA near the TS , but with a greater spatial extent in the case of U2AF1-S34F ( Figure 6B–C , Video 7 ) . Thus , all transcripts are post-transcriptionally spliced , albeit at a slower rate ( 27 ± 3 s; Figure 6D , Figure 4—figure supplement 1 ) . Splicing efficiency , poly ( A ) length , and 3′ UTR length were unchanged ( Figure 1—figure supplement 1C–D ) . In summary , these data suggest that the mutant delays splicing to post-release , slows splicing from freely diffusing transcripts , but has no detectable effect on splicing efficiency . 10 . 7554/eLife . 03939 . 026Figure 6 . The U2AF1-S34F mutant acts as a dominant negative by delaying splicing to post-release . ( A ) Expression of U2AF1-cerulean does not alter pre-release splicing compared to the un-transfected control . Expression of U2AF1-S34F-cerulean abolishes splicing at the TS ( horizontal slope of the intron-to-exon cross-correlation , blue curve ) . ( B ) Pre-mRNA ( red , marked by squares ) are enriched around the TS ( arrows ) indicating that splicing still occurs faster than diffusion . The enrichment is broader in the presence of U2AF1-S34F despite the similar spatial distributions of both proteins . ( C ) Gaussian fits onto pre-mRNA radial distance distributions from the TS . ( D ) The U2AF1-S34F mutant defers splicing to occur entirely away from the TS ( fractions obtained from model fits in Table 1 ) and increases post-release splicing time . ** p < 0 . 005 ( two-sided z-test vs untransfected control ) . ( E ) Two-color single-molecule FISH on endogenous FXR1 transcripts . Unspliced pre-mRNA ( co-localization of intronic and exonic probe ) appears in the vicinity of TSs ( the 4 bright dual-color spots ) . ( F ) The fraction of pre-mRNA transcripts in the nucleus in the presence of wt or mutant U2AF1 . ( G ) Spatial distribution of pre-mRNAs near TSs in the presence of wt or mutant U2AF1 ( N > 400 cells ) . Radial distributions show density of pre-mRNA normalized by density of mRNA . Bars: 4 µm . Error: SEM over cells ( bootstrap ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 02610 . 7554/eLife . 03939 . 027Video 7 . Spatial distribution of pre-mRNA with wild type or mutant U2AF1 . Left and right images show a cell that was transfected with the wild type ( wt ) or the mutant ( S34F ) version of splicing factor U2AF1 . Both the image show the intron channel . The enrichment of unspliced pre-mRNA ( red spots ) diffusing in the vicinity of the transcription site is broader in the case of the mutant , showing that splicing rate is slower . Imaging conditions are identical as in Video 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 03939 . 027 We then confirmed this kinetic effect on the endogenous FXR1 mRNA , which was shown to be alternatively spliced in the presence of U2AF1-S34F in both AML and lung adenocarcinoma ( Brooks et al . , 2014 ) . Using single-molecule FISH ( Femino et al . , 1998 ) ( Figure 6E ) , we examined the spatial distribution of intron and exon in fixed cells transfected with U2AF1 or U2AF1-S34F . As was the case for the β-globin reporter , we observed a population of unspliced FXR1 pre-mRNA in the vicinity of an active TS , indicating that at least some fraction of FXR1 transcripts are released before splicing ( Figure 6F ) . We then performed the same spatial analysis as above , except on fixed cells instead of live cells . Expression of the U2AF1-S34F mutant resulted both in an increase in the level of unspliced pre-mRNA in the nucleus ( 16 . 0 ± 0 . 4% compared to 6 . 0 ± 0 . 3% , Figure 6F ) and the radial distance from the TS ( 1 . 8 ± 0 . 5 μm compared to 0 . 5 ± 0 . 2 μm , Figure 6G ) . Taken together , both static and dynamic measurements on a reporter β-globin transcript and the endogenous FXR1 transcript suggest that the S34F mutation in U2AF1 acts in a dominant negative fashion to postpone splicing until after release and cause slower splicing from diffusing transcripts .
The picture that emerges from this study is one in which the β-globin-terminal intron can be spliced during multiple steps of the transcription cycle ( Figure 5 ) . A minority of transcripts are spliced while retained at the 3′ end of the gene , but the dominant pathway is the one in which splicing is contemporaneous with release or occurs shortly thereafter . These results are consistent with several studies which suggested that intron removal is enhanced upon cleavage ( Baurén et al . , 1998; Bird et al . , 2005 ) . In fact , we find that splicing is 10-fold faster on freely diffusing transcripts than on chromatin-bound transcripts . Our data are consistent with a model where commitment to splicing of the terminal β-globin intron occurs co-transcriptionally but suggest there may be a high energy barrier to completion of intron removal while transcripts are still tethered to chromatin , possibly due to steric constraints . Thus , post-release splicing may be the energetically favored process , but there is a long kinetic window in which the less-favorable pre-release intron removal can occur . The time required to remove an intron therefore becomes a central parameter in our understanding of RNA processing , with implications for both constitutive and alternative splicing ( Bentley , 2014 ) . This fundamental kinetic quantity has been elusive . Population-based measurements suggest a splicing time for several endogenous , un-modified introns of 5–10 min ( Singh and Padgett , 2009 ) . Bulk in vitro measurements on β-globin indicate a 40- to 50-min timescale . A single-molecule in vitro study on yeast transcripts measured ∼10-min splicing time . On the opposite end of the spectrum , previous live-cell imaging approaches relying on both direct and indirect measurements indicate splicing times in the order of ∼30 s ( Huranová et al . , 2010; Martin et al . , 2013 ) . One of the primary experimental advances in our study is the ability to observe the transcription and splicing process with a time resolution that spans three orders of magnitude . Since the time to splice is a distributed quantity , and splicing times vary from transcript to transcript , the variation in previously reported splicing rates may be strongly influenced by the temporal dynamic range of the method . Biologically , this variability in splicing rate may provide regulatory potential , as we discuss below . One potential criticism of the live-cell imaging approach is that the stem-loops and the coat protein may perturb kinetics . Several arguments stand against this view . First , our splicing kinetics are consistent with both population measurements ( Singh and Padgett , 2009 ) and in vitro single-molecule measurements ( Hoskins et al . , 2011 ) . Second , the splicing efficiency of the reporter is high ( Figure 1—figure supplement 1C ) , suggesting there are no dead-end intermediates . Third , endogenous metazoan RNAs are decorated with RNA-binding proteins from their inception ( Castello et al . , 2012 ) , suggesting that the spliceosome is well-equipped to handle bulky messages . Finally , since we can never exclude the possibility that a synthetic reporter may be missing features found in an endogenous gene , we have also recapitulated the results on the un-modified endogenous FXR1 message . Importantly , our single-molecule study reveals the existence of multiple pathways , indicating the absence of a strict checkpoint for intron removal ( Bird et al . , 2005; Alexander et al . , 2010; Brody et al . , 2011; Carrillo Oesterreich et al . , 2010; Bhatt et al . , 2012; Pandya-Jones et al . , 2013 ) . The presence of a delay at the 3′ end may be interpreted as a checkpoint mechanism to ensure that splicing takes place before transcript release , but there are several reasons to reject this interpretation of our data . First , none of the models which assume dependence between splicing and elongation/release fit better than one that assumes independence . Second , both splicing and elongation inhibition experiments support the view of the two processes being kinetically independent . Abolishing splicing with SSA does not lead to an increase in release times . Conversely , reducing elongation speed with CPT does not slow down splicing . Instead , it leads to an increase in pre-release splicing ( Figure 2E ) , as expected if the two processes are independent . Thus , splicing and release can happen in either order , with the order of events determined by kinetic competition . It is this competition which results in stochastic outcomes for the RNA . What are the physiological implications of this stochastic outcome ? Does the cell utilize changes in kinetic balance to alter gene expression ? Moreover , is the timing of intron removal of secondary importance to the timing of splicing commitment ? While , single-cell variation in alternative splicing has been observed , the mechanism behind this variability has remained elusive ( Waks et al . , 2011; Lee et al . , 2014 ) . We have found that a single point mutation in U2AF1 that is recurrent in multiple human cancers alters this kinetic balance to favor post-transcriptional splicing of both the β-globin reporter and also endogenous FXR1 mRNAs . Interestingly , single-molecule measurements on fixed cells suggest that post-release splicing may be the preferred pathway for alternatively spliced transcripts ( Vargas et al . , 2011 ) . Our studies on the mutant U2AF1 provide a mechanistic basis for this observation and suggest a role for mutations in the core splicing machinery for the increased levels of ‘noisy splicing’ which are observed in cancer ( Pickrell et al . , 2010; Chen et al . , 2011 ) . Furthermore , our time-resolved results indicate that post-release splicing is more efficient than pre-release splicing , which may explain how the timing of intron removal might lead to different outcomes for the message . Such a model relies on a degree of plasticity in spliceosome assembly and function , which has been suggested by in vitro single-molecule measurements ( Hoskins et al . , 2011; Shcherbakova et al . , 2013 ) . We speculate that the kinetic delay induced by U2AF1-S34F allows either for alternate pairing between 5′ and 3′ss during transcription or for post-release reconfiguration of pre-mRNA which is not yet committed to intron excision in the spliceosome . Other pathological RNA processing defects may also originate from a similar kinetic imbalance . In summary , the single-molecule approach developed here provides a blueprint for dissecting the many competing processes which take place at the earliest stages of gene expression .
The reporter gene vector ( Figure 1—figure supplement 1A ) was constructed with the human β-globin DNA sequence placed under the control of a Tet-responsive promoter , as described previously ( Janicki et al . , 2004; Darzacq et al . , 2007 ) . Briefly , the 3′ end of the human β-globin sequence was truncated 72 bp upstream of the endogenous stop codon . It was replaced by a cassette coding for the cyan fluorescent protein fused to the peroxisome-targeting sequence serine-lysine-leucine ( CFP-SKL ) , ending with a translation stop codon . A cassette containing 24 repeats of the MS2 stem loop sequence was inserted in the 3′ untranslated region ( 3′UTR ) , followed by a bovine growth hormone polyadenylation sequence ( BGH-PolyA ) . In our modified construct , we inserted a cassette containing 24 repeats of the PP7 stem loop sequence into the second and last intron of the gene , approximately halfway between the 5′ splice site ( 5′ss ) and the branch point . In order to be the least perturbing , the cassette was inserted 463 bp downstream of the 5′ss and 371 bp upstream of the branch point , and most of the endogenous DNA was conserved ( 8 bp in the intron were deleted ) . In addition , we replaced the MS2 cassette with one that is less prone to recombination . The repeating units of the PP7 and MS2 cassettes are composed of two-stem loop blocks , which are then multimerized 12× to get 24× total stem loops: MS2: GATCCTACGGTACTTATTGCCAAGAAAGCACGAGCATCAGCCGTGCCTCCAGGTCGAATC TTCAAACGACGACGATCACGCGTCGCTCCAGTATTCCAGGGTTCATCAG PP7: CTAAGGTACCTAATTGCCTAGAAAGGAGCAGACGATATGGCGTCGCTCCCTGCAGGTCGA CTCTAGAAACCAGCAGAGCATATGGGCTCGCTGGCTGCAGTATTCCCGGGTTCATTAGATC . The loops themselves are underlined . The construct was stably integrated into the genomic DNA of U2-OS cells ( human osteosarcoma cell line ) through transfection of the plasmid followed by a screen for genomic integration . The cell line was made by co-transfection of the reporter plasmid and a puromycin resistance plasmid , followed by selection with puromycin . The cell line also constitutively expresses the PP7 bacteriophage-coat protein fused to the mCherry fluorescent protein ( PP7-mCherry ) and the MS2 bacteriophage-coat proteins fused to the green fluorescent protein ( MS2-GFP ) . Both are under the control of a ubiquitin promoter and were stably introduced by lentivirus infection as described previously ( Larson et al . , 2013 ) . The clonal cell line used in this study was finally generated by single-cell cloning . The reporter is efficiently expressed and translated as evidenced from imaging of the protein product accumulating in peroxisomes ( Figure 1—figure supplement 1B ) . Cells were grown in DMEM medium ( Life technologies , Grand Island , NY ) supplemented with 10% fetal bovine serum ( FBS , Sigma-Aldrich , St Louis , MO ) . Cell were induced with 10 or 20 μM doxycycline ( Sigma Aldrich , St Louis , MO ) , at least 24 hr prior to imaging . Imaging was performed in Leibovitz L-15 phenol-free medium ( Life technologies , Grand Island , NY ) containing the same concentration of doxycycline and FBS . Pharmacological treatments were performed at 48 nM spliceostatin A ( Dr Minoru Yoshida , Chemical Genetics Lab , RIKEN Institute , Japan ) from 0 . 5 to 20 hr prior to imaging or 3 . 75 μM Camptothecin ( Sigma , CAS: 7689-03-4 ) from 1 to 19hr . U2AF1 ( a . k . a . U2AF35 ) constructs were made from a U2AF35-CFP plasmid provided by Angus Lamond ( University of Dundee , UK ) . The CFP cassette was replaced by a Cerulean one ( similar absorption/emission spectra but more photostable ) . The U2AF1 cassette was then swapped with the mutant version ( U2AF1-S34F ) provided by Peter Aplan ( NCI , NIH ) . Cells were transfected with mock or with plasmids expressing either the WT or S34F mutant of U2AF1 . Expression of β-globin reporter was induced for 15 hr . Cells were scraped and total RNA isolated using the Qiagen RNA isolation kit . 1 μg of total RNA was used to make first strand using ProtoScript II M-MuLV reverse transcriptase ( NEB , Ipsich , MA ) and random hexamers ( IDT , Coralville , IA ) according to the manufacturer's instructions in a final volume of 20 μl . 2 μl of the reverse transcription product was used in a qPCR reaction using IQ Syber Green mix ( Bio-Rad , Hercules , CA ) in a CFX96 qPCR machine ( Bio-Rad , Hercules , CA ) according to the manufacturer's instructions . We use varying amounts of G-block DNA ( IDT , Coralville , IA ) carrying primer pairs generating amplicons of the same size as those being tested . This G-block served as control and was used to generate the standard curves . Primer pairs spanning the junction of Exon2-Intron2 , and Intron2-Exon3 , and a primer pair amplifying a region of Exon3 served to measure unspliced and total RNA of β-globin reporter respectively . The unspliced fraction of the reporter RNA was calculated as 10^[ ( Cqintron-Kintron ) /Sintron - ( Cqtotal-Ktotal ) /Stotal] where Cq is the cycle number , and K and S are the constant and slope respectively of the standard curve for the primer pair ( Figure 1—figure supplement 1C ) . The splicing fractions calculated using Exon2-Intron2 primers or Intron2-Exon3 primers were similar and were therefore combined together . Error bars are standard errors over four measurements . U2-OS cells with the integrated β-globin reporter were transfected with mock or with plasmids expressing either the WT or S34F mutant of U2AF1 . Expression of β-globin reporter was induced for 15 hr . Cells were scraped and total RNA isolated using the Qiagen RNA isolation kit . 1 μg of total RNA was treated with RNAse H ( NEB , Ipswich , MA ) in the presence oligo dT according to the manufacturer's instructions . The digested RNA was then phenol:chloroform extracted , ethanol precipitated and suspended in 10 μl of water . A DNA adaptor ( 5-GGTCACCTTGATCTGAAGC , with a 5-phosphate and 3 amino modification to prevent further ligation ) was ligated to 1 μg of either total or RNAse H-treated RNA using T4 RNA ligase 1 ( NEB , Ipswich , MA ) in a final volume of 20 μl . 10 μl of the adaptor-ligated RNA was then used to synthesize the first strand using ProtoScript II RT ( NEB , Ipswich , MA ) and the reverse primer 5′-GCTTCAGATCAAGGTGACCTTTTT according to the manufacturer's instructions . The reaction was denatured for 20 min at 80°C and 2 . 5 μl of the first strand was used as template in a standard PCR reaction using the forward primer 5′-CCAGGGTTCATCAGATCCTATTCTATAGTGTCAC and the reverse primer 5-GCTTCAGATCAAGGTGACCTTTTT . Reaction products were then separated on a 3% agarose gel ( Figure 1—figure supplement 1D ) . Whole genome sequencing ( paired end sequencing ) was used to identify the integration sites of the reporter in the genome . The genomic library was prepared using TruSeq DNA sample preparation protocol . Samples were sequenced on HiSeq2000 using Illumina TruSeq v3 chemistry . The yield was 389 million reads after filtering . The reads were trimmed to remove low quality sequences ( Trimmomatic software ) and were aligned to both the human genome ( hg19 ) and the sequence of the β-globin reporter gene plasmid ( Bowtie2 and Illumina CASAVA Eland softwares ) . Paired reads that partially align to both genomic and plasmid sequences were extracted and used to identify possible insertion sites . Figure 1—figure supplement 2A shows an example of an insertion site: 4 read pairs align with both with genomic sequence ( up to chr8:145 , 074 , 337 ) and plasmid sequence ( all from the same position in the plasmid backbone ) . This defines the 5′ junction of the insertion . Similarly , three read pairs were found partially aligning a few bp downstream ( chr8:145 , 074 , 349 ) and with the plasimd up to the beginning of the MS2 cassette , defining the 3′ junction . Multiple copies of the plasmid may have been integrated . Three insertion sites were identified ( Figure 1—figure supplement 2B ) , including two where only one junction was found . All three were confirmed by PCR and the copy number of the reporter in the cell line was obtained as follows . Genomic DNA was isolated from cells ( 1 × 10 cm plate ) by standard protocols , and amplicons were PCR amplified using primer pairs overlapping the putative junctions . The PCR was performed for 24 cycles and the products separated on a 2% agarose gel ( Figure 1—figure supplement 2C ) . All four junctions show a band at the expected size . To estimate the copy number of the inserted plasmid , we also use two primer pairs that amplify a region that is internal to the plasmid construct . Each junction should be present in one copy per cell so that , in comparison , the amount of the internal amplicon should reflect the total copy number in the cell line . The gel was quantitated using the ImageJ software ( Figure 1—figure supplement 2D ) . To correct for primer efficiency , each primer pair was also used in the same PCR run to amplify varying amount of G-block DNA ( IDT , Coralville , IA ) . The product was quantitated as described above and the resulting calibration curves ( Figure 1—figure supplement 2E ) were used to correct the data . The internal-to-junction ratio of the corrected amounts of PCR products indicates between 4 and 7 total copies of the reporter ( Figure 1—figure supplement 2F ) . Microscopy data acquisition from Figure 1 , Figure 2 and Figure 6A , E–G ( and associated supplements ) and Videos 1 to 3 was performed using a custom build wide-field microscope ( described in greater details in Ferguson and Larson , 2013 ) . It consists in an AxioObserver inverted microscope ( Zeiss , Thornwood , NY ) with a high aperture objective ( Zeiss 63× C-Apochromat ) and two Evolve 512 EMCCD cameras ( Photometrics , Tucson , AZ ) . Excitation sources are 488 nm and 594 nm lasers ( Excelsior , Spectra Physics , Santa Clara , CA ) . Typical laser intensities are 250 μW and 50 μW respectively . For imaging Cerulean , we used a 445 LED ( Zeiss Colibri ) or an X-Cite lamp ( Lumen Dynamics ) . Cells were imaged in 35-mm MatTek dishes ( MatTek , Ashland , MA ) placed in a Tokai Hit stage incubator ( INUB-LPS , Shizuoka-ken , Japan ) . Average temperature inside the dish was measured at 37°C using a thermocouple . Images were taken every 10 s as z-stacks ( 7 or 9 images , Δz = 0 . 5 μm ) and in both color simultaneously with two cameras , using an exposure time of 100 ms and for a duration between 45 and 512 frames . Raw images were collected using MicroManager software ( Edelstein et al . , 2001 ) . Maximum intensity projections were computed ( e . g . Video 1 ) and used for tracking . Video 3 was obtained using a shorter exposure time ( 50 ms ) to observe the fluctuations due to diffusing RNAs . Bicolor fluorescence time traces at the transcription site were generated using a custom software written in IDL ( Source code 1 ) that was previously described , with minor modifications to handle bicolor data sets . In each image , diffraction-limited spots are detected using band-pass filtering and refined using an iterative Gaussian mask localization procedure ( Crocker and Grier , 1996; Thompson et al . , 2002; Larson et al . , 2005 , 2011 ) . Trajectories are then generated based on a nearest-neighbor method with a maximal jump distance threshold . If no spot is detected within the threshold distance , the previous location is used as the initial guess for the iterative Gaussian mask localization procedure . Integrated fluorescence intensity over the diffraction-limited spot is collected using a Gaussian mask fit after local background subtraction ( Thompson et al . , 2002 ) . An example of time-lapse video with tracking is shown in Videos 2 . Time traces were corrected for photobleaching as follows . In an ideal experiment , the fluorescence intensity histogram of the whole nucleus should stay roughly unchanged throughout the acquisition . We computed smoothed versions ( polynomial fit ) of the mean and standard deviation ( s . d . ) of this histogram in each color over time . Time traces from TS tracking were then normalized by the s . d . ( because of the background subtraction in the tracking procedure , only the s . d . of the nucleus histogram acts as a scale factor on a time trace ) . Traces were inspected for ( i ) accurate tracking ( portions of inacurate tracking were trimmed off; short traces ( <100 frames ) were discarded ) , ( ii ) in-focus TS ( traces where TS reaches the first or last z-plan were discarded ) , and ( iii ) signal to noise ratio ( highly noisy traces were discarded ) . Examples are shown on Figure 1C and Figure 2—figure supplement 1A . For each time trace , autocorrelation and crosscorrelation functions were computed asGab ( T ) =〈δa ( t ) δb ( t+T ) 〉 / 〈a ( t ) 〉〈b ( t ) 〉where <·> denotes time average , δa ( t ) means a ( t ) - áa ( t ) ñ , and a ( t ) and b ( t ) can be any combination of the red and green time traces r ( t ) and g ( t ) . Correlation functions were computed using a multi-tau algorithm ( Wohland et al . , 2001 ) , which iteratively down-samples the signals for increasing time delay . This yields a somewhat uniform spacing of time delay points on a logarithmic scale , reducing the sampling noise at long delays while keeping a high temporal resolution at short delays . When shifting the two signals , non-overlapping ends are not wrapped . See Figure 2—figure supplement 1A for examples of correlation functions from single traces . To reach better statistical convergence , correlation functions from single time traces were averaged together ( Figure 2—figure supplement 1B ) . Each point of the single-trace correlation functions was given a weight corresponding to the number of overlapping time points from the signals used in its computation . Bootstrapping was performed to obtain standard error of the mean correlation functions ( SEM ) . Two normalizations were performed prior to bootstrapping: ( i ) Baseline subtraction: Slow processes ( e . g . bursting , cell cycle ) may produce a slow decay which adds up to the fast transcription/splicing kinetics . These are usually well separated; see how the decay is much slower after 200–300 s on Figure 2—figure supplement 1B–F . The baseline subtraction gets rid of the slow decay , approximating it as a constant offset at short delays . ( ii ) Normalization with null-delay crosscorrelation Grg ( 0 ) : All 4 correlation functions were normalized using an estimate of Grg ( 0 ) . This was done for 2 reasons: ( a ) doing this normalization prior to bootstrapping promotes a good convergence of the crosscorrelation at short delays 0 ( the most informative part of the correlation functions; See Supplementary file 1—§2 ) . This constrains the fits to capture precisely the temporal features at short delays . ( b ) This normalization reduces by 1 the number of free parameters of all the models used to fit the data: On unnormalized correlation functions , changing the initiation rate of transcription simply scales up and down all 4 correlation functions together ( See Supplementary file 1—§1 ) . Normalizing them gets rid of this degree of freedom . Because most of the time Grg ( 0 ) is inaccurate due to shot noise and/or small tracking error ( e . g . Figure 2—figure supplement 1C , E , F ) , we use the estimate ( Grg ( Δt ) + Ggr ( Δt ) ) /2 instead ( Δt = 10 s is the sampling time ) . We performed imaging with a confocal laser scanning microscope to observe single RNAs diffusing in the nucleus ( Figure 4 , Figure 6B–C [associated supplements] , Figure 2—figure supplement 6 and Videos 5 to 7 ) . Imaging was performed on a Zeiss 780 confocal microscope with 488/594 nm excitation . To detect and localize RNAs in the nucleus , we used the same spot-localization algorithm as for widefield microscopy ( Figure 4—figure supplement 1B ) . False positive spots due to nucleoli or localized outside of the nucleus were eliminated using a standard masking procedure . Spots were detected independently in both channels ( intron and exon ) in 9 videos of 50 frames on average taken with a 3 . 26 s frame interval . Pairs of red and green spots colocalized by less than 250 nm ( Figure 4—figure supplement 1C ) were hence considered single bi-color particles . The TS was tracked and used as a reference for radial distributions of particles ( Figure 4C , Figure 4—figure supplement 1D ) . These distributions were normalized by the distribution obtained for random locations in the nucleus ( i . e . , a value of 1 observed indicates a purely uniform distribution ) . RNAs in the nucleus being relatively sparse , colocalized particles are expected to be mostly true positive , that is unspliced RNAs . On the other hand , a proportion of the red-only and green-only particles may be false positive ( e . g . , bi-color RNA only detected in one color ) . mRNAs being much more abundant than pre-mRNA ( Figure 4A ) , green-only spots should only have a small portion of false positives . It is however difficult to estimate what portion of the red-only spots truly represents free lariat and what is a wrong detection/categorization of pre-mRNA . The fact that the red-only radial distributions and the co-localized radial distribution are very similar suggests that red-only spots have a high proportion of false positives ( Figure 4—figure supplement 1D ) and that lariat lifetime is likely shorter than that of post-release pre-mRNA ( i . e . < 13 s ) . Depletion at distances shorter than 1 μm is due to the difficulty for the spot-localization algorithm to locate two spots closer than this distance . The radial distribution of colocalized red–green spots was fit to a Gaussian distribution with three parameters: standard deviation σ , height h , and baseline y0 ( Figure 4—figure supplement 1E ) . We determined the diffusion coefficient of RNAs in the nucleus using raster image correlation spectroscopy ( RICS ) ( Brown et al . , 2008 ) ( Figure 4—figure supplement 2 ) . Imaging was performed in photon counting mode . Time series of 10 frames and 512 × 512 pixels were taken with a 100 μs pixel dwell time , 61 ms line scanning time , and 52 nm pixel size . The 1/e2 beam waist was determined to be 246 nm in the green channel and 375 nm in the red channel by fitting a 2D Gaussian to a profile through the diffraction limited transcription site . Correlation functions were calculated and fit to a two component diffusion model using the Globals software package developed at the Laboratory for Fluorescence Dynamics at the University of California at Irvine ( http://www . lfd . uci . edu/globals/ ) . Based on the distribution of unspliced transcripts around the TS ( i . e . , normal distribution with σ = 2 . 421 μm ± 0 . 087 ) and a diffusion coefficient of D = 0 . 12 μm2/s , we deduce that the splicing time following release is on average σ2/4D = 12 . 71 s ± 0 . 92 . We applied the same analysis to confocal videos obtained on cells transfected with wild-type or mutant version of U2AF1 ( Figure 4—figure supplement 1E ) . We also used confocal images to normalize the intensity of the transcription site by the intensity of single transcripts , allowing us to count the number of nascent transcripts in both channels ( Figure 2—figure supplement 6 ) . The results were in agreement with the transcription and splicing kinetic parameters we found in our modeling analysis ( Table 1 ) . To confirm our results obtained on a reporter gene , we performed FISH on the endogenous gene FXR1 ( fragile X mental retardation ) . FXR1 was identified by Brooks et al . ( 2014 ) as being alternatively spliced in the presence of U2AF1-S34F mutation ( i . e . , higher retention of the second to last exon ) . According to our whole genome sequencing data ( See above ) , FXR1 is at a tetraploid locus in our line of U2-OS cells . We designed 48 probes in the second to last intron ( right upstream the alternatively spliced exon ) and 48 probes in the exons of FXR1 that are common between all the RefSeq variants , and excluding the two last exons ( Figure 6E ) . Probes were generally 20 nucleotides . Intronic and exonic probe sets were synthesized and labeled with cyanine dyes Cy3 and Cy5 , respectively , by Biosearch Technologies ( Petaluma , CA ) . U2-OS cells from the same cell line as in the rest of our study were grown on coverslips and transfected with wild type ( wt ) or mutant ( S34F ) U2AF1 labelled with Cerulean fluorescent protein , 24 hr prior to fixation . Fixation and hybridization were performed according to the Stellaris RNA FISH protocol ( Biosearch Technologies , Petaluma , CA ) . Coverslips were mounted on microscope slides using mounting media with DAPI ( ProLong Gold antifade reagent , Life Technologies ) . Imaging was performed on the same widefield microscope as described above . Light sources for imaging DAPI , Cy3 , Cy5 , and Cerulean were 365 nm , 530 nm , 625 nm , and 445 nm LEDs respectively , from a Zeiss Colibri ( Zeiss , Thornwood , NY ) . The detector was a Hamamatsu ORCA-R2 C10600 camera ( Hamamatsu Photonics , Japan ) . Fields of view were selected for cells with low levels of U2AF1-Cerulean and stacks of nine images with z-step of 0 . 5 μm were acquired in four colors . Maximum intensity projection images were used for analysis . Diffraction limited spots were identified in the Cy3 and Cy5 channels independently using the same software as above . Nuclear masks were generated from the DAPI channel using CellProfiler software ( Broad Institute , http://www . cellprofiler . org ) . Spots within 200 nm of each other were paired and considered a bicolor particle . Bicolor particles with a fluorescence more than a twofold of that of single RNAs were considered transcription sites ( TS ) . The measured number of mRNA ( exon only particles ) was similar with wild-type U2AF1 ( 29 . 9 mRNA/nucleus ) and with the mutant ( 35 . 5 mRNA/nucleus ) . However , the fraction of pre-mRNAs over total nuclear RNAs was significantly different: 6 . 0% ( ±0 . 3 ) with the wild type and 16 . 0% ( ±0 . 4 ) with the mutant ( Figure 6F ) . Radial distributions were obtained by computing the distance between single RNAs and all the TSs in a nucleus . Normalized density shown in Figure 6G is the density of pre-mRNA over the density of mRNA . The enrichment at short distances was fit with a Gaussian distribution as for the confocal data . The standard deviation parameter reveals that the local enrichment for pre-mRNA near TSs is broader with U2AF1-S34F ( 1 . 83 ± 0 . 49 μm ) than with U2AF1-wt ( 0 . 48 ± 0 . 19 μm ) . Note that the non-null baseline in the radial distributions also reveals a second population of pre-mRNAs which are spliced slower than diffusion in the nucleus . The proportion of this second population as well as its spatial distribution is not affected by U2AF1-S34F mutation . In Supplementary file 1—§2 , we showed how different features in the geometry of the correlation functions ( e . g . , position of specific angles , decay times , change of slope , … ) are reflecting different aspects of the transcription/splicing kinetics ( e . g . , elongation speed , intron and exon dwell times , stochastic co- and post transcriptional splicing , … ) . Here , we want to build mathematical models that fully predict the correlation functions given ( i ) a certain underlying mechanism and ( ii ) a set of parameters . In essence , we want to ask what molecular mechanisms are consistent with all the above-mentioned kinetic features—and possibly more—that the correlation functions are encoding . The caveat is however that these models necessarily make assumptions on the underlying mechanisms . Hence , we take a general approach in generating a series of simple models ( each assuming different mechanisms ) and assess which one ( s ) account for the experimental data . | To make a protein , part of a DNA sequence is copied to make a messenger RNA ( or mRNA ) molecule in a process known as transcription . The enzyme that builds an mRNA molecule first binds to a start point on a DNA strand , and then uses the DNA sequence to build a ‘pre-mRNA’ molecule until a stop signal is reached . To make the final mRNA molecule , sections called introns are removed from the pre-mRNA molecules , and the parts left behind—known as exons—are then joined together . This process is called splicing . However , it is not fully understood how the splicing process is coordinated with the other stages of transcription . For example , does splicing occur after the pre-mRNA molecule is completed or while it is still being built ? And what controls the order in which these processes occur ? One theory about how the different mRNA-making processes are coordinated is called kinetic competition . This theory states that the fastest process is the most likely to occur , even if the other processes use less energy and so might be expected to be preferred . Alternatively , the different steps may be started and stopped by ‘checkpoints’ that cause the different processes to follow on from each other in a set order . Coulon et al . used fluorescence microscopy to investigate how mRNA molecules are made during the transcription of a human gene that makes a hemoglobin protein . To make the RNA visible , two different fluorescent markers were introduced into the pre-mRNA that cause different regions of the mRNA to glow in different colors . Coulon et al . made the introns fluoresce red and the exons glow green . Unspliced pre-mRNA molecules contain both introns and exons and so fluoresce in both colors , whereas spliced mRNA molecules contain only exons and so only glow with a green color . By looking at both the red and green fluorescence signals at the same time , Coulon et al . could see when an intron was spliced out of the pre-mRNA . This revealed that in normal cells , splicing can occur either before or after the RNA is released from where it is transcribed . Thus , splicing and transcription does not follow a set pattern , suggesting that checkpoints do not control the sequence of events . Instead , the fact that a spliced mRNA molecule can be formed in different ways suggests kinetic competition controls the process . In some cancer cells , there are defects in the cellular machinery that controls splicing . When looking at cells with such a defect , Coulon et al . found that splicing only occurred after transcription was completed . This study thus provides insight into the complex workings of mRNA synthesis and establishes a blueprint for understanding how splicing is impaired in diseases such as cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2014 | Kinetic competition during the transcription cycle results in stochastic RNA processing |
Escherichia coli DNA polymerase V ( pol V ) , a heterotrimeric complex composed of UmuD′2C , is marginally active . ATP and RecA play essential roles in the activation of pol V for DNA synthesis including translesion synthesis ( TLS ) . We have established three features of the roles of ATP and RecA . ( 1 ) RecA-activated DNA polymerase V ( pol V Mut ) , is a DNA-dependent ATPase; ( 2 ) bound ATP is required for DNA synthesis; ( 3 ) pol V Mut function is regulated by ATP , with ATP required to bind primer/template ( p/t ) DNA and ATP hydrolysis triggering dissociation from the DNA . Pol V Mut formed with an ATPase-deficient RecA E38K/K72R mutant hydrolyzes ATP rapidly , establishing the DNA-dependent ATPase as an intrinsic property of pol V Mut distinct from the ATP hydrolytic activity of RecA when bound to single-stranded ( ss ) DNA as a nucleoprotein filament ( RecA* ) . No similar ATPase activity or autoregulatory mechanism has previously been found for a DNA polymerase .
DNA polymerase V is a low fidelity TLS DNA pol with the capacity to synthesize DNA on a damaged DNA template ( Reuven et al . , 1999; Tang et al . , 1999 ) . It is encoded by the LexA-regulated umuDC operon and is induced late in the SOS response in an effort to restart DNA replication in cells with heavily damaged genomes ( Goodman , 2002 ) . The enzyme is responsible for most of the genomic mutagenesis that classically accompanies the SOS response ( Friedberg et al . , 2006 ) . RecA protein plays a complex role in the induction of pol V . As a filament formed on DNA ( the form sometimes referred to as RecA* ) , RecA* facilitates the autocatalytic cleavage of the LexA repressor ( Little , 1984; Luo et al . , 2001 ) . This leads directly to the induction of the SOS response . Some 45 min after SOS induction , those same RecA* filaments similarly facilitate the autocatalytic cleavage of UmuD2 protein to its shorter but mutagenically active form UmuD′2 ( Burckhardt et al . , 1988; Nohmi et al . , 1988; Shinagawa et al . , 1988 ) . UmuD′2 then interacts with UmuC to form a stable UmuD′2C heterotrimeric complex ( Woodgate et al . , 1989; Bruck et al . , 1996; Goodman , 2002 ) . UmuD′2C copies undamaged DNA and performs TLS in the absence of any other E . coli pol ( Tang et al . , 1998 ) , but only when RecA* is present in the reaction . UmuD′2C ( Tang et al . , 1998; Karata et al . , 2012 ) or UmuC ( Reuven et al . , 1999 ) , have minimal pol activity in the absence of RecA* . Final conversion of the UmuD′2C complex to a highly active TLS enzyme requires the transfer of a RecA subunit from the 3′ end of the RecA* filament to form UmuD′2C-RecA-ATP , which we refer to as pol V Mut ( Jiang et al . , 2009 ) . ATP plays an essential but heretofore enigmatic role in the activation process . Activation can proceed with ATP or the poorly-hydrolyzed analogue ATPγS . ATP is part of the active complex , with approximately one molecule of ATP per active enzyme ( Jiang et al . , 2009 ) . Under some conditions , activated and isolated pol V Mut exhibited polymerase activity only when additional ATP or ATPγS was added to the reaction ( Jiang et al . , 2009 ) . The function of the ATP complexed with pol V Mut is delineated in this report .
Pol V Mut can be formed and effectively isolated by incubating UmuD′2C complexes with RecA* that is bound to ssDNA oligonucleotides tethered to streptavidin-coated agarose beads , spinning the beads out of solution to remove RecA* , and taking the now active pol V Mut from the supernatant . In this initially described protocol ( Jiang et al . , 2009 ) , a small amount of ATP or ATPγS is transferred adventitiously from the supernatant with the pol V Mut . When WT RecA protein is used in this activation , the isolated pol V Mut WT is active only if supplemental ATP or ATPγS is added to the reaction mixtures ( Jiang et al . , 2009 ) . However , when a RecA variant that provides more facile activation of pol V was used , RecA E38K/ΔC17 , the added ATP or ATPγS was apparently not needed for pol V Mut function ( Jiang et al . , 2009 ) . The reason for this disparity in the ATP requirement for pol V Mut function is resolved below . To explore the role of ATP , an amended protocol ( outlined in Figure 1A ) was used that employed a spin column to more rigorously remove exogenous ATP or ATPγS after pol V Mut formation . As shown in Figure 1B , pol V Mut function now depends completely on added ATP or ATPγS when this activation protocol was utilized , regardless of which RecA variant was used in the activation . Pol V Mut is not activated by GTP , ADP or dTTP , ( Figure 1—figure supplement 1A ) and does not incorporate ATP or ATPγS into DNA during synthesis ( Figure 1—figure supplement 1B ) . Thus , ATP or an ATP homolog is an absolute requirement for pol V Mut function . For pol V Mut WT , the addition of ATPγS supports synthesis , whereas ATP does not ( Figure 1B ) . The same ATP effect was observed for pol V Mut WT synthesis on DNA containing an abasic site ( Figure 1—figure supplement 2 ) . dATP activation does not result in appreciable DNA extension ( Figure 1—figure supplement 1 ) . For pol V Mut E38K/K72R , synthesis is observed with either ATPγS , ATP or dATP ( Figure 1B , Figure 1—figure supplement 1 ) . Pol V Mut E38K/ΔC17 can also use ATP , ATPγS or dATP as a required nucleotide cofactor ( Figure 1B , Figure 1—figure supplement 1 ) . Notably , the requirement for ATPγS/ATP was completely masked in earlier studies of transactivation of pol V by RecA* filaments that remain in the solution with pol V Mut , because the ATPγS or ATP needed to form RecA* is always present in the transactivation reaction ( Schlacher et al . , 2006 ) . 10 . 7554/eLife . 02384 . 003Figure 1 . Pol V Mut requires ATP/ATPγS for activity . Sketch of pol V Mut assembly; pol V is activated by RecA* bound to Cyanogen-Bromide Sepharose resin as described in ‘Materials and methods’ . The pol V Mut assembly protocol ensures the separation of pol V Mut from free RecA , ssDNA and ATPγS . ( B ) Pol V Mut ( 400 nM ) activity was detected on 5′-32P-labeled 3 nt oh HP ( 100 nM ) in the presence or absence of ATP/ATPγS and dNTPs . To detect free RecA in the pol V Mut solution , ssDNA and ATPγS was added to the reaction . Comparable activity levels between ATPγS alone and ATPγS + ssDNA indicate that pol V Mut is intact and free of RecA . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 00310 . 7554/eLife . 02384 . 004Figure 1—figure supplement 1 . Pol V Mut is not activated by GTP , ADP , or dTTP and does not incorporate ATP/ATPγS in to DNA during synthesis . Pol V Mut WT , pol V Mut E38K/K72R , and pol V Mut E38K/ΔC17 were assembled according to the protocol in Figure 1A . ( A ) ATP , dATP , GTP , ADP , dTTP or ATPγS ( 500 μM ) were used to activate pol V Mut for DNA synthesis and activity was checked in the presence of dNTPs ( 500 μM ) and 3 nt oh HP ( 50 nM ) . ( B ) A HP containing TTT as its 3 nt oh was employed to determine if the various pol V Muts insert ATP or ATPγS during DNA synthesis . DNA extension is only observed in reactions where dATP is included . Other dNTPs are not present in any of the reactions . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 00410 . 7554/eLife . 02384 . 005Figure 1—figure supplement 2 . Pol V Mut WT activity on DNA containing an abasic site . Pol V Mut ( 400 nM ) activity was detected on 5′-32P-labeled 12 nt oh HP ( 100 nM ) , containing an abasic site 3 nts upstream from the 3′-OH , in the presence or absence of ATP/ATPγS and dNTPs . Lesion bypass is only observed when ATPγS is present . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 005 Pol V Mut does not simply require ATP or ATPγS for activity; it possesses an intrinsic DNA-dependent ATPase activity ( Figure 2 ) . This is unprecedented for a DNA polymerase . A very sensitive ATPase assay , based on the fluorescence of a Pi binding protein , is used in this work . The assay permits observation of the first 5 μM of ATP hydrolyzed , which is limited by the concentration of the fluorescent Pi binding protein found in solution . A 30 nt ssDNA oligomer ( 1 μM ) is present in all reactions . In Figure 2A , results are shown with pol V Mut made with WT RecA protein ( pol V Mut WT ) . WT RecA protein alone exhibits limited stability on short oligonucleotides when present at sub-micromolar concentrations . Limited ATP hydrolysis by RecA WT alone is seen in this assay; the initial filaments produce a burst of ATP hydrolysis and then level off to a much slower rate , presumably reflecting filament binding , dissociation , and slow reassembly . Both phases of the reaction exhibit a dependence on RecA WT protein concentration ( Figure 2A , Figure 2—figure supplement 1A ) . In contrast , after detectable free RecA protein and RecA* have been removed , equivalent concentrations of pol V Mut WT exhibit higher levels of ATP hydrolysis than do similar amounts of RecA alone ( Figure 2A ) . The Pi release rate constant ( kcat ) in the presence of ssDNA , 12 nt and 3 nt over hang ( oh ) Hairpin ( HP ) and in the absence of DNA are summarized in Table 1 . For the calculation of rates see ‘Material and methods’ . 10 . 7554/eLife . 02384 . 006Figure 2 . Pol V Mut is a DNA-dependent ATPase . ( A and B ) ATP hydrolysis by pol V Mut and RecA ( 0 . 1 μM , 0 . 2 μM and 0 . 4 μM each ) was measured using MDCC-PBP ( 5 μM ) in the presence of 30 nt ssDNA ( 1 μM ) and ATP ( 500 μM ) . MDCC-PBP fluorescence increases as Pi is released due to ATP hydrolysis . The measurements were taken at a resolution of 1 point per sec for approximately 1000 s . ( C ) Binding of 400 nM pol V , RecA and pol V Muts to etheno-ATP at varied concentrations was measured using rotational anisotropy . The error bars correspond to 1 SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 00610 . 7554/eLife . 02384 . 007Figure 2—figure supplement 1 . RecA WT and RecA E38K/K72R-dependent ATP hydrolysis . ATP hydrolysis of ( A ) RecA WT and ( B ) RecA E38K/K72R was measured as a function of protein concentration in the presence of 1 µM 30 nt ssDNA and 500 µM ATP . Pi release resulting from ATP hydrolysis was measured as a change in fluorescence of MDCC-PBP ( 5 μM ) . The measurements were taken at 1 point per sec resolution for approximately 1000 s . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 00710 . 7554/eLife . 02384 . 008Table 1 . Pol V Mut and RecA ATP hydrolysis rate constantsDOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 008ATPase kcat ( s−1 ) *pol V Mut WT no DNA ( 1 . 5 ± 0 . 1 ) × 10−3 12 nt oh HP ( 9 . 0 ± 0 . 8 ) × 10−3 3 nt oh HP ( 4 . 3 ± 0 . 1 ) × 10−3 30 nt ssDNA ( 160 ± 5 ) × 10−3RecA WT 30 ssDNA ( 100 ± 5 ) × 10−3pol V Mut E38K/K72R no DNA ( 1 . 7 ± 0 . 2 ) × 10−3 12 nt oh HP ( 4 . 4 ± 0 . 7 ) × 10−3 3 nt oh HP ( 3 . 4 ± 0 . 7 ) × 10−3 30 nt ssDNA ( 17 ± 1 ) × 10−3RecA E38K/K72R ssDNA ( 0 . 6 ± 0 . 1 ) × 10−3†pol V Mut E38K/ΔC17 no DNA ( 7 . 0 ± 1 . 5 ) × 10−3 12 nt oh HP ( 54 ± 9 ) × 10−3 3 nt oh HP ( 46 ± 2 ) × 10−3 30 nt ssDNA ( 90 ± 10 ) × 10−3RecA E38K/ΔC17 30 nt ssDNA ( 120 ± 15 ) × 10−3*kcat is an average of at least three independent measurements; ± SEM . †kcat was measured at 2 . 5 μM concentration for RecA E38K/K72R; ATP hydrolysis was not detectable at lower protein concentrations . In principle , the ATPase properties of pol V Mut might be determined mainly , if not solely , by the properties of its RecA subunit . But that's in fact not the case . To address whether the pol V Mut -associated ATPase activity can be distinguished from the intrinsic DNA-dependent ATPase activity of RecA , we assembled pol V Mut with a RecA ( E38K/K72R ) mutant deficient in ATPase activity ( Gruenig et al . , 2008 ) . Using the same highly sensitive fluorescence-based assay as described for panel 2A , we were unable to detect DNA-dependent ATPase activity for RecA E38K/K72R above background ( Figure 2B ) . However , in contrast , pol V Mut E38K/K72R exhibits substantial ATPase activity . The observed ATPase scales with the concentration of pol V Mut E38K/K72R ( Figure 2B ) , as is also the case for pol V Mut WT ( Figure 2A ) . Therefore , pol V Mut is a DNA-dependent ATPase ( Figure 2 ) . As RecA E38K/K72R protein hydrolyzes little or no ATP on its own ( Figure 2B , Figure 2—figure supplement 1B ) , the pol V Mut activity cannot be attributed to a low level of contamination by RecA protein . A small amount of ATP hydrolysis above background can be observed with RecA E38K/K72R by increasing its concentration to 2 . 5 μM ( 84 nM Pi release/min , Figure 2—figure supplement 1B ) . We previously reported that the active complex of pol V Mut is composed of pol V-RecA-ATP ( Jiang et al . , 2009 ) . We can further distinguish pol V Mut from either pol V or free RecA in an assay measuring binding to etheno-ATP via anisotropy ( Figure 2C ) . Pol V Mut made with either RecA WT or RecA E38K/K72R registers a substantial and concentration-dependent signal in this assay , while binding of etheno-ATP to either pol V or RecA alone is much weaker . To further explore the requirement for the addition of a ribonucleotide cofactor , we measured pol V Mut binding to p/t DNA and DNA synthesis as a function of the concentration of added ATPγS or ATP ( Figure 3 ) . Binding is absolutely dependent on a nucleotide cofactor . Pol V Mut WT binding to p/t DNA with a 3 nt template overhang ( 3 nt oh ) increases roughly linearly up to about 180 μM ATPγS , saturating at about 220 μM ( Figure 3A ) . However , binding is not detectable in the presence of ATP . Increasing the length of the template overhang to 12 nt has essentially no effect on binding as a function of nucleotide concentration ( Figure 3—figure supplement 1 ) . However , with ATPγS pol V Mut binds with higher affinity to the p/t DNA 12 nt oh ( Table 2 ) . DNA synthesis corresponds closely to ATPγS-dependent pol V Mut WT binding , showing a linear increase in primer extension up to about 180 μM ATPγS , reaching about 70% total primer usage at about 500 μM ATPγS ( Figure 3B ) . There is no primer extension with ATP ( Figure 3B ) . Pol V Mut E38K/K72R binds to p/t DNA with ATPγS ( Figure 3C; Table 2 ) . Although it does not appear to bind DNA appreciably in the presence of ATP in this assay , some binding must occur as DNA synthesis is clearly observed ( Figure 3D ) . 10 . 7554/eLife . 02384 . 009Figure 3 . Pol V Mut binding and activity as a function of nucleotide . Binding of pol V Mut WT ( 1 μM ) ( A ) , pol V Mut E38K/K72R ( 400 nM ) ( C ) , and pol V Mut E38K/ΔC17 ( 400 nM ) ( E ) to 3 nt oh HP was measured as a change in rotational anisotropy . Activity of pol V Mut WT ( 400 nM ) ( B ) , pol V Mut E38K/K72R ( 400 nM ) ( D ) , and pol V Mut E38K/ΔC17 ( 400 nM ) ( F ) was quantified on 5′-32P-labeled 3 nt oh HP with varying concentrations of nucleotide and 500 μM dNTPs . A gel showing primer utilization ( %PU ) as a function of nucleotide is presented to the right of the graph . ATP ( open circle ) and ATPγS ( filled circle ) . The error bars correspond to 1 SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 00910 . 7554/eLife . 02384 . 010Figure 3—figure supplement 1 . Pol V Mut binding to 12 nt oh HP DNA as a function of ATP/ATPγS . Binding of ( A ) pol V Mut WT ( B ) pol V Mut E38K/K72R , and ( C ) pol V Mut E38K/ΔC17 ( 400 nM each ) to fluorescein-labeled 12 nt oh HP ( 50 nM ) was measured as a function of ATP ( open circles ) and ATPγS ( filled circles ) . DNA binding was observed as a change in rotational anisotropy . The error bars correspond to 1 SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 01010 . 7554/eLife . 02384 . 011Figure 3—figure supplement 2 . Pol V Mut WT activity as a function of ATPγS concentration in the presence of ATP . Pol V Mut WT ( 400 nM ) DNA synthesis is measured as a function of ATPγS ( filled circles ) on 3 nt oh HP ( 50 nM ) . When ATP ( 500 μM ) is present at each ATPγS concentration , primer utilization of pol V Mut WT is decreased ( open circles ) compared to reactions where no ATP was included . The error bars correspond to 1 SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 01110 . 7554/eLife . 02384 . 012Table 2 . Pol V Mut affinity to 12 nt oh HP DNADOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 012Pol V MutKd ( nM ) ATPATPγSpol V Mut WTnb876 ± 52pol V Mut E38K/K72Rnb920 ± 112polV Mut E38K/ΔC17312 ± 46469 ± 27nb–binding not detected . The same experiments performed with pol V Mut E38K/ΔC17 show that binding and activity correlate well with ATPγS and ATP . Here , a much higher affinity to p/t DNA with ATP ( Table 2 ) allows binding and primer extension ( 17% at 750 μM ATP , Figure 3E , F ) . Pol V Mut E38K/ΔC17 requires less ATPγS ( 78% at 100 μM ATPγS ) for optimal binding and activity compared to the other pol V Muts , corresponding to the more robust activation of pol V consistently seen with this RecA variant ( Schlacher et al . , 2006; Jiang et al . , 2009 ) . The data in Figure 3 establish an absolute requirement for an ATP cofactor for pol V Mut binding to p/t DNA . In no case is DNA binding detected in the absence of ATP or an ATP homolog . Since binding is a precursor to synthesis , the same nucleotide requirement holds for pol V Mut-catalyzed primer extension ( Figures 1B , Figure 3 ) . The link that's missing is the role of ATP hydrolysis in the DNA binding/synthesis process . Indirect evidence hinting at just such a link is the observation that pol V Mut WT hydrolyzes ATP about 9-fold more rapidly on ssDNA than pol V Mut E38K/K72R and about 2-fold more rapidly than pol V Mut E38K/ΔC17 ( Table 1 ) , suggesting perhaps that the more rapid ATP hydrolysis by pol V Mut WT diminishes its ability to remain bound to DNA long enough to catalyze synthesis under these in vitro conditions . Pol V Mut WT-dependent primer extension significantly decreases with a mixture of ATP/ATPγS ( Figure 3—figure supplement 2 ) , further suggesting that the fraction of enzyme that binds and hydrolyzes ATP is not associated with DNA long enough to catalyze appreciable primer elongation . With p/t DNA , ATP hydrolysis is slower for pol V Mut E38K/K72R compared to pol V Mut WT ( Table 1 ) , which could account for incorporation with added ATP . In the case of pol V Mut E38K/ΔC17 , although ATP hydrolysis is more rapid on p/t DNA compared to pol V Mut WT ( Table 1 ) , binding is much tighter ( Table 2 ) , which could plausibly explain primer extension observed with added ATP . To obtain direct evidence for a possible role of ATP hydrolysis in the pol V Mut reaction pathway , we used pol V Mut formed with RecA E38K/ΔC17 , which binds p/t DNA with higher affinity than either pol V Mut WT or pol V Mut E38K/K72R ( measured Kd for pol V Mut E38K/ΔC17 = 469 nM , compared to 876 nM for pol V Mut WT; Table 2 ) . There is a concomitant ∼5 to 10-fold greater DNA-dependent ATPase activity for pol V Mut E38K/ΔC17 compared with pol V Mut WT ( Table 1 ) . The pol V Mut E38K/ΔC17-p/t DNA complex formed in the presence of ATP was stable for a long enough time to conveniently monitor its dissociation ( Figure 4 ) . Fluorescent-labeled p/t DNA ( 12 nt oh , 50 nM ) has a rotational anisotropy ( RA ) of 0 . 05 when diffusing freely in solution ( Figure 4 ) . When incubated in the presence of eightfold excess pol V Mut E38K/ΔC17 ( 400 nM ) , essentially all of the DNA is bound in a complex with pol V Mut ( RA = 0 . 12 ) . Immediately following complex formation ( t ∼ 0 ) , 160-fold excess unlabeled p/t DNA ( 8 μM ) was added to trap pol V Mut as it dissociates . Pol V Mut E38K/ΔC17 forms a stable complex with p/t DNA in the presence of ATPγS , remaining bound for at least 4 min without dissociating . In contrast , when the complex is formed with ATP , pol V Mut E38K/ΔC17 dissociates exponentially as a function of time , asymptotically reaching ∼100% free DNA at about 75 s ( Figure 4 ) . The off-rate determined as the first-order rate constant is 0 . 053 s−1 . The magnitude of the off-rate is in remarkably close agreement with the DNA-dependent ATPase single-turnover rate constant for pol V Mut E38K/ΔC17 = 0 . 054 s−1 ( Table 1 ) . The data strongly suggest that ATP hydrolysis is responsible for pol V Mut-p/t DNA dissociation . This can also be seen during DNA synthesis , where longer segments of DNA are synthesized with ATPγS than with a similar concentration of ATP ( Figure 4—figure supplement 1 ) . The correspondence of dissociation rate to ATP hydrolysis rate further suggests that perhaps a single ATP turnover is sufficient to trigger dissociation of pol V Mut from a 3′-OH primer end . The absence of dissociation in the presence of the weakly hydrolyzed ATPγS strongly reinforces this conclusion . Pol V Mut WT and pol V Mut E38K/K72R , act similarly to pol V Mut E38K/ΔC17 , remaining stably bound to p/t DNA in the presence of ATPγS ( Figure 4—figure supplement 2 ) . Although ATP hydrolysis affects pol V Mut–DNA complex stability , the integrity of the pol V Mut protein complex is not affected by hydrolysis , with RecA remaining bound to UmuD′2C ( Figure 4—figure supplement 3A , B ) . 10 . 7554/eLife . 02384 . 013Figure 4 . Pol V Mut E38K/ΔC17 off-rate in the presence of ATP and ATPγS . Fluorescence depolarization of fluorescein-labeled ( 12 nt oh HP ) DNA was used to measure the dissociation constant of pol V Mut E38K/ΔC17 in the presence of ATP ( filled circles ) and ATPγS ( open circles ) . A stable protein–DNA complex ( 400 nM and 50 nM , respectively ) was preformed in the presence of nucleotide followed by the addition of excess ( 160 times ) trap DNA ( unlabeled 12 nt oh HP ) . The decrease in anisotropy over time was fit to an exponential decay to determine koff ( 0 . 053 ± 0 . 0025 s−1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 01310 . 7554/eLife . 02384 . 014Figure 4—figure supplement 1 . Pol V Mut E38K/ΔC17 primer extension length as a function of ATP or ATPγS concentration . DNA extension was measured with varying concentrations of ATP ( left panel ) and ATPγS ( right panel ) on 12 nt oh HP to determine the lengths of primer synthesized as a function of NTPs . The unextended primer is indicated as HP and the addition of a nucleotide is marked as nt 1–4 and nt 12 in the center of the panels . The length of DNA synthesized by pol V Mut E38K/ΔC17 increases with the concentration of ATP/ATPγS . Pol V Mut E38K/ΔC17 is far more active with ATPγS compared with ATP , which is consistent with in inability to dissociate from p/t DNA in the absence of ATP hydrolysis . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 01410 . 7554/eLife . 02384 . 015Figure 4—figure supplement 2 . Dissociation of pol V Mut WT and pol V Mut E38K/K72R in the presence of ATPγS . ( A ) pol V Mut WT and ( B ) pol V Mut E38K/K72R ( 400 nM ) were pre-bound to fluorescein-labeled 12 nt oh HP ( 50 nM ) in the presence of ATPγS . Excess trap DNA ( 8 μM unlabeled 12 nt oh HP ) was added to the stable protein–DNA complex and DNA binding of pol V Mut was monitored as a change in rotational anisotropy ( circles ) . Fluorescein-labeled 12 nt oh HP alone is indicated as a black line . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 01510 . 7554/eLife . 02384 . 016Figure 4—figure supplement 3 . Pol V Mut remains intact in the presence of ATP/ATPγS and during DNA synthesis . RecAF21AzF-Alexa Fluor 488 was used to form RecA* during pol V Mut assembly resulting in the formation of fluorescently-labeled pol V Mut ( pol V Mut WTF21AzF-Alexa Fluor 488 ) . ( A ) The rotational anisotropy of pol V Mut WTF21AzF-Alexa Fluor 488 ( 100 nM ) was measured under ATP hydrolysis and DNA synthesis conditions ( ‘Materials and methods’ ) . No change was observed upon addition of ATP/ATPγS , 12 nt oh HP ( 1 μM ) or during DNA synthesis ( 2 , 5 and 10 min ) indicating that pol V Mut remains intact . The anisotropy of free RecAF21AzF-Alexa Fluor 488 ( 100 nM ) was measured in parallel . The error bars correspond to 1 SEM . ( B ) DNA synthesis was measured for pol V Mut WT and pol V Mut WTF21AzF-Alexa Fluor 488 demonstrating that intact enzyme extends DNA under experimental conditions ( pol V Mut: 100 nM and HP 1 μM ) at the time periods ( 2 , 5 , 10 , 20 , and 30 min ) used for rotational anisotropy measurements ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 016 The failure of pol V Mut WT to synthesize DNA in the presence of ATP can presumably be attributed to its inability to form a sufficiently stable complex with p/t DNA ( Figure 3A ) . In contrast , pol V Mut E38K/ΔC17 , which forms a much more stable complex with p/t DNA can synthesize DNA in the presence of either ATP or ATPγS ( Figure 3F ) . To determine if pol V Mut WT can synthesize DNA in the presence of ATP ( as must presumably happen in vivo ) we have used the β sliding clamp to enhance pol V Mut binding to the 3′-primer end . Pol V Mut WT binds to β clamp and is able to synthesize DNA with moderately high processivity ( Karata et al . , 2012 ) . To determine if increased binding stability for pol V Mut WT might enable it to use ATP for synthesis , we used a p/t DNA with a 12 nt oh-containing streptavidin attached at the 5′-template end and the hairpin loop , preventing the β clamp from sliding off the DNA ( Schlacher et al . , 2006 ) . Pol V Mut WT in the presence of ATP ( in the absence of ATPγS ) incorporates 12 nt processively to reach the end of the template strand ( Figure 5 ) . The presence of ATP in the reaction , which is required to load β clamp ( Bertram et al . , 2000 ) , is able to support DNA synthesis . However , the addition of ATPγS significantly stimulates synthesis ( Figure 5 ) , presumably by further enhancing binding to p/t DNA ( Figure 5; Table 2 ) , and by reducing the unloading of β clamp by the γ complex ( Bertram et al . , 2000 ) . 10 . 7554/eLife . 02384 . 017Figure 5 . Pol V Mut WT is active with ATP only in the presence of β/γ complex . Sketch of the experimental set up is illustrated above the gels . To prevent the β clamp from sliding off the DNA , a 12 nt oh HP was designed containing biotin/streptavidin on both sides of the primer terminus substrate . The activity of pol V Mut WT was measured in the presence and absence of β/γ . In the presence of β/γ complex ( left panel ) pol V Mut WT is able to extend p/t with ATP , in contrast no DNA synthesis is observed with ATP in the absence of β/γ complex ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 017
Mutagenic DNA synthesis during the SOS response is an act of cellular desperation , and it comes with a price . However , when pol V Mut is activated and arrives on the scene , mutagenesis is not indiscriminate . Pol V Mut possesses a novel mechanism with which to limit processivity and restrict mutagenic DNA synthesis to those short DNA segments where it is required . In essence , the enzyme has evolved to do the absolute minimum to get cellular DNA synthesis restarted . No other DNA polymerase characterized to date has either an intrinsic ATPase activity or a similar autoregulatory mechanism . We have previously shown that the active form of DNA polymerase V is UmuD′2C-RecA-ATP ( Jiang et al . , 2009 ) , but the roles of ATP and RecA in polymerase function were unknown . In vitro results using RecA variants that provide more facile activation and/or lack intrinsic ATPase function now elucidate the role of ATP . ATP or an ATP homologue is required for pol V Mut function ( Figure 1B ) . Further , pol V Mut is a DNA-dependent ATPase ( Figure 2A , B ) , which binds ATP in the absence of DNA ( Figure 2C ) . Hydrolysis of ATP by pol V Mut results in dissociation of the enzyme from DNA ( Figure 4 ) . If ATPγS is used such that ATP is not hydrolyzed , then the enzyme remains stably bound to DNA ( Figure 4 , Figure 4—figure supplement 2 ) . This is the only DNA polymerase studied to date that is regulated by ATP binding and hydrolysis . Activation of pol V to pol V Mut requires transfer of a RecA monomer from the 3′-proximal tip of RecA* ( Schlacher et al . , 2006 ) ( see e . g . , Figure 1A ) . The ATP hydrolytic sites in RecA protein filament are situated at the interfaces of adjacent RecA subunits , and each neighboring subunit contributes key residues to the active site . There are no readily identifiable ATP-binding motifs present in either the UmuD′ or UmuC subunits of pol V . The avid ATPase activity observed with the ATPase-deficient RecA E38K/K72R indicates that the RecA subunit is not contributing a Walker A motif ( or P-loop ) to the aggregate site . We speculate that the ATPase active site in pol V Mut is newly created when RecA is added to the complex during activation , perhaps at the interface between the RecA subunit and UmuC or UmuD′ . This interface appears to play a central role in polymerase activation because the RecA ( S117F ) mutant ( initially referred to as RecA1730 ) , was shown to be SOS-non mutable in vivo ( Dutreix et al . , 1989 ) . The S117 residue is situated at the 3′-proximal tip of RecA* ( Sommer et al . , 1998 ) , and , notably , pol V Mut S117F has no measurable DNA synthesis activity in vitro ( Schlacher et al . , 2005 ) . In the cell pol V is post-transcriptionally regulated through proteolysis ( Frank et al . , 1993; Gonzalez et al . , 1998 , 2000 ) and by RecA* ( Burckhardt et al . , 1988; Nohmi et al . , 1988; Shinagawa et al . , 1988; Schlacher et al . , 2006 ) , presumably to ensure that this low fidelity pol ( Reuven et al . , 1999; Tang et al . , 1999 ) is used only in dire circumstances . When the regulation of pol V fails , as it does for the constitutive RecA E38K mutant , pol V Mut is induced in the absence of DNA damage generating ∼100-fold increase in mutations ( Witkin , 1967 ) . This huge increase in mutations likely occurs by copying undamaged DNA with exceptionally poor fidelity ( Tang et al . , 2000 ) . The regulation of pol V is needed to limit mutations , especially in stationary phase cells ( Corzett et al . , 2013 ) . Our biochemical data suggest that in vivo ATP adds another level to this regulation; not only is the timing of pol V activity regulated , but also its access to p/t DNA . A model for the role of ATP in pol V Mut activity is sketched in Figure 6 . ATP is required to bind pol V Mut to DNA . ATP hydrolysis releases the enzyme from DNA , which in vivo would ensure that tracts of DNA synthesized by pol V Mut are short , limiting the opportunity for misincorporation to the region immediately adjacent to the template lesion . Therefore , the internally regulated DNA-dependent ATPase of pol V Mut provides a way to limit mutational load . 10 . 7554/eLife . 02384 . 018Figure 6 . Model showing ATP regulation of pol V Mut activity . Pol V Mut is active for DNA synthesis only after binding a molecule of ATP ( green triangle ) to form UmuD′2C-RecA-ATP . The binding of ATP is required for polymerase association with p/t DNA . ATP-hydrolysis catalyzed by an intrinsic DNA-dependent ATPase triggers pol V Mut-p/t DNA dissociation , while leaving intact the UmuD′2C-RecA complex . DOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 018
DNA oligos were synthesized using a 3400 DNA synthesizer ( Applied Biosystems/Life Technologies , Carlsbad , CA ) . Oligo modifications ( Flourescein-dT Phosphoramidite , Biotin-dT , and 5′-Amino-Modifier C12 ) were purchased from Glen Research . DNA sequences are in Table 3 . 10 . 7554/eLife . 02384 . 019Table 3 . Sequences for p/t HP DNADOI: http://dx . doi . org/10 . 7554/eLife . 02384 . 019Sequences for p/t HP DNA 3 nt oh HP5′ AGA GCA GTT AGC GCA TTC AGC TCA TAC TGC TGA ATG CGC TAA CTG C 3′ 3 nt oh HP ( TTT ) 5′ TTT GCA GTT AGC GCA TTC AGC TCA TAC TGC TGA ATG CGC TAA CTG C 3′ Fluorescein ( FAM ) 3 nt oh HP5′ AGA GCA GTT AGC GCA T ( FAM ) C AGC TCA TAC TGC TGA ATG CGC TAA CTG C 3′ 12 nt oh HP5′ CGA AAC AGG AAA GCA GTT AGC GCA TTC AGC TCA TAC TGC TGA ATG CGC TAA CTG C 3′ Fluorescein ( FAM ) 12 nt oh HP5′ CGA AAC AGG AAA GCA GTT AGC GCA TTC AGC TCA TAC TGC TGA A ( FAM ) G CGC TAA CTG C 3′ Biotinylated ( Bio ) 12 nt oh HP5′ ( Bio ) GA AAC AGG AAA GCA GTT AGC GCA TTC AGC ( Bio ) CA TAC TGC TGA ATG CGC TAA CTG C 3′Sequences for pol V Mut assembly , activity and ATP hydrolysis DNA Amino C12-linked 45mer attached to cyanogen bromide-activated sepharose resin5′ C12TT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT TTT T 3′ ssDNA 30 mer for RecA transactivation and ATP hydrolysis5′ ACT GAC CCC GTT AAA ACT TAT TAC CAG TAA 3′ His-tagged pol V was purified from E . coli stain RW644 as described by Karata et al . ( 2012 ) and RecA WT was purified by a standard protocol ( Cox et al . , 1981 ) . RecA E38K/K72R and RecA E38K/ΔC17 were provided by Michael Cox at the University of Wisconsin , Madison . We incorporated p-azido-L-phenylalanine ( pAzF ) ( Chin et al . , 2002 ) into RecA WT to site specifically label the protein with Alexa Fluor 488 DIBO alkyne . For the cloning of RecA , we used the pAIR79 plasmid , which was a gift from Michael Cox at the University of Wisconsin , Madison . The Phe21 sequence in RecA WT was replaced with the amber codon via site-directed mutagenesis using Pfu Ultra polymerase ( Agilent Technologies ) . Once the sequence was confirmed , pAIR79 was cotransformed with the vector pEVOL-pAZF ( a gift from the Peter Schultz lab at The Scripps Research Institute , San Diego , CA ) into the BLR expression strains ( Young et al . , 2010 ) . The RecAF21AzF protein was purified using the same standard protocol for RecA WT ( Cox et al . , 1981 ) . RecAF21AzF was labeled with Alexa Flour 488 ( RecAF21AzF-Alexa Fluor 488 ) according to manufacturer's instructions ( Life Technologies ) . 5′ amino-modified 45 nt ssDNA was covalently attached to Cyanogen-Bromide Sepharose resin according to manufacturer's protocol ( Sigma–Aldrich ) and transferred to a spin column ( Biorad ) . Briefly , the resin was activated with 1 mM HCl then washed with water and equilibrated with coupling buffer ( 0 . 1 M NaHCO3 , 0 . 5 M NaCl ) . 5′amino-modified DNA ( 20 nmoles ) was incubated with 100 mg resin overnight at 4°C , and unbound oligomers were removed by washing resin seven times with coupling buffer . Reactive amino groups were blocked with Ethanolamine ( 1 M , pH 8; Sigma–Aldrich ) to prevent non-specific binding then stored in 1 M NaCl . The concentration of 45 nt ssDNA bound to the resin consistently provided about 6 nmoles per 100 mg of resin . Stable RecA* was assembled by incubating excess RecA or RecA mutants and ATPγS ( Roche ) or ATP ( Amersham-Pharmacia ) when stated with 0 . 5 nmole ssDNA-bound resin for 20 min at 37°C . Free RecA and ATPγS were separated from RecA*-resin by gentle centrifugation at 0 . 1×g for 1 min and collected in the flow through . Washes were repeated with reaction buffer ( 20 mM Tris–HCl pH 7 . 5 , 25 mM Sodium Glutamate , 8 mM MgCl2 , 8 mM DTT , 4% glycerol , 0 . 1 mM EDTA ) until no RecA was detected in the flow through . His-tagged pol V ( 2 nmole ) was resuspended in reaction buffer and mixed with RecA*-resin to form pol V Mut . The pol V-RecA*-resin suspension was incubated at 37°C for 15 min followed by centrifugation ( 0 . 1×g for 1 min ) to separate pol V Mut from the RecA*-resin in the spin column . The concentration of pol V Mut collected in the flow through was determined by SDS-PAGE gel . Briefly , pol V and RecA proteins at various known concentrations were resolved on an SDS-PAGE gel ( 10% ) as standards and gel band intensities were quantified using IMAGEQUANT software . Standard gel intensities of UmuC and RecA were then used to determine the concentration of pol V Mut . The activity of pol V Mut was detected via DNA extension on a 5′-32P-labeled primer template hairpin with a 3-nt overhang . Pol V Mut ( 400 nM ) was added to a 10-μl reaction mixture containing annealed template DNA ( 50 nM ) , ATP , ATPγS , dATP , GTP , dTTP , or ADP ( 500 μM , unless stated otherwise ) and dNTPs ( Amersham-Pharmacia ) ( 500 μM each ) . Substrate DNA was preincubated with first streptavidin ( 400 nM ) then β/γ complex ( 250 nM and 100 nM respectively ) ( a gift from Linda Bloom at the University of Florida , Gainesville ) when present . Reactions were carried out at 37°C for 30 min . To detect free RecA in the pol V Mut solution , ssDNA ( 50 nM ) was added to the reaction and activity was measured in the presence of ATPγS . Comparable activity levels between ATPγS alone and ATPγS + ssDNA indicate that pol V Mut is intact and free of RecA that is not part of the mutasome . Reactions were resolved on a 20% denaturing polyacrylamide gel allowing single nucleotide separation . Gel band intensities were detected by phosphorimaging and quantified using IMAGEQUANT software . Primer utilization was calculated as the unextended primer intensity subtracted from the total DNA intensity giving the percentage of primer utilized . E . coli phosphate-binding protein ( PBP ) was purified and labeled with MDCC fluorophore ( Life Technologies ) according to the protocol from Brune et al . ( 1998 ) . Binding of Pi ( phosphate ) by MDCC-PBP is rapid and tight ( Kd ∼ 0 . 1 µM ) resulting in a large increase in fluorescence ( Brune et al . , 1998 ) . The change in fluorescence of MDCC-PBP was detected in real-time using a QuantaMaster ( QM-1 ) fluorometer ( Photon Technology International ) . Wavelengths for excitation and emission of the MDCC were selected using monochromators with a 1-nm band pass width . Excitation and emission were set at 425 and 464 nm , respectively . A 65-μl aliquot of 5 μM MDCC-PBP was incubated with 0 . 05 units/ml PNPase , 100 μM 7-methylguanosine and various concentrations of pol V Mut or RecA . The PNPase and 7-methylguanosine were used to remove any traces of Pi in the reaction prior ATP hydrolysis . The time-based scan was initiated for about 1000 s . ATP hydrolysis was initiated by adding a 5-μl mixture of ATP and DNA to final concentrations of 500 μM and 1 μM , respectively and the measurements were taken at 1 point per sec resolution . The maximum rate ( Vmax ) of Pi release was derived from the linear slope of Pi release , and kcat was calculated by dividing Vmax by the enzyme concentration . Each measurement was repeated 2–3 times . Pol V Mut binding to p/t DNA was measured by changes in steady-state fluorescence depolarization ( rotational anisotropy ) . Reactions ( 70 μl ) were carried out at 37°C in standard reaction buffer , and contained a fluorescein-labeled hairpin DNA ( 50 nM ) , 500 μM ATP or ATPγS and varied concentrations of pol V Mut . For ATP titration experiments , fluorescein-labeled hairpin DNA ( 50 nM ) and pol V Mut were mixed together in reaction buffer and ATP or ATPγS was titered to a final concentration of 750 μM . Rotational anisotropy was measured using a QuantaMaster ( QM-1 ) fluorometer ( Photon Technology International ) with a single emission channel . Samples were excited with vertically polarized light at 495 nm and both vertical and horizontal emission was monitored at 520 nm . To measure the enzyme's off-rate , first pol V Mut E38K/ΔC17 was prebound to 50 nM fluorescein-labeled hairpin DNA with a 12 nt single-stranded overhang in the presence of 500 μM ATP or 500 μM ATPγS . The enzyme off-rate was measured by monitoring changes in rotational anisotropy after the addition of excess trap , unlabeled DNA ( 8 μM ) . The off-rate was fit to single exponential decay . Anisotropy for free DNA and pol V Mut E38K/ΔC17 in the absence of trap was measured in the same experiment . To determine the integrity of pol V Mut during ATP hydrolysis and DNA synthesis we employed the use of site-specifically labeled RecAF21AzF-Alexa Fluor 488 to assemble pol V Mut WTF21AzF-Alexa Fluor 488 . Pol V Mut WTF21AzF-Alexa Fluor 488 ( 100 nM ) was mixed with dNTPs ( 500 μM ) to a final volume of 70 μl and the rotational anisotropy was measured . To the same cuvette , ATP or ATPγS ( 500 μM ) was added for another measurement . To create DNA synthesis conditions , 12 nt oh HP ( 1 μM ) was added to the cuvette and rotational anisotropy was measured at 2 min , 5 min , and 10 min . The rotational anisotropy of RecAF21AzF-Alexa Fluor 488 ( 100 nM ) was measured in parallel to pol V Mut WTF21AzF-Alexa Fluor 488 . Pol V Mut binding to etheno-ATP ( Life Technologies ) was measured as a change in rotational anisotropy at three different etheno-ATP concentrations , 500 μM , 750 μM , and 1000 μM . In a 70-μl reaction , etheno-ATP was mixed in standard reaction buffer with pol V Mut , pol V or RecA . The concentration of protein used for measurements was 400 nM . Rotational anisotropy was measured using a QuantaMaster ( QM-1 ) fluorometer . Samples were excited with vertically polarized light at 410 nm and both vertical and horizontal emissions were monitored at 425 nm . | DNA polymerases are enzymes that copy the genetic material within a cell using a strand of an existing double helix as a template to guide the synthesis of a new DNA strand . Since DNA is continuously exposed to damaging agents , and damaged DNA can derail the DNA replication machinery; a cell must either repair or bypass ( via a process called translesion synthesis ) this damage to copy its genome . Most living things , from bacteria to humans , have specific DNA polymerases for translesion synthesis that can copy past damaged DNA , such as DNA Polymerase V in the bacterium E . coli . However , DNA polymerase V will also often introduce mistakes when it copies DNA that is not damaged—and cells will subsequently switch to use a different polymerase to accurately copy undamaged DNA . DNA polymerase V is activated by binding to a protein called RecA and a molecule of adenosine triphosphate ( ATP for short ) . ATP stores energy , which cells release by breaking down the molecule into simpler chemicals: but how do ATP and RecA work together to activate this polymerase ? Now , Erdem , Jaszczur et al . have addressed this question using biochemical techniques on purified polymerases , proteins and DNA fragments in a test tube . These experiments reveal that DNA polymerase V must bind to an ATP molecule before it can attach to the DNA template , and must remain bound to ATP while synthesizing the new DNA strand . After the activated polymerase has attached to the DNA , it will break down the molecule of ATP to free itself from the DNA . Furthermore , although the RecA protein can also break down ATP , Erdem , Jaszczur et al . found that a mutant RecA without this ability could still activate DNA polymerase V to break down this molecule itself . Binding to and breaking down a molecule of ATP by a DNA polymerase has not been observed before as a method of directly regulating these enzymes' activity . Erdem , Jaszcuzur et al . suggest that , in living cells , this extra level of control would limit how long the DNA polymerase V spends attached to the DNA . As such , this polymerase would only be used to copy stretches of damaged DNA , but would not continue on to copy neighboring stretches of undamaged DNA where it would likely introduce new errors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | DNA polymerase V activity is autoregulated by a novel intrinsic DNA-dependent ATPase |
Queens of social insects make all mate-choice decisions on a single day , except in honeybees whose queens can conduct mating flights for several days even when already inseminated by a number of drones . Honeybees therefore appear to have a unique , evolutionarily derived form of sexual conflict: a queen’s decision to pursue risky additional mating flights is driven by later-life fitness gains from genetically more diverse worker-offspring but reduces paternity shares of the drones she already mated with . We used artificial insemination , RNA-sequencing and electroretinography to show that seminal fluid induces a decline in queen vision by perturbing the phototransduction pathway within 24–48 hr . Follow up field trials revealed that queens receiving seminal fluid flew two days earlier than sister queens inseminated with saline , and failed more often to return . These findings are consistent with seminal fluid components manipulating queen eyesight to reduce queen promiscuity across mating flights .
Seminal fluid is a complex mixture of proteins and metabolites with multiple functions to enhance male reproductive success ( Poiani , 2006; Avila et al . , 2011 ) . It keeps sperm alive and motile , protects against pathogens , and regulates sperm capacitation , the final maturation step that enables sperm to fertilise eggs ( Chapman , 2001; Poiani , 2006; Otti et al . , 2009 ) . When females mate with multiple males , seminal fluid components can become agents of sexual selection and harm rival ejaculates while others manipulate female physiology to enhance a specific male’s reproductive success ( Parker , 1970; Birkhead and Møller , 1998; Chapman et al . , 2003a; den Boer et al . , 2010 ) . These interactions have been documented in detail in the fruit fly Drosophila melanogaster where seminal fluid promotes fast oviposition and reduces the willingness of females to seek additional copulations ( Chen et al . , 1988; Liu and Kubli , 2003; Chapman et al . , 2003b ) . A key molecule responsible for these effects is the sex peptide , a seminal fluid peptide that crosses the vaginal wall to enter the hemolymph and bind to a G-protein-coupled receptor on neurons that activate a signaling transduction cascade ( Yapici et al . , 2008 ) . Similar phenotypic effects have been reported in other insects , but without a detailed understanding of the molecular mechanisms involved ( e . g . Craig , 1967; Gillott and Langley , 1981; Baer et al . , 2001; Hayashi and Takami , 2014 ) . Recent proteomic characterizations of seminal fluid in honeybees revealed a number of proteins with the potential to interact with neurons ( Baer et al . , 2009; Grassl et al . , 2017 ) , suggesting that the seminal fluid of honeybee males ( known as drones ) might be able to manipulate queen mating behaviour . Several decades of research on ants , social bees and social wasps have shown that obligate multiple insemination of queens is always evolutionarily derived from single paternity ancestors ( Hughes et al . , 2008; Boomsma , 2013 ) . These studies imply that obligate polyandry evolved predominantly in lineages with large and long-lived colonies where genetically diverse workers make colonies more likely to survive and reproduce ( Mattila and Seeley , 2007; Mattila et al . , 2012 ) . Social insect males mating with a focal queen will benefit from shared paternity if that is their only route to reproductive success , but the optimal number of inseminations are expected to be higher for a queen than for the males inseminating her ( Koeniger , 1990 ) . This form of sexual conflict is unlikely to affect flight behaviour when queens depart as virgins and inseminations follow each other in quick succession . If queens never fly again , it is thus reasonable to assume they will store a genetically diverse fraction of all sperm received to maximise life-time reproductive success ( Jaffé et al . , 2012 ) while suppressing sperm competition between ejaculates ( Starr , 1984; Boomsma et al . , 2005 ) , a sequence of events that is increasingly well documented ( Mattila and Seeley , 2007; den Boer et al . , 2010; Mattila et al . , 2012 ) . Honeybees are the only social insect lineage so far known to deviate from this rule because a significant fraction of queens embark on additional flights on subsequent days when they are no longer virgins ( Woyke , 1964; Tan et al . , 1999 ) . Newly mated honeybee queens return to their hives and complete the process of sperm storage over several days , which corresponds with a physiological transition to become an established egg-laying mother queen ( Woyke , 1983; Winston , 1987 ) . If queens decide during this brief time period to leave for a second or third risky mating flight , their choice will reduce the fitness of drones whose sperm she has already acquired but not yet stored . Previous studies of the brains , ovaries and fat bodies of Apis mellifera queens demonstrated that complex physiological changes occur in response to mating , and that they are controlled by multiple , largely uncorrelated , mechanisms ( Koeniger , 1976; Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2009; Kocher et al . , 2010; Niño et al . , 2011; Vergoz et al . , 2012; Niño et al . , 2013a; Niño et al . , 2013b; Manfredini et al . , 2015; Brutscher et al . , 2019 ) . For example , both artificial insemination procedures and manipulation of insemination volume trigger ovary activation through mechanical stimuli of stretch receptors in the genital tract that also appear to affect a queen’s decision to pursue additional mating flights , independent of the type of substance transferred to the queen’s sexual tract ( Kocher et al . , 2009; Niño et al . , 2011 ) . Queen exposure to carbon dioxide induces similar changes as the ones induced by copulation , for example reducing mating flights number , triggering ovary development and changes in chemical composition of mandibular gland secretion , or altering gene expression in the brain ( Niño et al . , 2011; Vergoz et al . , 2012; Niño et al . , 2013b ) . These studies also found consistently altered expression of genes with known links to immune responses and visual perception of queens after mating ( Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2010; Manfredini et al . , 2015 ) . However , the factors inducing these latter effects have remained elusive and the ensuing reductions in visual perception have neither been phenotypically verified nor been interpreted as possible consequences of sexual conflict . Our present study provides an explicit test of the hypothesis that ejaculates contain molecules that affect the neurophysiology and behaviour of queens in such a way that these queens are less likely to acquire additional matings during subsequent mating flights . Unconstrained visual perception by queens is important for a successful return to their hives , but also expected to be essential for locating drone congregation areas during flight . We therefore predicted that seminal fluid compounds could be effective instruments to maximise the fitness interests of drones that already have achieved insemination success , if such compounds target queen photoreception . This conjecture could then be an example of sexual conflict mediated by sensory exploitation , which is known to have created sexual arms-race dynamics in other animals via male-manipulation that induced selection for compensatory traits in females ( Arnqvist , 2006; Hollis et al . , 2019 ) . We used RNA-sequencing to quantify gene expression changes in the brains of queens that we had artificially inseminated with seminal fluid ( solely or in combination with sperm ) and assessed whether this induced comparable changes in the expression of vision-related genes to those reported for naturally inseminated queens ( Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2010; Manfredini et al . , 2015 ) . Because we were able to confirm that the expression of genes involved in phototransduction was indeed altered by seminal fluid , we then phenotypically quantified the visual performance of queens that had been inseminated with seminal fluid with and without sperm , or a saline control fluid . We measured response amplitude and contrast sensitivity of the queens’ compound eyes and ocelli over two days after artificial insemination and found that those queens exposed to seminal fluid indeed experienced a significant decrease in visual perception . We completed our study by monitoring natural mating flight activities in an apiary , using queens that had received the same artificial insemination treatments . We found clear effects of seminal fluid on the timing of , and survival during , queen mating flights , consistent with ongoing antagonistic selection for manipulative seminal fluid traits and compensating behavioural defences by queens .
We found that 24 hr after queens had received seminal fluid – either pure or as part of ejaculates ( i . e . together with sperm ) – their brain gene expression profiles were substantially altered compared to queens that had either received mock inseminations ( i . e . the entire procedural sequence of artificial insemination but without injecting fluid into their genital tracts ) in a first RNA-seq experiment , or control inseminations with Hayes saline only in a second experiment ( Figure 1 ) . We identified 1327 ( 8 . 6% of the honeybee transcriptome ) differentially expressed genes ( DEGs ) across all pair-wise brain-comparisons between treatment groups in the two subsequent RNA-seq experiments , with an over-representation of up-regulated DEGs in queens exposed to pure seminal fluid or semen ( i . e . complete ejaculates ) ( Figure 1—figure supplement 1 and Figure 1—figure supplement 2; see Supplementary file 1 and Supplementary file 2 for the number and identity of DEGs identified in each pair-wise comparison , respectively ) . Functional enrichment analyses revealed that several Biological Process terms such as ‘signal transduction’ , ‘signaling’ , ‘cell communication’ and ‘response to stimuli’ were consistently enriched in all comparisons of semen and pure seminal fluid insemination treatments compared to controls ( Figure 1—figure supplement 3 and Supplementary file 3 ) . Enriched GO terms also included ‘pathogenesis’ ( semen vs . mock insemination ) , ‘ocellus pigmentation’ , ‘sleep’ , ‘proteolysis’ , ‘regulation of phagocytosis’ , ‘negative regulation of DNA replication’ , ‘RNA metabolic process’ and ‘carbohydrate catabolism’ ( seminal fluid vs . mock insemination ) , ‘lipid catabolic process’ ( both semen and seminal fluid vs . mock insemination ) , ‘mitotic cell cycle process’ , ‘vitamin transport’ , ‘DNA packaging’ , ‘catechol-containing compound metabolic process’ ( semen vs . Hayes saline ) , ‘establishment of epithelial cell polarity’ ( semen vs . mock insemination , seminal fluid vs . mock insemination and semen vs . Hayes saline ) , ‘negative regulation of MAPK cascade’ , ‘NAD metabolic process’ ( seminal fluid vs . Hayes saline ) , ‘tryptophan catabolic process to kynurenine’ ( both seminal fluid vs . mock-insemination and seminal fluid vs . Hayes saline ) , and several terms related to the transmembrane transport of ions ( seminal fluid vs . mock insemination and semen vs . Hayes saline; see Figure 1—figure supplement 3 and Supplementary file 3 for more detailed lists , including Molecular Function GO terms ) . Although we detected only a single DEG in the semen vs . seminal fluid comparison in the first RNA-seq experiment , the same comparison yielded 802 DEGs in RNA-seq experiment 2 , which had greater detection power for unknown reasons . However , all these DEGs had only relatively small expression changes ( −1 < log2 ( fold change ) < 1 ) and most of them were up-regulated in semen compared to seminal fluid ( Figure 1—figure supplement 1 ) , suggesting that the presence of sperm may have enhanced effects of seminal fluid on the expression of many of these genes . A GO enrichment analysis revealed that these genes were mostly involved in ‘signaling’ , ‘cell communication’ and ‘ion transport’ as with seminal fluid alone , but also mediated effects that were only recovered for this semen vs . seminal fluid comparison , such as the Biological Process terms ‘ATPase activity’ and ‘oxidation-reduction process’ ( Supplementary file 3 ) . We then used GAGE analyses ( Luo et al . , 2009 ) to identify which signaling and metabolic cascades were altered by the semen and seminal fluid insemination treatments . We found that the DEGs consistently mapped to the phototransduction pathway ( all comparisons , except seminal fluid vs . semen in experiment 1 ) and the neuroactive ligand-receptor interaction pathway ( all semen and seminal fluid comparisons vs . controls; Supplementary file 4 ) . Additionally , the Hippo signaling pathway was altered in both semen and seminal fluid comparisons against Hayes saline , the oxidative phosphorylation pathway only in semen vs . Hayes saline and seminal fluid vs . semen in experiment 2 ( suggesting the production of ATP through this mitochondrial pathway is enhanced by the presence of sperm ) , the phagosome and tyrosine metabolism pathways exclusively in semen vs . Hayes saline , and the ribosome pathway in both seminal fluid vs . Hayes saline and seminal fluid vs . semen comparisons ( with the pathway being consistently down-regulated in seminal fluid inseminated queens ) . The Hippo signaling pathway is known to control organ size during development by regulating cell-to-cell signaling and cell proliferation ( Halder and Johnson , 2011 ) , the phagosome pathway is linked to the process of particle engulfment by cells during inflammation , tissue remodelling , and defense against pathogens ( Stuart and Ezekowitz , 2005 ) , the tyrosine metabolism pathway can be involved in dopamine biosynthesis and plays a role in retinal pigmentation and associated diseases ( Molnár et al . , 2005; Yang et al . , 2017 ) , whereas the ribosome pathway is involved in protein synthesis , but also plays a role in DNA repair , replication , RNA processing and transcription , and development ( Yang et al . , 2005; Lai and Xu , 2007 ) . We consequently analysed in more depth the highly consistent effects we recovered across the two RNA-seq experiments for the phototransduction pathway - the process by which light is converted into electrical signals in photosensitive retinal cells ( Figure 2 ) . In Drosophila this process is mediated by a G-protein-coupled phospholipase C ( PLC ) that functions as the effector enzyme . This protein controls conductance changes in the plasma membrane of microvillar photoreceptors in the eye by activating two types of Ca2+-permeable cation channels , TRP and TRPL ( Hardie , 2012; Figure 2 ) . We found that the gene coding for PLC , no receptor potential A ( norpA ) , was always up-regulated in queens inseminated with semen or seminal fluid compared to control queens ( Figure 2 and Figure 3 ) , which implies phenotypic effects on queen vision . We also found that semen and seminal fluid affected the expression of several genes whose Drosophila and honeybee orthologs are involved in the development of retinal microvilli ( Baumann and Lautenschläger , 1994; Hicks et al . , 1996 ) , including an Actin gene ( Actin a in Figure 3 ) and the ninaC gene . Drosophila null mutants at the ninaC locus have reduced amounts of visual pigment , defects in response termination and light adaptation , increased dark noise , and light-dependent retinal degeneration ( Hardie , 2012 ) . Finally , we found more variable expression changes for a series of other genes in the same pathway , including those coding for the cation channels TRP and TRPL and the production of diacylglycerol lipase ( DAGL; Figure 2 ) . This latter enzyme is known to catalyse the hydrolysis of diacylglycerol ( DAG ) into polyunsaturated fatty acids ( PUFAs ) , which activates TRP and TRPL channels and is required for generating photoreceptor responses to light in Drosophila ( Hardie , 2012 ) . Across our two RNA-sequencing experiments , the annotation of the 37 genes with consistent differential expression in queens exposed to semen or pure seminal fluid inseminations compared to controls implies morphological changes in the retina ( Figure 1—figure supplement 2 and Supplementary file 5 ) . This list included the genes Hemicentin-1-like , Chaoptin , Thrombospondin , and Bardet-Biedl syndrome 2 , which have all been associated , among other roles , with retinal degenerative disorders in lineages ranging from Drosophila to humans ( Katsanis et al . , 2000; Schultz et al . , 2003; Stewart , 2006; Gurudev et al . , 2014 ) . This effect may be significant because the gene Hemicentin-1-like consistently showed the greatest fold change between queens inseminated with semen or seminal fluid compared to the controls . Our functional enrichment analyses further highlighted several Biological Process terms suggesting changes in cell structure , cell adhesion and tissue morphogenesis , such as ‘non-motile primary cilium assembly’ ( enriched in all our pair-wise comparisons ) , ‘cell-cell junction assembly’ and ‘eye-antennal disc development’ , consistent with phenotypic effects on eyes or remodelling of brain structures ( Figure 1—figure supplement 3 and Supplementary file 3 ) . The lists of significantly enriched Biological Process terms also contained genes related to the metabolism of cyclic guanosine monophosphate ( cGMP; Supplementary file 3 ) , a known derivate of phototransduction acting as regulator of ion channel conductance ( Hardie , 2012 ) . Finally , neuroactive receptor genes up-regulated in semen-treated and seminal-fluid-treated queens included the glutamate metabotropic receptor gene mGlu2R , the glutamate receptor gene NMDAR1 , and the serotonin receptor genes 5-HT2alpha and 5-HT1 ( Figure 3—figure supplement 1 , Figure 3—figure supplement 2 , Figure 3—figure supplement 3 and Supplementary file 4 ) . However , 5-HT2alpha was only up-regulated compared to mock inseminations , suggesting this gene is under control of stretch receptors in the queen genital tract rather than affected by male-derived secretions . To assess the extent to which our artificial insemination treatments produced changes similar to natural inseminations , we compared our DEGs with those of a previous study ( Manfredini et al . , 2015 ) that compared brain gene expression in naturally inseminated queens with that of virgin queens 48 hr after they were treated with CO2 or not . Although we measured brain gene expression after 24 hr rather than 48 hr , 153 of the 1 , 327 DEGs were shared with the 1 , 050 DEGs identified in the natural insemination comparisons of Manfredini et al . ( 2015 ) . This 12–15% overlap ( Hypergeometric test: representation factor = 1 . 7 , p<0 . 0001; Supplementary file 6 ) was comparable to the degree of overlap in brain transcriptomes of newly inseminated honeybee queens across independent studies ( Kocher et al . , 2010; Niño et al . , 2011; Niño et al . , 2013a; Manfredini et al . , 2015 ) . We therefore concluded that this overlap in DEGs is biologically relevant and consistent with artificial insemination with seminal fluid or semen inducing part of the gene expression changes induced by natural insemination . Accepting this partial match as a replication of previous results is reasonable because ( i ) the difference in experimental design ( i . e . different time-points and treatments ) between our study and the previous one was substantial , ( ii ) the mating process includes factors other than seminal fluid known to affect queen brain transcriptomes ( Koeniger , 1976; Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2009; Kocher et al . , 2010; Niño et al . , 2011; Vergoz et al . , 2012; Niño et al . , 2013a; Niño et al . , 2013b; Manfredini et al . , 2015; Brutscher et al . , 2019 ) , and ( iii ) RNA-seq with limited replication is unlikely to identify all DEGs in a given tissue so the overlap would not be expected to be more than partial . The 153 shared DEGs with the study by Manfredini et al . ( 2015 ) were generally enriched for Biological Process terms related to energy metabolism ( Supplementary file 7 ) , but we assessed whether the overlapping genes included those related to vision more specifically . Out of the eight vision-associated genes listed in the Additional File 6 of Manfredini et al . ( 2015 ) , four ( ninaC , ninaA , norpA , chaoptin ) were also differentially expressed in our study , which represented a statistically significant overlap ( Hypergeometric test: representation factor = 5 . 57 , p=0 . 0002 ) . A fifth gene ( Arr2 ) was only altered in our seminal fluid vs . semen comparison in RNA-seq experiment 2 ( Figure 2 ) . Our systematic comparisons with the DEG lists of Manfredini et al . ( 2015 ) revealed another six joint DEGs with a potential role in vision ( Actin ( a ) , Actin ( b ) , GNB1 , TRP , myosin-IIIb , Hemicentin-1-like ) , corroborating that the effects on vision between their natural inseminations and our artificial insemination experiments were similar . Taken together our gene expression results provide ample evidence that seminal fluid triggers the expression of vision-related genes similarly to what had been previously documented for naturally inseminated queens but without identifying the causal mating factors inducing these changes . We obtained electroretinograms ( ERGs ) to explore the phenotypic eyesight consequences of gene expression changes in queen brains by investigating the temporal contrast sensitivity functions of queen eyes one and two days after artificial insemination with semen , seminal fluid or Hayes saline control solution . For the compound eyes , we found that queens inseminated with semen or seminal fluid showed lower response amplitude to flickering light of low temporal frequencies ( the number of light flashes per second , expressed in Hz ) than queens inseminated with Hayes saline , and that the magnitude of this effect increased on the second day after insemination ( three-way statistical interaction term between stimulus frequency , number of days after insemination , and insemination treatment; N = 37 , χ2 = 38 . 88 , df = 8 , p<0 . 0001; Figure 4A and Supplementary file 8; see Figure 4—figure supplement 1 for the experimental set-up and Figure 4—figure supplement 2 for an example of ERG response and details on how we derived contrast sensitivity measurements ) . To confirm that the increased reductions in compound eye performance over the two consecutive days were not due to queens having spent the night attached to their holders ( see Methods for details ) , we compared the response amplitude 48 hr after insemination of the 18 queens that were measured both on day 1 and 2 with those of the nine queens that were only measured after 48 hr . We found no statistical difference ( N = 27 , χ2 = 8 . 566 , df = 6 , p=0 . 20; Supplementary file 9 ) , suggesting the effects were exclusively due to seminal fluid exposure . The reduction in response amplitude remained statistically significant after excluding all measurements of the semen ( full ejaculate ) treatment ( N = 25 , χ2 = 15 . 16 , df = 1 , p<0 . 0001; Supplementary file 10 ) , which is consistent with seminal fluid , rather than sperm , being primarily responsible for the reduction in phenotypic eyesight performance that we observed . For the ocelli , we found that inseminations with semen or seminal fluid induced lower response amplitude than control treatments at higher light stimulus contrasts ( statistical interaction between treatment and contrast; N = 35 , χ2 = 16 . 53 , df = 4 , p=0 . 0024; Figure 4B and Supplementary file 11 ) . Inseminating pure seminal fluid had the strongest effect , but treatment effects were less pronounced on the second day ( interaction between treatment and day; N = 35 , χ2 = 8 . 33 , df = 2 , p=0 . 0155; Supplementary file 11 ) . When we repeated the statistical analysis after excluding all queens inseminated with semen , we still found a significant interaction between light stimulus contrast and insemination treatment , confirming that reduced ocelli visual perception is induced by seminal fluid and that this effect is most pronounced at higher light stimulus contrasts ( N = 23 , χ2 = 6 . 69 , df = 2 , p<0 . 0001; Supplementary file 12 ) . Low signal to noise levels precluded calculation of differences in light contrast sensitivity for the ocelli , but for the compound eyes the contrast sensitivity at low temporal frequencies was always highest when queens were inseminated with control saline solution and lowest when they were inseminated with semen ( interaction between stimulus frequency and treatment; N = 37 , χ2 = 24 . 05 , df = 10 , p=0 . 008; Figure 4C and Supplementary file 13 ) . These effects appeared to develop over time , although the statistical interaction between day and treatment was only marginally significant ( N = 37 , χ2 = 11 . 02 , df = 5 , p=0 . 051; Supplementary file 13 ) . However , after excluding all semen measurements there was no significant insemination treatment effect on contrast sensitivity , implying that the presence of sperm may be required to induce this effect ( Supplementary file 14 ) . We also measured the response of the compound eyes to isolated , short ( 1 ms ) flashes of light , but this did not yield any significant differences in response duration , latency , or amplitude between treatment groups ( Figure 4—figure supplement 3; see Figure 4—figure supplement 4 for details on how amplitude , latency and duration were derived from the original ERG responses ) . Taken together our results offer compelling phenotypic evidence for the hypothesis that insemination induces reduced response to light stimulation in the compound eyes and , somewhat less consistently , in the ocelli of honeybee queens . As expected we found that most of these effects are primarily induced by seminal fluid and that sperm contributes to their enhancement . We equipped 36 queens with radio-frequency identification ( RFID ) tags after artificial insemination with semen , seminal fluid or Hayes saline and monitored their natural flight activity over several consecutive days . Among the 34 queens that left their hives , those inseminated with either pure seminal fluid or semen were more likely to get lost , i . e . they did not return to their hives when they flew again . They also triggered hive entrance sensors more often than their sister queens inseminated with Hayes saline ( Figure 5A; Binary Logistic Regression of”Not Found’ by”Treatment’ and”PingNr’ , the acronym for the number of times queens triggered the sensors located at hive entrances; Nagelkerke R2 = 0 . 535 , 79 . 4% Correct; Treatment: N = 34 , Wald-χ2 = 6 . 970 , df = 2 , p=0 . 031; PingNr: N = 34 , Wald-χ2 = 4 . 843 , df = 1 , p=0 . 028 ) . The latter effect is consistent with semen- and seminal-fluid-inseminated queens being disoriented or distressed by sunlight and spending more time at hive entrances than control queens . However , this difference may also reflect increased general activity induced by seminal fluid , similar to the sex peptide increasing female activity and reducing siesta sleep in Drosophila ( Isaac et al . , 2010 ) . Of a total of 13 queens that did not return to their hive , four triggered sensors for the last time two days after insemination ( seminal fluid: N = 2 , semen: N = 2 ) , two after three days ( seminal fluid: N = 1 , Hayes saline: N = 1 ) , four after four days ( seminal fluid: N = 1 , semen: N = 3 ) , and three between day six and seven ( seminal fluid: N = 1 , semen: N = 2 ) . Of the 21 queens returning to their hives , 17 ( 81 % ) performed flights that exceeded 7 min , which is a conservative threshold for assessing a complete mating flight based on the shortest mating flight performed by Hayes-inseminated control queens that later laid eggs in our study ( see Materials and Methods for details ) . We confirmed the presence of brood in worker comb in 11 of these queens ( 52% ) , indicating that they originated from fertilised eggs , because unfertilised eggs are laid into larger drone combs . Of these 17 queens attempting real mating flights , those inseminated with seminal fluid or semen left their colonies 1–2 days earlier than those inseminated with Hayes saline ( ANOVA , N = 17 , df = 2 , p=0 . 026 ) . This effect appeared to be general because we found the same statistical result when we analysed data of all 34 queens: those inseminated with either seminal fluid or semen triggered sensors for the first time earlier than control queens ( Figure 5B; Kruskal-Wallis , N = 34 , df = 2 , p=0 . 004 ) . Finally , semen- or seminal-fluid-inseminated queens also triggered sensors for the last time earlier than saline-treated queens ( Figure 5B; Kruskal-Wallis , N = 34 , df = 2 , p=0 . 033 ) .
One could argue that the loss of queen visual perception after insemination allows queens to reduce the substantial physiological costs for the maintenance of a metabolically complex trait ( Niven and Laughlin , 2008 ) and that the effects we documented would be adaptive if there are trade-offs between energetically demanding life history traits . Honeybee queens return to a hive where levels of luminance are low , so they do not require fully functional eyesight because all communication becomes non-visual and mediated by pheromones , direct contact , vibrations or sounds ( Billen , 2006 ) . Energetic trade-offs might be specifically important for newly inseminated honeybee queens , because their ability to produce large numbers of fertilised eggs shortly after their mating flight ( s ) is crucial to compensate for the lack of egg production in the hive and for the workers that were lost when the old mother queen left with a swarm a few days before ( Winston , 1987; Grozinger et al . , 2014 ) . Because older honeybee queens still have the capacity to initiate a new colony through swarming ( Winston , 1987 ) , a permanent loss of eyesight is not expected , even though swarming flights are probably less visually demanding than mating flights because swarming queens are always accompanied and guided by a large number of workers . If queens have to trade-off energetically demanding physiological traits such as visual perception and fecundity , they should be able to determine the optimal timing to modify the amount of energy allocated to these traits through mechanisms that provide reliable cues about the quantity , quality and diversity of the sperm they obtained , such as monitoring the number of copulations achieved , the degree of filling of the lateral oviducts with semen and the extent to which sperm have entered the spermatheca . However , our results show that the alterations in queen visual perception even occur in queens that only receive seminal fluid , which by itself does not deliver such information to queens . This is consistent with a conflict-mediating role of seminal fluid and not with queens harmoniously managing their optimal number of inseminations . We therefore conclude that the most parsimonious explanation of our findings is that seminal fluid inhibits future promiscuity of queens as part of a sexual conflict , because: 1 . We could show that changes in visual perception start much earlier than would be predicted if these changes were merely adaptive to queens , i . e . during the time window of repeated nuptial flights ( Winston , 1987 ) , and that visual perception is reduced well before the ca . 40 hr that sperm cells need to reach final storage in a queen’s spermatheca ( Woyke , 1983 ) . 2 . Our results confirmed that seminal fluid without sperm is capable of triggering both the genetic and the behavioural changes in the brains of honeybee queens , similar to those reported after natural mating flights ( Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2010; Manfredini et al . , 2015 ) . 3 . We also confirmed that these seminal fluid effects impose survival costs on queens during a period in which they would be expected to possess an unimpaired ability to engage in additional mating flights and return to the hive with high probability . A key condition for the presence of a sexual conflict and an arms race between queens and drones over paternity is that both sexes should pay fitness costs . Our results suggest this to be the case because successful ejaculates from a first mating flight ( in pre-storage for up till 40 hr ) risk collective death when their seminal fluids handicap queens in attempting additional flights . Queens have been under selection to accept mortality risks emanating from additional flights because of well-documented fitness benefits of permanently storing sperm of sufficient genetic diversity ( Mattila and Seeley , 2007; Mattila et al . , 2012 ) . However , their pre-stored ejaculates , competing within the queen sexual tract to obtain permanent storage in the spermatheca , are selected to take higher risks of queen-failure because their future paternal fitness will be reduced in a way that is proportional to the amount of sperm their mate obtains in a subsequent mating flight , especially when queens only store 3–5% of the sperm they received ( Baer , 2005 ) . Average mating frequencies of honeybee queens can thus be expected to be a compromise between the need for queens to obtain additional genetic diversity for their colonies and the mortality costs associated with additional flights resulting from longer exposure to predators , an increased risk of getting infected with sexually transmitted parasites ( Peng et al . , 2016 ) and aggravated by male manipulation of their visual perception . Where the arms race equilibrium settles will quantitatively depend on a number of environmental and genetic factors that may vary between geographic regions , seasons and species or subspecies of honeybees ( Kraus et al . , 2004; Kraus et al . , 2005; El-Niweiri and Moritz , 2011 ) . The sexual conflict hypothesis assumes that queens will have evolved mechanisms to neutralise seminal fluid compounds that affect their visual perception and that they vary in their effectiveness of counter-mechanisms to neutralise drone manipulations . The behaviours that we recorded in our apiary experiment and the fact that we detected substantial variation in queen visual perception one day after artificial inseminations provide further support for our interpretation of a sexually antagonistic arms race between drones and queens . These interactions would thus resemble the rapid evolution of adaptations and counter-adaptations in reproductive fluid molecules in other organisms ( Chapman , 2001; Swanson and Vacquier , 2002; Andrés et al . , 2006; Panhuis et al . , 2006; Haerty et al . , 2007; Findlay et al . , 2009; Walters and Harrison , 2010 ) . The interpretation of our results as being consistent with an ongoing sexual arms race over the number of mating flights rather than the number of copulations per se also agrees with previous research suggesting that honeybee queens adjust their flight number based on their insemination success during ( a ) previous flight ( s ) ( Schlüns et al . , 2005 ) . These studies already suggested that natural selection should act primarily at the level of queen flights , which represent greater efforts and risks than individual copulations occurring in quick succession do ( Tarpy and Page , 2000; Schlüns et al . , 2005 ) . The physiological and/or mechanical mechanisms mediating these responses remain poorly understood and the conceptual logic of our present study provides a novel framework and clear incentive for unravelling them . Although our genetic , phenotypic and behavioural results are all consistent with changes in queen vision , more research is needed to confirm whether the reduced performance of queens during mating flights is a consequence of perturbed visual perception only , or whether other physiological effects induced by seminal fluid might also play a role . One could argue that reduced queen survival in our apiary experiment resulted from harmful effects of seminal fluid on the general health of queens . This could for example occur if seminal fluid components affected female immunity as a form of ‘collateral damage’ of enhancing egg production ( Rolff and Siva-Jothy , 2002; Short et al . , 2012 ) . However , such arguments typically refer to studies of seminal fluid effects in non-social insects , where females re-mate at regular intervals throughout adult life . In these species , traits that enhance the paternity share of a focal male in the present clutch at the expense of later female health can evolve , because these female’s survival costs do not diminish the focal male’s fitness return ( Chapman et al . , 1995; Johnstone and Keller , 2000; Rolff and Siva-Jothy , 2002; Kemp and Rutowski , 2004; Wigby and Chapman , 2005 ) . However , a series of review papers and experimental studies have shown that this type of health effects are neither expected nor found in social insect queens ( Tsuji et al . , 1996; Boomsma et al . , 2005; Schrempf et al . , 2005; Heinze and Schrempf , 2008; Lopez-Vaamonde et al . , 2009; Rueppell et al . , 2015; Barribeau and Schmid-Hempel , 2017 ) . Natural selection will consistently eliminate traits mediating such ‘collateral damage’ because queens need to produce many cohorts of sterile workers before their colony can produce winged reproductives . This implies that even a slight negative effect of mating on the general physiological performance of queens would likely preclude survival until first reproduction and make males lose their paternity success together with the queen they inseminated . Hence , both the presence of seminal fluid effects on general health in female fruit flies and the absence of such effects in queens of social insects are direct consequences of the different evolutionary origins and temporal distribution of life-time female promiscuity . The exceptional mating system of the honeybee allows sexual conflict to be expressed for just a few days and can only target the odds of successful additional female promiscuity during this narrow time window , not their subsequent state of health . Although the ultimate ( evolutionary ) logic of an arms race over queen promiscuity during additional mating flights seems compelling and consistent with the evidence so far , it is also important to evaluate the degree to which our study provides insights into the proximate causation factors involved . Our data provide first and solid proof of concept evidence , but further work will be needed to unravel the complex interactions between differential gene expression and phenotypic effects on visual perception and flight behaviour . Our electrophysiological experiments for the compound eyes were consistent with visual perception loss accumulating quickly and gradually , and our apiary experiment showing that queens embarked on additional flights earlier , as if they actively responded to the fact of becoming visually handicapped , also matched our expectations . However , this pattern was different for the ocelli , suggesting not all complexities of seminal fluid effects on queen mating behaviour are straightforward . The drone congregation areas that honeybee queens need to localise in flight ( Winston , 1987 ) typically occur in spatially restricted areas , often associated with specific land-marks ( Galindo-Cardona et al . , 2012 ) and have a diameter of 30–200 m ( Baudry et al . , 1998 ) . Well-functioning compound eyes and ocelli thus appear to be essential for reaching drone congregations and for returning back to the hive . Honeybee workers use path integration with reference to the sun and a mental map based on learned visual landmarks when they navigate away from and back to the hive ( Menzel et al . , 2000; Menzel et al . , 2005 ) , but further research to confirm similar abilities in queens is required , if only because virgin queens have no previous flight experience to draw upon . Further research should also quantify whether there is a trade-off between cumulative mating flight effort and the initiation and scale of early egg-laying , because both are likely to affect the future reproductive success of queens . Additional causal factors that determine mating flight decisions of queens may emerge from such work , because our differential gene-expression analyses identified , apart from changes in phototransduction pathway genes , also a number of genes involved in energy metabolism , regulation of phagocytosis , DNA replication , RNA transcription , protein synthesis , and cellular adhesion , suggesting that metabolic effects and structural modifications or deterioration in photoreceptors or neurons may also occur . We also note that , although the queens used for our apiary experiment received a total volume of seminal fluid comparable to what they would likely receive during natural mating flights , the genetic diversity of these seminal fluid mixtures was likely higher compared to natural inseminations because we used hundreds of drones to collect the pure seminal fluid samples . We therefore generated qualitative evidence in the direction of our predictions , but we do not know whether the quantitative outcomes of our experiments reflected natural circumstances . For example our apiary experiments may have produced higher queen mortality than natural mating because queens were unusually handicapped by a larger diversity of seminal fluid molecules . Replication of our manipulative apiary experiment will therefore be needed to obtain a better quantitative understanding of the arms race dynamics for which we provide the first evidence . Finally , although we were careful to only expose queens to minimal amounts of CO2 and we were successful in recovering clear differences in gene expression between our treatment and control groups , the use of CO2 to stimulate ovary activation and to narcotise queens during artificial insemination may have masked additional changes induced by seminal fluid if they are also induced by CO2 exposure . Future research should clarify whether such confounding effects exist . Overall , our findings underline that polyandrous social insects provide intriguing testbeds for general sexual conflict theory and that honeybees offer a plethora of interesting research opportunities to unravel the proximate mechanisms that shape the practical implementation of the sexual conflict that we documented .
All queens used for experiments were bred at the University of Western Australia according to standard apicultural practices ( Laidlaw and Page , 1997 ) . We grafted honeybee ( Apis mellifera ligustica ) larvae at day four of their development ( i . e . one day after hatching ) from a single colony , transferred them into plastic queen cells ( Ceracell , Aukland , NZ ) and placed them into a queen-right cell-building colony , prepared 24 hr in advance by moving two frames of uncapped brood above the queen excluder and placing the graft bar in between these brood frames . Ten days later we placed the developing queens in their cells into 4-frame queen-less nucleus hives with queen excluders at the entrance to prevent natural mating flights and allowed them to hatch . Four days after hatching and one day before artificial inseminations , virgin queens were removed from their hives , caged , and exposed to CO2 for 1 min to stimulate ovary activation . Five nurse bees were added to the cages before returning queens to their original nucleus hives . Queens used in the apiary mating flight experiment were never exposed to CO2 but were cooled on ice prior to , and during , artificial inseminations ( see below ) . We collected semen ( consisting of seminal fluid and spermatozoa ) or pure seminal fluid using a previously-developed protocol ( Baer et al . , 2009 ) and keeping the collection procedures identical across all experiments performed . We randomly caught drones at hive entrances and kept them in cages before transferring them in two foster hives . After a maximum of two days , we re-collected the cages and anaesthetized drones with chloroform to initiate ejaculation . We then squeezed the drone’s abdomens between two fingers and collected the semen appearing on the tip of the endophallus in a glass capillary connected to a syringe ( Schley , Germany ) ( Baer et al . , 2009 ) and immediately used it for artificial inseminations for the ‘semen’ treatments . To obtain pure seminal fluid , we collected semen in batches of several hundred drones ( ~2000 in total ) as described above and pooled these samples in 1 . 5 ml Eppendorf tubes , which we centrifuged at 18 , 500 g and 4°C for 25 min . The supernatants were collected into new 1 . 5 ml Eppendorf tubes and centrifuged for 10 min at 18 , 500 g and 4°C ( Baer et al . , 2009 ) . We then collected the seminal fluid as supernatant , froze these aliquots in liquid nitrogen and stored them at −80°C prior to further experiments . This centrifugation method has previously been shown to be a reliable method to collect seminal fluid uncontaminated by sperm cells and/or major sperm proteins ( Baer et al . , 2009 ) . Artificial inseminations were performed as described in detail earlier ( Mackensen and Roberts , 1948 ) . For a first RNA-sequencing experiment we compared brain gene expression in queens that we artificially inseminated with either semen , seminal fluid or a mock insemination treatment , where no fluid was injected into the vaginal orifice . Virgin queens were sampled from their hives and randomly assigned to one of three experimental groups ( referred to as ‘Experiment 1’ ) : ( i ) instrumentally inseminated with 6 µl of semen pooled from approximately 10 males , ( ii ) instrumentally inseminated with 6 µl of seminal fluid , and ( iii ) mock-inseminated without injecting any fluid . In a second RNA-sequencing experiment ( referred to as ‘Experiment 2’ ) we further controlled for insemination of liquid into the queen reproductive tract and assessed to what extent gene expression changes in queen brains were dependent on reception of semen or seminal fluid rather than on the mechanical stimulation of the reproductive tract upon insemination . To do this , we randomly assigned queens to one of three experimental groups: ( i ) instrumentally inseminated with 6 µl of semen , ( ii ) instrumentally inseminated with 6 µl of seminal fluid , and ( iii ) instrumentally inseminated with 6 µl of Hayes saline ( 9 g NaCl , 0 . 2 g CaCl2 , 0 . 2 g KCl and 0 . 1 g NaHCO3 in 1000 ml H2O , adjusted to pH 8 . 7 and sterilised by filtration through a 0 . 22 µm syringe-filter , Membrane Solutions ) . To artificially inseminate queens , we sedated them with CO2 for a few seconds before placing them in a holder mounted onto a standard artificial insemination instrument ( Schley , Germany ) and inseminating them according to the treatments . Queens were afterwards allowed to recover for about 30 min before we returned them to their hives . We recollected queens after 24 hr , narcotised them with CO2 for a few seconds and flash froze their heads in liquid nitrogen . All heads were stored at −80°C until dissections . Brains of queens were dissected with Inox five watchmaker forceps under an Olympus SZX10 stereo microscope in ice-cold sterile Hayes saline . We ensured that brains were removed intact , including the optic lobes , and that all hypopharyngeal and other extraneous glandular tissue was removed . We combined the brains from three individual queens to obtain enough RNA for Illumina TruSeq sequencing ( see below ) , froze these pooled brain samples in liquid nitrogen and immediately stored them at −80°C . We consequently obtained three biological replicates ( each consisting of three pooled brains ) per treatment group in both experiments 1 and 2 ( 18 samples in total; Supplementary file 15 ) . To extract RNA from pooled brain samples , we briefly thawed them on ice , placed them in 10 µl of 0 . 25 M Tris pH 7 . 5 and homogenised them with a plastic pestle . We then added 50 µl of Trizol to each sample , incubated samples on ice for 15 min , thoroughly vortexed and returned them to ice for another 15 min , followed by centrifugation at 20 , 000 g for 15 min at 4°C . In the next step , we collected the supernatant , added one volume isopropanol , briefly vortexed the samples and incubated them at room temperature for 20 min , followed by centrifugation at 20 , 000 g for 15 mins at 4°C . After discarding the supernatants , we washed each pellet with 70% EtOH , followed by air-drying and resuspension in 20 µl of RNase-free H2O . To precipitate the RNA , we added 1 . 6 volumes of ice-cold 4 M LiCl and incubated samples on ice for 1 hr , followed by centrifugation at 20 , 000 g for 15 min at 4°C . After discarding the supernatants , each RNA pellet was washed with 500 µl 70% EtOH before resuspension in 10 µl RNase-free H2O . Sequencing libraries were generated from 1 µg total input RNA using the TruSeq Stranded mRNA Sample Kit ( Illumina , San Diego , CA ) and single-end sequencing by synthesis was performed on a HiSeq 2500 ( Illumina , San Diego , CA ) for 120 cycles , thus generating 120 bp reads . Samples were dispersed over four lanes of a single plate . The sequencing produced a mean of 54 , 574 , 574 reads per sample ( range 38 , 806 , 853–74 , 375 , 492; Supplementary file 15 ) . RNA-sequencing data have been deposited in NCBI's Gene Expression Omnibus ( Edgar et al . , 2002 ) and are accessible through GEO Series accession number GSE127185 ( https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE127185 ) . Reads were quality-controlled using FastQC ( http://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) and subsequently processed with Trimmomatic ( Bolger et al . , 2014 ) to remove adapters and low quality bases using the following parameters: LEADING: 3 ( trim the leading nucleotides until quality >3 ) , TRAILING: 3 ( trim the trailing nucleotides until quality >3 ) , SLIDINGWINDOW: 4:15 ( trim the window of size four for reads with local quality below a score of 15 ) , and MINLEN: 36 ( discard reads shorter than 36 bases ) . On average , 1 . 7% of the total reads were discarded during this step . Next , we removed reads that matched ribosomal RNA sequences ( rRNA ) using SortMeRNA ( Kopylova et al . , 2012 ) , which implied discarding an average of 2 . 3% of the reads . The remaining reads were aligned with STAR v . 2 . 4 . 2a ( Dobin et al . , 2013 ) to the latest version of the honeybee genome ( Apis mellifera assembly 4 . 5 ) available on BeeBase ( http://hymenopteragenome . org/beebase/ ? q=download_sequences ) , which resulted in an average mapping rate of 97% ( Supplementary file 15 ) . Mapped reads were converted into raw read counts with the htseq-count script ( http://www . huber . embl . de/users/anders/HTSeq/doc/count . html ) , and the R ( R Core Development Team , 2015 ) Bioconductor ( Huber et al . , 2015 ) package DESeq2 v . 1 . 10 . 1 ( Love et al . , 2014 ) was subsequently used to quantify differential gene expression between all pair-wise combinations of treatment groups for each experiment . P values of differential expression analyses were corrected for multiple testing with a false discover rate ( FDR ) of 10% . We used Non-metric Multidimensional Scaling ( NMDS ) of Bray-Curtis dissimilarities in Paleontological Statistics v . 3 . 04 ( Hammer et al . , 2004 ) to investigate overall sample clustering after removing experimental batch effects with the removeBatchEffect function implemented in the R Bioconductor package edgeR v . 3 . 12 . 1 ( Robinson et al . , 2010 ) . A Hypergeometric test ( http://nemates . org/MA/progs/overlap_stats . html ) was used to quantify overlap in differentially expressed genes between our study and Manfredini et al . ( 2015 ) . Conducting such a comparison was of interest to assess whether our artificial insemination treatments induced similar effects as those found in naturally inseminated queens . Manfredini et al . ( 2015 ) performed their experiments on Australian honeybees , and quantified gene expression in the brains of ( i ) virgin queens , ( ii ) naturally-inseminated queens , and ( iii ) CO2-treated queens two days after treatments were performed . They also used RNA-sequencing rather than microarrays as in previous studies ( Grozinger et al . , 2007; Kocher et al . , 2008; Kocher et al . , 2010; Niño et al . , 2011; Niño et al . , 2013a ) . For all of the above reasons the Manfredini data sets were the most appropriate to compare our results with those obtained from natural mating comparisons . To perform Gene Ontology ( GO ) enrichment analyses we first annotated the honeybee transcriptome ( OGS v3 . 2 available on BeeBase ) by running BLASTx against the NCBI Non-Redundant ( NR ) database ( performed in March 2016 ) , retaining the first 20 hits with a cutoff eValue of 10−3 . Blast2GO v . 3 . 2 ( Conesa et al . , 2005 ) was used to map the ensuing annotations to GO terms . We then used a hypergeometric test implemented in the R Bioconductor package GOstats v . 2 . 36 . 0 ( Falcon and Gentleman , 2007 ) to evaluate the differentially expressed gene lists for GO term associations , using the full transcriptome as background and retaining Biological Process and Molecular Function terms with P values < 0 . 05 . REVIGO ( Supek et al . , 2011 ) was subsequently used to reduce redundancy in significant GO terms and to summarise results by semantic similarity . Perturbed genetic pathways were identified with the R Bioconductor package Generally Applicable Gene-set Enrichment for Pathway Analysis ( GAGE v . 2 . 20 . 1 ) ( Luo et al . , 2009 ) by retaining Kyoto Encyclopedia of Genes and Genomes ( KEGG ) signaling and metabolic pathways ( accessed in August 2016 ) with q values < 0 . 2 . For significant pathways , we identified genes that showed expression changes over noise levels with the essGene function in GAGE using default parameters . To test whether exposure to seminal fluid resulted in a phenotypic alteration of queen visual perception , we reared virgin queens as described above and artificially inseminated them with either: ( i ) 6 µl of semen , ( ii ) 6 µl of seminal fluid or ( iii ) 6 µl of Hayes saline ( see above for details ) . After the insemination procedure , we caged queens individually and randomly placed them back into one of two foster colonies . The following day we recollected the queens and sedated them on ice , removed their legs and fixed them with bee wax on a plastic holder to minimise head movements . The holders with the queens were randomly assigned to , and mounted in , one of two Faraday cages . We recorded ERGs from both the queens’ compound eyes and their median ocellus using a differential amplifier ( DAM50 , World Precision Instruments ) connected to a standard PC via a 16-bit data acquisition card ( USB-6353 , National Instruments ) . All recordings were controlled by custom-made software in MATLAB R2014a ( Source code 1; Ogawa et al . , 2015 ) . A silver/silver-chloride wire of 0 . 1 mm diameter was inserted into the animal’s thorax and served as the reference electrode . The recording electrode was a platinum wire of 0 . 254 mm diameter covered with conductive , neutral pH gel ( ECGEL250 , Livingstone International ) , carefully positioned on the dorsal surface of one of the compound eyes or along the median ocellar lens ( Figure 4—figure supplement 1 ) . The electrical ground was connected to the Faraday cage . The light source was a ‘cool white’ LED light with 5 mm diameter ( C503C-WAS-CBADA151 , Cree Inc , Durham , NC , USA ) , powered by a custom-made LED driver using pulse width modulation ( PWM ) . All light stimuli were checked for linearity using a calibrated light metre ( ILT1700 , International Light Technology ) . The LED was positioned at an elevation of approximately 30° in the queen’s visual field and kept at a constant distance of 70 mm from the queen’s head . To reduce any electrical noise from the light source , two grounded metal shields with 3 and 1 mm holes were positioned 30 mm from the light source and 10 mm from the queen’s head , respectively . A total of 37 queens were tested , 12 inseminated with semen , 14 with seminal fluid and 11 with Hayes saline . From a total of 28 queens that were used to measure visual perception one day after the insemination treatments , 18 were re-used for measurements a day later – that is two days after they were artificially inseminated , and kept attached to their holders in a small plastic container in the dark overnight after pipette-feeding them with sugar water . Another nine queens were only measured two days after insemination and were collected directly from the hives where they had been placed after the inseminations . Queens were dark-adapted for 20 min prior to all recordings . To measure the eyes’ ability to detect temporal changes in brightness , we measured the temporal contrast sensitivity function , which is the inverse of the lowest detectable contrast at each temporal frequency . The stimulus contrasts were expressed as Michelson contrasts LMAX-LMINLMAX+LMIN where LMAX is maximum light intensity and LMIN is minimum light intensity of the square wave stimulation pattern . We used three light intensities ( 2 . 74*10−2 W/cm2 , 2 . 74*10−3 W/cm2 , 2 . 74*10−4 W/cm2; we also used a second Faraday cage/light source with 70% dimmer LED intensities , and randomly assigned queens to these two set-ups ) , and we tested all 80 combinations of eight temporal frequencies ( 2 , 4 , 8 , 16 , 32 , 64 , 128 , 256 Hz ) and 10 contrasts ( 0 . 0019 , 0 . 0039 , 0 . 0078 , 0 . 0156 , 0 . 0312 , 0 . 0625 , 0 . 125 , 0 . 25 , 0 . 5 , 1 ) at each light intensity . For an example of ERG response and further details on how we derived contrast sensitivity measurements see Figure 4—figure supplement 2 . We next recorded the impulse response of the compound eyes and ocelli to a 1 ms flash of light , at the same three light intensities as before , followed by 2 s of darkness . An averaged response of 100 times repetitions was taken as the impulse response for each individual . The average response per condition was then analysed for its latency , duration , and amplitude ( see Figure 4—figure supplement 4 for an example of original ERG response and further details on how amplitude , latency and duration were derived from the original responses ) . To test for significant differences between treatments in ERG measurements , we used linear mixed effects models within the R package lme4 ( Bates et al . , 2015 ) . All models included animal identity , date of measurement , and recording Faraday cage as random effects to account for repeated measures of some queens , for measurements performed on different days , and for measurements having been recorded in two different Faraday cages . The dependent variable contrast sensitivity was analysed as a function of the fixed effects: number of days after insemination , stimulus intensity , temporal frequency , contrast , and treatment group . The dependent variables amplitude , latency , and duration of the impulse response to a brief 1 ms light pulse were analysed as a function of the three fixed effects: number of days after insemination , stimulus intensity , and treatment group . Two queens belonging to the seminal fluid treatment were excluded from the ocelli dataset because their measurements represented clear outliers due to small signal sizes and large technical noise . Factors or interaction-terms were added stepwise and χ2 significance values were obtained by comparing nested models ( R function ANOVA ) . Only variables with p<0 . 05 were retained in the final model and all reported P values were tested against the final model . All models were also graphically checked for consistency with model assumptions of normality and homogeneity of variances . To corroborate our findings from our previous gene expression and electroretinogram experiments , we bred three rounds of 12 virgin sister queens and artificially inseminated them 8 days after hatching with either 6 µl of semen , Hayes saline or seminal fluid ( four queens per treatment in each round , 12 queens per treatment , 36 queens in total ) . We sedated queens on ice , because previous studies showed that using CO2 reduces the likelihood of queens embarking on mating flights ( Kocher et al . , 2010; Niño et al . , 2011; Niño et al . , 2013a ) . We fitted each queens with a RFID tag ( mic3-TAG 64-bit RO ) , and re-introduced queens individually into queen-less nucleus hives . In a set-up that we previously used to monitor honeybee behaviour ( Dosselli et al . , 2016 ) , we narrowed the hive entrances and forced individual queens to pass through a set of two RFID tag readers ( iID MAJA module 4 . 1 ) when they were leaving or returning to their hives to participate in mating flights . This set-up allowed us to reliably monitor the flight behaviour of queens as queens leaving their hive would trigger the inner reader closer to the entrance before the outer reader closer to the exit , while returning queens would trigger the readers in the opposite order . Raw data recoded by the readers were collected in XML format on a SD memory card in the database box ( ilD HOST type MAJA 4 . 1 ) from where they were downloaded to a PC and assembled in a MySQL database . To identify potential mating flights we: ( i ) only evaluated complete sequences of reader recordings ( in – out - out – in ) , ( ii ) only evaluated data recorded after 12:00 noon because honeybee queens only fly out in the afternoon to mate ( Winston , 1987 ) , and ( iii ) only retained data for flights of 406 s or longer , because the shortest average mating flight of a Hayes-inseminated queen that later laid eggs in our study was 406 s , and because similar thresholds have been applied in previous studies ( Heidinger et al . , 2014; Dosselli et al . , 2016 ) . We also recorded a queen's number of flights per day and her total flight duration . Two queens were discarded from subsequent analyses because the readers did not record any completed afternoon flight for them . This experiment was conducted at the University of Western Australia ( 31° 59' 5 . 143'' S , 115° 49' 15 . 553'' E ) from the end of January to the end of March 2017 . All hives were checked after the experiment for the presence of newly laid eggs and brood . To statistically compare effects in our field experiment we used IBM , SPSS version 23 for Mac . | For social insects like honeybees it is beneficial if their queens mate with many males , because genetic diversity can protect the hive against parasites . Early in life , a honeybee queen has a short period of time in which she can fly out to mate with males before returning to the hive with all the sperm needed to last for a lifetime . Queens that have mated on their first flight may embark on additional mating flights over a few consecutive days to further increase genetic variability in their offspring . This is problematic for a male that has already mated because the more males that inseminate the queen the fewer offspring will carry on his specific genes . This results in sexual conflict between males and queens over the number of mating flights . In many animals , males manipulate females using molecules in seminal fluid to reduce the chances of the female mating again and honeybee males may use a similar strategy . Previous studies revealed that insemination alters the activity of genes related to vision in a honeybee queen’s brain . This could be one way for the males to prevent queens from embarking on additional mating flights . Now , Liberti et al . find support for this idea by showing that seminal fluid can indeed trigger changes in the activity of vision-related genes in the brains of honeybee queens , which in turn reduce a queen’s opportunity to complete additional mating flights . Queens inseminated with seminal fluid were less responsive to light compared to queens that were exposed to saline instead . Electronic tracking devices affixed to queens showed that the seminal fluid-exposed queens left for mating flights sooner but were more likely to get lost and to not return to their hives compared to the saline-exposed queens . The experiments support the idea of a sexual arms race in honeybees . Males use seminal fluid to cause rapid deteriorating vision in queens , thus reducing their likelihood of leaving the hive to mate again and to find males when they do fly again . The queens try to counteract these effects by leaving for mating flights sooner , thereby increasing offspring genetic diversity and the success of their colonies . Further studies will be needed to find out how the honeybee sexual arms race varies across seasons , bee races , and geographic ranges . Such information will be useful for honeybee breeding programs , which rely on queen mating success and hive genetic diversity to ensure hive health . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2019 | Seminal fluid compromises visual perception in honeybee queens reducing their survival during additional mating flights |
Brain injuries can interrupt descending neural pathways that convey motor commands from the cortex to spinal motoneurons . Here , we demonstrate that a unilateral injury of the hindlimb sensorimotor cortex of rats with completely transected thoracic spinal cord produces hindlimb postural asymmetry with contralateral flexion and asymmetric hindlimb withdrawal reflexes within 3 hr , as well as asymmetry in gene expression patterns in the lumbar spinal cord . The injury-induced postural effects were abolished by hypophysectomy and were mimicked by transfusion of serum from animals with brain injury . Administration of the pituitary neurohormones β-endorphin or Arg-vasopressin-induced side-specific hindlimb responses in naive animals , while antagonists of the opioid and vasopressin receptors blocked hindlimb postural asymmetry in rats with brain injury . Thus , in addition to the well-established involvement of motor pathways descending from the brain to spinal circuits , the side-specific humoral signaling may also add to postural and reflex asymmetries seen after brain injury .
Brain lesions can interrupt descending neural pathways that convey motor commands from the cerebral cortex to motoneurons located in the brain stem and anterior horn of the spinal grey matter ( Cai et al . , 2019; Kuypers , 1981; Lemon , 2008; Purves et al . , 2001; Smith et al . , 2017; Tan et al . , 2012; Zörner et al . , 2014 ) . Brain injury-induced motor deficits typically develop on the contralateral side of the body and include motor weakness , loss of voluntary movements , spasticity , asymmetric limb reflexes , and abnormal posture . Evidence suggests that these deficits are related to the impaired signaling through the descending motor tracts and to deafferentation-induced spinal neuroplasticity ( Cai et al . , 2019; Deliagina et al . , 2014; Kanagal and Muir , 2009; Küchler et al . , 2002; Li and Francisco , 2015; Morris and Whishaw , 2016; Tan et al . , 2012; Whishaw et al . , 1998; Zelenin et al . , 2016; Zörner et al . , 2014; Zörner et al . , 2010 ) . Mechanisms of these impairments are not well understood . In animal experiments , a unilateral brain injury ( UBI ) induces hindlimb postural asymmetry ( HL-PA ) with contralesional limb flexion and asymmetry of the hindlimb withdrawal reflexes ( Watanabe et al . , 2021; Watanabe et al . , 2020; Zhang et al . , 2020 ) . Consistently , lateral hemisection of the spinal cord impairs postural functions ( Zelenin et al . , 2016 ) and enhances monosynaptic and polysynaptic hindlimb reflexes on the ipsilesional side ( Hultborn and Malmsten , 1983a; Hultborn and Malmsten , 1983b; Malmsten , 1983; Rossignol and Frigon , 2011 ) . Asymmetry in posture and reflexes persists after complete transection of the spinal cord ( Rossignol and Frigon , 2011; Watanabe et al . , 2020; Zhang et al . , 2020 ) . This may be due to neuroplastic changes in the lumbar spinal cord induced by brain or spinal cord injury through the descending neural tracts . In addition , the side-specific signals from a lesion site to the lumbar domains may be conveyed by endocrine messengers . Signaling that is not mediated by the descending neural tracts has been proposed but generally has been disregarded ( Bakalkin et al . , 1986; Cope et al . , 1980; Wolpaw and Lee , 1989 ) . Analysis of neurotransmitter mechanisms demonstrates that opioid peptides and Arg-vasopressin may induce the formation of HL-PA in rats with intact brains ( Bakalkin et al . , 1981; Bakalkin et al . , 1986; Chazov et al . , 1981; Klement'ev et al . , 1986; Watanabe et al . , 2020 ) . Either the left or the right hindlimb can be flexed , depending on the compound injected . The κ-opioid agonists dynorphin , bremazocine and U-50 , 488 , as well as the preferential endogenous µ-opioid agonist Met-enkephalin , induce flexion of the left hindlimb , whereas the δ-opioid agonist Leu-enkephalin and Arg-vasopressin cause the right limb to flex . These effects mimic the UBI-induced formation of the HL-PA , and suggest that these neurohormones may be involved in the left-right side specific postural and sensorimotor effects of the UBI . The left-right side-specific neurohormonal effects may be induced through lateralized receptors . Indeed , in the rat spinal cord , the expression of the opioid receptors and their co-expression patterns are different between the left and right sides ( Kononenko et al . , 2017; Watanabe et al . , 2021 ) . In this study , we tested the hypothesis that the effects of UBI on the hindlimb posture and sensorimotor functions are mediated by a side-specific neuroendocrine pathway that operates in parallel with the descending neural tracts . Our strategy was to disable the descending neural influences in order to reveal the endocrine signaling . For this purpose , the spinal cord was completely transected before the brain injury was performed . The HL-PA and asymmetry of withdrawal reflexes that are regulated by neurohormones were studied as the readouts of UBI effects . Hypophysectomy and pharmacological antagonists of opioid and vasopressin receptors were used to disable hormonal signaling . We also tested whether the administration of serum collected from animals with UBI , as well as the administration of pituitary neurohormones β-endorphin and Arg-vasopressin , may replicate the effects of brain injury by inducing HL-PA in rats with intact brain .
The hypothesis that a unilateral brain lesion may induce HL-PA through a pathway that bypasses the descending neural tracts was tested in rats that had complete transection of the spinal cord before the UBI was performed ( Figure 1A–E; Figure 1—figure supplements 1–5 ) . The spinal cord was transected at the T2-T3 level and then the hindlimb representation area of the sensorimotor cortex was ablated ( Figure 1A; Figure 1—figure supplement 1A ) . HL-PA was analyzed within 3 hr after the UBI by both the hands-on and hands-off methods of hindlimb stretching followed by photographic and / or visual recording of the asymmetry in animals under pentobarbital anesthesia ( for details , see ‘Materials and methods’ and Figure 1—figure supplement 2 ) . HL-PA data are presented as the median values of HL-PA in mm ( HL-PA size ) , and the probability to develop HL-PA ( denoted as PA on the figures ) that depicts the proportion of rats with HL-PA above the 1 mm threshold . The analysis was generally blind to the observer ( for details , see ‘Materials and methods’ ) . Control experiments demonstrated that this injury produced HL-PA with contralesional hindlimb flexion within 3 hr after the UBI in rats with intact spinal cord ( Figure 1F–H; Figure 1—figure supplement 1B–D ) , and contralesional hindlimb motor deficits in the beam-walking and ladder rung tests ( Figure 1—figure supplement 1E , F ) . Strikingly , in the rats with transected spinal cords the UBI also induced HL-PA ( Figure 1C–E ) . The HL-PA developed within 3 hr after the brain injury . Its size and probability were much greater than in rats with sham surgery ( Left UBI , n = 31; Right UBI , n = 15; sham surgery , n = 29 ) . An unanticipated observation was that in rats with HL-PA , the hindlimb was flexed on the contralesional side . The left or right hindlimb flexion was induced by the right and left UBI , respectively ( Figure 1D; Figure 1—figure supplement 3B , C; Figure 1—figure supplement 4B , C , F , G ) . Both Wistar rats ( Figure 1D–E; Figure 1—figure supplement 3 , 4F-I ) and Sprague Dawley rats ( Figure 1—figure supplements 4B–E and 5 ) that were used in further molecular and electrophysiological experiments , respectively , developed HL-PA with hindlimb flexion on the contralesional side . To ensure the completeness of the transection , a 3–4 mm spinal segment was excised at the T2-T3 level in a subset of rats ( Figure 1—figure supplement 5 ) . After the excision , the left-side UBI-induced hindlimb postural asymmetry with the right limb flexion that replicated the other findings . The HL-PA size and probability , the time course of HL-PA development and formation of contralesional hindlimb flexion in rats with transected spinal cords that received UBI ( Figure 1D , E; Figure 1—figure supplement 3 , Figure 1—figure supplement 4 ) were similar to those of the UBI animals with intact spinal cords ( Figure 1G , H; Figure 1—figure supplement 1C , D ) . We conclude that HL-PA formation in animals with transected spinal cord is mediated through a pathway that operates in parallel with the descending neural tracts and assures the development of contralesional flexion . The withdrawal reflexes are instrumental in the investigation of brain injury-induced functional changes in hindlimb neural circuits activated by afferent input ( Dewald et al . , 1999; Schouenborg , 2002; Serrao et al . , 2012; Spaich et al . , 2006; Zhang et al . , 2020 ) . We next sought to determine whether UBI in rats with transected spinal cords produces changes in the hindlimb withdrawal reflexes , and whether these changes are asymmetric . Special care was taken to ensure that EMG recordings obtained from the left and right hindlimbs were quantitatively comparable . To achieve this , a number of strict technical criteria , such as maximally symmetrical positioning of the stimulation and recording electrodes , were applied . The criteria used in this study are described in details in ‘Materials and methods’ , and are similar to those proposed by Hultborn and Malmsten ( Hultborn and Malmsten , 1983a; Hultborn and Malmsten , 1983b; Malmsten , 1983 ) . Furthermore , to minimize inter-individual variations , the asymmetry indices were used instead of the absolute values of the reflex size . This allowed double assessment: first , within both the UBI and control groups that identified asymmetric reflexes in each group , and , second , between these groups that revealed the effects of UBI vs . sham surgery . Because multiple responses were measured for the same animal , including two of its limbs , four muscles , and the varying stimulation conditions , and because they were analyzed within an animal group and between the groups , we applied mixed-effects models using Bayesian inference . Only strong and significant UBI effects were considered as biologically relevant . Electromyographic responses were recorded from the extensor digitorum longus , interosseous , peroneus longus , and semitendinosus muscles of the contra- and ipsilesional hindlimbs in the rats with UBI ( n = 18 ) or sham surgery ( n = 11 ) performed after complete spinal transection and analyzed as the asymmetry index ( AI = log2[Contra / Ipsi] , where Contra and Ipsi were values for muscles of the contralesional and ipsilesional limbs ) ( Figure 2; Figure 2—figure supplement 1; Figure 2—figure supplement 2 ) . When reflexes on both sides are equal ( i . e . the Contra / Ipsi ratio equals 1 ) , the asymmetry index is zero; if reflexes are doubled in size on the Contra or Ipsi side ( i . e . the Contra / Ipsi ratio equals 2 . 0 or 0 . 5 ) the asymmetry index is +1 or –1 , respectively . Analysis of the electrically evoked electromyographic responses revealed that the asymmetry index was different from zero in the current threshold for the semitendinosus muscle ( 3 . 6-fold lower on the contra- vs . ipsilesional side ) , and in the number of spikes for the extensor digitorum longus ( 3 . 5-fold higher on the contra- vs . ipsilesional side ) and semitendinosus ( 5 . 9-fold higher on the contra- vs . ipsilesional side ) muscles in UBI rats ( Figure 2C , D ) . No contra- vs . ipsilesional asymmetry was evident in the sham surgery group . Representative UBI-induced asymmetry in the number of spikes for the semitendinosus muscle is shown in Figure 2A , B ( for those of extensor digitorum longus , interosseous and peroneus longus muscles , see Figure 2—figure supplement 1 ) . When compared to sham surgery , UBI substantially decreased the asymmetry index for the current threshold of the semitendinosus ( 4 . 0-fold ) , and the asymmetry index for the number of spikes of the interosseous ( 2 . 8-fold ) that may be due to the decline in the responses on the contralesional side and/or their elevation on the ipsilesional side . Concomitantly , UBI elevated the asymmetry index for the number of spikes of the extensor digitorum longus ( 5 . 2-fold ) and semitendinosus ( 6 . 7-fold ) suggesting activation of the responses on the contralesional side and/or their inhibition on the ipsilesional side ( Figure 2E , F ) . No changes in peroneus longus were revealed . Each group consisted of rats with left and right sided surgeries ( Figure 2—figure supplement 2 ) . Analysis of the asymmetry index for the four groups ( i . e . the left UBI , left sham surgery , right UBI and right sham surgery groups ) revealed virtually the same asymmetries in the UBI group , and the same UBI vs . sham differences in the asymmetry index , but not for all comparisons ( Figure 2—figure supplement 3 ) . Thus , the right UBI produced higher responses of the left ( contralesional ) extensor digitorum longus ( 4 . 1-fold ) and the left ( contralesional ) semitendinosus ( 53 . 8-fold ) compared to those on the right side , while responses of the left vs . right interosseous were decreased ( 5 . 5-fold ) ( Figure 2—figure supplement 3D ) . Additionally , the interosseous muscle was found to be asymmetric after the right side UBI . No effects on thresholds were identified . Thus , in rats with transected spinal cord , UBI , but not sham surgery , induced asymmetry in withdrawal reflexes . The number of spikes of both flexor muscles , the extensor digitorum longus and semitendinosus was higher on the contra vs . ipsilateral side in the UBI rats . Consistently , the threshold was lower for the contra vs . ipsilesional semitendinosus . These effects may be due to ( i ) higher sensitivity of the afferent system reflected in a lower threshold on the contra vs . ipsilesional side for the semitendinosus; and ( ii ) an increased excitability of efferent systems for both muscles reflected in the increased number of spikes on the contra vs . ipsilesional side . At the cellular level , the increased excitatory drive may develop due to changes in local spinal circuits including those in presynaptic afferent inhibition , and/or changes in intrinsic membrane properties of motoneurons . Regardless of mechanism , robust differences in the asymmetry index between the UBI and sham groups suggested that the UBI markedly elevated both the sensitivity of semitendinosus afferents and the excitability of the extensor digitorum longus and semitendinosus efferents , all on the contralesional vs . ipsilesional side . The UBI also inhibited the contralesional interosseous . The UBI effects on the hindlimb withdrawal reflexes in rats with the transected spinal cords were similar in their range and contra-ipsilesional patterns to those of the UBI animals with intact spinal cords ( Watanabe et al . , 2021; Zhang et al . , 2020 ) . These effects corroborate clinical findings showing contralateral facilitation of withdrawal reflexes in stroke patients ( Dewald et al . , 1999; Serrao et al . , 2012; Spaich et al . , 2006 ) . We examined whether the UBI performed after complete spinal transection produced molecular changes in the lumbar spinal segments . Expression of 20 neuroplasticity-related , opioid and vasopressin genes , and the levels of three opioid peptides were analyzed in the ipsilesional and contralesional lumbar spinal cord of the rats with transected spinal cord that also had the left UBI ( n = 12 ) or left sham surgery ( n = 11 ) . Genes coding for regulators of axonal sprouting , synapse formation , neuronal survival and neuroinflammation ( Arc , Bdnf , Dlg4 , Homer-1 , Gap43 , Syt4 , and Tgfb1 ) , transcriptional regulators of synaptic plasticity ( cFos , Egr1 , and Nfkbia ) , and essential components of the glutamate system critical for neuroplastic responses and regulation of spinal reflexes ( GluR1 , Grin2a , and Grin2b ) were selected as neuroplasticity genes ( for detailed description , see ‘Materials and methods , Neuroplasticity-related genes’ ) . Genes of the opioid and vasopressin systems were included because of their involvement in asymmetric spinal responses to brain injury ( see next section ) . First , the mRNA levels and their median asymmetry index ( AI = log2[Contra/Ipsi] , where Contra and Ipsi were the levels in the contralesional and ipsilesional lumbar spinal cord ) were compared between the UBI and sham surgery groups . Gene expression was either elevated on the ipsilesional side ( Syt4 , Grin2a , Grin2b , and Oprk1; Figure 3—figure supplement 1A–D ) or decreased on the contralesional side ( Gap43 and Penk; Figure 3—figure supplement 1E , F ) ( for all six genes , Punadjusted < 0 . 05 ) . Consistently , the gene expression asymmetry index was decreased for Syt4 ( Padjusted = 0 . 004 ) , and for Oprk1 , Oprm1 , Dlg4 , and Homer1 ( for all four genes , Punadjusted < 0 . 05 ) ( Figure 3—figure supplement 1G–K ) . These differences were subtle , between 0 . 13 and 0 . 38-fold , and therefore were discarded as evidence of the UBI effects . These differences , however , pointed to the different direction in responses of the left and right spinal cord to the injury . Therefore , in the second step , we assessed whether the proportion of genes with lower expression on the contralesional vs . ipsilesional side was different between the UBI and sham surgery groups . The median gene expression asymmetry index of 19 out of 20 genes at the pairwise comparison was lower in the UBI rats compared to sham surgery group ( sign-test: p = 4×10−5 ) ( Figure 3A , B ) . Changes in the gene expression asymmetry index were consistent with decreased expression of 17 genes ( sign test: p = 0 . 003 ) in the contralesional half ( Figure 3—figure supplement 1E , F; Figure 3—figure supplement 2A–D ) concomitantly with elevated expression of 15 genes ( sign test: p = 0 . 041 ) in the ipsilesional half ( Figure 3—figure supplement 1A–D; Figure 3—figure supplement 2A–D ) . Gene co-expression patterns characterize regulatory interactions within and across tissues ( Dobrin et al . , 2009; Erola et al . , 2020; Gerring et al . , 2019; Zhang et al . , 2020 ) . Third , we examined whether the UBI induced changes in mRNA–mRNA correlations within the left and right half of the lumbar spinal cord ( intra-area correlations ) , and between these halves ( inter-area correlations ) . The proportion of intra-area positive correlations , which dominated in rats with sham surgery , was reduced after the UBI ( Fisher's Exact Test: all correlations in the right half , p = 3×10−5; significant correlations in the left and right areas , p = 0 . 008 and 0 . 009 , respectively ) ( Figure 3—figure supplement 2E–H ) . The inter-area gene-gene coordination strength was decreased after the UBI ( Wilcoxon signed-rank test; all and significant correlations: p = 4×10−7 and 3×10−4 , respectively ) ( Figure 3C , D ) . Positive inter-area correlations were predominant in rats with sham surgery ( 68% ) in contrast to the UBI rats ( 42% ) ( Fisher's Exact Test: all and significant correlations , p = 6×10−14 and 0 . 004 , respectively ) . Thus , the UBI robustly impairs coordination of expression of neuroplasticity-related and neuropeptide genes within and between the left and right halves of the lumbar spinal cord . Fourth , analysis of opioid peptides demonstrated that the UBI substantially elevated the levels of the proenkephalin marker Met-enkephalin-Arg-Phe in the ipsilesional ( Padjusted = 9×10−4 ) and contralesional ( Punadjusted = 0 . 020 ) spinal halves ( Figure 3E ) , and the prodynorphin-derived Dynorphin B and Leu-enkephalin-Arg in the ipsilesional spinal cord ( for both , Punadjusted < 0 . 05 ) ( Figure 3—figure supplement 3 ) . Altogether , the analysis of gene expression and of opioid peptides adds strong molecular evidence for the lateralized signaling from the injured brain to the lumbar neural circuits in rats with transected spinal cord . The left-right side specific mechanism that does not engage the descending neural tracts may operate through the neuroendocrine system by a release of pituitary hormones into the blood . Consistent with this hypothesis , no HL-PA developed in hypophysectomized animals that received left UBI after spinal transection ( n = 8 ) ; the HL-PA median values and PA were nearly identical to those in sham operated rats ( n = 8 ) ( Figure 4A; Figure 4—figure supplement 1A–E ) . We next examined whether left UBI stimulates the release of chemical factors that may induce the development of HL-PA , into the blood . Serum that was collected 3 hr after performing a left UBI in rats with transected spinal cord was administered either centrally ( into the cisterna magna; UBI serum , n = 13; sham serum , n = 7; Figure 4—figure supplement 1F–J ) or intravenously ( UBI serum , n = 13; sham serum , n = 7; Figure 4B; Figure 4—figure supplement 1K–O ) to rats after their spinalization . Serum administration by either route resulted in formation of HL-PA with its values and its probability similar to those induced by the UBI in rats with intact and transected spinal cords . Remarkably , animals injected with serum from rats with left UBI displayed hindlimb flexion on the right side , which was the same as the flexion side in the donor rats ( Figure 4B; Figure 4—figure supplement 1F–O ) . No HL-PA developed after administration of serum collected from rats with the left sham surgery . We conclude that the left UBI stimulates a release of chemical factors from the pituitary gland into the blood that induce HL-PA with contralesional flexion . Previous studies demonstrated that multiple peptide factors extracted from the brain , pituitary gland and serum may induce a side-specific hindlimb motor response . Several of them were identified as peptide neurohormones including opioid peptides ( Bakalkin and Kobylyansky , 1989; Bakalkin et al . , 1986; Chazov et al . , 1981 ) and Arg-vasopressin ( Klement'ev et al . , 1986 ) . It was found that Arg-vasopressin or Leu-enkephalin administered centrally induced HL-PA with right hindlimb flexion . The pituitary gland is the main source of the opioid neurohormone β-endorphin and the antidiuretic hormone Arg-vasopressin in the body . Here , we first replicated the effects of Arg-vasopressin that , consistent with a previous study ( Klement'ev et al . , 1986 ) , produced flexion of the right hindlimb after its intracisternal administration ( Figure 4—figure supplement 2; peptide , n = 22 , and saline , n = 9 at the 180 min time point ) . We then tested if β-endorphin and Arg-vasopressin may evoke asymmetric motor response after intravenous administration . Injection of these neurohormones , but not saline , to rats with transected spinal cords resulted in development of HL-PA with right hindlimb flexion ( Figure 4C; β-endorphin , n = 8; Arg-vasopressin , n = 7; saline , n = 4 ) . We next investigated whether opioid receptors and the vasopressin receptor V1B , that is expressed in the pituitary gland ( Roper et al . , 2011 ) , mediate formation of HL-PA in UBI rats or in animals treated with serum from UBI rats . Naloxone and SSR-149415 , the opioid and vasopressin V1B receptor antagonists , respectively , administered to animals with transected spinal cord that also received a left UBI ( naloxone , n = 6; SSR-149415 , n = 6; saline and vehicle , n = 11 ) inhibited HL-PA formation ( Figure 4D ) . Similarly , the HL-PA induced by serum from animals with left UBI was abolished by administration of either naloxone ( n = 6 ) or SSR-149415 ( n = 6 ) ( Figure 4E ) . Thus , the pituitary neurohormones β-endorphin and Arg-vasopressin released into the systemic circulation may serve as side-specific signals that mediate UBI effects on hindlimb motor circuits .
During embryonic development , the left–right asymmetry of the body is generated by multiple paracrine signaling molecules that enable communications between the left and right halves of the embryo ( Hamada et al . , 2002 ) . In the adult brain , functional lateralization is an organizing principle ( Duboc et al . , 2015; MacNeilage et al . , 2009 ) and lateralized functions may be regulated by paracrine signaling molecules including peptide neurohormones ( Deliagina et al . , 2000; Hussain et al . , 2012; Kononenko et al . , 2017; Marlin et al . , 2015; Nation et al . , 2018; Phelps et al . , 2019; Watanabe et al . , 2015; Watanabe et al . , 2021; Zink et al . , 2011 ) . Thus , oxytocin enables retrieval behavior by enhancing responses of the left , but not right , auditory cortex through its receptors expressed on the left side ( Marlin et al . , 2015 ) . Arg-vasopressin targets the left but not right hemispheric areas to modulate social recognition-related activity ( Zink et al . , 2011 ) . In the human brain , asymmetric distribution of the μ-opioid receptor along with opioid peptides that elicit euphoria and dysphoria may provide a basis for the lateralized processing of positive and negative emotions ( Kantonen et al . , 2020; Watanabe et al . , 2015 ) . Top-down control of the hypothalamic-pituitary-adrenal and hypothalamic-pituitary-gonadal axes , as well as the immune system , is left-right hemisphere specific ( Bakalkin et al . , 1984; Inase and Machida , 1992; Kiss et al . , 2020; Lueken et al . , 2009; Madsen et al . , 2012; Meador et al . , 2004; Sullivan and Gratton , 2002; Xavier et al . , 2013 ) . The right-sided injuries to the brain or spinal nerves compared to those on the left side produce stronger effects on the neuroendocrine systems , including hormonal levels in the peripheral circulation ( Hussain et al . , 2012; Inase and Machida , 1992; Kononenko et al . , 2017; Lueken et al . , 2009 ) . Cortisol ( corticosterone ) secretion is under excitatory control of the right hemisphere and , consistently , its phasic response is diminished in patients with right but not left sided stroke ( Lueken et al . , 2009 ) . These features may be related to the asymmetry of the hypothalamic – pituitary axis . Thus , basal and corticotropin-releasing hormone–induced secretion of Arg-vasopressin and ACTH by the pituitary gland is lateralized to the right petrosal sinus ( Kalogeras et al . , 1996 ) . Conversely , non-directional stimuli such as stress and stress-induced pain produce lateralized responses in the CNS ( Bakalkin et al . , 1982; Bakalkin et al . , 1984; Nation et al . , 2018; Sullivan and Gratton , 2002 ) . Both left- and right-sided nerve or body injuries elicit functional and molecular responses that are often lateralized to the right , but not to the left , in the brain and spinal cord ( Bakalkin et al . , 1984; Hussain et al . , 2012; Inase and Machida , 1992; Kononenko et al . , 2017; Phelps et al . , 2019 ) . Neuropathic pain induced by either the left- or right-sided nerve injuries is controlled through κ-opioid receptors in the right amygdala ( Phelps et al . , 2019 ) . Taken together , these findings suggest that feedforward and feedback interactions between the lateralized CNS features and the endocrine system controlling peripheral processes are mediated either by neural circuits with unusual , asymmetric organization , or by left-right sided neuroendocrine pathways that may be similar to the left-right sided paracrine mechanism operating in the development . An alternative pathway that does not engage descending neural tracts may convey side-specific signals from the brain to the paired endocrine glands , the left and right spinal cord , and the left and right extremities had been suggested ( Bakalkin et al . , 1984; Cope et al . , 1980; Wolpaw and Lee , 1989 ) , and supported by preliminary evidence ( Bakalkin et al . , 1986 ) but not elaborated . This study provides evidence for left-right side-specific humoral signaling that mediates the effects of UBI on the formation of HL-PA , asymmetry in withdrawal reflexes , and asymmetric changes in gene expression patterns in the lumbar spinal cord ( Figure 5 ) . Encoding of information about the injury side in a hormonal message , humoral transmission of this message to its target sites on peripheral nerve endings or spinal neurons , and translation of this message into the left-right side-specific response , are the key stages of this phenomenon . The left- and right-side-specific responses evoked by hormonal molecules circulating in the blood is a core of the humoral signaling pathway . Together with previous reports ( Bakalkin et al . , 1981; Bakalkin and Kobylyansky , 1989; Bakalkin et al . , 1986; Chazov et al . , 1981; Watanabe et al . , 2020 ) , this study demonstrates that peptide neurohormones and opioids administered intravenously , intrathecally or intracisternally induce HL-PA in rats with intact brain . The critical finding is that the side of the flexed limb depends on the compound administered . Endogenous and synthetic κ-opioid agonists dynorphin , bremazocine , and U-50 , 488 , along with the endogenous μ/δ-opioid agonist Met-enkephalin , induce flexion of the left hindlimb ( Bakalkin et al . , 1981; Bakalkin and Kobylyansky , 1989; Bakalkin et al . , 1986; Chazov et al . , 1981; Watanabe et al . , 2020 ) . In contrast , β-endorphin and Arg-vasopressin , and the δ-agonist Leu-enkephalin , cause the right limb to flex [the present study and Bakalkin et al . , 1981; Chazov et al . , 1981; Klement'ev et al . , 1986] . Thus , topographical information conveyed by the ‘non-directional’ molecular messengers circulating in the blood is converted into side-specific motor responses . Hypophysectomy disables the endocrine pathway including its opioid and Arg-vasopressin components and abolishes the HL-PA . Consistent with this finding , serum from rats with left UBI induces HL-PA with contralesional hindlimb flexion in rats with intact brain . The pituitary gland is the main source of Arg-vasopressin and β-endorphin in the body that are secreted into the bloodstream ( Autelitano et al . , 1989; Day and Akil , 1989 ) . Naloxone and SSR-149415 block the left UBI-induced formation of HL-PA . Altogether , these findings demonstrate that opioid and Arg-vasopressin neurohormones transmit side-specific signals from the injured brain to the spinal neural circuits . Theoretically , the paravertebral chain of sympathetic ganglia , which is the remaining neural connection after complete spinal cord transection , may convey supraspinal signals to the muscle vasculature and through this mechanism may differentially affect ipsi- and contralesional muscles . However , the sympathetic ganglia likely do not mediate control of lumbar neural circuits by the supraspinal structures ( Brodal , 1981; Wolpaw and Lee , 1989 ) . Furthermore , the sympathetic system has a limited capacity to independently regulate blood flow to the left and right hindlimbs ( Lee et al . , 2007 ) . Our present findings could not rule out a role of the sympathetic pathway . However , experiments with hypophysectomized rats , ‘pathological’ serum and neurohormones inducing the HL-PA strongly suggest the dominance of the humoral pathway in rats with transected spinal cord . Clinical studies revealed robust functional changes induced by stroke or traumatic brain injury ( TBI ) in contralateral withdrawal reflexes that also control posture and locomotion ( Dewald et al . , 1999; Sandrini et al . , 2005; Serrao et al . , 2012 ) . Patients with post-stroke motor deficits lose their ability to modulate the withdrawal reflexes that affects spatiotemporal interaction among joints and causes movement abnormalities during motor activities ( Bohannon and Smith , 1987; Serrao et al . , 2012 ) . Spinal withdrawal reflexes are regulated by the endogenous opioid peptides that may suppress the ipsilateral and contralateral segmental reflexes ( Clarke et al . , 1992; Duarte et al . , 2019; Jankowska and Schomburg , 1998; Schmidt et al . , 1991 ) . Opioid receptors are expressed both in the dorsal and ventral horns of the spinal cord ( Kononenko et al . , 2017; Wang et al . , 2018 ) and also in the periphery , in primary afferents including low-threshold mechanoreceptors that modulate cutaneous mechanosensation ( Bardoni et al . , 2014; Snyder et al . , 2018 ) . The left-right-specific endocrine mechanism may mediate the effects of UBI on the withdrawal reflexes through targeting peripherally or centrally located opioid receptors . A large fraction of patients with stroke and cerebral palsy do not relax their muscles – they are tonically contracted without any voluntary command ( Baude et al . , 2019; Gracies , 2005; Lorentzen et al . , 2018; Sheean and McGuire , 2009; Trompetto et al . , 2019 ) . This phenomenon is called ‘spastic dystonia’ and has a central mechanism that does not depend on afferent input ( Gracies , 2005; Lorentzen et al . , 2018; Sheean and McGuire , 2009 ) . In the HL-PA analysis , no nociceptive stimulation is applied and tactile stimulation is negligible [this study and Zhang et al . , 2020] . Stretch and postural limb reflexes are abolished immediately after complete spinal cord transection ( Frigon et al . , 2011; Miller et al . , 1996; Musienko et al . , 2010 ) and strongly inhibited by anesthesia ( Fuchigami et al . , 2011; Zhou et al . , 1998 ) . Therefore , in anesthetized rats with transected spinal cord , a role of nociceptive and stretch reflexes in the UBI-induced HL-PA formation may be limited . The finding that HL-PA is resistant to bilateral lumbar dorsal rhizotomy ( Zhang et al . , 2020 ) further supports this notion and suggests that the HL-PA and the clinical ‘spastic dystonia’ may be mechanistically similar . Classical hyperreflexia develops during several weeks after the impact and is considered as pathology of the corticospinal tract ( Williams et al . , 2017 ) . In contrast , the HL-PA is formed within 30 min after the UBI in rats with transected spinal cord . Whether the endocrine signaling has a role in exacerbation of stretch reflex after damage to the corticospinal tract would be interesting to study . Single injection of either naloxone or SSR-149415 inhibited HL-PA asymmetry formation . The vasopressin receptor V1B is largely expressed in the anterior pituitary by corticotrophs producing proopiomelanocortin ( Roper et al . , 2011 ) . A plausible scenario is that Arg-vasopressin released from neurohypophysis activates the V1B receptor and stimulates secretion of proopiomelanocortin-derived β-endorphin ( Roper et al . , 2011 ) that induces the HL-PA . Alternatively , Arg-vasopressin and β-endorphin may produce synergistic effects acting through the complex of the vasopressin receptor V1B and μ-opioid receptor that integrates two signaling pathways ( Koshimizu et al . , 2018 ) . The general opioid antagonists naloxone and naltrexone may normalize neurological functions that are impaired in animals and human patients after unilateral cerebral ischemia ( Baskin and Hosobuchi , 1981; Baskin et al . , 1984; Baskin et al . , 1994; Hans et al . , 1992; Hosobuchi et al . , 1982; Jabaily and Davis , 1984; Namba et al . , 1986; Skarphedinsson et al . , 1989; Wang et al . , 2019 ) , and also reduce spasticity in patients with primary progressive multiple sclerosis ( Gironi et al . , 2008 ) . Our findings suggest that the efficacy of pharmacological treatment may depend on topographical correspondence between the side of neuronal deficits and the side that is preferentially targeted by neurohormones and their antagonists . A number of anatomical , functional , and molecular studies revealed left–right asymmetry in the spinal cord organization ( de Kovel et al . , 2017; Deliagina et al . , 2000; Hultborn and Malmsten , 1983a; Hultborn and Malmsten , 1983b; Knebel et al . , 2018; Kononenko et al . , 2017; Malmsten , 1983; Nathan et al . , 1990; Ocklenburg et al . , 2017; Zhang et al . , 2020 ) . Three-quarters of cervical spinal cords are asymmetric with a larger right side ( Nathan et al . , 1990 ) . Spinal-muscular systems are asymmetric in human fetuses , and the asymmetry correlates with lateralized gene transcription ( de Kovel et al . , 2017; Ocklenburg et al . , 2017 ) . Mono- and polysynaptic segmental reflexes evoked by stimulation of the dorsal roots and recorded in the ventral roots in intact rats and cats display higher activity on the right side ( Hultborn and Malmsten , 1983a; Hultborn and Malmsten , 1983b; Malmsten , 1983 ) . Similarly , EMG recordings of hindlimb withdrawal reflexes evoked by electrical stimulation in control rats display higher activity on the right side ( Zhang et al . , 2020 ) . On this asymmetric background , neural circuits controlling the left and right limbs may be differently regulated by opioid peptides and Arg-vasopressin ( Bakalkin et al . , 1981; Bakalkin et al . , 1986; Chazov et al . , 1981; Klement'ev et al . , 1986 ) . The left-right side-specific neurohormonal effects may be mediated through lateralized receptors . In the rat spinal cord , the expression of three opioid receptors is lateralized to the left , and their proportions and co-expression patterns are different between the left and right sides ( Kononenko et al . , 2017; Watanabe et al . , 2021 ) . The asymmetry may be a critical feature of the spinal cord allowing translation of the ‘non-directional’ hormonal messages into the left-right side-specific response . The side-specific endocrine signaling was revealed in anaesthetized animals with transected spinal cords that were studied up to 180 min post UBI . Its biological and pathophysiological relevance has not been determined . Pathways from the injured brain area to the hypothalamic-pituitary system , neurohormones mediating effects of the right side injury , peripheral or central targets for the left- and right-side-specific endocrine messengers , and afferent , central or efferent mechanisms of the asymmetry formation have not been investigated . The study did not analyze forelimb postural asymmetry . It was not induced by lesion of the hindlimb sensorimotor cortex ( Zhang et al . , 2020 ) but it did develop after injury of the forelimb area ( unpublished data ) . The strategy was to selectively disable the neural and endocrine mechanisms by surgical means . In this approach , dissection of neural pathways does not allow us to assess a contribution of each pathway to asymmetric postural and motor deficits in awake animals . On the contrary , analysis of the UBI effects in the hypophysectomized rats with intact spinal cords may uncover a function of the left – right side-specific endocrine signaling . Rats with intact ( Figure 1—figure supplement 1C , D ) and transected ( Figure 1—figure supplements 3 and 4 ) spinal cords developed HL-PA during the first 30 min following UBI . It would be worthwhile to analyze in more details whether the time course of development of asymmetry differs between these groups; and also to ascertain whether signals mediated by the neural and endocrine pathways are additive , synergistic , or even antagonistic with respect to each other during the initial impairment phase and the recovery period using naive animals and the hypophysectomized rats with intact spinal cords . This study does not focus on clinical correlates and mechanisms of postural deficits . The withdrawal reflexes and hindlimb posture were studied as readouts of the UBI because they are regulated by neurohormones and may be analyzed after spinal cord transection . Furthermore , they are directed along the left-right axis , and , therefore , can reveal whether the endocrine system conveys the side-specific signals . At the same time , the HL-PA and withdrawal reflexes model several features of the brain injury-induced sensorimotor and postural deficits in humans . First , the changes induced by UBI have a contra-ipsilesional pattern . Second , the HL-PA is not dependent on the afferent input ( Zhang et al . , 2020 ) and in this regard it may be similar to ‘spastic dystonia’ , a tonic muscle overactivity that contributes to ‘hemiplegic posture’ ( Gracies , 2005; Lorentzen et al . , 2018; Sheean and McGuire , 2009 ) . Third , asymmetric exacerbated withdrawal reflexes that lead to flexor spasms in patients ( Bussel et al . , 1989; Dietz et al . , 2009; Lavrov et al . , 2006; Schouenborg , 2002 ) are similarly developed in rats . TBI and stroke cause dysfunction of the hypothalamic–pituitary system and hypopituitarism manifested as changes in secretion of pituitary hormones ( Bondanelli et al . , 2005; Emelifeonwu et al . , 2020; Klose and Feldt-Rasmussen , 2018; Lillicrap et al . , 2018 ) . Ischemic stroke activates the hypothalamus-pituitary-adrenal axis ( Anne et al . , 2007 ) , while a unilateral ablation of sensorimotor cortex elevates the level of circulating ACTH and induces morphological changes in the pituitary corticotrophs that produce ACTH and β-endorphin ( Lavrnja et al . , 2014 ) . Neurobiological mechanisms underlying these effects may be similar with those of the UBI-induced endocrine signaling described in this study . These mechanisms and anatomical pathways involved have not been identified . However , cortical projections to the hypothalamus that may potentially mediate effects of focal brain injury on secretion of pituitary hormones have been described ( Jeong et al . , 2016 ) . This study describes the left-right side-specific endocrine mechanism that , in addition to descending neural tracts , may mediate asymmetric effects of a unilateral brain injury on hindlimb postural asymmetry and spinal reflexes ( Figure 5 ) . Identification of features and the proportion of asymmetric sensorimotor deficits transmitted by neurohormonal signals vs . those mediated by neural pathways may be essential for understanding of stroke and TBI mechanisms .
Male Wistar Hannover rats ( Charles River Laboratories , Spain ) , with body weight of 150–200 g were used in behavioral , HL-PA and molecular experiments ( Figures 1 , 3 and 4B–E; Figure 1—figure supplement 1 , 3 , 4F-I; Figure 3—figure supplement 1 , 2 , 3; Figure 4—figure supplement 1G–O , 2 ) . Male Sprague Dawley rats were used for analysis of HL-PA ( Figure 1—figure supplements 2 , 4B–E and 5 ) , in electrophysiological experiments ( Figure 2 , Figure 2—figure supplement 1 , 2 , 3 ) ( Taconic , Denmark; 150–400 g body weight ) and for hypophysectomy ( Figure 4A and Figure 4—figure supplement 1B–E ) ( Charles River Laboratories , France; 115–125 g body weight ) . The animals received food and water ad libitum and were kept in a 12 hr day-night cycle at a constant environmental temperature of 21°C ( humidity: 65% ) . Approval for animal experiments was obtained from the Malmö/Lund ethical committee on animal experiments ( No . : M7-16 ) , and the ethical committee of the Faculty of Medicine of Porto University and Portuguese Direção-Geral de Alimentação e Veterinária ( No . 0421/000/000/2018 ) . The animals were anesthetized with sodium pentobarbital ( I . P . ; 60 mg/kg body weight , as an initial dose and then 6 mg/kg every hour ) . If needed , the anesthesia was reinforced with ≈1 . 5% isoflurane ( IsoFlo , Abbott Laboratories , Norway ) in a mixture of 65% nitrous oxide-35% oxygen . Core temperature of the animals was controlled using a feedback-regulated heating system . In the experiments involving electrophysiological recordings , the rats were ventilated artificially via a tracheal cannula and the expiratory CO2 and mean arterial blood pressure ( 65–140 mmHg ) was monitored continuously in the right carotid artery . The experimental design included rats with either the UBI alone or the UBI which was preceded by a complete spinal cord transection . In the UBI-only experiments , anesthetized rats were placed on a surgery platform with stereotaxic head holder . The rat head was fixed in a position in which the bregma and lambda were located at the same horizontal level . After local injection of lidocaine ( Xylocaine , 3 . 5 mg/ml ) with adrenaline ( 2 . 2 μg/ml ) , the scalp was cut open and a piece of the parietal bone located 0 . 5–4 . 0 mm posterior to the bregma and 1 . 8–3 . 8 mm lateral to the midline ( Paxinos and Watson , 2007 ) was removed . The part of the cerebral cortex located below the opening that includes the hind-limb representation area of the sensorimotor cortex ( HL-SMC ) was aspirated with a metallic pipette ( tip diameter 0 . 5 mm ) connected to an electrical suction machine ( Craft Duo-Vec Suction unit , Rocket Medical Plc , UK ) . Care was taken to avoid damaging the white matter below the cortex . After the ablation , bleeding was stopped with a piece of Spongostone and the bone opening was covered with a piece of TissuDura ( Baxter , Germany ) . For sham operations , animals underwent the same anesthesia and surgical procedures , but the cortex was not ablated . In the experiments in which UBI was preceded by the spinal cord transection , the anaesthetized animals were first placed on a surgery platform and the skin of the back was incised along the midline at the level of the superior thoracic vertebrae . After the back muscles were retracted to the sides , a laminectomy was performed at the T2 and T3 vertebrae . The spinal cord between the two vertebrae then was completely transected using a pair of fine scissors . A piece of Spongostan ( Medispon MDD sp . zo . o . , Toruń , Poland ) was placed between the rostral and caudal stumps of the spinal cord . The completeness of the transection was confirmed by ( i ) inspecting the cord during the operation to ensure that no spared fibers bridged the transection site and that the rostral and caudal stumps of the spinal cord were completely retracted; and ( ii ) examining the spinal cord in all animals after termination of the experiment . To further ensure the completeness of transection , a 3–4 mm spinal cord segment was dissected and removed after laminectomy at the T2-T4 level in a subset of rats . Inspection under the microscope demonstrated that the transection was complete . Following the surgery , the rats were mounted onto the stereotaxic frame and the UBI was performed as described above . After completion of all surgical procedures , the wounds were closed with the 3–0 suture ( AgnTho’s , Sweden ) and the rat was kept under an infrared radiation lamp to maintain body temperature during monitoring of postural asymmetry ( up to 3 hr ) and during EMG recordings . To verify the UBI site the rats were perfused with 4% paraformaldehyde and the brains were removed from the skulls . Following post fixation overnight in the same fixative , the brains were soaked in phosphate-buffered saline for 2 days , dissected into blocks and the blocks containing the lesion area were cut into 50 µm sections using a freezing microtome . Every fourth section was mounted on slides and stained for Nissl with modified Giemsa solution ( Sigma-Aldridge , USA; 1:5 dilution ) . The left drawing in Figure 1A shows the location of the right hindlimb representation area on the rat brain surface ( adapted from Frost et al . , 2013 ) . The hypophysectomy was performed at the Charles River Laboratories ( France ) site and all the surgery-related procedures , including postoperative care and transportation of animals , were performed according to the ethical recommendations of that company . The procedure for transauricular hypophysectomy , performed under isoflurane anesthesia , was described elsewhere ( Koyama , 1962 ) . Briefly , a hypodermic needle fitted to a plastic syringe was introduced into the external acoustic meatus until its tip reached the medial wall of the tympanic cavity . The needle was then pushed slightly further , so that its tip pierced the bone and entered the pituitary capsule . The hypophysis was then sucked into the syringe . The success of hypophysectomy was assessed by visual inspection of the hypophyseal region of the skull under a microscope following animal sacrifice and removal of the brain . Only data obtained in rats in which complete hypophysectomy was confirmed were included in the analysis . Sham-operated rats underwent an identical procedure except that the needle was not introduced into the pituitary capsule . Following the hypophysectomy , the animals were given a 3-week recovery period before initiating the UBI experiments . Experiments were performed between 10:00 and 15:00 hr over the course of the week preceding surgeries ( pre-training ) and one-day post-surgery ( testing ) . Analysis of the HL-PA by the hands-on method was described previously ( Bakalkin and Kobylyansky , 1989; Zhang et al . , 2020 ) . Briefly , the postural asymmetry measurement was performed under pentobarbital ( 60 mg/kg , I . P . ) anesthesia , or isoflurane anesthesia when rats with UBI were analyzed one or more days after the surgery . The level of anesthesia was characterized by a barely perceptible corneal reflex and a lack of overall muscle tone . The rat was placed in the prone position on 1 mm grid paper , and the hip and knee joints were straightened by gently pulling the hindlimbs back for 5–10 mm to reach the same level . Then , the hindlimbs were released and the magnitude of postural asymmetry was measured in millimeters as the length of the projection of the line connecting symmetric hindlimb distal points ( digits 2–4 ) on the longitudinal axis of the rat . The procedure was repeated six times in immediate succession , and the mean HL-PA value for a given rat was calculated and used in statistical analyses . The measurements were performed 0 . 5 , 1 , and 3 hr after the brain injury , or at other time points as shown on figures . In a separate group of rats ( Figure 1—figure supplement 1C , D ) , HL-PA was assessed 1 , 4 , 7 , and 14 days after the UBI or sham surgery under isoflurane anesthesia ( 1 . 5% isoflurane in a mixture of 65% nitrous oxide and 35% oxygen ) . The rat was regarded as asymmetric if the magnitude of HL-PA exceeded the 1 mm threshold ( see statistical section ) . The limb displacing a shorter projection was considered to be flexed . In a subset of the rats with UBI or sham surgery ( n = 11 and 10 , respectively ) , the hindlimbs were stretched by gently pulling two threads glued to the nails of the middle three toes of both legs . In another subset ( n = 6 ) , the skin of the hindlimbs including and distal to the ankle joints was fully anesthetized by a topical application of 5% lidocaine cream 10 min before the assessments of HL-PA in rats with UBI . The absence of the pedal withdrawal reflexes following lidocaine application was confirmed in awake rats by pinching the skin between the toes with blunt forceps . None of these two procedures affected the resulting HL-PA suggesting that HL-PA formation does not dependent on tactile input from the hind paw . For analysis in the supine position , the rat was placed in a V-shaped trough , a 90° - angled frame located on a leveled table surface with the 1 mm grid sheet; otherwise , the procedure was the same as for the prone position . The HL-PA values and the probability to develop asymmetry ( PA ) were essentially the same for both positions . In the hands-off analysis ( Figure 1—figure supplement 2 ) , the anesthetized rat was placed on the bench in prone position . Silk threads were glued to the nails of the middle three toes of both hindlimbs , and their other ends were tied to hooks attached to the movable platform that was operated by a micromanipulator . To reduce potential friction between the hindlimbs and the surface with changes in their position during stretching and after releasing them , the bench under the rat was covered with plastic sheet and the movable platform was raised up to form a 10° angle between the threads and the bench surface . Stretching was initiated at the ‘natural’ hindlimb position that was either symmetric or asymmetric , and performed for the 2 cm distance at a rate of 2 cm/s ( Variant 1 , V1; Figure 1—video 1 , episodes 1 and 2 ) . Alternatively , the limbs were adjusted to an approximately symmetric position by gently pulling the thread on the flexed limb and then stretching it at a rate of 2 cm/sec for 1 . 5 cm ( Variant 2 , V2; Figure 1—video 1 , episode 3 ) . The threads then were relaxed , the limbs were released and the resulting HL-PA was photographed . The procedure was repeated six times in succession , and the mean value of postural asymmetry for a given rat was calculated and used in statistical analyses . Both variants 1 and 2 ( V1 and V2 ) of the hands-off method , and the hands-on method produced virtually the same results; no differences ( p > 0 . 40 ) in the magnitude and its direction were revealed between them ( Figure 1—figure supplement 2 ) . The postural asymmetry analysis was blind to the observer excluding the analysis combined with the EMG . The ‘reverse design’ HL-PA results shown on Figure 1F were replicated by two groups in different laboratories ( Figure 1—figure supplement 4B–E and F–L , respectively ) and by both the hands-on and hands-off methods ( Figure 1—figure supplement 5 ) . The HL-PA was measured in mm with negative and positive HL-PA values that are assigned to rats with the left and right hindlimb flexion , respectively . This measure shows the flexion side and HL-PA value . However , it does not show the proportion of the animals with asymmetry in each group; we could not see whether all or a small fraction of animals display the asymmetry . Furthermore , its interpretation may not be straightforward for groups with the similar number of left or right flexion; in this case , the HL-PA value would be about zero . Data are also presented as the probability of postural asymmetry ( PA ) that shows the proportion of animals exhibiting HL-PA at the imposed threshold ( > 1 mm ) . The PA does not show flexion side and flexion size . These two measures are obviously dependent; however , they are not redundant and for this reason , both are required for data presentation and characterization of the HL-PA . Serum was collected from three animals in each group of rats with transected spinal cord 3 hr after the UBI or sham surgery , pooled , kept at −80°C until use , and administered intravenously ( 0 . 3 mL / rat ) to rats under pentobarbital anesthesia 10 min after complete spinal cord transection . Serum and Arg-vasopressin were administered into the cisterna magna ( intracisternal route; five microliters/rat ) ( Ramos et al . , 2019; Xavier et al . , 2018 ) of intact rats under pentobarbital anesthesia , which was followed by the spinal cord transection 10 min later . HL-PA was analyzed at the 0 . 5 , 1 and 3 hr time points after injection while the animals remained under pentobarbital anesthesia . SSR-149415 ( 5 mg/ml/kg , n=12 ) dissolved in a mixture of DMSO ( 10% ) and saline ( 90% ) , or vehicle alone ( n = 8 ) was administered I . P . 10 min before spinal cord transection . This was followed by either the left-side UBI ( SSR-149415: n = 6; vehicle: n = 5 ) or intravenous administration of serum from UBI rats ( SSR-149415: n = 6; vehicle: n = 3 ) . In rodents , the effects of SSR149415 develop within 0 . 5–1 hr and last for 4–6 hr after administration ( Ramos et al . , 2006; Serradeil-Le Gal et al . , 2002 ) . Naloxone ( 5 mg/ml/kg in saline ) or saline alone was injected I . P . 2 hr after delivering the UBI ( naloxone: n = 6; saline: n = 6 ) or after injecting the UBI serum ( naloxone: n = 6; saline: n = 3 ) to rats with transected spinal cord . Total RNA was purified by using RNasy Plus Mini kit ( Qiagen , Valencia , CA , USA ) . RNA concentrations were measured with Nanodrop ( Nanodrop Technologies , Wilmington , DE , USA ) . RNA ( 1 μg ) was reverse-transcribed to cDNA with the iScript cDNA Synthesis Kit ( Bio-Rad Laboratories , CA , USA ) according to manufacturer's protocol . cDNA samples were aliquoted and stored at –20°C . TagMan assay in 384-well format was applied . cDNAs were mixed with PrimePCR Probe assay and iTaq Universal Probes supermix ( Bio-Rad ) for qPCR with a CFX384 Touch Real-Time PCR Detection System ( Bio-Rad Laboratories , CA , USA ) according to manufacturer’s instructions . TagMan probes used are listed in Key resources table . All procedures were conducted strictly in accordance with the established guidelines for the qRCR based analysis of gene expression , consistent with the minimum information for publication of quantitative real-time PCR experiments guidelines ( MIQE ) ( Bustin et al . , 2009; Taylor et al . , 2019 ) . The raw qRT-qPCR data were obtained by the CFX Maestro Software for CFX384 Touch Real-Time PCR Detection System ( Bio-Rad Laboratories , CA , USA ) . mRNA levels of genes of interest were normalized to the geometric mean of expression levels of two reference genes Actb and Gapdh selected out of 10 genes ( Actb , B2m , Gapdh , Gusb , Hprt , Pgk , Ppia , Rplpo13a , Tbp , and Tfrc ) using the geNorm program [https://genorm . cmgg . be/ and Kononenko et al . , 2017; Vandesompele et al . , 2002] . The expression stability of candidate reference genes was computed for four sets of samples that were the left and right halves of the lumbar spinal cord obtained from the left-sided sham surgery group and the left-sided UBI group and was as follows ( from high to low ) : Actb , Gapdh , Tbp , Rplpo13a , Hprt , Pgk , B2m , Tfrc , Ppia , and Gusb . In each experiment , the internal reference gene-stability measure M did not exceed 0 . 5 at the threshold value imposed by the MIQE equal to 1 . 5 . The number of reference genes was optimized using the pairwise stability measure ( V value ) calculated by the geNorm program . The V value for Actb and Gapdh , the top reference genes was 0 . 12 that did not exceed the 0 . 15 threshold demonstrating that analysis of these two genes is sufficient for normalization . The procedure was described elsewhere ( Christensson-Nylander et al . , 1985; Merg et al . , 2006 ) . Briefly , 1 M hot acetic acid was added to finely powdered frozen tissues , and samples were boiled for 5 min , ultrasonicated , and centrifuged . Tissue extract was run through a SP-Sephadex ion exchange C-25 column , and peptides were eluted and analyzed by RIA . Anti-Dynorphin B antiserum showed 100% molar cross-reactivity with big dynorphin , 0 . 8% molar cross-reactivity with Leu-morphine ( 29 amino acid C-terminally extended Dynorphin B ) , and <0 . 1% molar cross-reactivity with Dynorphin A ( 1–17 ) , Dynorphin A ( 1–8 ) , α-neoendorphin , and Leu-enkephalin ( Yakovleva et al . , 2006 ) . Cross-reactivity of Leu-enkephalin-Arg antiserum with Dynorphin B and Leu- and Met-enkephalin was <0 . 1% molar , with α-neoendorphin 0 . 5% molar , with Dynorphin A ( 1–8 ) 0 . 7% molar , with Met-enkephalin-Arg-Phe 1% molar and with Met-enkephalin-Arg 10% molar . Cross-reactivity of Met-enkephalin-Arg-Phe antiserum with Met-enkephalin , Met-enkephalin-Arg , Met-enkephalin-Arg-Gly-Leu , Leu-enkephalin and Leu-enkephalin-Arg was <0 . 1% molar ( Nylander et al . , 1997 ) . Our variant of RIA readily detected Dynorphin B and Leu-enkephalin-Arg in wild-type mice ( Nguyen et al . , 2005 ) but not in Pdyn knockout mice; thus the assay was highly specific and not sensitive to the presence of contaminants . The peptide content is presented in fmol/mg tissue . Processing and statistical analysis of the HL-PA , EMG and molecular data was performed after completion of the experiments by the statisticians ( DS and VG ) , who were not involved in execution of experiments . Therefore , the results of intermediate statistical analyses could not affect acquisition of experimental data that otherwise might be biased . | Brain trauma or a stroke often lead to severe problems in posture and movement . These injuries frequently occur only on one side , causing asymmetrical motor changes: damage to the left brain hemisphere triggers abnormal contractions of the right limbs , and vice-versa . The injuries can disrupt neural tracts between the brain and the spinal cord , the structure that conveys electric messages to muscles . However , research has also shed light on new actors: the hormones released into the bloodstream by the pituitary gland . Similar to the effects of brain lesions , several of these molecules cause asymmetric posture in healthy rats . In fact , a group of hormones can trigger muscle contraction of the left back leg , and another of the right one . Could pituitary hormones mediate the asymmetric effects of brain injuries ? To investigate this question , Lukoyanov , Watanabe , Carvalho , Kononenko , Sarkisyan et al . focused on rats in which the connection between the brain and the spinal cord segments that control the hindlimbs had been surgically removed . This stopped transmission of electric messages from the brain to muscles in the back legs . Strikingly , lesions on one side of the brain in these animals still led to asymmetric posture , with contraction of the leg on the opposite side of the body . These effects were abolished when the pituitary gland was excised . Postural asymmetry also emerged when blood serum from injured rats was injected into healthy animals . The findings suggest that hormones play an essential role in signalling from the brain to the spinal cord . Further experiments identified that two pituitary hormones , β-endorphin and Arg-vasopressin , induced contraction of the right but not the left hindlimb of healthy animals . In addition , small molecules that inhibit these hormones could block the deficits seen on the right side after an injury on the left hemisphere of the brain . Taken together , these results show that neurons in the spinal cord are not just controlled by the neural tracts that descend from the brain , but also by hormones which have left-right side-specific actions . This unique signalling could be a part of a previously unknown hormonal mechanism that selectively targets either the left or the right side of the body . This knowledge could help to design side-specific treatments for stroke and brain trauma . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2021 | Left-right side-specific endocrine signaling complements neural pathways to mediate acute asymmetric effects of brain injury |
The nuclear pore complex ( NPC ) mediates nucleocytoplasmic transport through the nuclear envelope . How the NPC assembles into this double membrane boundary has remained enigmatic . Here , we captured temporally staged assembly intermediates by correlating live cell imaging with high-resolution electron tomography and super-resolution microscopy . Intermediates were dome-shaped evaginations of the inner nuclear membrane ( INM ) , that grew in diameter and depth until they fused with the flat outer nuclear membrane . Live and super-resolved fluorescence microscopy revealed the molecular maturation of the intermediates , which initially contained the nuclear and cytoplasmic ring component Nup107 , and only later the cytoplasmic filament component Nup358 . EM particle averaging showed that the evagination base was surrounded by an 8-fold rotationally symmetric ring structure from the beginning and that a growing mushroom-shaped density was continuously associated with the deforming membrane . Quantitative structural analysis revealed that interphase NPC assembly proceeds by an asymmetric inside-out extrusion of the INM .
The nuclear pore complex ( NPC ) is the largest non-polymeric protein complex in eukaryotic cells , embedded in a double membrane called the nuclear envelope ( NE ) , and mediates all macromolecular transport across the NE . The NPC has an octameric structure and is composed of multiple copies of over 30 different proteins termed nucleoporins ( Nups ) ( Grossman et al . , 2012; Schwartz , 2013; Strambio-De-Castillia et al . , 2010 ) . In metazoan cells NPCs are assembled in two cell-cycle stages , during nuclear assembly post anaphase and during nuclear growth in interphase . Both assembly pathways have distinct properties and are usually referred to as postmitotic and interphase NPC assembly ( Schooley et al . , 2012; Wandke and Kutay , 2013 ) . In postmitotic assembly , the double nuclear membrane and the NPC channel assemble concomitantly onto chromatin , and postmitotic formation of import competent nuclei with sealed nuclear membranes and functional NPCs is completed very rapidly within 15 min after anaphase onset ( Dultz et al . , 2008; Haraguchi et al . , 2000; Lu et al . , 2011; Otsuka et al . , 2014 ) . By contrast , interphase NPC assembly occurs only after the NE is fully sealed in late anaphase . This second assembly mechanism proceeds throughout telo- and interphase resulting in a doubling of the number of NPCs for the next division . In the context of the closed nucleus , NPCs must be formed by an insertion into the NE that fuses the outer and inner nuclear membranes ( ONM and INM ) . Interphase assembly is much slower compared to postmitotic assembly and shows differences in the molecular requirements and the order of recruited components ( D'Angelo et al . , 2006; Dultz and Ellenberg , 2010; Schellhaus et al . , 2015 ) . Several molecular requirements for interphase assembly have been reported in different model systems . Studies using in vitro assembled nuclei with Xenopus egg extracts have shown the requirement of RanGTP on both sides of the NE ( D'Angelo et al . , 2006 ) and the import of Nup153 to recruit the Nup107-160 complex to the INM ( Vollmer et al . , 2015 ) . In mammalian cells , the membrane curvature-sensing domain of Nup133 ( Doucet et al . , 2010 ) , the INM protein Sun1 ( Talamas and Hetzer , 2011 ) , and the targeting of the transmembrane nucleoporin Pom121 to the INM ( Funakoshi et al . , 2011 ) have been reported to be required . Although some of these studies as well as a study on the evolution of eukaryotic cells ( Baum and Baum , 2014 ) have suggested that interphase NPC assembly may initiate from the nuclear side , how and by what membrane deformation and fusion process NPC assembly takes place has remained enigmatic ( Doucet and Hetzer , 2010; Rothballer and Kutay , 2013 ) . Interestingly , INM deformations have been observed in yeast mutants lacking several nucleoporins , membrane proteins Apq12 and Brr6 , and the AAA-ATPase VPS4 and , while sometimes interpreted as pleiotropic consequences of transport defects , have also been suggested to be involved in nucleoporin quality control or NPC assembly ( Chadrin et al . , 2010; Hodge et al . , 2010; Makio et al . , 2009; Meszaros et al . , 2015; Murphy et al . , 1996; Scarcelli et al . , 2007; Webster et al . , 2014; Wente and Blobel , 1993 ) . However , it has remained unclear how NPC assembly takes place in wildtype cells and what the normal assembly intermediates might look like . Pioneering studies that used in vitro assembled and inhibitor treated nuclei ( Goldberg et al . , 1997 ) could unfortunately not establish the physiological nature of the partial NPC structures since they only examined the cytoplasmic side of the NE and were not able to analyze INM deformations . Despite this significant amount of indirect evidence and several competing hypotheses for interpreting it regarding NPC assembly ( Rothballer and Kutay , 2013 ) , progress in the field has been slow largely due to the experimental challenge of capturing the rare and sporadic interphase NPC assembly events and imaging them at single pore resolution in order to reliably distinguish newly-assembling from already-formed NPCs ( D'Angelo et al . , 2006; Dultz and Ellenberg , 2010 ) . To overcome this challenge and study the mechanism of interphase assembly in whole cells more effectively , we focused on the NPC-poor NE islands present in telophase nuclei that are populated with NPCs during nuclear expansion in the G1 phase of the cell-cycle ( Maeshima et al . , 2006 ) . These islands result from the so called 'core regions' where nuclear membrane sealing is locally delayed in mitosis due to removal of dense spindle microtubules from the DNA surface ( Vietri et al . , 2015 ) and therefore largely devoid of postmitotic NPC assembly , resulting in a low NPC density in the membrane of the core regions ( Dechat et al . , 2004; Haraguchi et al . , 2000 ) . Core regions therefore provide an almost 'virgin' double membrane surface , where de novo interphase NPC assembly is easier to observe . By systematically recording electron tomograms of core regions at different times of nuclear growth , using correlation with live imaging to determine the precise cell-cycle stage of each cell , we were indeed able to reliably capture intermediates of interphase NPC assembly . Three-dimensional ( 3D ) analysis of temporally ordered intermediates revealed that interphase NPC assembly proceeds by an inside-out INM evagination followed by fusion with the flat ONM . Averaging the structure of assembly intermediates at the same stage of membrane deformation showed that an eightfold symmetric nuclear ring underneath the INM already surrounds the base of the earliest detectable evaginations and that a mushroom-shaped density appears to drive the membrane deformation until fusion with the ONM .
Deformation and fusion of the nuclear membranes that must be present during interphase NPC assembly can only be reliably detected by high-resolution 3D electron microscopy ( EM ) . To target such EM observations , we established an assay that allowed us to estimate the position of the core region in the NE of telophase and G1 nuclei at any time during nuclear expansion post anaphase . To this end , we used 3D live confocal time-lapse imaging of the core marker Lap-2α tagged with YFP ( Dechat et al . , 2004 ) together with the chromatin marker histone 2B tagged with mCherry ( Figure 1—figure supplement 1A ) . 3D reconstruction of the core region surface in late anaphase allowed us to calculate the core regions at later times in the cell-cycle by modeling it onto the overall growth of the nuclear surface measured using histone 2B ( Figure 1—figure supplement 1B‒H ) . With this assay in hand , we then systematically imaged live cells after exiting mitosis on EM compatible sapphire disks with carbon coated landmarks ( Figure 1A ) , and natively fixed them by rapid high pressure freezing at defined times during G1 nuclear expansion . After cryo-substitution , we acquired high-resolution electron tomograms from sections cut through the core regions . The single cell correlation with live imaging allowed us to precisely determine the stage of nuclear expansion of each cell sampled by electron tomography and therefore temporally register all our samples ( Figure 1A , Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 19071 . 003Figure 1 . Interphase assembly intermediates of nuclear pore complexes ( NPCs ) . ( A ) Correlative live-cell imaging with electron microscopy ( EM ) . Cell-cycle progression of HeLa cells was monitored by confocal microscopy and the same cell was subjected to electron tomography . Tomograms were collected from different regions of the nuclear envelope ( NE ) . Inferred non-core , inner-core and outer-core regions are indicated in light blue , light green and dark green , respectively . C , cytoplasm; N , nucleoplasm . Scale bar , 1 µm . ( B ) An electron tomographic slice of the NE . An assembly intermediate and a mature pore are indicated by a red arrow and a blue arrowhead , respectively . Insets show enlarged images in which membranes are traced by white dotted lines . ONM , outer nuclear membrane; INM , inner nuclear membrane . Scale bars , 100 nm . ( C , D ) Immuno-EM with mAb414 antibody and 10 nm-gold particles . ( C ) The profile of the NE and the positions of gold particles are denoted in the bottom panel . A mature pore and an intermediate are indicated as in ( B ) . Scale bar , 100 nm . ( D ) The number of gold particles per assembly intermediate ( ‘observed’ ) and the one calculated assuming a random distribution of the particles ( ‘if random’ ) . 31 particles were found on 13 intermediates , whereas the random distribution estimated 1 . 6 particles to be on 13 intermediates . The p-value ( probability that the distribution is due to chance alone ) <10–100; a chi-square goodness of fit test . ( E ) Cryo-EM tomographic slices of isolated NEs of HeLa cells . A mature pore and an intermediate are indicated as in ( B ) . Other examples of intermediates are also indicated . Scale bars , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 00310 . 7554/eLife . 19071 . 004Figure 1—figure supplement 1 . Estimation of core regions . ( A , B ) Time-lapse three-dimensional ( 3D ) imaging of dividing HeLa cells which express Lap-2α-YFP and histone-H2b-mCherry by confocal microscopy . Scale bars , 10 µm . ( A ) Maximum intensity projection images . Time after anaphase onset is indicated . Lap-2α localizes in core regions only at late anaphase . ( B ) A single confocal section of another dividing cell at late anaphase . ( C–G ) A pipeline for estimating core regions after late anaphase . ( C ) The entire surface of the chromosomes was segmented and the areas where Lap-2α localizes are marked in dark and light green . ( D ) The segmented chromosomes were rotated as indicated by arrows in ( C ) . ( E ) The nuclei and core regions were approximated by the ellipsoid . ( F ) The area of core regions was measured in 20 different nuclei . The regions above the 95th , between the 5th and 95th , and below the 5th percentile of being core regions are inferred to be core ( dark and light green ) , indistinct ( gray ) , and non-core regions ( light blue ) , respectively . ( G ) The nucleus and core regions were estimated to grow isometrically after late anaphase . ( H ) Assignment of core regions in the electron microscopy ( EM ) image . Serial sections of plastic-embedded HeLa cells were cut and the Z position was determined for each section . The major axis of the nucleus was estimated and core and non-core regions were assigned based on the criteria defined in ( B–G ) . Scale bars , 5 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 00410 . 7554/eLife . 19071 . 005Figure 1—figure supplement 2 . Live-cell and EM images of cells analyzed by EM tomography . Cells cultured on carbon-patterned sapphire disks were imaged by confocal microscopy ( upper panels ) . After high-pressure freezing , plastic-embedding , and serial sectioning , the same cells were observed in EM . EM images of one of the sections are shown ( lower panels ) . Time after anaphase onset is indicated . The contrast of some of the live-cell images are enhanced by average intensity projection of a series of several images for presentation purposes . The sizes of live-cell and EM images are 58 × 58 µm and 20 × 20 µm , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 00510 . 7554/eLife . 19071 . 006Figure 1—figure supplement 3 . Galleries of interphase NPC assembly intermediates . ( A ) Other images of immuno-EM in Figure 1C . Assembly intermediates are indicated by white arrows . The profile of the nuclear envelope ( NE ) and the positions of gold particles in one EM image are denoted in a box . C , cytoplasm; N , nucleoplasm; ONM , outer nuclear membrane; INM , inner nuclear membrane . Scale bars , 100 nm . ( B ) Electron tomographic slices of mature pores and assembly intermediates of U2OS ( upper panels ) and NRK ( lower panels ) cells . Assembly intermediates are indicated by white arrows . The intermediates were found in both G1 staged and fully grown nuclei . Scale bars , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 006 In the resulting 158 tomograms , we consistently found approximately 50 nm evaginations of the INM ( Figure 1B , Table 1 , and Video 1 ) filled with electron dense material , that were clearly distinct from the ~200 nm nuclear egress structures transporting ribonucleoproteins and viruses reported recently ( Mettenleiter et al . , 2013; Speese et al . , 2012 ) . Immuno-EM showed that the evaginations were specifically enriched with at least one of the nucleoporins recognized by mAb414 ( Nup62 , Nup153 , Nup214 , and Nup358 ) ( Figure 1C , D , and Figure 1—figure supplement 3A ) , suggesting that they are pore assembly intermediates . Similar INM evaginations filled with electron dense material were also found in cryo-electron tomograms of vitrified isolated NEs ( Figure 1E ) , ruling out that they are artifacts of dehydration , heavy metal staining or resin-embedding during cryo-substitution , and demonstrating that they are stable membrane structures that persist even after in vitro isolation of the NE ( Bui et al . , 2013; Ori et al . , 2013 ) . Indistinguishable evaginations of the INM were also observed in U2OS ( human bone osteosarcoma epithelial ) and NRK ( normal rat kidney ) cells ( Figure 1—figure supplement 3B ) , ruling out that their occurrence is cell type , cancer or species specific . 10 . 7554/eLife . 19071 . 007Table 1 . Summary of EM tomography . A data table shows the surface area of the NE analyzed by EM tomography and the number of mature pores , assembly intermediates , and the outer and inner nuclear membrane ( ONM and INM ) fusion events found in each cell at a different time point after anaphase onset . The data obtained in non-core , inner- and outer-core regions are indicated separately . In total , 154 µm2 NE surface area was analyzed , and 279 intermediates and 1322 mature pores were found . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 007Time after anaphase onset ( min ) 19 . 224 . 428 . 436 . 342 . 053 . 261 . 065 . 673 . 682 . 9100116>180>180Non-coreAnalyzed surface area ( µm2 ) 5 . 294 . 045 . 063 . 995 . 175 . 544 . 043 . 723 . 063 . 733 . 323 . 174 . 032 . 80Number of mature pores8245484262515333333933314531Number of intermediates24635833215541Number of ONM/INM fusionInner-coreAnalyzed surface area ( µm2 ) 5 . 775 . 064 . 164 . 746 . 754 . 333 . 614 . 574 . 073 . 534 . 613 . 46Number of mature pores24163183016342727395332Number of intermediates21131283512134323Number of ONM/INM fusion21Outer-coreAnalyzed surface area ( µm2 ) 4 . 224 . 153 . 832 . 554 . 784 . 413 . 593 . 682 . 712 . 672 . 772 . 68Number of mature pores294027174435382830342528Number of intermediates23101991311141351Number of ONM/INM fusion210 . 7554/eLife . 19071 . 008Video 1 . EM tomographic slices of the nuclear envelope of a cell at 53 min post anaphase . One of the mature pores and an assembly intermediate are indicated by blue and red arrows , respectively . Scale bar , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 008 To test if the evaginations displayed a progression of structural changes consistent with the formation of mature NPCs , we analyzed their membrane shape in cells captured at different time points after the completion of postmitotic nuclear assembly . Quantitative analysis of 135 INM evagination membrane profiles from a time course of cells captured at 19 , 28 , and 53 min post anaphase revealed that evaginations progressively grow inside-out ( Figure 2 ) . Evagination depth increased significantly from 16 to 22 nm within 9 min ( 19 to 28 min post anaphase; Figure 2C ) and evagination diameter continuously and significantly increased from 51 to 58 nm within 34 min ( 19 to 53 min post anaphase; Figure 2D ) . Among the 279 total evaginations we found in 154 µm2 NE surface area ( Table 1 ) , we capture only five ONM/INM fusion events ( Figure 2 ) , where the INM evagination had reached the flat ONM surface . These fusion intermediates had an average evagination depth of 28 nm , similar to the ONM/INM distance and an average diameter of 61 nm , intermediate between late evaginations and mature nuclear pores ( Figure 2C , D ) , as expected for NPC assembly . The low number of fusion intermediates indicates that the fusion step must be very short-lived . 10 . 7554/eLife . 19071 . 009Figure 2 . Quantitative structural comparison of assembly intermediates . ( A ) Electron tomographic slices of assembly intermediates in cells captured at 53 , 28 , and 19 min after anaphase onset ( AO ) and ONM/INM fusion events . Profiles of ONM ( gray ) and INM ( blue ) in black and white boxes on EM images are depicted in the right panels . For the fusion , ONM is also depicted in blue . The image marked with a white asterisk was acquired on a differently embedded sample for enhancing membrane contrast ( see Materials and methods ) . Scale bar , 100 nm . ( B ) Membrane profiles of all the fusion events and intermediates at selected time points ( 53 , 28 , and 19 min ) . The bold lines indicate the averaged profiles . ( C–F ) Quantification of the evagination depth of INM ( C , E ) and the diameter of intermediates ( D , F ) as indicated by red bidirectional arrows in the left panels . ( C , D ) The plots are from 47 , 44 , and 39 intermediates at 19 , 28 , and 53 min , respectively , 5 ONM/INM fusions , and 45 mature pores . The ONM/INM distance was quantified near mature pores ( C ) . The median is depicted as a horizontal line and the whiskers show the 25th and 75th percentiles . *p<0 . 02 , **p<0 . 001; unpaired t-tests . ( E , F ) The depth and the diameter of intermediates in non-core , inner-core , and outer-core regions were indicated in light blue , light green and dark green , respectively . The median is depicted as a horizontal line . No statistical difference of the intermediate shape was observed between different regions of the NE at each time point ( p>0 . 1; unpaired t-tests ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 00910 . 7554/eLife . 19071 . 010Figure 2—source data 1 . Depth and diameter values used for Figure 2C–F . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 010 If the assembly intermediates we observed mature into fully assembled NPCs , their abundance should quantitatively explain the increased number of mature pores observed after nuclear expansion . To address this , we quantified the changes in density of intermediates and mature pores over time in EM tomograms of a time course of 12 cells correlatively fixed from 19 to 120 min post anaphase ( Figure 3 and Figure 1—figure supplement 2 , and Table 1 ) . This data showed that assembly intermediates are most abundant in core regions during the first hour , when this sealed membrane area still has a low density of mature pores due to the lack of postmitotic assembly ( Figure 3B ) . By contrast and as expected , non-core regions already exhibited a high density of mature pores that arose from postmitotic NPC assembly ( Figure 3B ) . Later than one hour post anaphase , the ratio of assembly intermediates to mature pores in the core regions had equilibrated to a similar level as found in non-core regions or fully grown nuclei ( Figure 3A , B ) , indicating that a period of frequent interphase pore assembly events during the first hour of G1 nuclear expansion populates the core regions of the NE with NPCs until the steady state interphase density is reached . 10 . 7554/eLife . 19071 . 011Figure 3 . Abundance of mature pores and assembly intermediates at different cell-cycle stages . ( A ) Measurement of nuclear pore density . Gray sheets are the NEs segmented from EM tomograms . Blue and red dots indicate the positions of mature pores and intermediates , respectively . Inner-core regions of cells at 19 and 100 min post anaphase are shown as examples . ( B ) Density of mature pores ( dark blue ) and intermediates ( red ) in non-core and core ( inner- plus outer-core ) regions of cells at different times . Error bars represent the s . d . from 6 , 6 , and 2 cells at <1 , 1–2 , and >3 hr post anaphase , respectively . **p<0 . 001; unpaired t-tests of the density difference of mature pores ( blue ) and intermediates ( red ) between core regions at <1 hr and the others . The modeled density of mature pores ( light blue ) and intermediates ( light red ) at >3 hr is also indicated ( see Figure 3—figure supplement 1 and Materials and methods for details ) . ( C , D ) Density of mature pores and intermediates in inner- ( C ) and outer-core ( D ) regions of cells at different time points . 3–7 tomograms were obtained in each region at each time point ( data are summarized in Table 1 ) . Dashed lines indicate the modeled density of mature pores and intermediates . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01110 . 7554/eLife . 19071 . 012Figure 3—source data 1 . Density values used for Figure 3B–D . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01210 . 7554/eLife . 19071 . 013Figure 3—figure supplement 2—source data 1 . Surface area values used for Figure 3— figure supplement 2C . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01310 . 7554/eLife . 19071 . 014Figure 3—figure supplement 1 . Modeling the density of nuclear pores . ( A ) Reaction scheme of the delay equation model ( Equations 1‒3 ) . V , production rate; KM , maturation initiation rate constant; τM , maturation delay; Kd , degradation rate constant . The details are described in Materials and methods . ( B ) Variants for NPC initiation ( Equation 7 ) . All mature NPCs coming from the interphase assembly process are assumed to be initiated 10 min after anaphase onset when the NE is sealed ( Dultz et al . , 2008; Otsuka et al . , 2014 ) . Variant 1: Pores are continuously initiated with a constant rate; Variant 2: NPCs are initiated with a time dependent rate that decreases with time to a basal rate; Variant 3: The majority of NPCs intermediates are initiated at 10 min after anaphase onset . ( C ) The discrepancy of simulations and intermediate/mature pore data which is given as the sum of squared residuals ( upper panel , Equation 8 ) . A log-likelihood ratio test indicates that models where the majority of NPC are initiated within 25 min after anaphase onset ( Variant 2 with Kv = 0 . 0641/min or Variant 3 ) significantly ( ***p<0 . 001 ) better fit the data than the model with a constant intermediate production ( Variant 1 ) . Variant 3 showed the best quality of the fit . Lower panel gives the maturation time TM ( Equation 5 ) in the different variants . ( D , E ) Simulations ( Variant 1‒3 , dashed lines ) and data for intermediate ( D ) and mature pore ( E ) densities in the inner core region . ( F , G ) Nuclear pore maturation modeled as a multi-step process ( Equations 9‒11 ) . ( F ) Reaction scheme of the multi-step model ( upper panel ) . Simulated density for mature pores ( blue lines ) and assembly intermediates ( red lines ) in the inner-core region is shown ( lower panel ) . Simulations are shown for kMj=kM as no significant improvement in fits was observed for non-equal rate constant . ( G ) Sum of squared residuals ( upper panel ) and modeled maturation time as function of the number of steps ( lower panel , computed from Equation 12 ) . Although the quality of the fit increases with the number of steps , the maturation time is consistently around 40 min . ( H , I ) Modeling with various maturation time ( Equations 13‒15 ) . ( H ) Different distributions of maturation times were allowed in the modeling ( upper panel ) . The modeled density for mature pores ( blue lines ) and assembly intermediates ( red lines ) in the inner-core region is shown in the lower panel . ( I ) Sum of squared residuals ( upper panel ) and average maturation time as function of the width w of the maturation time distribution ( lower panel ) . The data were reproduced for rather narrow maturation time distributions ( 43 ± 5 min ) as also indicated by the confidence interval for TM and Variant 3 ( see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01410 . 7554/eLife . 19071 . 015Figure 3—figure supplement 2 . Nuclear surface area measurement for the modeling . ( A ) Time-lapse 3D imaging of dividing HeLa cells which express histone-H2b-mCherry ( indicated in red ) . Images show single confocal sections ( upper ) and the segmented nuclei ( lower ) at indicated time points . Scale bars , 10 µm . ( B ) Quantification of nuclear surface area . Black and gray lines represent the average and s . d . of measurements from 9 cells , respectively . All cells undergo a second mitosis after 17–22 hr post anaphase . The data were cut after 17 hr and fitted with Equation 6 in Materials and methods ( red line ) . As indicated by an asterisk , G1 nuclear expansion during the first hour is non-linear . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 015 To test if the high abundance of assembly intermediates in core regions quantitatively explains the number of mature pores found in the same region at later times , we formulated a simple mathematical model for nuclear pore assembly . In this model , assembly intermediates are produced , enter a maturation phase , become mature pores after a typical maturation time , and are ultimately degraded ( Figure 3—figure supplement 1A and Materials and methods ) . NPC production and degradation rates were estimated from the measured steady state density of 11 NPCs/µm2 ( Figure 3B ) and a reported NPC lifetime of ~40 hr in cells with a similar cell-cycle duration ( Rabut et al . , 2004; Schwanhausser et al . , 2011 ) and are in line with the rare and rapid pore disassembly events that have been observed in mammalian cells ( Dultz and Ellenberg , 2010 ) . We modeled different scenarios for the appearance of intermediates ( Figure 3—figure supplement 1B , C ) . Provided that intermediates start to be initiated shortly after anaphase onset the model fits the experimental data of the core regions from 19‒120 min post anaphase well ( Figure 3C , D , and Figure 3—figure supplement 1D , E , Variant 2 , 3 ) , confirming that the abundance of assembly intermediates we observed at the beginning of nuclear growth quantitatively explains the number of mature NPCs observed one hour later . For the best model , where the majority of intermediates are initiated 10 starting minutes after anaphase onset ( Figure 3—figure supplement 1B , C , Variant 3 ) we can estimate the typical maturation time for interphase assembly to be 44 min ( 95% confidence interval [41‒50] ) . Similar average maturation time was obtained for alternative models where maturation steps are explicitly included ( Figure 3—figure supplement 1F , G ) or where the maturation time has a broader distribution ( Figure 3—figure supplement 1H , I ) , demonstrating the robustness of our results . The obtained maturation time is in good agreement with previous reports based on fluorescence microscopy ( ~25 min , D'Angelo et al . , 2006; ~60 min , Dultz and Ellenberg , 2010 ) . It is important to note that inside-out evaginations were present at lower density in non-core regions ( Figure 3B 'Non-core' ) and that we found no significant difference in the increase in depth or diameter of evaginations between non-core and core regions during G1 expansion ( Figure 2E , F ) . This rules out that inside-out assembly intermediates are specific to the core regions , or that core regions are delayed in their maturation . In addition , identical evaginating structures were also found at low density in fully grown nuclei sampled at later cell-cycle stages ( Figure 3B 'Mature NE' ) indicating that the inside-out assembly mechanism is not specific to G1 but occurs throughout interphase . Assuming the maturation time of 44 min , our model shows that the steady state abundance of assembly intermediates we observed in non-core regions and interphase nuclei would be sufficient to maintain the constant NPC density during nuclear growth in interphase that we observed ( Figure 3B “model” , and Figure 3—figure supplement 1 , 2 ) consistent with previous reports ( Dultz and Ellenberg , 2010; Maeshima et al . , 2010 ) . Taken together these results suggest that the assembly intermediates are present across the NE surface , and that the kinetics of NPC assembly are similar across the nuclear surface and throughout interphase . Most importantly , the abundance of intermediates can quantitatively explain both the increase in mature NPCs in the core regions during the rapid nuclear expansion in G1 as well as the homeostatic NPC assembly during nuclear growth later during interphase . The high abundance of intermediates in the core region during the first hour of G1 ( Figure 3C , D ) and their progressive increase in depth and diameter during this time ( Figure 2 ) indicates that the assembly process is relatively synchronous while the NPC-poor core region is populated to the same density as the rest of the nuclear surface ( Figure 3B ) . We should therefore be able to observe the maturation of assembly intermediates directly , by imaging nucleoporin accumulation in the NPC-poor core region in real time . To test this , we performed fast 3D live confocal time-lapse imaging and monitored the concentration of GFP-tagged nucleoporins in the core region during the first hour of G1 , the same time window we observed by correlative EM ( Figure 4A ) . Since the intermediates in inner- and outer-core regions grow in a similar manner ( Figure 2E , F ) and show similar abundance ( Figure 3C , D ) , we measured only the inner-core region where the proportion of intermediates to mature pores is much higher than in the outer-core region . 10 . 7554/eLife . 19071 . 016Figure 4 . Live imaging of nuclear pore assembly in core regions . ( A ) Time-lapse three-dimensional ( 3D ) imaging of GFP-Nup107 and GFP-Nup358 genome-edited cells . DNA was stained with silicon–rhodamine ( SiR ) Hoechst ( Lukinavicius et al . , 2015 ) . Single confocal sections of SiR and GFP channels are shown in the top and bottom panels , respectively . Segmented chromosomes ( light blue ) and inferred inner-core regions ( green ) are shown in the middle panels . Time after anaphase onset is indicated . Scale bars , 20 µm . ( B , C ) Quantification of Nup107 ( left ) and Nup358 ( right ) assembly in inner-core ( B ) and non-core ( C ) regions . The population of postmitotic and interphase NPC assembly measured in Figure 3A‒C is indicated in the left panels . Total intensities of Nup107 ( left ) and Nup358 ( right ) were quantified , normalized , and fitted with a sequential model of NPC assembly that allows for different rate constants for postmitotic and interphase assembly , respectively ( Equations 16‒18 in Materials and methods ) . Dots and shaded areas represent the average and s . d . of measurements from 30 cells for Nup107 and 25 cells for Nup358 , respectively . Black dashed and solid lines indicate the postmitotic and interphase assembly kinetics and gray solid lines show the combined kinetics . ( D ) Normalized densities of interphase Nup107 ( brown ) and Nup358 ( orange ) assembly . The density was measured by dividing the intensity obtained in ( B ) by the nuclear surface area . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01610 . 7554/eLife . 19071 . 017Figure 4—source data 1 . Intensity values used for Figure 4B , C . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01710 . 7554/eLife . 19071 . 018Figure 4—figure supplement 1 . Characterization of genome-edited cell lines expressing GFP-Nup107 and GFP-Nup358 . ( A ) Junction PCR of GFP-Nup107 ( left ) and GFP-Nup358 ( right ) cells with forward primers annealing the upstream of Nups and reverse primers annealing inside of Nups as indicated by black arrows in the bottom panels . Nup107 is homozygously tagged with GFP , whereas Nup358 is heterozygously tagged with GFP . ( B ) Immunoblot analysis of GFP-Nup107 ( left ) and GFP-Nup358 ( right ) cells using antibodies against Nup107 , Nup358 , GFP , and tubulin . An asterisk indicates unspecific bands . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 018 Since our EM analysis suggested an inside-out mechanism , we selected the nuclear and cytoplasmic ring component Nup107 ( Belgareh et al . , 2001 ) and the cytoplasmic filament component Nup358 ( Wu et al . , 1995; Yokoyama et al . , 1995 ) as candidate nucleoporins for genomic GFP-tagging ( Figure 4—figure supplement 1 ) and live imaging . As predicted , the accumulation kinetics differed substantially between core and non-core regions of the NE for both proteins ( Figure 4B , C ) . Since the non-core regions and inner-core regions contained 8% and 50% assembly intermediates in early G1 respectively ( Figure 3B , C ) , we used them to determine the postmitotic and interphase rates of accumulation that explain the different accumulation kinetics of the core region for both Nups resulting from the combined rates ( Figure 4B , C , and Materials and methods ) . Kinetic comparison of interphase accumulation of both Nups in the core region clearly revealed that Nup107 , a component of the nuclear and cytoplasmic ring , is recruited very early ( t1/2 = 15 min , see Materials and methods for details ) , while Nup358 , a component of the cytoplasmic filaments , is recruited only after a significant lag phase ( t1/2 = 51 min ) ( Figure 4D ) . The kinetically distinct and continuously increasing accumulation of two components of the NPC in the core region strongly support a maturation process of assembly intermediates into full pores . In addition , the late recruitment of the cytoplasmic Nup358 is consistent with an inside-out assembly mechanism on the molecular level . The kinetic analysis of bulk Nup accumulation across the inner-core region predicts that single NPC intermediates in early G1 cells should contain Nup107 but not Nup358 . To resolve single intermediates with bi-molecular labeling , we used live imaging to stage cells in G1 and then correlatively performed two-color super-resolution imaging using stimulated emission depletion ( STED ) microscopy with specific antibodies to detect Nup107 and Nup358 ( Materials and methods ) . This analysis indeed revealed many pore-sized discrete localizations of Nup107 in optical sections of the inner-core region NE in early G1 cells , which did not have significant Nup358 labeling ( Figure 5A , B '24 min , inner-core' ) , while non-core regions in the same nucleus contained only double labeled localizations with Nup358 appearing on top of the Nup107 labeling on the outside of the NE ( Figure 5A , B '24 min , non-core' ) , indicative of mature pores . After G1 expansion , also the inner-core region had almost only double-labeled structures ( Figure 5A , B '108 min' ) , consistent with the maturation of intermediates into mature pores . Quantification of the signal in segments along the NE profile allowed us to estimate the frequency of intermediates by the ratio of Nup107/Nup358 ( Figure 5C ) , showing that they are specific to the core region and occur only transiently during the first hour of G1 . These results are fully consistent with the EM observations that interphase NPC assembly intermediates populate the core region of NEs with an abundance that matches the number of mature pores found in this region an hour later after nuclear expansion ( Figure 3A , B ) , and suggest that cytoplasmic nucleoporins such as Nup358 are only recruited at the end of the maturation process , presumably after the growing INM evagination has fused with the ONM . 10 . 7554/eLife . 19071 . 019Figure 5 . Stimulated emission depletion ( STED ) imaging of assembly intermediates . GFP-Nup107 genome-edited cells were stained with anti-GFP and anti-Nup358 antibodies . ( A ) STED images of cells at 24 and 108 min after anaphase onset . Scale bar , 10 µm . ( B ) Flattened and enlarged images of the inferred non-core and inner-core regions indicated by white lines and arrows in ( A ) . The intensity ratios of Nup107 to Nup358 were quantified in every 300 nm segments along the NE and are shown in cyan-black-pink heat maps in the bottom panels . Scale bars , 1 µm . ( C ) The frequency of the segments with the Nup107/Nup358 ratio of >2 . 0 in non-core and inner-core regions at different times . The data are from 14 cells at <1 hr , 6 cells at 1–2 hr , and 4 cells at >3 hr after anaphase onset . Error bars represent the s . d . . **p<0 . 001; unpaired t-tests of the frequency difference between the inner-core region at < 1 hr and the others . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 01910 . 7554/eLife . 19071 . 020Figure 5—source data 1 . Frequency values used for Figure 5C . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 020 Nup107 is a component of the eight-fold rotationally symmetric cytoplasmic and nuclear rings of the NPC . Its early presence in assembly intermediates and their inside-out nature suggested that the nucleoplasmic ring might be one of the first structural elements to form during NPC assembly . To test this , we performed particle averaging of the electron densities of single INM evaginations isolated from tomograms , staged by time during G1 and picked by similarity of their membrane profile depth and diameter . Averaging of 11–36 intermediates revealed a ring structure composed of eight regularly spaced subunits underneath the INM ( Figure 6B and Figure 6—figure supplement 1 ) , which was strikingly similar to the nuclear ring of the mature NPC ( compare top views ( ii ) of the mature pore and intermediates in Figure 6B ) ( Bui et al . , 2013; Maimon et al . , 2012 ) . These eight-fold symmetric rings were already present in the shallowest evaginations we could detect during early G1 and could also be seen in individual tomograms ( Figure 6A , B , and Figure 6—figure supplement 1 ) . Interestingly , side views of the averaged particles revealed a progressive growth also of the mushroom-shaped protein density , whose cap closely matched the growing membrane evagination in depth and diameter and whose stalk was located centrally inside the nuclear ring and grew in length as the cap moved towards the outer membrane ( Figure 6B ) . The structure revealed by particle averaging of assembly intermediates is thus consistent with the observation that Nup107 ( a component of the nuclear ring ) assembles at an early stage and reveals a very interesting mushroom shaped density that might be the driving force of the membrane evagination . 10 . 7554/eLife . 19071 . 021Figure 6 . 3D structural comparison of assembly intermediates . ( A , B ) Electron tomographic slices of single ( A ) and averaged ( B ) mature pores and intermediates at selected time points ( 53 , 28 , and 19 min ) . The averaged images are from 36 mature pores and 14 , 11 , and 24 intermediates picked by similarity of their membrane profile depth and diameter at 53 , 28 , and 19 min , respectively . Red arrowheads i and ii on side-view images indicate the locations of the planes which are inclined at 90° in top views i and ii . The 8-fold symmetric rings observed in top views i and ii of the averaged mature pore are the spoke ring and the nuclear ring complexes , respectively ( Bui et al . , 2013; Maimon et al . , 2012 ) . Scale bars , 100 nm . ( C ) A schematic model of interphase nuclear pore assembly . The assembly intermediate is comprised of the nuclear ring and the central mushroom-shaped density . The assembly of the mushroom drives the INM deformation and it grows progressively inside-out . Once the ONM and INM fuse , it undergoes rapid and drastic structural rearrangements and finally becomes a mature pore . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 02110 . 7554/eLife . 19071 . 022Figure 6—figure supplement 1 . Stability of the subtomogram averaging . ( A ) Illustration of the intensity profile analysis . Images of the averaged assembly intermediate at 53 min post anaphase are shown as an example . Red arrowheads and a dotted line in the side view image indicate the location of the plane which is inclined at 90° in the top view . The intensity along the outer ring ( light blue dashed line ) in the top view image was measured and plotted . The peaks in the plot and the corresponding densities in the top view image are indicated by number . ( B , C ) Electron tomographic slices of single ( B ) and averaged ( C ) intermediates at selected time points ( 53 , 28 , and 19 min ) . Red arrowheads indicate the locations of the planes which are inclined at 90° in the top views . ( B ) Three examples of intermediates are shown at each time . ( C ) Different symmetry was imposed for the alignment of the particles for subtomogram averaging . The intensity along the outer ring was measured and plotted as described in ( A ) . As shown in the plot , the alignment with 8-fold symmetry gives highest intensity difference at regular intervals . Although not all the structures showed the clear 8-fold symmetry due to the noise of the images , this systematic symmetry analysis showed that the structures contain 8-fold symmetry . Scale bars , 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 19071 . 022 Taken together , we can conclude that the observed INM evaginations represent partially assembled NPCs , which contain minimally the nuclear ring with Nup107 and at least one of the central or nuclear O-glycosylated FG-repeat nucleoporins labeled by mAb414 , Nup62 and/or Nup153 , do not contain Nup358 , and are unlikely to contain the cytoplasmic filament base protein Nup214 due to their inside out nature and the lack of Nup358 . Our quantitative structural analysis of the membrane profiles and protein densities of a temporally ordered series of NPC assembly intermediates allowed us to reveal a novel mechanism for NPC biogenesis in intact nuclei of interphase cells by an inside-out extrusion of the NE ( Figure 6C ) . The first clearly detectable NPC intermediate is a shallow INM evagination surrounded at its base by an 8-fold rotationally symmetric nuclear ring complex , in whose center a dome-shaped density with a short stalk is embedded into the INM evagination . Subsequently , this shallow dome matures into a curved mushroom cap , always in direct contact with the growing evagination of the INM and supported by an elongating stalk on the nucleoplasmic side located in the center of the nuclear ring . We speculate that the mushroom-shaped density may use the membrane-attached nuclear ring to determine the site of NPC formation . It is further tempting to speculate that the mushroom-shaped density , potentially through connections to the nuclear ring , might generate the mechanical force needed for INM deformation and eventual fusion with the ONM . Interestingly , the mushroom-shaped structure is clearly distinct from the scaffold architecture of the mature NPC , indicating that interphase assembly cannot be explained by a simple collection of NPC subcomplexes over time but likely involves major structural rearrangements . Given that it had so far been unclear how interphase NPC assembly occurs , this inside-out extrusion mechanism , demonstrated in situ in human cells , provides a new framework to interpret existing genetic ( yeast ) and biochemical ( Xenopus ) data and to investigate the detailed molecular mechanism regulating the assembly process in the future .
Wildtype HeLa kyoto cell line was from Prof . Narumiya in Kyoto University ( RRID: CVCL_1922 ) , and the genome was sequenced previously ( Landry et al . , 2013 ) . Wildtype NRK ( RRID:CVCL_3758 ) and U2OS ( RRID:CVCL_0042 ) cell lines were purchased from ATCC ( Wesel , Germany ) . HeLa and NRK cells were grown in Dulbecco’s Modified Eagle’s Medium ( DMEM ) ( Sigma Aldrich , St . Louis , MO ) supplemented with 10% fetal calf serum ( FCS ) , 2 mM glutamine , 1 mM sodium pyruvate , and 100 µg/ml penicillin and streptomycin . U2OS cells were grown in McCoy's 5A Medium ( Sigma Aldrich ) supplemented with 10% fetal calf serum ( FCS ) , 1X non-essential amino acids solution ( Gibco , Waltham , MA ) , 5 mM glutamine , 1 mM sodium pyruvate , and 100 µg/ml penicillin and streptomycin . A plasmid carrying Lap-2α fused with YFP ( Dechat et al . , 2004 ) was introduced into HeLa cells with the transfection reagent , Fugene6 ( Promega , Madison , WI ) , according to the manufacturer’s protocol . A HeLa cell line stably expressing histone H2b-mCherry ( Neumann et al . , 2010 ) was maintained at 500 ng/ml puromycin ( Invitrogen , Carlsbad , CA ) . The mycoplasma contamination was tested by PCR every 2 or 3 months and was always negative . Cells were cultured on 2-well Lab-Tek Chambered Coverglass ( Thermo Fisher Scientific , Waltham , MA ) for live-cell imaging . For correlative light–electron microscopy , cells were grown on sapphire disks ( 0 . 05 mm thick , 3 mm diameter; Wohlwend GmbH , Sennwald , Switzerland ) , which had been carbon-coated in order to relocate cells on electron microscopy ( EM ) grids , and synchronized by double thymidine arrest ( Harper , 2005 ) . At least 30 min before imaging , the medium was replaced by imaging medium ( IM; CO2-independent medium without phenol red ( Invitrogen ) containing 20% FCS , 2 mM l-glutamine , and 100 µg/ml penicillin and streptomycin ) . Imaging was performed at 37°C in a microscope-body-enclosing incubator . Cells on carbon-coated sapphire disks were observed by confocal microscopy ( LSM510Meta or LSM780; Carl Zeiss , Oberkochen , Germany ) using 10 × 0 . 3 NA Plan-Neofluar or 20 × 0 . 8 NA Plan-Apochromat objective ( Carl Zeiss ) . The cell division process was monitored every 24 s by time-lapse imaging . For three-dimensional ( 3D ) time-lapse imaging , cells were observed by confocal microscopy ( LSM780 ) using 63 × 1 . 4 NA Plan-Apochromat objective ( Carl Zeiss ) . For Figure 1—figure supplement 1 , fluorescent chromatin and Lap-2α were recorded under the following conditions: 25 optical sections , section thickness of 2 . 0 µm , z-stacks of every 1 . 0 µm , the xy resolution of 0 . 13 µm , and a time-lapse interval of 30 s . For Figure 3—figure supplement 2 , fluorescent chromatin was monitored under the following conditions: 40 optical sections , section thickness of 1 . 4 µm , z-stacks of every 0 . 7 µm , the xy resolution of 0 . 13 µm , and a time-lapse interval of 10 min . For Figure 4 , DNA was stained with 0 . 2 µM silicon–rhodamine Hoechst ( Lukinavicius et al . , 2015 ) , and the nucleus and nucleoporins were monitored under the following conditions: 25 optical sections , section thickness of 2 . 5 µm , z-stacks of every 1 . 25 µm , the xy resolution of 0 . 25 µm , and a time-lapse interval of 1 min . Fluorescence images were filtered with a median filter ( kernel size: 0 . 25 × 0 . 25 µm ) for presentation purposes . A 3D computational pipeline was developed in MATLAB ( The MathWorks Inc . , Natick , MA ) that segments chromosomes and core regions from H2B-mCherry and Lap-2α-YFP channels , respectively and extracts different parameters . Original stacks were interpolated along z axis to obtain isotropic resolution and facilitate true 3D image analysis . A 3D Gaussian filter was applied to reduce the effects of high frequency noise . To detect chromosome regions , H2B-mCherry and SiR-Hoechst channels were binarized first using a multi-level thresholding method as described in Heriche et al . ( 2014 ) . Then , chromosome region of interest typically in metaphase in the first time point of the sequence was detected by analyzing the volume and location information of the connected components in the binary image . The detected chromosome was then tracked over the subsequent time points of the sequence and both of the daughter chromosomes were tracked after the division . The surface area of the chromosome was computed applying the method described in Legland et al . ( 2007 ) . For Lap-2α-YFP channel , a reference threshold was estimated by analyzing the intensity over time . This reference threshold was then adapted with a second threshold obtained from individual time points in order to segment the protein . The portion of the nuclear surface where Lap-2α localizes was marked to estimate the surface area of the core regions . Inner- and outer-core regions within nuclei were determined by dividing each nucleus with a cutting plane . The cutting plane was constructed from two vectors where the first one was directed towards the maximum elongation of nucleus and the second one was orthogonal to the first vector and was directed towards the upward z direction . These axes were determined by Eigen vector analysis on the pixel coordinates of the detected nucleus . For the measurement of the Nup intensity on the NE , segmented nuclear volume was dilated and eroded in 3D to define a nuclear membrane rim with 0 . 75 µm width . The areas of inner- and non-core regions were adjusted in individual time points based on the total surface area of the nuclei . Visualization of the chromosome surface in 3D was done in the Amira software package ( Pruggnaller et al . , 2008 ) . Cells at different cell-cycle stages were instantly frozen using a high-pressure freezing machine ( HPM 010; ABRA Fluid AG , Widnau , Switzerland ) . Just before freezing , cells were immersed in IM containing 20% Ficoll ( PM400; Sigma Aldrich ) for protecting cells from freezing damage . Freeze substitution into Lowicryl HM20 resin ( Polysciences Inc . , Warrington , PA ) was performed as described in a previous report ( Kukulski et al . , 2011 ) , with the following modifications: Frozen cells were incubated with 0 . 1% uranyl acetate ( UA ) in acetone at -90°C for 20–24 hr and , after infiltration into Lowicryl resin and UV-polymerization , samples were further polymerized by sunlight for 3–4 days . The cells were also embedded in EPON resin ( Serva , Heidelberg , Germany ) for enhancing membrane contrast as follows: Frozen cells were incubated with 1 . 0% osmium tetroxide ( OsO4 ) , 0 . 1% UA , and 5% water in acetone at −90°C for 20–24 hr . The temperature was raised to −30°C ( 5°C/hour ) , kept at −30°C for 3 hr , and raised to 0°C ( 5°C/hr ) . Samples were then washed with acetone , infiltrated with increasing concentrations of EPON in acetone ( 25 , 50 and 75% ) , embedded in 100% EPON and polymerized at 60°C for 2 days . Sections of 300 nm and 50 nm thickness were cut with a microtome ( Ultracut UCT; Leica , Wetzlar , Germany ) and collected on copper–palladium slot grids ( Science Services , München , Germany ) coated with 1% Formvar ( Plano , Wetzlar , Germany ) . As fiducial markers , 15 nm of gold-conjugated Protein A ( CMC university Medical Center Utrecht , Utrecht , Netherlands ) was absorbed on both sides of 300 nm sections . Sections were post-stained with 2% UA and lead citrate . Single or dual axis tilt series were acquired with a TECNAI TF30 transmission EM ( TEM; 300 kV; FEI , Hillsboro , OR ) equipped with a 2k x 2k Eagle camera ( FEI ) by using the Serial EM software ( Mastronarde , 2005 ) . The samples were pre-irradiated by an electron beam to minimize sample shrinkage during tilt series acquisition . Images were recorded over a −60° to 60° tilt range with an angular increment 1° at a pixel size of typically 0 . 75 nm or 1 . 0 nm . Tomograms were reconstructed using R-weighted backprojection method implemented in the IMOD software package ( version 4 . 5 . 6 ) ( Kremer et al . , 1996 ) . Dual axis tilt series were aligned using gold fiducial markers while single axis tilt series were aligned by patch tracking . It should be noted that tomographic resolution permits great advantages over classical approaches in which the thickness of a section is in the range of the diameter of a nuclear pore and it is thus not clear which parts of it are within the section and into which exact direction they are projected into in the electron micrograph . In contrast , 3D data permit a much more accurate mapping of orientation , membrane topology and substructures for each individual NPC , which is essential to study rare intermediates . Grids carrying 50 nm sections were pretreated with 0 . 1% Trition X-100 ( Sigma Aldrich ) in phosphate-buffered saline ( PBS ) for 10 min and blocked with 1% BSA and 0 . 1% fish skin gelatin ( Sigma Aldrich ) in PBS for 1 hr . Sections were then incubated with a primary antibody mAb414 ( Covance , Princeton , NJ; RRID:AB_10063490 ) , which recognizes four nucleoporins ( Nups 358 , 214 , 153 , and 62 ) , for 2 hr , a rabbit anti-mouse secondary antibody ( Cat . No . Z0259; Dako , Hamburg , Germany; RRID:AB_2532147 ) for 1 hr , and 10 nm of gold-conjugated Protein A ( CMC university Medical Center Utrecht ) for 30 min . The antibodies and Protein A beads were diluted in PBS with 0 . 2% BSA and the sections were washed for five times with PBS containing 0 . 2% BSA between steps . After multiple washes with PBS , sections were fixed in 2 . 5% glutaraldehyde in PBS for 20 min in order to immobilize the antibodies and Protein A-gold beads on sections . After washing with water , sections were post-stained with 2% UA and lead citrate for contrast enhancement . All steps were carried out at room temperature . Images were taken on a TEM ( CM 120 Biotwin; Phillips , Hillsboro , OR ) . For specificity analysis of immuno-EM labeling , the number of gold particles on assembly intermediates and ones nonspecifically attached within 50 nm under the inner nuclear membrane was counted . The nuclear envelope of HeLa cells was isolated and cryo-fixed as described previously ( Bui et al . , 2013; Ori et al . , 2013 ) . Tilt series of cryo-EM images were acquired using a Titan Krios TEM ( FEI ) at a pixel size of 0 . 34 or 0 . 43 nm and tomograms were reconstructed using the IMOD software package as described in ( Bui et al . , 2013 ) . The outlines of outer and inner nuclear membrane ( ONM and INM ) were manually marked by clicking points within the tomographic volume in the IMOD software package . The sets of clicked points were aligned to share an x-axis corresponding to INM and interpolated using a spline fit , and the resulting coordinates were fitted locally using a second degree polynomial fit as described in Kukulski et al . ( 2012 ) . The maximum depth of the INM evagination was determined from these two-dimensional profiles . For the ONM/INM distance , the median of the distance at 50 points between 45 and 90 nm away from mature pores was measured . The alignment , the interpolation , and the extraction of the parameter were done in MATLAB 7 . 4 . The maximum diameter of assembly intermediates and mature pores was measured manually from the top view images as illustrated in Figure 2D . The mature pores used for the measurement were the ones found in cells at >3 hr post anaphase . Unpaired t-tests with the assumption of equal variances were performed to compare two groups . The number of mature pores and intermediates was counted manually in the tomograms . The nuclear surface area was measured in each tomogram by manually tracing the NE using the IMOD software package . The shrinkage of the specimen was corrected by comparing the diameter of mature pores in EM tomograms of plastic resin with the one in cryo-EM tomograms . The shrinkage was 22 ± 2 . 7% ( the average and standard deviation , N = 13 sections ) and 15 ± 2 . 8% ( N = 11 sections ) in Lowicryl and EPON resin , respectively . Since non-core and core regions are hard to distinguish in the matured NE after late G1 , the pore density in cells >3 hr post anaphase was measured in any regions of the NE . Kinetic modeling of nuclear pore density is described in Materials and methods . We modeled pore maturation using delay equations ( Figure 3—figure supplement 1A ) . Assembly intermediates are generated with a rate V ( t ) and enter maturation with a rate constant kM . After a time interval KM an intermediate becomes fully matured to a NPC . We denote by τS the time required post anaphase to seal the NE and end the postmitotic assembly of the NPC . The simulation time t is related to the time after anaphase onset tAO by t=tAO−τS . For the data shown in Figure 3C , D we took τS = 10 min by assuming a start of interphase assembly to be 10 min after anaphase onset when the NE is sealed ( Dultz et al . , 2008; Otsuka et al . , 2014 ) . We simulated the process for τS= 0‒15 min , and found that τS+τM differed by less than 1% . The number of intermediates I and mature pores M is given by ( 1 ) dI ( t ) dt=V ( t ) −kMI ( t ) ( 2 ) dM ( t ) dt=kMI ( t−τM ) −kdM ( t ) , where kd is the degradation rate constant of mature pores . Consequently the number of intermediates in the process of maturation Im is given by ( 3 ) dIm ( t ) dt=kMI ( t ) −kMI ( t−τM ) . The total number of intermediates IT is the quantity we can measure which is given by ( 4 ) IT ( t ) =I ( t ) +Im ( t ) . Surface densities are computed from iT=IT/A , m=M/A , where A is the nuclear surface area . The overall maturation time is defined as ( 5 ) TM=1kM+τM We assume isotropic expansion of the nucleus where the surface area can be described byA ( tAO ) =a0+a1 ( 1−exp ( −kg1tAO ) ) +kg2tAO We obtained a0= 424 µm2 ( 95% confidence interval ( CI ) [212‒448] ) , a1 = 161 µm2 ( 95% CI [81‒185] ) , kg1 = 0 . 0722/min ( 95% CI [0 . 036‒0 . 093] ) , kg2 = 0 . 397/min ( 95% CI [0 . 199‒0 . 405] ) by fitting Equation 6 to the data in Figure 3—figure supplement 2 . We tested different intermediate production variants ( Figure 3—figure supplement 1B ) as ( 7 ) V ( t ) ={vA ( t ) , Variant 1 ( v1exp ( −kv ( t−tS ) ) +v0 ) A ( t ) , Variant 2i ( tS ) A ( tS ) δ ( t−ts ) +v0A ( t ) , Variant 3 . Variant 1 assumes a constant production rate density; Variant 2 assumes a time dependent production that decreases with time to a basal rate v0; finally in Variant 3 the majority of pores are initiated at tS . For Variant 3 the initial densities i ( tS ) and m ( tS ) , are estimated for the inner- and outer-core regions separately . The value of i ( tS ) is set to 0 for Variant 2 . We take im ( tS ) = 0 . The system of equations ( Equations 1‒6 ) is solved analytically to obtain the densities of intermediates and mature pores . The production and degradation rates are estimated from this and previous studies . For the mature pore degradation rate constant we take kd = 0 . 00042/min , which yields a pore life time of ~40 hr ( Rabut et al . , 2004; Schwanhausser et al . , 2011 ) . For Variant 2 and Variant 3 we take v0 = 0 . 0015 intermediates/µm2/min . This yields an NPC density in mature NEs ( 3‒20 hr post anaphase ) of 11 . 47 ± 1 . 33 NPCs/µm2 ( 0 . 65 ± 6e-4 intermediates/µm2 ) for Variant 3 and 12 . 1 ± 1 . 2 NPCs/µm2 ( 0 . 41 ± 1 . 36e-04 intermediates/µm2 ) for Variant 2 . Here the mean and standard deviation were estimated from inner-core*0 . 68 + outer-core*0 . 32 since the ratio of the surface area between inner- and outer-core regions is 0 . 68:0 . 32 . The other model parameters are estimated by minimizing the sum of squared residuals ( MATLAB routine lsqnonlin ) ( 8 ) F=1σ2∑j=1n/2 ( iT ( tj ) −Di ( tj ) ) 2+ ( m ( tj ) −Dm ( tj ) ) 2 , where Di and Dm are the measured densities of intermediate and mature pores ( Figure 3C , D ) , n is the number of data points , and σ2 = 2 . 18 pores/µm2 is the mean standard deviation estimated from all measurements . For Variant 1 we obtained kM = 7 . 13/min ( 95% CI [0 . 0379‒7 . 5730] ) , τM = 18 . 23 min ( 95% CI [0 . 725*‒27 . 59] ) and TM = 18 . 37 min ( 95% CI [16 . 12‒27 . 69] ) , for Variant 2 kM = 0 . 0996/min ( 95% CI [0 . 0274‒2 . 06] ) , τM = 17 . 75 min ( 95% CI [0 . 725*‒44 . 59] ) and TM = 27 . 78 min ( 95% CI [18 . 26‒44 . 72] ) , for Variant 3 kM = 1 . 357/min ( 95% CI [0 . 0408‒20*] ) , τM = 43 . 03 min ( 95% CI [18 . 78‒49 . 86] ) and TM = 43 . 76 ( 95% CI [41 . 32‒50] ) . For Variant 2 and 3 the model fit does not change for very low values τM or high values of kM , respectively . The asterisk indicates that the 95% boundary of the distribution has not yet been reached at the given value . The profile-likelihood method ( Venzon and Moolgavkar , 1988 ) has been used to estimate the 95% confidence . In this method the log-increase φ ( par ) =n[log ( F ( par ) n ) −log ( F ( parmin ) n ) ] of the mean squared distance F with respect to the best fit parmin was computed by varying the parameter of interest and optimizing the other parameters to the n data points . For |φ ( par ) |<χ1 , 0 . 952 = 3 . 84 the parameter is within its 95% CI . For TM ( Equation 5 ) the confidence interval is computed from the values of kM and τM . The quality of the fits ( lower sum of squared residuals , Figure 3—figure supplement 1C ) was slightly better for Variant 3 . Furthermore the obtained maturation time was more in agreement with previous reported values ( Dultz and Ellenberg , 2010 ) . We thus investigate alternatives to the maturation mechanism using Variant 3 only . The model with a multi-step maturation process ( Figure 3—figure supplement 1F ) reads ( 9 ) dI1dt=V ( t ) −kM1I1 ( 10 ) dIjdt=kM ( j−1 ) Ij−1−kMjIj , for j = 2 , . . . , N−1 ( 11 ) dMdt=kM ( N−1 ) IN−1−kdM . The sum of all intermediates ∑j=1N−1Ij/Aand M/A are fitted to the intermediate and mature pore densities , respectively . We modeled Variant 3 for the pore initiation . We found that there was no significant difference in the quality of the fits when assuming equal transition rate constants . Consequently the simulations shown are for kMj=kM . The maturation time defined as the characteristic time of mature pore appearance reads ( 12 ) TM=∑j=1N−11kMj= ( N−1 ) kM . We simulate different maturation times by allowing a maturation time distribution P ( τ ) , with finite positive support as ( 13 ) dI ( t ) dt=V ( t ) −kMI ( t ) ( 14 ) dIm ( t ) dt=kMI ( t ) −kM∫0∞P ( τ ) I ( t−τ ) dτ ( 15 ) dM ( t ) dt=kM∫0∞P ( τ ) I ( t−τ ) dτ−kdM ( t ) . The example shown in Figure 3—figure supplement 1H is for an uniform distribution of τM±w , where w is the half-width of the distribution . For tagging Nup107 at the N-terminus with monomeric enhanced GFP ( mEGFP ) , zinc finger nucleases ( ZFN ) containing DNA binding sequences in the 5’-3’ direction of TCAGTACTGATG and GCTGAGCCCGAAGTC were purchased from Sigma Aldrich . The donor plasmid consists of mEGFP cDNA sequence flanked by a left homology arm ( ENSEMBL release 75 , ENST00000229179 , chromosome 12: 68686269–68687065 ) and a right homology arm ( ENSEMBL release 75 , ENST00000229179 , chromosome 12: 68687065–68687892 ) . ZFN and the donor plasmid were transfected into HeLa cells as described in Mahen et al . ( 2014 ) . For tagging Nup358 at the N-terminus with mEGFP , CRISPR-Cas9 nickases were used . pX335-U6-Chimeric_BB-CBh-hSpCas9n ( D10A ) was a gift from Feng Zhang ( Addgene plasmid # 42335 , Cambridge , MA ) , and gRNAs were designed using the Feng Zhang Lab’s Target Finder ( http://crispr . mit . edu/ ) . The following gRNAs for Nup358 were cloned into pX335 ( Cong et al . , 2013 ) : 5’CCTGAGCGCTGGTCTCACGCGCC3’ and 5’GAGGCGCAGCAAGGCTGACGTGG3’ . CRISPR-Cas9 nickases and the donor plasmid were transfected using jetPRIME reagent ( Polyplus , New York , NY ) , according to the manufacture’s protocol . 7–10 days after transfection , cells were sorted with a MoFlo Legacy cell sorter ( Beckman Coulter , Brea , CA ) as described in Mahen et al . ( 2014 ) . Genomic DNA was prepared using ISOLATE II Genomic DNA Kit ( Bioline , Taunton , MA ) according to the supplier’s manual . Junction PCR was performed at endogenous loci to detect the insertion of mEGFP using separate sets of primers , one of which anneals inside mEGFP and the other one outside of the gene of interest . The primer sequences are as follows: Nup107 forward ( 5’ATTAATAAAAGGTATAAATGCCAGCAACAG3’ ) , Nup107 reverse ( 5’CACCTGGTCAACAACTACTTACTCCT3’ ) , NUP358 forward ( 5’GCATAAGACGGTGGTTCTGGAACCAATC3’ ) , and NUP358 reverse ( 5’AGCAAACTGACTCAAGATTCTGCGCA3’ ) . Touchdown PCR was performed using HotStar HiFidelity ( Qiagen , Hilden , Germany ) according to the supplier’s protocol . Cells were lysed for 20 min on ice in lysis buffer ( 10% glycerol , 1 mM DTT , 0 . 15 mM EDTA , 0 . 5% Triton X-100 , complete protease inhibitor cocktail and PhosSTOP ( Roche , Basel , Switzerland ) ) . Protein concentration was quantitated using the Bio-Rad Protein Assay ( Bio-Rad , Hercules , CA ) . 40 µg of total protein was run onto NuPAGE®4–12% Bis-Tris Gels ( Novex Life Technologies , Waltham , MA ) and transferred onto PVDF membrane using the Bio-Rad transfer system . After blocking with 5% milk solution ( nonfat milk powder in PBS + 0 . 1% Tween 20 ) , the following primary antibodies were used to label the proteins of interests: anti-Nup107 ( ab178399 , abcam , Cambridge , United Kingdom; RRID:AB_2620147 ) , anti-RanBP2 ( ab197044 , abcam; RRID:AB_2620148 ) , anti-tubulin ( DM-1A , Sigma; RRID:AB_521686 ) and anti-GFP ( Cat . No . 11814460001 , Roche; RRID:AB_390913 ) . Subsequently horseradish peroxidase ( HRP ) -conjugated secondary antibodies ( ECL anti-rabbit IgG HRP-linked whole antibody NA934V; RRID:AB_772206 , or ECL anti-mouse IgG HRP-linked whole antibody NA931V; RRID:AB_772210 , GE Healthcare , Little Chalfont , United Kingdom ) were used to detect the protein of interests with chemiluminescence reaction . Average intensities of Nup358 and Nup107 in the inferred inner-core and non-core regions were quantified . The total intensities were calculated by multiplying the average intensities by the nuclear surface area . Methods for the segmentation of core regions is described in ‘Segmentation of the nucleus and core regions’ section above . We formulated a sequential assembly model that describes the recruitment of Nup107 and Nup358 . Nup107 accumulates first with a rate constant k , and Nup358 assembles later with a rate constant l . The number of NPC intermediates can be described as ( 16 ) dN0xdt=−kxN0x ( 17 ) dN1xdt=kxN0x−lxN1x , wherex=pm , ip ( 18 ) dN2xdt=lxN1x , where N0 , 1 , 2pm , ip denote the number of NPCs that assemble through the interphase ( ip ) or postmitotic ( pm ) pathway without Nup107 nor Nup358 ( N0pm and N0ip ) , with Nup107 only ( N1pm and N1ip ) , or both Nup107 and Nup358 ( N2pm and N2ip ) . The rate constants for postmitotic and interphase recruitment of Nup107 and Nup358 are given by kpm and kip , and lpm and lip , respectively . The degradation and interphase production rates are small ( Rabut et al . , 2004; Dultz and Ellenberg , 2010; Schwanhausser et al . , 2011 ) and can be neglected in the time frame of 2 hr post anaphase ( see also ‘kinetic modeling of nuclear pore densities’ above ) . The total number of NPCs containing Nup107 and Nup358 are then given by ( 19 ) TNup107=[ ( N1pm+N2pm ) fpm+ ( N1ip+N2ip ) ( 1−fpm ) ] ( 20 ) TNup358=[N2pmfpm+N2ip ( 1−fpm ) ] where fpm is the fraction of postmitotic NPC . In the non-core region fpm = 0 . 92 , whereas in the inner-core region fpm = 0 . 5 ( Figure 3 ) . Since the nuclear membrane is not yet sealed before 10 min post anaphase ( Dultz et al . , 2008; Otsuka et al . , 2014 ) , we take for initial condition N0ip ( t = 10 min ) = 1 , and 0 for t < 10 min . For the postmitotic assembly we take N0pm ( t = 4 min ) = 1 , and 0 for t < 4 min . The four kinetic rate constants kpm , kip , lpm and lip are estimated by simultaneously fitting the total Nup107 and Nup358 intensities in non-core and inner-core regions from 4 min up to 125 min post anaphase ( Figure 4 ) . To match the experimental data normalization we also normalize the simulations . The normalization coefficients range from 1‒1 . 1 . The normalized pore densities are obtained by multiplying the normalized total number of pores by the nuclear surface area and subsequently dividing it by the maximal area . We obtained kpm = 0 . 355/min ( 95% CI [0 . 333‒0378] ) , lpm = 0 . 0437/min ( 95% CI [0 . 0425‒0 . 0449] ) , kip = 0 . 0374/min ( 95% CI [0 . 0335‒0 . 0417] ) , and lip = 0 . 0276/min ( 95% CI [0 . 0209‒0 . 0354] ) . Confidence intervals are obtained as explained in ‘kinetic modeling of nuclear pore densities’ above . After release from thymidine block , the division process of GFP-Nup107 genome-edited cells were monitored every 30 s by confocal microscopy ( LSM780; Carl Zeiss ) using 10 × 0 . 3 NA Plan-Neofluar objective ( Carl Zeiss ) . Cells were then fixed with paraformaldehyde and immunostained as described in the previous report ( Szymborska et al . , 2013 ) , with rabbit anti-Nup358 ( Cat . No . HPA018437 , The Human Protein Atlas; RRID:AB_2620151 ) and mouse anti-GFP ( Cat . No . 11814460001 , Roche; RRID:AB_390913 ) antibodies , and Abberior STAR RED-conjugated anti-rabbit IgG ( Cat . No . 2-0012-011-9 , Abberior GmbH , Göttingen , Germany; RRID:AB_2620152 ) and Abberior STAR 580-conjugated anti-mouse IgG ( Cat . No . 2-0002-005-1 , Abberior GmbH; RRID:AB_2620153 ) . Cells were mounted in Vectashield containing 4' , 6-diamidino-2-phenylindole ( DAPI ) ( Cat . No . H-1500 , Vector Laboratories Inc . , Burlingame , CA ) . Super-resolution imaging was performed on a Leica SP8 3X STED microscope , equipped with 775 nm pulsed wave depletion and white light pulsed lasers , Leica HCX 100 × 1 . 4 NA Plan Apochromat objective , and time-gated hybrid detectors ( Leica HyD ) . Excitation wavelength was adjusted to 580 and 633 nm , and bandpass filters were set to 585−630 and 650−702 nm , and the two channels were recorded pseudo-simultaneously by line switching . The fluorescent nuclei stained with DAPI were also recorded afterwards . The images were taken with a final optical pixel size of 20 nm , z-stacks of every 200 nm , and the optical section thickness of 550 nm . Images were filtered with a Gaussian filter ( kernel size: 0 . 5 × 0 . 5 pixel ) for presentation purposes . Lines with the width of 400 nm were drawn along the edge of the DAPI-stained nuclei . Non-core and core regions were inferred as described in Figure 1—figure supplement 1B‒G . The NEs on the lines were flattened and fluorescence intensity was quantified after binning of 15 pixels ( correspond to 300 nm width ) along the lines . The intensity difference between two channels was normalized using the images in non-core regions , and the intensity ratio of Nup107 to Nup358 was measured . All analyses were done in ImageJ ( http://rsbweb . nih . gov/ij/ ) . Assembly intermediates which have similar membrane profiles were selected at each time point and subjected to subtomogram averaging . The averaging was done on nuclear pores in freeze-substituted and plastic-embedded cells using the previously described averaging method ( Beck et al . , 2004 ) . Briefly , the subtomograms , which contain mature pores and intermediates , were extracted from the tomograms . The extracted subtomograms were aligned using iterative missing wedge compensation alignment procedure . Afterwards , the aligned subtomograms were averaged and visualized . The mature pores used for the averaging are the ones found in cells at >3 hr post anaphase . The overall structural similarity of the averaged nuclear pores to the respective cryo structures ( Figure 1E ) indicates a good structure preservation in freeze-substituted and plastic-embedded cells . For correlative light and electron microscopy , we first obtained one tomogram in each non- , inner- and outer-core region in 4 different cells at 19 , 28 , 53 , and 116 min after anaphase onset as pilot experiments . We then increased the number of dataset and eventually took 158 tomograms in 14 different cells . The exact value of the analyzed surface area and the number of nuclear pores found are listed in Table 1 . We picked up all the NE evaginations which were visible in the EM tomograms and did not perform any selection . Statistical analyses of the pore structure and density were performed only after all the data were taken . For immuno-EM , time-lapse 3D imaging , and STED microscopy , the data were from two independent experiments and the statistical analysis was carried out after all the data were obtained . Statistical analysis methods , sample sizes ( N ) and P values ( P ) for each experiment are indicated in figure legends . | The nucleus is the compartment within our cells that contains most of our genetic material . It is separated from the rest of the cell by a boundary called the nuclear envelope , which consists of two layers of membrane . All transport in and out of the nucleus has to pass through channels in the envelope , formed by large protein assemblies called the nuclear pore complexes . Each nuclear pore complex is composed of multiple copies of over 30 different proteins termed nucleoporins and there are several hundred proteins per pore . Before a cell divides in two , the nucleus has to grow and new nuclear pore complexes must be assembled into the double membrane barrier of the nuclear envelope . The assembly process would require the two nuclear membranes to fuse . However , exactly how nuclear pore complexes are assembled has been controversially debated for over 15 years , because no one has directly observed any of the intermediate stages during the assembly process . Now , Otsuka et al . have captured images of the different steps involved in assembling a nuclear pore complex in a human cell . This was achieved by observing living human cells in which the nucleus was growing and then studying them using advanced techniques such as high-resolution three-dimensional electron tomography and super-resolution microscopy . Otsuka et al . saw dome-shaped bumps or protrusions in the inner nuclear membrane that grew wider and deeper until they fused with the flat outer nuclear membrane . A ring of proteins surrounded the base of these protrusions from the beginning , and the membrane was deformed by a mushroom-shaped collection of proteins . Analysis of the molecules involved in these stages showed that assembly intermediates initially contained nucleoporins that face into the nucleus , and only later were nucleoporins that face into the rest of the cell added to the complex . The discovery that nuclear pore complexes assemble via an inside-out mechanism in human cells provides a new conceptual framework to interpret existing genetic and biochemical data . The findings also provide a new approach to explore the assembly process in much more detail and ask how nuclear pores first evolved . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [] | 2016 | Nuclear pore assembly proceeds by an inside-out extrusion of the nuclear envelope |
Bacterial swarming and biofilm formation are collective multicellular phenomena through which diverse microbial species colonize and spread over water-permeable tissue . During both modes of surface translocation , fluid uptake and transport play a key role in shaping the overall morphology and spreading dynamics . Here we develop a generalized two-phase thin-film model that couples bacterial growth , extracellular matrix swelling , fluid flow , and nutrient transport to describe the expansion of both highly motile bacterial swarms , and sessile bacterial biofilms . We show that swarm expansion corresponds to steady-state solutions in a nutrient-rich , capillarity dominated regime . In contrast , biofilm colony growth is described by transient solutions associated with a nutrient-limited , extracellular polymer stress driven limit . We apply our unified framework to explain a range of recent experimental observations of steady and unsteady expansion of microbial swarms and biofilms . Our results demonstrate how the physics of flow and transport in slender geometries serve to constrain biological organization in microbial communities .
Bacteria employ sophisticated surface translocation machinery to actively swarm , twitch , glide or slide over solid surfaces ( Kearns , 2010; Mattick , 2002; Spormann , 1999; Hölscher and Kovács , 2017 ) . Collectively , they also aggregate into multicellular communities on hydrated surfaces and exhibit large-scale coordinated movement ( Verstraeten et al . , 2008 ) . Surface motility in macroscopic colonies on hydrated surfaces such as gels occurs primarily via two distinct modes: either by rapid flagella-mediated swarming expansion ( Harshey , 1994; Harshey , 2003 ) , or alternatively by slow biofilm expansion driven by extracellular polymer matrix production ( Hall-Stoodley et al . , 2004 ) . In both cases , an interplay between mechanical constraints and biological organization sets limits on the overall colony morphology and expansion dynamics ( Persat et al . , 2015 ) . The forces driving colony expansion are generated by non-homogeneous patterns of biological activity , originating from spatial localizations in cell growth and division ( Hamouche et al . , 2017 ) , extracellular polymer matrix production ( Seminara et al . , 2012; Yan et al . , 2017; Srinivasan et al . , 2018 ) , osmolyte secretion ( Ping et al . , 2014 ) and active stresses ( Farrell et al . , 2013; Delarue et al . , 2016 ) . Conversely , the formation of localized biologically active zones is tightly coupled to the heterogeneity of the environment , including the diffusion and transport of nutrients ( Wang et al . , 2017 ) , accumulation of metabolic by-products ( Liu et al . , 2015; Gozzi et al . , 2017 ) and presence of quorum sensing and signaling agents that regulate cell-differentiation and development . Consequently , the dynamics of colony growth requires a mechanistic description that accounts for spatiotemporal inhomogeneities in biological activity , emergent forces , and flows that transport metabolic agents . In bacterial swarming , cells within the colony are actively propelled by the rotation of flagella in a thin layer of fluid extracted from the underlying soft tissue or gel ( Kearns , 2010 ) . In contrast , bacterial biofilms are surface aggregates of sessile bacteria embedded in a self-generated extracellular polymer matrix ( Flemming and Wingender , 2010 ) . Despite marked differences in regulatory genetic pathways , morphology and cell function ( Verstraeten et al . , 2008 ) , physical characteristics such as the fluidization of the substrate/tissue , gradients in nutrient availability , the low-aspect-ratio geometry and the existence of multiple phases ( i . e . cells , biopolymer and fluid ) are common to both bacterial film and swarm colonies . Motivated by these similarities , we present a unified multiphase framework that couples mechanics , hydrodynamics and transport to explain the dynamics of bacterial swarm and film expansion .
Experiments on swarming colonies of E . coli ( Darnton et al . , 2010; Wu and Berg , 2012; Ping et al . , 2014 ) , S . enterica ( Harshey and Matsuyama , 1994; Butler et al . , 2010; Kalai Chelvam et al . , 2014; Chen et al . , 2007 ) and P . aeruginosa ( Yang et al . , 2017 ) reveal certain reproducible features associated with this modality of collective behavior . For example , E . coli swarms on agarose gels have a steady front shape that propagates radially at a uniform speed ( Wu and Berg , 2012 ) . In these swarms , measurements of the osmotic pressure profiles were found to be consistent with the active secretion of wetting agents in regions of high cell density that serve to fluidize the swarm by extracting water from the underlying tissue , thus allowing it to spread ( Ping et al . , 2014 ) . These observations are not unique to E . coli; indeed our experiments with B . subtilis swarms , following ( Kearns and Losick , 2003 ) , indicate the same phenomena , that is a steady-state front shape and speed , as shown in Figure 1A–1E . Close to the spreading front , we observe a multilayer region of width W = 195 µm ± 35 µm , indicated by the dashed white lines in Figure 1B and 1C . The multilayer region correlates with increased colony thickness and local bacterial density ( Wu and Berg , 2012 ) . At the edge , and in the interior , there is just a monolayer of cells . The swarm radial expansion velocity is constant at V = 2 mm/hr ( see Figure 1D ) and the swarm front maintains a steady-state profile during expansion ( see Figure 1E ) . These observations raise a number of natural questions associated with the steady-state velocity and profile of the swarm colony . Given the observations of osmotic gradient-driven flow in the vicinity of the growing front ( Ping et al . , 2014 ) , coupled with variations in the thickness and activity of bacteria , any framework to explain these requires a consideration of a dynamic bacterial population interacting with ambient fluid , necessitating a multiphase description . In contrast with bacterial swarms , the spreading of bacterial biofilms is faciliated by the extracellular polymeric substance ( EPS ) matrix that expands via osmotic fluid influx , for example in B . subtilis ( Seminara et al . , 2012 ) and V . cholerae ( Yan et al . , 2017 ) biofilm colonies . However , EPS synthesis is not homogeneous , and depends on the local nutrient concentration and environmental heterogeneities experienced by cells within the same biofilm ( Vlamakis et al . , 2008; Berk et al . , 2012 ) . Recently , it was shown that the EPS matrix production is localized to cells in the propagating front of B . subtilis biofilms ( Srinivasan et al . , 2018 ) . In Figure 1F–1J , we show the results of repeating these experiments , but now focusing on a peripheral region of a biofilm colony using a B . subtilis strain ( MTC832 ) that harbors the P𝑡𝑎𝑝𝐴-cfp construct as a reporter for matrix production activity ( Wang et al . , 2016; Srinivasan et al . , 2018 ) . This highlights a ∼1 mm zone of matrix production activity at the periphery , seen in Figure 1G and H; indeed plots of averaged matrix production reporter intensity exhibit a distinct peak at the periphery , as shown in Figure 1J . The dynamics of radial expansion shows the existence of an initial acceleration regime followed by a transition to a second regime characterized by a monotonic decrease in expansion velocity , as plotted in Figure 1I . This transient mode of biofilm spreading driven by EPS production and swelling is quite different from that of bacterial swarming , and suggests that we might need a fundamentally different way to address its origins . However , if we now consider the EPS matrix and fluid as distinct phases ( Cogan and Keener , 2004; Cogan and Keener , 2005; Winstanley et al . , 2011; Seminara et al . , 2012 ) , with the bacterial population being relatively small , we are again led to a multiphase description of the system , but with a different dominant balance relative to that seen in bacterial swarms , which we now turn to .
For both swarms and biofilms , the active phase ( i . e . , swarm cells or the EPS matrix ) is generated within the bacterial colony by converting nutrient in the underlying substrate to biomass . The rate of change of nutrient concentration within the substrate depends on diffusion and nutrient uptake ( see Equation ( 3 ) and Equation ( A2 ) ) , and is derived in Appendix 2 . When the substrate concentration is scaled by the initial concentration c0 , the nutrient depletion rate depends on Γ/c0 , the ratio of the specific nutrient consumption rate to the initial concentration . Bacterial swarming is typically associated with nutrient rich conditions , where c0≫Γ . As a result , the nutrient uptake term can be neglected in bacterial swarming as g2→0 , and the concentration c≈c0 throughout swarm expansion . In contrast , biofilm growth occurs under nutrient limited conditions where Γ/c0∼O ( 1 ) , resulting in a corresponding uptake term shown in Table 1 . Therefore , biofilm expansion is necessarily unsteady and driven by the dynamics of the transient nutrient field . In both swarms and biofilms , the generation of the active phase drives colony expansion and is described by the growth term in Equation ( 1 ) using a logistic function g1=g0hϕ ( 1-hϕ/ ( Hϕ0 ) ) to model the active phase growth , where Hϕ0 is the limiting thickness , and g0 indicates a specific growth rate . In bacterial swarms , g0 is independent of the nutrient concentration ( as c≈c0 during swarm expansion ) . Therefore , the spreading swarm films have a steady-state structure that exhibits a central spatial plateau about hϕ=Hϕ0 . In contrast , biofilm growth corresponds to a nutrient poor environment . We model the biofilm growth dependence on nutrient concentration via a minimal Michaelis-Menten form g0=Gc/ ( K+c ) , . Unlike in nutrient rich conditions associated with swarms , this implies that biofilm growth is fundamentally transient; once the nutrient field at the interior is depleted as c→0 , biofilm growth term in that region is arrested and g1→0 independently of the vertical thickness ( i . e . , even if hϕ≠Hϕ0 ) . As a result , the biofilm does not give form a central plateau and the dynamics of the biofilm rim is fixed by the dynamics of nutrient depletion . Eventually the effect of the finite-size of the system ( the petri dish ) also becomes important it determines the overall dynamics of nutrient depletion . The terms Q1 ( x ) and Q2 ( x ) that represent the horizontal flux of the active and passive phases are obtained by depth integrating the momentum balance equations in the thin-film lubrication limit , as described in Appendix 2 ( c . f . Equations ( A9 ) - ( A11 ) ) . Within bacterial swarms , the passive aqeuous fluid phase is modeled as a Newtonian liquid with viscosity μ2 . The first term of Q1 ( x ) and Q2 ( x ) in Table 1 for swarms is generated by viscous and capillary stresses within the swarm fluid . The active swarmer cells are treated as inviscid and subjected to a hydrodynamic frictional drag force . Specifically , we assume that individual bacteria within the swarm are undergoing a random walk process with zero net displacement ( upon averaging over sufficiently large time-intervals ) . Even though there is no overall displacement , there is a net time-averaged drift that arises from viscous stokes drag interaction between the fluid and the active bacteria . The second term for Q1 ( x ) in Table 1 represents this time-averaged drift arising from frictional drag interaction of the bacteria with the swarm fluid . In biofilms , the EPS matrix phase constitutes an active viscous hydrogel network with viscosity μ1 , whereas the passive aqueous fluid phase is treated as a solvent with viscosity μ2 . The dominant stress within the EPS phase in the biofilm model arises from a Flory-Huggins swelling pressure in the polymer chains ( Cogan and Guy , 2010; Winstanley et al . , 2011 ) . In the fluid phase , the pressure pf is set by surface tension and curvature of the swarm fluid . Both these stresses contribute to the effective EPS phase pressure term Π ( x ) , as described in Appendix 2 . Consequently , the first term for Q1 ( x ) and Q2 ( x ) in Table 1 for biofilms is related to the gradient of the effective pressure . Moreover , following Winstanley et al . ( 2011 ) , we assume that the capillary and viscous stresses in the swarm fluid are negligible when compared to the frictional drag due to flow between water and the EPS polymer chain network in the biofilm model . Therefore , the second term for Q2 ( x ) in Table 1 represents a Darcy-type flow of the aqueous phase within the EPS matrix . The osmotic influx terms are considered separately in the following sections when describing the equations governing swarm and biofilm expansion .
To make sense of the scales in the problem , we use the dimensionless variables x^=x/L , z^=z/H and t^=tg0 where H is the vertical length scale , L is a horizontal length scale and 1/g0 is the time-scale associated with bacterial growth . The resultant horizontal velocity scale in the swarm colony is U=Lg0 . Swarm expansion is fluid driven , and therefore balancing the viscous stresses generated in the swarm fluid , with the curvature pressure due to surface tension ( Levich and Landau , 1942 ) results in μ2U/H2∼γH/L3 , where μ2 is the viscosity and γ is the surface tension of the aqueous phase . As a result , the natural horizontal length scale is L=H ( 𝐶𝑎 ) -1/3 , where 𝐶𝑎= ( μ2U/γ ) is a capillary number associated with the microbial swarm fluid . Consequently , in our model the expansion speed of the swarm colony , V=dR/dt , is determined by the product of the horizontal length scale and an effective growth rate , and is predicted to scale as , ( 5 ) V=C1g0H𝐶𝑎-1/3 . whereas , the swarm front itself is analogous to a capillary ridge in thin fluid film with a width W that is predicted to scale as , ( 6 ) W=C2𝐶𝑎-1/3 . where , C1 and C2 are dimensionless prefactors that require a detailed numerical calculation , and are discussed later . There are two important dimensionless parameters that describe swarm colony expansion . The first dimensionless parameter , α1 , relates the magnitude of capillary forces to the viscous drag acting on cells within the swarm and is defined as α1= ( γH/L2 ) / ( ζLU ) . Here , ζ=ζc/Vc where ζc is the friction coefficient of a single swarmer cell and Vc is its volume . The second dimensionless parameter α2 is defined as the ratio of a vertical fluid influx velocity Q0 , to a thickness velocity scale Hg0 associated with bacterial growth as α2=Q0/ ( Hg0 ) . The vertical length scale and equilibrium fluid volume fraction are estimated from the interior monolayer region as H = 0 . 5 μm and ϕ0=0 . 5 ( Wu and Berg , 2012 ) . We assume values of μ2=10-3 Pa . s for the ( aqeuous ) swarm fluid viscosity , and γ=10-2 N/m as its surface tension . The friction coefficient of a single cell is estimated from Stokes law as ζc=3πμ2a , and its volume is approximated as Vc=πa3/6 , where a = 1 μm is the cell diameter . Therefore , the friction coefficient is ζ=ζc/Vc≈18μ2/a2 . As a result of substituting the values of known parameters above , the dimensionless parameter α1 reduces to a constant geometric ratio , α1≈2a2/H2≈2/9≈0 . 22 . The value of α2 depends on the ratio Q0/g0 . Direct experimental measurements of the vertical influx fluid velocity profile V0 ( x ) and the spatial profiles of cell division in swarm colonies remain scarce ( Hamouche et al . , 2017 ) . In order to make progress in validating our model with real experimental data , the vertical fluid influx velocity scale is chosen as Q0=10−2μm/s . Consequently , we have chosen g0 as the only fitting parameter in our study , as detailed in Appendix 2 . As an example , in the following section we will show that a choice of g0=0 . 013 s-1 in our model reproduces the experimental swarm expansion speed shown in Figure 1D , and leads to a horizontal length scale of L=H ( Ca ) −1/3=100μm , velocity scale of U=Lg0=1 . 3μm s-1 , 𝐶𝑎=1 . 3×10-7 and a value of α2≈1 . 5 . A complete set of parameters for three experimental measurements of swarm expansion in B . subtilis , and two existing measurements in E . coli previously reported by Darnton et al . ( 2010 ) and Wu and Berg ( 2012 ) are summarized in Appendix 2 . With these assumptions , and assuming that the nutrient concentration is constant , Equations ( 1–3 ) reduce to the following scaled equations in the swarming limit , ( 7 ) ( h^ϕ ) t^+13 ( ϕh^3h^x^x^x^1-ϕ ) x^+α1 ( h^h^x^x^x^ϕ ) x^=h^ϕ ( 1-h^ϕ ) ( 8 ) ( h^ ( 1-ϕ ) ) t^+13 ( h^3h^x^x^x^ ) x^=α2ϕ-ϕ01-ϕ0 . To complete the formulation of the problem , we need five boundary conditions which are h^x^ ( 0 ) =h^x^ ( RP ) =0 , h^x^x^x^ ( 0 ) =h^x^x^x^ ( RP ) =0 , and ϕ ( 0 ) =ϕ0 , where RP is the dimensionless size of the petri-dish and is set much larger than the colony size ( R^=150 ) in our simulations . The initial condition corresponds to a circularly inoculated swarm colony , along with a thin pre-wetting film where no bacterial growth occurs ( see Appendix 3—figure 1 ) . Solving Equations ( 7 ) and ( 8 ) with the prescribed initial and boundary conditions numerically results in a steady state solution that advances at a constant speed ( see Figure 3 ) . In Figure 3 , we plot a representative steady state solution in the frame of the advancing front for α1=0 . 2 , α2=1 . 5 and ϕ0=0 . 5 . At the interior of the swarm , the average cell volume fraction is ϕ≈ϕ0 . Near the leading edge of the swarm , there is a region of enhanced thickness as indicated by the red line in Figure 3A . Immediately behind the leading edge , where the cell concentration is highest , so is the osmolyte concentration leading to fluid extraction from the substrate , while further behind , fluid is reabsorbed , as indicated by the arrows in Figure 3A . In Figure 3B , we show the steady-state osmotic flow solution and see that it correlates well with the experimentally measured osmotic pressure profile by Ping et al . ( 2014 ) in E . coli swarms . As shown in Appendix 3—figure 3 , our numerical horizontal flow profiles are also consistent the scaled radial fluid velocity measurements of Wu and Berg ( 2012 ) . In Figure 3C , we see that the radial expansion velocity scales as Hg0 and shows quantitative agreement with experiments and is insensitive to the fluid influx velocity scale when Q0≫g0H . Note that our model uses a coarse-graining procedure and represents the swarm thickness field using a continuum approximation . As a consequence , we are not able to quantitatively capture the decreasing height of the swarm ( i . e . , of the order of a few cells ) , that is experimentally observed over hundreds of micron towards the interior ( see Figure 1E ) . Furthermore , we corroborate our scaling law in Equation ( 5 ) by fitting our model to five independent experimental measurements of swarm expansion velocities for different systems , as shown in Figure 3D . These include measurements in B . subtilis swarms in this work , and in E . coli swarms previously reported by Darnton et al . ( 2010 ) and Wu and Berg ( 2012 ) that are summarized in Table A2 in Appendix 2 . The expansion velocity follows the -1/3 exponent predicted by Equation ( 5 ) for Ca varying from ∼ 5 × 10-8 to 10-6 . For each experiment , we have fit our theoretical model using the effective growth rate g0 as the fitting parameter and find that the numerical prefactor C1≈0 . 42 . However , as shown in Appendix 3—figure 5 in the Appendix , the measured multi-layer width does not follow the predicted scaling . From an experimental point of view , the width of the multi-layer region is not sharply defined in Figure 1E , and will depend on the choice of threshold . However , our multi-phase model is able to describe the zone of cellular and osmolyte activity near the leading edge that drives the advancing swarm front . This leads to a picture wherein the combination of a fluid-filled substrate and swarm front work together like a localized active circulatory system , quantitatively rationalizing the experimental observations of Wu and Berg ( 2012 ) and Ping et al . ( 2014 ) .
We consider dimensionless variables x^=x/L , z^=z/H , t^=tG , ϕ^=ϕ/ϕ0 and c^=c/c0 , where H is now the maximum biofilm thickness , G is the rate of EPS production , and c0 is the initial nutrient concentration in the substrate . As biofilm growth is nutrient limited ( Liu et al . , 2015 ) , the dimensionless length scale is determined from Equation ( 3 ) and is expected to scale as L= ( D/G ) 1/2 and the corresponding velocity scale is U= ( DG ) 1/2 . Using these scales , we can define the ratio of osmotic stresses relative to viscous stress in the EPS phase in terms of the dimensionless parameter , β1= ( Ψ0/L ) / ( μ1U/H2 ) , the ratio of capillary stresses relative to the EPS viscous stress in terms of another parameter , β2= ( γH/L3 ) / ( μ1U/H2 ) , the ratio of capillary stress to the interfacial drag in the aqueous fluid phase , β3= ( γH/L2 ) / ( ζUL ) , and the ratio of the fluid influx velocity to the EPS swelling velocity , β4=Q0/ ( HG ) . As shown in Appendix 2 , the effective nutrient uptake rate is S= ( ΓHϕ0 ) / ( c0d ) , where Γ is the nutrient consumption rate per unit concentration and d is the substrate thickness . Consequently , we define β5=S/G as the ratio of the effective nutrient uptake rate to the EPS production rate . We set the EPS production time-scale as G=1/40 min-1 , resulting in a horizontal length scale of L= ( D/G ) 1/2=1 . 1 mm and velocity scale U= ( DG ) 1/2=0 . 5μm/s . The effective nutrient uptake rate is estimated as S=1/25 min-1 , where we have taken d=7 mm as the substrate thickness ( Srinivasan et al . , 2018 ) , Γ=10-2 mM/s as the nutrient uptake rate ( Zhang et al . , 2010 ) , and c0=35 mM as the initial concentration of the carbon source . The friction coefficient is ζ∼μ2/ξ2 , where the EPS mesh size is ξ=50 nm ( Yan et al . , 2017 ) . Using measured estimates of the biofilm viscosity μ1=105 Pa . s ( Stoodley et al . , 2002; Lau et al . , 2009 ) , fluid phase viscosity μ2=10-3 Pa . s , surface tension γ=10-2 N/m , an osmotic scale Ψ0=2100 Pa ( Yan et al . , 2017 ) ( i . e . , ϕ0=0 . 04 ) , biofilm thickness H=400μm , and nutrient diffusivity in agarose gels of D=5×10-10m2/s ( Zhang et al . , 2010 ) implies that β1≈7 , β2≈0 . 01 , β3≈0 . 02 , β4≈1 and β5≈2 . Consequently , within the context of our model , it is evident that osmotic stresses , fluid influx and biomass growth are the dominant forces that drive colony expansion . Moreover , in the nutrient limited regime , our model predicts the transient maximum biofilm expansion velocity to scale as , ( 10 ) V=C3 ( DG ) 12whereas , the width of the propagating fronts of EPS production experimentally observed by Srinivasan et al . ( 2018 ) is predicted to scale according to , ( 11 ) W=C4 ( DG ) 12where C3 and C4 are once again dimensionless prefactors that require a detailed numerical calculation , as discussed later . With the above scaling assumptions , Equations ( 1–3 ) now reduces to the following partial differential equations that describe biofilm colony expansion , ( 12 ) ( h^ϕ^ ) t^−β13ϕ0 ( h^3 ( ϕ^3 ) x^ ) x^+β23ϕ0 ( h^3h^x^x^x^ ) x^=c^h^ϕ^ ( 1−h^ϕ^ ) K1+c^ , ( 13 ) ( h^ ( 1−ϕ0ϕ^ ) ) t^−β13 ( κ ( ϕ^ ) h^3 ( ϕ^3 ) x^ ) x^+β23 ( κ ( ϕ^ ) h^3h^x^x^x^ ) x^+β3 ( h^ ( 1−ϕ0ϕ^ ) κ ( ϕ^ ) h^x^x^x^ ) x^=β4 ( 1−ϕ0ϕ^ ) ( ϕ^3−1 ) , ( 14 ) c^t^-c^x^x^=-β5h^ϕ^c^K1+c^ . where κ ( ϕ^ ) = ( 1-ϕ0ϕ^ ) / ( ϕ0ϕ^ ) is a volume fraction dependent permeability term . The eight boundary conditions associated with Equations ( 12 ) – ( 14 ) are the symmetry boundary conditions h^x^ ( 0 ) =h^x^ ( RP ) =0 , h^x^x^x^ ( 0 ) =h^x^x^x^ ( RP ) =0 , ϕ^x^ ( 0 ) =ϕ^x^ ( RP ) =0 , ϕ^x^x^x^ ( 0 ) =ϕ^x^x^x^ ( RP ) =0 and c^x^ ( 0 ) =c^x^ ( RP ) =0 , where the dimensionless petri-dish size is chosen as RP=16 to match the size of typical 35 mm diameter petri dishes used in experiments ( Srinivasan et al . , 2018 ) . In Figure 4A , we plot the time evolution of the shape and nutrient concentration field for a biofilm colony of initial radius R^0=0 . 5 and thickness h^in=0 . 06 . Unlike in the case of swarms , the solutions to Equations ( 12 ) – ( 14 ) are transient , and exhibit two distinct expansion regimes: initial acceleration phase until t^c=5 , followed by a decelerating phase beyond . For t^<t^c , colony expansion arises as the microbes rapidly consumes locally available nutrient at the interior and synthesize fresh EPS matrix , generating spatial gradients in nutrient availability ( see Figure 4A ) . In Figure 4B , we show that the newly synthesized EPS generates a large osmotic pressure differential between the biofilm and the substrate , and osmotic fluid influx gradually relaxes the biofilm matrix to a swollen configuration . For t>tc , the localized zone of EPS production near the film front propagates with a fixed shape as shown in Figure 4C , consistent with the observed spatial localization in tapA gene activity ( see Figure 1J and Srinivasan et al . , 2018 ) . Moreover , the radial colony expansion profile in Figure 4D is also consistent with the non-monotonic front speed observed experimentally ( Srinivasan et al . , 2018 ) . For the specific experimental conditions we consider , our detailed theory allows us to estimate the prefactors in the scaling laws Equations ( 10 ) – ( 11 ) so that C3≈0 . 2 and C4≈1 . 8 . These results are hallmarks of a transition from a bulk to an edge biofilm growth mode , triggered by nutrient limitation ( Pirt , 1967 ) . In the deceleration regime , diffusive transport of nutrients from a region external to the colony continues to sustain EPS production at the biofilm periphery , analogous to Stefan-like problems in solidification . Our generalized multiphase model is thus able to quantitatively rationalize the expansion curves , transition time and localized biological activity observed experimentally , and demonstrates that nutrient availability and diffusive transport governs the dynamics of Bacillus subtilis macrocolonies grown on agar .
Analysis of collective microbial expansion in thin film geometries often prioritizes biological mechanisms , such as genetic regulation , developmental programs and cellular signaling/competition , over the role of the heterogeneous physical micro-environments . Here we have presented a multi-phase theory that quantitatively describes the expansion dynamics of microbial swarms and biofilms and considers variations in the colony thickness , an aspect of colony expansion that has often been overlooked in many theories ( Korolev et al . , 2012; Ghosh et al . , 2015; Wang et al . , 2017 ) . The resulting unified description of both steady-state swarms and transient biofilm spreading leads to simple estimates and scaling laws for the colony expansion rate that are validated via comparison with experimental measurements for different systems . In swarms , exudation of water from the permeable substrate via bacterial osmolyte secretion facilitates steady state colony expansion . Numerical solutions of our model demonstrate that the shape of the swarm front is determined by capillarity , and its expansion speed by cell-division and growth , leading to scaling laws validated by comparison with previous experiments . In contrast , transient biofilm macrocolony expansion on agar is driven by osmotic polymer stresses generated via EPS matrix production in a spatially localized zone at the periphery . Nutrient transport and depletion leads to the formation of these heterogenous zones , and results in two regimes in biofilm expansion . However our depth-integrated theory also has certain limitations . For example , we are unable to capture discrete thickness variations of the order of a few cells , which might require an agent-based approach . For bacterial swarms , our model is unable to quantitatively account for the region of enhanced thickness ( i . e . , the multilayer region in Figure 1C and E ) , likely because the multilayer width is difficult to experimentally ascertain , owing to the large tail distribution seen in the mean intensity trace in Figure 1E , and the arbitrariness in the choice of threshold in Appendix 3—figure 5 . Similarly , in the context of biofilm colony expansion , our model does not account for sliding and frictional contact between the cells/EPS matrix and the substrate ( Farrell et al . , 2013 ) . More generally , our mean-field picture neglects fluctuation-driven effects during colony expansion , such as the formation multicellular raft structures ( Kearns , 2010 ) and synchronized long-range interactions ( Chen et al . , 2017 ) . Natural next steps of our approach include ( i ) adding three-dimensional effects by allowing for spatial variations in the mechanical stresses , flows and nutrient fields in the vertical direction , ( ii ) accounting for orientational order in the bacterial swarms and films , and ( iii ) accounting for interfacial tension on the stability of the growing swarm/biofilm-fluid interface , especially in the context of fingering instabilities in microbial colonies Trinschek et al . ( 2018 ) . A rigorous multi-phase approach may also be relevant in revisiting pattern formation phenomena in microbial colony expansion ( Matsushita et al . , 1999 ) , that so far been addressed primarily using various non-linear diffusion models ( Golding et al . , 1998; Allen and Waclaw , 2019 ) that ignore the third dimension . Finally , from an experimental and theoretical perspective , our results naturally raise the question of controlling biofilm and swarm expansion by manipulating water and nutrient availability , complementing the better studied approaches of manipulating colonies by the genetic regulation of EPS production , cell division , and chemical signaling in microbial colonies . | Bacteria can grow and thrive in many different environments . Although we usually think of bacteria as single-celled organisms , they are not always solitary; they can also form groups containing large numbers of individuals . These aggregates work together as one super-colony , allowing the bacteria to feed and protect themselves more efficiently than they could as isolated cells . These colonies move and grow in characteristic patterns as they respond to their environment . They can form swarms , like insects , or biofilms , which are thin , flat structures containing both cells and a film-like substance that the cells secrete . Availability of food and water influences the way colonies spread; however , since movement and growth are accompanied by mechanical forces , physical constraints are also important . These include the ability of the bacteria to change the water balance and their local mechanical environment , and the forces they create as they grow and move . Previous research has used a variety of experimental and theoretical approaches to explain the dynamics of bacterial swarms and biofilms as separate phenomena . However , while they do differ biologically , they also share many physical characteristics . Srinivasan et al . wanted to exploit these similarities , and use them to predict the growth and shape of biofilms and bacterial swarms under different conditions . To do this , a unified mathematical model for the growth of both swarms and biofilms was created . The model accounted for various factors , such as the transport of nutrients into the colony , the movement of water between the colony and the surface on which it grew , and mechanical changes in the environment ( e . g . swelling/softening ) . The theoretical results were then compared with results from experimental measurements of different bacterial aggregates grown on a soft , hydrated gel . For both swarms and biofilms , the model correctly predicted how fast the colony expanded overall , as well as the shape and location of actively growing regions . Biofilms and other bacterial aggregates can cause diseases and increase inflammation in tissues , and also hinder industrial processes by damage to submerged surfaces , such as ships and waterpipes . The results described here may open up new approaches to restrict the spreading of bacterial aggregates by focusing on their physical constraints . | [
"Abstract",
"Introduction",
"Experimental",
"background",
"Theoretical",
"framework",
"Bacterial",
"swarms",
"Bacterial",
"films",
"Discussion"
] | [
"physics",
"of",
"living",
"systems"
] | 2019 | A multiphase theory for spreading microbial swarms and films |
Protein arginine methyltransferase 5 ( Prmt5 ) is the major type II enzyme responsible for symmetric dimethylation of arginine . Here , we found that PRMT5 was expressed at high level in ovarian granulosa cells of growing follicles . Inactivation of Prmt5 in granulosa cells resulted in aberrant follicle development and female infertility . In Prmt5-knockout mice , follicle development was arrested with disorganized granulosa cells in which WT1 expression was dramatically reduced and the expression of steroidogenesis-related genes was significantly increased . The premature differentiated granulosa cells were detached from oocytes and follicle structure was disrupted . Mechanism studies revealed that Wt1 expression was regulated by PRMT5 at the protein level . PRMT5 facilitated IRES-dependent translation of Wt1 mRNA by methylating HnRNPA1 . Moreover , the upregulation of steroidogenic genes in Prmt5-deficient granulosa cells was repressed by Wt1 overexpression . These results demonstrate that PRMT5 participates in granulosa cell lineage maintenance by inducing Wt1 expression . Our study uncovers a new role of post-translational arginine methylation in granulosa cell differentiation and follicle development .
Follicles are the basic functional units in the ovaries . Each follicle consists of an oocyte , the surrounding granulosa cells , and theca cells in the mesenchyme . The interaction between oocytes and somatic cells is crucial for follicle development . Follicle maturation experiences primordial , primary , secondary , and antral follicular stages . Primordial follicles are formed shortly after birth via breakdown of oocyte syncytia . Each primordial follicle is composed of an oocyte surrounded by a single layer of flattened pregranulosa cells that remains in a dormant phase until being recruited into the primary stage under the influence of two main signaling pathways ( Monniaux , 2016 ) . Once activated , flattened granulosa cells become cuboidal , and follicles continue to grow through proliferation of granulosa cells and enlargement of oocytes . Development of high-quality oocytes is important for female reproductive health and fertility ( Jagarlamudi and Rajkovic , 2012; Liu et al . , 2010; Monniaux , 2016; Richards and Pangas , 2010 ) . Although gonadotropin , follicle-stimulating hormone ( FSH ) , and luteinizing hormone ( LH ) are important for the growth of antral follicles , the early stages of follicle development are driven by a local oocyte-granulosa cell dialog . Abnormalities in this process may lead to follicle growth arrest or atresia ( Oktem and Urman , 2010; Richards and Pangas , 2010 ) . Granulosa cells are derived from progenitors of the coelomic epithelium that direct sexual differentiation at the embryonic stage and support oocyte development postnatally ( Liu et al . , 2010; Smith et al . , 2014 ) . Theca-interstitial cell differentiation occurs postnatally along with the formation of secondary follicles . The steroid hormone produced by theca-interstitial cells plays important roles in follicle development and maintenance of secondary sexual characteristics ( Liu et al . , 2010 ) . The Wilms’ tumor ( WT ) suppressor gene Wt1 is a nuclear transcription factor indispensable for normal development of several tissues . In gonads , Wt1 is mainly expressed in ovarian granulosa cells and testicular Sertoli cells . During follicle development , Wt1 is expressed at high levels in granulosa cells of primordial , primary , and secondary follicles , but its expression is decreased in antral follicles , suggesting that it might be a repressor of ovarian differentiation genes in the granulosa cells ( Hsu et al . , 1995 ) . Our previous studies demonstrated that Wt1 is required for the lineage specification and maintenance of Sertoli and granulosa cells ( Cen et al . , 2020; Chen et al . , 2017 ) . However , the underlying mechanism that regulates the expression of Wt1 in granulosa cells is unknown . Protein arginine methyltransferase 5 ( PRMT5 ) is a member of the PRMT family that catalyzes the transfer of methyl groups from S-adenosylmethionine to a variety of substrates and is involved in many cellular processes , such as cell growth , differentiation , and development ( Di Lorenzo and Bedford , 2011; Karkhanis et al . , 2011; Stopa et al . , 2015 ) . PRMT5 is the predominant type II methyltransferase that catalyzes the formation of most symmetric dimethylarginines ( SDMAs ) in the cells and regulates gene expression at the transcriptional and posttranscriptional levels ( Karkhanis et al . , 2011 ) . PRMT5 forms a complex with its substrate-binding partner , the WD-repeat protein MEP50 ( or WDR77 ) , which greatly enhances the methyltransferase activity of PRMT5 by increasing its affinity for protein substrates ( Stopa et al . , 2015 ) . In gonad development , inactivation of Prmt5 specifically in primordial germ cells ( PGCs ) causes massive loss of PGCs ( Kim et al . , 2014; Li et al . , 2015; Wang et al . , 2015 ) . PRMT5 promotes PGC survival by regulating RNA splicing ( Li et al . , 2015 ) and suppressing transposable elements at the time of global DNA demethylation ( Kim et al . , 2014 ) . In this study , we found that PRMT5 is expressed at high level in ovarian granulosa cells of growing follicles and the expression level changes with follicle development , suggesting that PRMT5 in granulosa cells plays a role in follicle development . To test the function of PRMT5 in granulosa cells , we specifically inactivated Prmt5 in granulosa cells using Sf1+/cre . The Sf1+/cre mouse expresses Cre recombinase in the adrenogonadal primordium at 10 dpc , the precursors for cortical cells in the adrenals and somatic cells in the gonads ( Bingham et al . , 2006; Huang and Yao , 2010 ) . We found that Prmt5flox/flox;Sf1+/cre female mice were infertile and that follicles were arrested at the secondary stage . The expression of WT1 was dramatically reduced , and the granulosa cells in secondary follicles began to express steroidogenic genes . Further studies revealed that PRMT5 regulates follicle development by facilitating Wt1 translation .
The expression of PRMT5 in ovarian granulosa cells was examined by immunofluorescence . As shown in Figure 1—figure supplement 1 , PRMT5 ( red ) was expressed in oocytes , but no PRMT5 signal was detected in the granulosa cells of primordial follicles ( A , A’ , white arrows ) . PRMT5 started to be expressed in granulosa cells of primary follicles ( B , B’ , white arrows ) and was continuously expressed in granulosa cells of secondary follicles ( C , C’ , white arrows ) , and antral follicles ( D , D’ , white arrows ) , but its expression decreased significantly in the corpus luteum ( E , E’ , white arrows ) . To test the functions of PRMT5 in granulosa cell development , we specifically deleted Prmt5 in granulosa cells by crossing Prmt5flox/flox mice with Sf1+/cre transgenic mice . In Prmt5flox/flox;Sf1+/cre female mice , PRMT5 expression was completely absent from granulosa cells ( Figure 1—figure supplement 2B , D , arrows ) , whereas the expression of PRMT5 in oocytes and interstitial cells was not affected , suggesting that Prmt5 was specifically deleted in granulosa cells . No obvious developmental abnormalities were observed in adult Prmt5flox/flox;Sf1+/cre mice ( Figure 1A ) . However , the female mice were infertile with atrophic ovaries ( Figure 1B , Figure 1—figure supplement 3 ) . The results of immunohistochemistry showed growing follicles at different stages in control ovaries at 2 months of age ( Figure 1C ) . In contrast , only a small number of follicles and few corpora lutea were observed in Prmt5flox/flox;Sf1+/cre mice ( Figure 1D ) . We further examined follicle development in Prmt5flox/flox;Sf1+/cre mice at different developmental stages . As shown in Figure 1 , a large number of growing follicles at the primary and secondary stages were observed in Prmt5flox/flox;Sf1+/cre mice ( F ) at 2 weeks , which was comparable to the situation in control ovaries ( E ) . Many follicles progressed to antral follicle stages in control mice at 3 weeks ( G ) , whereas the development of follicles in Prmt5flox/flox;Sf1+/cre mice was arrested at the secondary stage . Almost no antral follicle was observed ( Figure 1—figure supplement 4B ) and aberrant granulosa cells were evident ( Figure 1H ) . The defects in follicle development were more obvious at 4 and 5 weeks ( Figure 1J and L , Figure 1—figure supplement 4C and D ) . The number of granulosa cells around oocytes was dramatically reduced , and follicle structure was disrupted . To explore the underlying mechanism that caused the defects in follicle development in Prmt5flox/flox;Sf1+/cre mice , the expression of granulosa cell-specific genes was analyzed by immunohistochemistry . As shown in Figure 2 , FOXL2 protein was expressed in the granulosa cells of both control ( A , C ) and Prmt5flox/flox;Sf1+/cre mice ( B , D ) at P14 and P18 . WT1 protein was expressed in granulosa cells of primordial , primary , and secondary follicles in control mice at P14 and P18 ( E , E’ , G , G’ , arrows ) . WT1 was also detected in the follicles of Prmt5flox/flox;Sf1+/cre mice at P14 ( F , arrow ) . However , not all the granulosa cells were WT1-positive; some of them were WT1-negative ( F’ , arrowheads ) . The WT1 signal was almost completely absent from the majority of granulosa cells in Prmt5flox/flox;Sf1+/cre mice at P18 ( H , H’ ) ; very few granulosa cells were WT1-positive ( H’ , arrowheads ) . We also found that the granulosa cells in control ovaries were cuboidal and well-organized ( G’ , arrow ) . In contrast , the granulosa cells in Prmt5flox/flox;Sf1+/cre mice were flattened ( H’ , dashed line circle ) and were indistinguishable from surrounding stromal cells . The decreased WT1 protein expression in Prmt5-deficient granulosa cells was also confirmed by FOXL2/WT1 double staining ( Figure 2—figure supplement 1 ) . Wt1 plays a critical role in granulosa cell development , and mutation of Wt1 leads to pregranulosa cell-to-steroidogenic cell transformation ( Cen et al . , 2020; Chen et al . , 2017 ) . Therefore , we further examined the expression of steroidogenic genes in Prmt5-deficient granulosa cells at P18 . As shown in Figure 3 , in control ovaries , 3β-HSD ( 3β-hydroxysteroid dehydrogenase , also known as Hsd3B1 ) and CYP11A1 ( cytochrome P450 , family 11 , subfamily a , polypeptide 1 , also known as P450scc ) were expressed in theca-interstitial cells ( A , C , arrowheads ) . In addition to theca-interstitial cells , 3β-HSD ( B , green , arrows ) and CYP11A1 ( D , red , arrows ) were also detected in the granulosa cells of Prmt5flox/flox;Sf1+/cre mice . We also examined the expression of SF1 ( steroidogenic factor 1 , also known as NR5A1 ) , which is a key regulator of steroid hormone biosynthesis ( Ikeda et al . , 1993 ) . As expected , SF1 was expressed only in theca-interstitial cells of control ovaries ( E , red , arrowheads ) , whereas a high level of SF1 expression was detected in Prmt5-deficient granulosa cells ( F , red , arrows ) , suggesting that the identity of granulosa cells was changed . The follicle structure was destroyed as indicated by disorganized Laminin staining ( H , arrows ) . The proliferation and apoptosis of Prmt5-deficient granulosa cells were analyzed by BrdU incorporation and TUNEL assays . As shown in Figure 3—figure supplement 1 , the numbers of TUNEL-positive cells and BrdU-positive granulosa cells were not changed in Prmt5flox/flox;Sf1+/cre ovaries compared to control ovaries at P14 and P18 . To further confirm the above results , follicles were dissected from the ovaries of 2-week-old mice and cultured in vitro . As shown in Figure 3—figure supplement 2 , the histology of follicles from Prmt5flox/flox;Sf1+/cre mice was comparable to that of control follicles at D2 . Proliferation of granulosa cells in control follicles was observed at D4 , and the follicles developed to the preovulatory stage with multiple layers of granulosa cells after 9 days of culture ( A–C , G , H ) . The granulosa cells were detached from oocytes in Prmt5flox/flox;Sf1+/cre follicles at D4 ( E , and a magnified view in L ) , and no colonized granulosa cells were observed after 9 days of culture ( D–F , J , K ) . Most of the granulosa cells were attached to the culture dishes just like the interstitial cells , and denuded oocytes were observed after 3 days of culture ( E , F , J , K , and a magnified view in L ) . To further verify the differential expression of granulosa cell-specific and steroidogenic genes in Prmt5-deficient granulosa cells , granulosa cells were isolated at P18 , and gene expression was analyzed by western blot and real-time PCR analyses . As shown in Figure 4A , B , the protein levels of PRMT5 and its interacting partner MEP50 were decreased dramatically in Prmt5flox/flox;Sf1+/cre granulosa cells , as expected . The protein level of WT1 was significantly reduced in Prmt5-deficient granulosa cells . Surprisingly , the mRNA level of Wt1 was not changed in Prmt5-deficient granulosa cells ( C ) . FOXL2 expression was also decreased , but the difference was not significant . The expression of the steroidogenic genes CYP11A1 , StAR , and NR5A1 was significantly increased in Prmt5-deficient granulosa cells , consistent with the immunostaining results . The mRNA levels of these genes were also significantly increased ( C ) . Cyp19a1 is expressed in granulosa cells and its expression is increased in antral and preovulatory follicles ( Doody et al . , 1990; Stocco , 2008 ) . We found that the mRNA level of Cyp19a1 was also significantly increased in Prmt5-deficient granulosa cells ( C ) . We also examined the functions of PRMT5 by treating granulosa cells with the PRMT5-specific inhibitor EPZ015666 . The protein level of WT1 was significantly reduced after EZP015666 treatment , whereas the mRNA level was not changed . The expression of steroidogenic genes was significantly increased at both the protein and mRNA levels after EZP015666 treatment ( Figure 4D , E and F ) . These results were consistent with those in Prmt5-deficient granulosa cells , indicating that the effect of PRMT5 on granulosa cells was dependent on its methyltransferase activity . These results suggest that PRMT5 is required for maintenance of granulosa cell identity and that inactivation of this gene causes premature luteinization of granulosa cells . PRMT5 has been reported to regulate the translation of several genes in an IRES-dependent manner ( Gao et al . , 2017; Holmes et al . , 2019 ) . Internal ribosome entry sites ( IRESs ) are secondary structures in the 5′UTR that directly recruit the ribosome cap independently and initiate translation without cap binding and ribosome scanning ( Baird et al . , 2006; Coldwell et al . , 2000; Stoneley and Willis , 2004 ) . Wt1 5′UTR is 268 bp , GC-rich ( 68% ) and contains seven CUG codons and one AUG codon . These features usually act as strong barriers for ribosome scanning and conventional translation initiation . Translation initiation in a number of these mRNAs is achieved via IRES-mediated mechanisms ( Stoneley and Willis , 2004 ) . To test whether the Wt1 5′UTR contains an IRES element , we utilized a pRF dicistronic reporter construct in which upstream Renilla luciferase is translated cap-dependently , whereas downstream Firefly luciferase is not translated unless a functional IRES is present . A stable hairpin structure upstream of Renilla luciferase minimizes cap-dependent translation ( Coldwell et al . , 2000 , Figure 5A ) . Wt1 5′UTR was inserted into the intercistronic region between Renilla and Firefly luciferase ( named pRWT1F ) , and primary granulosa cells were transfected with pRF or pRWT1F . The Firefly/Renilla luciferase activity ratio was analyzed 24 hr later . As shown in Figure 5B , the Firefly/Renilla luciferase activity ratio was dramatically increased in pRWT1F-transfected cells compared to pRF-transfected cells . In contrast , the Firefly/Renilla luciferase activity ratio was not increased when Wt1 5′UTR was inserted in the reverse direction ( pRWT1-RevF; Figure 5B ) . The luciferase activity was dramatically increased with insertion of Ccnd1 5′UTR as a positive control , which has been reported to contain an IRES element in the 5′UTR ( Shi et al . , 2005 ) . These results suggest that Wt1 5′UTR probably contains an IRES element . To verify that Firefly luciferase protein was synthesized by translation of an intact dicistronic transcript instead of a monocistronic mRNA generated by cryptic splicing or promoter within the dicistronic gene ( Kunze et al . , 2016 ) , mRNA from pRF- or pRWT1F-transfected cells was treated with DNase , reverse-transcribed , and then amplified with primers binding to the 5′ end of Renilla luciferase and 3′ end of Firefly luciferase open reading frame spanning the whole transcript . Only one band was detected in both cells with the expected molecular weight ( Figure 5—figure supplement 1A ) . Moreover , qPCR analysis of Firefly and Renilla luciferase mRNA levels also showed that the Firefly/Renilla luciferase mRNA ratio was not different between pRF- and pRWT1F-transfected cells ( Figure 5—figure supplement 1B ) , further excluding the possibility that insertion of the Wt1 5′UTR into pRF generated a monocistronic Firefly ORF . To examine which part of Wt1 5′UTR contributes to its IRES activity , Wt1 5′UTR was divided into three fragments and respectively inserted into pRF construct ( pRWT1F –268 to –158 , pRWT1F –198 to –58 , pRWT1F –105 to 1 ) . These constructs were transfected into primary granulosa cells ( Figure 5C ) . No IRES activity was detected for middle fragment ( –198 to –58 ) . Although the luciferase activity of pRWT1F –268 to –158 and pRWT1F –105 to –1 was significantly increased compared with control pRF , they were much lower than that of the full-length 5′UTR , suggesting that the full length of Wt1 5~UTR is required for maximal IRES activity . To investigate the effect of PRMT5 on Wt1 IRES activity , granulosa cells were treated with EPZ015666 for 4 days , we found that Wt1 IRES activity was decreased significantly after EPZ015666 treatment ( Figure 5D ) . As a positive control , the IRES activity of Ccnd1 5′UTR was also significantly decreased after EPZ015666 treatment , which was consistent with previous study ( Gao et al . , 2017 , Figure 5D ) . We also checked Wt1 IRES activity in Prmt5flox/flox;Sf1+/cre granulosa cells . As expected , Wt1 IRES activity was significantly decreased in Prmt5-deficient granulosa cells compared with control granulosa cells ( Figure 5—figure supplement 1C ) . These results indicate that PRMT5 regulates Wt1 expression at the translational level by inducing its IRES activity in granulosa cells . IRES-mediated translation depends on IRES trans-acting factors ( ITAFs ) , which function by associating with the IRES and either facilitate the assembly of initiation complexes or alter the structure of the IRES ( Jo et al . , 2008; Kunze et al . , 2016 ) . Heterogeneous nuclear ribonucleoprotein A1 ( HnRNPA1 ) is a well-studied RNA binding protein that plays important roles in pre-mRNA and mRNA metabolism ( Dreyfuss et al . , 2002 ) . HnRNPA1 is also an ITAF that has been reported to regulate the IRES-dependent translation of many genes , such as Ccnd1 , Apaf1 ( Cammas et al . , 2007 ) , Myc ( Jo et al . , 2008 ) , Fgf2 ( Bonnal et al . , 2005 ) , and Xiap ( Lewis et al . , 2007; Wall and Lewis , 2017 ) . HnRNPA1 can be methylated by PRMT1 ( Wall and Lewis , 2017 ) or PRMT5 ( Gao et al . , 2017; Holmes et al . , 2019 ) , which regulates the ITAF activity of HnRNPA1 . To test whether PRMT5 interacts with HnRNPA1 in granulosa cells , coimmunoprecipitation experiments were conducted . We found that in control granulosa cells HnRNPA1 and PRMT5 were pulled down by antibody against the PRMT5 main binding partner MEP50; conversely , PRMT5 and MEP50 could be pulled down by the HnRNPA1 antibody ( Figure 5E ) . Although HnRNPA1 protein expression was not changed between control and Prmt5flox/flox;Sf1+/cre granulosa cells , the level of symmetric dimethylation of HnRNPA1 in Prmt5flox/flox;Sf1+/cre granulosa cells was significantly reduced compared with that in control granulosa cells ( Figure 5E ) . To test whether HnRNPA1 functions during PRMT5-mediated Wt1 translation , HnRNPA1 was knocked down in granulosa cells via siRNA transfection . Western blot analysis results showed that HnRNPA1 protein levels were significantly decreased with siRNA transfection ( Figure 6A , Figure 6—figure supplement 1 ) . We found that WT1 protein level was increased significantly in granulosa cells after knockdown of HnRNPA1 . The decreased WT1 expression in EPZ015666-treated granulosa cells was partially reversed by knockdown of HnRNPA1 ( Figure 6A , Figure 6—figure supplement 1 ) . The luciferase activity of pRWT1F was increased in granulosa cells with HnRNPA1 siRNA treatment and decreased in those with EPZ015666 treatment . The decreased luciferase activity in EPZ015666-treated granulosa cells was partially reversed by knockdown of HnRNPA1 ( Figure 6B ) . To further confirm the effect of HnRNPA1 on Wt1 IRES activity , HnRNPA1 was overexpressed in granulosa cells , and we found that Wt1 IRES activity was significantly decreased ( Figure 6D ) . These results indicated that as an ITAF the effect of HnRNPA1 on Wt1 IRES activity was repressive . There are five arginine residues in the HnRNPA1 glycine/arginine-rich ( GAR ) motif , which can be symmetrically or asymmetrically dimethylated by PRMT5 ( Gao et al . , 2017 ) or PRMT1 ( Rajpurohit et al . , 1994; Wall and Lewis , 2017 ) , respectively . R206 , R218 , R225 , and R232 are required for HnRNPA1 ITAF activity ( Gao et al . , 2017; Wall and Lewis , 2017 ) . To determine the role of HnRNPA1 arginine methylation in Wt1 IRES activity , the four arginine residues were mutated to lysines ( Figure 6C ) , and flag-tagged HnRNPA1 or mutant plasmids were cotransfected with pRWT1F into granulosa cells . We found that Wt1 IRES activity was further decreased in granulosa cells overexpressing mutant HnRNPA1 compared to those overexpressing wild-type HnRNPA1 ( Figure 6D ) . However , the difference in Wt1 IRES activity between cells overexpressing mutant HnRNPA1 and cells overexpressing wild-type HnRNPA1 disappeared when the granulosa cells were treated with EPZ015666 ( Figure 6D ) . These results indicate that the repressive function of HnRNPA1 on Wt1 IRES activity is inhibited by PRMT5-mediated arginine symmetric dimethylation . To test the interaction between HnRNPA1 and Wt1 mRNA , RNA immunoprecipitation was performed with HnRNPA1 antibody in primary granulosa cells . As shown in Figure 6E , Wt1 mRNA was pulled down by the HnRNPA1 antibody in granulosa cells . Next , granulosa cells were infected with flag-tagged wild-type or arginine-mutant HnRNPA1 adenovirus ( Figure 6F–I ) and RNA immunoprecipitation was conducted with a FLAG antibody . The results showed that mutation of arginines did not affect the interaction between HnRNPA1 and Wt1 mRNA ( Figure 6J ) . To test whether the upregulation of steroidogenic genes in Prmt5-deficient granulosa cells is due to downregulation of WT1 , granulosa cells from Prmt5flox/flox;Sf1+/cre mice were infected with control or GFP-tagged WT1-expressing adenovirus ( Figure 7A and B ) . Wt1 protein ( Figure 7D arrow , E ) and mRNA ( Figure 7C ) levels were dramatically increased in Prmt5-deficient granulosa cells after Wt1 overexpression . We found that the expression of steroidogenic genes was significantly decreased in these cells . These results suggest that the aberrant differentiation of Prmt5-deficient granulosa cells can be rescued by WT1 .
Protein arginine methylation is one of the most important epigenetic modifications and is involved in many cellular processes . In this study , we found that protein arginine methylation plays important roles in granulosa cell development . The development of ovarian follicles is a dynamic process . The granulosa cells in antral follicles express gonadotropin receptors . Before ovulation , granulosa cells begin to express steroidogenic enzymes that are necessary for progesterone and estradiol synthesis ( Irving-Rodgers et al . , 2004; Smith et al . , 2014 ) . In this study , we found that Prmt5-deficient granulosa cells began to express steroidogenic genes in secondary follicles and the upregulation of the steroidogenic genes in Prmt5-deficient granulosa cells was reversed by Wt1 overexpression , indicating that PRMT5 is required for preventing the premature differentiation of granulosa cells via regulation of WT1 expression . Coordinated interaction between granulosa cells and oocytes is required for successful follicle development and production of fertilizable oocytes . The premature luteinized granulosa cells will lose their structural and nutritional support for oocytes , which will lead to follicle growth arrest or atresia at early stages of folliculogenesis . Nuclear receptor Sf1 plays a critical role in the regulation of steroid hormone biosynthesis by inducing the expression of steroidogenic enzymes in steroidogenic cells ( Ikeda et al . , 1993 ) . Our previous study demonstrated that WT1 represses Sf1 expression by directly binding to the Sf1 promoter region and that inactivation of Wt1 causes upregulation of Sf1 , which in turn activates the steroidogenic program ( Chen et al . , 2017 ) . In the present study , the mRNA and protein levels of Sf1 were significantly upregulated after WT1 loss . Therefore , we speculate that the upregulation of steroidogenic genes in Prmt5-deficient granulosa cells is most likely due to the increased expression of Sf1 gene . As an important nuclear transcription factor , the function of WT1 in granulosa cell development has been investigated . However , the molecular mechanism that regulates the expression of Wt1 gene is unknown . In this study , we found that the expression of WT1 at the protein level was dramatically reduced in Prmt5-deficient granulosa cells , whereas the mRNA level was not changed , indicating that PRMT5 regulates Wt1 expression at the post-transcriptional level . In our mouse model , Prmt5 was inactivated in granulosa cells at the early embryonic stage . However , defects in follicle development were not observed until 2 weeks after birth . This outcome probably occurred because Prmt5 is not expressed in granulosa cells before the development of primary follicles ( Figure 1—figure supplement 1 ) . During the early stage , Wt1 expression is also maintained in pregranulosa cells . Therefore , we speculate that there must be other factor ( s ) involved in regulating Wt1 expression before primary follicle stage . More than 100 mRNAs in mammals contain IRES elements in their 5′UTRs ( Jaud et al . , 2019 ) , which are involved in various physiological processes , such as differentiation , cell cycle progression , apoptosis , and stress responses ( Godet et al . , 2019 ) . The 5′UTR sequence of Wt1 mRNA is highly conserved , with more than 85% homology among the sequences of 29 mammalian species . Our study indicates that the Wt1 5′UTR has IRES activity . HnRNPA1 belongs to the HnRNP family , which comprises at least 20 members associated with RNA processing , splicing , transport , and metabolism ( Godet et al . , 2019; Roy et al . , 2017 ) . As a main ITAF , HnRNPA1 either activates the translation of Fgf2 ( Bonnal et al . , 2005 ) , Srebp-1a ( Damiano et al . , 2013 ) , and Ccnd1 ( Shi et al . , 2005 ) or inhibits the translation of Xiap ( Lewis et al . , 2007 ) , Apaf ( Cammas et al . , 2007 ) , and Bcl-xl ( Bevilacqua et al . , 2010 ) . The underlying mechanism by which HnRNPA1 activates some IRESs but suppresses other IRESs is still unknown . HnRNPA1 may compete with other ITAFs for binding or may modify IRES structure and thus regulate IRES activity ( Cammas et al . , 2007; Lewis et al . , 2007 ) . It has been reported that the expression of several genes is regulated by PRMT5 at the protein level ( Gao et al . , 2017; Nicholas et al . , 2013 ) . Gao et al . reported that PRMT5 regulates IRES-dependent translation via methylation of HnRNPA1 in the 293T and MCF-7 cell lines . They found that HnRNPA1 activates the IRES-dependent translation and that methylation of HnRNPA1 facilitates the interaction of HnRNPA1 with IRES mRNA to promote translation ( Gao et al . , 2017 ) . In the present study , we found that Wt1 IRES activity was repressed by HnRNPA1 ( Figure 8A ) and that the repressive effect of HnRNPA1 was reversed by PRMT5-mediated arginine methylation; thus , Wt1 IRES-dependent translation was promoted by PRMT5 ( Figure 8B ) . The ITAF activity of HnRNPA1 can be regulated by post-translational modifications ( Godet et al . , 2019 ) . Phosphorylation of HnRNPA1 on serine 199 by Akt inhibits IRES-dependent translation of c-myc and cyclin D1 ( Jo et al . , 2008; Shi et al . , 2005 ) . Symmetric dimethylation of HnRNPA1 by PRMT5 enhances HnRNPA1 ITAF activity and promotes the translation of target mRNAs ( Gao et al . , 2017 ) . Asymmetric dimethylation of HnRNPA1 by PRMT1 inhibits its ITAF activity ( Wall and Lewis , 2017 ) . These results suggest that arginine methylation has different effects on the ITAF activity of HnRNPA1 according to different IRESs and cell contexts . Our study demonstrated that HnRNPA1 ITAF activity toward Wt1 mRNA was repressed by PRMT5-mediated arginine methylation . However , the affinity between HnRNPA1 and Wt1 mRNA was not affected after mutation of arginine residues , consistent with the findings of a previous study ( Wall and Lewis , 2017 ) . Therefore , the inhibition of HnRNPA1 ITAF activity by PRMT5 does not occur through changes in the binding of HnRNPA1 to Wt1 mRNA . Arginine–glycine–glycine ( RGG ) -motif region is also reported to be involved in mediating the interactions between homo- and heterotypic proteins . It is possible that arginine methylation of HnRNPA1 changes the interactions of HnRNPA1 and its protein partners , which affects the ITAF activity of HnRNPA1 ( Wall and Lewis , 2017 ) . The underlying mechanism needs further investigation . Epigenetic modification is involved in numerous cellular processes . However , the functions of epigenetic modification in granulosa cell development have not been well studied . In this study , we demonstrated that Prmt5 is required for maintenance of granulosa cell identity in follicle development and that inactivation of Prmt5 causes premature luteinization of granulosa cells . Our study also demonstrates that PRMT5 regulates WT1 expression at the translational level by methylating HnRNPA1 . This study provides very important information for better understanding the regulation of gonad somatic cell differentiation .
All animal experiments were carried out in accordance with the protocols approved by the Institutional Animal Care and Use Committee ( IACUC ) of the Institute of Zoology , Chinese Academy of Sciences ( CAS; SYXK 2018-0021 ) . All mice were maintained on a C57BL/6;129/SvEv mixed background . Prmt5flox/flox;Sf1+/cre female mice were obtained by crossing Prmt5flox/flox mice with Prmt5+/flox;Sf1+/cre mice . Prmt5flox/flox and Prmt5+/flox female mice were used as controls . For breeding experiment , both control and Prmt5flox/flox;Sf1+/cre female mice were crossed with wild-type male mice when they reached 8 weeks of age . Each pair was kept in a cage for 4 months . The number of pups delivered during this period was counted . The dicistronic construct pRF was a generous gift from Professor Anne Willis , University of Cambridge . pRWT1F , pRCCND1F , and pRWT1-RevF were constructed by inserting the mouse Wt1 5′UTR , human Ccnd1 5′UTR , or mouse Wt1 5′UTR in reverse orientation into EcoRI and NcoI sites of the pRF vector . Mouse Wt1 5′UTR and human Ccnd1 5′UTR sequence were amplified by PCR . The primers amplifying the whole transcript of pRF binding to the 5′ end of Renilla and 3′ end of Firefly ORF: pRF-F: GCCACCATGACTTCGAAAGTTTATGA; pRF-R: TTACACGGCGATCTTTCCGC . FLAG-tagged HnRNPA1 and mutant plasmids were generated by inserting the coding sequence and a mutant sequence of mouse HnRNPA1 , respectively , into NheI and BamHI sites of the pDC316-mCMV-ZsGreen-C-FLAG vector . Adenoviruses containing WT1 coding sequence , HnRNPA1 , or the mutant sequence were generated using the Gateway Expression System ( Invitrogen ) . The primers used for constructing the plasmids are as follows: Granulosa cells were isolated from mice at 16–18 days old . After mechanical dissection , ovaries were cut into several parts and incubated in PBS containing 1 mg/ml collagenase IV ( VETEC , V900893 ) in a water bath with circular agitation ( 85 rpm ) for 5 min at 37°C . Follicles were allowed to settle and washed in PBS . The supernatant were discarded . A second enzyme digestion was performed in PBS containing 1 mg/ml collagenase IV , 1 mg/ml hyaluronidase ( SIGMA , SIAL-H3506 ) , 0 . 25% Trypsin , and 1 mg/ml DNase I ( AppliChem , A37780500 ) for 15 min . FBS was added to stop the digestion and cell suspension was filtered through a 40 μm filter . Cells were centrifuged , washed , and then plated in 24-well plate in DMEM/F12 supplemented with 5% FBS . For EPZ015666 treatment , granulosa cells were incubated in the medium with the addition of 5 μM EPZ015666 ( MedChemExpress , HY-12727 ) for 4–5 days . When cells were approximately 70% confluent , granulosa cells were transfected with plasmids or infected with adenovirus according to the experiments . At the end of culture , cells were lysed for RT-qPCR , western blot analysis , or luciferase activity analysis using a dual luciferase reporter assay system ( Promega , E1910 ) . Control siRNA or siRNA to HnRNPA1 was purchased from ThermoFisher ( S67643 , S67644 ) and transfected into granulosa cells with Lipofectamine 3000 transfection reagent without P3000 . 48 hr later , pRF or pRWT1F were transfected and luciferase activities were measured the following day . Follicles were dissected and cultured as previously described ( Gao et al . , 2014 ) . Briefly , ovaries of 14-day-old mice were dissected aseptically using the beveled edges of two syringe needles . Follicles with 2–3 layers of granulosa cells , a centrally placed oocyte , an intact basal membrane , and attached theca cells were selected and cultured individually in 20 μl droplets of culture medium ( αMEM supplemented with 5% FBS , 1% ITS , and 100 mIU/ml recombinant FSH ) . The culture were maintained in 37°C and 5% CO2 in air . The medium was replaced every other day . The histology of the follicles was recorded daily under a microscope . Granulosa cells isolated from mice at 16–18 days old were cultured in 10 cm dishes and lysed with lysis buffer ( 50 mM Tris–HCl [pH 7 . 5] , 150 mM NaCl , 1 mM EDTA , 1% Nonidet P-40 ) supplemented with protease inhibitors cocktail ( Roche ) and 1 mM PMSF . 1 mg of protein were first pre-cleared with protein A/G agarose beads ( GE , 17-0618-01 , 17-5280-01 ) for 1 hr at 4°C , then incubated with HnRNPA1 antibody ( Abcam , ab5832 ) , or MEP50 antibody ( Abcam , ab154190 ) for 4 hr at 4°C . Then protein A and G agarose beads were added and incubated overnight . The immunoprecipitates were washed four times in lysis buffer supplemented with cocktail and PMSF , resolved in loading buffer , incubated for 5 min at 95°C , and then analyzed by western blotting . The antibodies used in western blotting include PRMT5 ( Millipore , 07-405 ) , SYM10 ( Millipore , 07-412 ) , HnRNPA1 ( Abcam , ab5832 ) , and MEP50 ( Abcam , ab154190 ) . Granulosa cells were washed with PBS , lysed with RIPA buffer ( 50 mM Tris–HCl [pH 7 . 5] , 150 mM NaCl , 1% NP-40 , 0 . 1% SDS , 1% sodium deoxycholate , 5 mM EDTA ) supplemented with protease inhibitors cocktail ( Roche ) and 1 mM PMSF . Equal amounts of total protein were separated by SDS/PAGE gels , transferred to nitrocellulose membrane , and probed with the primary antibodies . The images were captured with the ODYSSEY Sa Infrared Imaging System ( LI-COR Biosciences , Lincoln , NE ) . Densitometry was performed using ImageJ software . The protein expression was normalized to that of GAPDH . Blots are representative of three independent experiments . The antibodies used were PRMT5 ( Millipore , 07-405 ) , MEP50 ( Abcam , ab154190 ) , WT1 ( Abcam , ab89901 ) , FOXL2 ( Abcam , ab5096 ) , CYP11A1 ( Proteintech , 13363-1-AP ) , StAR ( Santa Cruz , sc-25806 ) , SF1 ( Proteintech , 18658-1-AP ) , and FLAG ( Sigma , F1804 ) . Granulosa cells were isolated from mice at 16–18 days old and cultured in 10 cm dishes . The cells were then lysed with RIP buffer ( 50 mM Tris-HCl [pH 7 . 5] , 150 mM NaCl , 5 mM EDTA , 1% NP-40 , 0 . 5% sodium deoxycholate ) supplemented with protease inhibitor cocktail and 200 U/ml RNase inhibitor . 5% of the cell lysate supernatants were used as the input and the remaining were incubated with 1 . 5 μg of IgG ( mouse , Santa Cruz , sc-2025 ) , HnRNPA1 antibody ( Abcam , ab5832 ) , or FLAG antibody ( Sigma , F1804 ) for 4 hr at 4°C . Then protein A and G agarose beads were added to immunoprecipitate the RNA/protein complex . The conjugated beads were thoroughly washed with lysis buffer ( 50 mM Tris–HCl [pH 7 . 5] , 500 mM NaCl , 5 mM EDTA , 1% NP-40 , 0 . 5% sodium deoxycholate ) supplemented with cocktail and 200 U/ml RNase inhibitor . Bound RNA was extracted using a RNeasy Kit and analyzed with RT-qPCR analysis . Total RNA was extracted using a RNeasy Kit ( Aidlab , RN28 ) in accordance with the manufacturer’s instructions . 1 µg of total RNA was used to synthesize first-strand cDNA ( Abm , G592 ) . cDNAs were diluted and used for the template for real-time SYBR Green assay . Gapdh was used as an endogenous control . All gene expression was quantified relative to Gapdh expression . The relative concentration of the candidate gene expression was calculated using the formula 2-ΔΔCT . Real-time RT-PCR primers are as follows: Immunohistochemistry procedures were performed as described previously ( Gao et al . , 2006 ) . Stained sections were examined with a Nikon microscope , and images were captured by a Nikon DS-Ri1 CCD camera . For immunofluorescence analysis , the 5 μm sections were incubated with 5% BSA in 0 . 3% Triton X-100 for 1 hr after rehydration and antigen retrieval . The sections were then incubated with the primary antibodies for 1 . 5 hr and the corresponding FITC-conjugated donkey anti-goat IgG ( 1:150 , Jackson ImmunoResearch , 705-095-147 ) and Cy3-conjugated donkey anti-rabbit IgG ( 1:300 , Jackson ImmunoResearch , 711-165-152 ) for 1 hr at room temperature . The following primary antibodies were used: WT1 ( Abcam , ab89901 ) , FOXL2 ( Abcam , ab5096 ) , CYP11A1 ( Proteintech , 13363-1-AP ) , SF1 ( Proteintech , 18658-1-AP ) , and 3β-HSD ( Santa Cruz , sc-30820 ) . After being washed three times in PBS , the nuclei were stained with DAPI . The sections were examined with a confocal laser scanning microscope ( Carl Zeiss Inc , Thornwood , NY ) . For follicle counting analysis , whole ovaries from control and Prmt5flox/flox;Sf1+/cre female mice at 2 , 3 , 4 , and 5 weeks of age were serially sectioned at 5 μm thickness ( n = 3/time point/genotype ) , and follicles were counted on every five sections . All experiments were repeated at least three times . 3–5 mice for each genotype at each time point were used for immunostaining or quantitative experiments . For immunostaining , one representative picture of similar results from 3 to 5 mice for each genotype at each time point is presented . The quantitative results are presented as the mean ± SEM . All granulosa cell culture experiments were repeated at least three times by using three different cell preparations . Statistical analyses were conducted using GraphPad Prism version 9 . 0 . 0 . Unpaired two-tailed Student’s t-tests were used for comparison between two groups . For three or more groups , data were analyzed using one-way ANOVA . p-Values<0 . 05 were considered to indicate significance . | Infertility in women can be caused by many factors , such as defects in the ovaries . An important part of the ovaries for fertility are internal structures called follicles , which house early forms of egg cells . A follicle grows and develops until the egg is finally released from the ovary into the fallopian tube , where the egg can then be fertilised . In the follicle , an egg is surrounded by other types of cells , such as granulosa cells . The egg and neighbouring cells must maintain healthy contacts with each other , otherwise the follicle can stop growing and developing , potentially causing infertility . The development of a follicle depends on an array of proteins . For example , the transcription factor WT1 controls protein levels by activating other genes and their proteins and is produced in high numbers by granulosa cells at the beginning of follicle development . Although WT1 levels dip towards the later stages of follicle development , insufficient levels can lead to defects . So far , it has been unclear how levels of WT1in granulose cells are regulated . Chen , Dong et al . studied mouse follicles to reveal more about the role of WT1 in follicle development . The researchers measured protein levels in mouse granulosa cells as the follicles developed , and discovered elevated levels of PRMT5 , a protein needed for egg cells to form and survive in the follicles . Blocking granulosa cells from producing PRMT5 led to abnormal follicles and infertility in mice . Moreover , mice that had been engineered to lack PRMT5 developed abnormal follicles , where the egg and surrounding granulosa cells were not attached to each other , and the granulosa cells had low levels of WT1 . Further experiments revealed that PRMT5 controlled WT1 levels by adding small molecules called methyl groups to another regulatory protein called HnRNPA1 . The addition of methyl groups to genes or their proteins is an important modification that takes place in many processes within a cell . Chen , Dong et al . reveal that this activity also plays a key role in maintaining healthy follicle development in mice , and that PRMT5 is necessary for controlling WT1 . Identifying all of the intricate mechanism involved in regulating follicle development is important for finding ways to combat infertility . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2021 | PRMT5 regulates ovarian follicle development by facilitating Wt1 translation |
IgE can trigger potent allergic responses , yet the mechanisms regulating IgE production are poorly understood . Here we reveal that IgE+ B cells are constrained by chronic activity of the IgE B cell receptor ( BCR ) . In the absence of cognate antigen , the IgE BCR promoted terminal differentiation of B cells into plasma cells ( PCs ) under cell culture conditions mimicking T cell help . This antigen-independent PC differentiation involved multiple IgE domains and Syk , CD19 , BLNK , Btk , and IRF4 . Disruption of BCR signaling in mice led to consistently exaggerated IgE+ germinal center ( GC ) B cell but variably increased PC responses . We were unable to confirm reports that the IgE BCR directly promoted intrinsic apoptosis . Instead , IgE+ GC B cells exhibited poor antigen presentation and prolonged cell cycles , suggesting reduced competition for T cell help . We propose that chronic BCR activity and access to T cell help play critical roles in regulating IgE responses .
Of all the antibody isotypes , immunoglobulin E ( IgE ) can elicit the most rapid immune responses in immediate-type hypersensitivity , contributing to the pathogenesis of numerous allergic diseases ( Gould et al . , 2003 ) . IgE-mediated hypersensitivity responses are typically localized to specific tissues such as the skin , nose , lung , or intestine , whereas systemic responses can result in life-threatening anaphylaxis . Only a small fraction of individuals with allergic diseases will experience anaphylaxis , however , suggesting that IgE responses are normally restricted . Indeed , IgE is the least abundant antibody isotype in serum . While the availability of IgE in serum is limited in part by a short half-life and binding to Fc receptors , the production of secreted IgE appears to be tightly regulated ( Geha et al . , 2003 ) . In order to understand the regulation of IgE production , recent attention has focused on IgE-expressing ( IgE+ ) B cells . Little had been known about these cells due to their low abundance and technical difficulties in their detection . These challenges have largely been overcome by the generation of fluorescent IgE reporter mice as well as improved technical methods ( He et al . , 2013; Talay et al . , 2012a; Wesemann et al . , 2011; Yang et al . , 2012 ) . Initial studies of IgE+ B cells in mice have revealed several key differences from B cells expressing IgG1 , the other major isotype induced in type 2 immune responses . IgE+ B cells appeared only transiently and at low frequencies in germinal centers ( GCs ) ( He et al . , 2013; Talay et al . , 2012b; Yang et al . , 2012 ) . These structures form during immune responses in lymphoid tissues and are major sites for antibody affinity maturation as well as the generation of long-lived plasma cells ( PCs ) and memory B cells ( Allen et al . , 2007a; Victora and Nussenzweig , 2012 ) . Consistent with the limited participation of IgE+ B cells in GCs , IgE+ responses typically exhibit reduced affinity maturation compared with IgG1+ responses , and most of the affinity maturation that does occur requires an IgG1+ B cell intermediate ( Erazo et al . , 2007; He et al . , 2013; Xiong et al . , 2012; Yang et al . , 2012 ) . In addition , we observed a relative paucity of long-lived IgE+ PCs ( Yang et al . , 2012 ) and other groups reported that memory IgE responses were largely initiated by non-IgE-expressing B cells ( He et al . , 2013; Katona et al . , 1991; Turqueti-Neves et al . , 2015 ) . Taken together , it appears that IgE+ B cells undergo an abortive GC phase that limits downstream IgE responses . In contrast , a larger proportion of IgE+ B cells were observed to have a PC phenotype compared with IgG1+ B cells in several studies ( note that here we use a broad definition of PCs to refer to all antibody secreting cells , including plasmablasts ) ( Erazo et al . , 2007; He et al . , 2013; Laffleur et al . , 2015; Yang et al . , 2012 ) . We reported that this observation could be recapitulated in cell culture of primary mouse B cells , suggesting the increased PC differentiation of IgE+ B cells was B cell intrinsic ( Yang et al . , 2012 ) . We proposed that the propensity of IgE+ B cells to undergo terminal differentiation into PCs may directly contribute to the low frequency and disappearance of IgE+ B cells from GCs ( Yang et al . , 2012 ) . Another group proposed that IgE+ GC B cells exhibit diminished BCR signaling and undergo increased rates of apoptosis ( He et al . , 2013 ) . Both intrinsic and extrinsic mechanisms could account for the distinct features of IgE+ B cells . A likely candidate for intrinsic regulation is the expression of the IgE B cell receptor ( BCR ) . Each isotype of BCR has a different constant region sequence which may confer different signaling capabilities . IgG BCRs can promote enhanced responses compared with IgM BCRs , most notably enhanced PC differentiation in recall responses ( Martin and Goodnow , 2002 ) . This is thought to be due , at least in part , to the extended intracellular cytoplasmic tail of the IgG BCR , compared with the short three amino acid sequence ( KVK ) in IgM ( Martin and Goodnow , 2002 ) . A conserved tyrosine motif in the cytoplasmic tail of the IgG BCR , which was reported to be primarily responsible for its differential signaling , is also present in the IgE BCR ( Engels et al . , 2009 ) . The cytoplasmic tail of the IgE BCR has also been implicated in promoting apoptosis through binding the mitochondrial protein Hax1 ( Laffleur et al . , 2015 ) . Here we show that the IgE BCR is a major determinant of IgE+ B cell fate . In the presence of T cell help signals , the IgE BCR promoted PC differentiation . This cell fate predisposition occurred in the absence of cognate antigen , whereas the IgG1 BCR promoted PC differentiation only in the presence of cognate antigen . Multiple domains of the BCRs were responsible for the difference between the IgE and IgG1 isotypes . The propensity of the IgE BCR to induce antigen-independent PC differentiation was associated with a weak , constitutive activity of the IgE BCR . Genetic or pharmacological perturbations in BCR signaling led to reduced PC differentiation of IgE+ B cells in the absence of cognate antigen in cell culture . In immunized mice , reductions in BCR signaling led to consistently increased IgE+ GC B cell responses with variable effects on PCs . The effects of BCR signaling on IgE+ GC B cell responses could not be explained by differential rates of apoptosis , as we found no evidence that the IgE BCR directly promotes apoptosis . Instead , BCR signaling slowed the cell cycle progression of IgE+ GC B cells , and low IgE BCR expression limited antigen uptake and presentation . Thus , IgE B cell responses are restrained by a predisposition toward early PC differentiation , prolonged cell cycles , and limited access to T cell help , leading to reduced affinity maturation and memory cell generation .
Several studies in mice have observed that a larger fraction of IgE+ cells have a PC phenotype compared with IgG1+ cells after immunization ( Erazo et al . , 2007; He et al . , 2013; Yang et al . , 2012 ) . We previously reported that this in vivo observation could be recapitulated in primary B cell cultures with anti-CD40 antibodies and IL-4 ( Yang et al . , 2012 ) , which promote class switch recombination ( CSR ) to IgE and IgG1 . In these B cell cultures , a substantially greater fraction of IgE+ cells were PCs compared with IgG1+ cells . This result was confirmed again here using CD138 ( Syndecan-1 ) as a marker of PCs ( Figure 1A and B ) . We found that a larger fraction of IgE+ cells than IgG1+ cells had a PC phenotype regardless of the concentration of anti-CD40 antibody , although we noted that stronger CD40 stimulation was inhibitory toward PC differentiation ( Figure 1A and B ) . We also previously established that the increased PC differentiation occurs in IgE+ B cells that have undergone the same number of cell divisions as IgG1+ B cells ( Yang et al . , 2012 ) . These observations established a strong correlation between CSR to IgE and PC differentiation , but the cause of this correlation was unknown . Notably , these culture conditions mimic T cell help but do not stimulate the BCR . We hypothesized that the IgE BCR itself promotes PC differentiation in the absence of cognate antigen . We sought to test this model by ectopically expressing the IgE BCR in primary B cells . 10 . 7554/eLife . 21238 . 003Figure 1 . The IgE BCR promotes antigen-independent PC differentiation of mouse B cells in culture . ( A and B ) Representative flow cytometry ( A ) and quantification ( B ) of PC differentiation ( CD138+ ) from wild-type B cells cultured for 4 d with IL-4 and anti-CD40 ( αCD40 ) . Cells were pre-gated as IgD–IgM– and with a broad B220+ gate . ( C ) Diagram of the retroviral construct for the ectopic expression of BCR heavy chains of various isotypes . A detailed vector diagram is provided in Figure 1—figure supplement 1 . ( D ) Frequency of PCs ( CD138+ ) among AID-deficient B cells ectopically expressing different BCR isotypes or Thy1 . 1 ( Thy1 ) as a control . Cells were cultured for a total of 4 d with anti-CD40 and IL-4 and were retrovirally transduced on d 1 . ( E ) Diagram of the retroviral construct for the ectopic expression of both heavy and light chains of BCRs specific for TNP . ( F ) Frequency of PCs ( CD138+ ) among AID-deficient B cells ectopically expressing different isotypes of TNP-specific BCRs in the absence or presence of TNP-OVA antigen ( Ag ) . Cells were cultured for 4d with anti-CD40 and IL-4 , were retrovirally transduced on d 1 , and antigen was added on d 2 . Similar data were obtained with B1-8flox/+ Cγ1Cre/+ B cells ( Figure 1—figure supplement 2 ) . Transduced cells for ( D ) and ( F ) were identified as Cerulean+ . Dots represent data points from individual experiments . Bars represent the mean . LTR , long terminal repeat , which on the 3’ end is self-inactivating ( white x ) ; ψ+ , extended packaging signal; VH and Vκ , coding sequences for the variable region of the heavy and light chains , respectively; CH and Cκ , coding sequences for constant regions of the heavy chain and light chain , respectively; M1 , M2 , coding sequences for the M1 and M2 exons; n . s . , not significant; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 ( t-tests with the Holm-Sidak correction for multiple comparisons ( B , F ) , or one-way ANOVA with Dunnett’s post-test comparing each heavy chain to the Thy1 . 1 control sample ( D ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 00310 . 7554/eLife . 21238 . 004Figure 1—figure supplement 1 . Vector for retroviral transduction of BCRs . Diagram of the retroviral vector for ectopic expression of different isotypes of BCR in primary B cells and B cell lines . LTR , long terminal repeat , which on the 3’ end is self-inactivating ( white x ) ; ψ+ , extended packaging signal; EF1 , EF1α promoter , AmpR , ampicillin resistantance gene . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 00410 . 7554/eLife . 21238 . 005Figure 1—figure supplement 2 . Antigen-independent versus antigen-dependent PC differentiation with B1-8flox/+ Cγ1Cre/+ B cells . Quantification of the frequency of PCs ( CD138+ ) in B1-8flox/+ Cγ1Cre/+ B cells ectopically expressing different isotypes of TNP-specific BCRs in the absence or presence of TNP-OVA antigen ( Ag ) . Transduced cells were identified as Cerulean+ . Dots represent data points from individual experiments , bars represent the mean . n . s . , not significant; *p<0 . 05; ****p<0 . 0001 ( t-tests with the Holm-Sidak correction for multiple comparisons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 005 In order to determine whether the IgE BCR directly promoted PC differentiation , we developed an approach to ectopically express the IgE BCR versus other BCR isotypes . We chose retroviral transduction , which is a robust method to express genes in primary B cells ( Wu et al . , 1987 ) . However , initial experiments with standard retroviral vectors were hampered by variable expression . We therefore engineered a retroviral vector using the EF1α promoter ( Figure 1C , Figure 1—figure supplement 1 ) , which we found gave much more uniform and robust expression . A heavy chain variable region specific for 2 , 4 , 6-trinitrophenyl ( TNP ) was linked to the heavy chain constant regions of each BCR isotype ( Figure 1C ) . As a reporter of transduction , Cerulean ( Rizzo et al . , 2004 ) , a derivative of cyan fluorescent protein , was placed upstream of the heavy chain , linked by a 2A sequence ( de Felipe et al . , 2006 ) which allows translationally-linked expression of multiple proteins from a single transcript ( Figure 1C and Figure 1—figure supplement 1 ) . With this optimized retroviral vector , we expressed BCRs of various isotypes in B cells from AID-deficient ( Aicda–/– ) mice that cannot undergo class switch recombination ( Muramatsu et al . , 2000 ) . B cell proliferation was again stimulated with anti-CD40 and IL-4 . Transduced cells were identified as Cerulean+ and PC differentiation was measured by the expression of the marker CD138 ( Syndecan-1 ) . Strikingly , among all isotypes tested , the IgE BCR promoted the highest frequency of PC differentiation in the absence of cognate antigen ( Figure 1D ) . The IgM and IgG1 BCRs did not promote PC differentiation in the absence of cognate antigen , whereas the IgD and IgA BCRs had intermediate effects ( Figure 1D ) . Since the IgG1 BCR has been reported to be much more efficient than IgM at inducing PC differentiation in vivo ( Martin and Goodnow , 2002 ) , we modified our system to test antigen-dependent effects . Specifically , we generated retroviral constructs encoding both the light chain and heavy chain from a TNP-specific monoclonal antibody , in which the heavy chain variable region was linked to the constant regions of various heavy chain isotypes ( Figure 1E ) . These TNP-specific BCR constructs were then transduced into AID-deficient B cells . Upon the addition of TNP-ovalbumin ( OVA ) , the TNP-specific IgG1 BCR promoted robust PC differentiation , similar to the IgE BCR ( Figure 1F ) . The addition of TNP-OVA also resulted in a more moderate increase in PC differentiation in cells transduced with TNP-specific IgM BCRs , with intermediate results with IgD BCRs ( Figure 1F ) . However , the addition of TNP-OVA did not cause a further increase in PC differentiation in cells transduced with TNP-specific IgE or IgA BCRs ( Figure 1F ) . Taken together , these data suggest that the IgE BCR is a strong inducer of PC differentiation in an antigen-independent manner , mimicking the behavior of an antigen-engaged IgG1 BCR . Having established that the IgE BCR promotes antigen-independent PC differentiation which can be recapitulated by ectopic expression , we sought to determine which domain ( s ) of IgE were responsible for this activity by domain swap experiments with IgG1 . Initial efforts focused on the intracellular cytoplasmic tail ( CT ) region , which is thought to be responsible for major differences in signaling among BCR isotypes ( Wienands and Engels , 2016 ) . Surprisingly , antigen-independent PC differentiation was unaffected when the IgE CT was replaced with that of IgG1 ( Figure 2A ) . The transmembrane ( TM ) region of the BCR is thought to mediate association with the signaling adapters Igα and Igβ ( Reth , 1992 ) , yet swapping the TM region also had no impact on antigen-independent PC differentiation ( Figure 2A ) . Membrane BCRs also contain a short extracellular segment proximal to the membrane that is unique to each isotype , known as the membrane Ig isotype-specific ( migis ) segment or the extracellular membrane proximal domain ( Davis et al . , 1991; Major et al . , 1996 ) . The expression of constructs in which the migis of IgE was swapped with that of IgG1 led to a profound loss of antigen-independent PC differentiation ( Figure 2A ) . In reverse swaps , in which these domains of IgE were introduced into IgG1 , the migis enhanced antigen-independent PC differentiation , but more striking results were seen when the IgE migis was combined with the IgE TM and CT , but not with the TM alone , suggesting that the CT may also be able to contribute to antigen-independent PC differentiation ( Figure 2A ) . Taken together , the IgE migis region appears to be necessary , but not sufficient , for antigen-independent PC differentiation mediated by the IgE BCR . Full antigen-independent activity of the IgE BCR required the IgE migis region to be combined with either the extracellular domains of IgE or the CT of IgE . 10 . 7554/eLife . 21238 . 006Figure 2 . Contribution of different domains of the IgE BCR to antigen-independent PC differentiation . Cells were retrovirally transduced with constructs in which the domains of IgE ( orange ) and IgG1 ( blue ) were swapped as illustrated ( see legend in upper-right for numbered constructs in ( B , D , and E ) ) . Primary B cells were cultured with anti-CD40 and IL-4 in ( A–C and F–G ) . ( A ) Frequency of PC differentiation ( CD138+ ) among transduced AID-deficient B cells . ( B ) Representative flow cytometry of the surface abundance of IgM in transduced AID-deficient B cells . Numbers in the plots show the gMFI of surface IgM . ( C ) Inverse correlation of cell surface IgM with PC differentiation ( CD138+ ) in transduced AID-deficient B cells . Each dot represents the results from B cells transduced with a distinct chimeric BCR from the constructs shown in ( A ) ; dots in orange represent chimeric BCRs with the migis derived from IgE; dots in blue represent chimeric BCRs with the migis derived from IgG1 . ( D ) Representative flow cytometry of surface BCR ( λ light chain ) expression on transduced J558L cells . The lower panels depict cells that had been stably transduced with Cd79a ( Igα ) . ( E ) Quantification of surface BCR ( λ light chain ) expression on transduced ( Cerulean+ ) J558L cells ( Cd79a– ) . ( F and G ) Frequency of PC differentiation ( CD138+ ) among transduced B1-8flox/+ Cγ1Cre/+ B cells . In ( G ) the individual CH domains of IgE and IgG1 were swapped as illustrated . Surface BCR and Cerulean reporter expression are provided in Figure 2—figure supplement 1 . Transduced cells were identified as Cerulean+ ( A–E ) with the addition of IgM–IgD– ( F–G ) . Dots represent data points from individual experiments , except in ( C ) where the data are from a single experiment representative of three independent experiments . Bars represent the mean . VH , coding sequence for the variable region of the heavy chain; CH , coding sequence for the constant region of the heavy chain; TM , transmembrane region; CT , cytoplasmic tail; gMFI , geometric mean fluorescence intensity; n . s . , not significant . **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 ( for each group of related constructs , one-way ANOVA with Dunnett’s post-test comparing each construct to the leftmost parent construct ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 00610 . 7554/eLife . 21238 . 007Figure 2—figure supplement 1 . Surface expression of BCRs on naturally class-switched versus transduced primary B cells . B cells were cultured for 4 d with anti-CD40 and IL-4 . ( A ) Representative histograms of surface BCR expression measured by isotype-specific antibodies ( surface IgE , surface IgG1 ) or anti-κ light chain ( surface Igκ ) on wild-type B cells that underwent natural CSR to IgE or IgG1 ( class-switched ) versus on B1-8flox/+ Cγ1Cre/+ B cells retrovirally transduced with IgE or IgG1 . Cells were pre-gated as IgM–IgD– and transduced cells were identified as Cerulean+ . ( B ) Representative flow cytometry of B1-8flox/+ Cγ1Cre/+ B cells retrovirally transduced with IgE or IgG1 BCRs . Cells were pre-gated as IgM–IgD–Cerulean+ . ( C ) Quantification of surface BCR expression measured with isotype-specific antibodies ( upper panels ) and Cerulean fluorescence intensity ( lower panels ) . AID-deficient B cells were retrovirally transduced with the indicated constructs ( related to Figure 2A ) . Dots represent data points from individual experiments , bars represent the geometric mean fluorescence intensity ( gMFI ) . VH , coding sequence for the variable region of the heavy chain; CH , coding sequence for the constant region of the heavy chain; TM , transmembrane region; CT , cytoplasmic tail . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 007 In the course of these experiments , we also measured the surface expression of IgM , which remains genetically encoded in the AID-deficient B cells that we transduced with IgE versus IgG1 BCRs . We observed that IgM was downmodulated in B cells transduced with IgE ( Figure 2B , construct 1 ) but not with IgG1 ( Figure 2B , construct 3 ) . The IgM downmodulation was dependent on the migis region , as revealed by transducing constructs with the migis regions swapped ( Figure 2B , constructs 2 and 4 ) . Indeed , in cells transduced with each of the constructs shown in Figure 2A , all of the constructs that contained the IgE migis resulted in downmodulation of surface IgM , whereas the constructs that contained the IgG1 migis did not ( Figure 2C ) . We hypothesized that this IgM downmodulation might reflect competition of the transduced BCRs versus IgM for binding to Igα . It was reported that similar to IgM , the IgE BCR depends on binding to Igα for export from the endoplasmic reticulum , whereas the IgG BCR does not ( Venkitaraman et al . , 1991 ) . This difference among isotypes had been attributed to the TM domain ( Reth , 1992; Venkitaraman et al . , 1991 ) , but we considered whether the migis , which is proximal to the membrane , could also be involved in Igα association . We transduced the IgE versus IgG1 BCRs , or constructs in which the migis regions were swapped , into J558L cells , which lack Igα ( Hombach et al . , 1990 ) . For comparison , we transduced the BCRs into J558L cells in which we had stably transduced Igα . Remarkably , the IgE migis made BCR surface expression completely dependent on Igα , whereas the IgG1 migis permitted substantial surface BCR localization in the absence of Igα ( Figure 2D and E ) . These data indicate that the migis region is a major site of interaction with Igα . In order to determine whether residual IgM expression in AID-deficient B cells was affecting our analysis of IgE versus IgG1 BCR domains , we repeated our experiments in B cells in which the pre-existing IgM BCR is deleted , analogous to natural CSR . Specifically , we made use of mice with a loxP-flanked B1-8 Ig heavy chain variable region allele ( B1-8flox ) , which could be deleted with Cre recombinase ( Lam et al . , 1997 ) . These mice were bred to mice carrying Cγ1 ( Ighg1 ) -Cre ( Casola et al . , 2006 ) , which is efficiently induced by anti-CD40 and IL-4 , resulting in Cre-mediated deletion of the existing B1-8flox BCRs . We retrovirally transduced these cells with new BCRs close in time to the deletion of the existing BCR , mimicking natural CSR . Ectopic expression of BCRs of different isotypes in B1-8flox/+ Cγ1Cre/+ B cells gave similar results to AID-deficient B cells , with IgE promoting a high frequency of antigen-independent PC differentiation , whereas IgG1 promoted antigen-dependent PC differentiation ( Figure 1—figure supplement 2 ) . In domain swap experiments in B1-8flox/+ Cγ1Cre/+ cells , the IgE migis region still contributed to antigen-independent PC differentiation , but played a less prominent role than in the AID-deficient B cells , presumably since no IgM was present to compete for Igα ( Figure 2F ) . Specifically , in IgE BCR constructs with the IgG1 migis , antigen-independent PC differentiation was reduced compared with constructs with the IgE migis , but was still elevated compared with the full IgG1 BCR , suggesting only a partial requirement of the migis region for antigen-independent PC differentiation . In reverse domain swaps in which regions of the IgG1 BCR were substituted with their counterparts in IgE , we observed a synergistic contribution of both the IgE migis and IgE CT to antigen-independent PC differentiation ( Figure 2F ) . However , when the IgG1 extracellular domains were coupled with the IgE migis , IgE TM , and IgE CT , the frequency of antigen-independent PC differentiation was intermediate between that of the full IgE BCR and IgG1 BCRs ( Figure 2F ) , again suggesting a contribution of the IgE extracellular domains , which were further explored below . Taken together , these data indicated that the IgE migis and CT both contributed to , but could not fully account for , the antigen-independent activity of the IgE BCR . We therefore tested the contribution of the extracellular domains of the IgE BCR to antigen-independent PC differentiation . The IgE BCR has four extracellular constant region domains ( CH1-4 ) , whereas the IgG1 BCR has three domains ( CH1-3 ) , with the second domain ( CH2 ) of IgE replaced by a hinge in IgG1 ( Gould et al . , 2003 ) . As expected , in the context of the IgE migis and CT , which we established above were major contributors to antigen-independent PC differentiation , there was no significant effect of swapping the extracellular domains , although there was a trend suggesting a contribution of the IgE CH2 and CH3 domains ( Figure 2G ) . We therefore considered whether a contribution of the extracellular domains could be further revealed in hybrid constructs containing the IgG1 CT , to remove the contribution of the IgE CT . Indeed , both the IgE CH2 and IgE CH3 reproducibly contributed to antigen-independent PC differentiation , as revealed by transducing hybrid constructs in which these domains had been swapped with the IgG1 hinge and CH2 , respectively ( Figure 2G ) . Thus , multiple parts of the IgE molecule , specifically the CH2 , CH3 , migis , and CT , all contribute to antigen-independent PC differentiation , making this BCR distinct from the IgG1 BCR . In order to further validate the results from our isotype and domain swap BCR comparisons , we measured the surface expression of the ectopically-expressed BCRs . The surface expression of the transduced IgE and IgG1 BCRs were equivalent to each other , as measured with an antibody to the light chain which pairs with these heavy chains ( Figure 2—figure supplement 1A ) . We also compared the surface expression of the transduced BCRs with the endogenous BCRs in normal B cells that had been induced to undergo natural CSR to IgE versus IgG1 . We observed that the surface expression of the transduced BCRs was overlapping with the surface expression of normal endogenous IgE BCRs , but less than that of normal endogenous IgG1 BCRs ( Figure 2—figure supplement 1A ) . This difference was due to the fact that membrane IgE normally has lower expression than membrane IgG1 ( Figure 2—figure supplement 1A ) , as previously reported ( He et al . , 2013 ) . Therefore , in the transduction system we did not ‘overexpress’ the BCRs but rather achieved a surface abundance similar to normal membrane IgE . We also noted that the induction of PC differentiation by the IgE BCR occurred over the entire range of surface BCR expression , indicating that small changes in surface expression would be unlikely to impact our results ( Figure 2—figure supplement 1B ) . All domain swap constructs had equivalent expression of the Cerulean reporter and achieved measurable surface IgE and/or IgG1 expression within approximately a 4-fold range ( Figure 2—figure supplement 1C ) . We therefore conclude that our system allowed a fair comparison of IgE versus IgG1 BCR domains for the ability to promote antigen-independent PC differentiation . We next sought to determine whether the antigen-independent PC differentiation mediated by IgE BCR was due to antigen-independent BCR signaling that differed from the IgG1 BCR . Initial attempts to look at phosphorylation of the downstream signaling adapters such as Syk , Btk , Erk , and Akt , by phosflow failed to show striking differences between IgE+ and IgG1+ B cells ( data not shown ) , presumably because many of these phosphorylation events are transient and can only be observed within minutes of strong acute stimulation , whereas the antigen-independent activity may be weaker and constitutive . We therefore considered cumulative readouts of BCR activity . We found that a larger fraction of IgE+ B cells than IgG1+ B cells expressed the activation marker CD69 ( Figure 3A ) , which was particularly apparent with low concentrations of anti-CD40 antibody , since strong CD40 stimulation could itself promote CD69 upregulation ( data not shown ) . Recently , it has been reported that Nur77 upregulation is a readout that is very sensitive to antigen-receptor signaling but only weakly induced by CD40 stimulation ( Zikherman et al . , 2012 ) . IgE+ B cells had higher constitutive Nur77 expression than IgG1+ B cells , as revealed by a Nur77-GFP reporter ( Figure 3B ) . The elevated CD69 and Nur77 expression suggest that IgE+ B cells have higher constitutive BCR activity than IgG1+ B cells . 10 . 7554/eLife . 21238 . 008Figure 3 . The IgE BCR exhibits differential constitutive activity compared with the IgG1 BCR . ( A and B ) Primary B cells were cultured for 4 d with IL-4 and anti-CD40 followed by flow cytometric evaluation of IgE+ and IgG1+ cells as in Figure 1A . Representative flow cytometry ( left ) and quantification ( right ) of the frequency of cells that were CD69+ ( A ) or Nur77-GFP+ ( B ) . Dots represent individual samples pooled from four ( A ) or five ( B ) experiments . ( C–H ) TIRF microscopy of J558L cells that were transduced with the IgE or IgG1 BCRs together with Igα-YFP ( ‘BCR’ ) and stained to show the plasma membrane ( PM ) . ( C and D ) . Representative TIRF microcopy images of single cells ( C ) and individual BCR clusters ( D ) . Scale bars , 1 . 5 µm . ( E and F ) Quantification of the fluorescence intensity ( E ) and size ( F ) of BCR microclusters . Dots indicate individual measurements . ( G and H ) Characterization of the Brownian diffusion coefficient of IgE and IgG1 BCRs by single molecule tracking , displayed as mean squared displacement ( MSD ) versus time ( G ) and cumulative probability distribution ( H ) plots . Bars show the mean ( A , B , and F ) or the geometric mean ( E ) . Error bars ( G ) indicate the SEM . **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 ( Mann-Whitney U-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 008 To gain further insights into the constitutive activity of the IgE versus IgG1 BCRs , we imaged these BCRs on the plasma membrane of quiescent B cells by total internal reflection fluorescence ( TIRF ) microscopy . BCR clustering on the surface of B cells is thought to be important for BCR signaling after antigen activation as well as for tonic BCR signaling ( Pierce and Liu , 2010 ) , which is essential for B cell survival ( Kraus et al . , 2004; Lam et al . , 1997 ) . We visualized BCR clustering through yellow fluorescent protein ( YFP ) -tagged Igα , which was expressed in J558L cells together with either IgE or IgG1 BCRs . J558L cells expressing similar amounts of IgE and IgG1 both showed high fluorescence intensity BCR cluster spots on the plasma membrane ( Figure 3C ) , consistent with previous reports that a minor fraction of BCRs are clustered to support tonic signaling in resting B cells ( Pierce and Liu , 2010 ) . However , some of the IgE BCR clusters were noticeably higher in fluorescence intensity ( Figures 3C , D and E ) and tended to be larger in size ( Figure 3F ) than the IgG1 BCR clusters . These results suggest that IgE BCRs intrinsically form more prominent clusters compared with IgG1 BCRs in quiescent B cells . The mobility feature of BCRs is another dimension which may indicate the status of receptors and downstream signaling . We used single molecule tracking to quantify the diffusion rate of IgE or IgG1 BCR molecules on the surface of J558L cells . This tracking analysis illuminated that the Brownian diffusion of IgE BCRs was more confined compared to that of IgG1 BCRs ( Figure 3G ) . Moreover , the short-range diffusion coefficients of single BCR molecules were processed and derived into a cumulative distribution probability plot , which confirmed that IgE BCRs have decreased Brownian diffusion coefficients compared with IgG1 BCRs ( Figure 3H ) . Taken together , these data suggest that IgE BCRs form more prominent clusters with reduced mobility compared with IgG1 BCRs in the absence of cognate antigen , providing further evidence that these BCRs differ in their constitutive activity . To determine whether BCR signaling was actually mediating the antigen-independent PC differentiation , we tested whether genetic or pharmacological disruption of BCR signaling pathways would affect PC differentiation in the cell culture assay . Treatment of the primary B cell cultures with ibrutinib to inhibit Btk , an important kinase in BCR signaling , led to reduced antigen-independent PC differentiation ( Figure 4A ) , while cell viability and CSR were maintained ( data not shown ) . Inhibitors of Syk and PI3Ks , which are critical kinases in BCR signal transduction , also partially inhibited PC differentiation but led to a marked loss of cell viability and thus were not further evaluated ( data not shown ) . To test the role of Syk by a genetic approach without disrupting B cell development , we generated Syk heterozygous B cells in vitro by culturing B cells from mice carrying a single loxP-flanked allele of Syk ( Sykflox/+ ) ( Saijo et al . , 2003 ) and Cγ1-Cre , which is induced in B cells cultured with anti-CD40 and IL-4 ( Casola et al . , 2006 ) . Syk heterozygosity led to reduced PC differentiation in the absence of antigen ( Figure 4B ) . The BCR co-receptor CD19 has been implicated in tonic BCR signaling ( Mattila et al . , 2013 ) , as has one of its major targets PI3K ( Srinivasan et al . , 2009 ) . Strikingly , antigen-independent PC differentiation was completely abrogated in CD19-deficient B cells ( Figure 4C ) . In contrast , the BCR signaling adapter BLNK ( BASH , SLP-65 ) only partially contributed to antigen-independent PC differentiation , with a two-fold reduction observed in BLNK-deficient B cells ( Figure 4D ) . These results suggest that antigen-independent PC differentiation has a differential reliance on particular BCR signaling pathways . Taken together , these data in general demonstrate that BCR signaling is needed for antigen-independent PC differentiation , providing further evidence that this is mediated by constitutive activity of the IgE BCR . 10 . 7554/eLife . 21238 . 009Figure 4 . Antigen-independent PC differentiation mediated by the IgE BCR is sensitive to perturbations in BCR signaling . B cells were cultured with IL-4 and anti-CD40 for 4 d . ( A–D ) Representative flow cytometry ( left ) and quantification ( right ) of PC differentiation ( CD138+ ) among B cells that were treated with DMSO solvent control versus 12 . 5 nM ibrutinib ( Ib ) ( A ) , from control ( Ctrl ) Syk+/+ Cγ1Cre/+ versus Sykflox/+ Cγ1Cre/+ ( Syk het ) mice ( B ) , from wild-type ( WT ) control versus Cd19Cre/Cre ( Cd19 ko ) mice ( C ) , or from wild-type ( WT ) control versus Blnk–/– ( Blnk ko ) mice ( D ) . Cells were gated as in Figure 1A . See also Figure 4—figure supplement 1 . ( E ) Quantification of the frequency of PCs ( CD138+ ) among B1-8flox/+ Cγ1Cre/+ B cells retrovirally transduced with TNP-specific IgE or IgG1 BCRs . Ibrutinib ( Ib ) was added immediately after spinfection ( d 1 ) , antigen ( TNP-OVA ) was added on d 2 , and cells were analyzed on d 4 . Transduced cells were identified as IgM–IgD–Cerulean+ . ( F ) Flow cytometry of GFP expression in B1-8i , Nur77-GFP B cells . 12 . 5 nM ibrutinib ( Ib ) was added on d 2 and then the cognate antigen NP-APC ( Ag ) was added on d 3 , and cells were analyzed on d 4 with further staining on ice with NP-APC to detect antigen-specific cells . Data are representative of two experiments . Dots represent data points from individual experiments . Bars represent the mean . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 ( t-tests with the Holm-Sidak correction for multiple comparisons ( A–D ) , one-way ANOVA followed by Dunnett’s post-test ( E ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 00910 . 7554/eLife . 21238 . 010Figure 4—figure supplement 1 . IRF-4 contributes to antigen-independent PC differentiation mediated by the IgE BCR in cell culture . Representative flow cytometry ( A ) and quantification ( B ) of PC differentiation ( CD138+ ) of control Irf4+/+ Cγ1Cre/+ ( Ctrl ) versus Irf4flox/+ Cγ1Cre/+ ( Irf4 het ) B cells cultured for 4 d with IL-4 and anti-CD40 . Cells were gated as in Figure 1A . Dots represent data points from separate experiments , bars represent the mean . *p<0 . 05 , **p<0 . 01 , ( t-tests with the Holm-Sidak correction for multiple comparisons ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 010 To further evaluate the constitutive activity of the IgE BCR , we compared the effects of perturbing BCR signaling on antigen-independent versus antigen-dependent PC differentiation . With our retroviral transduction system described above , we ectopically expressed TNP-specific light chains together with TNP-specific heavy chains coupled to IgE versus IgG1 constant regions ( with the construct shown in Figure 1E ) . We then treated cells with ibrutinib in order to inhibit Btk , prior to antigen stimulation with TNP-OVA . In the absence of TNP-OVA , ibrutinib treatment reduced antigen-independent PC differentiation mediated by the transduced BCRs , as we had previously observed in normal primary B cells that had undergone natural class switch recombination to IgE and IgG1 ( Figure 4E ) . Interestingly , however , when we added TNP-OVA , antigen-dependent PC differentiation was not significantly affected by ibrutinib treatment ( Figure 4E ) . To further evaluate the effects of Btk inhibition on constitutive versus antigen-dependent BCR signals , we used the Nur77-GFP reporter to measure BCR signaling activity in B cells carrying the B1-8 Ig heavy chain variable region knock-in specific for 4-hydroxy-3-nitrophenylacetyl ( NP ) when paired with λ light chains . Consistent with our previous results , in the absence of antigen , IgE+ B cells exhibited higher Nur77-GFP expression than IgG1+ B cells ( Figure 4F ) . The addition of cognate antigen ( NP-OVA ) resulted in much stronger GFP expression in antigen-specific B cells of both isotypes ( Figure 4F ) . Ibrutinib treatment abrogated Nur77-GFP expression in the absence of antigen , whereas ibrutinib treatment had less pronounced effects on Nur77-GFP expression in the presence of antigen ( Figure 4F ) . Taken together , these data indicate that the constitutive activity of the IgE BCR is weaker than antigen-dependent BCR stimulation and is more sensitive to pharmacological inhibition . Based on our above findings that the IgE BCR has a weak but constitutive activity that is distinct from the IgG1 BCR , we anticipated that perturbing BCR signaling in vivo might have differential effects on IgE versus IgG1 responses . After immunization , BLNK-deficient mice showed a striking increase in IgE+ B cell frequencies within GCs , compared with no change in IgG1+ B cell frequencies , in the context of relatively normal total GC B cell numbers ( Figure 5A and Figure 5—figure supplement 1A ) . BLNK-deficient mice also had an increase in IgE+ PCs , but not IgG1+ PCs ( Figure 5A ) . This result differed from cell culture , which we consider in depth below ( see the Discussion section ) . While CD19-deficient mice are defective in T-dependent immune responses in vivo ( Rickert et al . , 1995 ) and could not be studied , we tested the role of CD19 by immunizing mice heterozygous for Cd19 ( Rickert et al . , 1995 ) . The two-fold reduction in Cd19 in these mice also resulted in a selective increase in the frequency of IgE+ B cells in GCs , but did not affect the frequency of IgG1+ B cells nor total GC B cell numbers ( Figure 5A and Figure 5—figure supplement 1A ) . Cd19 heterozygous mice also showed a modest elevation in IgE+ PCs ( Figure 5A ) . Since both BLNK and CD19 are involved in B cell development , which could potentially have an indirect effect on IgE responses , we also tested the effects of transient pharmacological inhibition of BCR signaling . Wild-type mice were immunized and then 6 d later , at the beginning of the GC and PC responses , were treated with ibrutinib to inhibit Btk for 3 d . This inhibition of BCR signaling during an ongoing GC response resulted in a general two-fold reduction in GC cell numbers ( Figure 5—figure supplement 1B ) as previously reported ( Mueller et al . , 2015 ) , but there was again a selective increase in the frequency of IgE+ B cells in GCs , whereas the frequency of IgG1+ GC B cells was unaffected ( Figure 5B ) . A trend toward an increase in IgE+ PCs was also observed in ibrutinib-treated mice , although this did not reach statistical significance ( Figure 5B ) . Since ibrutinib can also inhibit Itk ( Dubovsky et al . , 2013 ) , which has been implicated in T cell regulation of IgE responses ( Felices et al . , 2009 ) , we sought to test the effects of BCR signaling perturbation specifically in activated B cells in mice with normal B cell development . In Sykflox/+ Cγ1Cre/+ mice , two functional copies of the Syk gene are present in all cells except activated B cells , which express Cγ1-Cre and become Syk heterozygous . This two-fold decrease in Syk specifically in activated B cells also resulted in a selective increase in the frequency of IgE+ B cells within GCs , whereas IgG1+ B cell frequencies were not significantly affected , in the context of slightly elevated total GC B cell numbers ( Figure 5C and Figure 5—figure supplement 1C ) . IgE+ PCs , but not IgG1+ PCs , were elevated as well when activated B cells were Syk heterozygous ( Figure 5C ) . Thus , with four different BCR signaling perturbations , IgE+ B cell frequencies were selectively elevated in GCs , indicating that in general , BCR signaling negatively regulates IgE+ GC B cell responses . The effects of these signaling perturbations on IgE+ PC numbers were variable , possibly reflecting differential requirements for these signaling adapters in antigen-independent versus antigen-dependent PC differentiation ( see the Discussion section ) . 10 . 7554/eLife . 21238 . 011Figure 5 . BCR signaling negatively regulates in vivo IgE+ responses . Mice were immunized subcutaneously with NP-CGG in alum adjuvant and draining lymph nodes were analyzed by flow cytometry 9 d later . ( A to D ) Quantification of frequency of IgE+ and IgG1+ cells among GC B cells and the total number of IgE+ and IgG1+ PCs in wild-type ( WT ) versus Blnk–/– ( Blnk ko ) versus Cd19Cre/+ ( Cd19 het ) mice ( A ) , Captex 355 vehicle ( Veh ) versus ibrutinib ( Ib ) -treated mice ( B ) , control Syk+/+ Cγ1Cre/+ ( Ctrl ) versus Sykflox/+ Cγ1Cre/+ ( Syk het ) mice ( C ) , control Irf4+/+ Cγ1Cre/+ ( Ctrl ) versus Irf4flox/+ Cγ1Cre/+ ( Irf4 het ) mice ( D ) . The total number of lymph node cells and GC B cells are provided in Figure 5—figure supplement 1 . An analysis of somatic hypermutation related to ( C ) is provided in Figure 5—figure supplement 2 . PCs were CD138+B220lo-intCD38loIgDlo and intracellular IgEhi or total IgG1hi . GC B cells were CD138–B220hiPNAhiCD38loIgDlo and intracellular IgEint or total IgG1int . Dots represent individual mice . Bars represent the mean ( % of GC B cells ) or geometric mean ( cell number ) . Data are representative of two ( A ) , three ( C ) , and four ( B ) independent experiments . n . s . , not significant; **p<0 . 01; ***p<0 . 001; ****p<0 . 0001 ( one-way ANOVA followed by Dunnett’s post-test ( A ) , t-tests with the Holm-Sidak correction for multiple comparisons ( B–D ) ; the numbers of PCs were log transformed for all statistical tests ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 01110 . 7554/eLife . 21238 . 012Figure 5—figure supplement 1 . Number of total lymph node cells and GC B cells in mice with perturbations in BCR signaling . Mice were immunized subcutaneously with NP-CGG in alum adjuvant and draining lymph nodes were analyzed by flow cytometry 9 d later . ( A to D ) Quantification of the total LN cellularity ( left ) and total number ( # ) of GC B cells ( right ) in wild-type ( WT ) versus Blnk–/– ( Blnk ko ) versus Cd19Cre/+ ( Cd19 het ) mice ( A ) , Captex 355 vehicle ( Veh ) versus ibrutinib ( Ib ) -treated mice ( B ) , control Syk+/+ Cγ1Cre/+ ( Ctrl ) versus Sykflox/+ Cγ1Cre/+ ( Syk het ) mice ( C ) , control Irf4+/+ Cγ1Cre/+ ( Ctrl ) versus Irf4flox/+ Cγ1 Cre/+ ( Irf4 het ) mice ( D ) . GC B cells were CD138–B220hiPNAhiCD38loIgDlo . Dots represent individual mice . Bars represent the median ( LN cellularity ) or geometric mean ( GC cell number ) . n . s . , not significant; *p<0 . 05 , ( one-way ANOVA followed by Dunnett’s post-test ( A ) , t-tests with the Holm-Sidak correction for multiple comparisons ( B-D ) ; the numbers of cells were log transformed for all statistical analyses ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 01210 . 7554/eLife . 21238 . 013Figure 5—figure supplement 2 . Frequency of somatic mutations in GC B cells versus PCs . Sequence analysis of VH186 . 2 amplified from single NP-specific GC B cells or PCs from draining lymph nodes of control Syk+/+ Cγ1Cre/+ ( Ctrl ) versus Sykflox/+ Cγ1Cre/+ ( Syk het ) mice . The numbers of mutations compared with the germline sequence are shown on the outer edges of the circles . The pie chart is shaded to represent the fraction of cells with the indicated number of mutations . An inset in the center of each pie chart shows the total number of sequences analyzed ( # seq ) and the frequency of germline sequences . GC B cells were identified as CD138–B220hiPNAhiCD38loIgDlo , total NPint , and intracellular IgEint or total IgG1int . PCs were CD138+B220lo-intPNAloCD38loIgDlo , total NPhi , and intracellular IgEhi or total IgG1hi . Mice were immunized subcutaneously with NP-CGG in alum adjuvant and draining lymph nodes were harvested 9 d later . Cells were sorted over two experiments from a total of 16 mice ( Ctrl ) and 17 mice ( Syk het ) and the sequences were pooled for analysis . We note that while there appear to be a higher frequency of GC B cells with germline sequences in Syk het mice , this was not consistent finding when the data were disaggregated by experiment . The high affinity mutation W33L was only present in a negligible number of sequences at this timepoint , as expected . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 013 One of the targets of antigen receptor signaling is the transcription factor IRF4 , which has been implicated in both GC and PC responses depending on its abundance and association with other transcription factors ( Nutt et al . , 2011; Ochiai et al . , 2013 ) . While IRF4 is necessary for CSR to IgE and IgG1 , precluding the analysis of Irf4-deficient B cells , we were able to evaluate Irf4-heterozygous B cells , which underwent relatively normal CSR . A single copy of Irf4 was conditionally deleted in activated B cells from Irf4flox/+ Cγ1Cre/+ mice . Cell culture of Irf4-heterozygous B cells reduced the antigen-independent PC differentiation of IgE+ B cells ( Figure 4—figure supplement 1 ) , similar to our BCR signaling perturbations in Figure 4 . These data confirm a role for IRF4 in antigen-independent PC differentiation mediated by the IgE BCR . In vivo , the effects of Irf4 heterozygosity were subtle , but we observed a two-fold increase in the frequency of IgE+ B cells compared with a 10% increase in the frequency of IgG1+ B cells in GCs ( Figure 5D ) . There was also a trend toward increased IgE+ PCs when Irf4 was made heterozygous , although this did not reach statistical significance ( Figure 5D ) . The selective increase in IgE+ B cells in GCs and slight increase in IgE+ PCs when Irf4 was heterozygous resembled our above findings with perturbations in BCR signaling . These findings suggest that Irf4 may be one target responsible for the specific effects of BCR signaling on IgE GC and PC responses . In general , our observations with the various perturbations above showed that changes in IgE+ GC B responses were not always coupled with equivalent changes in IgE+ PC responses , which could be due to heterogeneity in PC responses . Indeed , IgE+ PCs can arise both via extrafollicular and GC-derived pathways ( Yang et al . , 2014 ) . We therefore sought to gain insight into the origin of the IgE+ PCs that accumulated in some mice with perturbed BCR signaling . The extent of somatic mutation of the antibody variable regions in PCs and GC B cells was assessed in conditional Syk heterozygous and control mice . We sequenced VDJs containing the VH186 . 2 variable region , which dominates the NP-specific response on the C57BL/6 background ( Bothwell , 1984 ) , from individual NP-specific cells . The majority of IgE+ and IgG1+ GC B cells carried somatic mutations , as expected , whereas the vast majority of IgE+ and IgG1+ PCs had germline sequences at this timepoint ( Figure 5—figure supplement 2 ) . This result indicates that the majority of IgE+ PCs were likely derived from the extrafollicular pathway , suggesting that a two-fold reduction in Syk led to an increase in the IgE+ PC extrafollicular response . However , there was an increase in the fraction of PCs carrying somatic mutations when the cells were Syk heterozygous . In addition , a substantial fraction of GC B cells still had germline sequences at this timepoint . Therefore , we cannot formally exclude that some of the increase in IgE+ PCs may have arisen from the enhanced IgE GC response of Syk heterozygous B cells . Overall , our data implicate BCR signaling in the regulation of both extrafollicular and GC pathways of the IgE response . The IgE BCR may therefore have additional influences on in vivo responses distinct from promoting antigen-independent PC differentiation , which we explored further below . A recent study concluded that the IgE BCR negatively regulates IgE responses by promoting high levels of apoptosis through a mitochondrial pathway ( Laffleur et al . , 2015 ) . The apoptosis was linked to relocalization of the mitochondrial protein Hax1 ( Laffleur et al . , 2015 ) , which was reported to bind a sequence found in the CT of the IgE BCR ( Oberndorfer et al . , 2006 ) . Laffleur et al . ( 2015 ) reported that in primary B cell cultures with anti-CD40 and IL-4 , upon withdrawal of these stimuli , IgE+ B cells exhibited increased apoptosis compared with IgG1+ B cells . However , we were unable to reproduce this result . While withdrawal of anti-CD40 and/or IL-4 led to a general increase in apoptotic cells as measured by annexinV staining , the frequency of apoptotic IgE+ and IgG1+ B cells were similar ( Figure 6A ) . Ectopic expression of the IgE BCR in primary B cells also resulted in similar frequencies of apoptosis to ectopic expression of the IgG1 BCR ( Figure 6B and C ) . Swapping ( constructs 2 and 5 ) or truncating the IgE CT domain to the short KVK sequence found in IgM ( construct 3 ) also had no significant impact on apoptosis ( Figure 6B and C ) . Laffleur et al . ( 2015 ) also reported that transfection of the IgE BCR into the A20 B cell line promoted high levels of apoptosis , making it difficult to obtain stable transfectants . While the A20 cell line is not amenable to retroviral transduction , we were able to use this approach to express our constructs in WEHI-231 cells , which are known to readily undergo apoptosis upon BCR cross-linking ( Benhamou et al . , 1990; Hasbold and Klaus , 1990 ) . Similar frequencies of apoptosis were observed in WEHI-231 cells transduced with the IgE BCR versus IgG1 BCR or with CT mutant constructs ( Figure 6D ) , as we had observed in primary B cells . In contrast , anti-IgM treatment induced robust apoptosis of WEHI-231 cells , as expected , confirming the validity of the assay ( Figure 6D , rightmost panels ) . The frequency of cells transduced with the IgE BCR was also stable from 3 d to 7 d after transduction ( Figure 6D ) , arguing against any deleterious effects of expressing the IgE BCR or its CT . Similar results were also obtained in two other B cell lines , BAL 17 and M12 ( Kim et al . , 1979; data not shown ) . Notably , Laffleur et al . ( 2015 ) had expressed human IgE heavy chains in a mouse B cell line without verifying that the human heavy chains can pair appropriately with mouse BCR components , whereas in our study we expressed normal mouse IgE heavy chains in mouse B cell lines , avoiding this potential technical issue . 10 . 7554/eLife . 21238 . 014Figure 6 . The IgE BCR does not promote intrinsic apoptosis in vitro or in vivo . ( A ) Representative flow cytometry of the frequency of apoptosis ( annexin V+ ) of wild-type B cells cultured for 4 d with IL-4 and anti-CD40 ( αCD40 ) . The culture media was left unchanged , or replaced with fresh media without anti-CD40 ( αCD40 withdrawn ) , without IL-4 ( IL-4 withdrawn ) , or without both anti-CD40 and IL-4 ( αCD40 and IL-4 withdrawn ) 5 hr before annexin V staining . Cells were gated as B220+IgD–IgM– CD138– . ( B and C ) Representative flow cytometry ( B ) and quantification ( C ) of the frequency of apoptosis ( annexin V+ ) of B1-8flox/+ Cγ1Cre/+ B cells that were retrovirally transduced with the indicated constructs ( refer to the legend in the upper-right ) on d 1 of culture . Cells were gated as Cerulean+ . ( D ) Representative flow cytometry of the frequency of apoptosis ( annexin V+ ) of WEHI-231 cells retrovirally transduced with the indicated constructs ( refer to the legend in the upper-right ) compared to negative ( untreated ) or positive controls ( αIgM treated ) and then analyzed after the indicated number of days . Data are representative of three independent experiments . ( E–F ) Representative flow cytometry ( E ) and quantification ( F ) of the frequency of apoptosis ( positive for an antibody to activated caspase ( Cas . ) 3 ) among IgE+ versus IgG1+ GC B cells in draining lymph nodes 7 d after immunization . Similar results were observed in three independent experiments . ( G , H ) Representative flow cytometry ( G ) and quantification ( H ) of the frequency of apoptosis ( ‘Casp-Glow’ VAD-FMK+ , annexin V+ ) among IgE+ versus IgG1+ GC B cells 7 d after immunization . ( I ) Quantification of frequency of IgE+ and IgG1+ cells among GC B cells in littermate control ( Ctrl ) versus Eµ-Bcl2-22 transgenic ( Tg ) mice the indicated number of days after immunization . Data were pooled from four experiments . Mice were immunized subcutaneously with NP-CGG ( E–H ) or NP-KLH ( I ) in alum and draining lymph nodes were analyzed by flow cytometry . GC B cells were gated as CD138–B220hiPNAhiCD38loIgDlo and intracellular IgEint or total IgG1int . Dots represent samples from individual experiments ( C ) , or mice ( F , H , I ) . Bars represent the mean ( C , F , H ) or median ( I ) . n . s . , not significant ( one-way ANOVA ( C ) , paired t-tests ( F , H ) , and t-tests with the Holm-Sidak correction for multiple comparisons ( I ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 014 An in vivo study also reported that IgE+ GC B cells exhibited higher rates of apoptosis than IgG1+ GC B cells , as revealed with an active caspase antibody or active caspase substrate ( He et al . , 2013 ) . However , when we stained with an active caspase-3 antibody , we found similar frequencies of apoptosis among IgE+ and IgG1+ GC B cells ( Figure 6E and F ) . We also observed similar rates of apoptosis when we incubated the cells with an active caspase substrate and when we stained the cells with annexin V ( Figure 6G and H ) . Taken together , we find no evidence that the IgE BCR directly promotes intrinsic apoptosis in primary B cells in culture , in B cell lines , nor in vivo GC B cells . We also confirmed by a genetic approach that intrinsic apoptosis was not the major mechanism responsible for the progressive loss of IgE+ B cells from GCs over time . In transgenic mice overexpressing the anti-apoptotic gene Bcl2 in B cells , we had previously reported a dramatic increase in IgE+ PC but not GC B cell responses ( Yang et al . , 2012 ) . We further characterized the kinetics of IgE+ GC B cell responses in Bcl2 transgenic mice for four weeks after immunization . IgE+ B cells were lost from GCs at a similar rate in Bcl2 transgenic mice and littermate controls ( Figure 6I ) . Taken together , our data are inconsistent with the model that IgE+ B cells are eliminated from GCs primarily due to increased rates of intrinsic apoptosis mediated by the IgE BCR . Our findings that the IgE BCR does not directly promote apoptosis suggested that other mechanisms are likely responsible for the selective effects of BCR signaling on IgE+ GC B cell responses . Recent studies have demonstrated that selection for high affinity B cells in the GC is linked to differential proliferation rates ( Gitlin et al . , 2015 , 2014 ) . Cells with faster cell cycles would outcompete cells with slower cell cycles , which can be measured by administering a pulse of a thymidine analog during S phase and then evaluating cell cycle progression ( Gitlin et al . , 2015 ) . Using this approach , we immunized mice and 6–7 d later , at the peak of the GC response , we injected the thymidine analog EdU to pulse-label cells in S phase . We then waited 3 . 5 hr to allow cells to variably progress from S phase to the G2 , M , and/or G1 phases of the cell cycle , which was assessed by measuring DNA content ( Figure 7A ) . In wild-type mice , we observed that IgE+ GC B cells tended to have slower cell cycles than IgG1+ GC B cells , as revealed by a smaller proportion of EdU+ cells reaching the G1 phase of the cell cycle in this time window ( Figure 7A and B ) . We then tested whether inhibition of Btk with ibrutinib affected the cell cycle speed . Interestingly , ibrutinib treatment resulted in a greater fraction of EdU-labeled cells progressing to the G1 phase of the cell cycle in 3 . 5 hr , indicating faster cell cycles ( Figure 7C ) . In ibrutinib-treated mice , IgE+ and IgG1+ GC B cells showed equivalent cell cycle speeds ( Figure 7C ) . Thus , inhibiting BCR signaling actually made IgE+ GC B cells more competitive with IgG1+ GC B cells . The observed differences in cell cycle speeds are likely to magnify over multiple cell cycles , as GC B cells have been reported to undergo cell division every 6–12 hr ( Allen et al . , 2007b; Hauser et al . , 2007; MacLennan , 1994 ) . These data suggest that BCR signaling normally constrains IgE+ GC B cell responses through prolonged cell cycle times . 10 . 7554/eLife . 21238 . 015Figure 7 . IgE+ GC B cells exhibit delayed cell cycle progression and reduced antigen presentation . ( A–C ) Mice were injected with EdU to pulse-label cells in S phase 6 . 5 d after subcutaneous immunization with NP-CGG in alum . Draining lymph nodes were harvested 3 . 5 hr later . ( A ) Representative flow cytometry of EdU staining and DAPI labeling ( DNA content ) in IgE+ and IgG1+ GC B cells . EdU+ cells were gated ( upper panels ) and G1 versus S/G2/M phase cells were resolved by DNA content ( lower panels ) . ( B ) Quantification of the frequency of EdU+ cells in G1 phase among IgE+ versus IgG1+ GC B cells . ( C ) Quantification of the frequency of EdU+ cells in G1 phase among IgE+ versus IgG1+ GC B cells from mice treated with vehicle ( Ctrl ) or ibrutinib ( Ib ) twice daily for 3 d , starting 4 d after immunization . ( D–G ) HEL-specific ( Hy10 ) B cells were transferred into wild-type congenic recipient mice and then mice were immunized subcutaneously with DEL-OVA in alum and draining lymph nodes were analyzed 7 d later . ( D and E ) Representative flow cytometry ( D ) and quantification ( E ) of BCR surface expression by HEL-Alexa647 labeling of Hy10 GC B cells ( CD45 . 2+ ) compared with recipient endogenous ( Endo . ) GC B cells ( CD45 . 1+ ) . ( F and G ) Representative flow cytometry ( F ) and quantification ( G ) of Eα peptide-MHC II antigen presentation by Y-Ae antibody staining of Hy10 GC B cells ( CD45 . 1+ ) compared with recipient endogenous GC B cells ( CD45 . 1– ) . GC B cells were gated as CD138–B220hiPNAhiCD38loIgDlo and intracellular IgEint or total IgG1int in all panels , with the addition of IgMlo ( F and G ) . Similar results were obtained in two independent experiments . Dots represent individual mice and bars represent the median ( B , E , G ) or mean ( C ) . gMFI , geometric mean fluorescence intensity; n . s . , not significant; **p<0 . 01; ****p<0 . 0001 ( Wilcoxon matched-pairs signed rank test ( B ) , and t-tests with the Holm-Sidak correction for multiple comparisons ( C , E , G ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21238 . 015 As previous studies have linked GC B cell proliferation and selection to the acquisition of T cell help ( Gitlin et al . , 2015 , 2014 ) , we hypothesized that the IgE+ B cells were at a disadvantage for T cell interactions . It is thought that the ability of the membrane BCR to bind and endocytose antigen , which is then processed and presented on MHC class II , determines the extent of T cell interactions in the GC ( Allen et al . , 2007a; Victora and Nussenzweig , 2012 ) . The BCR was reported to be in lower abundance on the surface of IgE+ GC B cells compared with IgG1+ GC B cells ( He et al . , 2013 ) . In order to confirm this finding , we used an adoptive transfer system in which we tracked antigen-specific B cells . Specifically , Hy10 knock-in B cells ( Allen et al . , 2007b ) , specific for avian egg lysozyme , were transferred into congenically-marked mice , which were then immunized with the low affinity antigen duck egg lysozyme ( DEL ) , conjugated to OVA to provide T cell help . We then evaluated BCR surface expression by measuring the capacity for antigen binding . Hy10 IgE+ and IgG1+ GC B cells were labeled with fluorescent hen egg lysozyme ( HEL ) , a high affinity antigen . We observed that antigen-specific IgE+ GC B cells bound 4 . 4-fold less HEL than antigen-specific IgG1+ GC B cells ( Figure 7D and E ) , consistent with a four-fold difference reported with a different antigen system ( He et al . , 2013 ) . We then tested antigen uptake and presentation with the well-established Y-Ae system ( Germain and Jenkins , 2004 ) , which has recently been applied to discern differences in the abundance of peptide-MHC complexes on the surface of GC B cells ( Bannard et al . , 2016 ) . In this system , the Y-Ae antibody recognizes the Eα peptide presented on the I-Ab MHC molecule ( Murphy et al . , 1989; Rudensky et al . , 1991 ) . We conjugated the Eα peptide to HEL and then administered this reagent in the aforementioned adoptive transfer system . After several hours , allowing time for antigen processing to occur , IgE+ GC B cells had reduced Y-Ae antibody binding compared with IgG1+ GC B cells , indicating that IgE+ GC B cells displayed fewer peptide-MHC complexes ( Figure 7F and G ) . Taken together , these data suggest that the low BCR surface expression on IgE+ GC B cells limits the ability of these cells to compete for T cell help , illuminating an additional mechanism by which the IgE BCR may limit IgE+ GC B cell responses .
The findings we have reported here establish that distinct properties of the IgE BCR regulate the fate of IgE+ B cells . In the presence of stimuli mimicking T cell help , the constitutive activity of the IgE BCR promoted PC differentiation in an antigen-independent manner . In contrast , the IgG1 BCR promoted PC differentiation only when the receptor was ligated with cognate antigen . This difference between the IgE and IgG1 BCRs was attributed to multiple domains , particularly the extracellular membrane-proximal migis region , with further contributions from the IgE CH2 and CH3 extracellular domains and the IgE CT . In cell culture , the BCR signaling adapters BLNK , Btk , CD19 , and Syk contributed to antigen-independent PC differentiation mediated by the IgE BCR , with an essential role for CD19 . In mice , genetic deficiency or pharmacological inhibition of these signaling components consistently resulted in elevated IgE GC responses , yet IgE PC responses were variably affected . The increased IgE+ GC B cell responses upon Btk inhibition were associated with accelerated cell cycle progression , suggesting that the reduced competitive fitness of IgE+ GC B cells is in part due to BCR signaling . In addition , low surface expression of the IgE BCR on GC B cells led to reduced antigen uptake and presentation , likely limiting access to T cell help . While previous studies have largely focused on isotype-dependent BCR responses to antigen stimulation , here we revealed that BCR isotypes also differ in antigen-independent activity . BCRs were known to exhibit tonic signaling , as deletion of the BCR results in rapid B cell death ( Kraus et al . , 2004 ) . In this study , we reported that the IgE BCR stood out among all isotypes tested as exhibiting the highest capability of inducing antigen-independent PC differentiation . However , our data suggest that the constitutive activity of the IgE BCR is substantially weaker in magnitude than acute antigen stimulation . Interestingly , weak , constitutive BCR activity also promoted the PC differentiation of B cells from mice in which the BCR had been substituted with the latent membrane protein 2A ( LMP2A ) from Epstein Barr virus ( Lechouane et al . , 2013 ) . Increased PC differentiation was also reported in a BCR transgenic line with low surface abundance of IgG2a and elevated constitutive activity ( Man et al . , 2010 ) . We therefore propose that weak , constitutive BCR activity may be highly effective at promoting PC differentiation . Our data implicated the extracellular membrane-proximal migis region as a key difference between the IgE and IgG1 BCRs . The migis was also a major determinant of BCR surface localization in the presence versus the absence of Igα . This finding suggests that the interaction between the BCR and Igα extends beyond the TM domains into the extracellular membrane proximal region . A possible association of the migis region with Igα had been speculated based on experiments with human versus mouse Igα ( Reth , 1992 ) . In addition , the migis regions of IgM and IgD were reported to control the glycosylation pattern of Igα ( Pogue and Goodnow , 1994 ) . Since our data identified the migis region as an important component of the antigen-independent activity of the IgE BCR , we propose that this region is normally involved in signal transduction upon antigen binding . The migis of IgE had been implicated in signal transduction in a study of simplified BCR molecules in which receptors were cross-linked ( Poggianella et al . , 2006 ) . Human IgE BCRs contain one of two possible migis regions by alternative splicing , a long versus short form ( Poggianella et al . , 2006 ) . It will be interesting in future studies to examine whether these splice variants differentially regulate the interaction of the IgE BCR with Igα . Our finding that the CT region of IgE could contribute to , but was not essential for , antigen-independent PC differentiation , is consistent with a role for this region in modulating or enhancing IgE BCR signaling ( Engels et al . , 2009; Sato et al . , 2007 ) . Some functions of the IgE CT might be equivalent to the IgG CT due to the conserved tyrosine motif; thus , our domain swap experiments would highlight isotype-specific functions of these domains ( Engels et al . , 2009 ) . A contribution of the IgE CT to in vivo responses was noted in a mouse strain engineered to delete this region , however , notably , the effect was quantitative ( 2–4 fold ) rather than absolute ( Achatz et al . , 1997 ) , again suggesting the IgE CT contributes to but is not essential for the distinct features of the IgE BCR . The domain swap experiments also revealed a partial contribution of the IgE extracellular CH2 and CH3 domains to antigen-independent PC differentiation . Structural studies of secreted IgE indicated that an asymmetric bend can occur in IgE at the CH2-CH3 linker region , causing one of the two CH2 domains to fold back onto the CH3 and CH4 domains , making particularly extensive contacts with the CH3 domain ( Wan et al . , 2002 ) . Recent evidence suggests that this bent shape can also be extended , which can subsequently allow CH2 to flip from one side of CH3-CH4 to the other ( Drinkwater et al . , 2014 ) . It seems plausible that for the membrane IgE BCR , the ability to adopt an asymmetric bend conformation mimics antigen binding . The IgE CH2 domain versus IgG hinge regions are also analogous to the domain structures of IgM versus IgD , respectively , which recently have been implicated in differential antigen responsiveness ( Übelhart et al . , 2015 ) . Providing further evidence for the regulation of in vivo B cell responses by the IgE extracellular domains , a recent study found that memory responses were abrogated when the extracellular domains of IgG1 were replaced with those of IgE ( Turqueti-Neves et al . , 2015 ) . Of the BCR signaling adapter molecules that we examined , we observed that CD19 was essential for antigen-independent PC differentiation . Signal transduction via PI3K , a target of CD19 , has also been identified as the major pathway by which tonic BCR signaling maintains BCR survival ( Srinivasan et al . , 2009 ) . PI3K may therefore be a general transducer of constitutive BCR signals which can potentially mediate different outcomes including PC differentiation and survival . Since PI3K signaling is also downstream of numerous other receptors , future studies may provide further insights into how these signals are integrated to determine cell fate . Our studies also support a role for Syk , Btk , and BLNK in antigen-independent PC differentiation . In future studies , it would be interesting to delineate how these findings relate to studies on the role of ERK1/2 in antigen-dependent PC differentiation ( Yasuda et al . , 2011 ) . In physiologic immune responses to T-dependent antigens , B cells may receive both antigen-independent and antigen-dependent BCR signals . Since the IgE BCR exhibited antigen-independent activity that was distinct from the IgG1 BCR and that was particularly sensitive to perturbations in BCR signaling , we hypothesized that modulating BCR signaling would have a selective effect on IgE+ B cell responses in vivo . Indeed , we found it striking that in four different perturbations of BCR signaling , the frequency of IgE+ B cells within GCs was selectively increased , whereas the frequency of IgG1+ B cells was not affected . Together with our above findings , these results suggest that the antigen-independent activity of the IgE BCR normally limits the magnitude of IgE+ B cell responses in the GC . Unexpectedly , some of the perturbations in BCR signaling that we tested resulted in increased numbers of IgE+ PCs in immunized mice , whereas we had observed reductions in IgE+ PCs in cell culture . This result could be related to the timing of PC differentiation driven by antigen-independent versus antigen-dependent BCR signals . Specifically , the antigen-independent activity of the IgE BCR is likely to promote the ‘premature’ terminal differentiation of IgE+ B cells into PCs . The BCR signaling perturbations may have inhibited this premature terminal differentiation , thereby enabling IgE+ B cells to undergo further rounds of proliferation , prior to normal antigen-dependent PC differentiation . This model is consistent with our observations that the BCR signaling perturbations did not affect IgG1+ PC numbers , which our data suggest are generated primarily by antigen-dependent signals . In addition , this model is consistent with our findings regarding IRF4 , a transcription factor that is essential for PC differentiation and can be induced by BCR signaling ( Nutt et al . , 2011; Ochiai et al . , 2013 ) . When activated B cells were made Irf4 heterozygous , this resulted in reduced antigen-independent IgE+ PC differentiation in cell culture but a slight increase , rather than a decrease , in the total number of IgE+ PCs in immunized mice . Some of the increase in IgE+ PCs that we observed with various perturbations in mice could be a consequence of enhanced IgE+ GC B cell responses , but our somatic hypermutation analysis suggests this cannot fully explain the results , since most IgE+ PCs had germline sequences whereas most IgE+ GC B cells had somatic mutations . Another possibility is that the BCR signaling perturbations could have a direct impact on the proliferation or longevity of the IgE+ PCs , which have high surface expression of the IgE BCR . The BCR signaling perturbations could also have affected the rate of CSR to IgE , as PI3K signaling has been reported to negatively regulate CSR ( Omori et al . , 2006 ) . Overall , our data are most consistent with the model that perturbations in BCR signaling led to delayed terminal differentiation of IgE+ B cells , enabling further rounds of proliferation and the generation of longer-lived PCs . Since the perturbations in BCR signaling led to a very consistent increase in IgE+ GC B cells , whereas the effects on IgE+ PC numbers were less consistent , it seemed likely that IgE BCR signaling does not solely affect PC differentiation . Our mutational analysis revealed that most of the IgE+ PCs were derived from the extrafollicular pathway at the timepoint in which we could observe increases in IgE+ GC B cells , suggesting that BCR signaling also has a direct impact on the GC response . One group proposed that IgE+ B cells are progressively lost from GCs due to low BCR surface expression , leading to reduced BCR signaling and apoptosis ( He et al . , 2013 ) . However , using similar methods to detect apoptotic B cells , we were unable to replicate the finding that IgE+ B cells had higher rates of apoptosis compared with IgG1+ B cells in GCs . A caveat to these disparate findings is that the ex vivo analysis of the apoptosis of GC B cells may be hampered by the rapid uptake of apoptotic cell fragments by tingible body macrophages ( MacLennan , 1994 ) . In contrast , the analysis of apoptosis in cultured B cells should be more robust . Yet , in cell culture studies , we were also unable to observe an increased rate of apoptosis in IgE+ B cells compared with IgG1+ B cells , in contrast to a recent study ( Laffleur et al . , 2015 ) . We further verified that the ectopic expression of the IgE BCR in primary B cells did not promote apoptosis in cell culture . In addition , while a recent study reported that IgE BCR expression promoted the apoptosis of a B cell line ( Laffleur et al . , 2015 ) , we were unable to replicate this result in three different B cell lines . Consistent with our observations , a study of isotype-specific BCR signaling reported no difficulty in expressing the IgE BCR in B cell lines ( Sato et al . , 2007 ) . A possible explanation for the apoptosis due to the IgE BCR observed in the Laffleur et al . ( 2015 ) study was that the human IgE BCR was expressed in mouse B cells . The human IgE heavy chain may not appropriately pair with key accessory proteins in mouse cells , such as the mouse Ig light chains or mouse Igα , which could affect BCR surface localization . We also observed that the progressive loss of IgE+ B cells from GCs could not be prevented by overexpression of the anti-apoptotic gene Bcl2 , in contrast to our prior results demonstrating that short-lived IgE+ PCs were rescued from apoptosis . Similarly , another group reported that IgE+ GC B cells were very rare in mice deficient in the death receptor Fas , despite exaggerated IgE+ PC responses ( Butt et al . , 2015 ) . Taken together , our data are inconsistent with the notion that the IgE BCR directly promotes apoptosis in B cells . We therefore considered other mechanisms by which IgE+ GC B cell responses might be restricted by the IgE BCR . Recent studies of the GC have revealed that differential rates of proliferation may contribute to the selection of high affinity B cells ( Gitlin et al . , 2015 , 2014 ) . On average , IgE+ GC B cells exhibited slower cell cycles than IgG1+ GC B cells , yet these rates increased and became equivalent when BCR signaling was inhibited . Interestingly , in a recent study , most GC B cells were observed to have dampened BCR signaling due to high expression of the phosphatase SHP-1 , yet an exception was the cells in the G2/M phases ( Khalil et al . , 2012 ) . We favor a model in which constitutive signaling by the IgE BCR delays the progression of IgE+ GC B cells through the G2/M phases of the cell cycle . After multiple cell cycles , this delayed proliferation may result in the IgE+ GC B cells being outcompeted by IgG1+ GC B cells . The low abundance of the IgE BCR on GC B cells , reported in a recent study ( He et al . , 2013 ) and confirmed here , is likely to also influence GC B cell responses . In addition to an unusual polyadenylation sequence ( Anand et al . , 1997; Karnowski et al . , 2006 ) , the low expression of the IgE BCR may be related to its constitutive activity , as has been observed with the downregulation of IgM on autoreactive B cells that receive chronic activation signals ( Goodnow et al . , 1988 ) . Indeed , a recent study obtained evidence for spontaneous internalization of the IgE BCR ( Laffleur et al . , 2015 ) , which is likely related to its constitutive activity . Since we could not observe increased apoptosis of IgE+ GC B cells , we considered the impact of low BCR expression on antigen uptake and presentation . We observed that a smaller proportion of IgE+ GC B cells captured antigen and displayed peptide-MHC complexes compared with IgG1+ GC B cells . This finding suggests that IgE+ GC B cells have limited access to T cell help compared with IgG1+ GC B cells , due to low expression of the IgE BCR . Overall , our findings suggest that expression of the IgE BCR may result in a competitive disadvantage for IgE+ B cells in the GC , due to premature differentiation , delayed cell cycle progression , and limited antigen uptake and presentation . These findings may thus help explain the progressive loss of IgE+ GC B cells over time observed by multiple groups ( He et al . , 2013; Talay et al . , 2012b; Yang et al . , 2012 ) . However , we note that our observations were at a population level and do not directly reflect the behavior of individual cells . A very small fraction of IgE+ GC B cells could still acquire the high affinity mutations necessary to surpass the stringent threshold of selection during affinity maturation , as has been observed by two groups ( He et al . , 2013; Yang et al . , 2012 ) . Our findings may also have implications for the dynamics of IgE+ GC B cells . For example , signals provided by T cell help are thought to promote the transit of GC B cells from the light zone to the dark zone ( Victora and Mesin , 2014 ) . IgE+ B cells , at a disadvantage for T cell help , might thus show reduced evidence of light zone to dark zone transit . However , the net impact on positioning within the GC may be hard to predict . Indeed , the signals that regulate the timing at which cells transit from the dark zone to the light zone are not well defined ( Bannard et al . , 2013; Victora and Mesin , 2014 ) . IgE+ GC B cells also exhibited different cell cycle characteristics than IgG1+ GC B cells , which might influence dark and light zone positioning ( Victora and Nussenzweig , 2012 ) . One study reported a paucity of IgE+ B cells in the GC light zone ( He et al . , 2013 ) , whereas we had readily observed IgE+ GC B cells within the dense follicular dendritic cell network corresponding to the light zone ( Yang et al . , 2012 ) . It would be interesting in future studies to determine whether and how the IgE BCR influences the positioning of cells in dark and light zones over the course of an immune response . While our manuscript was being finalized , another study was published ( Haniuda et al . , 2016 ) reporting that autonomous signaling of the IgE BCR induced PC differentiation and apoptosis . In this study , ectopic expression of the IgE BCR promoted PC differentiation in cultured B cells , similar to our findings . This group also reported that the IgE BCR promoted apoptosis; however , we were not able to observe this effect in any of our in vitro or in vivo assays . As in our domain swap studies , this study identified the IgE migis region as critical for antigen-independent PC differentiation . The IgE migis was reported by Haniuda et al . to influence the BCR association with CD19 , which we propose would be secondary to the interaction with Igα that we identified here . The extracellular domains of IgE were also implicated in the autonomous activity of the IgE BCR in this paper , which here we have further defined as specifically the CH2 and CH3 domains of IgE . The authors also reported a modest impact of deleting the IgE CT but did not test swapping this domain with that of IgG1 , as we reported here . Another distinct feature of our work presented here is the microscopy studies of the mobility and clustering of the IgE BCR , which were not examined by Haniuda et al . While both groups observed increased IgE+ GC B cells in Cd19-heterozygous and BLNK-deficient mice , our conclusions differed with respect to IgE+ PC responses . Haniuda et al . reported a decreased IgE PC/GC ratio in Cd19-heterozygous mice as evidence of decreased PC generation; however , this ratio was likely affected primarily by the increased number of IgE+ GC B cells , since we actually observed a modest increase , rather than a decrease , in IgE+ PC numbers in these mice . We were also not able to reproduce the decrease in IgE+ PCs reported in BLNK-deficient mice and instead we observed a striking increase in IgE+ PCs . We further tested the in vivo impact of inhibiting Btk as well as deletion of a single copy of Syk and Irf4 , which also led to increased IgE+ GC B cell responses but inconsistent effects on IgE+ PCs . While we support the idea that the autonomous activity of the IgE BCR can promote short-lived PC differentiation , the uncoupling of IgE+ GC B cell and PC responses in our studies suggested that BCR signaling has other impacts on IgE+ GC responses . Specifically , we established that BCR signaling affects the cell cycle progression of IgE+ GC B cells and that low BCR surface expression led to reduced antigen uptake and presentation . Taken together , we propose the following summary model to account for our findings . Upon CSR to IgE , a B cell will begin to express the IgE BCR , which has distinct constitutive activity . If this B cell receives T cell help , this BCR activity will promote PC differentiation . However , if this B cell does not receive T cell help , this BCR activity will delay cell cycle progression , causing the B cell to be outcompeted by clones expressing other isotypes . In this way , the constitutive activity of the IgE BCR makes cell fate relatively independent of antigen stimulation of the BCR , but rather highly dependent on antigen presentation and T cell help . Thus , the availability of T cell help may be a major mechanism for controlling IgE B cell responses . The chronic activity of the IgE BCR leads to surface BCR downmodulation , reducing antigen uptake and presentation , thus making it less likely that the cell will receive T cell help . Premature PC differentiation , delayed cell cycle progression , and limited T cell help due to IgE BCR expression all result in the progressive decline in IgE+ GC B cells and prevent the generation of IgE+ memory B cells and long-lived PCs , thereby reducing the average affinity and overall duration of the IgE response . A general implication of our work is that BCR signaling negatively regulates IgE responses . Interestingly , Btk-deficient mice had been observed to undergo enhanced IgE responses , although this was not thought to be due to BCR signaling ( Kawakami et al . , 2006 ) . In genetic experiments that specifically affect signaling in B cells , we were able to demonstrate that diminished BCR signaling leads to exaggerated IgE+ responses . Our data therefore suggest that genetic variations or pharmacological treatments that alter the strength of BCR signaling may have a selective effect on IgE+ responses , which may be clinically important in the development of allergy .
C57BL/6J mice ( RRID:IMSR_JAX:000664 ) , Boy/J CD45 . 1 mice ( RRID:IMSR_JAX:002014; B6 . SJL-PtprcaPepcb/BoyJ ) , B1-8i mice ( RRID:IMSR_JAX:012642; B6 . 129P2 ( C ) -Ightm2Cgn/J ) , Blnk–/– mice ( RRID:IMSR_JAX:004524; B6 . 129-Blnktm1Achn/J ) , Cγ1-Cre mice ( RRID:IMSR_JAX:010611; B6 . 129P2 ( Cg ) -Ighg1tm1 ( cre ) Cgn/J ) , Irf4flox mice ( RRID:IMSR_JAX:009380; B6 . 129S1-Irf4tm1Rdf/J ) , and Sykflox mice ( RRID:IMSR_JAX:017309; B6 . 129P2-Syktm1 . 2Tara/J ) were originally from The Jackson Laboratory . Aicda–/– mice ( RRID:MGI:2654846; Aicdatm1Hon; [Muramatsu et al . , 2000] ) , B1-8flox mice ( RRID:MGI:3693006; Ightm4Cgn; [Lam et al . , 1997] ) , Cd19Cre mice ( RRID:MGI:1931143; Cd19tm1 ( cre ) Cgn; [Rickert et al . , 1995] ) , Eµ-Bcl2-22 mice ( RRID:MGI:3052827; B6 . Cg-Tg ( BCL2 ) 22Wehi; [Strasser et al . , 1991] ) , Hy10 mice ( RRID:MGI:3702732; B6 . Cg-Ightm1Cys/+ , Tg ( Igk ) 5Cys; [Allen et al . , 2007b] ) , and Nur77-GFP ( RRID:MGI:4847273; Tg ( Nr4a1-EGFP ) GY139Gsat; [Zikherman et al . , 2012] ) mice were maintained on the C57BL/6 background and/or bred to Boy/J CD45 . 1 congenic mice . Otherwise , C57BL/6 CD45 . 1 congenic mice , which were used as wild-type mice in some experiments , were from the National Cancer Institute / Charles River Frederick National Laboratory ( 01B96; B6-Ly5 . 2/Cr , later renamed to B6-Ly5 . 1/Cr ) . Mice with significant skin lesions or other signs of poor health were excluded from the study . Mice were housed in specific-pathogen-free facilities and protocols were approved by the Institutional Animal Care and Use Committee of the University of California , San Francisco . NP-CGG ( estimated conjugation ratio of 30–33 ) and NP-KLH ( estimated conjugation ratio of 27–29 ) were purchased from Biosearch Technologies . For the preparation of DEL-OVA , DEL ( Worthington Biochemical ) was resuspended at 11 mg/mL in 50 mM sodium phosphate buffer pH 7 . 1 with 2 . 2 mM EDTA ( Thermo Fisher Scientific ) , then diluted with 10% v/v 1 M sodium bicarbonate . Endograde OVA ( Hygros ) was resuspended in PBS at 2 mg/mL . Lysines of DEL were labeled with a free thiol residue by incubating DEL with Traut’s reagent ( Thermo Fisher Scientific ) at an 0 . 8:1 ratio of Traut’s:DEL for 60 min . Concomitantly , a maleimide group was linked to OVA lysines using sulfosuccinimidyl 4- ( N-maleimidomethyl ) cyclohexane-1-carboxylate ( Sulfo-SMCC; Thermo Fisher Scientific ) at a 15:1 ratio of Sulfo-SMCC:OVA . DEL-SH and OVA-maleimide conjugates were respectively purified through Bio-Spin 6 and Bio-Spin 30 columns ( Bio-Rad ) . DEL-SH was added to OVA-maleimide at an estimated 10:1 excess of DEL-SH: OVA-maleimide based on the original concentrations of reagents . Free DEL was removed from the preparation by gel filtration in DPBS over a Superdex 200 Increase 10/300 GL column ( GE Healthcare ) , followed by concentration through Amicon Ultra 30k MWCO concentrator columns ( EMD Millipore ) . The final concentration was determined from the A280 using a 0 . 1% Cε of 1 . 18 , assuming an average 1:1 stoichiometry of DEL:OVA . For the preparation of HEL conjugated to Eα peptide , HEL ( Sigma-Aldrich ) was resuspended at 4 mg/mL in H2O with 2 mM EDTA . A maleimide group was linked to HEL lysines using Sulfo-SMCC at a ratio of 0 . 7:1 Sulfo-SMCC:HEL for 60 min at room temperature with rolling and tilting . Free Sulfo-SMCC was removed by passing the HEL Sulfo-SMCC mixture through either 7 MWCO Zeba-Spin ( Thermo Fisher Scientific ) or Bio-Spin 6 ( Bio-Rad ) columns equilibrated with 50 mM sodium phosphate buffer pH 7 . 1 with 2 mM EDTA . The resulting HEL-maleimide conjugate was incubated with the Eα peptide carrying an additional terminal cysteine residue ( H2N-ASFEAQGALANIAVDKAC-OH; New England Peptide ) at a 2:1 ratio of Eαcys:HEL , based on the original concentrations . Stoichiometric ( 1:1 ) HEL-Eα was purified from the soluble fraction over a HiTrap SP FF cation exchange column ( GE Healthcare ) on an ÄKTApure chromatography system , using 50 mM sodium phosphate buffer pH 7 . 1 ( solution A ) with a gradient into 50 mM sodium phosphate + 1 M sodium chloride ( solution B ) . Under these conditions , stoichiometric HEL-Eα conjugates eluted at ≈6 . 5% solution B as determined by SDS-PAGE ( unpublished data ) . Stoichiometric HEL-Eα was concentrated through Amicon Ultra 10 kDa MWCO concentrator columns ( EMD Millipore ) , and the concentration was determined from the A280 using a predicted 0 . 1% Cε of 2 . 116 . NP-CGG or NP-KLH were administered subcutaneously in alum adjuvant ( Alhydrogel; Accurate Chemical and Scientific ) . The antigens were brought to a concentration of 1 mg/ml in PBS and then mixed with an equal volume of alum adjuvant . In most experiments , 25 µl per site were injected into the upper flanks , above the shoulders , and scruff of the neck to generate responses in the draining axillary , brachial , and facial LNs . In some experiments , 25 µl per site were injected into both lower flanks and bilaterally proximal to the base of the tail , generating an immune response in the draining inguinal LNs . For experiments involving Hy10 B cells , 5 × 104 HEL-binding naïve follicular B cells were adoptively transferred into congenic ( CD45 . 1 or CD45 . 2 ) wild-type mice by intravenous injection into the retro-orbital plexus , 1 d prior to immunization as described ( Allen et al . , 2007b ) . For Figure 7D and E , immunizations were in both ear pinnae with 6 . 25 µg DEL-OVA/alum , followed by analysis 7 d later of the draining facial lymph nodes . For Figure 7F and G , mice were immunized subcutaneously on the left side at the lower flank and base of the tail with 15 µg DEL-OVA/alum . 6 . 5 d later , mice were challenged with 10 µg HEL-Eα on the left side at the lower flank and base of the tail , followed by analysis 9 hr later of the draining inguinal lymph nodes . Ibrutinib was synthesized as per published protocols ( Pan et al . , 2007 ) and dissolved in Captex 355 ( ABITEC ) at 1 . 56 mg/mL by sonication . Mice were immunized with NP-CGG subcutaneously as described above , and then starting 3 . 5 or 6 d after immunization they were injected intraperitoneally twice each day for 3 d with ibrutinib at 12 . 5 mg/kg dose or vehicle control . Draining lymph nodes were harvested 0 . 5 d after the last injection . Primary B cells were purified from spleens by CD43 and CD11c negative selection as described ( Sullivan et al . , 2011 ) . To enrich for NP-specific B cells from B1-8 mice , Igλ+ B cells were purified by negative selection using biotin-conjugated anti-Igκ ( RRID:AB_345328 , clone RMK-12 , BioLegend ) in addition to anti-CD43-biotin ( RRID:AB_2255226 , clone S7 , BD Biosciences ) and anti-CD11c-biotin ( RRID:AB_313773 , clone N418 , BioLegend ) . In order to induce CSR and/or proliferation in vitro , purified splenic B cells or crude splenocytes ( 5 × 105 cells/ml ) were cultured in RPMI media containing 10% fetal bovine serum ( Life Technologies ) , 1x penicillin-streptomycin-glutamine ( Life Technologies ) , 50 µM β-mercaptoethanol ( Thermo Fisher Scientific ) , 10 mM HEPES ( Life Technologies ) , 62 . 5–250 ng/ml rat anti-mouse CD40 antibody ( RRID:AB_871724 , FGK-45 , Miltenyi Biotec ) , and 25 ng/ml recombinant mouse IL-4 ( R and D Systems ) in 96-well Microtest U-bottom plates ( BD Falcon ) with a volume of 200 µl per well . Cells were cultured for 4–5 d in a humidified incubator at 37°C in 5% CO2 . Data from entire culture experiments were discarded if viability was unusually poor or positive controls failed to show robust PC differentiation . For treatment of cultured B cells , ibrutinib dissolved in DMSO was added to final concentration of 12 . 5 nM or as indicated 2 d after initiation of the culture . The final concentration of DMSO in the culture was always 0 . 025% ( w/v ) . To stimulate B cells with cognate antigen , NP-APC or TNP12-OVA ( Biosearch Technologies ) was added to a final concentration 0 . 5 µg/ml on d 3 of culture . The mouse myeloma-derived J558L cell line ( RRID:CVCL_3949 , obtained from J . Cyster ) was maintained in DMEM high glucose media with 10% FBS , 10 mM HEPES , and 1x penicillin-streptomycin-glutamine ( Life Technologies ) in a humidified incubator at 37°C in 10% CO2 . The mouse B lymphoma cell lines BAL 17 ( RRID:CVCL_9474 , obtained from A . DeFranco ) , M12 ( obtained from J . Cyster ) , and WEHI-231 ( RRID:CVCL_0577 , obtained from J . Cyster ) were maintained in RPMI media containing 10% fetal bovine serum ( Life Technologies ) , 1x penicillin-streptomycin-glutamine ( Life Technologies ) , 10 mM HEPES ( Life Technologies ) in a humidified incubator at 37°C in 5% CO2 . Anti-IgM stimulation of WEHI-231 cells was with 5 µg/ml F ( ab’ ) 2 fragment goat anti-mouse IgM , µ chain-specific ( RRID:AB_2338474 , Jackson ImmunoResearch ) . Phoenix-Eco cells ( RRID:CVCL_H717 , obtained from K . M . Ansel , [Swift et al . , 2001] ) , which were used for retroviral packaging , were seeded into multi-well plates with DMEM high glucose media containing 10% FBS , 10 mM HEPES , 1x penicillin-streptomycin-glutamine , adjusting the cell density to reach 50–70% confluency at the time of transfection . We confirmed that cell lines had characteristics consistent with published results , by flow cytometric evaluation of surface marker expression and/or functional assays . As most of our study involved short-term cultures of primary cells isolated from mouse tissues , testing for mycoplasma in cell cultures was not performed . The VH ( designated V4 ) and Vκ genes of the IgG1 hybridoma clone 1B7 . 11 ( M . Wabl lab , UCSF ) , specific for TNP , were cloned by degenerate PCR as described ( Bradbury et al . , 1995 ) . The coding sequences of membrane IgG1 and secreted IgG1 were amplified from cDNA of the 1B7 . 11 hybridoma as well . The coding sequences of membrane and secreted isoforms of IgM , IgD , IgA , and IgE were cloned from cDNA prepared from C57BL/6 B cells activated in vitro . The coding sequence of V4 was fused to the constant regions of different BCR isotypes by overlapping PCR . Expression constructs of chimeric BCRs were obtained using the Gibson Assembly Master Mix ( New England BioLabs ) . Expression constructs of BCRs with site-directed mutations were obtained using the Q5 site-directed mutagenesis kit ( New England BioLabs ) . The identity of all expression constructs have been verified by Sanger sequencing . To allow relatively high and constitutive expression of exogenous genes in primary lymphocytes and B cell lines , we modified a self-inactivating retroviral vector pQCXIN ( Clontech ) . The CMV immediate early promoter , the internal ribosome entry site , and the neomycin resistance gene in the vector were replaced with the human EF1α promoter from the pEF/myc/nuc plasmid ( Life Technologies ) . To allow assessment of the expression of BCR constructs by retroviral transduction , the coding sequences of the fluorescent protein Cerulean ( Rizzo et al . , 2004 ) ( obtained from Addgene ) , light chain , and heavy chains of different isotypes of BCR were linked in-frame by T2A peptide sequences without internal stop codons and cloned downstream of the EF1α promoter using the restriction sites indicated in Figure 1—figure supplement 1 . The T2A-light chain cassette was only included in the retroviral vectors for expression of TNP-specific BCRs ( Figure 1E ) . Phoenix-Eco cells ( Swift et al . , 2001 ) were transfected with a mixture of 70% retroviral plasmid DNA and 30% MSCV ecotropic gag-pol-env plasmid DNA ( from J . Cyster , UCSF ) using TransIT -LT1 Transfection Reagent ( Mirus Bio ) according to the manufacturer’s instructions . The medium of the transfected cells were replaced with fresh medium the next morning and again with fresh medium containing 1x ViralBoost ( Alstem ) in the evening . The next day , the retroviral supernatant was added to primary B cells or B cell lines together with 10 mM HEPES and 5 µg/ml polybrene , followed by centrifugation at 1100x g at room temperature for 90 min . The spinfected cells were then resuspended in the original growth media for further culture . Primary B cells were analyzed by flow cytometry 3 d after spinfection , whereas B cell lines were analyzed at multiple time points as indicated . Cell suspensions were prepared from LNs and were stained with antibodies ( Supplementary file 1 ) essentially as described ( Yang et al . , 2012 ) . With the exception of the apoptosis analyses described below , nonviable cells were excluded by labeling cells during surface staining with the fixable viability dye eFluor780 ( eBioscience ) as described ( Yang et al . , 2012 ) . Intracellular IgE staining was as described ( Yang et al . , 2012 ) . Briefly , cell surface IgE was first blocked with a large excess of unconjugated anti-IgE antibody RME-1 ( RRID:AB_315073 , BioLegend ) during surface staining , then cells were fixed and permeabilized using the Cytofix/Cytoperm Fixation/Permeabilization solution kit ( BD Biosciences ) . Then intracellular IgE was stained with fluorescently-labeled RME-1 ( see Supplementary file 1 ) . For the detection of NP-binding cells , APC ( ProZyme ) was conjugated to the succinimidyl ester of NP ( NP-Osu; Biosearch Technologies ) at a ratio of 1 mg APC to 80 µg NP-Osu as described ( McHeyzer-Williams and McHeyzer-Williams , 2004 ) and then purified on Bio-Spin 6 or Bio-Spin 30 spin columns ( Bio-Rad ) equilibrated with PBS . For the preparation of HEL-Alexa 647 , lyophilized HEL ( Sigma-Aldrich ) was resuspended at 1 . 1 mg/mL in DPBS ( Life Technologies ) , and then diluted with 10% v/v 1 M sodium bicarbonate . To this solution , an 8 . 8-fold molar excess of Alexa Fluor 647 carboxylic acid , succinimidyl ester ( Life Technologies ) was added . After incubation at room temperature for two hours with gentle mixing , excess free Alexa Fluor 647 dye was removed by passing the solution through two Bio-Spin 6 columns ( Bio-Rad ) equilibrated with PBS . Flow cytometry data were collected on an LSR Fortessa ( BD ) and analyzed with FlowJo v10 . All samples were gated on FSC-A versus SSC-A , over a broad range of FSC-A to include blasting lymphocytes , followed by FSC-W versus FSC-H and then SSC-W versus SSC-H gates to exclude doublets . Some two dimensional plots are shown with ‘large dots’ for better visualization of rare events . For the analysis of apoptosis of cultured primary B cells and B cells lines , annexin V staining and cell surface antigen staining and washing were done in annexin staining buffer ( 10 mM HEPES ( pH7 . 4 ) , 0 . 14 M NaCl , 2 . 5 mM CaCl2 ) . The primary B cells were then fixed and permeabilized for IgE and IgG1 staining . For the ex vivo analysis of apoptosis in GC B cells , after surface staining , fixed and permeabilized cells from draining LNs were stained with anti-activated caspase 3 antibody ( RRID:AB_1727414 , clone C92-605 , BD Biosciences ) . Alternatively , cell suspensions from draining LNs were incubated with FITC-VAD-FMK ( BioVision ) at 37°C for 30’ , washed once with the buffer included in the CaspGLOW kit ( BioVision ) , and then surface stained with annexin V and for other markers , followed by intracellular staining for IgE and total IgG1 , as described above . Our FSC-A and SSC-A gating strategy involved a broad FSC-A gate for lymphocytes that included both viable and recently apoptotic cells , but did not include small cell fragments which tended to be autofluorescent and nonspecifically bound to antibodies , which would have confounded analysis . For cell cycle analysis , each mouse was injected intravenously with 1 mg EdU ( Life Technologies ) in 1xPBS 6 . 5 d after immunization . After 3 . 5 hr , draining LNs were collected and single cell suspensions were processed for cell surface staining followed by EdU detection , using the Click-iT plus EdU-Alexa Fluor 647 kit ( Life Technologies ) , according to the manufacturer’s protocol , except that 10 million cells , instead of 1 million cells were used for one test . To ensure acquiring sufficient events of IgE+ B cells , 50 million cells from the draining LNs of each mouse were processed . Consequently , intracellular IgE and total IgG1 were stained as described above , but in the Click-iT saponin-based permeabilization and wash buffer . Finally DAPI was added to label DNA to determine the fraction of cells in the G1 , S , and G2/M phases of the cell cycle . The final concentration of DAPI was adjusted to achieve a similar fluorescence intensity of the G1 peak and the signal was collected on a linear scale at the LO speed setting to maximize resolution . Doublets were excluded by FSC-W , SSC-W , and DAPI-W measurements . Bulk populations of NP-specific , IgE+ or IgG1+ GC B cells and PCs were first enriched by magnetic bead-based depletion and then sorted on a FACS Aria 3u ( BD ) , as described ( Yang et al . , 2012 ) . Single cells were then sorted into 96-well plates and VH186 . 2 sequences were amplified by nested PCR , followed by Sanger sequencing , and analyzed as described ( Yang et al . , 2012 ) . To quantify the BCR clustering on the surface of J558L cells transduced with IgE or IgG1 by TIRF microscopy , BCRs were visualized by Venus ( Nagai et al . , 2002 ) , a YFP derivative , fused with the cytoplasmic domain of Cd79a as described ( Liu et al . , 2010a , 2010b; Tolar et al . , 2009 ) . Cells were stained with 10 µM of the hydrophobic membrane dye DiD on ice for five mins , washed and then loaded onto coverslips that had been pre-coated with Poly-L-Lysine . Cells were allowed to adhere for five mins and then imaged live at 37°C . TIRF microscopy images were captured by an Olympus IX-81 microscope supported by ANDOR iXon+ DU-897D electron-multiplying EMCCD camera , a 514 nm laser , a 568 nm laser , Olympus 100 × 1 . 45 N . A . objective lens and a TIRF port . TIRF microscopy image capture was controlled by Metamorph software ( Molecular Devices ) and the exposure time was 100 ms for 512 × 512 pixel images . BCR microcluster fluorescence intensity and size were analyzed by Matlab ( MathWorks ) ( Source code 1 ) or Image J ( NIH ) as described ( Liu et al . , 2010a , 2010b; Tolar et al . , 2009 ) . Single BCR molecule tracking experiments were performed as described ( Liu et al . , 2010a , 2010b; Tolar et al . , 2009 ) . In brief , BCRs labeled with Venus were first photobleached with a high power laser for 5–10s , and then imaged by TIRF microscopy . A 100 × 100 pixel sub-region of the electron-multiplying CCD chip with an exposure time of 30 ms per frame was used , the time resolution of which was sufficient to track the single-molecule BCRs as described . Short-range diffusion coefficients and MSD for individual BCR molecule trajectories were processed as described . GraphPad Prism v6 or v7 were used for statistical analyses , with appropriate tests chosen based on experimental design after consulting the GraphPad Statistics Guide . All tests were two-tailed . In order to achieve sufficient power to discern meaningful differences , experiments were performed with multiple biological replicates and/or multiple times , with details provided in each individual figure legend . The number of samples chosen for each comparison was determined based on past similar experiments or by performing pilot experiments to assess the expected magnitude of differences . Full statistical results with exact p values for all figures are provided in Supplementary file 2 . | Antibodies are proteins that recognize and bind to specific molecules , and so help the immune system to defend the body against foreign substances that are potentially harmful . In some cases , harmless substances – such as pollen , dust or food – can trigger this response and lead to an allergic reaction . A type of antibody called immunoglobulin E ( IgE ) is particularly likely to trigger an allergic response . In general , immune cells called plasma cells produce antibodies and release them into the body . However , in B cells – the cells from which plasma cells develop – the antibodies remain on the surface of the cells . Here , the antibody acts as a “receptor” that allows the B cell to tell when its antibody has bound to a specific substance . Generally , B cells only activate when their B cell receptors bind to a specific substance . This binding triggers signals inside the cell that determine its fate – such as whether it will develop into a plasma cell . Recent studies have shown that B cells that have IgE on their surface ( IgE+ B cells ) are predisposed to develop rapidly into plasma cells . To investigate why this is the case , Yang et al . have now studied B cells both in cell culture and in mice . The results show that the IgE B cell receptor autonomously signals to the cell even when it is not bound to a specific substance , in a manner that differs from other types of B cell receptors . This increases the likelihood that the IgE+ B cell will develop into a plasma cell and limits the competitive fitness of IgE+ B cells . These findings provide new insights into how IgE responses are regulated by the B cell receptor . The next step will be to determine , at a molecular level , the basis for the autonomous signaling produced by the IgE B cell receptor when it is not bound to a specific substance . It will then be possible to investigate how this mechanism compares with the way that signals are normally transmitted when a B cell receptor binds to a specific substance . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2016 | Regulation of B cell fate by chronic activity of the IgE B cell receptor |
Advances in fluorescence microscopy enable monitoring larger brain areas in-vivo with finer time resolution . The resulting data rates require reproducible analysis pipelines that are reliable , fully automated , and scalable to datasets generated over the course of months . We present CaImAn , an open-source library for calcium imaging data analysis . CaImAn provides automatic and scalable methods to address problems common to pre-processing , including motion correction , neural activity identification , and registration across different sessions of data collection . It does this while requiring minimal user intervention , with good scalability on computers ranging from laptops to high-performance computing clusters . CaImAn is suitable for two-photon and one-photon imaging , and also enables real-time analysis on streaming data . To benchmark the performance of CaImAn we collected and combined a corpus of manual annotations from multiple labelers on nine mouse two-photon datasets . We demonstrate that CaImAn achieves near-human performance in detecting locations of active neurons .
Here we present CaImAn , an open source pipeline for the analysis of both two-photon and one-photon calcium imaging data . CaImAn includes algorithms for both offline analysis ( CaImAn batch ) where all the data is processed at once at the end of each experiment , and online analysis on streaming data ( CaImAn online ) . Moreover , CaImAn requires very moderate computing infrastructure ( e . g . , a personal laptop or workstation ) , thus providing automated , efficient , and reproducible large-scale analysis on commodity hardware . Our contributions can be roughly grouped in three different directions: Methods: CaImAn batch improves on the scalability of the source extraction problem by employing a MapReduce framework for parallel processing and memory mapping which allows the analysis of datasets larger than would fit in RAM on most computer systems . It also improves on the qualitative performance by introducing automated routines for component evaluation and classification , better handling of neuropil contamination , and better initialization methods . While these benefits are here presented in the context of the widely used CNMF algorithm of Pnevmatikakis et al . ( 2016 ) , they are in principle applicable to any matrix factorization approach . CaImAn online improves and extends the OnACID prototype algorithm ( Giovannucci et al . , 2017 ) by introducing , among other advances , new initialization methods and a convolutional neural network ( CNN ) based approach for detecting new neurons on streaming data . Our analysis on in vivo two-photon and light-sheet imaging datasets shows that CaImAn online approaches human-level performance and enables novel types of closed-loop experiments . Apart from these significant algorithmic improvements CaImAn includes several useful analysis tools such as , a MapReduce and memory-mapping compatible implementation of the CNMF-E algorithm for one-photon microendoscopic data ( Zhou et al . , 2018 ) , a novel efficient algorithm for registration of components across multiple days , and routines for segmentation of structural ( static ) channel information which can be used for component seeding . Software: CaImAn is a complete open source software suite implemented primarily in Python , and is already widely used by , and has received contributions from , its community . It contains efficient implementations of the standard analysis pipeline steps ( motion correction - source extraction - deconvolution - registration across different sessions ) , as well as numerous other features . Much of the functionality is also available in a separate MATLAB implementation . Data: We benchmark the performance of CaImAn against a previously unreleased corpus of manually annotated data . The corpus consists of 9 mouse in vivo two-photon datasets . Each dataset is manually annotated by 3–4 independent labelers that were instructed to select active neurons in a principled and consistent way . In a subsequent stage , the annotations were combined to create a ‘consensus’ annotation , that is used to benchmark CaImAn , to train supervised learning based classifiers , and to quantify the limits of human performance . The manual annotations are released to the community , providing a valuable tool for benchmarking and training purposes . Before presenting the new analysis features introduced with this work , we overview the analysis pipeline that CaImAn uses and builds upon .
We compared the performance of each human annotator against a consensus annotation . The performance was quantified with a precision/recall framework and the results of the performance of each individual labeler against the consensus annotation for each dataset is given in Table 1 . The range of human performance in terms of F1 score was 0 . 69–0 . 94 . All annotators performed similarly on average ( 0 . 84 ± 0 . 05 , 0 . 87 ± 0 . 07 , 0 . 85 ± 0 . 06 , 0 . 83 ± 0 . 08 ) . We also ensured that the performance of labelers was stable across time ( i . e . their learning curve plateaued , data not shown ) . As shown in Table 1 ( see also Figure 4b ) the F1 score was never 1 , and in most cases it was less or equal to 0 . 9 , demonstrating significant variability between annotators . Figure 3 ( bottom ) shows an example of matches and mismatches between individual labelers and consensus annotation for dataset K53 , where the level of agreement was relatively high . The high degree of variability between human responses indicates the challenging nature of the source extraction problem and raises reproducibility concerns in studies relying heavily on manual ROI selection . This process may have generated slightly biased results in favor of each individual annotator as the consensus annotation is always a subset of the union of the individual annotations . We also used an alternative cross-validation approach , where the labels of each annotator were compared with the combined results of the remaining annotators . The combination was constructed using a majority vote when a dataset was labeled from 4 annotators , or an intersection of selections when a dataset was labeled by 3 . The results ( see Table 3 in Materials and methods ) indicate an even higher level of disagreement between the annotators with lower average F1 score 0 . 82 ± 0 . 06 ( mean ± STD ) and range of values 0 . 68-0 . 90 . More details are given in Materials and methods ( Cross-Validation analysis of manual annotations ) . We first benchmarked CaImAn batch and CaImAn online against consensus annotation for the task of identifying neurons locations and their spatial footprints , using the same precision recall framework ( Table 1 ) . Figure 4a shows an example dataset ( K53 ) along with neuron-wise matches and mismatches between CaImAn batch vs consensus annotation ( top ) and CaImAn online vs consensus annotation ( bottom ) . The results indicate a similar performance between CaImAn batch and CaImAn online; CaImAn batch has F1 scores in the range 0 . 69–0 . 78 and average performance 0 . 75 ± 0 . 03 ( mean ± STD ) . On the other hand CaImAn online had F1 scores in the range 0 . 70–0 . 82 and average performance 0 . 76 ± 0 . 05 . While the two algorithms performed similarly on average , CaImAn online tends to perform better for longer datasets ( e . g . , datasets J115 , J123 , K53 that all have more than 40000 frames; see also Table 2 for characteristics of the various datasets ) . CaImAn batch operates on the entire dataset at once , representing each spatial footprint with a constant in time vector . In contrast , CaImAn online operates at a local level looking at a short window over time to detect new components , while adaptively changing their spatial footprint based on new data . This enables CaImAn online to adapt to slow non-stationarities that can appear in long experiments . CaImAn approaches but is in most cases below the accuracy levels of human annotators ( Figure 4b ) . We attribute this to two primary factors: First , CNMF detects active components regardless of their shape , and can detect non-somatic structures with significant transients . While non-somatic components can be filtered out to some extent using the CNN classifier , their existence degrades performance compared to the manual annotations that consist only of neurons . Second , to demonstrate the generality and ease of use of our tools , the results presented here are obtained by running CaImAn batch and CaImAn online with exactly the same parameters for each dataset ( see Materials and methods ( Implementation details ) ) : fine-tuning to each individual dataset can significantly increase performance ( Figure 4b ) . To test the later point we measured the performance of CaImAn online on the nine datasets , as a function of 3 parameters: ( i ) the trace SNR threshold for testing the traces of candidate components , ( ii ) the CNN threshold for testing the shapes of candidate components , and ( iii ) the number of candidate components to be tested at each frame ( more details can be found in Materials and methods ( Implementation details for CaImAn online ) ) . By choosing a parameter combination that maximizes the value for each dataset , the performance generally increases across the datasets with F1 scores in the range 0 . 72–0 . 85 and average performance 0 . 78±0 . 05 ( see Figure 4 ( orange ) and Figure 4—figure supplement 1 ( magenta ) ) . This analysis also shows that in general a strategy of testing a large number of components per timestep but with stricter criteria , achieves better results than testing fewer components with looser criteria ( at the expense of increased computational cost ) . The results also indicate different strategies for parameter choice depending on the length of a dataset: Lower threshold values and/or larger number of candidate components ( Figure 4—figure supplement 1 ( red ) ) , lead to better values for shorter datasets , but can decrease precision and overall performance for longer datasets . The opposite also holds for higher threshold values and/or smaller number of candidate components ( Figure 4—figure supplement 1 ( blue ) ) , where CaImAn online for shorter datasets can suffer from lower recall values , whereas in longer datasets CaImAn online can add neurons over a longer period of time while maintaining high precision values and thus achieve better performance . A similar grid search was also performed for the CaImAn batch algorithm where four parameters of the component evaluation step ( space correlation , trace SNR , min/max CNN thresholds ) were optimized individually to filter out false positives . This procedure led to F1 scores in in the range 0 . 71–0 . 81 and average performance 0 . 774±0 . 034 ( Figure 4 ( red ) ) . We also compared the performance of CaImAn against Suite2p ( Pachitariu et al . , 2017 ) , another popular calcium imaging data analysis package . By using a small grid search around some default parameters of Suite2p we extracted the set of parameters that worked better in the eight datasets where the algorithm converged ( in the dataset J123 Suite2p did not converge ) . CaImAn outperformed Suite2p in all datasets with the latter obtaining F1 scores in the range 0 . 41–0 . 75 , with average performance 0 . 55±0 . 12 . More details about the comparison are shown in Figure 4—figure supplement 2 and Materials and methods ( Comparison with Suite2p ) . Testing the quality of the inferred traces is more challenging due to the unavailability of ground truth data in the context of large scale in vivo recordings . As mentioned above , we defined as ‘ground truth’ the traces obtained by running the CNMF algorithm seeded with the binary masks obtained by consensus annotation procedure . After spatial alignment with the results of CaImAn , the matched traces were compared both for CaImAn batch and for CaImAn online . Figure 5a , shows an example of 5 of these traces for the dataset K53 , showing very similar behavior of the traces in these three different cases . To quantify the similarity we computed the correlation coefficients of the traces ( consensus vs CaImAn batch , and consensus vs CaImAn online ) for all the nine datasets ( Figure 5b–c ) . Results indicated that for all but one dataset ( Figure 5b ) CaImAn batch reproduced the traces with higher fidelity , and in all cases the mean correlation coefficients was higher than 0 . 9 , and the empirical histogram of correlation coefficients peaked at the maximum bin 0 . 99–1 ( Figure 5c ) . The results indicate that the batch approach extracts traces closer to the consensus traces . This can be attributed to a number of reasons: By processing all the time points simultaneously , the batch approach can smooth the trace estimation over the entire time interval as opposed to the online approach where at each timestep only the information up to that point is considered . Moreover , CaImAn online might not detect a neuron until it becomes strongly active . This neuron’s activity before detection is unknown and has a default value of zero , resulting in a lower correlation coefficient . While this can be ameliorated to a great extent with additional passes over the data , the results indicate the trade-offs inherent between online and batch algorithms . We tested CaImAn online with a 380 GB whole brain dataset of larval zebrafish ( Danio rerio ) acquired with a light-sheet microscope ( Kawashima et al . , 2016 ) . The imaged transgenic fish ( Tg ( elavl3:H2B-GCaMP6f ) jf7 ) expressed the genetically encoded calcium indicator GCaMP6f in almost all neuronal nuclei . Data from 45 planes ( FOV 820 × 410 μm2 , spaced at 5 . 5 μm intervals along the dorso-ventral axis ) was collected at 1 Hz for 30 min ( for details about preparation , equipment and experiment refer to Kawashima et al . ( 2016 ) ) . With the goal of simulating real-time analysis of the data , we run all the 45 planes in parallel on a computing cluster with nine nodes ( each node is equipped with 24 CPUs and 128–256 GB RAM , Linux CentoOS ) . Data was not stored locally in each machine but directly accessed from a network drive . The algorithm was initialized with CaImAn batch run on 200 initial frames and looking for 500 components . The small number of frames ( 1885 ) and the large FOV size ( 2048×1188 pixels ) for this dataset motivated this choice of increased number of components during initialization . In Figure 6 we report the results of the analysis for plane number 11 of 45 . For plane 11 , CaImAn online found 1524 neurons after processing 1685 frames . Since no ground truth was available for this dataset , it was only possible to evaluate the performance of this algorithm by visual inspection . CaImAn online identified all the neurons with a clear footprint in the underlying correlation image ( higher SNR , Figure 6a ) and missed a small number of the fainter ones ( low SNR ) . By visual inspection of the components the authors could find very few false positives . Given that the parameters were not tuned and that the classifier was not trained on zebrafish neurons , we hypothesize that the algorithm is biased towards a high precision result . Spatial components displayed the expected morphological features of neurons ( Figure 6b–c ) . Considering all the planes ( Figure 6e and Figure 6—figure supplement 1 ) CaImAn online was able to identify in a single pass of the data a total of 66108 neurons . See Video 1 for a summary across all planes . The analysis was performed in 21 min , with the first three minutes spent in initialization , and the remaining 18 in processing the data in streaming mode ( and in parallel for each plane ) . This demonstrates the ability of CaImAn online to process large amounts of data in real-time ( see also Figure 8 for a discussion of computational performance ) . We tested the CNMF-E implementation of CaImAn batch on in vivo microendosopic data from mouse dorsal striatum , with neurons expressing GCaMP6f . 6000 frames were acquired at 30 frames per second while the mouse was freely moving in an open field arena ( for further details refer to Zhou et al . , 2018 ) . In Figure 7 we report the results of the analysis using CaImAn batch with patches and compare to the results of the MATLAB implementation of Zhou et al . ( 2018 ) . Both implementations detect similar components ( Figure 7a ) with an F1-score of 0 . 89 . 573 neurons were found in common by both implementations . 106 and 31 additional components were detected by Zhou et al . ( 2018 ) and CaImAn batch respectively . The median correlation between the temporal traces of neurons detected by both implementations was 0 . 86 . Similar results were also obtained by running CaImAn batch without patches . Ten example temporal traces are plotted in Figure 7b . We examined the performance of CaImAn in terms of processing time for the various analyzed datasets presented above ( Figure 8 ) . The processing time discussed here excludes motion correction , which is highly efficient and primarily depends on the level of the FOV discretization for non-rigid motion correction ( Pnevmatikakis and Giovannucci , 2017 ) . For CaImAn batch , each dataset was analyzed using three different computing architectures: ( i ) a single laptop ( MacBook Pro ) with 8 CPUs ( Intel Core i7 ) and 16 GB of RAM ( blue in Figure 8a ) , ( ii ) a linux-based workstation ( CentOS ) with 24 CPUs ( Intel Xeon CPU E5-263 v3 at 3 . 40 GHz ) and 128 GB of RAM ( magenta ) , and ( iii ) a linux-based HPC cluster ( CentOS ) where 112 CPUs ( Intel Xeon Gold 6148 at 2 . 40 GHz , four nodes , 28 CPUs each ) were allocated for the processing task ( yellow ) . Figure 8a shows the processing of CaImAn batch as a function of dataset size on the four longest datasets , whose size exceeded 8 GB , on log-log plot . Results show that , as expected , employing more processing power results in faster processing . CaImAn batch on a HPC cluster processes data faster than acquisition time ( Figure 8a ) even for very large datasets . Processing of an hour long dataset was feasible within 3 hr on a single laptop , even though the size of the dataset is several times the available RAM . Here , acquisition time is computed based on the assumption of imaging a FOV discretized over a 512×512 grid at a 30 Hz rate ( a typical two-photon imaging setup with resonant scanning microscopes ) . Dataset size is computed by representing each measurement using single precision arithmetic , which is the minimum precision required for standard algebraic processing . These assumptions lead to a data rate of ∼105 GB/hr . In general performance scales linearly with the number of frames ( and hence , the size of the dataset ) , but we also observe a dependency on the number of components , which during the solution refinement step can be quadratic . This is expected from the properties of the matrix factorization approach as also noted by past studies ( Pnevmatikakis et al . , 2016 ) . The majority of the time ( Figure 8b ) required for CaImAn batch processing is taken by CNMF algorithmic processing either during the initialization in patches ( orange bar ) or during merging and refining the results of the individual patches ( green bar ) . To study the effects of parallelization we ran CaImAn batch several times on the same hardware ( linux-based workstation with 24CPUs ) , limiting the runs to different numbers of CPUs each time ( Figure 8c ) . In all cases we saw significant performance gains from parallel processing , with the gains being similar for all stages of processing ( patch processing , refinement , and quality testing , data not shown ) . We saw the most effective scaling with our 50 G dataset ( J123 ) . For the largest datasets ( J115 , ∼100GB ) , the speedup reaches a plateau due to limited available RAM , suggesting that more RAM can lead to better scaling . For small datasets ( ∼5 GB ) the speedup factor is limited by increased communications overhead ( indicative of weak scaling in the language of high performance computing ) . The cost of processing 1p data in CaImAn batch using the CNMF-E algorithm ( Zhou et al . , 2018 ) is shown ( Figure 8d ) for our workstation-class hardware . Splitting in patches and processing in parallel can lead to computational gains at the expense of increased memory usage . This is because the CNMF-E introduces a background term that has the size of the dataset and needs to be loaded and updated in memory in two copies . This leads to processing times that are slower compared to the standard processing of 2 p datasets , and higher memory requirements . However ( 8 c ) , memory usage can be controlled enabling scalable inference at the expense of slower processing speeds . Figure 8a also shows the performance of CaImAn online ( red markers ) . Because of the low memory requirements of the streaming algorithm , this performance only mildly depends on the computing infrastructure , allowing for near real-time processing speeds on a standard laptop ( Figure 8a ) . As discussed in Giovannucci et al . , 2017 processing time of CaImAn online depends primarily on ( i ) the computational cost of tracking the temporal activity of discovered neurons , ( ii ) the cost of detecting and incorporating new neurons , and ( iii ) the cost of periodic updates of spatial footprints . Figure 8e shows the cost of each of these steps for each frame , for one epoch of processing of the dataset J123 . Distributing the spatial footprint update more uniformly among all frames removes the computational bottleneck appearing in Giovannucci et al . , 2017 , where all the footprints where updated periodically at the same frame . The cost of detecting and incorporating new components remains approximately constant across time , and is dependent on the number of candidate components at each timestep . In this example five candidate components were used per frame resulting in a relatively low cost ( ∼7 ms per frame ) . A higher number of candidate components can lead to higher recall in shorter datasets but at a computational cost . This step can benefit by the use of a GPU for running the online CNN on the footprints of the candidate components . Finally , as noted in Giovannucci et al . , 2017 , the cost of tracking components can be kept low , and increases mildly over time as more components are found by the algorithm ( the analysis here excludes the cost of motion correction , because the files where motion corrected before hand to ensure that manual annotations and the algorithms where operating on the same FOV . This cost depends on whether rigid or pw-rigid motion correction is being used . Rigid motion correction taking on average 3–5 ms per frame for a 512×512 pixel FOV , whereas pw-rigid motion correction with patch size 128×128 pixel is typically 3–4 times slower ) . Figure 8f shows the overall processing speed ( in frames per second ) for CaImAn online for the nine annotated datasets . Apart from the number of neurons , the processing speed also depends on the size of the imaged FOV and the use of spatial downsampling . Datasets with smaller FOV ( e . g . , YST ) or datasets where spatial downsampling is used can achieve higher processing speeds for the same amount of neurons ( blue dots in Figure 8f ) as opposed to datasets where no spatial downsampling is used ( orange dots in Figure 8f ) . In most cases , spatial downsampling can be used to increase processing speed without significantly affecting the quality of the results , an observation consistent with previous studies ( Friedrich et al . , 2017a ) . In Figure 8g the cost per frame is plotted for the analysis of the whole brain zebrafish recording . The lower imaging rate ( 1 Hz ) allows for tracing of neural activity with computational costs significantly lower than the 1 s between volume imaging time ( Figure 8e ) , even in the presence of a large number of components ( typically more than 1000 per plane , Figure 6 ) and the significantly larger FOV ( 2048×1188 pixels ) . Figure 9 shows an example of tracking neurons across six different sessions corresponding to six different days of mouse cortex in vivo data using our multi-day registration algorithm RegisterMulti ( see Materials and methods , Algorithm 8 ) . 453 , 393 , 375 , 378 , 376 , and 373 active components were found in the six sessions , respectively . Our tracking method detected a total of 686 distinct active components . Of these , 172 , 108 , 70 , 92 , 82 , and 162 appeared in exactly 1 , 2 , 3 , 4 , 5 , and all six sessions respectively . Contour plots of the 162 components that appeared in all sessions are shown in Figure 9a , and parts of the FOV are highlighted in Figure 9d showing that components can be tracked in the presence of non-rigid deformations of the FOV between the different sessions . To test the stability of RegisterMulti for each subset of sessions , we repeated the same procedure running backwards in time starting from day 6 and ending at day 1 , a process that also generated a total of 686 distinct active components . We identified the components present in at least a given subset of sessions when using the forward pass , and separately when using the backwards pass , and compared them against each other ( Figure 9b ) for all possible subsets . Results indicate a very high level of agreement between the two approaches with many of the disagreements arising near the boundaries ( data not shown ) . Disagreements near the boundaries can arise because the forward pass aligns the union with the FOV of the last session , whereas the backwards pass with the FOV of the first session , potentially leading to loss of information near the boundaries . A step by step demonstration of the tracking algorithm for the first three sessions is shown in Figure 9—figure supplement 1 . Our approach allows for the comparison of two non-consecutive sessions through the union of components without the need of a direct pairwise registration ( Figure 9—figure supplement 1f ) , where it is shown that registering sessions 1 and 3 directly and through the union leads to nearly identical results . Figure 9c compares the registrations for all pairs of sessions using the forward ( red ) or the backward ( blue ) approach , with the direct pairwise registrations . Again , the results indicate a very high level of agreement , indicating the stability and effectiveness of the proposed approach . A different approach for multiple day registration was recently proposed by Sheintuch et al . ( 2017 ) ( CellReg ) . While a direct comparison of the two methods is not feasible in the absence of ground truth , we tested our method against the same publicly available datasets from the Allen Brain Observatory visual coding database . ( http://observatory . brain-map . org/visualcoding ) . Similarly to Sheintuch et al . ( 2017 ) the same experiment performed over the course of different days produced very different populations of active neurons . To measure performance of RegisterPair for pairwise registration , we computed the transitivity index proposed in Sheintuch et al . ( 2017 ) . The transitivity property requires that if cell 'a’ from session one matches with cell 'b’ from session 2 , and cell 'b’ from session two matches with cell 'c’ from session 3 , then cell 'a’ from session one should match with cell 'c’ from session 3 when sessions 1 and 3 are registered directly . For all ten tested datasets the transitivity index was very high , with values ranging from 0 . 976 to 1 ( 0 . 992±0 . 006 , data not shown ) . A discussion between the similarities and differences of the two methods is given in Materials and methods .
Significant advances in the reporting fidelity of fluorescent indicators , and the ability to simultaneously record and modulate neurons granted by progress in optical technology , have made calcium imaging one of the two most prominent experimental methods in systems neuroscience alongside electrophysiology recordings . Increasing adoption has led to an unprecedented wealth of imaging data which poses significant analysis challenges . CaImAn is designed to provide the experimentalist with a complete suite of tools for analyzing this data in a formal , scalable , and reproducible way . The goal of this paper is to present the features of CaImAn and examine its performance in detail . CaImAn embeds existing methods for preprocessing calcium imaging data into a MapReduce framework and augments them with supervised learning algorithms and validation metrics . It builds on the CNMF algorithm of Pnevmatikakis et al . ( 2016 ) for source extraction and deconvolution , extending it along the lines of ( i ) reproducibility and performance improvement , by automating quality assessment through the use of unsupervised and supervised learning algorithms for component detection and classification , and ( ii ) scalability , by enabling fast large scale processing with standard computing infrastructure ( e . g . , a commodity laptop or workstation ) . Scalability is achieved by either using a MapReduce batch approach , which employs parallel processing of spatially overlapping , memory mapped , data patches; or by integrating the online processing framework of Giovannucci et al . , 2017 within our pipeline . Apart from computational gains both approaches also result in improved performance . Towards our goal of providing a single package for dealing with standard problems arising in analysis of imaging data , CaImAn also includes an implementation of the CNMF-E algorithm of Zhou et al . ( 2018 ) for the analysis of microendoscopic data , as well as a novel method for registering analysis results across multiple days . To evaluate the performance of CaImAn batch and CaImAn online , we used a number of distinct labelers to generate a corpus of nine annotated two-photon imaging datasets . The results indicated a surprising level of disagreement between individual labelers , highlighting both the difficulty of the problem , and the non-reproducibility of the laborious task of human annotation . CaImAn reached near-human performance with respect to this consensus annotation , by using the same parameters for all the datasets without dataset dependent parameter tweaking . Such tweaking can include setting the SNR threshold based on the noise level of the recording , the complexity of the neuropil signal based on the level of background activity , or specialized treatment around the boundaries of the FOV to compensate for eventual imaging artifacts , and as shown can significantly improve the results on individual datasets . As demonstrated in our results , optimal parameter setting for CaImAn online can also depend on the length of the experiment with stricter parameters being more suitable for longer datasets . We plan to investigate parameter schemes that increase in strictness over the course of an experiment . CaImAn has higher precision than recall when run on most datasets . While more balanced results can be achieved by appropriately relaxing the relevant quality evaluation thresholds , we prefer to maintain a higher precision as we believe that the inclusion of false positive traces can be more detrimental in any downstream analysis compared to the exclusion of , typically weak , true positive traces . This is true especially in experiments with low task dimensionality where a good signal from few neurons can be sufficient for the desired hypothesis testing . Apart from being used as a benchmarking tool , the set of manual annotations can also be used as labeled data for supervised learning algorithms . CaImAn uses two CNN based classifiers trained on ( a subset of ) this data , one for post processing component classification in CaImAn batch , and the other for detecting new neurons in residual images in the CaImAn online . The deployment of these classifiers resulted in significant gains in terms of performance , and we expect further advances in the future . The annotations are made freely available to the community for benchmarking and training purposes . Our results suggest similar performance between CaImAn batch and CaImAn onine when evaluated on the basis of processing speed and quality of results , with CaImAn online outperforming CaImAn batch on longer datasets in terms of neuron detection , possibly due to its inherent ability to adapt to non-stationarities arising during the course of a large experiment . By contrast , CaImAn batch extracts better traces compared to CaImAn online with respect to the traces derived from the consensus annotations . While multiple passes over the data with CaImAn online can mitigate these shortcomings , this still depends on good initialization with CaImAn batch , as the analysis of the whole brain zebrafish dataset indicates . In offline setups , CaImAn onine could also benefit from the post processing component evaluation tools used in batch mode . for example using the batch classifier for detecting false positive components at the end of the experiment . CaImAn online differs from CaImAn batch in that the former has lower memory requirements and it can support novel types of closed-loop all-optical experiments ( Packer et al . , 2015; Carrillo-Reid et al . , 2017 ) . As discussed in Giovannucci et al . , 2017 , typical all-optical closed-loop experiments require the pre-determination of ROIs that are monitored/modulated . Indeed , CaImAn online allows identification and modulation of new neurons on the fly , greatly expanding the space of possible experiments . Even though our simulated online processing setup is not integrated with hardware to an optical experimental setup , our results indicate thatCaImAn online performed close to real-time in most cases . Real time can be potentially achieved by using parallel computational streams for the three steps of frame processing ( motion correction and tracking , detecting new neurons , updating shapes ) , since these steps can be largely run in an asynchronous mode independently . This suggests that large scale closed-loop experiments with single cell resolution are feasible by combining existing all-optical technology and our proposed analysis method . While CaImAn uses a highly scalable processing pipeline for two-photon datasets , processing of one-photon microendoscopic imaging data is less scalable due to the more complex background model that needs to be retained in memory during processing . Adapting CaImAn online to the one-photon data processing algorithm of Zhou et al . ( 2018 ) is a promising way for scaling up efficient processing in this case . The continuing development and quality improvement of neural activity indicators has enabled direct imaging of neural processes ( axons/dendrites ) , imaging of synaptic activity ( Xie et al . , 2016 ) , or direct imaging of voltage activity in vivo conditions ( Piatkevich et al . , 2018 ) . While the approach presented here is tuned for somatic imaging through the use of various assumptions ( space localized activity , CNN classifiers trained on images of somatic activity ) , the technology of CaImAn is largely transferable to these domains as well . We will pursue these extensions in future work .
CaImAn batch uses memory mapping for efficient parallel data access . With memory mapped arrays , arithmetic operations can be performed on data residing on the hard drive without explicitly loading it to RAM , and slices of data can be indexed and accessed without loading the full file in memory , enabling out-of-core processing ( Toledo , 1999 ) . On modern computers tensors are stored in linear format , no matter the number of the array dimensions . Therefore , one has to decide which elements of an array are contiguous in memory: in row-major order , consecutive elements of a row ( first-dimension ) are next to each other , whereas in column-major order consecutive elements of a column ( last dimension ) are contiguous . Such decisions significantly affect the speed at which data is read or written on spinning disks ( and to a lesser degree on solid state drives ) : in column-major order reading a full column is fast because memory is read in a single sequential block , whereas reading a row is inefficient since only one element can be read at a time and all the data needs to be accessed . In the context of calcium imaging datasets , CaImAn batch represents the datasets in a matrix form Y , where each row corresponds to a different imaged pixel , and each column to a different frame . As a result , a column-major order mmap file enables the fast access of individual frames at a given time , whereas a row-major order files enables the fast access of an individual pixel at all times . To facilitate processing in patches CaImAn batch stores the data in row-major order . In practice , this is opposite to the order with which the data appears , one frame at a time . In order to reduce memory usage and speed up computation CaImAn batch employs a MapReduce approach , where either multiple files or multiple chunks of a big file composing the original datasets are processed and saved in mmap format in parallel . This operation includes two phases , first the chunks/files are saved ( possibly after motion correction , if required ) in multiple row-major mmap format , and then chunks are simultaneously combined into a single large row-major mmap file . The CNMF framework ( Figure 1d ) for calcium imaging data representation can be expressed in mathematical terms as ( Pnevmatikakis et al . , 2016 ) ( 1 ) Y=AC+B+E . Here , Y∈Rd×T denotes the observed data written in matrix form , where d is the total number of observed pixels/voxels , and T is the total number of observed timesteps ( frames ) . A∈Rd×N denotes the matrix of the N spatial footprints , A=[𝐚1 , 𝐚2 , … , 𝐚N] , with ai∈Rd×1 being the spatial footprint of component i . C∈RN×T denotes the matrix of temporal components , C=[𝐜1 , 𝐜2 , … , 𝐜N]⊤ , with ci∈RT×1 being the temporal trace of component i . B is the background/neuropil activity matrix . For two-photon data it is modeled as a low rank matrix B=𝐛𝐟 , where b∈Rd×nb , f∈Rnb×T correspond to the matrices of spatial and temporal background components , and nb is the number of background components . For the case of micro-endoscopic data the integration volume is much larger and the low rank model is inadequate . For this we use the CNMF-E algorithm of Zhou et al . ( 2018 ) where the background is modeled as ( 2 ) B=W ( Y-AC ) , where W∈Rd×d is an appropriate weight matrix , where the ( i , j ) entry models the influence of the neuropil signal of pixel j to the neuropil signal at pixel i . To combine results from different patches we first need to account for the overlap at the boundaries . Neurons lying close to the boundary between neighboring patches can appear multiple times and must be merged . With this goal , we optimized the merging approach used in Pnevmatikakis et al . ( 2016 ) : Groups of components with spatially overlapping footprints whose temporal traces are correlated above a threshold are replaced with a single component , that tries to explain as much of the variance already explained by the ‘local’ components ( as opposed to the variance of the data as performed in Pnevmatikakis et al . ( 2016 ) ) . If Aold , Cold are the matrices of components to be merged , then the merged component 𝐚m , 𝐜m are given by the solution of the rank-1 NMF problem: ( 3 ) minam≥0 , cm≥0‖AoldCold−amcm⊤‖ . Prior to merging , the value of each component at each pixel is normalized by the number of patches that overlap in this pixel , to avoid counting the activity of each pixel multiple times . We follow a similar procedure for the background/neuropil signals from the different patches . When working with two-photon data , the spatial background/neuropil components for each patch can be updated by keeping their spatial extent intact to retain a local neuropil structure , or they can be merged when they are sufficiently correlated in time as described above to promote a more global structure . For the case of one-photon data , CNMF-E estimates the background using a local autoregressive process ( see Equation 2 ) ( Zhou et al . , 2018 ) , a setup that cannot be immediately propagated when combining the different patches . To combine backgrounds from the different patches , we first approximate the backgrounds Bi from all the patches i with a low rank matrix using non-negative matrix factorization of rank gb to obtain global spatial , and temporal background components . ( 4 ) [bi , fi]=NNMF ( Bi , gb ) . The resulting components are embedded into a large matrix B∈Rd×T that retains a low rank structure . After the components and backgrounds from all the patches have been combined , they are further refined by running CNMF iteration of updating spatial footprints , temporal traces , and neuropil activity . CaImAn batch implements these steps in parallel ( as also described in Pnevmatikakis et al . ( 2016 ) ) : Temporal traces whose corresponding spatial traces do not overlap can be updated in parallel . Similarly , the rows of the matrix of spatial footprints A can also be updated in parallel ( Figure 2b ) . The process is summarized in algorithmic format in Algorithms 1–2 . When working with one-photon data , instead of producing a low-rank approximation of B that would underfit the background , we increase patch overlap and run the full pipeline on each patch . In the final phase , when neurons overlap we retain only the variant with the highest quality rather than merging them . Source extraction using matrix factorization requires solving a bi-convex problem where initialization plays a critical role . The CNMF/CNMF-E algorithms use initialization methods that exploit the locality of the spatial footprints to efficiently identify the locations of candidate components ( Pnevmatikakis et al . , 2016; Zhou et al . , 2018 ) . CaImAn incorporates these methods , extending them by using the temporal locality of the calcium transient events . The available initialization methods for CaImAn batch include: GreedyROI: This approach , introduced in Pnevmatikakis et al . ( 2016 ) , first spatially smooths the data with a Gaussian kernel of size comparable to the average neuron radius , and then initializes candidate components around locations where maximum variance ( of the smoothed data ) is explained . This initialization strategy is fast but requires manual specification of the number of components by the user . RollingGreedyROI: The approach , introduced in this paper , operates like GreedyROI by spatially smoothing the data and looking for points of maximum variance . Instead of working across all the data , RollingGreedyROI looks for points of maximum variance on a rolling window of a fixed duration , for example 3 s , and initializes components by performing a rank one NMF on a local spatial neighborhood . By focusing into smaller rolling windows , RollingGreedyROI can better isolate single transient events , and as a result detect better neurons with sparse activity . RollingGreedyROI is the default choice for processing of 2-photon data . GreedyCorr: This approach , introduced in Zhou et al . ( 2018 ) , initializes candidate components around locations that correspond to the local maxima of an image formed by the pointwise product between the correlation image and the peak signal-to-noise ratio image . A threshold for acceptance of candidate neurons is used , making it unnecessary to pre-specify the neuron count . This comes at the expense of a higher computational cost . GreedyCorr is the default choice for processing of one-photon data . SparseNMF: Sparse NMF approaches , when ran in small patches , can be effective for quickly uncovering spatial structure in the imaging data , especially for neural processes ( axons/dendrites ) whose shape cannot be easily parametrized and/or localized . SeededInitialization: Often locations of components are known either from manual annotation or from labeled data obtained in a different way , such as data from a static structural channel recorded concurrently with the functional indicator . CaImAn can be seeded with binary ( or real valued ) masks for the spatial footprints . Apart from A , these masks can be used to initialize all the other relevant matrices C and B as well . This is performed by ( i ) first estimating the temporal background components 𝐟 using only data from parts of the FOV not covered by any masks and , ( ii ) then estimating the spatial background components 𝐛 , and then estimating A , C ( with A restricted to be non-zero only at the locations of the binary masks ) , using a simple NMF approach . Details are given in Algorithm 3 . Here we present the unsupervised and supervised quality assessment tests in more detail ( Figure 2 ) . We collected manual annotations from four independent labelers who were instructed to find round or donut shaped neurons of similar size using the ImageJ Cell Magic Wand tool ( Walker , 2014 ) . We focused on manually annotating only cells that were active within each dataset and for that reason the labelers were provided with two summary statistics: ( i ) A movie obtained by removing a running 20th percentile ( as a crude background approximation ) and downsampling in time by a factor of 10 , and ( ii ) the max-correlation image . The correlation image ( CI ) at every pixel is equal to the average temporal correlation coefficient between that pixel and its neighbors ( Smith and Häusser , 2010 ) ( eight neighbors were used for our analysis ) . The max-correlation image is obtained by computing the CI for each batch of 33 s ( 1000 frames for a 30 Hz acquisition rate ) , and then taking the maximum over all these images ( Figure 3—figure supplement 1a ) . Neurons that are inactive during the course of the dataset will be suppressed both from the baseline removed video ( since their activity will always be around their baseline ) , and from the max-correlation image since the variation around this baseline will mostly be due to noise leading to practically uncorrelated neighboring pixels ( Figure 3—figure supplement 1a ) . Nine different mouse in vivo datasets were used from various brain areas and labs . A description is given in Table 2 . To create the final consensus , the labelers were asked to jointly resolve the inconsistencies between their annotations . To this end , every ROI selected by at least one but not all labelers was re-considered by a group of at least two labelers that decided whether it corresponds to an active neuron or not . The annotation procedure provides a binary mask per selected component . On the other hand , the output of for each component is a non-negatively valued vector over the FOV ( a real-valued mask ) . The two sets of masks differ not only in their variable type but also in their general shape: Manual annotation through the Cell Magic Wand tool tends to produce circular shapes , whereas the output of CaImAn will try to accurately estimate the shape of each active component ( Figure 3—figure supplement 1b ) . To construct the consensus components that can be directly used for comparison , the binary masks from the manual annotations were used to seed the CNMF algorithm ( Algorithm 3 ) . This produced a set of real valued components with spatial footprints restricted to the areas provided by the annotations , and a corresponding set of temporal components that can be used to evaluate the performance of CaImAn ( Figure 4 ) . Registration was performed using the RegisterPair algorithm ( Algorithm 7 ) and match was counted as a true positive when the ( modified ) Jaccard distance ( Equation 11 ) was below 0 . 7 . Details of the registration procedure are given below ( see Component registration ) . As mentioned in the results section , comparing each manual annotation with the consensus annotation can create slightly biased results in favor of individual annotators since the consensus annotation is chosen from the union of individual annotations . To correct for this we performed a cross-validation analysis where the annotations of each labeler were compared against an automatically generated combination of the rest of the labelers . To create the combined annotations we first used the RegisterMulti procedure to construct the union of each subset of N-1 labelers ( where N is the total number of labelers for each dataset ) . When N=4 then the combined annotation consisted of the components that were selected by at least two labelers . When N=3 a stricter intersection approach was used; the combined annotation consisted of the components that were selected by both remaining labelers . The procedure was repeated for all subsets of labelers and all datasets . The results are shown in Table 3 While individual scores for specific annotators and datasets vary significantly compared to using the consensus annotation as ground truth ( Table 1 ) , the decrease in average performance was modest indicating a low bias level . CaImAn uses two CNN classifiers; one for post processing component screening in CaImAn batch , and a different one for screening candidate components in CaImAn online . In both cases a four layer CNN was used , with architecture as described in Figure 2e . The first two convolutional layers consist of 32 3×3 filters each , whereas each of the latter two layers consist of 64 3×3 filters , all followed by a rectifier linear unit ( ReLU ) . Every two layers a 2×2 max-pool filter is included . A two layer dense network with 512 hidden units is used to compute the predictions ( Figure 2e ) . To efficiently distribute the cost of updating shapes across all frames we derived a simple algorithm that ( i ) ensures that every spatial footprint is updated at least once every Tu steps , where Tu is a user defined parameter , for example Tu=200 , and ( ii ) no spatial component is updated during a step when new components were added . The latter property is used to compensate for the additional computational cost that comes with introducing new components . Whenever a new component is added the algorithm collects the components with overlapping spatial footprints and makes sure they are updated at the next frame . This property ensures that the footprints of all required components adapt quickly whenever a new neighbor is introduced . The procedure is described in algorithmic form in Algorithm 6 . Fluorescence microscopy methods enable imaging the same brain region across different sessions that can span multiple days or weeks . While the microscope can visit the same location in the brain with reasonably high precision , the FOV might might not precisely match due to misalignments or deformations in the brain medium . CaImAn provides routines for FOV alignment and component registration across multiple sessions/days . Let 𝐚11 , 𝐚21 , … , 𝐚N11 and 𝐚12 , 𝐚22 , … , 𝐚N22 the sets of spatial components from sessions 1 and 2 respectively , where N1 and N2 denote the total number of components from each session . We first compute the FOV displacement by aligning some summary images from the two sessions ( e . g . , mean or correlation image ) , using a non-rigid registration method , for example NoRMCorre ( Pnevmatikakis and Giovannucci , 2017 ) . We apply the estimated displacement field to the components of A1 to align them with the FOV of session 2 . To perform the registration , we construct a pairwise distance matrix D∈RN1×N2 with D ( i , j ) =d ( 𝐚i1 , 𝐚j2 ) , where d ( ⋅ , ⋅ ) denotes a distance metric between two components . The chosen distance corresponds to the Jaccard distance between the binarized versions of the components . A real valued component 𝐚 is converted into its binary version m ( 𝐱 ) by setting to one only the values of 𝐚 that are above the maximum value of 𝐚 times a threshold θb , for example θb=0 . 2: ( 10 ) m ( a ) ( ( x ) ) ={1 , a ( ( x ) ) ≥θb‖a‖∞0 , otherwise . To compute the distance between two binary masks m1 , m2 , we use the Jaccard index ( intersection over union ) which is defined as ( 11 ) J ( m1 , m2 ) =|m1∩m2||m1∪m2| , and use it to define the distance metric as ( 12 ) d ( ai1 , aj2 ) ={1−J ( m ( ai1 ) , m ( aj2 ) ) 1−J ( m ( ai1 ) , m ( aj2 ) ) ≤θd0 ( m ( ai1 ) ⊆m ( aj2 ) ) OR ( m ( aj2 ) ⊆m ( ai1 ) ) , ∞otherwise . where θd is a distance threshold , for example 0 . 5 above which two components are considered non-matching and their distance is set to infinity . This is done to prevent false matchings between only marginally overlapping components . After the distance matrix D has been completed , an optimal matching between the components of the two sessions is computed using the Hungarian algorithm to solve the linear assignment problem . As infinite distances are allowed , it is possible to have components from both sessions that are not matched with any other component , preventing false assignments and enabling the registration of sessions with different number of neurons . Moreover , the use of infinite distances speeds up the Hungarian algorithm as it significantly restricts the space of possible pairings . This process of registering components across two sessions ( RegisterPair ) is summarized in Algorithm 7 . To register components across multiple sessions , we first order the sessions chronologically and register session 1 against session 2 . From this registration we construct the union of the distinct components between the two sessions by keeping the matched components from session two as well as the non-matched components from both sessions aligned to the FOV of session 2 . We then register this union of components to the components of session three and repeat the procedure until all sessions are have been registered . This process of multi session registration ( RegisterMulti ) is summarized in Algorithm 8 . At the end of the process the algorithm produces a list of matches between the components of each session and the union of all active distinct components , allowing for efficient tracking of components across multiple days ( Figure 9 ) , and the comparison of non-consecutive sessions through the union without the need of direct pairwise registration ( Figure 9—figure supplement 1 ) . An alternative approach to the problem of multiple session registration ( CellReg ) was presented recently by Sheintuch et al . ( 2017 ) where the authors register neurons across multiple days by first constructing a similar union set of all the components which is then refined using a clustering procedure . RegisterMulti differs from the CellReg method of Sheintuch et al . ( 2017 ) along the following dimensions: Each dataset was processed using the same set of parameters , apart from the expected size of neurons ( estimated by inspecting the correlation image ) , the size of patches and expected number of neurons per patch ( estimated by inspecting the correlation image ) . For the dataset N . 01 . 01 , where optical modulation was induced , the threshold for merging neurons was slightly higher ( the stimulation caused clustered synchronous activity ) . For shorter datasets , rigid motion correction was sufficient; for longer datasets K53 , J115 we applied non-rigid motion correction . Parameters for the automatic selection of components were optimized using a grid search approach . The global default parameters for all datasets were obtained by performing a grid search on the nine datasets over the following values: trace peak SNR threshold on the set {1 . 75 , 2 , 2 . 25 , 2 . 5} , spatial correlation threshold on the set {0 . 75 , 0 . 8 , 0 . 85} , lower threshold on CNN classifier ( reject if prediction is below a certain value ) on the set {0 . 05 , 0 . 1 , 0 . 15} , and upper threshold on classifier ( accept if prediction is above a certain value ) on the set {0 . 9 , . 95 , 0 . 99 , 1} . The best overall parameters ( used for the results reported in Table 1 ) were given for the choice ( 2 , 0 . 85 , 0 . 1 , 0 . 99 ) . For all datasets the background neuropil activity was modeled as a rank two matrix , and calcium dynamics were modeled as a first order autoregressive process . The remaining parameters were optimized so that all the datasets could be run on a machine with less than 128 GB RAM . Datasets were processed for two epochs with the exception of the longer datasets J115 , K53 where only one pass of the data was performed to limit computational cost . For all datasets the background neuropil activity was modeled as a rank two matrix , and calcium dynamics were modeled as a first order autoregressive process . For each dataset , CaImAn online was initialized on the first 200 frames , using the BareInitialization on the entire FOV with only two neurons , so in practice all the neurons were detected during the online mode . We did not post-process the results ( which could have further enhanced performance ) in order to demonstrate performance levels with fully online practices . As in batch processing , the expected size of neurons was chosen separately for each dataset after inspecting the correlation image . Several datasets ( N . 03 . 00 . t , N . 02 . 00 , J123 , J115 , K53 ) were spatially decimated by a factor of 2 to enhance processing speed , a step that did not lead to changes in detection performance . To select global parameters for all datasets we performed a grid search on all nine datasets by varying the following parameters: The peak SNR threshold for accepting a candidate component on the set {0 . 6 , 0 . 8 , 1 , 1 . 2 , 1 . 4 , 1 . 6 , 1 . 8 , 2} , the online CNN classifier threshold for accepting candidate components on the set {0 . 5 , 0 . 55 , 0 . 6 , 0 . 65 , 0 . 7 , 0 . 75} , and the number of candidate components per frame on the set {5 , 7 , 10 , 14} . The best overall parameters ( reported in Table 1 ) were given for the choice ( 1 . 2 , 0 . 65 , 10 ) . This parameter choice was also the best when conditioning on the shorter six datasets . For the three longer datasets , the best parameter choice was ( 2 , 0 . 6 , 5 ) , corresponding to a stricter set of parameters with less candidate components per frame ( Figure 4—figure supplement 1 ) . For the analysis of the whole brain zebrafish dataset , CaImAn online was run for one epoch with the same parameters as above , with only differences appearing in the number of neurons during initialization ( 600 vs 2 ) , and the value of the threshold for the online CNN classifier ( 0 . 75 vs 0 . 5 ) . The former decision was motivated by the goal of retrieving with a single pass neurons from a preparation with a denser level of activity over a larger FOV in this short dataset ( 1885 frames ) . To this end , the number of candidate neurons at each timestep was set to 10 ( per plane ) . The threshold choice was motivated by the fact that the classifier was trained on mouse data only , and thus a higher threshold choice would help diminish potential false positive components . Rigid motion correction was applied online to each plane . To compare CaImAn with Suite2p we used the MATLAB version of the Suite2p package ( Pachitariu et al . , 2017 ) . To select parameters for Suite2p we used a small grid search around the default values for the variables nSVDforROI , NavgFramesSVD , and sig . The classifier used by Suite2p was not re-trained for each dataset but used with the default values . For each case ( with the classifier and without the classifier ) , the values that give the best F1 score in average are reported in Figure 4—figure supplement 2 . The dataset J123 was excluded from the comparison since ( due its low SNR ) Suite2p did not converge and kept adding a large number of neurons in each iteration . Use of the classifier improved the results on average , from F1 score 0 . 51±0 . 12 without the classifier to 0 . 55±0 . 12 , however the use of the classifier improved only four of the eight tested datasets in terms of the F1 score . As with CaImAn it is possible that dataset specific parameter choice can lead to improved results . To quantify performance as a function of SNR we construct the consensus traces by running CaImAn batch on the datasets seeded with the ‘consensus’ binary masks obtained from the manual annotators . Afterwards we obtain the average peak-SNR of a trace 𝐜 with corresponding residual signal 𝐫 ( Equation 5 ) is obtained as ( 13 ) SNR ( 𝐳 ) =-Φ-1 ( pmin ) , where Φ-1 ( ⋅ ) denotes the probit function ( quantile function for the standard Gaussian distribution ) , 𝐳 is the z-scored version of 𝐜+𝐫 ( Equation 6 ) and pmin is given by Equation 8 . Let c1gt , c2gt , … , cNgt be the consensus traces and c1cm , c2cm , … , cNcm be their corresponding CaImAn inferred traces . Here we assume that false positive and false negative components are matched with trivial components that have 0 SNR . Let also SNRgti=SNR ( cigt ) and SNRcmi=SNR ( cicm ) , respectively . After we compute the SNR for both consensus and inferred traces the performance algorithm can be quantified in multiple ways as a function of a SNR thresholds θSNR: Precision: Precision at level θSNR , can be computed as the fraction of detected components with SNRcm>θSNR that are matched with consensus components . It quantifies the certainty that a component detected with a given SNR or above corresponds to a true component . PREC ( θSNR ) =|{i: ( SNRcmi>θSNR ) & ( SNRgti>0 ) }||{i: ( SNRcmi>θSNR ) }| Recall: Recall at level θSNR , can be computed as the fraction of consensus components with SNRgt>θSNR that are detected by the algorithm . It quantifies the certainty that a consensus component with a given SNR or above is detected . RECALL ( θSNR ) =|{i: ( SNRgti>θSNR ) & ( SNRcmi>0 ) }||{i: ( SNRgti>θSNR ) }| F1 Score: An overall F1 score at level θSNR , can be obtained by computing the harmonic mean between precision and recallF1 ( θSNR ) =2PREC ( θSNR ) ×RECALL ( θSNR ) PREC ( θSNR ) +RECALL ( θSNR ) The cautious reader will observe that the precision and recall quantities described above are not computed in the same set of components . This can be remedied by recomputing the quantities in two different ways: AND framework: Here we consider a match only if both traces have SNR above the given threshold:PRECAND ( θSNR ) =|{i: ( SNRcmi>θSNR ) & ( SNRgti>θSNR ) }||{i: ( SNRcmi>θSNR ) }|RECALLAND ( θSNR ) =|{i: ( SNRgti>θSNR ) & ( SNRcmi>θSNR ) }||{i: ( SNRgti>θSNR ) }| OR framework: Here we consider a match if either trace has SNR above the given threshold and its match has SNR above 0 . RECALLOR ( θSNR ) =|{i: ( max ( SNRgti , SNRcmi ) >θSNR ) & ( min ( SNRgti , SNRcmi ) >0 ) }||{i: ( SNRcmi>0 ) }|RECALLOR ( θSNR ) =|{i: ( max ( SNRgti , SNRcmi ) >θSNR ) & ( min ( SNRgti , SNRcmi ) >0 ) }||{i: ( SNRgti>0 ) }| It is easy to show thatPRECAND ( θSNR ) ≤PREC ( θSNR ) ≤PRECOR ( θSNR ) RECALLAND ( θSNR ) ≤RECALL ( θSNR ) ≤RECALLOR ( θSNR ) F1AND ( θSNR ) ≤F1 ( θSNR ) ≤F1OR ( θSNR ) , with equality holding for θSNR=0 . As demonstrated in Figure 4d , these bounds are tight . CaImAn contains a number of additional features that are not presented in the results section for reasons of brevity . These include: In the following we present in pseudocode form several of the routines introduced and used by CaImAn . Note that the pseudocode descriptions do not aim to present a complete picture and may refer to other work for some of the steps . | The human brain contains billions of cells called neurons that rapidly carry information from one part of the brain to another . Progress in medical research and healthcare is hindered by the difficulty in understanding precisely which neurons are active at any given time . New brain imaging techniques and genetic tools allow researchers to track the activity of thousands of neurons in living animals over many months . However , these experiments produce large volumes of data that researchers currently have to analyze manually , which can take a long time and generate irreproducible results . There is a need to develop new computational tools to analyze such data . The new tools should be able to operate on standard computers rather than just specialist equipment as this would limit the use of the solutions to particularly well-funded research teams . Ideally , the tools should also be able to operate in real-time as several experimental and therapeutic scenarios , like the control of robotic limbs , require this . To address this need , Giovannucci et al . developed a new software package called CaImAn to analyze brain images on a large scale . Firstly , the team developed algorithms that are suitable to analyze large sets of data on laptops and other standard computing equipment . These algorithms were then adapted to operate online in real-time . To test how well the new software performs against manual analysis by human researchers , Giovannucci et al . asked several trained human annotators to identify active neurons that were round or donut-shaped in several sets of imaging data from mouse brains . Each set of data was independently analyzed by three or four researchers who then discussed any neurons they disagreed on to generate a ‘consensus annotation’ . Giovannucci et al . then used CaImAn to analyze the same sets of data and compared the results to the consensus annotations . This demonstrated that CaImAn is nearly as good as human researchers at identifying active neurons in brain images . CaImAn provides a quicker method to analyze large sets of brain imaging data and is currently used by over a hundred laboratories across the world . The software is open source , meaning that it is freely-available and that users are encouraged to customize it and collaborate with other users to develop it further . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2019 | CaImAn an open source tool for scalable calcium imaging data analysis |
The precise regulation of microtubule dynamics is essential during cell division . The kinesin-13 motor protein MCAK is a potent microtubule depolymerase . The divergent non-motor regions flanking the ATPase domain are critical in regulating its targeting and activity . However , the molecular basis for the function of the non-motor regions within the context of full-length MCAK is unknown . Here , we determine the structure of MCAK motor domain bound to its regulatory C-terminus . Our analysis reveals that the MCAK C-terminus binds to two motor domains in solution and is displaced allosterically upon microtubule binding , which allows its robust accumulation at microtubule ends . These results demonstrate that MCAK undergoes long-range conformational changes involving its C-terminus during the soluble to microtubule-bound transition and that the C-terminus-motor interaction represents a structural intermediate in the MCAK catalytic cycle . Together , our work reveals intrinsic molecular mechanisms underlying the regulation of kinesin-13 activity .
The Kinesin-13 protein family is a class of microtubule depolymerases that regulate microtubule dynamics . Kinesin-13 family members are essential for correct interphase microtubule organization , cell polarity , and chromosome segregation during mitosis ( reviewed in Walczak et al . , 2013 ) . Kinesin-13 proteins induce the catastrophe of microtubule polymers by stabilizing the curved protofilament conformation found at the free ends of microtubules ( Gardner et al . , 2011 ) . Unlike processive kinesin motors , which have a motor domain at one end followed by a long coiled-coil and a globular tail , Kinesin-13 proteins possess a conserved motor domain containing the ATPase activity , flanked by two non-structured regions ( Figure 1A ) . The neck region , N-terminal to the motor , and the motor domain form the minimal region necessary for robust microtubule depolymerization ( Maney et al . , 1998; Ovechkina et al . , 2002 ) . The divergent regions flanking the motor domain are important for regulating its enzymatic activity , spatial targeting , dimerization , and creating unique kinesin functional specificity ( reviewed in Welburn , 2013 ) . The N terminus of the kinesin-13 MCAK ( Figure 1A , also known as Kif2c ) is responsible for its localization at kinetochores where it binds Sgo2 , and to the plus ends of microtubules where it associates with the end binding ( EB ) proteins ( Walczak et al . , 1996; Maney et al . , 1998; Mennella et al . , 2005; Tanno et al . , 2010; Welburn and Cheeseman , 2012 ) . Interestingly , the last 9 amino acids within the C terminus of MCAK are also necessary for plus tip tracking ( Moore et al . , 2005 ) . The region C-terminal to the motor domain ( residues 584–725 ) has been proposed to enable MCAK dimerization , but also to interact with the N terminus independently of the motor region ( Maney et al . , 2001; Hertzer et al . , 2006; Ems-McClung et al . , 2007; Zhang et al . , 2011; Ems-McClung et al . , 2013 ) . Additional work suggests the existence of long-range interactions between non-motor regions of MCAK in the context of full-length MCAK ( Moore and Wordeman , 2004; Hertzer et al . , 2006; Zhang et al . , 2011; Ems-McClung et al . , 2013 ) . The nature and the function of these inter- and intra-molecular interactions within the MCAK dimer are not known . 10 . 7554/eLife . 06421 . 003Figure 1 . The C terminus of MCAK binds to the motor domain . ( A ) Top: schematic diagram showing the different functional domains of full-length MCAK . Bottom: table representing the constructs used and given names . ( B ) Coomassie-stained gel showing a resin-based binding assay for purified His-M , His-NM , and CT domains to either glutathione agarose beads containing GST ( as a control ) or the GST-CT domain . The star represents residual GST . ( C ) Top , gel filtration elution profile of MCAK motor alone ( M , red ) and MCAK motor bound to the CT domain ( M + CT , cyan ) . Bottom , coomassie-stained gel showing the size-exclusion chromatography profile of M and M + CT . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 003 MCAK is the most potent microtubule depolymerase in the Kinesin-13 family ( Ogawa et al . , 2004 ) . Consequently , its tight regulation is critical for its proper function . Although MCAK depletion causes chromosome segregation defects and lagging chromosomes , MCAK overexpression results in spindle defects and is associated with taxol resistance in cancer cells ( Ganguly et al . , 2011a , 2011b ) . The regions that ultimately regulate MCAK targeting and fine-tune the catalytic activity of full-length MCAK lie outside of the motor region . Aurora B phosphorylation at the MCAK N terminus decreases its depolymerase activity ( Andrews et al . , 2004; Lan et al . , 2004; Ohi et al . , 2004 ) . CDK1 , Plk1 , and Aurora A have also been proposed to regulate MCAK activity in vitro ( Zhang et al . , 2007; Sanhaji et al . , 2010; Zhang et al . , 2011 ) . Removal of the last 9 amino acids of MCAK from Chinese hamster cells alleviates auto-inhibition of its depolymerase activity by increasing lattice-stimulated ATPase activity , and increases its microtubule binding in vitro ( Moore and Wordeman , 2004 ) . However , conflicting studies have proposed that the C-terminal tail of MCAK can either inhibit or activate the MCAK depolymerase activity ( Moore and Wordeman , 2004; Hertzer et al . , 2006; Zhang et al . , 2011 ) . Overall , the molecular mechanisms that regulate full-length MCAK activity remain unclear . Until recently , molecular studies on MCAK have focused on the interaction of the monomeric motor domain with microtubules to dissect the mechanism of MCAK-induced microtubule catastrophe ( Moores et al . , 2002 , 2003; Ogawa et al . , 2004; Shipley et al . , 2004; Mulder et al . , 2009; Asenjo et al . , 2013; Zhang et al . , 2013 ) . However , the monomeric motor domain does not function in isolation , as full-length MCAK is a physiological dimer ( Maney et al . , 2001 ) . Recent studies utilizing a FRET probe fused to the neck linker region of MCAK and the C terminus revealed that full-length MCAK switches from a ‘closed’ to ‘open’ conformation upon microtubule binding , but the trigger for this conformational change is unknown . MCAK is also thought to adopt a ‘closed’ conformation at microtubule ends ( Ems-McClung et al . , 2013 ) . The nucleotide state also influences the structure of full-length MCAK and induces a conformational change , as measured by deuterium exchange ( Burns et al . , 2014 ) . These studies suggest that full-length MCAK undergoes large dynamic structural changes during its catalytic cycle and upon binding to microtubules . However , the structure and organization of these flanking regions , and the trigger of the conformational changes remain uncharacterized . Here , we sought to define the molecular basis for the regulation of MCAK by its inter- and intra-molecular interactions . Our data establish that the MCAK motor domain binds to a 25 residue peptide from the extreme C terminus , termed the C-terminal tail ( CT ) domain . The CT domain induces motor dimerization in solution , reminiscent of the self-interaction mechanism of Kinesin-1 with its C terminus ( Kaan et al . , 2011 ) . The crystal structure of the MCAK C-terminal tail bound to the motor domain reveals how the C-terminal domain stabilizes a dimeric MCAK motor configuration . We also show that the MCAK C terminus controls the affinity of full-length MCAK for microtubules and reduces its association with the lattice to ensure maximal recruitment to microtubule ends , where MCAK can act as a depolymerase . When present in solution , the MCAK C-terminus binds to the motor domain . However , upon microtubule binding , the C terminus is displaced from the motor . This step is triggered by the microtubule itself , independently of the E-hook of tubulin and is necessary to allow binding of the motor to microtubules , and stimulate MCAK depolymerase activity . Within the context of the full-length MCAK , this indicates that MCAK undergoes long-range conformational changes driven by its C terminus during its soluble to microtubule-bound transition . Overall , our work presents a new paradigm for kinesin regulation by microtubules rather than their cargos , and provides important insights into the mechanism and regulation of MCAK to control microtubule dynamics and ensure proper genome stability .
Kinesin-1 is regulated through an auto-inhibitory mechanism whereby one C-terminal tail binds at the interface of the two motors , such that it creates a second point of attachment in addition to the coiled coil region . This limits the head movement of one kinesin with respect to the other ( Stock et al . , 1999; Hackney and Stock , 2000; Kaan et al . , 2011 ) . We sought to test whether the C-terminal tail domain of MCAK , which has been proposed to regulate MCAK activity ( Moore and Wordeman , 2004 ) , was sufficient to interact with the motor domain . To do this , we expressed the N-terminal and motor region ( residues 1–583 , termed NM ) or the motor region of MCAK along with the neck linker region ( residues 181–583 , termed M ) as His-tagged proteins , and the C-terminal MCAK tail ( residues 700–725 , termed CT domain ) as a GST fusion protein ( Moore and Wordeman , 2004; Hertzer et al . , 2006 ) ( Figure 1A ) . Following cleavage and removal of the GST fusion , the CT domain alone was unable to interact with itself through dimerization , based on the absence of binding between the CT domain and GST-CT domain ( Figure 1B , right lane ) . We cannot however exclude a very tight interaction between two CT domains . In contrast , the GST-CT domain protein bound to both the NM and M domains of MCAK as a stable complex ( Figure 1B ) . In addition , the CT domain bound the motor domain independently of the GST ( Figure 1C ) . Together , these experiments reveal that the MCAK CT domain interacts with its catalytic domain in solution . Above , we demonstrated that the MCAK CT domain interacts with the MCAK motor domain . Full-length MCAK is a dimer in solution ( Maney et al . , 2001 ) . Therefore , we sought to test whether the CT domain interacts with one or two motor domains . To define the stoichiometry of this interaction , we subjected the complex to analytical size-exclusion chromatography . The motor domain alone behaves as a monomer and eluted with an apparent size of ∼45 kDa . When the MCAK motor and the CT domain were incubated together and subjected to analytical size-exclusion chromatography , the elution peak shifted to an earlier fraction , suggesting dimerization of the MCAK motor ( Figure 1C ) . Using SDS-PAGE analysis , we confirmed that the shift to a larger complex was due to the interaction of the motor domain with the CT domain ( Figure 1C , bottom ) . Size-exclusion chromatography coupled with multi-angle light scattering ( SEC-MALS ) experiments further indicated that , in the absence of the CT domain , over 90% of the motor domain was monomeric ( Figure 2A , B ) , with a molecular weight of ∼45 . 3 kDa measured with under 3 kDa accuracy , in agreement with the theoretical molecular weight of ∼46 . 1 kDa ( Figure 2C ) . The predicted masses for complexes of one and two CT domains bound to two motor domains are ∼94 . 2 and 97 . 7 kDa , respectively . The measured molecular weight for the motor-CT domain complex was ∼91 . 5 kDa , suggesting the complex consists of two motors bound to one CT domain ( Figure 2A–C ) . Since the CT domain is unlikely to dimerize alone ( Figure 1B ) , we conclude that the MCAK CT domain induces the dimerization of the motor domains cooperatively . 10 . 7554/eLife . 06421 . 004Figure 2 . The C terminus of MCAK induces motor domain dimerization . ( A ) Size-exclusion chromatography elution profiles of motor domain alone ( red ) and motor domain-CT domain complex ( cyan ) . The horizontal red and cyan lines correspond to SEC-MALS calculated masses for motor domain alone and motor domain-CT domain complex , respectively . ( B ) Calibration curve for estimation of Stokes radii of motor domain alone ( red ) and motor domain-CT domain complex ( cyan ) . ( C ) Table to show the calculated apparent masses and stoke radii of the motor domain-CT domain complex and motor domain alone . The motor domain is drawn in orange and the CT domain in red , to represent the formation of the possible complexes and their predicted size . ( D ) Steady state intrinsic tryptophan fluorescence emission spectra profile for the titration of the CT domain ( CT ) ranging from 0 to 15 . 6 μM , into 1 μM of motor domain after excitation at 295 nm . ( E ) Effect of the CT domain titration on tryptophan ( magenta ) and aromatic residue ( green ) fluorescence quenching of the motor domain . The extent of fluorescence quenching of the motor domain is represented as a percentage of fluorescence change measured for aromatic residues ( 280 nm ) and tryptophan ( 295 nm ) with increasing concentration of wild type CT domain . Relative change in fluorescence after background correction is shown as a function of CT domain concentration . Error bars represent the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 004 We next determined the affinity of the MCAK CT domain for the motor domain using intrinsic fluorescence spectroscopy . We titrated increasing amounts of the CT domain with 1 μM motor domain and measured the corresponding change in the fluorescence intensity of aromatic residues ( Figure 2D ) . The change in fluorescence upon peptide binding corresponded to ∼1 . 1 μM affinity of the CT domain for the motor domain , although this measurement does not take into account any existing equilibrium between the motor domains ( Figure 2E ) . Overall , this affinity reflects the sum of the dimerization affinity of the motor domains and the affinity of the CT domain for the motor domains , as CT domain binding is cooperative with motor dimerization . Taken together , our data demonstrate that upon binding , the CT domain of MCAK engages with two motor domains . To test how the motor domain of MCAK interacts with the C terminus at the molecular level , we co-crystallized and determined the structure of the motor domain bound to a chemically synthesized peptide corresponding to the CT domain ( 709QLEEQASRQISS720 ) using molecular replacement to a resolution of 3 Å with good stereochemical parameters ( Table 1 , Figure 3A ) . The asymmetric unit contains four molecules of the motor domain assembled into two dimers ( chains A and B , C and D ) and in a spacegroup distinct from the MCAK motor crystallized alone . The dimerization interface involves packing of the MCAK motor domains along their helix α1 and β3 sheet , close to helix α0 and loop L1 , which form the neck linker region ( Figure 3A , Figure 3—figure supplement 1A ) . From the Fo-Fc electron density map , we could observe interpretable electron density close to the interface between chains A and B and build residues 710 to 716 of the CT domain ( Figure 3B ) . We found that a single CT domain binds to both motor domains , close to their neck linker regions . This structural arrangement provides a structural explanation for how the CT domain promotes motors dimerization , reminiscent of the Kinesin-1-tail domain interaction ( Kaan et al . , 2011 ) . The head-to-head motor arrangement is quasi-symmetrical , with the CT domain stabilizing the interface between two motor domains . However , the peptide does not sit on a twofold crystallographic axis and binds asymmetrically to the dimer , unlike the Kinesin-1 tail . Our data reveal the molecular basis for the CT domain-induced dimerization of the motor domains , binding along the motor dimer interface to stabilize the complex . 10 . 7554/eLife . 06421 . 005Table 1 . Data collection , structure determination , and refinement statistics for the X-ray crystal structure of the CT domain of MCAK bound to its motor domainDOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 005StatisticsMCAK motor domain-peptide complexUnit cell dimensionsa = 46 . 31 Å , b = 245 . 64 Å , c = 79 . 40 Å , α = 90 . 00° , β = 95 . 84° , γ = 90 . 00°Space groupP21Molecules per asymmetric unit4Resolution range ( Å ) 30 . 0–3 . 0Total reflections155983Unique reflections35 , 146Completeness ( % ) 99 . 0 ( 99 . 2 ) Multiplicity4 . 4 ( 4 . 5 ) Rsym ( % ) 9 . 1 ( 68 . 2 ) I/σ ( I ) 10 . 2 ( 2 . 0 ) Rwork/Rfree ( % ) 26 . 4/28 . 6Wilson B ( Å2 ) 77 . 5Average B ( Å2 ) : Overall71 . 0 Main chain72 . 05 Side chain and solvent70 . 66 Peptide56 . 98r . m . s . d . bond lengths ( Å ) 0 . 095r . m . s . d . bond angles ( ° ) 1 . 53Ramachandran plot statistics ( % ) : Favoured87 . 6 Allowed11 . 7 Outliers0 . 710 . 7554/eLife . 06421 . 006Figure 3 . Structure of a human motor-CT domain MCAK complex . ( A ) Kinesin motor domain dimers ( cyan and green ) bound to the CT domain ( yellow , spacefill ) of MCAK . ADP is in red . ( B ) Motor-CT domain interface showing the electron density map ( 2Fobs − Fcalc ) , contoured at σ = 1 . 00 for the CT domain of MCAK . ( C ) Interactions within the motor-CT domain complex of less than 4 Å are represented by dotted lines . Oxygen and nitrogen atoms are colored red and blue . ( D ) Overlay of the human motor domain and C terminus structure ( blue ) with the structure of murine MCAK ( pink , PDB: 1V8J ) . The respective neck regions containing the α0 and neck linker are in royal blue and magenta , respectively . The CT domain of MCAK is drawn in yellow as a sphere model with oxygen and nitrogen atoms in blue and red . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 00610 . 7554/eLife . 06421 . 007Figure 3—figure supplement 1 . Structural analysis of the MCAK motor-CT domain . ( A ) Dimeric interface of MCAK motors ( cyan and green ) . The respective neck regions containing α0 and loop L1 and the dimerization interface including α1 and β3 are indicated . ( B ) Overlay of chain A ( green ) and B ( blue ) with chain C and D ( salmon ) , showing that Glu244 in chain C points towards the peptide binding site . Glu244/A is repositioned and stabilized by Lys286/B through a salt bridge interaction in presence of the CT domain . His257/B is also repositioned in chain B in the presence of the CT domain . ( C ) Overlay of our MCAK motor-CT domain structure ( cyan ) with the structure of murine MCAK ( pink , PDB: 1V8J ) showing the switch I ( yellow ) , switch II regions ( orange ) , and the ATP-binding P-loop site ( red ) . The neck regions are shaded in darker blue and pink , respectively . ( D ) Orientation of the neck regions for overlaid mouse and human MCAK structures ( pink and blue , respectively ) . The change in direction of the neck linker occurs around His257 and Arg258 . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 007 A second potential CT domain-binding site is present in the dimeric motor arrangement for chain A and B . However , it is obstructed by symmetry-related molecules in the asymmetric unit . Chain C and D assemble similarly as dimers , with one potential binding site also obstructed by a symmetry-related molecule . Interestingly , at the second site , the Glu244/C side chain points outwards to the solvent and is incompatible with binding of the CT domain ( Figure 3—figure supplement 1B ) . In the CT domain-bound dimer , the side chain of Glu244/A is rearranged and points towards chain B and is stabilized by a salt bridge with Lys286/B . The imidazole ring of His257/B also moves backwards , allowing the CT domain to bind . Thus the motor domains can dimerize , but the CT domain stabilizes this dimeric motor assembly after rearrangement of Glutamate 244 and Histidine 257 ( Figure 3—figure supplement 1B ) . The CT domain contributes the side chains of Glu711 and Glu712 , forming charged ‘fingers’ that dip into the cavity lining the dimeric motor interface to further stabilize the interface ( Figure 3C ) . The carboxyl side chain of Glu711 forms a hydrogen bond with His257/B , while the carboxyl group of Glu712 is hydrogen bonded to the amide group of Thr 242/A . Additional hydrogen bonds stabilize the peptide-motor complex through main chain interactions . The backbone amide group of Glu712 is stabilized with the backbone carbonyl group of Ala241/A and the backbone amide of Ser715 hydrogen bonds to the backbone carbonyl group of Cys245/A . Although the hydroxyl group of Ser715 does not interact directly with the motor domain , it does point inwards towards the motor domain . The binding of the CT domain occurs far from the P-loop , which forms the ATP binding site , and the switch I and switch II regions ( Figure 3—figure supplement 1C ) . In addition , binding of the CT domain does not cause any changes in the ATP binding site or in the overall structure of MCAK ( RMSD: 0 . 863 Å ) . Interestingly in our structure , the L1 and the α0 helix part of the neck linker have swung away from the microtubule binding site with respect to the previously published mouse MCAK/Kif2c structure . This suggests that the neck region has conformational flexibility around a hinge region at Arg258 , and can adopt at least two states ( Figure 3D , Figure 3—figure supplement 1C , D ) ( Ogawa et al . , 2004 ) . Overlaying our structure with the mouse MCAK structure reveals that the conformation of the neck linker region in the mouse MCAK/Kif2c structure does not allow binding of the CT domain , due to steric hindrance ( Figure 3D , Figure 3—figure supplement 1D ) . The neck region of MCAK has been shown previously to be critical for the depolymerase activity of MCAK ( Maney et al . , 2001; Ovechkina et al . , 2002 ) . It is therefore possible that disruption of the CT domain-motor interaction allows conformational changes in the neck region that are necessary for catalysis . Taken together , our work reveals that one MCAK CT domain acts directly to stabilize the formation of a dimeric MCAK through an extended interface , where the neck linker lies on the face opposite of the microtubule binding site . To validate the residues implicated in generating the interface between the MCAK motor domain and CT domain , we generated a series of point mutants to selectively disrupt the binding of the CT domain to the motor domain . Based on our crystal structure and the sequence conservation of the C terminus , we predicted that Glu711 and Glu712 would be critical for the CT domain-motor interaction , whereas Arg716 and Ile718 would not prevent CT domain-motor binding ( Figures 3C , 4A ) . As expected from the crystal structure , a CTE711A , E712A domain no longer bound to the motor domain of MCAK , whereas a CTR716A or CTI718A domain bound robustly ( Figure 4C , D ) . As revealed in the structure , these two negatively charged glutamic acid residues are critical for the interaction between the CT and motor domains . These two amino acids are conserved from Drosophila to human , and are also present in the kinesin-13 family member Kif2a , suggesting that the motor-tail domain interaction is conserved ( Figure 4A , B ) . 10 . 7554/eLife . 06421 . 008Figure 4 . Sequence requirement for the formation of a motor-CT tail complex . ( A ) Sequence alignment of the conserved CT domain of MCAK for various species alongside the Drosophila kinesin-13 Klp10A and human Kif2a . The conserved residues are highlighted in red . The three amino acids that are critical for binding to the motor domain are marked with a green star . ( B ) Sequence alignment of the C terminus of human Kif2a , Kif2b , and MCAK/Kif2c . Amino acid numbering is relative to the Kif2a sequence . The MCAK CT domain binding to the motor domain is boxed in green . The sequences were aligned using the program T-coffee ( EBI ) and formatted with ESPRIPT ( Gouet et al . , 1999 ) . ( C ) Coomassie-stained gel showing a resin-based binding assay using glutathione agarose beads for purified His-M , binding to the GST-CT and GST-CT point mutants . ( D ) Size-exclusion chromatography elution profile of the motor domain alone ( red dashes ) , motor incubated with the CT , CTS715E , CTE711A , E712A domains ( cyan , green dashes , and purple , respectively ) . Bottom , coomassie-stained gel showing the size-exclusion chromatography elution of the motor incubated with the CTS715E and CTE711A , E712A domains ( green and purple , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 008 Interestingly , in addition to these key structural residues , we found that Ser715 in the CT domain is highly conserved across species and is present in the related kinesin , Kif2a , suggesting that this residue could play a role in the tail-motor interaction . Ser715 has been reported to be phosphorylated in vitro by Aurora A and Plk1 ( Zhang et al . , 2008 , 2011 ) . In our crystal structure , the hydroxyl group of Ser715 is in close proximity to His246 and Glu244 . A larger side chain would cause steric hindrance and prevent the CT domain-motor domain association . To test whether the nature of the side chain at position 715 can regulate the interaction between the motor domain of MCAK and its CT domain , we generated GST-CTS715E and GST-CTS715A constructs . Although the GST-CT and GST-CTS715A domains bound to the motor domain , the GST-CTS715E domain did not interact with the motor domain ( Figure 4C ) . In addition , the motor domain did not co-migrate with the CTS715E domain by gel filtration ( Figure 4D ) . Based on the crystal structure of the CT domain-motor complex , post-translational modification of this residue would destabilize the interaction through electrostatic repulsion and steric hindrance . This demonstrates that the conserved side chains of Glu711 , Glu712 , and Ser715 are critical for stabilizing the binding of the CT domain to the motor domain . Taken together , our data suggest that the molecular mechanism underlying the interaction between the MCAK C terminus and the motor domain is highly conserved across species . The CT domain induces dimerization of the MCAK motor . To test whether the CT domain was the major dimerization region within MCAK , we generated full-length MCAKS715E , in which the CT domain cannot bind to the motor domains . The gel filtration profile of MCAKS715E was similar to MCAK , indicating that MCAKS715E was of similar size to full-length dimeric MCAK in solution ( Figure 5A ) . This indicates that there is a second dimerization region within MCAK , independent of the CT domain . 10 . 7554/eLife . 06421 . 009Figure 5 . Full-length MCAK remains dimeric upon disruption of the motor-CT tail interaction but retains its depolymerase activity . ( A ) Size-exclusion chromatography elution profiles of full-length MCAK ( green ) and full-length MCAKS715E ( magenta ) . ( B ) Graph plotting the microtubule depolymerase activity of 100 nM MCAK and MCAKS715E by measuring the distribution of 2 μM microtubules in the pellet ( P ) and free soluble tubulin ( S ) over time . Error bars represent the standard deviation . Experiments were repeated three times . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 009 We next asked whether the CT domain affects MCAK depolymerase activity and MCAK function . Removal of the last 9 amino acids at the MCAK C terminus has been reported to increase the lattice-stimulated ATPase activity but not its ATPase activity in solution ( Moore and Wordeman , 2004 ) . However , conflicting studies have reported that removal of the last 28 amino acids in Xenopus MCAK results in a decrease in MCAK depolymerase activity ( Hertzer et al . , 2006 ) . Thus , the role of the CT domain in the context of full-length MCAK remains unclear . The microtubule depolymerase activity of full-length MCAKS715E , in which the CT domain can no longer bind to the motor domain , appeared similar to wild type MCAK in microtubule depolymerization assays ( Figure 5B ) . However there are limitations to this assay , as we were only able to measure the rate of microtubule depolymerization using cosedimentation assays for a given MCAK concentration rather than examining single MCAK molecules at microtubule ends . It is possible that a change in MCAK microtubule binding affinity will have a counteracting effect on MCAK diffusion rate or the rate of tubulin removal at ends as previously shown ( Cooper et al . , 2010 ) . In this case , the overall depolymerase activity that our assay measures may remain unchanged because the increase in affinity of MCAK for the microtubule lattice may cause a reduction in two-dimensional diffusion and consequently a reduction in microtubule end targeting . As MCAK and MCAKS715E displayed a similar depolymerase activity in our in vitro depolymerase assay , this raises the possibility that the CT domain acts indirectly as an inhibitor and has an additional distinct cellular function . To test the contribution of the CT domain to MCAK activity and function , we designed a system to generate an inducible covalent CT domain-motor complex in vitro based on our structure to control for the displacement of the CT domain from the motor domain . Cys287/A in Loop 1 of the motor domain is in close proximity to the CT domain , with the side chain of Glu712 and the sulfhydryl group of Cys287 pointing towards each other ( Figure 6—figure supplement 1A ) . Therefore , we mutated Glu712 to a cysteine to generate a disulfide bridge between the peptide and the motor domain , estimated to be ∼3 Å under oxidizing conditions . First , we purified full-length MCAKE712C . In presence of reducing agent ( DTT ) , full-length MCAKE712C eluted as one complex , of similar size to MCAK ( Figure 6—figure supplement 2A ) . Under oxidizing conditions ( without DTT ) MCAKE712C ran similarly to MCAK on an SDS-PAGE gel ( Figure 6—figure supplement 2B ) . However , we were not able to determine the efficiency of the covalent attachment between Cysteine 712 and Cysteine 287 . To test that a covalent linkage had been achieved , we expressed the cleavable GST-CTE712C domain . Under oxidizing conditions ( without DTT ) , the motor and both the GST-CTE712C domain and untagged CTE712C domain formed a covalent complex ( Figure 6—figure supplement 1B , C ) . Analytical gel filtration of the GST-CTE712C—bound motor complex eluted as a single peak , earlier than the peak for the motor alone ( Figure 6—figure supplement 1D ) . However , SDS-PAGE analysis indicated that , within the assembled complex , there was one free motor and one motor covalent bound to the GST-CTE712C domain . This indicates that one CT domain binds to two motors , only one of which is crosslinked ( Figure 6—figure supplement 1B , E , F ) . Based on our structural analysis , binding of one CT domain to one of the motors in the dimer would not obstruct the solvent accessibility of the second Cys287 . Thus , this experiment suggests that within a CT domain-motor complex , one CT domain binds to two motor domains , consistent with the stoichiometry we determined using SEC-MALS ( Figure 2C ) . We also noted that the covalent attachment of the CT domain to the motor domain would also prevent any conformational rearrangement and repositioning of the neck region close to the microtubule-binding interface ( Figure 3D ) and may thus decrease its microtubule depolymerase activity . Full-length MCAK has been proposed to undergo large conformational changes upon binding to microtubules , although the underlying mechanism is unclear ( Ems-McClung et al . , 2013; Burns et al . , 2014 ) . Based on our data , we hypothesized that in solution , the CT domain binds to the motor , but that the CT domain is displaced when the motor binds to microtubules . To test whether MCAK has a reduced ability to bind to microtubules when the CT domain is bound to the motor , we first performed cosedimentation assays with full-length MCAKE712C . In the presence of DTT , MCAKE712C bound to microtubules similarly to wild type full-length MCAK . However , under oxidizing conditions ( absence of DTT ) , the affinity of MCAKE712C for microtubules was reduced and a fraction of MCAKE712C did not bind microtubules , even at saturating microtubule concentrations ( Figure 6—figure supplement 2C , D ) . This indicates that the binding of the CT domain of MCAK to the motor interferes with MCAK binding to microtubules . To further dissect the effect of the CT domain on the motor domain in the context of microtubules , we performed cosedimentation assays with the CTE712C domain-motor domain complex with increasing concentrations of microtubules . If tubulin within the microtubule is necessary to displace the CT domain and allow binding of the motor to microtubules , we hypothesized that only the non-covalently bound MCAK motor would be able to undergo the conformational change necessary for binding to microtubules , whereas the CT domain-bound MCAK motor ( M-CTE712C ) fraction would be in a locked conformation and would not bind or only bind weakly . Cosedimentation of the motor domain in the presence of the CTE712C domain and DTT was similar to the MCAK motor alone with Kds of 0 . 44 and 0 . 64 μM respectively , indicating that the CTE712C domain did not interfere with the motor under reducing conditions ( Figure 6A , B ) . Similarly , the addition of DTT did not affect the affinity of MCAK motor in presence of the CT domain ( Figure 6—figure supplement 3A , B ) . In contrast , addition of the CTE712C domain to the MCAK motor under oxidizing conditions reduced the fraction of MCAK bound to microtubules by ∼50% , indicating that half of the CT domain-bound MCAK motor ( M-CTE712C ) sample did not bind to microtubules ( Figure 6A , B ) . In these samples , only MCAK motor that was not bound to the CT domain cosedimented with microtubules . Also we did not detect the CT and CTE712C domains in the microtubule-bound , pelleted samples ( Figure 6A , Figure 6—figure supplement 3A ) . This demonstrates that the binding of MCAK to its C terminal tail region and to microtubules is mutually exclusive . 10 . 7554/eLife . 06421 . 010Figure 6 . The binding of the CT domain to the motor prevents MCAK binding to microtubules and reduces MCAK depolymerase activity . ( A ) Western blot showing the cosedimentation of 50 nM the motor domain of MCAK alone and in the presence of the cleaved CTE712C domain , with and without the addition of DTT to control the formation of the disulphide bridge , at the indicated concentration of microtubules . Detection of the MCAK motor and the CT domain were done using an anti-His and anti-MCAK CT domain antibody , respectively . When the CTE712C domain is covalently bound to the motor domain , we observe free CTE712C domain ( ∼3 . 5 kD ) and motor-bound CTE712C domain ( ∼47 kD ) when probing for the CT domain . The CTE712C-bound motor remained in the supernatant . ( B ) Graph plotting the microtubule binding activity of the complexes in ( A ) in absence of nucleotide . Data were fitted with a modified Hill equation ( Welburn et al . , 2010 ) . Error bars represent the standard deviation . ( C and D ) Western blot showing the cosedimentation of 100 nM full-length MCAKS715E incubated in the presence of 2 μM taxol-stabilized microtubules with increasing concentration of the cleaved free CT ( C ) and CTS715E ( D ) domains . The western blots were probed with the antibody directed against the CT domain . All experiments were repeated three times . ( E ) Coomassie-stained gel showing the microtubule depolymerization activity of 50 nM MCAK motor alone and 50 nM MCAK motor-CTE712C domain in presence and absence of DTT , over time on 2 μM taxol-stabilized microtubules . Free tubulin and microtubule polymers were separated using a cosedimentation assay . ( F ) Graph plotting the quantified microtubule depolymerase activity for conditions in ( E ) . The data were fitted with linear regression . The specific depolymerase activity of a covalent MCAK-CTE712C complex was calculated by subtracting the activity of MCAK motor alone , which represents ∼50% of the population . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 01010 . 7554/eLife . 06421 . 011Figure 6—figure supplement 1 . Tunable covalent linkage of the CT domain of MCAK to the motor . ( A ) Model showing that mutation of Glutamate 712 to Cysteine can create a disulphide bond between Cysteine 287 and Cysteine 712 . ( B ) Coomassie-stained gel showing that in the absence of reducing agent such as DTT , the GST-CTE712C domain binds specifically the motor domain through a disulphide bridge with a ∼50% efficiency . ( C ) Western blot probing for the CT domain and ponceau stain showing total protein indicate that in absence of DTT the motor and the CT domain form a covalent complex . There are two bands for the motor domain , one of them coupled to the CT domain , with similar stoichiometry to ( B ) . ( D ) Coomassie-stained gel showing the size-exclusion chromatography profile of the motor and the GST-CTE712C domain ( M + GST-CTE712C ) . For one motor-CT complex , there is one free motor and one CT-bound motor . ( E ) Gel filtration elution profile of MCAK motor alone ( M , red ) and MCAK motor bound to the GST-CTE712C domain ( M + GST-CTE712C , cyan ) . ( F ) Schematic diagram of the efficiency of disulphide bridge formation for MCAK motor dimers , quantified from ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 01110 . 7554/eLife . 06421 . 012Figure 6—figure supplement 2 . The affinity of MCAK for microtubules decreases when the CT domain of MCAK is not displaced from the motor . ( A ) Size-exclusion chromatography elution profile of full-length MCAK ( green ) and MCAKE712C ( magenta ) in absence of DTT . ( B ) Coomassie-stained gel showing MCAK and MCAKE712C in the absence of reducing agent such as DTT . ( C ) Western blot showing the cosedimentation of 50 nM MCAKE712C in absence of nucleotide , with and without the addition of DTT to control the formation of the disulphide bridge , at the indicated concentration of microtubules . ( D ) Graph plotting the microtubule binding activity of the proteins in ( C ) and fitted with a modified Hill equation ( Welburn et al . , 2010 ) . All experiments were repeated at least three times . Error bars represent the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 01210 . 7554/eLife . 06421 . 013Figure 6—figure supplement 3 . Absence of reducing agent does not affect MCAK motor properties . ( A ) Western blot showing the cosedimentation of 50 nM motor domain of MCAK in presence of the CT domain ( M + CT ) and in absence of nucleotide , with and without the addition of DTT to control the formation of the disulphide bridge , at the indicated concentration of microtubules . ( B ) Graph plotting the microtubule binding activity of the complexes in ( A ) . ( C ) Coomassie-stained gel showing the microtubule depolymerization activity of 50 nM MCAK motor with the CT domain in presence and absence of DTT , over time on 2 μM taxol-stabilized microtubules . Free tubulin and microtubule polymers were separated using a cosedimentation assay . ( D ) Graph plotting the quantified microtubule depolymerase activity for conditions in ( C ) . All experiments were repeated at least three times . Error bars represent the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 013 To further test the effect of the CT domain on MCAK binding to the microtubule lattice , we tested the effect of free CT domain on full-length MCAKS715E in which its own CT domain is unable to bind the motor . We found that the addition of free CT domain decreased the affinity of MCAKS715E for microtubules ( Figure 6C ) . In contrast , titration of free CTS715E did not interfere with MCAKS715E binding to microtubules ( Figure 6D ) . This indicates that the CT domain specifically competes with microtubules for MCAK binding and effectively reduces the affinity of MCAK for microtubules . Our data suggest that to function as an active depolymerase , MCAK must undergo a large conformational change in which the CT domain of MCAK dissociates from the motor domain and releases the motor domains from each other . In total , our findings demonstrate that the CT domain acts through an allosteric mechanism to prevent MCAK binding microtubules until the CT domain is displaced , thereby enabling the MCAK depolymerase activity . We next tested the MCAK depolymerase activity when the CT domain is covalently bound to the motor domain . We first tested whether the specific reducing conditions had an effect on MCAK depolymerase activity in the presence of the native CT that could not covalently bind the motor domain ( Figure 6—figure supplement 3C , D ) . In both presence and absence of DTT , the MCAK motor could depolymerize microtubules , leading to an increase in free tubulin in the supernatant ( S ) and a decrease in microtubules in the pellet ( P ) over time . Next , we incubated the CTE712C domain with the motor in presence and absence of DTT to generate unbound and CTE712C-bound MCAK motor . In presence of DTT , the CTE712C domain did not bind the MCAK motor and the depolymerase activity was similar to that of the wild type MCAK motor alone ( Figure 6E , F ) . However , in absence of DTT , the CTE712C domain bound-MCAK motor displayed reduced depolymerase activity ( Figure 6E , F ) . The activity of this motor-CTE712C domain complex is likely to be lower than that of the observed apparent activity due to the presence of a non-covalently bound MCAK fraction , which functions as a fully active depolymerase ( ∼45–50% , Figure 6—figure supplement 1F ) . We calculated the specific activity by taking into account the fraction of active MCAK ( ∼50% , Figure 6F–blue dotted line ) . This shows that the covalent binding of CTE712C to MCAK motor strongly inhibits the depolymerase activity of MCAK . Overall , this demonstrates that the displacement of the CT domain is necessary for the full microtubule depolymerization activity of MCAK . Above , we found that the displacement of the CT domain is required for MCAK motor association with the microtubule ( Figure 6 ) . However , the molecular mechanism that triggers the displacement of the CT domain from the motor was unclear . To test whether the negatively charged E-hook of tubulin or the lattice itself triggers the removal of the CT domain from the motor , we performed cosedimentation assays of the motor bound to the CT domain with microtubules in absence of the tubulin tails . To test this , we treated microtubules for 10 and 120 min with subtilisin to remove the C-terminal tails of β and α/β-tubulin , respectively ( Figure 7—figure supplement 1A ) . Cosedimentation of the motor-CT domain complex in presence of subtilisin-treated microtubules revealed that the CT domain was displaced from the motor and remained in the supernatant , while the motor domain bound with a high affinity to the tubulin lattice ( Kd = 0 . 2 μM ) ( Figure 7—figure supplement 1B , C ) . Removal of the α-tubulin tail did not further modify the affinity of the motor domain for microtubules that also lacked the β-tubulin tail . Taken together , the microtubule lattice itself rather than the acidic tails of tubulin trigger the release of the CT domain from the motor . Removal of the entire C-terminal domain of MCAK has been shown to increase the affinity of MCAK for microtubules and prevent plus end targeting , although the mechanism is not defined ( Moore and Wordeman , 2004; Moore et al . , 2005 ) . In addition , we found that the CT domain reduces the ability of full-length MCAK to bind to microtubules ( Figure 6C ) . Therefore , we predicted that full-length MCAKS715E , in which the CT domain is unable to interact with the motor domains would have a higher affinity for microtubules than wild type full-length MCAK . To test this , we measured the affinity of full-length wild type MCAK and MCAKS715E for microtubules using a cosedimentation assay . We found that MCAKS715E showed a 10-fold increase in the apparent affinity for microtubules compared to wild type MCAK ( ∼0 . 2 μM and 1 . 5 μM , respectively; Figure 7A ) . MCAK has been reported previously to bind to microtubules lacking the acidic tails ( Niederstrasser et al . , 2002; Helenius et al . , 2006 ) . However , these studies indicated that the ability of MCAK to diffuse on the lattice was reduced in this case . To test the effect of the acidic C-terminal tails of tubulin on MCAK binding and on the function of the CT domain , we tested the affinity of MCAK and MCAKS715E for subtilisin-treated microtubules ( Figure 7B ) . Removal of β-tubulin C termini increased the affinity of full-length MCAK , whereas the affinity of full-length MCAKS715E for microtubules remained comparably high ( Figure 7C ) . This indicates that both the CT domain of MCAK and the C termini of tubulin cooperate to reduce the affinity of MCAK for microtubules and ensure that MCAK does not become trapped on the lattice , away from its microtubule ends substrate . 10 . 7554/eLife . 06421 . 014Figure 7 . The CT domain reduces the affinity of MCAK to microtubules . ( A and B ) Western blot showing the cosedimentation of 50 nM MCAK with microtubules at the indicated concentrations . In panel B , the microtubules have been treated with subtilisin for 10 min prior to the cosedimentation assay . ( C ) Graph plotting the average microtubule binding activity of MCAK and MCAKS715E in absence of nucleotide . The dashed and full curves correspond to subtilisin-treated and untreated microtubules , respectively . The data were fitted using a modified Hill equation . Error bars represent the standard deviation . ( D ) Representative images of HeLa cells transiently transfected with mCherry-EB3 and GFP-MCAK or GFP-MCAKS715E , alongside the respective average normalized fluorescence intensity linescan profiles at microtubule plus tips . Grey shading of the linescans represents the standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 01410 . 7554/eLife . 06421 . 015Figure 7—figure supplement 1 . The displacement of the CT domain from the motor is triggered by the microtubule lattice but is independent of the E-hook of tubulin . ( A ) Western blot showing the efficiency of the α- and β-tubulin tails removal over time . ( B ) Western blot showing the cosedimentation of 50 nM MCAK motor domain with microtubules at the indicated concentrations probed with antibodies detecting the C-terminal tails of α- and β-tubulin . The microtubules have been treated with subtilisin for 10 or 120 min prior to the cosedimentation assay . ( C ) Graph plotting for the average microtubule binding activity of MCAK motor in absence of nucleotide after 10 or 120 min subtilisin-treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 015 MCAK utilizes weak tethering to diffuse on the negatively charged C-terminal tails of the microtubule lattice ( Helenius et al . , 2006 ) . The neck region was originally proposed to promote MCAK diffusion , similarly to Kif1a ( Thorn et al . , 2000; Wang and Sheetz , 2000; Ovechkina et al . , 2002; Helenius et al . , 2006 ) . The idea that the neck was the electrostatic tether supporting E-hook mediated diffusion was subsequently disproven ( Cooper et al . , 2010 ) . Thus to date the regions of MCAK responsible for diffusion remain unclear . Based on the CT domain controlling the affinity of MCAK for microtubules redundantly with the C-terminal tail of tubulin , we hypothesized that this electrostatically charged CT region may play also a role in MCAK diffusion on the lattice and targeting to microtubule ends ( Figure 8 ) . To test whether the CT domain of MCAK controls the targeting of MCAK by decreasing MCAK affinity for the microtubule lattice , we examined the localization of full-length MCAK and MCAKS715E in HeLa cells . GFP-MCAKS715E targeted weakly to microtubule plus ends but also accumulated on the microtubule lattice ( Figure 7D ) , confirming our in vitro observation ( Figure 7A , C ) . In contrast , GFP-MCAK was robustly targeted to microtubule ends and co-localized with mCherry-EB3 . Future work will address whether the CT domain is the main region providing direct lattice diffusion properties to MCAK through electrostatic interactions . In total , these data suggest that the CT domain reduces the affinity of MCAK for microtubules and may be the electrostatic tether that allows MCAK specific targeting to microtubule ends . 10 . 7554/eLife . 06421 . 016Figure 8 . Model for MCAK conformation in solution and when bound to microtubules . MCAK has a compact structure in solution , with one C terminus binding at the interface between two motor domains . MCAK can bind to microtubules through the microtubule-binding region , which allosterically triggers release of the C-terminus of MCAK . The motor domains can then efficiently bind to and depolymerize the microtubule end , through possible repositioning of the neck linker region . Both the C terminus of MCAK and the negatively charged E-hook of tubulin , reduce the binding of MCAK to microtubules , enabling MCAK to diffuse efficiently to microtubule ends . DOI: http://dx . doi . org/10 . 7554/eLife . 06421 . 016
MCAK is a powerful microtubule depolymerase , whose activity must be tightly regulated through phosphorylation and self-interaction . Our results reveal a regulatory paradigm for the Kinesin-13 microtubule depolymerases , which are functionally and structurally distinct from processive kinesins . Previously , the molecular organization of full-length dimeric kinesin depolymerases and the inhibitory mechanisms for kinesin depolymerases were unclear . Here , we show that in solution , the C terminus of MCAK interacts with the two motor domains through long-range interactions . Binding of the CT domain and microtubules to the motor is mutually exclusive . While the acidic tails of tubulin control the affinity of MCAK for microtubules , they are not necessary for the displacement of the MCAK CT domain from the motor . The tubulin subunit itself triggers the removal of the CT domain from the motor , most likely through a conformational change within the microtubule-binding region in the motor domain . Disruption of this interaction causes MCAK to bind more strongly to microtubules , which leads to the accumulation of MCAK along the microtubule lattice and is disadvantageous for a microtubule-depolymerizing enzyme that acts at microtubule ends . Removal of the tubulin C termini also increases the affinity of MCAK for the microtubule lattice . Therefore , we propose that both the CT domain of MCAK and the C termini tubulin have important functions to reduce the affinity of MCAK for the microtubule lattice and facilitate MCAK diffusion , in part through electrostatic repulsion , in agreement with previous observations on a tailless MCAK ( Moore and Wordeman , 2004; Helenius et al . , 2006 ) . The CT domain of MCAK has a predominant effect on controlling MCAK affinity through intramolecular interactions , possibly through weak intramolecular interactions with the motor not engaged with the lattice or by interfering with the E-hook of tubulin . Taken together , both the CT domain of MCAK dimers and the acidic tails of tubulin effectively contribute to efficient microtubule lattice engagement , plus tip targeting , and activation of the depolymerase ( Helenius et al . , 2006 ) . This model could explain why MCAK activity is stimulated by the microtubule lattice and requires both its CT domain and the C terminus of tubulin for optimal activity ( Niederstrasser et al . , 2002; Helenius et al . , 2006; Cooper et al . , 2010 ) . The MCAK C-terminal binding motif ‘EEXXS’ is conserved across species from Drosophila to Human and is present in the kinesin-13 family member Kif2a , suggesting that this regulatory targeting mechanism is highly conserved ( Cameron et al . , 2006 ) . Interestingly , the C terminus of the Kinesin-13 member Kif2b diverges dramatically from Kif2a and MCAK ( Figure 4B ) . Kif2b binds to Cep170 through its C terminus to enhance its targeting to the spindle . The Kif2b C-terminal tail regulates kinesin activity through an alternate mechanism based on an association with binding partners ( Welburn and Cheeseman , 2012 ) . Our work reveals that MCAK undergoes long-range conformational changes during its transition from soluble to microtubule-bound state . The extreme C terminus of MCAK binds to the motor domain in solution and this interaction is abrogated upon MCAK binding to microtubules . This implies that a major microtubule-induced conformational change in MCAK occurs by disrupting the regulated interaction of the motor with the CT domain , which is triggered by the microtubule lattice itself . This event may also allow and require rearrangement of the neck region , which can swing into two distinct conformations on opposite faces of the MCAK motor domain . Recent work reported that MCAK undergoes long-range conformational changes upon binding to microtubules based on FRET ( Ems-McClung et al . , 2013 ) , although the nature of the change was unknown . Aurora B phosphorylation of the neck region has been proposed to control the long-range interactions with a C-terminal non-motor region of MCAK , however the molecular basis for this regulatory mechanism was lacking . Recent low resolution studies using deuterium-exchange and mass spectrometry also indicated that the C terminus of MCAK within the context of the full-length MCAK is more stable in solution than in the presence of microtubules ( Ems-McClung et al . , 2013; Burns et al . , 2014 ) . Thus , our studies reveal the molecular basis for this microtubule-induced change in conformation . An increasing number of kinesins also appear to be regulated by self-interactions ( reviewed in Welburn , 2013 ) . Kinesin-1 , Kif17 , and CENP-E can each undergo self-inhibition in solution to limit squandering of ATP ( Coy et al . , 1999; Friedman and Vale , 1999; Hackney and Stock , 2000; Espeut et al . , 2008 ) . We currently only have molecular insights into the inhibitory mechanism for Kinesin-1 , where one C-terminus binds at the interface between two motor domains to inhibit the molecular motor ( Hackney et al . , 2009; Kaan et al . , 2011 ) . Here , we demonstrate that certain features of the molecular inhibitory mechanism for processive kinesins can be extended to depolymerizing kinesins despite their different structural arrangement but that the function of this self-interaction is distinct . In both cases , the C terminus acts allosterically and stabilizes a motor domain dimer through a second dimerization site , distinct from the major dimerization domain ( Kaan et al . , 2011 ) . However , in the structure of the Kinesin-1 tail complex , the tail binds on a twofold symmetry axis utilizing two ionic interactions . The tail binds symmetrically and in both directions on the motor around a twofold symmetry axis . We found that the MCAK CT domain binds asymmetrically with multiple interactions with the motor domain and adopts only one potential orientation ( Figure 3 ) . Once this occurs , the interaction of the tail with the MCAK motor displays a reduced affinity for microtubules , similarly to Kinesin-1 . However , while Kinesin-1 auto-inhibition reduces ATPase activity , the C terminus of MCAK does not interfere with ATP hydrolysis in solution ( Hackney and Stock , 2000; Moore and Wordeman , 2004 ) . In addition , unlike motile kinesins , the alleviation of MCAK auto-inhibition is not stimulated by cargo proteins , but rather by the microtubule lattice itself , although the removal of the tail is in both cases electrostatically-driven ( Stock et al . , 1999 ) . In total , our work reveals that regulation of self-interactions in the kinesin superfamily emerges as a conserved feature , but that the nature of their regulation is distinct between processive and depolymerizing kinesins . Future structural work on MCAK will help us understand how this potent non-canonical kinesin functions in vivo .
His-MCAK ( 1–725 ) , His-MCAKS715E , and His-MCAKE712C were cloned in the pFL vector and subsequently used for Sf9 cell expression using the BEVS baculovirus expression system and protocol ( Fitzgerald et al . , 2006 ) . Full-length MCAK proteins were purified as described earlier ( Moore and Wordeman , 2004 ) . His-MCAK ( 183–583 , M ) and His-MCAK ( 1–583 , NM ) were subcloned in pET3aTr vector . For the CT domain of MCAK ( 700–725 ) , two long primers with BamHI and XhoI restriction sites; forward: 5′ – CCCGGATCCATCAAGGCCT TGCGCCTGGCCATGCAGCTGGAAGAGCAGGCTAGCAGACAAATAAGCAGCAAGAAACGGCCCCAGTGACTCGAGCCC – 3′ and reverse: 5′ – GGGCTCGAGTCACTGGGGCCGTTTCTTGCTGCTTATTTGTCTGCTAGCCTGCTCTTCCAGCTGCATGGCCAGGCGCAAGGCCTTGATGGATCCGGG – 3′ were first annealed together as double stranded DNA . The insert was then ligated into pGEX 6p1 ( GE Healthcare Life Sciences , UK ) . For the MCAK CTEEEEE , a gene encoding the C terminus of MCAK was synthesised by LifeTechnologies and was subcloned into the MCAK vector , described in Welburn and Cheeseman ( 2012 ) . Amino acids 716 and 718–721 were mutated to glutamates . Protein expression was induced by addition of 0 . 5 mM IPTG to BL21 ( DE3 ) Codon plus cells transformed with respective constructs at OD600 of 0 . 7–0 . 8 for 16 hr at 18°C . Cells were lysed by sonication in lysis buffer ( 50 mM Hepes , pH 7 . 4 , 200 mM NaCl , 1 mM MgCl2 , 1 mM PMSF , 1 mg/ml DNaseI , 2 mg/ml lysozyme , 10 mM Imidazole ) and clarified at 20 , 000 rpm for 1 hr at 4°C . His-tagged and GST-tagged proteins were subsequently purified using Ni-NTA–agarose beads and glutathione-sepharose beads , respectively ( GE Healthcare Life Sciences , UK ) ) according to the manufacturer's guidelines . MCAK constructs containing the motor domain were eluted with elution buffer ( 50 mM Hepes , pH 7 . 4 , 200 mM NaCl , 1 mM MgCl2 , 1 mM ATP , and 300 mM imidazole ) . Cleavage of the GST tag was performed using the GST-3C protease overnight at 4°C . Proteins were further purified using gel filtration chromatography pre-equilibrated in gel filtration buffer ( For full-length MCAK: 100 mM HEPES , pH 7 . 3 , 200 mM NaCl , 200 mM KCl , 1 mM DTT , 1 mM MgCl2 , 1 mM Na-EGTA , 1 mM ATP; for the motor domain constructs: 50 mM HEPES , pH 7 . 2 , 150 mM NaCl , 1 mM DTT , 1 mM MgCl2 , 1 mM Na-EGTA , 1 mM ATP; for CT domain constructs: 50 mM HEPES , pH 7 . 2 , 150 mM NaCl , 1 mM DTT , 1 mM MgCl2 , 1 mM Na-EGTA ) . Analytical gel filtration chromatography was performed using either a Superdex 75 or a Superose 6 10/300 GL column ( GE Healthcare , UK ) . To purify the CT domain alone , we cleaved GST and CT after gel filtration and a concentration step , and performed a glutathione affinity-purification third step to remove the GST and collect the CT domain . The CT domain was then further purified by separating it from remaining GST using a concentrator with a 3 kD-cutoff . Protein concentrations were determined with a combination of Bradford protein assays and densitometry of Coomassie-stained gels relative to a BSA standard . To visualize both the motor and CT domains on protein gels , 16% Tricine gels were used , according to the manufacturer's instructions ( Invitrogen , Life Technologies , Paisley , UK ) . MCAK motor domain was pre-treated with spectroscopy buffer ( 100 mM HEPES , pH 7 . 4 , 150 mM NaCl ) supplemented with 5 mM EDTA to remove any bound ADP . The protein was then desalted into spectroscopy buffer using a Disposable PD-10 Desalting Columns ( GE Healthcare Life Sciences , UK ) . The experiment was performed with a modified protocol as previously described ( Chadborn et al . , 1999 ) . A Cary 2200 spectrophotometer was used to measure absorption spectra; fluorescence was measured using an ISS K2 spectrofluorometer at 25°C . The intrinsic fluorescence of tryptophan and aromatic amino acids after excitation at 295 nm and 280 nm , respectively was recorded through an Ealing 340 nm centre-wavelength filter . The emission spectra were measured from 300 to 400 nm . The motor domain was diluted to 1 μM in spectroscopy buffer . First , the emission spectrum for the motor domain alone was recorded . Then the following concentrations of CT domain peptide , cleaved from GST and further purified , were titrated: 0 , 60 , 120 , 240 , 480 , 960 , 1920 , 3840 , 7680 , and 15 , 660 nM . The starting volume was 3 ml before peptide addition and was never increased more than 1% to negate any effect on fluorescence measurements . Because of the presence of GST ( 5% of the total peptide ) all measurement was corrected with measurement of buffer containing the same concentration of peptide . The change in fluorescence was calculated after they were normalized against each concentration of the CT domain alone in the spectroscopy buffer , to correct for non-specific fluorescence . Full-length MCAK complex was diluted in S buffer ( 50 mM NaCl , 20 mM Hepes pH 7 . 0 ) to 50 nM , in absence of nucleotide to prevent MCAK-dependent microtubule depolymerisation . To assemble an MCAK motor-CT domain complex , 50 nM of the motor domain and 100 nM of the CT domain were used . Microtubule binding assays were performed as described ( Cheeseman et al . , 2006 ) using equal volumes of taxol-stabilized microtubules in BRB80 and MCAK in S buffer . MCAK was quantified using anti-MCAK antibody against the C terminus of MCAK ( 709QLEEQASRQISS720 ) , generated by GL Biochem ( Shangai ) Ltd ( China ) or anti-His antibody to probe for the motor domain alone ( GE Healthcare Lifesciences , UK ) . The data from at least three independent experiments were fitted to a modified Hill equation to determine the apparent Kd . Microtubule depolymerization assays were performed essentially as described previously ( Hertzer et al . , 2006 ) . MCAK was diluted to 100 nM in S-buffer containing 1 mM DTT and 2 mM Mg-ATP . For depolymerization assays of the motor domain in presence of the CT domain after GST cleavage and removal , 50 nM of motor domain and 100 nM of the CT domain were used . The microtubule depolymerization assay was initiated by the addition of 2 μM taxol-stabilized microtubules to a reaction buffer containing MCAK . Reactions were incubated at room temperature with increasing times and followed by centrifugation to separate microtubules from free tubulin . The data are represented as mean ± SD from three independent experiments . The GST-CTE712C and CTE712C domains were cross-linked to the motor domain by incubating them with the motor domain in a buffer containing 100 mM HEPES , pH 7 . 4 , 150 mM NaCl but lacking DTT for 1 hr at 4°C . As a negative control the motor domain and the CTE712C domain were incubated in the identical buffer , supplemented with 5 mM DTT . Tubulin ( 5 mg/ml ) was first polymerized into MTs in the presence of 1 mM GTP and gradual addition of 0 . 05 µM , 0 . 5 µM and 2 µM taxol for 1 hr at 37°C . The polymerized MTs were then treated with 100 μg/ml subtilisin and incubated at 37°C for 10 min to cleave β-tubulin tails and 120 min to cleave both α- and β-tubulin tails . Each reaction was then terminated with the addition of 3 mM PMSF . DM1A ( Abcam , UK ) and c-terminal β-tubulin ( Sigma , UK ) were used to detect the α- and β-tails , respectively by western blotting . The subtilisin treated microtubules were then pelleted at 28°C in a TLA100 rotor at 80 , 000 rpm for 10 min and the microtubule pellets were resuspended in warm BRB80 buffer to obtain subtilisin treated microtubules . The cosedimentation assays were then performed as described before . Size-exclusion chromatography with on-line multi-angle light scattering ( SEC-MALS ) was performed using a GE Superdex 200 10/300 GL column on an ÄKTA FPLC system . MALS measurements were performed using a MiniDAWN in-line detector ( Wyatt Technology , Santa Barbara , CA , USA ) . MCAK motor domain and C terminus were at 2 mg/ml in 100 mM HEPES , pH 7 . 2 , 150 mM NaCl . Protein concentration was monitored using a UV monitor at 280 nm and a refractive index detector was set at 690 nm ( Optilab DSP , Wyatt Technology , Santa Barbara , CA , USA ) . Data were analyzed using Astra software ( Wyatt Technology , Santa Barbara , CA , USA ) using the refractive index detector and a refractive index increment ( dn/dc ) value of 0 . 185 ml/g . Gel phase distribution coefficients ( Kav ) were determined from the equation Kav = ( Ve − Vo ) / ( Vt − Vo ) , where Ve , Vo , and Vt represent the elution volume of the protein of interest , the column void volume and the total bed volume of the column , respectively . 1 mM MCAK motor domain ( PDB:2HEH , Addgene , Cambridge MA , USA ) was incubated with the CT peptide 709QLEEQASRQISS720 ( China peptides Co , Ltd , China ) in a ratio of 1:2 for 1 hr at 4°C before setting up crystallization trials . Elongated rectangular crystals appeared by vapor diffusion after two days in sitting drops using 24% wt/vol PEG 1500 and 20% vol/vol Glycerol as a precipitant . Crystals were grown in MRC 2 Well Crystallization Plate ( Hampton Research , Aliso Viejo , CA , USA ) at 19°C . Crystals were cryoprotected in a solution containing 28% wt/vol PEG 1500 and 30% vol/vol Glycerol and flash-frozen in dry liquid nitrogen . Diffraction data were recorded at Diamond Light Source on beamline ID24 at 100 K . Data were processed using XDS package ( Kabsch , 2010 ) and SCALA operated through the CCP4 suite GUI ( Collaborative Computational Project , Number 4 , 1994 ) . The structure of the MCAK motor-tail complex was solved by molecular replacement using the program MOLREP . The MCAK motor domain structure ( PDB code: 2HEH ) was used as a search model . Structure refinement was performed using Refmac5 and Phenix ( Adams et al . , 2010 ) . Model quality statistics are summarized in Table 1 . Figures were prepared using PyMOL ( Delano , 2002 ) . The final model and the structure factor amplitudes have been submitted to the Protein Data Bank under the accession code 4UBF . Transfection of GFP-MCAK and mCherry-EB3 constructs in HeLa cells was performed using Effectene ( Quiagen , Dusseldorf , Germany ) according to manufacturer's instructions . Images were acquired on a Nikon TIRF inverted microscope system with a perfect focus , with a 100× TIRF Apo 1 . 49 objective ( Nikon , UK ) using an Andor Zyla technology Scmos camera . Imaging was carried out at 37°C . Images were analyzed using ImagePro software and OMERO . Linescan averages were calculated from over 100 comets . | Within a cell , there is a scaffold-like structure called the cytoskeleton that provides shape and structural support , and acts as a transport network for the movement of molecules around the cell . This scaffold contains highly dynamic polymers called microtubules that are made from a protein called tubulin . The constant growth and shrinking of the ends of the microtubules is essential to rebuild and adapt the cytoskeleton according to the needs of the cell . A protein called MCAK belongs to a family of motor proteins that can move along microtubules . It generally binds to the ends of the microtubules to shorten them . Previous studies have found that a single MCAK protein binds to another MCAK protein to form a larger molecule known as a dimer . Part of the MCAK protein forms a so-called motor domain , which enables this protein to bind to the microtubules . One end of the protein , known as the C-terminus , controls the activity of this motor domain . However , it is not clear how this works . Talapatra et al . have now revealed the three-dimensional structure of MCAK's motor domain with the C-terminus using a technique called X-ray crystallography . The experiments show that the C-terminus binds to the motor domain , which promotes the formation of the dimers . A short stretch of amino acids—the building blocks of proteins—in the C-terminus interacts with two motor molecules . This ‘motif’ is also found in other similar proteins from a variety of animals . However , once MCAK binds to a microtubule , the microtubule triggers the release of the C-terminus from the motor domain . This allows MCAK to bind more strongly to the microtubule . The experiments also show that the binding of the C-terminus to the motor domain alters the ability of MCAK to associate with microtubules , which encourages the protein to reach the ends of the polymers . Future work is required to see whether other motor proteins work in a similar way . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | The C-terminal region of the motor protein MCAK controls its structure and activity through a conformational switch |
Little is known about the excess mortality caused by multidrug-resistant ( MDR ) bacterial infection in low- and middle-income countries ( LMICs ) . We retrospectively obtained microbiology laboratory and hospital databases of nine public hospitals in northeast Thailand from 2004 to 2010 , and linked these with the national death registry to obtain the 30-day mortality outcome . The 30-day mortality in those with MDR community-acquired bacteraemia , healthcare-associated bacteraemia , and hospital-acquired bacteraemia were 35% ( 549/1555 ) , 49% ( 247/500 ) , and 53% ( 640/1198 ) , respectively . We estimate that 19 , 122 of 45 , 209 ( 43% ) deaths in patients with hospital-acquired infection due to MDR bacteria in Thailand in 2010 represented excess mortality caused by MDR . We demonstrate that national statistics on the epidemiology and burden of MDR in LMICs could be improved by integrating information from readily available databases . The prevalence and mortality attributable to MDR in Thailand are high . This is likely to reflect the situation in other LMICs .
The emergence of antimicrobial resistance ( AMR ) is of major medical concern , particularly in low- and middle-income countries ( LMICs ) ( World Health Organization , 2014; Laxminarayan et al . , 2013 ) . In LMICs , antibiotic use is increasing with rising incomes , affordable antimicrobials and the lack of stewardship in hospital and poor control of over-the-counter sales . This is driving the emergence and spread of multidrug-resistant ( MDR ) pathogens in community and hospital settings . Hospital data from LMICs suggest that the cumulative incidence of community-acquired Extended-Spectrum Beta-Lactamase ( ESBL ) producing Escherichia coli and Klebsiella pneumoniae infections are increasing over time ( Kanoksil et al . , 2013; Ansari et al . , 2015 ) . A recent report from the International Nosocomial Infection Control Consortium ( INICC ) also showed that the prevalence of AMR organisms causing hospital-acquired infections ( HAI ) in ICUs in LMICs is much higher than those in the United States ( US ) ( Rosenthal et al . , 2014 ) . Attributable mortality , generally defined as the difference in mortality between those with and without the condition of interest , is an important parameter used to estimate the burden of AMR . In the US , it is estimated that mortality from infection attributable to AMR is 6 . 5% , ( Roberts et al . , 2009 ) leading to an estimate of 23 , 000 deaths attributable to AMR each year ( Center for Disease Controls and Prevention and U . S . Department of Health and Human Services , 2013 ) . In the European Union , it is estimated that the number of deaths attributable to selected antibiotic-resistant bacteria is about 25 , 000 each year ( European Centre for Disease Prevention and Control and European Medicines Agency , 2009 ) . There is limited information on mortality attributable to AMR in LMICs . The mortality attributable to ventilator-associated pneumonia in ICUs in Colombia , Peru , and Argentina is estimated to be 17% , 25% , and 35% , respectively , and is associated with a high percentage of AMR organisms ( Moreno et al . , 2006; Cuellar et al . , 2008; Rosenthal et al . , 2003 ) . The mortality attributable to ESBL and methicillin-resistance Staphylococcus aureus ( MRSA ) is estimated to be 27% and 34% in neonatal sepsis in Tanzania , respectively , ( Kayange et al . , 2010 ) which has been used to postulate an estimate that 58 , 319 deaths could be attributable to ESBL and MRSA in India alone ( Laxminarayan et al . , 2013 ) . In an effort to harmonize the surveillance systems of AMR , a joint initiative between the European Centre for Disease Prevention and Control ( ECDC ) and the Centres for Disease Prevention and Control ( CDC ) have developed standard definitions of multidrug-resistance ( MDR ) ( Magiorakos et al . , 2012 ) . We recently combined large data sets from multiple sources including microbiology databases , hospital admission databases , and the national death registry from a sample of ten public hospitals in northeast Thailand from 2004 to 2010 ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . We defined community-acquired bacteraemia ( CAB ) as the isolation of a pathogenic bacterium from blood taken in the first 2 days of admission and without a hospital stay in the 30 days prior to admission , healthcare-associated bacteraemia ( HCAB ) as the isolation of a pathogenic bacterium from blood taken in the first 2 days of admission and with a hospital stay within 30 days prior to the admission , and hospital-acquired bacteraemia ( HAB ) as the isolation of a pathogenic bacterium from blood taken after the first 2 days of admission ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . We reported an increase in the incidence of CAB , HCAB and HAB over the study period , and that bacteraemia was associated with high case fatality rates ( 37 . 5% , 41 . 8% and 45 . 5% , respectively ) ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . Here , we apply ECDC/CDC standard definitions of MDR to this large data set to evaluate the prevalence , trends , and mortality attributable to MDR bacteria isolated from the blood . We then estimate the number of deaths attributable to MDR in Thailand nationwide .
Of CAB , HCAB and HAB caused by S . aureus , 8% , 28% , and 50% were caused by MDR S . aureus , respectively ( p<0 . 001 ) . Almost all MDR S . aureus were MRSA ( 92% [357/389] , Table 2 ) . We did not observe a trend in the proportion of S . aureus bacteraemia being caused by MRSA ( Figure 2 ) . Vancomycin non-susceptible S . aureus was found in <1% of tested isolates ( 6/1380 ) . 10 . 7554/eLife . 18082 . 005Figure 2 . Trends in proportions of Staphylococcus aureus bacteraemia being caused by MRSA in Northeast Thailand . ( A ) community-acquired , ( B ) healthcare-associated and ( C ) hospital-acquired Staphylococcus aureus bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 00510 . 7554/eLife . 18082 . 006Table 2 . Antibiogram of S . aureus causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 006Antibiotic categoryAntibiotic agentsCAB ( n = 1176 patients ) HCAB ( n = 259 patients ) HAB ( n = 446 patients ) p valuesAminoglycosidesGentamicin24/484 ( 5% ) 16/84 ( 19% ) 66/151 ( 44% ) <0 . 001AnsamycinsRifampin2/129 ( 2% ) 1/19 ( 5% ) 0/38 ( 0% ) 0 . 37Anti-MRSA cephalosporinsCeftarolineNANANA-CefamycinsOxacillin *80/1145 ( 7% ) 67/247 ( 27% ) 210/441 ( 48% ) <0 . 001FluoroquinolonesCiprofloxacin3/45 ( 7% ) 2/8 ( 25% ) 4/10 ( 40% ) 0 . 01MoxifloxacinNANANA-Folate pathway inhibitorsTrimethoprim-sulphamethoxazole99/1139 ( 9% ) 57/251 ( 23% ) 185/438 ( 42% ) <0 . 001FucidanesFusidic acid33/618 ( 5% ) 4/170 ( 2% ) 12/291 ( 4% ) 0 . 26GlycopeptidesVancomycin †4/833 ( 0 . 5% ) 0/190 ( 0% ) 2/357 ( 1% ) 0 . 86Teicoplanin2/66 ( 3% ) 1/17 ( 6% ) 0/17 ( 0% ) 0 . 72TelavancinNANANA-GlycylcyclinesTigecyclineNANANA-LincosamidesClindamycin118/1147 ( 10% ) 77/251 ( 31% ) 202/438 ( 46% ) <0 . 001LipopeptidesDaptomycinNANANA-MacrolidesErythromycin138/1116 ( 12% ) 76/240 ( 32% ) 222/429 ( 52% ) <0 . 001OxazolidinonesLinezolid0/81 ( 0% ) 0/16 ( 0% ) 0/32 ( 0% ) -PhenicolsChloramphenicol6/86 ( 7% ) 4/24 ( 17% ) 2/14 ( 14% ) 0 . 21Phosphonic acidsFosfomycin14/361 ( 4% ) 10/66 ( 15% ) 24/141 ( 17% ) <0 . 001StreptograminsQuinupristin-dalfopristinNANANA-TetracyclinesTetracyclineNANANA-DoxycyclineNANANA-MinocyclineNANANA-MDR94/1176 ( 8% ) 73/259 ( 28% ) 222/446 ( 50% ) <0 . 001NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR ( one or more of these have to apply ) : ( i ) an MRSA is always considered MDR by virtue of being an MRSA ( ii ) non-susceptible to ≥1 agent in ≥3 antimicrobial categories . * Defined by using a 30 μg cefoxitin disc and an inhibition zone diameter of <21 mm . † Defined by using a 30 μg vancomycin disc and an inhibition zone diameter of <15 mm . MDR Enterococcus spp . were not found in CAB ( 0/176 ) and HCAB ( 0/49 ) , while 3% ( 4/117 ) of Enterococcus spp . causing HAB were MDR . Of CAB caused by Enterococcus spp . , 15% ( 20/134 ) and 23% ( 35/153 ) was non-susceptible to ampicillin and gentamicin , respectively ( Table 3 ) , while 42% ( 34/81 ) and 62% ( 63/101 ) of HAB caused by Enterococcus spp . were non-susceptible to those agents , respectively ( both p<0 . 001 ) . Vancomycin non-susceptible Enterococcus spp . was found in 4% of tested isolates ( 15/338 ) . Of CAB , HCAB and HAB caused by E . coli , 35% , 58% and 63% were caused by MDR E . coli , respectively ( p<0 . 001 ) . Of E . coli causing CAB , 79% ( 2246/2843 ) , 16% ( 501/3076 ) , 24% ( 728/3000 ) , 58% ( 1738/3007 ) , and 17% ( 559/3346 ) were non-susceptible to commonly-used antimicrobials for community-acquired infections such as ampicillin , cefotaxime , ciprofloxacin , trimethoprim-sulphamethoxazole , and gentamicin , respectively ( Table 4 ) . From 2004 to 2010 , the proportions of community-acquired E . coli bacteraemia being caused by E . coli non-susceptible to extended-spectrum cephalosporins rose from 5% ( 9/169 ) to 23% ( 186/815 ) ( p=0 . 04 ) ( Figure 3 ) . The proportions of healthcare-associated and hospital-acquired E . coli bacteraemia being caused by E . coli non-susceptible to extended-spectrum cephalosporins were high ( 44% [204/465] and 52% [190/368] , respectively ) , but a significant trend over time was not observed ( p=0 . 18 and p=0 . 63 , respectively ) . Carbapenem non-susceptible E . coli was found in <1% of tested isolates ( 12/3838 ) . 10 . 7554/eLife . 18082 . 007Figure 3 . Trends in proportions of Escherichia coli bacteraemia being caused by E . coli non-susceptible to extended-spectrum cephalosporins in Northeast Thailand . ( A ) community-acquired , ( B ) healthcare-associated and ( C ) hospital-acquired E . coli bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 00710 . 7554/eLife . 18082 . 008Table 3 . Antibiogram of Enterococcus spp . causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 008Antibiotic categoryAntibiotic agentsCAB ( n = 176 patients ) HCAB ( n = 49 patients ) HAB ( n = 117 patients ) p valuesAminoglycosidesGentamicin ( high level ) 35/153 ( 23% ) 24/45 ( 53% ) 63/101 ( 62% ) <0 . 001StreptomycinStreptomycin ( high level ) NANANA-Carbapenems*ImipenemNANANA-Meropenem1/1 ( 100% ) NA3/5 ( 60% ) >0 . 99DoripenemNANANA-FluoroquinolonesCiprofloxacin37/44 ( 84% ) 9/10 ( 90% ) 31/37 ( 84% ) >0 . 99Levofloxacin5/18 ( 28% ) 1/6 ( 17% ) 11/15 ( 73% ) 0 . 01MoxifloxacinNANANA-GlycopeptidesVancomycin9/176 ( 5% ) 0/49 ( 0% ) 6/113 ( 5% ) 0 . 27Teicoplanin0/11 ( 0% ) 0/4 ( 0% ) 0/10 ( 0% ) -GlycylcyclinesTigecyclineNANANA-LipopeptidesDaptomycinNANANA-OxazolidinonesLinezolid0/8 ( 0% ) 0/2 ( 0% ) 0/4 ( 0% ) -PenicillinsAmpicillin20/134 ( 15% ) 6/37 ( 16% ) 34/81 ( 42% ) <0 . 001Streptogramins*Quinupristin-dalfopristinNANANA-TetracyclineDoxycyclineNANANA-MinocyclineNANANA-MDR0/176 ( 0% ) 0/49 ( 0% ) 4/117 ( 3% ) 0 . 02NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR: non-susceptible to ≥1 agent in ≥3 antimicrobial categories . *Intrinsic resistance in E . faecium against carbapenems and in E . faecalis against streptogramins . When a species has intrinsic resistance to an antimicrobial category , that category is removed prior to applying the criteria for the MDR definition and is not counted when calculating the number of categories to which the bacterial isolate is non-susceptible . 10 . 7554/eLife . 18082 . 009Table 4 . Antibiogram of E . coli causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 009Antibiotic categoryAntibiotic agentsCAB ( n = 3382 patients ) HCAB ( n = 494 patients ) HAB ( n = 403 patients ) p valuesAminoglycosidesGentamicin559/3346 ( 17% ) 166/484 ( 34% ) 178/398 ( 45% ) <0 . 001TobramycinNANANA-Amikacin72/2685 ( 3% ) 26/397 ( 7% ) 32/326 ( 10% ) <0 . 001Netilmicin68/1394 ( 5% ) 25/259 ( 10% ) 42/254 ( 17% ) <0 . 001Anti-MRSA cephalosporinsCeftarolineNANANA-Antipseudomonal penicillins + β lactamase inhibitorsTicarcillin-clauvanic acidNANANA-Piperacillin-tazobactam23/511 ( 5% ) 10/103 ( 10% ) 15/89 ( 17% ) <0 . 001CarbapenemsErtapenem4/1325 ( <1% ) 1/235 ( <1% ) 4/205 ( 2% ) 0 . 02Imipenem3/2449 ( <1% ) 0/386 ( 0% ) 3/344 ( 1% ) 0 . 04Meropenem0/1988 ( 0% ) 1/314 ( <1% ) 1/244 ( <1% ) 0 . 05Non-extended spectrum cephalosporinsCefazolin468/1095 ( 43% ) 115/174 ( 66% ) 80/102 ( 78% ) <0 . 001Cefuroxime219/1438 ( 15% ) 96/226 ( 42% ) 102/202 ( 50% ) <0 . 001Extended-spectrum cephalosporinsCefotaxime501/3076 ( 16% ) 199/455 ( 44% ) 185/361 ( 51% ) <0 . 001Ceftazidime392/3020 ( 13% ) 165/446 ( 37% ) 164/351 ( 47% ) <0 . 001Cefepime30/293 ( 10% ) 12/42 ( 29% ) 18/53 ( 34% ) <0 . 001CephamycinsCefoxitin36/1200 ( 3% ) 16/215 ( 7% ) 16/195 ( 8% ) <0 . 001CefotetanNANANA-FluoroquinolonesCiprofloxacin728/3000 ( 24% ) 221/452 ( 49% ) 171/384 ( 45% ) <0 . 001Folate pathway inhibitorsTrimethoprim-sulphamethoxazole1738/3007 ( 58% ) 294/442 ( 67% ) 225/350 ( 64% ) <0 . 001GlycylcyclinesTigecycline0/7 ( 0% ) NA0/1 ( 0% ) -MonobactamsAztreonamNANANA-PenicillinsAmpicillin2246/2843 ( 79% ) 371/420 ( 88% ) 301/342 ( 88% ) <0 . 001Penicillins + β lactamase inhibitorsAmoxicillin-clavulanic acid790/3074 ( 26% ) 191/463 ( 41% ) 158/373 ( 42% ) <0 . 001Ampicillin-sulbactam83/296 ( 28% ) 18/48 ( 38% ) 12/25 ( 48% ) 0 . 06PhenicolsChloramphenicol14/63 ( 22% ) 1/4 ( 25% ) 3/5 ( 60% ) 0 . 14Phosphonic acidsFosfomycinNANANA-PolymyxinsColistin*2/34 ( 6% ) 0/6 ( 0% ) 1/6 ( 17% ) 0 . 61MDR1177/3382 ( 35% ) 288/494 ( 58% ) 252/403 ( 63% ) <0 . 001NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR: non-susceptible to ≥1 agent in ≥3 antimicrobial categories . *Defined by using an inhibition zone of <11 mm . Of CAB , HCAB and HAB caused by K . pneumoniae , 14% , 36% , and 66% were caused by MDR K . pneumoniae , respectively ( p<0 . 001 ) . Of K . pneumoniae causing CAB , 16% ( 146/902 ) , 16% ( 143/894 ) , 23% ( 198/876 ) , and 9% ( 94/999 ) were non-susceptible to cefotaxime , ciprofloxacin , trimethoprim-sulphamethoxazole and gentamicin , respectively ( Table 5 ) . From 2004 to 2010 , the proportions of community-acquired K . pneumoniae bacteraemia being caused by K . pneumoniae non-susceptible to extended-spectrum cephalosporins rose from 12% ( 6/50 ) to 24% ( 64/263 ) ( p=0 . 04 ) ( Figure 4 ) . The proportions of healthcare-associated and hospital-acquired K . pneumoniae bacteraemia being caused by K . pneumoniae non-susceptible to extended-spectrum cephalosporins were also high ( 40% [71/177] and 71% [304/431] , respectively ) , but a significant trend over time was not observed ( p=0 . 16 and p=0 . 35 , respectively ) . Carbapenem non-susceptible K . pneumoniae was found in <1% of tested isolates ( 11/1555 ) . 10 . 7554/eLife . 18082 . 010Figure 4 . Trends in proportions of Klebsiella pneumoniae bacteraemia being caused by K . pneumoniae non-susceptible to extended-spectrum cephalosporins in Northeast Thailand . ( A ) community-acquired , ( B ) healthcare-associated and ( C ) hospital-acquired K . pneumoniae bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 01010 . 7554/eLife . 18082 . 011Table 5 . Antibiogram of K . pneumoniae causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 011Antibiotic categoryAntibiotic agentsCAB ( n = 1010 patients ) HCAB ( n = 196 patients ) HAB ( n = 455 patients ) p valuesAminoglycosidesGentamicin94/999 ( 9% ) 53/193 ( 27% ) 265/444 ( 60% ) <0 . 001TobramycinNANANA-Amikacin17/815 ( 2% ) 12/157 ( 8% ) 109/398 ( 27% ) <0 . 001Netilmicin20/450 ( 4% ) 23/112 ( 21% ) 124/320 ( 39% ) <0 . 001Anti-MRSA cephalosporinsCeftarolineNANANA-Antipseudomonal penicillins + β lactamase inhibitorsTicarcillin-clauvanic acidNANANA-Piperacillin-tazobactam24/166 ( 14% ) 14/32 ( 44% ) 73/121 ( 60% ) <0 . 001CarbapenemsErtapenem2/432 ( 0% ) 1/100 ( 1% ) 5/264 ( 2% ) 0 . 17Imipenem1/778 ( 0% ) 1/164 ( 1% ) 2/408 ( 0% ) 0 . 24Meropenem0/583 ( 0% ) 1/113 ( 1% ) 2/317 ( 1% ) 0 . 10Non-extended spectrum cephalosporinsCefazolin76/319 ( 24% ) 30/60 ( 50% ) 101/127 ( 80% ) <0 . 001Cefuroxime81/478 ( 17% ) 35/98 ( 36% ) 161/231 ( 70% ) <0 . 001Extended-spectrum cephalosporinsCefotaxime146/902 ( 16% ) 71/173 ( 41% ) 298/424 ( 70% ) <0 . 001Ceftazidime124/927 ( 13% ) 63/176 ( 36% ) 295/430 ( 69% ) <0 . 001Cefepime5/100 ( 5% ) 8/22 ( 36% ) 25/51 ( 49% ) <0 . 001CephamycinsCefoxitin15/396 ( 4% ) 10/95 ( 11% ) 14/230 ( 6% ) 0 . 03CefotetanNANANA-FluoroquinolonesCiprofloxacin143/894 ( 16% ) 66/176 ( 38% ) 187/430 ( 43% ) <0 . 001Folate pathway inhibitorsTrimethoprim-sulphamethoxazole198/876 ( 23% ) 69/171 ( 40% ) 219/407 ( 54% ) <0 . 001GlycylcyclinesTigecyclineNANANA-MonobactamsAztreonamNANANA-Penicillins + β lactamase inhibitorsAmoxicillin-clavulanic acid131/945 ( 14% ) 68/183 ( 37% ) 291/443 ( 66% ) <0 . 001Ampicillin-sulbactam20/105 ( 19% ) 9/17 ( 53% ) 23/38 ( 61% ) <0 . 001PhenicolsChloramphenicol4/19 ( 21% ) 0/2 ( 0% ) 0/3 ( 0% ) >0 . 99Phosphonic acidsFosfomycinNANANA-PolymyxinsColistin *0/6 ( 0% ) 0/2 ( 0% ) 0/5 ( 0% ) -MDR146/1010 ( 14% ) 71/196 ( 36% ) 301/455 ( 66% ) <0 . 001NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR: non-susceptible to ≥1 agent in ≥3 antimicrobial categories . * Defined by using an inhibition zone of <11 mm . Of CAB , HCAB and HAB caused by P . aeruginosa , 5% , 10% , and 25% were caused by MDR P . aeruginosa , respectively ( p<0 . 001 ) . Of P . aeruginosa causing HAB , 38% ( 68/179 ) , 27% ( 48/177 ) , 23% ( 39/169 ) and 26% ( 42/164 ) were non-susceptible to commonly-used antimicrobials for HAI such as ceftazidime , amikacin , ciprofloxacin and carbapenems , respectively ( Table 6 ) . We did not observe a trend in the proportions of P . aeruginosa being caused by P . aeruginosa that were non-susceptible to any specific antibiotic group ( Figure 5 ) . 10 . 7554/eLife . 18082 . 012Figure 5 . Trends in proportions of Pseudomonas aeruginosa bacteraemia being caused by P . aeruginosa non-susceptible to carbapenem in Northeast Thailand . ( A ) community-acquired , ( B ) healthcare-associated and ( C ) hospital-acquired Pseudomonas aeruginosa bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 01210 . 7554/eLife . 18082 . 013Table 6 . Antibiogram of P . aeruginosa causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 013Antibiotic categoryAntibiotic agentsCAB ( n = 286 patients ) HCAB ( n = 103 patients ) HAB ( n = 179 patients ) p valuesAminoglycosidesGentamicin29/235 ( 12% ) 13/88 ( 15% ) 60/140 ( 43% ) <0 . 001TobramycinNANANA-Amikacin27/284 ( 10% ) 13/100 ( 13% ) 48/177 ( 27% ) <0 . 001Netilmicin8/155 ( 5% ) 5/67 ( 7% ) 34/120 ( 28% ) <0 . 001Antipseudomonal carbapenemsImipenem14/238 ( 6% ) 6/86 ( 7% ) 37/154 ( 24% ) <0 . 001Meropenem9/163 ( 6% ) 8/73 ( 11% ) 24/125 ( 19% ) 0 . 001Doripenem2/17 ( 12% ) 0/3 ( 0% ) 2/2 ( 100% ) 0 . 04Antipseudomonal cephalosporinsCeftazidime29/280 ( 10% ) 16/103 ( 16% ) 68/179 ( 38% ) <0 . 001Cefepime2/36 ( 6% ) 2/18 ( 11% ) 10/28 ( 36% ) 0 . 01Antipseudomonal fluoroquinolonesCiprofloxacin24/275 ( 9% ) 12/101 ( 12% ) 39/169 ( 23% ) <0 . 001Levofloxacin0/1 ( 0% ) 1/1 ( 100% ) 1/1 ( 100% ) >0 . 99Antipseudomonal penicillins + β lactamase inhibitorsTicarcillin-clauvanic acidNANANA-Piperacillin-tazobactam8/85 ( 9% ) 6/38 ( 16% ) 8/46 ( 17% ) 0 . 37MonobactamsAztreonamNANANA-Phosphonic acidsFosfomycin1/1 ( 100% ) NANA-PolymyxinsColistin0/7 ( 0% ) 0/3 ( 0% ) 1/7 ( 14% ) >0 . 99Polymyxin BNANANA-MDR13/286 ( 5% ) 10/103 ( 10% ) 45/179 ( 25% ) <0 . 001NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR: non-susceptible to ≥1 agent in ≥3 antimicrobial categories . Of CAB , HCAB and HAB caused by Acinetobacter spp . , 28% , 50% , and 75% were caused by MDR Acinetobacter spp . , respectively ( p<0 . 001 ) . Of Acinetobacter spp . causing HAB , 75% ( 377/500 ) , 63% ( 310/495 ) , 67% ( 322/481 ) and 64% ( 315/490 ) were non-susceptible to ceftazidime , amikacin , ciprofloxacin and carbapenems , respectively ( Table 7 ) . There was borderline evidence that the proportion of hospital-acquired Acinetobacter spp . bacteraemia being caused by Acinetobacter spp . non-susceptible to carbapenem rose from 49% ( 19/39 ) in 2004 to 65% ( 70/108 ) in 2010 ( p=0 . 10 ) ( Figure 6 ) . Non-susceptibility to colistin was observed in 3% of tested isolates ( 2/63 ) . 10 . 7554/eLife . 18082 . 014Figure 6 . Trends in proportions of Acinetobacter spp bacteraemia being caused by Acinetobacter spp non-susceptible to carbapenem in Northeast Thailand . ( A ) community-acquired , ( B ) healthcare-associated and ( C ) hospital-acquired Acinetobacter spp bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 01410 . 7554/eLife . 18082 . 015Table 7 . Antibiogram of Acinetobacter spp . causing bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 015Antibiotic categoryAntibiotic agentsCAB ( n = 449 patients ) HCAB ( n = 115 patients ) HAB ( n = 501 patients ) p valuesAminoglycosidesGentamicin112/390 ( 29% ) 45/105 ( 43% ) 310/455 ( 68% ) <0 . 001TobramycinNANANA-Amikacin123/442 ( 28% ) 45/112 ( 40% ) 310/495 ( 63% ) <0 . 001Netilmicin44/203 ( 22% ) 24/64 ( 38% ) 224/381 ( 59% ) <0 . 001Antipseudomonal carbapenemsImipenem87/397 ( 22% ) 37/102 ( 36% ) 293/459 ( 64% ) <0 . 001Meropenem65/284 ( 23% ) 32/81 ( 40% ) 229/348 ( 66% ) <0 . 001Doripenem16/45 ( 36% ) 9/10 ( 90% ) 6/7 ( 86% ) 0 . 001Antipseudomonal fluoroquinolonesCiprofloxacin84/413 ( 20% ) 53/106 ( 50% ) 322/481 ( 67% ) <0 . 001Levofloxacin2/5 ( 40% ) 2/2 ( 100% ) 8/9 ( 89% ) 0 . 11Antipseudomonal penicillins + β lactamase inhibitorsTicarcillin- clauvanic acidNANANA-Piperacillin-tazobactam22/98 ( 22% ) 13/28 ( 46% ) 74/106 ( 70% ) <0 . 001Extended-spectrum cephalosporinsCefotaxime242/291 ( 83% ) 89/94 ( 95% ) 407/420 ( 97% ) <0 . 001Ceftazidime133/448 ( 30% ) 61/114 ( 54% ) 377/500 ( 75% ) <0 . 001Cefepime18/53 ( 34% ) 10/22 ( 45% ) 95/133 ( 71% ) <0 . 001Folate pathway inhibitorTrimethopri-sulphamethoxazole119/356 ( 33% ) 55/99 ( 56% ) 333/435 ( 77% ) <0 . 001Penicillins + β lactamase inhibitorsAmpicillin-sulbactam43/134 ( 32% ) 16/29 ( 55% ) 79/115 ( 69% ) <0 . 001PolymyxinsColistin *2/16 ( 13% ) 0/14 ( 0% ) 0/33 ( 0% ) 0 . 11Polymyxin BNANANA-TetracyclinesTetracyclineNANANA-DoxycyclineNANANA-MinocyclineNANANA-MDR125/449 ( 28% ) 58/115 ( 50% ) 374/501 ( 75% ) <0 . 001NOTE: Data are number of isolates demonstrating non-susceptible to the antimicrobial over the total number of isolates tested ( % ) . CAB = Community-acquired bacteraemia , HCAB = Healthcare-associated bacteraemia , HAB = Hospital-acquired bacteraemia , and NA = Not available . The first isolate of each patient was used . MDR: non-susceptible to ≥1 agent in ≥3 antimicrobial categories . * Defined by using an inhibition zone of <11 mm . The 30-day mortality in patients with CAB , HCAB and HAB caused by MDR bacteria was 35% ( 549/1555 ) , 49% ( 247/500 ) , and 53% ( 640/1198 ) , compared with 32% ( 1595/4924 ) , 37% ( 264/716 ) , and 42% ( 383/903 ) in CAB , HCAB , and HAB caused by non-MDR bacteria , respectively ( Figure 7 ) . In the final multivariable logistic regression model , gender , age group , year of admission and time to bacteraemia ( for HAB ) were included ( Supplementary file 2 ) . 10 . 7554/eLife . 18082 . 016Figure 7 . Forest plot of mortality in patients with MDR bacteraemia compared with non-MDR bacteraemia in Northeast Thailand . ( A ) Community-acquired bacteraemia . ( B ) Healthcare-associated bacteraemia . ( C ) Hospital-acquired bacteraemia . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 01610 . 7554/eLife . 18082 . 017Figure 7—source data 1 . Mortality in patients with MDR and non-MDR bacteraemia in Northeast Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 017 If excess mortality in patients infected with MDR bacteria after adjusting for confounding factors in the final multivariable model is assumed to be caused by MDR , the mortality attributable to MDR was 7% ( 95%CI 4% to 10% , p<0 . 001 ) in CAB , 15% ( 95%CI 5% to 24% , p<0 . 001 ) in HCAB and 15% ( 95%CI 2% to 27% , p<0 . 001 ) in HAB ( Figure 7 ) . Heterogeneity between different organisms was clearly observed in HAB ( p<0 . 001 ) , and borderline evidence of heterogeneity was observed in HCAB ( p=0 . 09 ) . The heterogeneity observed in HCAB and HAB was largely caused by MDR Acinetobacter spp . ( Figure 7B and C ) . Mortality attributed to MDR was highest for hospital-acquired MDR Acinetobacter bacteraemia ( 41% ) . Using our estimated mortality attributed to MDR bacteraemia ( Figure 7C ) and national statistics of HAI caused by MDR bacteria , we further estimated that 19 , 122 of 45 , 209 ( 43% ) deaths in patients with HAI due to MDR bacteria in Thailand in 2010 represented excess mortality caused by MDR ( Table 8 ) . All parameters used to estimate the number of excess deaths in Thailand are shown in Supplementary file 2 . 10 . 7554/eLife . 18082 . 018Table 8 . Estimates of mortality attributable to multidrug-resistance ( MDR ) in hospital-acquired infection ( HAI ) in Thailand . DOI: http://dx . doi . org/10 . 7554/eLife . 18082 . 018PathogensNo of patients*Estimated mortality ( % ) †Estimated mortality if the infections were caused by non-MDR organisms ( % ) † , ‡Estimated excess mortality caused by MDR ( % ) † , ‡MDR Staphylococcus aureus18 , 7258262 ( 44% ) 5463 ( 29% ) 2799 ( 15% ) MDR Escherichia coli11 , 1162163 ( 19% ) 1566 ( 14% ) 597 ( 5% ) MDR Klebsiella pneumoniae15 , 2395267 ( 35% ) 4979 ( 33% ) 288 ( 2% ) MDR Pseudomonas aeruginosa61183966 ( 65% ) 3696 ( 60% ) 270 ( 4% ) MDR Acinetobacter spp36 , 55325 , 551 ( 70% ) 10 , 383 ( 28% ) 15 , 168 ( 41% ) Total87 , 75145 , 209 ( 52% ) 26 , 087 ( 30% ) 19 , 122 ( 22% ) *Cumulative incidence of antimicrobial resistant HAI in Thailand 2010 estimated by Pumart et al . ( 2012 ) . †All parameters used to estimate the mortality and excess mortality are shown in Supplementary file 2 . ‡Excess mortality caused by MDR ( mortality attributable to MDR ) was defined as the difference in mortality of patients with MDR infection and their mortality if they were infected with non-MDR infections .
This study presents detailed antimicrobial susceptibility data on common pathogenic bacteria , the association of MDR with infection acquisition ( community-acquired , healthcare-associated and hospital-acquired ) , and excess mortality from MDR in a developing country . Our estimate of excess deaths caused by MDR in HAI patients in Thailand ( 19 , 122 deaths per year in a country of about 66 million population in 2010 ) is large compared to those estimated in USA ( 23 , 000 death per year in a country of 316 million population in 2013 ) ( Center for Disease Controls and Prevention and U . S . Department of Health and Human Services , 2013 ) and the European Union ( 25 , 000 deaths per year in EU of about 500 million population in 2007 ) ( European Centre for Disease Prevention and Control and European Medicines Agency , 2009 ) . Our study highlights the need for public health officials and international health organizations to improve systems to track and reduce the burden of AMR in LMICs . Our estimated mortality for those with MDR HAI ( 45 , 209 , Table 2 ) is higher than those previously published by Pumart et al . ( 38 , 481 ) ( Pumart et al . , 2012 ) , probably because we used 30-day mortality rather than in-hospital mortality . Acinetobacter spp . is increasingly recognized as an important cause of HAI , ( Munoz-Price and Weinstein , 2008; Peleg and Hooper , 2010 ) and our study confirms the importance of this species as a leading cause of hospital-acquired MDR infection in a developing tropical country ( Hongsuwan et al . , 2014; Nhu et al . , 2014 ) . The high mortality observed in MDR Acinetobacter spp . bacteraemia is because treatment options are limited and those available are associated with toxicity ( Fishbain and Peleg , 2010 ) . The high proportions of S . aureus , E . coli and K . pneumoniae bactaeremia being caused by MRSA and E . coli and K . pneumoniae non-susceptible to extended-spectrum cephalosporins , respectively , are consistent with previous reports from other tropical countries ( Moreno et al . , 2006; Cuellar et al . , 2008; Rosenthal et al . , 2003 ) . The rising proportions of community-acquired E . coli and K . pneumoniae bacteraemia being caused by E . coli and K . pneumoniae non-susceptible to extended-spectrum cephalosporins , and the rising proportion of hospital-acquired Acinetobacter bacteraemia being causing Acinetobacter non-susceptible to carbapenem suggest that the burden of AMR in Thailand is deteriorating over time . A limitation of this study is that more complete clinical data were not available . Mortality attributable to MDR could be overestimated if MDR infection was associated with more severely ill patients in ICUs . However , our estimated attributable mortality is comparable to the previous reports . For example , our estimated mortality attributable to MDR Acinetobacter bacteraemia ( 40 . 6% ) is comparable to the mortality attributable to imipenem resistant Acinetobacter bacteraemia reported by Kwon et al . in Korea ( 41 . 1% ) , which was adjusted by severity of illness ( Kwon et al . , 2007; Falagas and Rafailidis , 2007 ) . In addition , data on hospitalization in other hospitals not participating in the study ( for example , a smaller community hospital or a private hospital in the province ) were not available , which could have resulted in a misclassification of CAB , HCAB and HAB in some cases . We also note that data on attributable mortality from different countries is difficult to compare because of the differing study designs . For example , our mortality outcome is the overall 30-day mortality , including both directly and indirectly contributed to MDR , while an EU study only considered directly attributable deaths ( European Centre for Disease Prevention and Control and European Medicines Agency , 2009 ) . The p values for trends were generated by the stratification method; therefore , the analysis was not biased towards the increasing availability of the hospital data over the study period . Nonetheless , the trends could be affected by an increasing use of blood culture , changes in antimicrobial agents tested for susceptibility , and greater standardization of laboratory methodologies over time ( Opartkiattikul and Bejrachandra , 2002 ) . It is likely that the burdens of MDR similar to that observed in our study are present in many secondary and tertiary hospitals in tropical LMICs , particularly where extended-spectrum cephalosporins and carbapenem are widely used . Nonetheless , resources for diagnostics , methodologies used in the laboratories , and study designs need to be carefully considered when performing a comparison between different settings . Despite the increasing global focus on AMR in LMICs , considerable gaps remain in our understanding of the scale of the problem . We have demonstrated that the integration of information from readily available routinely collected databases can provide valuable information on the trends and mortality attributable to AMR in Thailand . The methodology used in our study could be applied to explore the burden of AMR in other LMICs where microbiological facilities and hospital admission database are available .
From 2004 to 2010 , Thailand was classified as a lower-middle income country with an average income of $4782 per person per year in 2010 ( WorldBank , 2015 ) . Northeast Thailand consists of 20 provinces covering 170 , 226 km and had an estimated population of 21 . 4 million in 2010 . A large proportion of the population in this area lives in rural settings , with most adults engaging in agriculture with an emphasis on rice farming . Healthcare in Thailand is mainly provided by government-owned hospitals . Each province has a provincial hospital , which provides services and care to individuals within its catchment area . Additionally , provincial hospitals act as referral hospitals for smaller community hospitals for severely ill patients . All provincial hospitals receive comparable resources , which are proportional to the respective populations of the provinces . Provincial hospitals , unlike smaller community hospitals , are equipped with a microbiology laboratory capable of performing bacterial culture using standard methodologies for bacterial identification and susceptibility testing provided by the Bureau of Laboratory Quality and Standards , Ministry of Public Health ( MoPH ) , Thailand ( Opartkiattikul and Bejrachandra , 2002 ) . During the study period , antimicrobial susceptibility was determined in all study hospitals using the disc diffusion method according to Clinical and Laboratory Standards Institute ( CLSI ) ( National Committee for Clinical Laboratory Standards , 2004 ) . We conducted a retrospective , multicentre surveillance study of all provincial hospitals in Northeast Thailand . From the hospitals that agreed to participate , data were collected from microbiology and hospital databases between January 2004 and December 2010 . Hospital number ( HN ) and admission number ( AN ) were used as a record linkage between the two databases and to identify individuals who had repeat admissions . The death registry for Northeast Thailand was obtained from the Ministry of Interior ( MoI ) , Thailand , and used to identify patients who were discharged from hospital but died at home shortly after , which is a common practice in Thailand ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . Ethical permission for this study was obtained from the Ethical and Scientific Review Committees of the Faculty of Tropical Medicine , Mahidol University , and of the MoPH , Thailand . Written consent was given by the directors of the hospitals to use their routine hospital database for research . Consent was not sought from the patients as this was a retrospective study , and the Ethical and Scientific Review Committees approved the process . The microbiology laboratory data collected included hospital number ( HN ) , admission number ( AN ) , specimen type , specimen date , culture result , and antibiotic susceptibility profile ( antibiogram ) . We consulted with study sites when the results of antimicrobial susceptibility testing were unclear . Hospital admission data were collected from the routine in-patient discharge report , which is regularly completed by attending physicians and reported to the MoPH , Thailand , as part of the national morbidity and mortality reporting system . The data collected included HN , AN , national identification 13-digit number , admission date , and discharge date . Date of death was also extracted from this record . Data collected from the national death registry obtained from the MoI included the national identification 13-digit number and the date of death . Bacteraemia was classified as CAB , HAB and HCAB as described previously ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . Polymicrobial infection was defined in patients who had more than one species of pathogenic organisms isolated from the blood during the same episode , and was excluded from the analysis . Information on the incidence of CAB , HCAB and HAB from all pathogenic organisms has been published previously ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . The 30-day mortality of CAB and HCAB was defined as death within 30 days of the admission date . The 30-day mortality of HAB was determined on the basis of a record of death within 30 days of the positive blood culture taken as recorded in the routine hospital database or by a record of death in the national death registry . In the event that a patient had more than one episode of bacteraemia , only the first episode was included in the study . The standard definition of MDR proposed by ECDC/CDC was used ( Magiorakos et al . , 2012 ) . In brief , MDR was defined as non-susceptibility to at least one agent in three or more antimicrobial categories . Additionally , methicillin-resistant Staphylococcus aureus ( MRSA ) were automatically described as MDR ( Magiorakos et al . , 2012 ) . Pearson’s chi-squared test and Fisher’s Exact test were used to compare categorical variables . A non-parametric test for trends was used to assess changes in prevalence of antimicrobial resistance over time stratified by hospital ( using the npt_s command in STATA ) . Mortality of patients with a first episode of HAB , HCAB and HAB caused by S . aureus , Enterococcus spp . , E . coli , K . pneumoniae , Pseudomonas aeruginosa , and Acinetobacter spp . were evaluated in relation to MDR . We selected these organisms based on guidelines for MDR proposed by ECDC/CDC , ( Magiorakos et al . , 2012 ) and the fact that E . coli and K . pneumoniae were the most common causes of bacteraemia caused by Enterobacteriaceae in our setting ( Kanoksil et al . , 2013; Hongsuwan et al . , 2014 ) . Isolates tested for less than three antimicrobial categories were excluded from the analysis because they were not applicable to ECDC/CDC standard definitions of MDR . To examine the association between MDR and mortality , we performed a multivariable logistic regression analysis adjusting for a priori selected baseline confounders . To take account of the fact that patients with CAB , HCAB , and HAB were different populations with different definitions of 30-day mortality , we applied models to each group ( CAB , HCAB and HAB ) separately . Multivariable logistic regression models were developed using a purposeful selection method ( Bursac et al . , 2008 ) . Potential confounding variables evaluated included age , gender and admission year . In the model for HAB , time to bacteraemia was also evaluated as a potential confounder because there was evidence suggesting that time to HAI was associated with mortality from HAI ( Moine et al . , 2002; Nguile-Makao et al . , 2010 ) . Time to bacteremia was defined as the duration between hospital admission and the date positive blood culture was taken . All models were stratified by hospital . The mortality attributable to MDR was calculated using adjusted odds ratios ( aORs ) estimated by the final multivariable logistic regression models . If X was the observed mortality in patients with MDR infection , the estimated odds of mortality if they were infected with non-MDR organisms ( O ) would be ( 1/aOR ) * ( X/ ( 1-X ) ) . Assuming that excess mortality was due to MDR , then the mortality attributable to MDR would be the absolute difference between mortality in patients with MDR infection ( X ) and the predicted mortality if they were infected with non-MDR organisms ( O/ ( 1+O ) ) , which would be X – ( O/ ( 1+O ) ) ( Benichou , 2001; Greenland and Robins , 1988 ) . Heterogeneity between different organisms within each group of patients ( CAB , HCAB , and HAB ) was assessed using the chi-squared test , and the percentage of variation due to heterogeneity ( I-square ) was calculated . The number of deaths attributable to MDR in Thailand was determined using the methodology described previously ( European Centre for Disease Prevention and Control and European Medicines Agency , 2009 ) . Data used included our estimated mortality attributable to MDR bacteraemia and cumulative incidence of HAI bacteraemia , lower respiratory track infection ( LRTI ) , urinary tract infection ( UTI ) , skin and soft tissue infection ( SSTI ) , and other sites of infection caused by MDR S . aureus , E . coli , K . pneumoniae , P . aeruginosa , and Acinetobacter spp . in Thailand in 2010 , which have been described previously ( Pumart et al . , 2012 ) . Death attributable to MDR Enterococcus spp . was not included as the cumulative incidence of MDR Enterococcus infection in Thailand was not available ( Pumart et al . , 2012 ) . Attributable mortality by site of infection ( LRTI , UTI , SSTI and other site ) was estimated by applying correction factors corresponding to the relative mortality from infections of those sites compared to bacteraemia ( Martone et al . , 1998 ) . All analyses were performed using STATA version 14 . 0 ( StataCorp LP , College station , Texas , USA ) . | Antimicrobial resistance is a global problem . Each year , an estimated 23 , 000 deaths in the United States and 25 , 000 deaths in the European Union are extra deaths caused by bacteria resistant to antibiotics . People in low- and middle-income countries are also using more antibiotics , in part because of rising incomes , lower costs of antibiotics , and a lack of control of antimicrobial usage in the hospitals and over-the-counter sales of the drugs . These factors are thought to be driving the development and spread of bacteria that are resistant to multiple antibiotics in countries such as China , India , Indonesia and Thailand . However , a lack of information makes it difficult to estimate the size of the problem and , then , to track how antimicrobial resistance and multi-drug resistance is changing over time in these and other low- and middle-income countries . Now , by integrating routinely collected data from a range of databases , Lim , Takahashi et al . estimate that around an extra 19 , 000 deaths are caused by multi-drug resistant bacteria in Thailand each year . Thailand has a population of about 70 million , and so , per capita , this estimate is about 3 to 5 times larger than those for the United States and European Union ( which have a populations of about 300 million and 500 million , respectively ) . Lim , Takahashi et al . also show that more of the bacteria collected from patients are resistant to multiple antimicrobial drugs and that the burden of antimicrobial resistance in Thailand is worsening over time . These findings suggest that more studies with a systematic approach need to be done in other low- and middle-income countries , especially in countries where microbiological laboratories are readily available and routinely used . Further work is also needed to identify where resources and attentions are most needed to effectively fight against antimicrobial resistance in low- and middle-income countries . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2016 | Epidemiology and burden of multidrug-resistant bacterial infection in a developing country |
A general method is described for the site-specific genetic encoding of cyanine dyes as non-canonical amino acids ( Cy-ncAAs ) into proteins . The approach relies on an improved technique for nonsense suppression with in vitro misacylated orthogonal tRNA . The data show that Cy-ncAAs ( based on Cy3 and Cy5 ) are tolerated by the eukaryotic ribosome in cell-free and whole-cell environments and can be incorporated into soluble and membrane proteins . In the context of the Xenopus laevis oocyte expression system , this technique yields ion channels with encoded Cy-ncAAs that are trafficked to the plasma membrane where they display robust function and distinct fluorescent signals as detected by TIRF microscopy . This is the first demonstration of an encoded cyanine dye as a ncAA in a eukaryotic expression system and opens the door for the analysis of proteins with single-molecule resolution in a cellular environment .
Fluorescent reporters are useful for the study of protein dynamics in a live cell , because they can inform on the conformational dynamics of a protein . A common strategy to incorporate fluorophores into target proteins is the fusion to fluorescent proteins , such as GFP or one of its spectral variants ( Miranda et al . , 2013; Zachariassen et al . , 2016 ) . This approach has the built-in convenience of encoding but their positioning within the target protein can be limited by their large size , which may impact protein function or trafficking . Alternatively , post-translational chemical labeling with compact fluorescent dyes through reactive side-chains ( Cys , Lys ) or with bio-orthogonal labeling allows for the use of more diverse fluorophores and experimental applications ( Mannuzzu et al . , 1996; Cha and Bezanilla , 1997; Priest et al . , 2015; Debets et al . , 2013 ) . Nonetheless , Cys- or Lys-labeling in live cells is limited to extracellular residues and often results in a substantial background signal due to native reactive residues of the cell . This shortcoming is further limiting when labelling side-chains in eukaryotic membrane proteins which can have many ‘background’ cysteine residues , that when mutated can lower expression profiles or have functional consequences . Bio-orthogonal labeling is advantageous in this respect ( i . e . reduced off-target labeling , increased specificity ) but requires that the residues or regions for modification are chemically accessible . Fluorescent non-canonical amino acids ( ncAA ) present a solution to these drawbacks as they are relatively small , and theoretically , can be encoded at any site within the protein in cellular and cell-free environments ( Turcatti et al . , 1996; Cohen et al . , 2002; Zhang et al . , 2004; Summerer et al . , 2006; Wang et al . , 2006; Kajihara et al . , 2006; Pantoja et al . , 2009; Kalstrup and Blunck , 2013 ) . However , of the available encodable fluorophores , most are excited by near UV light , a spectroscopic property that leads to competing cellular fluorescence , and have limited , if any , utility for single-molecule studies e . g . they have very short fluorescence lifetimes . Further , ongoing advances in protein structure determination are revealing the macromolecular structures of membrane protein complexes and in some cases have started to provide mechanistic glimpses into their function ( Eisenstein , 2016 ) . However , a structure represents a snapshot of a single conformation , yet proteins constantly cycle though various states , all of which have the potential to be functionally significant . Therefore , although high-resolution protein structures are becoming more commonplace , assigning functional correlates remains a significant challenge . This is true especially for eukaryotic membrane proteins , many of which cannot be easily produced in biochemical scales and for which expression in a live cell is a prerequisite , e . g . as is the case for many voltage-gated plasma-membrane channels . Thus , there is a growing general need for improved encoded fluorescent reporters , ideally those that are well suited to single-molecule studies and are applicable to eukaryotic expression systems . Herein a method is described for genetic encoding of organic Cy dyes as non-canonical amino acids ( Cy-ncAAs ) , in cell-free and cellular environments . We first report a synthetic approach for the production of orthogonal misacylated tRNAs that carry a Cy-ncAA . The tolerance of the eukaryotic ribosome for these fluorophores is demonstrated in a eukaryotic cell-free protein expression system with a novel luciferase rescue assay . We then show in Xenopus oocytes that these Cy-dye-based optical probes ( Cy3 , Cy5 and the self-healing Cy3 variant LD550 ) can be incorporated into membrane proteins ( providing two examples: a chloride ion channel and a voltage-gated sodium channel auxiliary subunit ) . These proteins were properly folded , functional and trafficked to the plasma membrane , albeit in reduced levels compared to wild-type channels . TIRF microscopy was then used to obtain single-molecule images from cells expressing membrane proteins containing Cy3 and Cy5 , further confirming that the encoding is successful .
A method is presented for the cellular encoding of single-molecule fluorophores as non-canonical amino acids based on widely employed organic Cy dyes for single-molecule studies ( Figure 1 ) . These encoded Cy-ncAAs were produced through a hybrid strategy based on the design elements of misacylated tRNA for cell-free expression systems ( Kajihara et al . , 2006 ) and the genetic code expansion technique of nonsense suppression in Xenopus laevis oocytes ( Leisle et al . , 2015 ) . Briefly , a free primary amine containing amino acid , such as para-amino-L-phenylalanine ( AF ) , is chemically conjugated to the dinucleotide phospho-desoxy-cytosine phospho-adenosine ( short pCA ) and subsequently reacted with a commercially available Cy dye succinimide ester under acidic conditions to ensure selective labeling of the para-amino group ( Figure 1; Figure 1—figure supplements 1–4 ) . This dinucleotide-ncAA product is HPLC purified and then enzymatically coupled in vitro to an orthogonal tRNA which is competent to suppress an in-frame nonsense codon in a cell-free translation reaction or in the context of the Xenopus oocyte , as shown in Figure 1 . The latter allows for electrophysiological determination of ion channel function at the plasma membrane by two-electrode voltage clamp ( TEVC ) recordings , and oocytes are an established experimental platform to visualize single ion channel complexes using total internal reflection fluorescence ( TIRF ) microscopy ( Ulbrich and Isacoff , 2007; Sonnleitner et al . , 2002 ) . 10 . 7554/eLife . 19088 . 003Figure 1 . Experimental flow for synthesis and incorporation of Cy fluorophores as ncAAs for single-molecule imaging of plasma membrane proteins in X . laevis oocytes . Dashed box ( top ) shows fluorophores used in this study . For details of the procedure see main text . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00310 . 7554/eLife . 19088 . 004Figure 1—figure supplement 1 . HPLC chromatogram ( A ) , UV absorbance spectrum ( B ) and LRMS ( C ) of AF-pCA . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00410 . 7554/eLife . 19088 . 005Figure 1—figure supplement 2 . HPLC chromatogram ( A ) , UV absorbance spectrum ( B ) and LRMS ( C ) of Cy3-AF-pCA . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00510 . 7554/eLife . 19088 . 006Figure 1—figure supplement 3 . HPLC chromatogram ( A ) , UV absorbance spectrum ( B ) and LRMS ( C ) of Cy5-AF-pCA . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00610 . 7554/eLife . 19088 . 007Figure 1—figure supplement 4 . HPLC chromatogram ( A ) and UV absorbance spectrum ( B ) of LD550-AF-pCA . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 007 The fluorescence of the Cy-ncAAs provided a straightforward way to assess the tRNA ligation efficiency through the measurement of relative absorbance of the nucleic acid and the cyanine chromophore of an acylated tRNA or by HPLC . Surprisingly , the standard ligation conditions ( 37°C , 40 min ) used for nonsense suppression by our lab and others ( Pantoja et al . , 2009; Pless et al . , 2013 , 2014; Nowak et al . , 1998 ) resulted in relatively poor tRNA acylation yields with Cy-ncAAs ( Figure 2A ) . We reasoned that this low ligation efficiency could , in part , be due to hydrolysis of the Cy-ncAA from the tRNA occurring during the enzymatic ligation reaction at 37°C; hydrolysis is a highly temperature sensitive process ( Stepanov and Nyborg , 2002 ) . This possibility was tested by performing the misacylation reaction at a lower temperature ( 4°C ) . The incubation time was increased ( to 2 , 5 or 11 hr ) to overcome the reduced enzymatic activity of the RNA ligase induced by the lower temperature . This approach identified conditions at 4°C for 2 hr with higher ligation efficiencies: 68 ± 10 . 2% for Cy3-ncAA and 79 ± 2 . 5% for Cy5-ncAA ( Figure 2A ) . The results for ligations at 4°C , 5 hr were not significantly different from samples obtained after 2 hr ( Cy3: p=0 . 76; Cy5: p=0 . 08 ) and the acylation efficiency decreased significantly after an incubation time of 11 hr . Consequently , for the following experiments , all tRNA acylation reactions were done at 4°C for 2 hr . These acylation yields were also confirmed by HPLC where Cy3 and Cy5 ligation reactions were similar and displayed high ( >80% ) acylated yields ( Figure 2—figure supplement 1 ) . However , the HPLC purified tRNA fraction diminished the yield , roughly 10% of the starting material , resulting in amounts that were insufficient for proper experimental validation . Hence , non-HPLC purified aminoacylated tRNA were employed for all subsequent non-sense suppression experiments . 10 . 7554/eLife . 19088 . 008Figure 2 . Optimization of tRNA acylation and genetically encoding Cy-ncAAs into soluble proteins in a eukaryotic cell-free expression system . ( A ) In vitro acylation efficiency of tRNA with Cy3- or Cy5-AF was determined by absorbance measurements for different RNA ligation conditions ( Cy3: N = 5; Cy5: N ≥ 3 ) . Highest efficiencies were found at 4°C after 2 or 5 hr . 4°C , 2 hr was chosen as optimal reaction condition for all further experiments . ( B ) In a eukaryotic in vitro translation system , Luciferase activity of NanoLuc Amber was rescued by Cy3-AF-tRNA produced 56% of the Phe suppression signal ( 826 ± 126 AU versus 466 ± 85 AU for Phe and Cy3 , respectively ) . Both of which were well above background ( bg ) signal or from the Nanoluc Amber plus empty tRNA ( no AA ) . All were quantified by bioluminescence ( AU , arbitrary units ) . The read-through control ( no AA ) was not significantly different from the background luminescence ( bg; p>0 . 1 ) . NanoLuc Amber Phe is ~11 fold above the no AA condition and Cy3-AF ~6 fold ( N = 5 , individual values with mean ± s . e . m . ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00810 . 7554/eLife . 19088 . 009Figure 2—figure supplement 1 . HPLC purification of acylated pyrrolysine tRNA ( PylT ) . ( A ) HPLC chromatogram at 260 nm showing a trace of non-acylated and Cy3-AF-acylated PylT at 36 . 8 min , with 96% acylation efficiency based on area-under-peaks . ( B ) UV absorbance spectrum of Cy3-AF-acylated PylT . ( C ) HPLC chromatogram at 260 nm showing a trace of non-acylated PylT and Cy5-AF-acylated PylT at 42 . 8 min , with 98% acylation efficiency based on area-under-peaks . ( D ) UV absorbance spectrum of Cy5-AF-acylated PylT . Utilizing a PylT ε260 value of 690 , 000 M-1cm-1 ( http://www . idtdna . com/calc/analyzer ) the ratio of tRNA to functionally ligated Cy dyes ( using Cy3 ε550 of 150 , 000 M−1cm−1 and Cy5 ε650 of 250 , 000 M−1cm−1 ) was calculated . The ratios of the HPLC UV absorbances at 260 nm and 550/650 nm were compared to the ε260 and ε550/ ε650 values and were both found to be within 30% of the theoretical 1:1 tRNA:CyX ratio . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 00910 . 7554/eLife . 19088 . 010Figure 2—figure supplement 2 . In eukaryotic cell-free translation system , the yield of nonsense suppressed NanoLuc Amber ( Phe , N = 5 , same data as in Figure 2B ) is roughly 7000 times lower than for wild-type NanoLuc ( WT , N = 3 ) . Individual values are shown with corresponding mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 010 To determine if large amino acids like the Cy-ncAAs could be tolerated by the eukaryotic ribosome , they were encoded into target proteins in a rabbit reticulocyte cell-free protein expression system ( Figure 2B ) . An amber ( TAG ) version of the high-activity NanoLuc , a small luciferase subunit from the deep-sea shrimp Oplophorus gracilirostris ( Hall et al . , 2012 ) , was generated by inserting the amber stop codon TAG between Gyl159 and Val160 and named NanoLuc Amber . A pyrrolysine tRNA from Methanosarcina mazei ( PylT ) was chosen for nonsense suppression given that it has been shown to be orthogonal in eukaryotic cells ( Hall et al . , 2012; Mukai et al . , 2008 ) ; however , to our knowledge , this is the first use in a cell-free protein synthesis system in an in vitro misacylated form . Applying the optimized tRNA ligation conditions , PylT was employed to encode Cy3-AF or Cy5-AF or the natural amino acid Phe within NanoLuc Amber . Rescue of the NanoLuc Amber full length protein via TAG suppression was quantified in a luminescence assay ( Figure 2B ) . Eukaryotic protein synthesis reactions supplemented with NanoLuc Amber cRNA in the presence of PylT lacking an appended amino acid ( no AA ) produced no measurable bioluminescence above the background signal ( p>0 . 1 ) , consistent with a truncated , nonfunctional NanoLuc Amber protein , and demonstrates orthogonality of the Pyl tRNA ( Figure 2B ) . In contrast , supplementing the in vitro translation reactions with NanoLuc Amber cRNA and PylT carrying Phe or Cy3-AF resulted in NanoLuc Amber rescue as shown by bioluminescence ( Figure 2B ) . Here , Cy3 incorporation resulted in 56% of the expression of Phe suppression ( 826 ± 126 AU versus 466 ± 85 AU for Phe and Cy3 , respectively ) . Notably , in the cell-free production , NanoLuc amber rescue was poor for Phe and Cy3 , which had yields roughly 7000 and 13 , 000-fold , respectively , lower than wild-type NanoLuc ( Figure 2—figure supplement 2 ) . These data indicate that large fluorescent amino acids based on Cy-dyes can be encoded by the eukaryotic translation machinery through nonsense suppression . We next tested the possibility that Cy-ncAAs could be encoded in the cellular environment of the Xenopus oocyte expression system given its ease for translation of cRNA transcripts and nonsense suppression with in vitro misacylated tRNA in combination with electrophysiological and single-molecule examination ( Leisle et al . , 2015; Dougherty and Van Arnam , 2014 ) . The tRNA variant most often used for amber codon suppression in oocytes is THG73 ( Tetrahymena thermophila G73 ) ( Saks et al . , 1996 ) . The chloride channel ClC-0 was chosen as a representative membrane protein as previous experiments on ClC-0 have shown that the gating glutamate ( E166 ) , a transmembrane residue in the ion permeation pathway , is widely tolerant to natural amino acid substitutions , including large-volume side chains like Phenylalanine , while showing robust expression of chloride conducting ion channels ( Zhang et al . , 2009 ) . Available crystal structures of gating glutamate mutants of EcClC ( a prokaryotic CLC homologue ) suggest that substitutions at this position cause the side chain to orient towards the extracellular space with access to aqueous solution , resulting in an open pore phenotype ( Dutzler et al . , 2003; Lobet and Dutzler , 2006 ) . This position also turned out to be amenable to nonsense codon suppression as suggested by rescued channel function of ClC-0 E166TAG by Phe-tRNA ( Figure 3A–C ) . To test that the rescued currents were mediated by ClC-0 , chloride was substituted by iodide , which has been shown to inhibit ClC-0 currents ( Pusch et al . , 1995; Ludewig et al . , 1997; Zifarelli and Pusch , 2007 ) , resulting in a strong reduction of the measured current amplitudes ( Figure 3—figure supplement 1 ) . Phenylalanine was used as a positive control for nonsense suppression of ClC-0 E166TAG because Phe is a close mimic of the core structure of the Cy-ncAA . However , we also note that under our current experimental conditions , the wild-type glutamate amino acid was not successfully coupled to pCA , barring its use for nonsense suppression . The ‘read-through’ at the introduced TAG codon in position 166 , a spurious process in genetic code expansion techniques , was minimal at 24 hr post-injection in the presence of a nonacylated ( no AA ) tRNA ( Figure 3A–C ) . However , these bleed-through currents generally increase with time , in a site and channel specific manner , and this was also the case with E166TAG ClC-0 expression past 24 hr . Further , max . 7 . 5 ng of ClC-0 E166TAG cRNA was used per oocyte as higher cRNA amounts led to spurious but increased read-through expression . Consistent with results obtained in cell-free protein synthesis , Cy3 and Cy5-ncAAs could be encoded at E166TAG , resulting in functional channels with robust expression levels , constitutively open gating and chloride selectivity , similar in each parameter to Phe incorporation ( Figure 3A–C ) . To further test the ncAA size limits of the eukaryotic ribosome , a ‘self-healing’ Cy3 variant LD550 ( Zheng et al . , 2014 ) was encoded at the E166 position in ClC-0 . The macroscopic currents elicited by the fluorophore-incorporated ClC-0 channels were all well above the background currents ( Cy3: 8 . 6-fold; Cy5: 5 . 8-fold; LD550: 4 . 8-fold; p<0 . 001 for all Figure 3B ) , thus demonstrating significant suppression efficiencies of the encoded Cy-ncAA-tRNA . The importance of optimizing the tRNA ligation reaction is highlighted when comparing the expression of channels rescued by tRNA produced at 37°C versus 4°C where the former condition yields less efficient channel rescue levels ( Figure 3B ) . Further , the amount of tRNA used was also optimized for suppression at ClC-0 E166TAG ( Figure 3—figure supplement 2 ) . Of note , due to the central location of the E166 encoding site within the ClC-0 reading frame , truncated protein transcripts that lack an encoded Cy-ncAA would be non-functional and biologically inert . Thus , the whole-cell ionic currents are produced by a substantial population ( ~750 , 000 channels per µA [Ludewig et al . , 1996] ) of functionally expressed , full-length chloride channels at the plasma membrane that contain the encoded Cy-ncAA . 10 . 7554/eLife . 19088 . 011Figure 3 . Genetic encoding of Cy-ncAAs into membrane proteins in Xenopus laevis oocytes . ( A ) to ( C ) Nonsense suppression of ClC-0 E166TAG . ( A ) The chloride channel function of ClC-0 was successfully reconstituted with the misacylated tRNAs ( Phe , Cy3 , Cy5 , LD550 ) while nonacylated tRNA ( no AA ) yielded no functional channels . Representative TEVC current recordings and the voltage clamp protocol are shown . Horizontal scale bars indicate time ( 50 ms ) , vertical scale bars the current amplitude ( 10 µA ) . Note that traces were scaled to unity for clearer presentation of current properties . ( B ) For quantification , currents elicited by a +80 mV pulse in oocytes expressing reconstituted chloride channels were normalized to currents of the background control ( no AA ) . Black bars indicate rescue by suppressor tRNA acylated at 4°C for 2 hr , white bars show results for tRNA acylated at 37°C for 40 min . For Cy3 and Cy5 , rescue of ClC-0 E166TAG has been significantly increased by using tRNAs acylated at 4°C ( Cy3: p<0 . 001; Cy5: p=0 . 01; LD550: p=0 . 11 ) . Values for the fold increase in conductance are shown as mean ± s . e . m . ; numbers of individual oocytes tested are indicated ( for 4°C results were pooled from 5–11 different batches of oocytes; for 37°C from 2–7 batches ) . ( C ) Steady-state current-voltage relationships for all conditions shown in ( A ) and recordings quantified in ( B ) for 4°C . As predicted , the reconstituted channels behave as voltage-independent , constitutively open conductances with a reversal potential around the chloride equilibrium potential , indicative of a chloride-selective channel . ( D ) and ( E ) Nonsense suppression of NaVβ1 E27TAG . ( D ) Representative TEVC recordings of oocytes coinjected with NaV1 . 4 and β1 E27TAG + Cy3-tRNA ( black ) or β1 E27TAG + nonacylated tRNA ( grey ) . Accelerated NaV1 . 4 fast-inactivation kinetics ( black ) indicate successful incorporation of Cy3 into β1 . Currents were elicited by a 50 ms test pulse to −20 mV from a holding potential of −120 mV . They were scaled to unity for illustration purposes; scale bars: horizontal indicates time ( 20 ms ) , vertical indicates current amplitude ( 1 µA ) . ( E ) For quantification of the effect on inactivation kinetics the time period till the peak current ( IPEAK ) decayed to 10% of its initial value was estimated . Results for NaV1 . 4 + β1 E27TAG + no AA ( 23 ± 2 . 1 ms ) were not significantly different from NaV1 . 4 expressed alone ( 26 ± 1 . 1 ms; p=0 . 36 ) . Both , wild-type β1 and the reconstituted β1 , significantly accelerated the inactivation kinetics of NaV1 . 4 to 5 ± 0 . 9 ms and 8 ± 1 . 0 ms , respectively , while not being significantly different from each other ( p=0 . 08 ) . Analysis includes data from two oocyte batches . Values are presented as mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 01110 . 7554/eLife . 19088 . 012Figure 3—figure supplement 1 . Iodide blocks ClC-0 E166Phe currents , indicating that they are indeed ClC-0 mediated . ( A ) Oocytes injected with ClC-0 E166TAG cRNA and Phe-tRNA ( left ) or ClC-0 E166TAG cRNA and non-acylated tRNA ( center ) or with water ( H2O , right ) were perfused with the following sequence of ions: Chloride ( blue ) , Iodide ( red ) , Chloride ( green ) . Representative current traces for each condition and the corresponding current-voltage ( I/V ) relationships are shown ( vertical scale bar: current , 5 µA; horizontal scale bar: time , 50 ms ) . In Cl--containing saline reversal potentials were estimated to be ~−30 mV for Phe-tRNA ( left ) , ~−40 mV for tRNA and ~−65 mV for H2O ( right ) injected oocytes . ( B ) Averaged currents evoked by a +80 mV pulse in Chloride and Iodide , respectively , are shown as mean ± s . e . m . for each condition ( n = 7 for Phe-tRNA , n = 5 for tRNA , n = 5 for H2O; from 2–3 different batches of oocytes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 01210 . 7554/eLife . 19088 . 013Figure 3—figure supplement 2 . Nonsense suppression efficiency at position E166 in ClC-0 . ( A ) Optimization of tRNA amounts for nonsense suppression of ClC-0 E166TAG . Oocytes were injected with 7 . 5 ng of ClC-0 E166TAG cRNA and decreasing Val-tRNA amounts ( 200/100/50/25 ng per oocyte ) . As negative control non-acylated tRNA ( no AA ) was coinjected ( 200 ng per oocyte ) . Values represent current amplitudes as mean ± s . e . m . ( N = 4 for 1; N = 4 for 1/2; N = 3 for 1/4; N = 3 for 1/8; N = 5 for no AA; from 1 batch of oocytes ) . ( B ) Injecting a 100-fold dilution of wild-type ( WT ) ClC-0 cRNA ( 0 . 075 ng per oocyte ) produced currents at +80 mV similar in amplitude to ClC-0 E166TAG rescued currents with Phe , where 7 . 5 ng of cRNA was injected per oocyte . Values are shown as mean ± s . e . m . ( N = 30 for Phe , N = 5 for WT from at least three independent batches ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 013 To further validate the approach , Cy3-AF was encoded in the eukaryotic sodium channel ( NaV ) complex . NaVs contribute to the upstroke of the action potential in the excitable tissues of nerve and muscle ( Ahern et al . , 2016 ) , are widely implicated in inherited and acquired human diseases and are high value therapeutic targets ( Ahuja et al . , 2015 ) . The sodium ion selective pore forming α-subunit displays modulated expression and gating by the single pass transmembrane β subunit ( O'Malley and Isom , 2015 ) . Specifically , in the context of the Xenopus oocyte , co-expression of NaV1 . 4 and β1 produces currents with accelerated inactivation compared to NaV1 . 4 expressed alone ( Makita et al . , 1996 ) . The data show that encoding Cy3 at position E27 in the extracellular domain of β1 resulted in acceleration of NaV1 . 4 inactivation kinetics comparable to wild-type β1 with minimal read-through , consistent with the functional expression of the full-length auxiliary β1 subunit ( Figure 3D and E ) . Thus , in tandem , the electrophysiological data demonstrate the successful site-specific incorporation of Cy-ncAAs into two different membrane proteins – ClC-0 and NaV β1 subunit . To confirm plasma membrane expression of ClC-0 channels containing encoded Cy-ncAAs , oocytes were examined using a single fluorescent molecule TIRF microscope . Under TIRF illumination multiple fluorescent spots could be detected in the region of the oocyte plasma membrane in contact with the coverslip glass ( Figure 4A ) . Oocytes were coinjected with ClC-0 E166TAG cRNA + Cy3-AF-tRNA + Cy5-AF-tRNA or for the mock condition with ClC-0 wild-type cRNA + Cy3-AF-tRNA + Cy5-AF-tRNA to estimate non-specific background fluorescence . The analysis yielded three significant observations that demonstrate the functional encoding of the Cy-ncAAs into ClC-0 . First , overall Cy3 and Cy5 spot densities were higher for the encoded ( Cy-ncAA-tRNA plus ClC-0 E166TAG cRNA ) versus ‘mock’ conditions ( Cy-ncAA-tRNA plus ClC-0 wild-type cRNA; Figure 4A and B ) . Specifically , the number density for the encoded Cy3 was 0 . 14 ± 0 . 02 spots per µm2 and 0 . 08 ± 0 . 01 for the encoded Cy5 , while in mock 0 . 03 ± 0 . 01 spots per µm2 were detected for Cy3 ( p=0 . 001 ) and 0 . 02 ± 0 . 01 for Cy5 ( p=0 . 004; Figure 4B ) . The fluorescent spots observed in the mock condition may be ascribed to non-encoded Cy-ncAA-tRNA or hydrolyzed Cy-ncAA that may have escaped the cytosol during peeling the vitelline membrane ( a prerequisite to TIRF imaging ) or that are trapped within the small cytosolic volume of the oocyte microvilli ( Sonnleitner et al . , 2002 ) . The likelihood that those spots contain CyX-ncAAs encoded within the truncation codons of endogenous plasma membrane proteins that use TAG as a native stop codon is very low because their expression level well below of an overexpressed protein . Further , analysis of EST databases for Xenopus laevis oocytes stage V and VI shows that the list of possible candidates is limited ( Supplementary file 1 ) . However , most importantly , read-through of a stop codon of an endogenous protein would likely lead to translation of the 3’ UTR and ubiquitin-mediated protein degradation ( Bengtson and Joazeiro , 2010 ) . The second observation supporting specific encoding of CyX-ncAAs into ClC-0 is the significant colocalization between Cy3 and Cy5 fluorescent spots in oocytes co-injected with Cy3 , Cy5 and ClC-0 E166TAG cRNA . Of all imaged spots , 14 ± 2 . 2% of those with Cy5 fluorescence co-localized with Cy3 spots in the encoded reaction , compared to only 3 ± 1 . 0% for randomly co-localized Cy3/Cy5 spots , an outcome similar to mock conditions with 3 ± 0 . 8% ( Figure 4A and C ) . Thus , in dual Cy3- and Cy5-ncAA encoding conditions , multiple fluorophores appear within single spots suggesting either the encoding of Cy3- and Cy5-ncAA in single or multiple ClC-0 dimers within a distinct fluorescent spot . Lastly , a photo-bleaching analysis of fluorescent spots in encoded and mock condition was performed ( Figure 5 , Figure 5—figure supplement 1 ) . Multiple examples of step-wise bleaching events of the encoded condition are presented in Figure 5A and B . The data reveal that while the mock bleaching profile was strongly dominated by single step events , the encoded reaction showed a clear shift towards multiple steps ( 2+ ) , as shown as absolute values in a histogram ( Figure 5C ) and in relative values ( Figure 5—figure supplement 1 ) . Further , under encoding conditions some spots showed upwards of 4- and 5-step bleaching events , an observation that was unmatched in the mock condition . Thus , the photo-bleaching data further demonstrate encoding of the Cy-ncAA , likely within single and multiple ClC-0 dimers , in a manner that is consistent with expression beyond the single-molecule level or clustering of rescued full-length channels . Indeed , clustering of ClC-0 wild-type channel has been observed previously in excised patch experiments ( Bauer et al . , 1991 ) . 10 . 7554/eLife . 19088 . 014Figure 4 . Single-molecule TIRF microscopy imaging of cellular encoded Cy3 and Cy5 fluorophores in ClC-0 , a dimeric plasma membrane chloride channel . ( A ) TIRF images show distinct fluorescent spots for Cy3 and Cy5 encoded into ClC-0 at position E166 ( Encoded ( top ) , Mock ( bottom ) ) . A subpopulation of Cy3 and Cy5 spots colocalized ( highlighted in white boxes ) . The encoded and mock injected images were grayscaled to identical levels to enable viewing of both bright and dim spots . ( B ) Number density of fluorescent spots per area ( µm2 ) of plasma membrane was significantly different between Encoded ( black ) and Mock ( white ) . p values see main text . Total numbers of counted spots included in this analysis are: N = 4406 for encoded Cy3 , N = 2459 for encoded Cy5 , N = 1154 for mock Cy3 , N = 717 for mock Cy5; they result from ≥5 oocytes ( ≥3 batches ) per condition ( C ) Same images as in ( B ) were examined for colocalization between Cy3 and Cy5 . In Encoded , 14 ± 2% of the Cy5 signal merged with Cy3 . Random colocalization in those images was estimated to 3 ± 1% and colocalization in Mock was not significantly different from that ( p=0 . 96 ) . All values are mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 01410 . 7554/eLife . 19088 . 015Figure 5 . Photo-bleaching profiles for Cy3 and Cy5 spots in Encoded and Mock conditions . ( A ) and ( B ) Photo-bleaching events of Cy3 and Cy5 fluorescent spots in the Encoded sample . ( A ) Example of a 5-step photo-bleaching event of a fluorescent spot showing Cy3 ( top ) or Cy5 fluorescence ( bottom ) , respectively , in an oocyte co-injected with ClC-0 E166TAG cRNA as well as Cy3- and Cy5-tRNA . A single representative frame was selected from the video sequence showing step-wise reduction in spot-intensity . Presented area size is 20 pixels x 20 pixels . ( B ) Examples of fluorescence traces ( in arbitrary units , AU ) as a function of time show step-wise photo-bleaching for Cy3 ( left ) and Cy5 ( right ) , respectively , until complete photo-destruction of the fluorophore was achieved . Arrows indicate observed steps during the bleaching process; N is equal to the total number of identified steps . ( C ) Photo-bleaching histogram for Cy3 and Cy5 spots in Encoded and Mock reveal a distribution distribution of steps: while in Mock 1-step events strongly dominated , in Encoded the distribution was clearly shifted towards 2+ steps . Values represent absolute numbers for the identified bleaching step event analyzed from an equal number of frames for each condition ( five oocytes , three batches ) . In total , following numbers were included: N = 1564 for encoded Cy3 , N = 762 for encoded Cy5 , N = 296 for mock Cy3 , N = 300 for mock Cy5 . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 01510 . 7554/eLife . 19088 . 016Figure 5—figure supplement 1 . Photo-bleaching profiles for Cy3 and Cy5 spots in Encoded and Mock . For clarity , results from Figure 5C are shown here as relative photo-bleaching event distributions . In each case , the Encoded conditions produce a statistically significant difference ( p<0 . 0001 ) over Mock distributions . All values are mean ± s . e . m . In total , following numbers were included: N = 1564 for encoded Cy3 , N = 762 for encoded Cy5 , N = 296 for mock Cy3 , N = 300 for mock Cy5 . DOI: http://dx . doi . org/10 . 7554/eLife . 19088 . 016 None the less , the photo-bleaching data , while strongly supportive of the encoding of the Cy-ncAA , also indicate that there remain a number of issues that must be overcome before using this method generally for stoichiometry measurements . First , the basis for the observation of spots that bleach with >2 steps merits further investigation . This could arise , as noted , from over-expression of ClC-0 in the membrane , a possibility that could be tested by titrating down to single-molecule levels by reducing the amount of tRNA injected or reducing expression time . Another possibility is that there are truly higher-order assemblies of ClC-0 in the membrane , as has been observed before and described as clustering ( Bauer et al . , 1991 ) . In this case , ClC-0 may not be a good system for studying stoichiometry . However , ClC-0 was chosen as a model membrane protein with proven expression in the oocyte , with the added benefit that its permeation pathway has been well studied with the identification of residues such as E166 that are known to be tolerant of a variety of side-chains . A second issue is that the labeling yield is too low to allow for reliable stoichiometry analysis . The data from HPLC analysis of the tRNA suggest that this decrease in fluorescence yield occurs after injection into the oocyte , but it could be occurring after the dye is in the extracellular oxygen rich environment of the solution or during preparation for imaging . During incorporation , the Cy-ncAA may be exposed to reductive conditions in the cell that lead to formation of dark state adducts ( Dempsey et al . , 2009 ) . Furthermore , imaging was performed in the absence of typical oxygen scavenger systems , as this has been shown to lead to improved photo-bleaching traces ( Chadda et al . , 2016 ) but could reduce the overall fluorescent yield . In this regard , future encoding experiments performed in the presence of oxygen quenchers or more stable fluorophores ( Zheng et al . , 2014 ) may be required for specific applications of the approach . Overall , in its current iteration , the method will require more development in order to be useful for stoichiometry measurements of protein complexes mainly because of uncertainty in labelling efficiency . None the less , once these issues have been overcome , the use of such probes will be a significant advance for stoichiometry studies over current approaches that rely of chemical labeling of proteins with Cy-dyes or encoded probes , such as GFP . However , examining dynamic single-molecule FRET is possible , in principle , provided only spots containing both Cy3 and Cy5 fluorescence signals are examined . Yet , due to low degree of colocalization this may be impractical depending on the target protein . The method could also be applied to extract time-regimen information from single-molecule experiments . In fact , estimating lateral diffusion in cellular plasma membrane by single molecule tracking necessitates sparse labelling . In the past , these studies have led to discovery of dynamic nano lipid-raft formation ( Suzuki et al . , 2012 ) or elucidation of organization of focal adhesions ( Kusumi et al . , 2014 ) . It is also useful for other single-molecule studies such as ligand binding ( e . g . kinetics of toxin binding , or other high-affinity ligands ) or a single-molecule version of voltage fluorometry . Thus , while additional effort is needed to maximize the experimental possibilities for encoding Cy-ncAA , there are a number of applications at this stage . Further , we anticipate that ultimately it will be applicable to mammalian cell lines once approaches for tRNA delivery have been established . The incorporation of noncanonical amino acids via nonsense suppression with in vitro acylated tRNAs results in significantly reduced yields compared to wild-type expression ( Leisle et al . , 2015; Dougherty and Van Arnam , 2014 ) . However , the eventual goal of encoding Cy-ncAA is for single-molecule studies , thus , decreased yield is not necessarily a barrier to the application of this approach . In the case of ClC-0 , for example , the injection of 100-fold less of the wild-type cRNA was needed to match the lower current amplitudes observed in the ‘rescued’ ClC-0 channels ( Figure 3—figure supplement 2 ) . There are multiple factors that impact the overall efficiency of suppression . For one , competition between the supplied suppressor tRNA with the endogenous translation terminator , release factor 1 ( RF-1 ) , for TAG codons . Second , the intrinsic suppression competency of in vitro transcribed and folded tRNA compared to that of native tRNA is not known . Consequently , the in vitro synthesized , orthogonal tRNA might display a lower translation efficiency compared to native tRNAs . Third , amino acid hydrolysis from the tRNA is pH and temperature dependent and might occur during incubation of the injected cells prior to forming a complex with the endogenous elongation factors ( Peacock et al . , 2014; Stepanov and Nyborg , 2002 ) . Taken together , the data demonstrate that Cy-ncAA-based optical probes were genetically incorporated into membrane proteins ( providing two examples: a chloride ion channel and a voltage-gated sodium channel auxiliary subunit ) , and that these proteins were functional and trafficked to the cell surface . TIRF microscopy was used to obtain single-molecule imaging and photo-bleaching data from cells expressing membrane proteins containing Cy3 and Cy5 , further confirming that encoding was successful . Thus , the study provides the first description of the genetic encoding of organic cyanine dyes by a eukaryotic ribosome in a cell-free translation system as well as in live cells . The observation that Cy-ncAAs can be encoded into a nascent transcript by the eukaryotic ribosome reveals that the ribosome is surprisingly plastic , and thus opens the door to study proteins by single-molecule approaches in cellular environments . The current fluorescent yields make studies such as ligand binding , or voltage-dependent fluorescence experiments possible . Once fluorescent yields are improved , the approach has applications for the future study of multisubunit stoichiometry through photo-bleaching of encoded Cy-ncAAs , or for obtaining real-time insights into conformational dynamics by single-molecule FRET studies in live cells .
The ncAAs Cy3- , Cy5- and LD550-para-amino-L-phenylalanine were synthesized conjugated to the dinucleotide phospho-desoxy-cytosine phospho-adenosine ( pdCpA or short pCA ) . For nonsense suppression during cell-free translation , pyrrolysine tRNA ( PylT ) was used . A modified ( G73 ) version of Tetrahymena thermophila tRNA , THG73 was employed for nonsense suppression in Xenopus leavis oocytes . To yield a full-length ( acylated ) tRNA , first a construct lacking the last two nucleotides was transcribed in vitro using CellScript T7-Scribe Standard RNA IVT Kit ( CELLSCRIPT , Madison , WI ) . The DNA templates ( coding for the tRNA preceded by a T7 promoter ) were ordered as PAGE-purified primers from Integrated DNA Technologies ( Coralville , IA ) . THG73 forward: ATTCGTAATACGACTCACTATAGGTTCTATAGTATAGCGGTTAGTACTGGGGACTCTAAATCCCTTGACCTGGGTTCGAATCCCAGTAGGACCGC; THG73 reverse: GCGGTCCTACTGGGATTCGAACCCAGGTCAAGGGATTTAGAGTCCCCAGTACTAACCGCTATACTATAGAACCTATAGTGAGTCGTATTACGAAT; PylT forward: ATTCGTAATACGACTCACTATAGGAAACCTGATCATGTAGATCGAACGGACTCTAAATCCGTTCAGCCGGGTTAGATTCCCGGGGTTTCCGC; PylT reverse: GCGGAAACCCCGGGAATCTAACCCGGCTGAACGGATTTAGAGTCCGTTCGATCTACATGATCAGGTTTCCTATAGTGAGTCGTATTACGAAT . 20 µg of annealed oligonucleotides were used as a template . The total reaction volume was adjusted to 100 µl and the kit reagents were added in the following amounts: 10 µl of 10X T7-Scribe transcription buffer , 7 . 5 µl of each nucleotide ( 100 mM stocks ) , 10 µl of 100 mM Dithiothreitol , 2 . 5 µl ScriptGuard RNase Inhibitor , 10 µl T7-Scribe enzyme solution . After the reaction was incubated for 4–5 hr at 37°C , the DNA template was digested with 5 µl DNase ( 1 U/µl ) provided with the kit for 30–60 min . The tRNA was extracted from the reaction with acidic phenol chlorophorm ( 5:1 , pH 4 . 5 ) and precipitated with ethanol . The precipitates tRNA was pelleted , washed , dried and resuspended in 100 µl DEPC-treated water and further purified with Chroma Spin-30 columns ( Clontech , Mountainview , CA ) . The procedure yielded roughly 100 µl of ~5 µg/µl tRNA that was stored in aliquots at −80°C . Prior to the ligation reaction of the 73-mer THG73 to the ncAA-pCA conjugate ( or just the pCA to yield a complete , nonacylated tRNA ) , the tRNA was folded in 10 mM HEPES ( pH 7 . 4 ) by heating at 94°C for 3 min and subsequent gradual cool-down to ~10°C . 25 µg of folded tRNA in 30 µl of 10 mM HEPES ( pH 7 . 4 ) were mixed with 28 µl of DEPC-treated water , 8 µl of 3 mM ncAA-pCA ( or pCA ) dimethylsulphoxide stock , 8 µl of 10X T4 RNA Ligase one buffer , 1 µl 10 mM ATP and 5 µl of T4 RNA Ligase 1 ( New England Biolabs , Ipswich , MA ) then incubated at 4°C for 2 hr ( for experiments to optimize the acylation efficiency the following conditions were also applied: 37°C , 40 min; 4°C , 5 hr; 4°C , 11 hr ) . The ( misacylated ) tRNA was extracted from the samples with acidic phenol chlorophorm ( 5:1 , pH 4 . 5 ) and precipitated with ethanol . The precipitated tRNA pellets were washed , dried in a Speedvac and stored at −80°C . The ligation efficiencies of the CyX-tRNAs were determined by absorbance measurements at 260 nm ( RNA ) and 550 nm ( Cy3/LD550 ) and 650 nm ( Cy5 ) , respectively , in serial dilutions and duplicates for each resuspended tRNA pellet . The calculations were performed according to the Lambert-Beer law using the following molar extinction coefficients: 696 , 100 M−1 cm−1 for THG73 tRNA ( http://www . idtdna . com/calc/analyzer ) , 150 , 000 M−1 cm−1 for Cy3 and 250 , 000 M−1 cm−1 for Cy5; and the following equation: ligation efficiency ( % ) = [ACyX*εRNA/ ( ARNA*εCyX ) ]*100% . In total 3–5 acylated tRNA pellets were used independently per condition for each fluorophore . In the attempt to purify acylated from non-acylated tRNA after the ligation reaction the following HPLC purification protocol was employed . CyX-acylated tRNA pellets were dissolved in 100 µL of buffer A ( 20 mM ammonium acetate , pH 5 . 0 containing 10 mM magnesium acetate and 400 mM sodium chloride ) , loaded onto a POROS R2/10 column ( 2 . 1 mm x 100 mm , Applied Biosystems , Carlsbad , CA ) and separated utilizing a series of linear gradients with increasing amounts of buffer B ( 70% buffer A , 30% ethanol ) . A Waters 2998 photodiode array detector enabled detection and isolation of peaks corresponding to both free and acylated tRNA ( Figure 2—figure supplement 1 ) . The efficiency of the CyX-tRNA ligation was estimated based on the area under the HPLC peaks at 260 nm . Due to extremely high loss of material during recovery from purification , the HPLC-purified tRNA was not utilized in experiments presented , thus the data serves to confirm the determination of the ligation efficiencies by absorbance measurements ( see above ) . All cell-free translation reactions were performed using the nuclease-treated Rabbit Reticulocyte Lysate System ( Promega , Madison , WI ) . The cRNAs for NanoLuc Amber ( NanoLuc luciferase with an inserted amber stop codon between G159 and V160 ) was generated from a linearized pcDNA3 . 1 plasmid containing the gene of interest using the mMessage mMachine T7 Ultra Kit ( Thermo Fisher Scientific , Grand Island , NY ) . Subsequently , the cRNA was purified with RNeasy Mini Kit ( Quiagen , Hilden , Germany ) , the concentration was determined by absorbance measurements at 260 nm and the quality was confirmed on an RNase-free 1% agarose gel . The tRNA transcription and misacylation is described above . For nonsense suppression of the luciferase construct PylT tRNA was used . The cell-free translation reactions were assembled according to the user’s manual: 35 µl lysate , 0 . 5 µl amino acid mixture minus Leu ( 1 mM ) , 0 . 5 µl amino acid mixture minus Met ( 1 mM ) , 1 µl RNasin Ribonuclease inhibitor ( 40 U/µl ) , 1 µl cRNA ( 2 µg/µl ) , 1 . 5 µl nonacylated or acylated tRNA ( 10 µgl/µl ) , 10 . 5 µl DEPC-treated water . The reaction took place at 22°C for 90 min . To quantify the luciferase activity the reactions were diluted 1:10 in PBS ( pH 7 . 4 ) and mixed 1:1 with the reagent from Nano-Glo Luciferase Assay System ( Promega , Madison , WI ) . The measurements were conducted in triplicates on the plate reader Spectramax i3 ( Molecular Devices , Sunnyvale , CA ) at room temperature and 1 s integration time . All experiments were performed three times . The technique was applied as previously described ( Pless and Ahern , 2013; Dougherty and Van Arnam , 2014; Leisle et al . , 2015 ) with modifications . The tRNA transcription and misacylation is described above . Regarding the in vitro cRNA transcription , the cRNA for ClC-0 and ClC-0 E166TAG was transcribed from a pTLN vector using the mMessage mMachine SP6 Kit ( Thermo Fisher Scientific , Grand Island , NY ) . Rat NaVβ1 and β1 E27TAG were transcribed from a pcDNA3 . 1 plasmid using the mMessage mMachine T7 Kit ( Thermo Fisher Scientific , Grand Island , NY ) and NaV1 . 4 from a pBSTA vector with the mMessage mMachine T7 Ultra Kit ( Thermo Fisher Scientific , Grand Island , NY ) . All reactions were set up according to the user manuals . Purification of the cRNA from the transcription reaction was conducted on columns from the RNeasy Mini Kit ( Quiagen , Hilden , Germany ) . Concentration was determined by absorbance measurements at 260 nm and quality was confirmed on a 1% agarose gel ( RNase-free ) . TEVC was performed as described before extensively ( Pless et al . , 2011; Pusch et al . , 1995 ) . In brief , voltage-clamped chloride currents were recorded in ND96 solution ( in mM: 96 NaCl , 2 KCl , 1 . 8 CaCl2 , 1 MgCl2 , 5 HEPES , pH 7 . 5 ) and sodium currents in standard Ringer ( in mM: 116 NaCl , 2 KCl , 1 MgCl2 , 0 . 5 CaCl2 , 5 HEPES , pH 7 . 4 ) using an OC-725C voltage clamp amplifier ( Warner Instruments , Hamden , CT ) . For Iodide exchange experiments of ClC-0 currents , NaCl was substituted by NaI equimolar . Glass microelectrodes backfilled with 3 M KCl had resistances of 0 . 5–3 MΩ . Data were filtered at 1 kHz and digitized at 10 kHz using a Digidata 1322A ( Molecular Devices , Sunnyvale , CA ) controlled by the pClamp 9 . 2 software ( Molecular Devices , Sunnyvale , CA ) . Chloride currents were elicited by +20 mV voltage steps from −120 mV to +80 mV preceded by a +80 mV pre-pulse from a holding potential of −30 mV . Sodium currents were elicited by a 50 ms test pulse to −20 mV from a holding potential of −120 mV . Clampfit 9 . 2 software was used for current analysis . Numbers of oocytes and batches used , details on current analysis and statistical significance of effects are indicated in the appropriate figure legends or the main text . All values are presented as mean ± s . e . m . To determine statistical significance Student’s t-test ( two-tailed distribution; two-sample equal variance ) was performed . The threshold for significance was p=0 . 05 . To classify the proteins that are expressed in Xenopus laevis ooctyes , we started with an annotated , non-normalized cDNA library as expressed sequence tags ( ESTs ) for X . laevis oocytes in stages V and VI , constructed by Blumberg et . al ( Blumberg et al . , 1991; Hawley et al . , 1995 ) . Each of the ESTs were then queried against Xenopus laevis’s Unigene database that contains 31306 sequence clusters ( Build #94 ) ( Pontius et al . , 2003; Boguski and Schuler , 1995; Schuler , 1997 ) . We identified 1305 unique mRNA sequences that were expressed in stage V/VI X . laevis oocytes . The open reading frame ( ORF ) was identified and transcribed from these mRNA sequences and the protein sequence was used to predict the number of transmembrane ( TM ) helices using TMHMM 2 . 0 ( Krogh et al . , 2001 ) . TMHMM algorithm is currently the best performing algorithm and it correctly predicts 97–98% of the transmembrane helices . We would like to point out that that our goal was not to correctly predict transmembrane helices but to classify the protein as soluble or membrane proteins . We classified all proteins with more than one predicted transmembrane segment as a membrane protein . With this method we found 43 membrane proteins whose mRNA sequence showed a TAG stop codon . We call attention to the possibility that of 18 protein sequences with a single predicted transmembrane helix , the N-terminally situated TM segment could be a signal peptide that may or may not get cleaved post-translationally . Hence , our bioinformatics analysis might overestimate the number of membrane proteins expressed in Xenopus laevis oocytes by a small amount . | Many scientists would argue that the leading edge of biological exploration is playing out at the level of individual molecules . On this scale , the essential molecular players of life are so small that they simply cannot be seen with a normal light microscope . While technology that can capture static snapshots of individual proteins frozen in time continues to advance , the choice of tools to observe individual proteins in action remains limited . Moreover , each of the existing tools for studying protein dynamics in living cells also has its own caveats . These issues led Leisle et al . to set out to develop a new method that would allow researchers to study individual proteins in live cells . This goal required a probe that was relatively small , bright , stable and compatible with biological samples . Fluorescent probes called “Cy dyes” meet all these criteria . Leisle et al . turned these probes into amino acids , the building blocks of proteins , and then supplied them to cells that were genetically programmed to incorporate the probes into a protein of interest as it was being built . This new technique allows the protein to be marked at specific sites and stops other proteins from being labeled by mistake ( a common problem with other protein labeling methods ) . This new approach was confirmed to work , firstly , in a cell extract and , secondly , in an intact cell with two unrelated proteins found in the cell membrane . The cells used were frog eggs , a type of cell that is widely used in biological experiments . Yet this approach should be easy to apply to any protein and , in theory , to any cell type after it has been optimized . The next challenges include finding ways to get the probe incorporated more efficiently into the protein of interest and to protect the probes from losing their brightness – a phenomenon called quenching . Finally , studies of single molecules provide the deepest possible insight into how a protein in a living cell carries out its activities . Better tools for single-molecule studies will lead to a more complete understanding of the dynamic life of proteins in action . Moreover , in the case for those proteins that are related to diseases , these kinds of studies may in future guide the development of new or improved drugs to treat disease . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources"
] | 2016 | Cellular encoding of Cy dyes for single-molecule imaging |
The ability to read a page of text or recognize a person's face depends on category-selective visual regions in ventral temporal cortex ( VTC ) . To understand how these regions mediate word and face recognition , it is necessary to characterize how stimuli are represented and how this representation is used in the execution of a cognitive task . Here , we show that the response of a category-selective region in VTC can be computed as the degree to which the low-level properties of the stimulus match a category template . Moreover , we show that during execution of a task , the bottom-up representation is scaled by the intraparietal sulcus ( IPS ) , and that the level of IPS engagement reflects the cognitive demands of the task . These results provide an account of neural processing in VTC in the form of a model that addresses both bottom-up and top-down effects and quantitatively predicts VTC responses .
How does visual cortex work ? One approach to answering this question consists in building functional models that characterize the computations that are implemented by neurons and their circuitry ( Hubel and Wiesel , 1963; Heeger et al . , 1996 ) . This approach has been fruitful for the front end of the visual system , where relatively simple image computations have been shown to characterize the spiking activity of neurons in the retina , thalamus , and V1 ( Carandini et al . , 2005; Wu et al . , 2006 ) . Based on this pioneering work in electrophysiology , researchers have extended the modeling approach to characterize responses in human visual cortex , as measured by functional magnetic resonance imaging ( fMRI ) ( Wandell , 1999; Dumoulin and Wandell , 2008; Kay et al . , 2008 ) . Models of early visual processing have been able to offer accurate explanations of low-level perceptual functions such as contrast detection ( Ress et al . , 2000; Ress and Heeger , 2003 ) and orientation discrimination ( Bejjanki et al . , 2011 ) . However , these models are insufficient to explain high-level perceptual functions such as the ability to read a page of text or recognize a face . These abilities are believed to depend on category-selective regions in ventral temporal cortex ( VTC ) , but the computations that give rise to category-selective responses are poorly understood . The goal of the present study is to develop a model that predicts fMRI responses in high-level visual cortex of human observers while they perform different cognitive tasks on a wide range of images . We seek a model that is fully computable—that is , a model that can operate on any arbitrary visual image and quantitatively predict BOLD responses and behavior ( Kay et al . , 2008; Huth et al . , 2012; Khaligh-Razavi and Kriegeskorte , 2014; Yamins et al . , 2014 ) . Achieving this goal requires four innovations: First , we need to develop a forward model that characterizes the relationship between visual inputs and the BOLD response in word- and face-selective cortex . Second , we need to dissociate bottom-up stimulus-driven effects from modulation by top-down cognitive processes and characterize how these processes alter the stimulus representation . Third , we need to localize the source of the top-down effects and integrate bottom-up and top-down computations into a single consolidated model . Finally , the neural computations should be linked to the measured behavior of the visual observer . In this study , we make progress on these four innovations and develop a model that characterizes bottom-up and top-down computations in word- and face-selective cortex .
Ventral temporal cortex ( VTC ) is divided into a mosaic of high-level visual regions that respond selectively to specific image categories , and are believed to play an essential role in object perception ( Kanwisher , 2010; Dehaene and Cohen , 2011 ) . We focus on two specific VTC regions , the visual word form area ( VWFA ) , which selectively responds to words ( Cohen et al . , 2000 , 2002; Wandell et al . , 2012 ) , and the fusiform face area ( FFA ) , which selectively responds to faces ( Kanwisher et al . , 1997; Grill-Spector and Weiner , 2014 ) . We measured blood oxygenation level dependent ( BOLD ) responses to a set of carefully controlled images while manipulating the cognitive task that the subjects performed on the stimuli . The first task was designed to minimize the influence of cognitive processes on sensory processing of the stimulus . Subjects performed a demanding perceptual task on a small dot ( 0 . 12° × 0 . 12° ) presented at fixation . In this fixation task , the presented stimuli are irrelevant to the subject , and we interpret evoked activity as reflecting primarily the intrinsic , bottom-up response from VTC . We acknowledge that the fixation task may not perfectly isolate bottom-up responses . For example , high-contrast stimuli may automatically attract attention . Moreover , there are other potential interpretations of the fixation task: for example , allocating attention to the small fixation dot might engage active suppression of responses to the presented stimuli . To a first approximation , much of the variance in the bottom-up fixation responses from VWFA and FFA is explained by the category of stimulus ( Figure 1d , red lines ) . However , we find that responses are not invariant to low-level properties of the stimulus: both image contrast and phase coherence modulate response amplitudes . For example , the response to a word in VWFA is 2 . 4 times stronger when the word is presented at 100% contrast as compared to 3% contrast . These bottom-up effects ( see also Rainer et al . , 2001; Avidan et al . , 2002; Yue et al . , 2011; Nasr et al . , 2014 ) may be somewhat surprising given that theories of word recognition generally posit that the VWFA response is invariant to low-level features ( Dehaene and Cohen , 2007 , 2011; Price and Devlin , 2011 ) . In fact , it is currently debated whether the VWFA should be considered a visual area or a ‘meta modal’ language region ( Reich et al . , 2011; Striem-Amit et al . , 2012 ) . Our measurements indicate that when top-down signals are minimized , word- and face-selective cortex is sensitive to low-level image properties , and that an accurate model of the computations performed by these regions must consider not only the stimulus category but also low-level features of the stimulus . 10 . 7554/eLife . 22341 . 003Figure 1 . VTC responses depend on both stimulus properties and cognitive task . ( a ) Stimuli . Stimuli included faces , words , and noise patterns presented at different contrasts and phase-coherence levels , as well as full-contrast polygons , checkerboards , and houses . ( b ) Trial design . Each trial consisted of four images drawn from the same stimulus type . ( c ) Tasks . On a given trial , subjects performed one of three tasks . ( d ) Evoked responses in VWFA ( top ) and FFA ( bottom ) for different stimuli and tasks . Color of x-axis label indicates the perceived stimulus category as reported by the subjects . Error bars indicate bootstrapped 68% CIs . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 00310 . 7554/eLife . 22341 . 004Figure 1—figure supplement 1 . Comprehensive summary of fMRI measurements . Black bars indicate responses ( beta weights ) evoked by different stimuli and tasks . Red lines indicate the average response across stimuli , computed separately for each task . Error bars indicate bootstrapped 68% CIs ( resampling subjects with replacement ) . Percentages in ROI labels indicate the strength of the response observed during the categorization and one-back tasks relative to the fixation task . For example , in FFA , the average response across stimuli during the one-back task is 79% stronger than the average response across stimuli during the fixation task . Task effects are substantially stronger in VWFA and FFA than in early visual areas V1–V3 . The larger apparent task modulation in V1 compared to V2 and V3 might due to small eye movements that may have been made during the categorization and one-back tasks . Our interpretation of the observed IPS activity during the fixation task is that this activity reflects the decision-making process involved in judging the color of the fixation dot . Support for this interpretation comes from the fact that the root-mean-square contrast of the stimuli , computed over a small region surrounding the fixation dot ( 0 . 36° × 0 . 36° ) , correlates strongly with IPS responses during the fixation task ( r = 0 . 86 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 004 We also measured VTC responses while subjects performed a categorization task , in which the subject reports the perceived category of the stimulus , and a one-back task , in which the subject detects consecutive repetitions of stimulus frames . Despite the presentation of identical stimuli across the three tasks , there are substantial changes in evoked VTC responses . Responses are larger for the categorization ( Figure 1d , green lines ) and one-back tasks ( Figure 1d , blue lines ) compared to the fixation task ( Figure 1d , red lines ) , and we interpret these response increases as reflecting top-down modulation . In some cases , the top-down modulation is even larger than the modulation achieved by manipulation of the stimulus . For example , the VWFA response to 3%-contrast words during the one-back task exceeds the response to 100%-contrast words . Note that the task effects cannot be explained simply by differences in spatial attention: the one-back task produces substantially larger responses than the categorization task despite the fact that both tasks require the locus of spatial attention to be on the stimulus . Task effects in lower-level areas exist but are smaller in size ( Figure 1—figure supplement 1 ) . A potential explanation of the top-down modulation is differences in task difficulty ( Ress et al . , 2000 ) . For example , it is presumably more difficult to perceive low-contrast stimuli than high-contrast stimuli , and this may explain why there is a large response enhancement for low- but not high-contrast stimuli ( see VWFA contrast-response function for word stimuli in Figure 1d ) . Later in this paper , we provide a computational mechanism that could underlie the psychological concept of task difficulty . In summary , our measurements indicate that VTC responses cannot be interpreted without specifying the cognitive state of the observer . A complete model of the computations performed by VWFA and FFA must consider the cognitive task in addition to stimulus properties . Before addressing the influence of top-down factors , we first develop a model of bottom-up responses in VWFA and FFA . Although the field has long understood that stimulus category is a good predictor of evoked responses ( Kanwisher et al . , 1997; Kriegeskorte et al . , 2008; Grill-Spector and Weiner , 2014 ) , we do not yet have a computational explanation of this phenomenon . In other words , although we are able to use our own visual systems to assign a label such as ‘word’ or ‘face’ to describe the data , we have not yet identified the operations that enable our visual systems to derive these labels in the first place . An additional limitation of our conceptual understanding is that it fails to account for the sensitivity of VWFA and FFA to low-level image properties . We therefore ask: Is it possible to develop a quantitative characterization of the bottom-up computations that can reproduce observed stimulus selectivity in human VTC ? Extending an existing computational model of fMRI responses in the visual system ( Kay et al . , 2008 , 2013b , 2013c ) , we conceive of a model involving two stages of image computations ( Figure 2a ) . The first stage consists of a set of local oriented filters , akin to what has been used to model physiological responses in V1 ( Jones and Palmer , 1987; Carandini et al . , 2005 ) . The second stage consists of a normalized dot product applied to the outputs of the first stage . This dot product computes how well a given stimulus matches a category template ( for example , a word template for VWFA , a face template for FFA ) . We construct category templates directly from the stimulus set used in the experiment; in a later section we explore how well this approach generalizes . The present model , termed the Template model , is almost certainly an oversimplification of the complex nonlinear processing performed in VTC . Nevertheless , the model is theoretically motivated , consistent with hierarchical theories of visual processing ( Fukushima , 1980; Heeger et al . , 1996; Serre et al . , 2007; DiCarlo et al . , 2012; Rolls , 2012 ) , and provides a useful starting point for characterizing the computations that underlie word- and face-selectivity . Moreover , unlike recently popular deep neural network models that also involve hierarchical processing ( Khaligh-Razavi and Kriegeskorte , 2014; Yamins et al . , 2014; Güçlü and van Gerven , 2015 ) , the model we propose is parsimonious with only three free parameters , and is therefore straightforward to fit and interpret ( see Figure 2a and Materials and methods ) . 10 . 7554/eLife . 22341 . 005Figure 2 . Model of bottom-up computations in VTC . ( a ) Model architecture . The predicted response of the Template model is given by a series of image computations ( see Materials and methods ) . ( b ) Cross-validation performance . Black bars indicate bottom-up stimulus-driven responses measured during the fixation task , dark lines and dark dots indicate model predictions ( leave-one-stimulus-out cross-validation ) , and light lines and light dots indicate model fits ( no cross-validation ) . Scatter plots in the inset compare model predictions against the data . The Template model is compared to the Category model which simply predicts a fixed response level for stimuli from the preferred stimulus category and a different response level for all other stimuli ( the slight decrease in response as a function of contrast is a result of the cross-validation process ) . ( c ) Comparison of performance against control models . Bars indicate leave-one-stimulus-out cross-validation performance . Error bars indicate 68% CIs , obtained by bootstrapping ( resampling subjects with replacement ) . Solid horizontal lines indicate the noise ceiling , that is , the maximum possible performance given measurement variability in the data . Dotted horizontal lines indicate the cross-validation performance of a model that predicts the same response level for each data point ( this corresponds to R2 = 0 in the conventional definition of R2 where variance is computed relative to the mean ) . The performance of the Template model degrades if the second stage of nonlinearities is omitted ( Template model ( only subtractive normalization ) ) or if the first stage of the model involving V1-like filtering is omitted ( Template model ( omit first stage ) ) . The plot also shows that the precise configuration of the template is important for achieving high model performance ( Template model ( non-selective , mixed , random templates ) ) . ( d ) Performance as a function of spatial frequency tuning . Here we manipulate the spatial frequency tuning of the filters in the Template model ( while fixing spatial frequency bandwidth at one octave ) . The Template model uses a single set of filters at a spatial frequency tuning of 4 cycles/degree . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 00510 . 7554/eLife . 22341 . 006Figure 2—figure supplement 1 . Testing the Template model on a wide range of stimuli . ( a ) Stimuli . We collected an additional dataset consisting of 92 images from a previous study by Kriegeskorte et al . ( 2008 ) ( all images shown ) , along with 22 images from the original experiment ( three images shown ) . We assessed model accuracy using 20-fold cross-validation across stimuli ( see Materials and methods for details ) . ( b ) Performance of Template model ( original ) . Black bars indicate data from FFA , with error bars indicating 68% CIs ( error across trials ) . Red lines and red dots indicate model predictions . Inset shows the category template used in the model . The model performs poorly . ( c ) Performance of Template model ( half-max average ) . This model derives the category template by computing ( in the V1-like representation ) the centroid of all stimuli in the training set that evoke at least half of the maximum response . Performance improves . ( d ) Performance of Template model ( half-max cluster ) . This model derives multiple category templates by performing k-means clustering ( in the V1-like representation ) on all stimuli in the training set that evoke at least half of the maximum response . Performance further improves , resolving both underprediction of responses ( for example , green arrow in panel b ) and overprediction of responses ( for example , blue arrow in panel ( b ) . ( e ) Results for VWFA . Similar responses are observed across the 92 Kriegeskorte images . Responses are well predicted by the original Template model , up to the level of measurement noise in this region . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 006 Applying the Template model to responses measured during the fixation task , we find that the Template model accurately predicts a large amount of variance in the responses of VWFA and FFA ( Figure 2b ) . The model outperforms a phenomenological model , termed the Category model , that posits that perceived stimulus category is sufficient to predict the response of category-selective regions . Notably , the Template model is able to predict the response to non-preferred stimulus categories in each ROI . This suggests that responses to non-preferred stimuli are meaningful and the result of a well-defined computation performed by the visual system ( Haxby et al . , 2001 ) . The model also outperforms simplified versions of the Template model that include only one of the two processing stages , as well as versions of the Template model in which the category template lacks tuning ( non-selective template ) , is equally weighted between words and faces ( mixed template ) , or is constructed randomly ( random template ) ( Figure 2c ) . The experiment we have conducted explores a limited range of stimuli . To further assess how well the Template model generalizes , we collected an additional dataset that includes 92 images taken from a previous study of object representation ( Kriegeskorte et al . , 2008 ) . In its original instantiation ( Figure 2a ) , the Template model uses category templates tailored to the stimuli in the main experiment , and we find that this instantiation of the Template model does not generalize well to the larger stimulus set ( Figure 2—figure supplement 1b ) . However , by implementing a simple model extension in which we use a data-driven approach to estimate category templates , we find that the Template model achieves a reasonable level of accuracy on the new stimulus set ( Figure 2—figure supplement 1d ) . This finding validates the basic architecture of the Template model , demonstrates how the Template model might be extended to account for increasingly large ranges of measurements , and provides a promising method to model response properties of other high-level visual regions not investigated here ( for example , place- and limb-selective cortex ) . The Template model advances us towards a computational understanding of VTC by demonstrating that BOLD responses in VTC can be predicted based on a template-matching operation on incoming visual inputs filtered by early visual cortex . The present results indicate that although high-level representations are not identical to low-level properties , they are built from , and fundamentally tied to , low-level properties through a series of linear and nonlinear operations . This conclusion is consistent with classic hierarchical theories of visual cortex ( Fukushima , 1980; Heeger et al . , 1996; Serre et al . , 2007; DiCarlo et al . , 2012; Rolls , 2012 ) and recent evidence that visual features may explain semantic representations found in high-level visual cortex ( Jozwik et al . , 2016 ) . Our model can be viewed as a potential mechanism for how semantic tuning properties emerge in visual cortex ( Huth et al . , 2012 ) . Our results indicate when studying high-level sensory representations in the brain , a precise characterization of the stimulus still matters . While the Template model explains bottom-up responses of VWFA and FFA as indexed by the fixation task , it does not explain why responses are higher in these areas during the categorization and one-back tasks ( see Figure 1d ) . This is simply because the stimuli are identical across the three tasks and the response of the Template model , like that of many computational models of visual processing , is solely a function of the stimulus . Before we can design a model to capture the top-down effects , we must first understand exactly how top-down signals shape the VTC response . By visualizing VTC responses as points in a multi-dimensional neural space with VWFA , FFA , and hV4 BOLD response amplitudes as the axes , we see that the responses to words and faces lie on specific manifolds , appearing as ‘arms’ that emanate from the origin ( Figure 3 ) . Importantly , we observe that the categorization and one-back tasks act as a scaling mechanism on the representation observed during the fixation task . The scaling mechanism moves the representation of each stimulus along the arms and away from the origin . Moreover , the amount of scaling is not constant across stimuli but is stimulus-specific , and this is most evident when considering the lowest contrast stimuli ( Figure 3 , black dots ) . 10 . 7554/eLife . 22341 . 007Figure 3 . Top-down stimulus-specific scaling of VTC representation . ( a ) Responses plotted in multi-dimensional neural space . Each dot indicates ROI ( VWFA , FFA ) responses to a stimulus . In each plot , the black line indicates a linear decision boundary separating words and faces ( nearest-centroid classifier , angular distance ) . ( b ) Schematics of potential top-down mechanisms ( these models are formally evaluated in Figure 5c; see Materials and methods section ‘IPS-scaling model’ for details ) . ( c ) Categorization and one-back tasks produce stimulus-specific scaling . Arrows indicate the change in representation compared to the fixation task . ( d ) Scaling improves readout . Each data point indicates the signed Euclidean distance between the word-face decision boundary ( as determined from the one-back task ) and the neural response to a single stimulus . Lines join data points that correspond to the same stimulus . The scaling observed during the categorization and one-back tasks moves responses away from the decision boundary , thereby improving signal-to-noise ratio . ( e ) Separation of other stimulus categories . Including hV4 as a third dimension reveals that stimuli categorized as neither words nor faces manifest as a third ‘arm’ that emanates from the origin . Although not reported to be a word by the subjects , the polygon stimulus behaves similarly to word stimuli . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 007 The visualization also shows that substantial responses to non-preferred categories are present in each ROI ( for example , faces in VWFA , words in FFA ) and that these responses are scaled during the stimulus-directed tasks . Thus , not only is information regarding non-preferred categories present in each ROI , but this information is actively modulated when subjects perform a perceptual task on those categories . These observations support the view that the brain uses a distributed strategy for perceptual processing and that category-selective regions are components of a more general network of regions that coordinate to extract visual information ( Haxby et al . , 2001; Cox and Savoy , 2003 ) . An alternative scheme , more in line with a modular view of perceptual processing ( Kanwisher and Wojciulik , 2000; Baldauf and Desimone , 2014 ) , is area-specific enhancement , in which the representation of a stimulus is enhanced only in the region that is selective for that stimulus ( for example , enhancement of words only in VWFA , enhancement of faces only in FFA ) . This scheme is not supported by our measurements ( Figures 3b and 3c; formal model evaluation is performed in a later section ) . Rather , response scaling occurs even for non-preferred stimulus categories , and the amount of scaling varies as a function of stimulus properties such as image contrast . A simple interpretation of the scaling effects is that they serve to increase signal-to-noise ratio in visually evoked responses in VTC ( Brouwer and Heeger , 2013 ) . For example , assuming that one use of the stimulus representation in VTC is to discriminate whether the presented stimulus is a word or face ( or , more generally , identify the category of the stimulus [DiCarlo et al . , 2012] ) , the scaling induced by the stimulus-directed tasks serves to increase the distance of neural responses from a linear decision boundary that separates words and faces ( Figure 3d ) . Interestingly , the categorization and one-back tasks appear to act via the same scaling mechanism . The stronger scaling observed for the one-back task might be a consequence of increased amplitude or duration of neural activity . These results suggest that , at least for the perceptual tasks sampled here and the spatial scale of neural activity measured in this study , top-down cognitive processes do not impart additional tuning or selectivity but serve to amplify the selectivity that is already computed by visual cortex . To design a plausible model that can predict top-down effects , we next turn to identifying the neural circuitry that generates task modulations in VTC . There are two candidate mechanisms . The first is that sensitivity to task is locally generated from the neuronal architecture of VTC itself . We explore an alternative hypothesis whereby top-down modulation is induced by input from another brain region that is sensitive to task demands . To identify this region , we perform a connectivity analysis in which we first subtract the bottom-up signal in VTC , as given by responses measured during the fixation task , from responses measured during the categorization and one-back tasks . We then correlate these residuals , which isolate the top-down signal , against the responses of every cortical location . Applying this connectivity analysis to our data , we find that responses in the intraparietal sulcus ( IPS ) predict the top-down enhancement of VTC responses ( Figure 4b ) better than responses in any other region of cortex . As a control , if we omit the subtraction step and simply correlate raw VTC responses with the responses of different cortical locations , we find that the correlation is instead strongest with a range of areas spanning occipital cortex ( Figure 4a ) . This indicates that the VTC response is a mixture of bottom-up and top-down effects and that the top-down influence from the IPS becomes clear only when bottom-up effects are removed . Comparing our results to a publicly available atlas ( Wang et al . , 2015 ) , we estimate that the source of top-down modulation is localized to the IPS-0 and IPS-1 subdivisions of the IPS ( see also Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 22341 . 008Figure 4 . IPS is the source of top-down modulation to VTC . ( a ) Correlation with raw VTC response . This map depicts the correlation between the VTC response observed during the categorization and one-back tasks with the response at each cortical location ( inset shows an unsmoothed and unthresholded map ) . Positive correlations are broadly distributed across occipital cortex . Results are shown for subjects with whole-brain coverage ( n = 3 ) ; results for other subjects with partial-brain coverage ( n = 6 ) are shown in Figure 4—figure supplement 1 . ( b ) Correlation with top-down component of VTC response . After removing bottom-up responses ( fixation task ) , the correlation is spatially localized to a hotspot in IPS-0/1 . ( c ) Tractography using diffusion MRI . We find that the vertical occipital fasciculus ( Yeatman et al . , 2014 ) connects VWFA and FFA to the IPS hotspot in each subject for which diffusion data were collected ( n = 8 ) ( rendering shows a representative subject ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 00810 . 7554/eLife . 22341 . 009Figure 4—figure supplement 1 . Maps of top-down connectivity to VTC . This figure shows thresholded and unthresholded maps for individual subjects and group averages ( same format as Figure 4b; all maps shown on the fsaverage surface ) . At the lower right of each map is the range of values used for the colormap . The left two columns show the results obtained for the six subjects with partial brain coverage . Group average results for these subjects are shown in the last row . The right two columns show the results obtained for the three subjects with full brain coverage . Group average results for these subjects are shown in the third to last row . Group average results for all subjects are shown in the second to last row . The last row shows the results obtained from a control analysis in which we generate individual-subject maps by correlating cortical responses with random Gaussian noise and then average these maps across subjects . This control analysis produces no substantial correlations . Notice that the peak correlation is found in and around IPS-0/1 for both the group of subjects with partial brain coverage ( red arrow ) and the group of subjects with full brain coverage ( green arrow ) . Some variability in the location of the peak correlation is expected given that there are limits on the degree to which functional areas can be aligned across subjects based solely on anatomical features . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 009 Previous research has identified IPS as playing a key role in controlling spatial attention ( Saalmann et al . , 2007; Lauritzen et al . , 2009 ) . Our results extend these findings by showing that , despite the fact that spatial attention is always directed towards the foveal stimulus during the categorization and one-back tasks , the amount of modulation from the IPS is flexible and varies depending on properties of the stimulus and demands of the task . For example , during the categorization task , the observed enhancement for low-contrast stimuli is much larger than that for high-contrast stimuli . This mechanism could explain the finding that difficult tasks enhance visual responses ( Ress et al . , 2000 ) . The direct influence of IPS on neural responses in VTC is consistent with anatomical measurements demonstrating the existence of a large white-matter pathway connecting dorsal and ventral visual cortex , called the vertical occipital fasciculus ( VOF ) ( Yeatman et al . , 2013 , 2014; Takemura et al . , 2016 ) . Using diffusion-weighted MRI and tractography ( data acquired in 8 of 9 subjects ) , we show that the VOF specifically connects the VWFA and FFA with the functionally identified peak region in the IPS ( Figure 4c ) . The VWFA falls within the ventral terminations of the VOF for seven subjects and , for the eighth , the VWFA is 2 . 7 mm anterior to the VOF , well within the margin of error for tractography ( Jeurissen et al . , 2011 ) . The FFA falls within the ventral terminations of the VOF for all eight subjects . These results provide an elegant example of how anatomy subserves function , and sets the stage for a circuit-level computational model that , guided by anatomical constraints , characterizes the computations that emerge from interactions between multiple brain regions . The previous two sections provide critical insights that set the stage for building a quantitative model that predicts top-down effects in VTC . Building upon the observation that top-down modulation acts as a scaling mechanism on responses in VWFA and FFA ( see Figure 3 ) and the observation that top-down effects are correlated with the IPS signal ( see Figure 4 ) , we propose that the magnitude of the IPS response to a stimulus indicates the amount of top-down scaling that is applied to bottom-up sensory responses in VTC ( Figure 5a ) . We implement this model , termed the IPS-scaling model , using response magnitudes extracted from a broad anatomical mask of the IPS . This strategy helps avoids the overfitting that might ensue from a more specific voxel-selection procedure tailored to the fine-scale and potentially idiosyncratic pattern of results from the connectivity analysis . For example , if we were to select the single cortical location in the IPS that best correlates with the top-down modulation of VTC , this would make voxel selection a critical part of the model and render the modeling analysis circular ( Kriegeskorte et al . , 2009 ) . Nevertheless , the selection procedure is not completely independent , so the modeling results should not be viewed as providing independent evidence for the involvement of the IPS . 10 . 7554/eLife . 22341 . 010Figure 5 . Model of top-down computations in VTC . ( a ) Model architecture . The predicted response during the stimulus-directed tasks ( categorization task , one-back task ) is given by scaling the bottom-up response , with the amount of scaling proportional to the IPS signal . ( b ) Cross-validation performance . Same format as Figure 2b . The arrows highlight an example of how the bottom-up response ( red arrow ) is multiplied by the IPS signal ( green arrow ) to produce the predicted response ( blue arrow ) . ( c ) Comparison of performance against alternative models . Same format as Figure 2c ( some error bars do not include the bar height; this is a consequence of the bootstrap procedure ) . Although the Additive and Scaling models perform well , note that these are ad hoc , phenomenological models . For instance , the Scaling model ( task-specific ) posits separate parameters for the amount of scaling under the categorization and one-back tasks . However , such a model does not explain why there is a different amount of scaling , whereas the IPS-scaling model provides such an explanation . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 010 We find that the IPS-scaling model accurately characterizes the observed data ( Figure 5b ) . For example , notice that the FFA response to faces increases gradually for each contrast increment during the fixation task ( relatively unsaturated contrast-response function , red arrow ) . When subjects perform the one-back task , we observe a U-shaped contrast-response function in IPS ( green arrow ) ; multiplication of the two functions predicts a contrast-response function that is highly saturated and accurately matches the observed contrast-response function in FFA during the one-back task ( blue arrow ) . Importantly , the IPS-scaling model uses a single set of scale and offset parameters on the IPS response and accurately predicts scaling of VTC responses across the categorization and one-back tasks ( Figure 5b , top plot ) . This finding suggests that the scaling of VTC by IPS is a general mechanism supporting perception and is independent of the specific cognitive task performed by the observer . Furthermore , the scale and offset parameters that are estimated from the data show that when IPS exhibits close to zero evoked activity ( for example , FACE at 100%-contrast; see Figure 1—figure supplement 1 ) , the corresponding scaling factor is close to one . This has a sensible interpretation: when IPS is inactive , we observe only the bottom-up response in VTC and no top-down modulation . We assessed the cross-validation performance of the IPS-scaling model in comparison to several alternative models of top-down modulation ( including those schematized earlier in Figure 3b ) . In line with earlier observations ( Figure 3b and c ) , we find that a model positing enhancement for only the preferred stimulus category of each area ( Area-specific enhancement model ) does not optimally describe the data . We find that a phenomenological scaling model ( Scaling model ( task-specific ) ) outperforms a phenomenological additive model ( Additive model ( task-specific ) ) , confirming earlier observations that the top-down modulation is a scaling effect ( Figure 5c ) . This conclusion is further supported by the higher performance observed when the IPS interacts with VTC multiplicatively ( IPS-scaling model ) compared to when it interacts additively ( IPS-additive model ) . Finally , we find that the performance of the IPS-scaling model degrades if the IPS input into the model is shuffled across conditions ( IPS-scaling model ( shuffle , shuffle within task ) ) , confirming that top-down modulation from the IPS is dependent on the stimulus and task . Is the IPS the only region that induces top-down modulation of VTC ? Inspection of the connectivity results ( see Figure 4b ) reveals that the top-down residuals in VTC are correlated , to a lesser extent , with the responses of other regions . These weaker correlations might be incidental , or might capture other important signals . Given that the IPS-scaling model accounts for nearly all of the variance induced by top-down modulation of VTC ( see Figure 5b and c ) , we suggest that it is sufficient to consider only the IPS for the current set of measurements . However , future measurements that employ new stimulus manipulations and other cognitive tasks may reveal the role of a more extensive brain network . The IPS-scaling model can be extended to account for new measurements by systematically parameterizing the connectivity with additional brain regions . For example , some models of reading posit that language-related regions can directly influence the VWFA ( Twomey et al . , 2011 ) , suggesting that to account for measurements made during a more naturalistic reading task , it may be necessary to include Broca’s area in the model . Although informative , the finding that IPS provides top-down stimulus-specific scaling of VTC is an incomplete explanation , as the burden of explaining the top-down effects is simply shifted to the IPS . We are thus left wondering: is it possible to explain the response profile of the IPS ? In particular , can we explain why the IPS is more active for certain stimuli compared to others ? Answering these questions will provide a critical link between cognitive state and IPS activity . Inspired by previous research on perceptual decision-making ( Shadlen and Newsome , 2001; Heekeren et al . , 2004; Gold and Shadlen , 2007; Kayser et al . , 2010a ) , we implement a Drift diffusion model that attempts to account for IPS responses measured during the categorization task ( Figure 6a ) . The model uses VTC responses during the fixation task as a measure of sensory evidence , and posits that the IPS accumulates evidence from VTC over time and exhibits an activity level that is monotonically related to accumulation time . For example , when VTC responses are small , as is the case for low-contrast stimuli , sensory evidence for stimulus category is weak , leading to long accumulation times ( indexed by measurements of reaction time during the experiment ) , and large IPS responses . 10 . 7554/eLife . 22341 . 011Figure 6 . Model of perceptual decision-making in IPS . ( a ) Model architecture . We implement a model that links the stimulus representation in VTC to a decision-making process occurring in IPS . The model first uses the bottom-up VTC response as a measure of sensory evidence and predicts reaction times in the categorization task . The model then predicts the IPS response as a monotonically increasing function of reaction time . Note that this model does not involve stochasticity in the evidence-accumulation process , and is therefore a simplified version of the classic drift diffusion model ( Ratcliff , 1978 ) . ( b ) Cross-validation performance . Same format as Figure 2b ( except that reaction times are modeled in the left plot ) . ( c ) Comparison of performance against control models . The performance of the Drift diffusion model does not degrade substantially if a single threshold is used , thus justifying this simplification . Performance degrades if axis-aligned category vectors are used , supporting the assertion that responses of multiple VTC regions are used by subjects in deciding image category . ( d ) Overall model architecture . This schematic summarizes all components of our computational model ( Figures 2a , 5a and 6a ) . Bottom-up visual information is encoded in the VTC fixation response ( green box; Template model ) , fixation responses are routed to the IPS for evidence accumulation ( purple box; Drift diffusion model ) , and then feedback from the IPS to VTC causes top-down modulation during the categorization and one-back tasks ( yellow box; IPS-scaling model ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22341 . 011 Our implementation of the Drift diffusion model involves two steps . First , we use VTC responses during the fixation task ( reflecting sensory evidence ) to predict reaction times measured in the categorization task . The quality of the predictions is quite high ( Figure 6b , left ) . Second , we apply a simple monotonic function to the reaction times measured during the categorization task to predict the level of response in the IPS ( see Materials and methods ) . The rationale is that neural activity in IPS is expected to be sustained over the duration of the decision-making process ( Shadlen and Newsome , 2001 ) , and so the total amount of neural activity integrated over time should be larger for longer decisions . Assuming that the BOLD signal reflects convolution of a sluggish hemodynamic response function and fine-scale neural activity dynamics , small differences in the duration of neural activity ( for example , between 0 and 2 s ) are expected to manifest in differences in BOLD amplitudes ( Kayser et al . , 2010a ) and only minimally in the shapes of BOLD timecourses . The cross-validated predictions of our proposed model explain substantial variance in IPS ( Figure 6b , right ) . It is possible to offer a psychological explanation of IPS activity as reflecting task difficulty—for example , we can posit that IPS activity is enhanced for low-contrast stimuli because the observer works harder to perceive these stimuli . The value of the model we have proposed is that it provides a quantitative and formal explanation of the computations that underlie ‘difficulty’ . According to the model , categorization of low-contrast stimuli is difficult because the IPS computations required to perform the task involve longer accumulation time , and this is reflected in the fact that IPS response magnitudes increase monotonically with reaction time . Thus , our model performs several critical functions: it relates the cognitive task performed by the subject to IPS activity , proposes a computational explanation of task difficulty , and posits that top-down modulation of VTC by IPS is a direct consequence of fulfilling task demands . We have substantiated this hypothesis for the categorization task and suggest that this will serve as a foundation for modeling more complex cognitive tasks such as one-back .
In summary , we have measured and modeled how bottom-up and top-down factors shape responses in word- and face-selective cortex . A template operation on low-level visual properties generates a bottom-up stimulus representation , while top-down modulation from the IPS scales this representation in the service of the behavioral goals of the observer . We develop a computational approach that posits explicit models of the information processing performed by a network of interacting sensory and cognitive regions of the brain and validate this model on experimental data . We make publicly available data and open-source software code implementing the model at Kay , 2017 ( with a copy archived at https://github . com/elifesciences-publications/vtcipsmodel ) . The model we propose is valuable because it integrates and explains a range of different stimulus and task manipulations that affect responses in VWFA , FFA , and IPS . Response properties in these regions can now be interpreted using a series of simple , well-defined computations that can be applied to arbitrary images . However , it is also important to recognize the limitations of the model . First , we have tested the model on only a limited range of stimuli and cognitive tasks . Second , the accuracy with which the model accounts for the data is reasonable but by no means perfect . For instance , the Template model does not capture the step-like response profile of VTC as phase coherence is varied ( see Figure 2b ) , and the accuracy with which the model accounts for a wide range of stimuli that includes faces , animals , and objects , is moderate at best ( see Figure 2—figure supplement 1 ) . Third , we have thus far characterized VTC and IPS responses at only a coarse spatiotemporal scale ( that is , BOLD responses averaged over specific regions-of-interest ) . Given these limitations , the present work constitutes a first step towards the goal of developing a comprehensive computational model of human high-level visual cortex . We have provided data and code so that other researchers can build on our approach , for example , by testing the generalizability of the model to other stimuli and tasks , extending and improving the model , and comparing the model against alternative models . The fact that cognitive factors substantially affect stimulus representation in visual cortex highlights the importance of tightly controlling and manipulating cognitive state when investigating stimulus selectivity . In the present measurements , the most striking example comes from stimulus contrast . When subjects perform the fixation task , the contrast-response function ( CRF ) in VWFA is monotonically increasing , whereas during the one-back task , the CRF flips in sign and is monotonically decreasing ( see Figure 1d ) . This effect ( similar to what is reported in Murray and He , 2006 ) is puzzling if we interpret the CRFs as indicating sensitivity to the stimulus contrast , but is sensible if we interpret the CRFs as instead reflecting the interaction of stimulus properties and cognitive processes . The influence of cognition on visual responses forces us to reconsider studies that report unexpected tuning properties in VTC and IPS , such as tuning to linguistic properties of text in VWFA ( Vinckier et al . , 2007 ) and object selectivity in parietal cortex ( Sereno and Maunsell , 1998 ) . In experiments that do not tightly control the cognitive processes executed by the observer , it is impossible to distinguish sensory effects from cognitive effects . Our quantitative model of VTC-IPS interactions provides a principled baseline on which to re-interpret past findings , design follow-up experiments , and guide data analysis . There is a large body of literature on characterizing and modeling the effect of spatial attention on contrast-response functions ( Luck et al . , 1997; Boynton , 2009; Reynolds and Heeger , 2009; Itthipuripat et al . , 2014 ) . Although several models have been proposed , none of these straightforwardly account for the present set of measurements . The response-gain model ( McAdams and Maunsell , 1999 ) posits that attention causes a multiplicative scaling of contrast-response functions . Our observations are consistent with the general notion of response scaling ( see Figure 3 ) , but importantly , we find that the amount of scaling differs for different stimuli . Whereas the response-gain model implies that scaling is constant and therefore contrast-response functions should grow steeper during the stimulus-directed tasks , we find the opposite ( see Figure 1d ) . The contrast-gain model ( Reynolds et al . , 2000 ) posits that attention causes a leftward shift of contrast-response functions ( as if contrast were increased ) . This model does not account for our measurements , since the stimulus-directed tasks can generate responses to low-contrast stimuli that are larger than responses to high-contrast stimuli ( see Figure 1d ) . Finally , the additive-shift model ( Buracas and Boynton , 2007 ) posits that attention causes an additive increment to contrast-response functions; we find that our observations are better explained by a scaling , not additive , mechanism ( see Figures 3 and 5c ) . Thus , the effects we report are novel , and previous models of attention cannot explain these effects . Furthermore , by investigating responses to a wide range of stimuli ( including manipulations of not only contrast but also phase coherence and stimulus category ) and by characterizing the source of attentional signals , our work develops a more comprehensive picture of information processing in the visual system . There are a number of research questions that remain unresolved . First , our connectivity analysis and modeling of VTC-IPS interactions are based on the correlation of BOLD responses and do not provide information regarding the directionality , or timing , of neural interactions . In other words , our correlational results do not , in and of themselves , prove that the IPS causes VTC modulation; rather , we are imposing an interpretation of the results in the context of a computational model . We note that our interpretation is in line with previous work on perceptual decision-making showing top-down influence ( Granger causality ) of IPS on visual cortex ( Kayser et al . , 2010b ) . Our working hypothesis is that sensory information arrives at VTC ( as indexed by fixation responses ) , these signals are routed to IPS for evidence accumulation , and then feedback from the IPS modulates the VTC response ( as indexed by categorization and one-back responses ) . Temporally resolved measurements of neural activity ( for example , EEG , MEG , ECoG ) will be necessary to test this hypothesis . Second , the scaling of BOLD response amplitudes by IPS is consistent with at least two potential mechanisms at neural level: the IPS may be inducing a scaling on neural activity in VTC or , alternatively , a sustainment of neural activity in VTC . Some support for the latter comes from a recent study demonstrating that ECoG responses in FFA exhibit sustained activity that is linked to long reaction times in a face gender discrimination task ( Ghuman et al . , 2014 ) . Finally , IPS is part of larger brain networks involved in attention ( Corbetta and Shulman , 2002 ) and decision-making ( Gold and Shadlen , 2007 ) , and identifying the computational roles of other regions in these networks is necessary for a comprehensive understanding of the neural mechanisms of perception .
Eleven subjects participated in this study . Two subjects were excluded due to inability to identify VWFA in one subject and low signal-to-noise ratio in another subject , leaving a total of nine usable subjects ( age range 25–32; six males , three females ) . All subjects were healthy right-handed monolingual native-English speakers , had normal or corrected-to-normal visual acuity , and were naive to the purposes of the experiment . Informed written consent was obtained from all subjects , and the experimental protocol was approved by the Washington University in St . Louis Institutional Review Board . Each subject participated in 1–3 scanning sessions , over the course of which anatomical data ( T1-weighted high-resolution anatomical volume , diffusion-weighted MRI data ) and functional data ( retinotopic mapping , functional localizer , main experiment ) were collected . Stimuli were presented using an NEC NP-V260X projector . The projected image was focused onto a backprojection screen and subjects viewed this screen via a mirror mounted on the RF coil . The projector operated at a resolution of 1024 × 768 at 60 Hz , and the viewing distance was 340 cm . A Macintosh laptop controlled stimulus presentation using code based on the Psychophysics Toolbox ( Brainard , 1997; Pelli , 1997 ) . Approximate gamma correction was performed by taking the square root of pixel intensity values before stimulus presentation . Behavioral responses were recorded using a button box . The experiment consisted of 22 types of stimuli . All stimuli were small grayscale images ( approximately 2° × 2° ) presented at fixation . Each stimulus type consisted of 10 distinct images ( for example , 10 different faces for a face stimulus ) , and a subset of these images were presented on each given trial . Stimuli were presented in 4 s trials , one stimulus per trial . In a trial , four images from a given stimulus type ( for example , FACE , 10% contrast ) were presented sequentially using an 800 ms ON , 200 ms OFF duty cycle . To generate the sequence of four images , we first randomly selected four distinct images out of the ten images associated with the stimulus type . Then , for certain trials ( details below ) , we modified the sequence to include a repetition by randomly selecting one of the images ( excluding the first ) and replacing that image with the previous image . Throughout stimulus presentation , a small dot ( 0 . 12° × 0 . 12° ) was present at the center of the display . The dot switched to a new randomly selected color every 600 ms using a set of six possible colors: magenta , red , yellow , green , cyan , and blue . In the experiment , two of the stimuli were duplicated ( FACE and WORD ) , yielding a total of 24 stimulus conditions . Data corresponding to these duplicate stimuli are not used in this paper . Each run began and ended with a 16 s baseline period in which no stimuli were presented . During a run , each of the 24 stimulus conditions was presented three times . Six blank trials ( no stimulus ) were also included . The order of stimulus and blank trials was random , subject to the constraints that blank trials could not occur first nor last , blank trials could not occur consecutively , and no stimulus condition could occur consecutively . During the baseline periods and blank trials , the small central dot was still present . A randomly selected two of the three trials associated with each stimulus condition were modified to include an image repetition ( as described previously ) . Each run lasted 344 s ( 5 . 7 min ) . For each run , subjects were instructed to maintain fixation on the central dot while performing one of three tasks . In the fixation task , subjects were instructed to press a button whenever the central dot turned red , and were additionally reminded to not confuse the red and magenta colors . In the categorization task , subjects were instructed to report for each stimulus trial whether they perceived a word , a face , or neither ( ‘other’ ) . Responses were made using three different buttons , and subjects were reminded to make only one response for each 4 s trial . Note that it is possible that responses are made prior to the completion of the four images that comprise a trial . In the one-back task , subjects were instructed to press a button whenever an image was repeated twice in a row , and were informed that repetitions occurred only within stimulus trials and not across trials . Subjects were warned that although some stimuli are faint ( low contrast ) , they should still try their best to perform the categorization and one-back tasks . Subjects were also informed that some trials are blank trials and that responses were not expected on these trials . Subjects were familiarized with the stimuli and tasks before the actual experiment was conducted . Subjects performed each of the three tasks four times during the course of the experiment , yielding a total of 3 tasks × 4 runs = 12 runs . The physical stimulus sequence ( including the temporal ordering of stimulus images and dot colors ) was held constant across tasks . This was accomplished by generating four distinct stimulus sequences and cycling through the sequences and tasks . Specifically , the order of stimulus sequences was ABCD ABCD ABCD , where each letter corresponds to a distinct sequence , and the order of tasks was XYZ XYZ XYZ XYZ , where each letter corresponds to a distinct task . The order of tasks was counterbalanced across subjects . Each stimulus and task combination ( for example , CHECKERBOARD during one-back task ) occurred a total of 3 trials × 4 runs = 12 times over the course of the experiment . MRI data were collected at the Neuroimaging Laboratory at the Washington University in St . Louis School of Medicine using a modified 3T Siemens Skyra scanner and a 32-channel RF coil . For functional data , 28 oblique slices covering occipitotemporal cortex were defined: slice thickness 2 . 5 mm , slice gap 0 mm , field-of-view 200 mm × 200 mm , phase-encode direction anterior-posterior . A T2*-weighted , single-shot , gradient-echo EPI sequence was used: matrix size 80 × 80 , TR 2 s , TE 30 ms , flip angle 77° , nominal spatial resolution 2 . 5 mm × 2 . 5 mm × 2 . 5 mm . Fieldmaps were acquired for post-hoc correction of EPI spatial distortion . To achieve comprehensive coverage for localization of top-down effects , a whole-brain version of the protocol involving 58 slices and a multiband ( Feinberg et al . , 2010 ) factor of 2 was used in three of the nine subjects . In addition to functional data , T1-weighted anatomical data ( MPRAGE sequence , 0 . 8 mm resolution ) and diffusion-weighted data ( spin-echo EPI sequence , 2 mm resolution , 84 directions , b-values of 1500 and 3000 ) were acquired . The diffusion sequence was acquired twice , reversing the phase-encode direction , in order to compensate for spatial distortions . Diffusion data were not acquired for one subject due to time constraints . Behavioral results for the categorization task are used in the present study . We analyzed both reaction times ( RT ) and category judgments . We defined RT as the time elapsed between the onset of the first of the four images in a given trial and the button press . Trials in which no buttons were pressed were ignored . For each subject , we summarized RTs by computing the median RT across trials for each stimulus . To obtain group-averaged RTs , we added a constant to each subject's RTs in order to match the mean RT to the grand mean across subjects and then computed the mean and standard error across subjects ( this normalization procedure compensates for additive offsets in RT across subjects ) . Category judgments were analyzed by calculating percentages of trials on which a given subject categorized a given stimulus into each of the three categories ( word , face , other ) . Subjects were highly consistent in their judgments: for each stimulus , the most frequently reported category was the same across subjects and was reported more than 85% of the time . Category judgments obtained from the categorization task are used in the labeling and interpretation of experimental results ( for example , Figures 1 and 3 ) . Subject motion was corrected by co-registering each volume to the average of the non-diffusion-weighted b = 0 images . Gradient directions were adjusted to account for the co-registration . From pairs of volumes acquired with reversed phase-encode directions , the susceptibility-induced off-resonance field was estimated using a method similar to that described in Andersson et al . ( 2003 ) as implemented in FSL ( Smith et al . , 2004 ) . Eddy currents were corrected using FSL’s eddy tool . The b = 3000 measurements were used to estimate fiber orientation distribution functions for each voxel using constrained spherical deconvolution as implemented in mrtrix ( Tournier et al . , 2007 ) ( CSD , lmax = 4 ) , and fiber tracts were estimated using probabilistic tractography ( 500 , 000 fibers ) . For each subject , we identified the vertical occipital fasciculus ( VOF ) using a previously published algorithm ( Yeatman et al . , 2014 ) , and then quantified the Euclidean distance from the VOF terminations to word- and face-selective regions in VTC and the task-related hotspot in the IPS . The T1-weighted anatomical volume acquired for each subject was processed using FreeSurfer ( Fischl , 2012 ) . The results were used to create a cortical surface reconstruction positioned halfway between the pial surface and the boundary between gray and white matter . We used the fsaverage surface from FreeSurfer to define anatomical ROIs ( details below ) . These ROIs were transformed to native subject space by performing nearest-neighbor interpolation on the spherical surfaces created by FreeSurfer ( these surfaces reflect folding-based alignment of individual subject surfaces to the fsaverage surface ) . Functional data were pre-processed by performing slice time correction , fieldmap-based spatial undistortion , motion correction , and registration to the subject-native anatomical volume . The combined effects of distortion , motion , and registration were corrected using a single cubic interpolation of the slice time corrected volumes . Interpolations were performed directly at the vertices of the subject’s cortical surface , thereby avoiding unnecessary interpolation and improving spatial resolution ( Kang et al . , 2007 ) . The pre-processed fMRI data were analyzed using GLMdenoise ( Kay et al . , 2013a ) ( http://kendrickkay . net/GLMdenoise/ ) , a data-driven denoising method that derives estimates of correlated noise from the data and incorporates these estimates as nuisance regressors in a general linear model ( GLM ) analysis of the data . For our experiment , we coded each stimulus and task combination as a separate condition and also included the blank trials , producing a total of ( 24 stimulus + 1 blank ) × 3 tasks = 75 conditions . The response to blank trials was interpreted as the response to a 0%-contrast stimulus . Estimates of BOLD response amplitudes ( beta weights ) were converted to units of percent BOLD signal change by dividing amplitudes by the mean signal intensity observed at each vertex . To obtain ROI responses , beta weights were averaged across the vertices composing each ROI . Error bars ( 68% CIs ) on beta weights were obtained by bootstrapping runs . Group-averaged beta weights were calculated using a procedure that compensates for large intrinsic variation in percent BOLD change across subjects . First , the beta weights obtained for each subject in a given ROI were normalized to be a unit-length vector ( for example , b~i=bi/‖bi‖ where bi indicates beta weights for the ith subject ( 1 x n ) , ‖ ‖ indicates L2-norm , and bi~ indicates normalized beta weights for the ith subject ) . Next , normalized beta weights were averaged across subjects , using bootstrapping to obtain error bars ( 68% CIs ) . Finally , the resulting group-averaged beta weights were multiplied by a scalar such that the mean of the beta weights is equal to the mean of the original unnormalized beta weights obtained from all subjects . The motivation of this last step is to produce interpretable units of percent BOLD change instead of normalized units . Note that in some cases , beta weights are repeated for easier visualization ( for example , in Figure 1 , NOISE at 100% contrast is the same data point as FACE at 0% phase coherence ) . Group-averaged beta weights were used in computational modeling . Visual field maps were defined using the population receptive field ( pRF ) technique applied to retinotopic mapping data ( Dumoulin and Wandell , 2008; Kay et al . , 2013b ) . Subjects participated in 2–4 runs ( 300 s each ) in which they viewed slowly moving apertures ( bars , wedges , rings ) filled with a colorful texture of objects , faces , and words placed on an achromatic pink-noise background . The aperture and texture were updated at 5 Hz , and blank periods were included in the design ( Dumoulin and Wandell , 2008 ) . A semi-transparent fixation grid was superimposed on top of the stimuli ( Schira et al . , 2009 ) . Stimuli occupied a circular region with diameter 10° and the viewing distance was 251 cm . A small semi-transparent central dot ( 0 . 15° × 0 . 15° ) was present throughout the experiment and changed color every 1–5 s . Subjects were instructed to maintain fixation on the dot and to press a button whenever its color changed . The time-series data from this experiment were modeled using the Compressive Spatial Summation model ( Kay et al . , 2013b ) as implemented in analyzePRF ( http://kendrickkay . net/analyzePRF/ ) . Angle and eccentricity estimates provided by the model were then visualized on cortical surface reconstructions and used to define V1 , V2 , V3 , and hV4 ( Brewer et al . , 2005 ) . Due to the limited amount of pRF data acquired , there was insufficient signal-to-noise ratio to define visual field maps in parietal cortex . Category-selective regions FFA and VWFA were defined using functional localizers ( Weiner and Grill-Spector , 2010 , 2011 ) . Subjects participated in two runs ( 336 s each ) in which they viewed blocks of words , faces , abstract objects , and noise patterns . Each block lasted 16 s and consisted of 16 images presented at a rate of 1 Hz . The images differed from those in the main experiment . In each run , the four stimulus types were presented four times each in pseudorandom order , with occasional 16 s blank periods . A semi-transparent fixation grid was superimposed on top of the stimuli ( Schira et al . , 2009 ) . Stimuli occupied a 4° × 4° square region , with the words , faces , and objects occupying the central 3° × 3° of this region . The viewing distance was 340 cm . Subjects were instructed to maintain central fixation and to press a button when the same image is presented twice in a row . The time-series data from this experiment were analyzed using a GLM to estimate the amplitude of the BOLD response to the four stimulus categories . To define FFA and VWFA , responses to the four stimulus categories were visualized on cortical surface reconstructions . FFA and VWFA were defined based on stimulus selectivity , anatomical location , and topological relationship to retinotopic areas ( Weiner and Grill-Spector , 2010; Yeatman et al . , 2013; Weiner et al . , 2014 ) . We defined FFA as face-selective cortex ( responses to faces greater than the average response to the other three categories ) located on the fusiform gyrus . We included in the definition both the posterior fusiform gyrus ( pFus-faces/FFA-1 ) and middle fusiform gyrus ( mFus-faces/FFA-2 ) subdivisions of FFA ( Weiner et al . , 2014 ) . We defined VWFA as word-selective cortex ( responses to words greater than the average response to the other three categories ) located in and around the left occipitotemporal sulcus . In some subjects , multiple word-selective patches were found , and all of these patches were included in the definition of VWFA . Anatomically-defined ROIs were also created ( see Figure 4a and b ) . Based on curvature values on the fsaverage surface , we created an anatomical mask of the IPS by selecting the posterior segment of the intraparietal sulcus ( Pitzalis et al . , 2012 ) . Using the atlas of visual topographic organization provided by Wang et al . ( Wang et al . , 2015 ) , we estimate that this IPS mask overlaps V3A , V3B , IPS-0 , IPS-1 , and IPS-2 . The locations of IPS-0/1/2/3 from the atlas are shown in Figure 4 and Figure 4—figure supplement 1 . We also created an anatomical mask of VTC by computing the union of the fusiform and inferiortemporal parcels provided by the FreeSurfer Desikan-Killiany atlas ( Desikan et al . , 2006 ) and trimming the anterior extent of the result to include only visually responsive cortex . The VTC mask includes both FFA and VWFA as well as surrounding cortex . In our data , we find that word-selective visual cortex in some subjects is confined to the left hemisphere , consistent with previous studies ( Yeatman et al . , 2013 ) . Therefore , to ease interpretation , we restricted our analysis to VWFA , FFA , VTC , and IPS taken from the left hemisphere . In addition , we restricted the definition of V1 , V2 , V3 , hV4 , VTC , and IPS to include only vertices exhibiting response amplitudes in the main experiment that are positive on average . This procedure excludes voxels with peripheral receptive fields which typically exhibit negative BOLD responses to centrally presented stimuli . To identify the cortical region that generates top-down effects in VWFA and FFA , we performed a simple connectivity analysis . First , we averaged BOLD responses across our VTC mask , given that top-down effects appear broadly across VTC . Next , we identified the component of the VTC response that is of no interest , specifically , the bottom-up stimulus-driven response . Our estimate of this component is given by our measurement of VTC responses during the fixation task ( 22 stimuli + 1 blank = 23 values ) . We then subtracted the bottom-up component from the VTC response measured during the categorization task ( 23 values ) and one-back task ( 23 values ) . This produced a set of residuals ( 46 values ) that reflect the top-down effect in VTC . Finally , we correlated the residuals with the responses of each cortical location in our dataset during the categorization and one-back tasks ( 46 values ) . The cortical location that best correlates with the residuals is interpreted as a candidate region that supplies top-down modulation to VTC . The results were visualized by averaging correlation values across subjects based on the fsaverage cortical alignment and plotting the results on the fsaverage surface . Results from the three subjects for which whole-brain fMRI data were acquired are shown in Figure 4a and b . Results from the remaining six subjects with limited fMRI coverage are provided in Figure 4—figure supplement 1 . Note that the correlation-based analysis we have used is most suitable for connectivity effects that are additive in nature ( for example , IPS providing additive enhancement to VTC ) . However , the modulation is more accurately characterized as a multiplicative , or scaling , effect ( see Figure 3b and c ) . The advantage of correlation is that it is robust to noise and computationally efficient; we perform a more precise evaluation of different top-down mechanisms in the computational modeling section below . There are three important differences between the connectivity analysis described here and conventional correlation-based resting-state functional connectivity ( RSFC ) ( Buckner et al . , 2013 ) and the psychophysiological interactions ( PPI ) technique ( O'Reilly et al . , 2012 ) . One is that our connectivity is performed on data that have explicit manipulation of stimulus and task ( unlike RSFC ) . Another is that we analyze the data explicitly in terms of information-processing operations performed by the brain ( unlike PPI ) . In other words , functional connectivity is characterized , not as correlated signal fluctuations , but as a direct consequence of information-processing operations . A third difference is that our connectivity is performed on beta weights that pool across trials ( Rissman et al . , 2004 ) , as opposed to raw BOLD time-series . This concentrates the analysis on brain responses that are reliably driven by the stimulus and task , and de-emphasizes trial-to-trial fluctuations in cognitive performance ( Donner et al . , 2013 ) . We developed a computational model to account for BOLD responses measured in VTC and IPS . The model is composed of three components , each of which addresses a different aspect of the data ( Figure 6d ) . The first component ( Template model ) specifies how a given stimulus drives bottom-up VTC responses as measured during the fixation task; the second component ( IPS-scaling model ) specifies how top-down modulation from the IPS during the categorization and one-back tasks affects VTC responses; and the third component ( Drift-diffusion model ) specifies how accumulation of evidence from VTC predicts reaction times and IPS responses during the categorization task . Note that although the three model components could be yoked together ( for example , the output from the Template model could serve as the input to the Drift-diffusion model ) , in our model implementations , we adopt the approach of isolating each model component so that the quality of each component can be assessed independently of the others . For all three model components , computational modeling was performed using nonlinear least-squares optimization ( MATLAB Optimization Toolbox ) . Leave-one-stimulus-out cross-validation was used to assess model accuracy ( thus , we assess the ability of models to generalize to stimuli that the models have not been trained on ) . Note that the use of cross-validation enables fair comparison of models that have different levels of flexibility ( or , informally , different numbers of free parameters ) . This is because models that are overly complex will tend to fit noise in the training data and thereby generalize poorly to the testing data ( Hastie et al . , 2001 ) . Accuracy was quantified as the percentage of variance explained ( R2 ) between cross-validated predictions of the data ( aggregated across cross-validation iterations ) and the actual data . In the case of beta weights , variance was computed relative to 0% BOLD signal change ( Kay et al . , 2013b ) . In certain cases , accuracy is reported using Pearson’s correlation ( r ) ; this metric assesses performance relative to the mean . To assess reliability of cross-validation results , model fitting and cross-validation were repeated for each bootstrap of the group-averaged data ( resampling subjects with replacement ) . For benchmarks on cross-validation performance , we calculated noise ceilings using Monte Carlo simulations ( Kay et al . , 2013b ) and quantified the performance of a flat-response model that predicts the same response level for each data point . Software code implementing the model proposed in this paper is available at http://cvnlab . net/vtcipsmodel/ . | As your eyes scan this page , your visual system performs a series of computations that allow you to derive meaning from the printed words . The visual system solves this task with such apparent ease that you may never have thought about the challenges that your brain must overcome for you to read a page of text . The brain must overcome similar challenges to enable you to recognize the faces of your friends . Two factors affect how the neurons in the visual system respond to what you are looking at: the physical features of the object and your cognitive state ( for example , your knowledge , past experiences , and the cognitive demands of the task at hand ) . To figure out exactly how these factors influence the responses of the neurons , Kay and Yeatman used functional magnetic resonance imaging to scan the brains of human volunteers as they viewed different images ( some of which were of faces or words ) . The volunteers had to perform various tasks while viewing the images . These tasks included focusing their attention on a small dot , categorizing the image , and stating whether the image had previously been shown . From the brain imaging data , Kay and Yeatman developed a model of the brain circuits that enable faces and words to be recognized . The model separately characterizes the influence of physical features and the influence of cognitive state , and describes several different types of processing: how the brain represents what is seen , how it makes a decision about how to respond , and how it changes its own activity when carrying out the decision . The model has been made freely available online so that other researchers can reproduce and build upon Kay and Yeatman’s findings . The current model is not perfect , and it does not describe neural activity in fine detail . However , by obtaining new experimental measurements , the model could be systematically improved . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2017 | Bottom-up and top-down computations in word- and face-selective cortex |
Many reef-building corals participate in a mass-spawning event that occurs yearly on the Great Barrier Reef . This coral reproductive event is one of earth's most prominent examples of synchronised behavior , and coral reproductive success is vital to the persistence of coral reef ecosystems . Although several environmental cues have been implicated in the timing of mass spawning , the specific sensory cues that function together with endogenous clock mechanisms to ensure accurate timing of gamete release are largely unknown . Here , we show that moonlight is an important external stimulus for mass spawning synchrony and describe the potential mechanisms underlying the ability of corals to detect environmental triggers for the signaling cascades that ultimately result in gamete release . Our study increases the understanding of reproductive chronobiology in corals and strongly supports the hypothesis that coral gamete release is achieved by a complex array of potential neurohormones and light-sensing molecules .
Adaptation to environmental cycles including daily , tidal , lunar , or annual cycles over millions of years , has enabled marine organisms to synchronize many aspects of their biology , thereby creating important biological rhythms for coping with a variable world ( Tessmar-Raible et al . , 2011 ) . One of nature’s greatest examples of synchronized behavior is the coral spawning that occurs on the Great Barrier Reef ( GBR ) . The ‘mass-spawning events’ on the GBR is the Earth’s largest reproduction event . During these annual events ( Babcock et al . , 1986; Harrison et al . , 2011 ) changes in the intensity of moonlight trigger the spawning of more than 130 species of scleractinian corals as well as hundreds of other invertebrates over a couple of nights ( Babcock et al . , 1986; Harrison et al . , 2011 ) . Broadcast spawning requires coral colonies to carefully synchronize the release of egg and sperm into the water column in order to optimize fertilization success ( Harrison et al . , 2011; Levitan , 2005 ) . Environmental factors , such as sea surface temperature , lunar phase and the daily light cycle , influence reproductive timing in coral and induce spawning ( Babcock et al . , 1986; Babcock et al . , 1994; Harrison et al . , 1984 ) . However , the mechanism by which corals fine-tune and coordinate spawning remains unclear . Seasonal patterns are crucial for inducing gametogenesis and spawning in broadcast spawning species ( Mendes and Woodley , 2002 ) , where rising sea temperature is the most likely environmental driver behind the maturation of eggs and sperm ( Babcock et al . , 1986; Harrison et al . , 1984; Oliver et al . , 1988 ) . The phase of the moon , on the other hand , coordinates the timing of mass spawning , selecting neap tides that reduce gamete dilution in the water column ( Babcock et al . , 1986; Babcock et al . , 1994; Mendes and Woodley , 2002; Willis et al . , 1985 ) . Spawning occurs a few days after the full moon and at a precise time after sunset , although the time of day and night ( s ) on which spawning and gamete release occurs is highly species-specific . In addition to being influenced by moonlight ( Babcock et al . , 1986; Harrison et al . , 1990; Glynn et al . , 1991 ) , the timing of spawning is thought to be regulated by the onset of darkness ( Babcock et al . , 1986 ) , the duration of regionally calm weather ( van Woesik , 2010 ) , food availability ( Fadlallah , 1981 ) , twilight chromaticity ( Sweeney et al . , 2011 ) , and salinity ( Jokiel , 1985 ) . How these exogenous factors function together with endogenous mechanisms in corals to achieve mass spawning harmonization is unknown .
To explore the interaction of exogenous and endogenous factors , we characterized the transcriptome from 16 Acropora millepora colonies over 8 days leading up to the spawning night , as well as during and after spawning ( Figure 1 ) . These results were compared with transcriptomes of four A . millepora colonies sampled 3 months prior to the month of spawning , during full and new moon days at midday and night ( 12:00 , 18:00 and 24:00 ) . A comparison of new and full moon samples , when spawning did not occur , revealed changes in gene expression according to the day within the month , and relative to levels of moonlight ( Figure 1—figure supplement 1 , Supplementary file 1 , 2 ) , thereby demonstrating a response by RNA transcription to the lunar cycle . Notably , many of the genes showing higher expression levels at midnight on full moon days compared with midnight on new moon days , were linked to rhythmic processes ( circadian clock related genes ) and light signalling , including cryptochrome 1 , cryptochrome 2 , and thyrotroph embryonic factor . 10 . 7554/eLife . 09991 . 003Figure 1 . Changes in gene expression during mass spawning of the coral , Acropora millepora . ( A ) A . millepora colony releasing gametes . ( B ) Schematic representing the sampling regime of coral branches from A . millepora colonies ( colonies were sampled for 8 days leading up to the spawning night , during the spawning night and 1 day following the spawning event ) , blue arrows indicate timing of experiment start , full moon and observed spawning . Additionally , colonies were sampled prior to the November spawning month ( the colonies were sampled at 12:00 , 18:00 and 24:00 in August on new moon and full moon days ) . ( C ) Hierarchical clustering of A . millepora gene expression data for the 184 coral transcripts that were only variable during the spawning day using sampling points in August and November . ‘New’ and ‘Full’ denote new moon and full moon conditions , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00310 . 7554/eLife . 09991 . 004Figure 1—figure supplement 1 . Hierarchical clustering of Acropora millepora gene expression during full moon versus new moon days at 12:00 , 18:00 and 00:00 . ‘New’ and ‘full’ refers to new moon and full moon days , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00410 . 7554/eLife . 09991 . 005Figure 1—figure supplement 2 . Hierarchical clustering of the expression patterns of genes that are both expressed and non-constant ( Materials and methods ) , indicates that a sizable number of genes are highly variable only on the spawning day . New and full denotes New moon and full moon conditions and A denotes ambient treatment . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 005 To elucidate the role of natural moonlight phases and the effect of ‘light pollution’ on mass spawning synchronization , 16 large A . millepora colonies were collected from the Heron Island reef flat ( GBR , Australia ) and transferred to large outdoor aquaria , which were exposed to natural sunlight , moonlight and flow-through seawater from the reef flat . Coral colonies were exposed to one of the following treatments: ambient ( maintained under ambient conditions of daylight and moonlight , N=6 ) , light ( illuminated using artificial light , low intensity ~ 5 µmole quanta m-2 s-1 daily from sunset to midnight , then kept in the dark until sunrise , N=5 ) , or dark ( covered with a black shade from sunset to sunrise to block moonlight , N=5 ) ( Figure 2A ) beginning 8 days prior to the spawning night . We sampled corals from all treatments during the days leading up to the spawning event at midday , at moonrise and at hourly intervals during the spawning night after sunset; this included sampling of released gametes ( Figure 1B ) . During the spawning night , colonies exposed to ambient conditions spawned in a similar manner to those on the reef ( gametes were released at approximately 21:30–22:30 ) ; however , no sign of spawning behavior occurred in either the light or dark treatments . For ambient corals , we identified 184 transcripts that were only variable during the spawning day and showed a clear change in expression ( mostly induction ) just prior to and at the time of gamete release ( Figure 1C ) ; however , such changes were not observed in corals that were kept in the dark ( Figure 2B ) . Colonies exposed to the light treatment showed a mismatch of gene expression profiles , and the expression of the genes that were highly variable during spawning occurred prematurely ( around sunset at 18:15 and 19:30 , Figure 2B ) . Moreover , these colonies did not release gametes , implying that ‘light pollution’ disrupted spawning synchronization . Gene enrichment analysis revealed that coral transcripts , which were only variable during spawning , were enriched in Gene Ontology ( GO ) processes of the cell cycle , GTPase activity and signal transduction ( Figure 3A ) . An analysis of transcripts that were highly up-regulated or down-regulated during spawning showed a larger proportion of transcripts ( 177 ) that increased their expression levels during spawning ( Figure 2—figure supplement 1A ) , a substantially smaller portion of 29 transcripts that were down-regulated ( Figure 2—figure supplement 1B ) and no correlation with gene expression profiles from the released gametes . Fifty four out of the 177 up-regulated genes , and none of the 29 down-regulated genes , were also found to be variable during the spawning day ( Supplementary file 1 ) . The up-regulated coral genes were enriched in G protein-coupled processes , signal transduction and respiratory processes ( Figure 3B ) . Down-regulated genes were enriched in processes which generate and maintain rhythms in the physiology of organisms ( rhythmic processes ) such as circadian clock related genes ( Figure 3C ) and included some additional genes , which were not clustered . A qPCR validation ( Supplementary file 3 ) of both up-regulated and down-regulated genes showed strong correlation between the qPCR and RNA-seq data based on log2 fold change measured on the spawning night in ambient conditions at 19:30 versus 22:00 ( r2=0 . 92 , p<0 . 0001 ) ( Figure 2—figure supplement 2 ) . 10 . 7554/eLife . 09991 . 006Figure 2 . Disrupting synchronized mass spawning in the coral , Acropora millepora . ( A ) Beginning 8 days prior to the spawning night , A . millepora colonies were exposed to one of the following treatments: ambient ( A ) , in which colonies were exposed to natural day and night cycles with full exposure to moon light; light ( L ) , in which colonies were exposed to natural daylight during the day and artificial photosynthetically active radiation ( PAR ) light ( ~5 µmol quanta m-2 s-1 ) at sunset every day for ~6 hrs ( between 18:15 and 24:00 ) and then left in the dark until sunrise; or dark ( D ) , in which colonies were exposed to natural daylight during the day and left in the dark from 18:15 to sunrise . ( B ) Hierarchical clustering of A . millepora gene expression data for the 184 coral transcripts that were only variable during the spawning night , depicting gene expression changes between treatments A , L and D denote ambient , light and dark treatments , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00610 . 7554/eLife . 09991 . 007Figure 2—figure supplement 1 . Hierarchical clustering of Acropora millepora gene expression data for ( A ) the 177 coral transcripts that were up-regulated and ( B ) the 29 coral transcripts that were down-regulated during the spawning night . A , L and D denote ambient , light and dark treatments , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00710 . 7554/eLife . 09991 . 008Figure 2—figure supplement 2 . Correlation of gene expression Log2 fold change between values obtained from RNA-seq analysis and expression values obtained using quantitative PCR ( qPCR ) . Twelve genes were arbitrarily chosen based on either high up- or down-regulation in corals sampled on the spawning night in ambient conditions at 22:00 versus ambient corals on the spawning night at 19:30 . Gene expression measured by qPCR and RNA-seq was closely correlated ( r2 = 0 . 92; p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00810 . 7554/eLife . 09991 . 009Figure 2—figure supplement 3 . The effect of light quantity and quality on the timing of broadcast spawning in the coral Acropora millepora at Heron Island . ( A ) Represents the percentage of corals that released gametes on the night of the spawning and the percentage of late spawners ( shaded area ) , as compared to the control corals and corals monitored at the field reef site . ( B ) Represents the percentages of colonies which released gametes on the night of the spawning November 13 , 2006 and the phase shift ( blue line ) that is number of hours in the timing of spawning from the control group due to the different light treatments . Total number of colonies was N =90 , one-way ANOVA followed by post hoc Tamhane test showed significant differences between the groups with regard to the phase shift in the spawning time and the light and color treatments , p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 00910 . 7554/eLife . 09991 . 010Figure 3 . Gene ontology enrichment analysis for Acropora millepora gene variability during mass spawning . Gene enrichments ( false discovery rate<0 . 1 ) across GO categories are shown . GOseq was used to test for enriched GO categories of genes that were ( A ) variable during spawning , ( B ) up-regulated during spawning or ( C ) down-regulated during spawning . The left and right identifiers represent the gene names and UniProt accession numbers , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 010 Since corals do not have specialized visual structures , light detection is likely mediated through photosensitive molecules such as opsins or cryptochromes ( Reitzel et al . , 2013 ) that help the corals to adapt and to be synchronized with the external irradiance levels ( Panda , 2002 ) . Furthermore , specific classes of G-proteins can be activated by opsins in response to light in coral larvae ( Mason et al . , 2012 ) . Neuropeptides are another diverse class of signalling molecules that act through G-protein coupled receptors . These peptides have been implicated in an array of biological processes such as reproduction , metabolism , feeding , circadian rhythms , adaptive behaviors and cognition ( Grimmelikhuijzen and Hauser , 2012 ) . The importance of neuropeptides in life phase transitions has been recorded across a range of taxa such as annelids , cnidarians , insects and mammals ( Schoofs and Beets , 2013 ) . Our results indicate that gamete release from a broadcast spawning coral is potentially controlled by a series of G-protein-coupled signalling pathways and that the trigger for release might be a non-visual ocular photoreceptor such as the melanopsin-like coral homolog and/or a range of neuropeptide candidates ( Figure 4 , Supplementary file 1 ) . Genes induced during spawning included two melanopsin ( opn4b ) -like homologs , a battery of G-protein-coupled receptors ( GPCRs ) , and a homolog of synaptotagmin 7 ( Figure 3 , Figure 4 , Supplementary file 1 ) . Melanopsin ( Opn4 ) has been shown to be required for light-induced circadian phase shifting in mammals ( Panda , 2002 ) . Given the broad changes in expression patterns in GPCRs and their potential triggers , it seems that neuropeptide signalling appears to be central in controlling gamete release in broadcast coral spawning , where such a simple nervous system likely evolved in an ancestor common to all cnidarians ( Grimmelikhuijzen and Hauser , 2012 ) . Our data are consistent with the activation of GPCR receptors and consequent GPCR signalling cascades followed by changes in cell migration ( eg , RAP2B , UNC5D , IF2B1 ) , immunity/cell death ( eg . TLR2 , TLR6 , TRAF5 , DAB2P , FAK1 , NLRC5 ) respiration ( NU4M , NU5M , COX1 , NDUA4 , CYB ) and cytoskeletal organization ( eg . PAXI , INF2 , NUP43 ) which together lead to synchronized gamete release ( Figure 4 ) . 10 . 7554/eLife . 09991 . 011Figure 4 . Proposed model for signalling events during spawning and gamete release in Acropora millepora . The release of gametes occurs in the presence of moonlight or another signal ( such as an olfactory signal ) that stimulates melanopsin-like homologs and/or other neuropeptides and commences GPCR signalling cascades upon receptor activation . This effect results in cell migration , follicle rupture , changes in immunity/cell death , respiration and cytoskeletal organization , and the combination of which results in a synchronized gamete release . Blue labels indicate genes being up-regulated; orange labels indicate genes down-regulated . DOI: http://dx . doi . org/10 . 7554/eLife . 09991 . 011 Our study does not provide direct evidence as to whether the melanopsin-like homologs that are up-regulated during spawning are true photopigments , as they have been shown to be in mammals ( Peirson et al . , 2009 ) . The sequential expression pattern of these melanopsin-like homologs , however , suggests that they are important for signalling events during mass spawning in corals regardless of whether they convey a light , pheromone or other signal . The discovery of potential neuropeptide/GPCR-coupled signalling mechanisms are consistent with neurohormones playing a role in synchronized spawning events in tropical abalone ( York et al . , 2012 ) and in the settlement behavior of coral larvae ( Grasso et al . , 2011 ) . Additionally , an increase in a tachykinin-like peptide receptor suggests that a pheromonal trigger may be involved ( Winther et al . , 2006 ) . Given that the transient receptor potential protein superfamily represents calcium permeable channels which have important roles in phototransduction in Drosophila and in mammalian vision ( Panda , 2002; Pan et al . , 2011 ) , our results showing an up-regulation of Trp-related protein 4 fits with the proposed hypothesis that cytoplasmic calcium may act as a secondary messenger for coral photoreceptors ( Hilton et al . , 2012 ) . Although the cues for the synchronization of gamete release in broadcast spawning corals are complex , our results indicate that nocturnal illumination is an important factor and that changes in the light regime surrounding corals can desynchronize the timing of gamete release . In this respect , only colonies exposed to ambient light conditions at night spawned . We further tested the concept of light regime changes causing a mismatch in the timing of gamete release through another experiment in which A . millepora colonies were exposed to treatments where the length of the day was extended artificially ( between sunset and midnight ) , through exposure to a suite of light quantity and quality regimes . We found that colonies exposed to ambient conditions and treatments with light sources in the red light region ( 620–700 nm ) of the spectrum spawned at 21:30 , which was synchronous with A . millepora colonies spawning in the field . However , colonies exposed to light in the green ( 500–620 nm ) and blue ( 400–500 nm ) regions of the spectrum and to PAR ( 400–700 nm ) at low-dim , medium and high light intensities phase shifted their spawning time by a 6–8 hr delay or spawned 2 nights later as compared with colonies under ambient conditions ( Figure 2—figure supplement 3 ) . Our results suggest that detection of light in the blue and green parts of the spectrum is important for gamete release , which is congruent with suggestions that blue light detection is a key element for coral spawning behavior ( Gorbunov and Falkowski , 2002 ) and evidence that A . millepora larvae settle more readily under blue and green light as compared with red light ( Strader et al . , 2015 ) . Furthermore , our results from both experiments ( Figure 2 , Figure 2—figure supplement 3 ) on the effect of ‘light pollution’ on coral spawning behavior suggest that disruption or phase shift delay in spawning time can occur rapidly , that is , within 7 days of exposure to changes in nocturnal light regimes . These findings differ from previous studies which suggested that entrainment to moonlight rhythms occurs over several months ( Willis et al . , 1985 ) . Our results on the effects of light on the timing of spawning are crucial because sexual reproduction is one of the most important processes for the persistence of reefs . The interplay between endogenous clocks and external cues in an era of industrialization and global change ( when artificial lights compete with moonlight to affect reproductive timing and fertilization success in broadcasting species ) should be considered in plans to protect coral reefs and marine ecosystems .
Twenty whole colonies of A . millepora were collected on November 9 , 2011 from the Heron Island reef flat ( 23 33’S , 151 54’E ) , GBR , Australia . Small branches were cut from the central portion of each colony prior to specimen collection to confirm pink-coloured eggs , which are present when a colony is reproductively mature ( Harrison et al . , 1984 ) . Four of the colonies were transported close to the shore but were left in the field on the reef flat ( treatment F ) . The remaining 16 colonies were transferred to large , outdoor flow-through aquaria and were exposed to natural sunlight , moonlight and flow-through seawater from the reef flat . Colonies in the outside aquaria were divided into the following three treatments: ambient ( A ) , in which the colonies ( N=6 ) were exposed to natural day and night cycles and full exposure to moonlight; light ( L ) , in which the colonies ( N=5 ) were exposed to natural daylight during the day and to artificial PAR light ( ~5 µmol quanta m-2 s-1 ) post sunset every day for ~6 hrs between 18:15 and 24:00 and then left in the dark until sunrise; and dark ( D ) , in which the colonies ( N=5 ) were exposed to natural daylight during the day and left in the dark between 18:15 and sunrise . The experiment was conducted at Heron Island Research Station in an area that was maintained in darkness at night to avoid artificial light contamination from non-experimental sources . Branches from colonies were collected from each of the three treatments on 4 d ( November 10 , 12 , 14 , and 15 , 2011 ) leading up to the spawning night ( November 16 , 2011 ) ( Figure 2A ) . The corals were sampled on these days at noon and during moonrise , which was between 21:00 and 23:00 depending on the day . On November 16 ( the spawning night ) , we sampled the corals at noon , 18:15 , 19:30 , 21:00 , 22:00 , 22:30 and 00:20 . The release of gametes occurred between 21:30 and 22:30 , and the colonies in the Ambient treatment began to show signs of setting at 19:30 . We also sampled released gametes from colonies in the Ambient treatment . Sampled coral branches were snap-frozen in liquid nitrogen and stored at -80°C until processing for total RNA extraction . Additional branches were sampled ( N=4 ) during August from the reef flat at a 2-m depth during new moon and full moon days at 12:00 , 18:00 , and 24:00 . Total RNA from the coral branches was isolated by homogenizing 100 mg coral tissue in 1 ml TRIzol ( Invitrogen ) according to the manufacturer’s instructions . The RNA was then extracted once with 1 volume chloroform and precipitated in ½ volume isopropanol , washed in 1 volume of 75% ethanol and subsequently dissolved in RNase-free water . These samples were then processed through a 5 M LiCl precipitation overnight at −20°C , washed three times with 75% ethanol and subsequently dissolved in RNase-free water . The integrity and quality of the total RNA was assessed using a Bioanalyzer ( Agilent Technology ) . Only the samples showing intact RNA ( RNA integrity number >8 ) were used for the RNA-seq analysis . The Illumina TruSeq protocol was used to prepare libraries from the RNA samples . Overall , 12 libraries of full moon and new moon samples were run on one lane and 24 libraries of samples from the spawning experiment were run on two additional lanes in the Illumina HiSeq2000 machine using the multiplexing strategy of the TruSeq protocol . On average , ~10 million single-end reads were obtained for each sample of the full moon and new moon conditions , and ~15 million paired-end reads were obtained for each sample in the spawning experiment . The sequencing data reported in this study was deposited to the Sequence Read Archive ( SRA ) , under accession SRP055723 . The reads were 100 bases long . TopHat ( Trapnell et al . , 2009 ) was used to align the reads against the A . millepora genome , keeping only uniquely aligned reads with up to two mismatches per read . Only reads that were aligned to the protein coding regions of the A . millepora genes ( Moya , 2012 ) were used . A custom script written in Perl was used to parse the output of TopHat ( given in the Sequence Alignment/Map ( SAM ) format [http://samtools . sourceforge . net/] ) and to convert it into the raw number of reads aligned to each A . millepora gene ( available via Dryad data repository: doi:10 . 5061/dryad . 541g6 ) . The A . millepora gene information was downloaded from the NCBI database , we retained only genes that were found to be significant similar ( Blastx E-value<1e-6 ) to the Swiss-Prot proteins data set . On average , ~40% of the reads in the full moon and new moon conditions and ~30% of the reads of the spawning conditions passed all filters and were mapped to the coding regions of the 12 , 384 A . millepora genes with Swiss-Prot homology . These numbers are comparable to the ones obtained by a similar analysis in a better annotated animal , the zebrafish , in which between 10% to 56% of the reads mapped to the coding regions of known genes ( Ben-Moshe at al . , 2014; Tovin et al . , 2012 ) . Statistically significant differences between the number of reads aligned to each A . millepora gene ( the expression profile ) in the tested conditions were identified as described ( Alon et al . , 2011 ) . Briefly , the expression profiles were normalized using a variation of the TMM normalization method ( Robinson and Oshlack , 2010 ) . Subsequently , we searched for expression differences between samples associated with the full versus new moon conditions that could not be explained by the expected Poisson noise with a p-value <0 . 05 and using a Bonferroni correction for multiple testing ( Alon et al . , 2011 ) . Each one of the full moon and new moon samples was sequenced twice ( technical duplicates ) , and as the correlation between the gene’s expression levels in these duplicates was very high ( average Pearson's correlation >0 . 999 ) the average expression level is presented in the figures . The data from the full moon and new moon experiments and the spawning experiment were combined and normalized as described above . We then performed a hierarchical clustering ( MATLAB , MathWorks ) of all of the genes that ( 1 ) had expression levels above the median of all of the gene expression levels in at least one sample and ( 2 ) had a variance in the expression level that was above the median variance of all of the genes . The gene's variance was normalized by the gene's mean expression level ( the expected variance assuming Poisson noise ) . Overall , 4756 genes out of the original list of 12 , 384 genes passed these two filters and were therefore considered to be both expressed and to have an expression profile that was not constant . Clustering of the expression patterns of these genes indicates that a sizable number of genes are highly variable only on the spawning day ( Figure 1—figure supplement 2 ) . These genes are characterized and identified below . The genes that were highly variable during the spawning day in ambient light conditions ( regular light-dark cycles ) , which ultimately led to the spawning event , could be responding to at least two factors: ( 1 ) moonlight ( either directly via light or indirectly through the time of the month ) or ( 2 ) the time of day . Alternatively , these genes can also be variable on other days , such as the days prior to or after the spawning day . Finally , these highly variable genes could also be part of the molecular mechanism of spawning . To identify the latter group , we selected all of the genes that were highly variable during the spawning day in ambient conditions but significantly less variable in the full moon and new moon conditions at different times of the day ( without spawning events ) and the days prior to the spawning events ( all under ambient conditions ) . Variability was estimated using a gene's variance divided by its expected variance assuming Poisson noise in the gene expression levels , which is equal to the gene's mean expression level . All of the gene variability measures were sorted , and we selected genes that were in the top 10% on the spawning day ( i . e . , highly variable ) but not in the top quartile ( i . e . , not as variable ) in the previous days and in the full moon and new moon experiments . We noted that changing the cut-offs produced similar results , for example by selecting genes that were in the top 5–25% on the spawning day but not in the top 25–50% in the previous days and in the full moon and new moon experiments . Additionally , we pursued high top expression levels ( more than 100 reads ) of these genes in any one of the mentioned conditions ( only one quarter of the genes showed this expression level or higher ) . We also identified genes with higher or lower expressions on the spawning day compared with the previous days and those of the full moon and new moon experiments . We selected for at least a twofold difference between the mean expression level on the spawning day and ( A ) the mean expression on the previous days , as well as ( B ) the mean expression in the full moon and new moon experiments . Thus , the difference in expression level had to be at least twofold in respect to both ( A ) and ( B ) . Additionally , we filtered for greater than 100 reads in the top expression level in the higher condition . The genes detected were analyzed using the software GOseq ( Young et al . , 2010 ) to find statistically significant over-represented functional groups with a Benjamini-Hochberg false discovery rate equal to or less than 0 . 1 . We compared the gene's expression pattern in the described conditions with the expression pattern of the released gametes that were sequenced , and found no significant correlation between the gene's expression patterns ( average Pearson's correlation of 0 . 22 between the released gametes and all the other described conditions ) . Specifically , for the genes detected above as having higher or lower expression on the spawning day , the correlation with the released gametes sample was even lower ( average Pearson's correlation of 0 . 05 between the released gametes and all the spawning day conditions ) . The expression patterns of the RNA-seq data for selected genes with variable expressions during spawning ( Supplementary file 1 ) were validated via quantitative PCR ( qPCR ) . Primers were designed from RNA-seq data using the Primer Express Software v3 . 0 ( Applied Biosystems , USA ) . Total RNA ( 1000 ng ) was reverse transcribed with a SuperScript VILO cDNA Synthesis Kit ( Invitrogen ) following the manufacturer’s instructions . Transcript levels were determined via qPCR assays using an Eppendorf 5075 ( Applied Biosystems , USA ) robot to dispense SYBR Green PCR master mix ( Applied Biosystems , UK ) into 384-well plates , and assays were run in a 7900HT Fast Real-time PCR System ( Applied Biosystems , USA ) . The PCR conditions included an initial denaturation for 10 min at 95°C , followed by 45 cycles of 95°C for 15 s and 60°C for 1 min . Finally , a dissociation step included 95°C for 2 min , 60°C for 15 s and 95°C for 15 s . The final reaction volume was 10 µl and included 300 nM of primers . All reactions were carried out with two technical replicates ( which showed no difference in expression levels ) . For each candidate gene sample ( from five replicates ) , the ambient colonies at 19:30 were tested against the ambient colonies at 22:00 on the spawning night . A no-template control and a no-reverse transcription control were performed for each gene and treatment to ensure that the cDNA samples and PCR reagents did not undergo DNA contamination . Additionally , to ensure the specificity of the primers for the coral cDNA , they were tested on cDNA and genomic DNA from Symbiodinium sp . as a template in a PCR . This procedure avoided any amplification of non-coral DNA . The comparative delta CT method ( Walker , 2002 ) and a maximal PCR efficiency for each gene ( E=2 ) were used to determine the relative quantities of mRNA transcripts from each sample . Each value was normalized to two reference genes , adenosylhomocysteinase ( AdoHcyase ) and ribosomal protein L7 ( Rpl7 ) . The selection of reference genes for this experiment was performed using a pool of reference genes ( Supplementary file 3 ) and analysing their expression stability via the geNorm software ( Vandesompele et al . , 2002 ) . For this study , the most stable expression was found for AdoHcyase and Rpl7 ( M value=0 . 225 ) , and a minimum of two reference genes was recommended ( V2/3=0 . 102 ) . The real-time dissociation curve was used to check for the presence of a unique PCR product . Relative expression values for each gene were calculated by showing a ratio of the relative expression at 22:00 over the relative expression at 19:30 . The results of the qPCR and RNA-seq analyses are presented on a log2 scale . A Welch t-test was used to test for significant differences in the expression levels ( obtained via quantitative real-time PCR ) of each gene at 19:30 compared with that at 22:00 ( Supplementary file 3 ) . Ninety healthy colonies of A . millepora ( which were checked for maturity as described above ) were transferred from the Heron Island reef flat and placed in the outdoor flow-through aquaria which was continuously flushed with seawater obtained from the Heron Island reef flat . Five colonies were placed into each aquarium , which were shaded during the day to simulate PAR levels at a 3-m depth . The coral colonies were acclimated to ambient seawater temperatures for 48 hrs prior to the experiment . At night , the shading was removed after sunset to expose the corals to natural moonlight . We simulated later ‘sunsets’ to determine if light intensity and light spectra can phase shift-spawning time . The coral colonies were divided between aquaria exposed to the light intensity treatments ( n=15 ) and light spectra treatments ( n=10 ) ; the coral colonies were divided into three aquaria for each of the seven treatments . The coral colonies were irradiated with artificial light every day for 6 hrs between 18:00 and 24:00 ( midnight ) . These colonies also received the ambient light/dark cycle during the day . We tested three light intensities ( 100 [‘high’] , 50 [‘medium’] and 0 . 75–1 [‘low-dim’] μmole quanta m-2 s-1 [using T8 fluorescent lamps] ) and three light spectra ( blue [400–500 nm , Lee filter model , deep blue 120 40% transmission] , green [500–620 nm , Lee filter model , dark green 124 60% transmission] and red [620–700 nm , Lee filter model , bright red 026 , 80% transmission] ) . The light intensity was adjusted to 1 μmol quanta m-2 s-1 for all three light spectra using a neutral density filter ( Lee filter model , neutral density 210 ( 0 . 6 ) 20% transmission ) when needed . The light was measured using a LI-COR ( LI-192S ) light meter . As a control , we maintained another set of corals that were exposed to ambient light/dark cycles . The corals were divided into four large tables; the first table included the control tanks , the second the ‘high’ light intensity treatment , the third the ‘medium’ light and the fourth table the ‘low’ light treatment corals . At the ends of each table , we added another four tanks into which were projected the blue , green and red light spectra . To avoid any light ‘leaks’ , black plastic screens were placed between the treatments . The experiment was conducted on a roof deck at the Heron Island Research Station to avoid artificial light ‘contamination’ and increase exposure to moonlight when the corals were not artificially irradiated . The experiment began on October 31 , 2006 and ended on November 15 , 2006 . The corals were monitored at night each day after the full moon ( November 5 , 2006 ) ; moonrise began at 17:53 , and moonset was at 04:22 ( November 6 ) . Each night , the corals were monitored between 18:00 and 03:00 or until spawning ceased . The times when the gamete bundles appeared in the polyp mouths ( ‘setting’ ) were recorded , as well as the times at which the first and last gamete bundles were released into the water column . The data were analyzed based on the timing of gamete release . Spawning in the field for the same species was observed via snorkeling or scuba diving . The major spawning on the reef and in our control aquaria occurred on the night of November 13 , 2006 , which was 8 nights after the full moon night . After the spawning event on November 13 , the corals were held in the aquaria and monitored for later spawning in the ambient light/dark cycles with no artificial light until November 15 . At the end of the experiment , the coral colonies were returned to the Heron Island reef flat . | Sexual reproduction in corals is possibly the most important process for replenishing degraded coral reefs . Most corals are “broadcast spawners” that reproduce by releasing their egg cells and sperm cells into the sea water surface . To maximize their chances of reproductive success , most coral in the Great Barrier Reef – over 130 species – spawn on the same night , during a time window that is approximately 30-60 minutes long . This is the largest-scale mass spawning event of coral in the world , and is triggered by changes in sea water temperature , tides , sunrise and sunset and by the intensity of the moonlight . How corals tune their spawning behavior with the phases of the moonlight was an unanswered question for decades . Now , Kaniewska , Alon et al . have exposed the coral Acropora millepora – which makes up part of the Great Barrier Reef – to different light treatments and sampled the corals before , during and after their spawning periods . This revealed that light causes changes to gene expression and signaling processes inside cells . These changes are specifically related to the release of egg and sperm cells , and occur only on the night of spawning . Furthermore , by exposing corals to light conditions that mimic artificial urban “light pollution” , Kaniewska , Alon et al . caused a mismatch in certain cellular signaling processes that prevented the corals from spawning . Reducing the exposure of corals to artificial lighting could therefore help to protect and regenerate coral reefs . Future work will involve comparing these results with information about a coral species from another part of the world to investigate whether there is a universal mechanism used by corals to control when they spawn . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"short",
"report"
] | 2015 | Signaling cascades and the importance of moonlight in coral broadcast mass spawning |
Serotonin ( 5-HT ) is associated with mood and motivation but the function of endogenous 5-HT remains controversial . Here , we studied the impact of phasic optogenetic activation of 5-HT neurons in mice over time scales from seconds to weeks . We found that activating dorsal raphe nucleus ( DRN ) 5-HT neurons induced a strong suppression of spontaneous locomotor behavior in the open field with rapid kinetics ( onset ≤1 s ) . Inhibition of locomotion was independent of measures of anxiety or motor impairment and could be overcome by strong motivational drive . Repetitive place-contingent pairing of activation caused neither place preference nor aversion . However , repeated 15 min daily stimulation caused a persistent increase in spontaneous locomotion to emerge over three weeks . These results show that 5-HT transients have strong and opposing short and long-term effects on motor behavior that appear to arise from effects on the underlying factors that motivate actions .
Serotonin ( 5-HT ) is a major neuromodulator that is one of the most important pharmacological targets in the treatment of psychiatric disorders such as anxiety and depression ( Vaswani et al . , 2003 ) . Yet 5-HT has also numerous other clinically-important actions on functions including eating ( Brewerton , 1995 ) , aggression ( Lesch and Merschdorf , 2000 ) , sex ( Rosen et al . , 1999 ) , pain ( Jung et al . , 1997 ) , and perception ( Geyer and Vollenweider , 2008 ) . However , the dominant conception of the 5-HT system still places mood and affect in a central position and considers other actions as side effects . To some degree different behavioral effects have been mapped to specific subsets of the 17-member receptor family , but a coherent pharmacologically-grounded theory of 5-HT function centered on affective states has been elusive ( Dayan and Huys , 2009 ) . Optogenetic tools provide novel viewpoints on neuromodulatory function by allowing the neurons that release a neuromodulator to be monitored and manipulated with high anatomical and genetic precision ( Yizhar et al . , 2011 ) . This allows a neuromodulatory system to be studied in terms of firing patterns with a temporal precision unavailable by pharmacological interventions . 5-HT neurons were classically thought to be regular-firing , ‘clock-like’ neurons ( Jacobs and Azmitia , 1992 ) . It is now known that , like other neuromodulatory neurons , 5-HT neurons exhibit phasic bursts of firing locked to significant behavioral events , such as stimuli , movements and rewards ( Ranade and Mainen , 2009; Cohen et al . , 2015; Liu et al . , 2014; Nakamura et al . , 2008 ) . Tonic and phasic activation of the dopamine ( Goto et al . , 2007; Niv et al . , 2007 ) and norepinephrine ( Aston-Jones and Cohen , 2005 ) systems have distinct functional consequences . Opposite functions for phasic and tonic 5-HT have been proposed as well ( Daw et al . , 2002 ) , but little is in fact known about the effects of phasic 5-HT firing . Optogenetic manipulations , which afford both genetic specificity and temporal control , might access phasic effects that are inaccessible by pharmacological manipulation ( e . g . , Tsai et al . , 2009 ) . Here , we used an optogenetic approach to target 5-HT neurons in the dorsal raphe nucleus ( DRN ) , the main source of 5-HT to the forebrain , for activation . Our starting point was experiments in the open field arena , a widely used assay for spontaneous behavior ( Hall , 1934; Seibenhener and Wooten , 2015 ) . A long-standing theory of 5-HT suggests that it is a mediator of ‘behavioral inhibition’ ( Soubrié , 1986 ) . This theory was motivated by data showing that 5-HT depletion increases startle responses ( Davis and Sheard , 1974; Davis et al . , 1980 ) and locomotor activity ( Gately et al . , 1985; Eagle et al . , 2009 ) . On the other hand , it has also been theorized , largely based on recordings of 5-HT neurons , that the primary function of the 5-HT system is instead the facilitation of rhythmic motor behaviors ( Jacobs and Fornal , 1993 ) . We found that phasic activation of DRN 5-HT neurons caused a rapid and striking decrease in locomotion , an effect of order 50% , while sparing other behaviors such as grooming . We then examined the detailed kinematics of locomotor gait using a linear track ( Machado et al . , 2015 ) . Surprisingly , in this assay we saw no effect of 5-HT activation , arguing against a motor hypothesis and suggesting a higher level underlying cause leading to motor effects . Indeed , 5-HT is strongly implicated in the control of impulsivity ( Dalley and Roiser , 2012 ) , suppressing actions because it helps avoiding future punishments ( Dayan and Huys , 2008 , 2009 ) or obtaining delayed rewards ( Miyazaki et al . , 2011a , 2011b , 2012 , 2014; Fonseca et al . , 2015 ) . In either case , serotonin is thought to drive behavioral inhibition by altering the impact of the future motivating outcome . Anxiety is a prominent motivation in the open field , and mice show strong center-avoidance or thigmotaxis ( reviewed in Prut and Belzung ( 2003 ) ) . Yet , we found effects on locomotion that were not accompanied by spatial biases and yielded no positive or negative effects on anxiety measures . It has also been theorized that 5-HT signals worse-than expected outcomes and might drive aversive learning ( Daw et al . , 2002; Dayan and Huys , 2009 ) . Conversely , 5-HT has also been proposed to signal reward ( Liu et al . , 2014 ) . To test if phasic 5-HT encodes or modulates reward or punishment , we repeatedly paired phasic 5-HT stimulation with a specific location within the open field , but we found neither appetitive or aversive place preference learning . Instead we found that slowing of movement could masquerade as a place preference depending on how occupancy was assessed . These findings support previous studies showing that DRN 5-HT activation does not itself cause reinforcement learning ( Miyazaki et al . , 2014; Fonseca et al . , 2015 ) . Unexpectedly , however , we found that repeated daily phasic 5-HT activation resulted in a long-term enhancement of locomotion , an effect similar in magnitude to the transient effect but opposite in sign . This observation recalls the two-to-three week window for selective serotonin re-uptake inhibitors ( SSRI ) therapeutic effects to develop ( Machado-Vieira et al . , 2010 ) and studies linking 5-HT activation to plasticity ( Santarelli et al . , 2003; Maya Vetencourt et al . , 2008 ) .
To activate DRN 5-HT neurons , we expressed the light-sensitive ion channel channelrhodopsin-2 ( ChR2 ) in DRN 5-HT neurons using an AAV2/9 viral vector ( AAV2/9-Dio-ChR2-EYFP ) injected into the DRN of SERT-Cre mice or wild-type littermate controls ( WT ) and implanted an optical fiber in the same location ( Figure 1A ) ( see Dugué et al . ( 2014 ) for more details ) . Experimenters were blind to the mice’s genotype throughout training and testing in this and all experiments below . Histological analysis performed at the end of testing confirmed ChR2-YFP expression was localized to the DRN in SERT-Cre animals ( Figure 1B ) and that there was no expression in WT controls ( data not shown ) . 10 . 7554/eLife . 20975 . 003Figure 1 . Optogenetic DRN 5-HT activation reduces spontaneous locomotion in the open field . ( A ) Schematic of the optogenetic approach . DRN neurons are infected with AAV2/9-Dio-ChR2-EYFP . In SERT-Cre mice , 5-HT neurons will express ChR2-YFP ( green cells ) and can be photoactivated with blue light delivered through an implanted optical fiber . ( B ) Fluorescence image of a parasagittal section showing ChR2-YFP expression ( green ) localized to the DRN . Scale bar , 500 μm . ( C ) Schematic drawing of the open field paradigm . ( D ) Schematic diagram of photostimulation protocol . Each 30 min session consisted of stimulated ( stim , blue ) and non-stimulated ( nostim , white ) blocks of 5 min . A session always starting with a non-stimulated block and blocks always alternated , for a total of 30 min . During stim blocks , 3 s pulse trains of light were delivered every 10 s . Pre and post intervals shown were used to calculate stimulation effects . ( E ) Probability of being in a specific behavioral state for non-stimulated ( pre ) and stimulated ( post ) intervals for the population of SERT-Cre mice ( N = 15 ) . Note that probabilities do not sum to 100% because scoring does not include all time points . Individual mice shown in grey lines and averages across mice in filled circles . Error bars indicate SEM . In some cases , the error bars are too small to be visible . n . s: not significant . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , with paired t-test . ( F ) Probability of being in a mobile state ( walking , rearing , jumping ) , as a function of time relative to stimulation onset . The shaded area indicates SEM across mice . ( G ) Same as ( F ) but for the immobile states ( resting , digging , grooming and scratching ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 003 We first tested mice in an open field assay in which the effect of stimulation on relatively spontaneous behaviors ( not driven by overt reward or punishment ) could be characterized ( Figure 1C ) . SERT-Cre and WT mice received blue light pulses in alternating 5 min blocks with 3 s light on , interleaved with 7 s light off , during stimulation blocks ( Figure 1D ) . To characterize the effects of DRN 5-HT activation we first performed a video based ethological analysis ( Choleris et al . , 2001; Video 1 ) . The most frequent behaviors observed in the open field were walking and rearing , with the remainder of time spent grooming , digging or otherwise awake but relatively stationary ( Figure 1E ) . Scratching and jumping were observed very infrequently and were not further analyzed . DRN 5-HT activation induced a profound decrease in locomotion , ( p=3 . 61 × 10−5 , paired t-test , SERT-Cre mice , N = 15 ) , Figure 1E ) and a robust increase in the probability of the ‘resting’ state ( p=4 . 75 × 10−3 , paired t-test , SERT-Cre mice , N = 15 ) , Figure 1E ) . Rearing was also reduced ( p=0 . 0495 ) , while digging and grooming were unaffected ( Figure 1E , digging , p=0 . 536; grooming , p=0 . 582; paired t-test , SERT-Cre mice , N = 15 ) . The ‘resting’ state did not correspond to freezing , as the animals did not exhibit a crouching position and often continued to make small movements ( Blanchard and Blanchard , 1969 ) , sometimes accompanied by a lowering of the head ( Video 1 ) . To summarize the dynamics of DRN 5-HT activation effects on behavioral state , we grouped the ‘mobile’ states ( walking , jumping , rearing ) and the relatively ‘immobile’ ( resting , grooming , digging , scratching ) and plotted their probability as a function of time relative to stimulation ( Figure 1F–G ) . This showed that the onset of DRN 5-HT activation effects were very rapid , peaking within 2 s and returning to baseline within 5 s ( Figure 1F–G ) . 10 . 7554/eLife . 20975 . 004Video 1 . Example SERT-Cre mouse behavior in the open field experiment during a photostimulation block . Also shown are the behavioral state ( red ) , photostimulation condition ( on/off , filled/empty blue square respectively ) , and speed ( green rectangle , normalized to the maximum value in that session ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 004 To better quantify these effects in a simple and robust manner , we extracted the mouse’s spatial position over time from video data using standard automated tracking methods , allowing us to obtain more precise measurements over a larger data set ( see Materials and methods ) . Optogenetic activation occurred heterogeneously in space , as illustrated by the position traces for representative WT and SERT-Cre mice ( Figure 2A ) . As suggested by the ethological analysis , light delivery in SERT-Cre , but not WT , mice resulted in a rapid decrease in movement speed , with >90% of the decrease being reached within <1 s ( Figure 2B ) . The reduction in average speed was clearly visible for every SERT-Cre mouse tested and highly robust at the group level ( 6 . 14 ± 0 . 31 cm·s-1 pre-stimulation and 2 . 94 ± 0 . 24 cm·s-1post-stimulation ( mean ± SEM ) ; p=4 . 80 × 10−7 , paired t-test , SERT-Cre mice , N = 15 mice , Figure 2C ) . The effect of stimulation could be seen as a shift in the distribution of movement speeds toward slower speeds ( Figure 2D ) . The difference between post- and pre-stimulation speed was also highly significantly different between SERT-Cre and WT animals ( t-test , p=7 . 45 × 10−7 , Figure 2E ) . The effect of optogenetic activation was dose-dependent over a range of lower stimulation frequencies ( 5 , 15 and 25 Hz; Pearson correlation coefficient , r = −0 . 837 , p=6 . 84 × 10−4 , Figure 2F ) . No significant effects were found in any measure for WT mice ( N = 9 ) . 10 . 7554/eLife . 20975 . 005Figure 2 . Optogenetic DRN 5-HT activation slows down animals in the open field , independently of previous locomotion speed . ( A ) Position tracking of example WT ( top ) and SERT-Cre ( bottom ) mice . All positions visited in the session are shown in gray; the positions visited during each 3 s stimulation periods are shown for one WT ( green ) and one SERT-Cre ( blue ) mouse . ( B ) Time course or speed relative to stimulation onset . Here , and below , WT ( N = 9 ) are shown in green and SERT-Cre mice ( N = 15 ) in blue and data mean ± SEM across animals is shown . Pre and post intervals are indicated by horizontal lines . ( C ) Average speed in pre- and post-stimulation intervals for individual mice ( gray lines ) and for the population of mice ( mean ± SEM ) . In some cases , the error bars are too small to be visible . n . s: not significant . ***p<0 . 001 with paired t test . ( D ) Probability distribution of speed in pre- and post-stimulation intervals for the population of WT and SERT-Cre mice . ( E ) Difference between speed in post- and pre-stimulation intervals ( delta speed ) *** , p<0 . 001 with two-sample t test . ( F ) Dependence of delta speed on frequency of stimulation for the individual mice ( gray lines ) and for the subset of SERT-Cre mice tested ( mean ± SEM , N = 3 ) . ( G ) Speed probability distribution in the pre interval for all stimulation periods within stimulated blocks ( as well as equivalent measures for non-stimulated blocks ) for an example SERT-Cre mouse . ( H ) Average post speed in stimulated blocks , conditioned by pre speed for SERT-Cre mice . The four colors indicate the speed ranges used for pre conditioning , as indicated in the distribution in ( G ) . ( I ) The same as ( H ) but for equivalent period in non-stimulated blocks . ( J ) The difference between the stimulated and non-stimulated blocks shows the effect of stimulation conditioned on pre speed . ( K ) Average difference between stimulated and non-stimulated blocks for delta speed ( difference between post- and pre-intervals ) for each quartile for individual mice ( gray lines ) and for the population of SERT-Cre mice ( mean ± SEM , N = 15 ) . In some cases , the error bars are too small to be visible . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 005 These results indicate that optogenetic activation of DRN 5-HT neurons rapidly switches animals away from engaging in active behaviors such as locomotion or rearing and suggest that they may slow the speed of locomotion . We next examined whether these effects were specific to the state of the animal at the onset of stimulation . We first tested whether the effects depend on the prior speed of the animal by conditioning the speed plots by the pre-stimulation speed ( Figure 2G ) . This conditioning procedure produces a tendency of the speed to revert back toward the mean , which can be seen even in non-stimulated blocks ( Figure 2H–I ) . We therefore calculated the stimulation effect by subtracting speed values in equivalent non-stimulation blocks from values in the stimulation blocks ( 'delta speed' , Figure 2J ) . There were significant effects on the average delta speed ( post-pre intervals ) for all pre-stimulation speeds ( Figure 2K ) , as indicated by comparing delta speed between non-stimulated and stimulated trials across different speed quartiles ( 2-way ANOVA , stimulation , prior speed; SERT-Cre mice , N = 15; stimulation , F ( 1 , 110 ) =123 , p=1 . 18 × 10−4; prior speed , F ( 3 , 110 ) =160 , p=6 . 53 × 10−40; stimulation x prior speed F ( 3 , 110 ) =18 . 3 , p=1 . 02 × 10−9 ) , followed by paired t-tests , with Bonferroni correction ( stimulation , p<0 . 05; prior speed quartiles 1–2 , n . s; other quartiles , p<0 . 05 ) . This analysis shows that decrease in speed produced by DRN 5-HT neuron activation was larger when the speed of the animal just prior to stimulation onset was faster . No significant changes were found for WT mice ( data not shown , N = 9 ) . This indicates that the effects of optogenetic activation of DRN 5-HT neurons were not limited to biasing animals against initiating movement , but also rapidly stopped mice even when they were already moving at relatively high speed . These observations suggest an unconditional effect on active motor behavior and indicate a possible impairment in motor coordination during active movements . To test for possible effects of optogenetic activation of DRN 5-HT neurons on motor output and coordination , we first used the accelerating rotarod assay ( Figure 3A ) , a widely used assay for motor coordination in rodents ( Carter et al . , 2001 ) . After allowing 2–3 weeks for virus expression , mice were trained for two consecutive days on the rotarod and on the third day DRN 5-HT neurons were optogentically activated ( 5 mW , 10 ms , 490 nm light pulses at 25 Hz ) in a randomly-selected 50% of trials , i . e . from placement of the mouse on the rotarod until the mouse fell ( Figure 3B–C ) . We found no difference in latency to fall between stimulated and non-stimulated trials for SERT-Cre mice ( p=0 . 743 , paired t-test , N = 7 ) or between WT and SERT-Cre mice ( stim . and non-stim . trials , p=0 . 513 , two-sample t-test , SERT-Cre mice , N = 7 , WT mice , N = 5 , Figure 3D ) . Thus optogenetic stimulation was without obvious effects on motor coordination assessed by this test . 10 . 7554/eLife . 20975 . 006Figure 3 . Optogenetic DRN 5-HT activation does not affect motor coordination in the rotarod assay . ( A ) Schematic of the accelerating rotarod assay . ( B ) Latency to fall for one example SERT-Cre mouse , with randomly interspersed stimulated trials . ( C ) Average latency to fall , showing learning and testing period . Note , stimulation ( cyan bar ) only occurred after training , WT ( green , n = 5 ) and SERT-Cre ( blue , N = 7 ) mice ( mean ± SEM ) are shown . ( D ) Average latency to fall in the testing session for individual mice ( gray lines ) and for the population of WT mice ( green , N = 5 ) and SERT-Cre mice ( blue , N = 7 , mean ± SEM ) . In some cases , the error bars are too small to be visible . n . s: not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 006 It is possible that the rotarod assay failed to reveal a more subtle effect on motor coordination . Therefore we next studied a subset of mice using a linear track assay with high-speed videography , allowing tracking the trajectory of the four paws , nose and tail ( LocoMouse system , Machado et al . , 2015 ) . In this assay , mice cross back and forth in a narrow brightly-lit corridor ( Figure 4A–B ) . Mice are motivated to cross the track by the tendency to avoid light and by water rewards they received at either side ( mice were mildly water-restricted ) . DRN 5-HT optogenetic activation was delivered starting when the animal entered the track and until it reached the other end ( stimulation parameters as in the rotarod assay ) . Surprisingly , despite the effects on locomotor behavior in the open field , we observed no effect of stimulation on speed , measured either as time to cross the track ( p=0 . 446 , paired t-test , SERT-Cre mice , N = 7 , Figure 4C ) or the speed ( p=0 . 387 , paired t-test , SERT-Cre mice , N = 7 , Figure 4D ) , for the population of SERT-Cre mice . We also ran the same mice in our standard open field protocol ( Figures 1–2 ) and found a similar magnitude of effects on speed as described above ( pre-speed 7 . 32 ± 0 . 73 cm·s-1 , post-speed 5 . 39 ± 0 . 65 , p=9 . 63 × 10−4 , paired t-test , SERT-Cre mice , N = 7 ) . 10 . 7554/eLife . 20975 . 007Figure 4 . Optogenetic DRN 5-HT activation does not induce motor impairment in the LocoMouse assay . ( A ) Schematic drawing of the LocoMouse apparatus . Water deprived animals walk freely across a glass corridor connected to two boxes with water ports . A mirror below at 45° angle allows a single high-speed camera to capture side and bottom views at 400 frames per second . DRN 5-HT photostimulation occured randomly in 50% of crossings . ( B ) Paws , nose and tail segments were automatically segmented and tracked in 3D . Individual strides were divided into swing and stance phases for further analysis . ( C ) Average time to cross the linear track for individual mice ( gray lines ) and for the population of SERT-Cre mice ( N = 7 , mean ± SEM ) in non-stimulated ( black ) and stimulated ( blue ) trials . In some cases , the error bars are too small to be visible . n . s: not significant . ( D ) Average whole-body speed ( center of mass ) for individual mice ( gray lines ) and for the population of SERT-Cre mice ( N = 7 , mean ± SEM ) in non-stimulated ( black ) and stimulated ( blue ) trials . In some cases , the error bars are too small to be visible . n . s: not significant . ( E ) Instantaneous forward speed of front-right paw during swing phase at stride speed of 15–20 cm·s-1 for stimulated ( blue ) and non-stimulated ( black ) crossings . ( F ) x-y position of four paws relative to the body center during swing . ( G ) Vertical ( z ) position of front-right paw relative to ground during swing . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 00710 . 7554/eLife . 20975 . 008Figure 4—figure supplement 1 . DRN 5-HT activation does not affect motor coordination and locomotion . ( A ) Polar plots indicating the phase of the step cycle in which each limb enters stance , aligned to stance onset of front-right paw ( FR , red ) , for the population of SERT-Cre mice ( mean , N = 7 ) . Distance from the origin represents walking speed . Left , non-stimulated trials; right , stimulated trials . ( B ) Nose trajectory in the z dimension , corresponding to height above the floor , for SERT-Cre mice ( N = 7 ) , conditional on walking speeds of 15–20 cm·s-1 . Dark lines and shading show mean ± SEM . ( C ) Same as ( B ) , but for tail segment eight trajectory along the y dimension . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 008 To test for more subtle possible effects , we next analyzed detailed locomotor kinematics ( Figure 4A–B ) . We performed single-limb gait analyses , characterizing continuous 3D paw trajectories in forward ( x-axis ) , side-to-side ( y-axis ) and vertical ( z-axis ) directions , for different speed intervals . We found no difference in peak swing velocity between stimulated and non-stimulated trials ( p=0 . 878 , paired t-test , SERT-Cre mice , N = 7 , Figure 4E ) . Neither the base of support between stimulated and non-stimulated trials ( p=0 . 586 , paired t-test , SERT-Cre mice , N = 7 , Figure 4F ) , nor the vertical ( z ) trajectory peak ( p=0 . 723 , paired t-test , SERT-Cre mice , N = 7 , Figure 4G ) were affected by stimulation . Polar plots indicating the phase of the step cycle in which each limb enters stance ( aligned to stance onset of the front right paw ) , showed no differences between DRN 5-HT activated and non-activated trials ( 3-way ANOVA with paw , speed and stimulation as main factors; stimulation: F ( 185 , 1 ) =0 . 209 , p=0 . 648 , SERT-Cre mice , N = 7 , Figure 4—figure supplement 1A ) . Furthermore , neither vertical ( z ) movement of the nose ( peak values ) ( p=0 . 641 , paired t-test , SERT-Cre mice , N = 7 , Figure 4—figure supplement 1B ) nor tail side-to-side oscillations ( y ) ( p=0 . 310 , paired t-test , SERT-Cre mice , N = 7 , Figure 4—figure supplement 1C ) trajectories were affected in stimulated trials . The results with the linear track suggest that DRN 5-HT optogenetic activation does not effect motor coordination or the detailed kinematics of locomotor behavior . However , this is secondary to the unanticipated finding that the primary effect on locomotor speed is highly context-dependent . This might be attributed to the strong drive to perform specific actions in those contexts . Therefore , we next turned to consider whether the effects on locomotion could be a consequence of specific affective or motivational states that might be induced by optogenetic activation of DRN 5-HT neurons . The open field assay has long been used as a screen for anxiety-related behavior in rodents ( Bailey and Crawley , 2009 ) . Thigmotaxis , the tendency to remain close to the walls and avoid the center of the arena , is a widely used measure of anxiety with ethological face validity ( Choleris et al . , 2001 ) . If the observed locomotor effects were secondary to an increase ( or decrease ) in fear or anxiety they should be accompanied by changes in thigmotaxis or by more subtle interaction between location and stimulation effects . We first characterized the effect of DRN 5-HT activation conditioned on the location of the mouse within the open field arena at the time of stimulation onset ( center , periphery and corners; Figure 5A ) . Mice spent more time in the corners and periphery and less in the center ( Prut and Belzung , 2003 ) and showed distinct speed probability distributions depending on their location within the arena , with higher speeds mainly being exhibited in the center ( Figure 5B–D ) . Nevertheless , DRN 5-HT activation caused a robust decrease in speed regardless of the spatial area where the animal was located at stimulation onset ( Figure 5B–D ) , as indicated by comparing speed between pre- and post-stimulation intervals across the different areas ( 2-way ANOVA , stimulation , areas; SERT-Cre mice , N = 15; stimulation , F ( 1 , 84 ) =250 , p=2 . 58 × 10−17; areas , F ( 2 , 84 ) =18 . 9 , p=3 . 99 × 10−4; stimulation x areas , F ( 2 , 84 ) =8 . 58 , p=0 . 0238 ) . The effect was confirmed with paired t-tests comparing speed pre- and post-stimulation ( center , p=6 . 19 × 10−5; periphery , p=2 . 50 × 10−7; corners p=6 . 08 × 10−7; SERT-Cre mice , N = 15 ) . Given the different behaviors and speeds in each area it is not surprising that there was a positive interaction between stimulation and areas factors . We found no significant effects between stimulated and non-stimulated trials across areas for WT animals ( data not shown , N = 9 ) . 10 . 7554/eLife . 20975 . 009Figure 5 . Effect of DRN 5-HT optogenetic activation does not induce anxiety-like behavior in the open field . ( A ) Position tracks of an example SERT-Cre mouse , depicting the main areas of the open field: corners ( dark pink ) , periphery ( light blue ) , center ( light pink ) and edges ( gray ) . Filled circles represent the position of the mouse at the beginning of each stimulus train . For this and subsequent panels , blue indicates stimulated blocks and black indicates equivalent times in non-stimulated blocks . ( B ) Speed probability distributions during pre and post stimulation intervals that began when the mouse was in the center area of the open field mice . For this and subsequent panels , data is pooled or averaged across all SERT-Cre mice ( N = 15 ) . ( C , D ) The same as ( B ) but for the periphery and corners areas . ( E ) Average speed across blocks within a session for individual mice ( gray lines ) and for the population of SERT-Cre mice ( N = 15 , mean ± SEM ) in non-stimulated ( black ) and stimulated ( blue ) blocks . In some cases , the error bars are too small to be visible . Each pair of stimulated and non-stimulated block was compared ( ***p<10−3 , paired t test with Bonferroni correction for multiple comparisons ) . ( F ) Fractional occupancy in each area of the open field , for the population of SERT-Cre mice ( N = 15 ) , color code as in ( A ) . ( G ) Fraction center area occupancy as a function of duration within the session . ( H ) Same as G but showing individual mice averaged over the entire session duration . ( I ) The cumulative distribution of the average distance from the geometric center point of the open field normalized to the distance of the walls to the center . ( J ) Fraction of the total distance travelled that was in the center area as function of time within the session . ( K ) Same as J but showing individual mice averaged over the entire session duration . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 009 Measures of anxiety may be sensitive to the amount of exposure to the environment , as animals become more accustomed to it . We next examined whether stimulation effects were dependent on the amount of time the animal was in the arena ( Figure 5F ) . Animals showed a tendency to move more slowly with time spent in the box over 30 min ( Figure 5E ) . Optogenetic activation caused a decrease in speed in SERT-Cre mice across all 5 min blocks ( Figure 5E ) as indicated by a 2-way ANOVA ( stimulation , blocks; SERT-Cre mice , N = 15; stimulation , F ( 1 , 84 ) =67 . 6 , p=0 . 0218; blocks , F ( 2 , 84 ) =4 . 36 , p=0 . 0158; stimulation x blocks , F ( 2 , 84 ) =1 . 18 , p=0 . 314 ) , followed by paired t-tests for comparison between non-stimulated and stimulated trials across blocks ( blocks 1–3 , p<0 . 001 ) . There were no significant effects for WT animals ( data not shown , N = 9 ) . The above results show that optogenetic activation of DRN 5-HT neurons is insensitive to the state of the animal when it occurs within the open field , but is somewhat more robust in its effects on rapid excursions ( Figure 2H–K ) that occur primarily in the center of the box ( Figure 5B ) . Taken together , these observations might suggest that DRN 5-HT activation could tend to decrease the frequency or duration of center excursions , increasing thigmotaxis . However , we observed no change in center occupancy between stimulated and non-stimulated blocks ( 2-way ANOVAs , with block and stimulation condition; stimulation , F ( 1 , 2 ) =0 . 611 , p=0 . 516; blocks , F ( 2 , 2 ) =3 . 08 , p=0 . 245 , SERT-Cre , N = 15; Figure 5G ) . A post-hoc comparison of total center occupancy between stimulated and non-stimulated blocks again showed no difference ( p=0 . 763 , paired t-test , Figure 5H ) . Similarly , we observed no effect in two other measures of thigmotaxis: distance from the center of the open field arena ( p>0 . 999 , Kolmogorov-Smirnov test comparing stim . and non-stim . distributions , Figure 5I ) and ratio of central vs . peripheral movement distance ( 2-way ANOVA , with block and stimulation condition; stimulation , F ( 1 , 2 ) =4 . 01 , p=0 . 183; blocks , F ( 2 , 2 ) =2 . 22 , p=0 . 311 , SERT-Cre , N = 15 , Figure 5J and post-hoc comparison of the ratio of central vs . peripheral distance between stimulated and non-stimulated blocks , p=0 . 139 , paired t-test , Figure 5K ) . No significant changes were observed in any measure for WT mice ( data not shown , N = 9 ) . Thus , photostimulation effects were both largely independent of the mouse’s location and while they biased action selection and speed , they did not bias mice toward or away from the center or periphery of the arena . These results show that DRN 5-HT activation does not affect thigmotaxis in the open field arena on the time scale of seconds to minutes . Therefore the effects of phasic optogenetic activation of these neurons do not appear to result from attenuation or enhancement of anxiety insofar as that can be measured by center avoidance . However , they do not exclude the possibility of longer-term effects . For example , some important affective and motivational effects of the dopamine system build up over longer periods and require contingency with environmental events , such as the conditioned place preference that can be produced by stimulating dopamine neurons contingent on an animal being in a certain location ( McDevitt et al . , 2014; Tsai et al . , 2009 ) . Indeed , at least one report has reported reinforcing effects of DRN 5-HT activation based on a ‘real time’ conditioned optogenetic stimulation assay ( Liu et al . , 2014 ) . In order to test for such effects , we ran an open field assay with photostimulation delivered contingent on the mouse entering a spatial region of interest ( ROI ) within the arena ( Figure 6 ) . Following Lui et al . ( 2014 ) , we defined the ROI as a sub-region of the center area ( Figure 6A ) ; animals had no visible cues for the ROI . On each day mice freely explored the arena for 15 min . After one habituation session , SERT-Cre ( N = 4 ) and WT ( N = 5 ) mice were subjected to three days of conditioning , with photostimulation being delivered during the entire excursion within the defined ROI ( 10 ms , 20 mW pulses at 25 Hz ) . They were then tested for one day without photostimulation . Note that we used this increased laser power and frequency of pulses in order to match the protocol of Liu et al . ( 2014 ) . We also ran a subset of mice in our standard open field protocol ( Figures 1–2 ) , in which stimulation occurs in random locations , using the same amplitude and found a similar magnitude of effects on speed ( pre-speed 7 . 4 ± 0 . 67 cm·s-1 , post-speed 4 . 0 ± 0 . 16 cm·s-1 , p=0 . 0103 , paired t-test , SERT-Cre mice , N = 4 ) . Therefore we believe the ROI and standard protocol are comparable except for the presence of spatial contingency of stimulation in the former . 10 . 7554/eLife . 20975 . 010Figure 6 . Optogenetic DRN 5-HT activation in a specific region of interest does not produce aversive or appetitive responses . ( A ) Position tracks for an example SERT-Cre mouse for sessions before ( pre ) , during ( T1 , T2 , T3 ) and after ( post ) photostimulation . Red square indicates the ROI in which photostimulation occurred ( indicted by blue lines ) . ( B , C ) Heat maps depicting the normalized occupancy and average speed for the same sessions depicted in A . ( D–G ) Population data ( mean ± SEM ) for SERT-Cre mice ( N = 4; above , blue are stimulated sessions ) or WT mice ( N = 5; below , green are stimulated sessions ) . Different measures as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 010 Given the decrease of speed induced by optogenetic activation of DRN 5-HT neurons in the standard protocol , we expected that animals would spend more time in the ROI simply as a consequence of slowing . Locomotion traces , occupancy and speed heat maps for all five days of the assay for a representative SERT-Cre mouse are shown in Figure 6A–C . Optogenetic activation of DRN 5-HT neurons produced a significant increase in ROI occupancy ( F ( 4 , 15 ) =5 . 56 , p=6 . 00 × 10−3 , 1-way ANOVA across days; Figure 6D ) . Consistent with a movement slowing effect , we found a decrease in the average speed inside the ROI ( F ( 4 , 15 ) =68 . 9 , p=1 . 80 × 10−9 , ANOVA; Figure 6G ) and the rate of exits from the ROI ( F ( 4 , 15 ) =26 . 6 , p=1 . 17 × 10−6 , ANOVA; Figure 6F ) across stimulated sessions for the population of SERT-Cre mice ( N = 4 ) . Post-hoc tests comparing the effect of each significant variable across sessions for SERT-Cre animals showed a significant increase in ROI occupancy by the third stimulated session and increases in rate of exits from the ROI and average speed in the ROI by the first stimulated session ( Figure 6D , F , G , p<0 . 05 , N = 4 , t-tests with Bonferroni correction ) . No significant effects were observed in WT mice ( N = 5 , Figure 6D–G bottom ) and no significant differences were found for average speed outside the ROI across sessions for either SERT-Cre or WT mice ( data not shown ) . These effects might be interpreted as either a place preference or as a stimulation-dependent reduction in movement speed . However , two additional observations argue for the latter interpretation . First , if stimulation reinforced the value of the ROI , we would expect mice to return more frequently to it . Contrary to this expectation , there was no change in the rate of entries into the ROI in stimulated mice ( F ( 4 , 15 ) =0 . 195 , p=0 . 937 , ANOVA; Figure 6E ) . The lack of change in rate of entries into the ROI despite the strong increase in occupancy indicates that mice do not approach or avoid the ROI , contrary to what is expected from either appetitive or aversive reinforcement learning . Second , if ROI stimulation induced a place preference , the effects should persist after pairing ( Liu et al . , 2014 ) . This was also not the case . ROI occupancy and rate of exits assayed on the day immediately after the pairing protocol returned to baseline levels ( p>0 . 05 , post-hoc t-tests , with Bonferroni correction , comparing pre and post day , N = 4 ) . Thus , the mice did not learn to prefer or avoid being in the ROI . Overall , these results appear to be consistent with the hypothesis that the entire effect of ROI stimulation ( Figure 6 ) is due to the same transient slowing effect seen in the standard open field assay in which stimulation is delivered in a spatially random fashion ( Figures 1–2 ) . Activation of DRN 5-HT neurons appears to promote ROI occupancy due to a decrease in movement speed within the ROI , resulting in a decrease in the rate of ROI exits . Interestingly , however , there appeared to be an increase in the effect of stimulation across days , as indicated by the fact that ROI occupancy increase did not reach significance before the third day of stimulation . In addition , there was a decrease in speed within the ROI between the pre- and post-stimulation sessions ( p<0 . 05 , post-hoc t-test , Figure 6G ) that was not seen in WT mice ( F ( 4 , 20 ) =0 . 183 , p=0 . 945 , 1-way ANOVA across days ) . Together , these differences suggested the possibility that in addition to transient effects of phasic stimulation there might be longer-term , stimulation-dependent accumulation of effects in SERT-Cre animals . In order to test for possible long-term effects of optogenetic activation of DRN 5-HT neurons and to characterize the dynamics of stimulation effects over this longer period , two sets of SERT-Cre and WT mice were tested in the open field arena for >3 weeks . Group 1 ( G1 , 3 SERT-Cre and 2 WT mice ) received DRN 5-HT photostimulation for 24 consecutive days ( Figure 7A–B ) . Group 2 ( G2 , 3 SERT-Cre and 2 WT mice ) were exposed to the arena for 23 days with no stimulation; for this group stimulation commenced on the 24th day and lasted for six consecutive days ( Figure 7B ) . 10 . 7554/eLife . 20975 . 011Figure 7 . Long-term optogenetic DRN 5-HT activation induces an increase in speed in the open field . ( A ) Schematic diagram of the photostimulation protocol , as in Figure 1D; a total of 270 s of 20 Hz stimulation is delivered over a 30 min session . ( B ) Experimental protocol . Group 1 ( G1 , 3 SERT-Cre and 2 WT mice ) received photostimulation from session 1 to 24 . Group 2 ( G2 , 3 SERT-Cre and 2 WT mice ) was exposed to the arena for 23 days , receiving photostimulation beginning only on the 24th day for six consecutive days . ( C ) Average speed across sessions for Group 1 ( N = 3 SERT-Cre mice ) . Non-stimulated ( pre , black ) and stimulated ( post , blue ) intervals . Lines and shading indicate mean ± SEM here and E , G , I , J . ( D ) Speed in early and late sessions ( as indicated in ( B ) ) for mice in ( C ) . Dark lines and error bars indicate ( mean ± SEM ) . Error bars are too small to see in some cases . Gray lines indicate individual mice . *p<0 . 05; n . s: not significant , with paired t test . Same applies to F , H , K , L . ( E , F ) Same as ( C , D ) but for Group 2 ( N = 3 SERT-Cre mice ) . ( G ) Group one stimulation effect ( delta speed , difference between post- and pre-stimulation intervals ) across all sessions . ( H ) Stimulation effect in early vs . late sessions . ( I ) Same as ( G ) but for Group 2 . ( J ) Group one speed across sessions assessed during the pre interval only in the first block , i . e . prior to receiving any stimulation during that session ( K ) Speed in the first block as in ( I ) for early vs . late sessions ( as indicated in ( B ) ) . ( L ) Center occupancy in early vs . late sessions for Group one and Group 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 01110 . 7554/eLife . 20975 . 012Figure 7—figure supplement 1 . Lack of correlation between short and long-term effects of DRN 5-HT activation . ( A ) Difference in speed during the pre-stimulation interval between sessions n+1 and n-1 is plotted against the stimulation effect on day n ( delta speed , difference between post- and pre-stimulation intervals ) for all mouse-session pairs ( excluding the first and last sessions; n = 66 ) . Red line indicates linear regression curve . R2=3 . 89 × 10−3 , p=0 . 270 against a constant model . ( B ) Residuals of a linear regression between pre-stimulation speeds and sessions are plotted against the stimulation effect ( delta speed , difference between post- and pre-stimulation intervals ) for all mouse-session pairs ( excluding the last session; n = 69 ) . Red line indicates linear regression curve . R2=0 . 030 , p=0 . 080 against a constant model . DOI: http://dx . doi . org/10 . 7554/eLife . 20975 . 012 Unexpectedly , we observed that the stimulated group ( G1 ) showed a progressive increase in movement speed over the course of 24 days , ( Figure 7C ) while the non-stimulated group ( G2 ) maintained a constant average speed over the same period ( Figure 7E ) . This long-term effect was , after 24 days , comparable in magnitude to the transient decreased evoked by stimulation , and manifested both during stimulation and non-stimulation epochs . The average speed increased in G1 from 3 . 50 ± 0 . 29 cm·s-1 ( day 1 ) to 7 . 11 ± 0 . 21 cm·s-1 ( day 24 ) during stimulation and 6 . 28 ± 0 . 74 cm·s-1 ( day 1 ) to 9 . 03 ± 0 . 36 cm·s-1 ( day 24 ) during control period . These effects were confirmed with a paired t-test , comparing pre-stim . speed in early ( 1–6 ) vs . late ( 19–24 ) sessions ( p=0 . 00210 for G1 , SERT-Cre mice , N = 3 , Figure 7D and p=0 . 736 for G2 , SERT-Cre mice , N = 3 , Figure 7F ) and by the difference in speed in the non-stimulated epochs between G1 and G2 mice ( p=0 . 0192 , paired t-test comparing pre-stim . speed in late sessions , SERT-Cre mice , N = 3 , Figure 7D , F ) . The results were further corroborated using generalized linear regression analysis ( see Materials and methods ) . We found that the slope of the regression , which corresponds to a long-term change in speed , was significantly higher than zero only in the mice that were photostimulated repeatedly throughout the experiment ( G1; fitted coefficient for session , −0 . 03 , p=0 . 180; intercept , 6 . 5 , p=6 . 51 × 10−42; group −0 . 59 , p=0 . 208; session X group interaction term , 0 . 15 , p=1 . 43 × 10−5 ) . Long-term effects were not likely due to changes in effectiveness of stimulation , such as an increase in expression levels , since the magnitude of the transient effect was unaltered over the course of the experiment for G1 ( delta speed in early vs . late sessions: −2 . 69 ± 0 . 23 vs . −2 . 50 ± 0 . 22 cm·s-1 , p=0 . 212 , paired t-test , SERT-Cre mice , N = 3 , Figure 7G–H ) . Moreover , even after the induction of the increase in speed in G1 , there was no difference in the magnitude of the transient decrease between G2 and G1 ( Figure 7C , E , I; difference in speed in pre- vs . post-stimulation epochs in the initial sessions of stimulation: 2-way ANOVA with groups and sessions as main factors , sessions , F ( 5 , 5 ) =0 . 317 , p=0 . 884; group , F ( 1 , 5 ) =4 . 10 , p=0 . 0991 ) . Overall , the transient and long-term effects appear to interact roughly additively . These results indicate that repeated phasic DRN 5-HT optogenetic activation induces two independent and opposite effects: a transient slowing and a sustained speeding of movement . We infer that the sustained effect is a long-term consequence of daily stimulation . However , we considered that the increase in speed might depend on the presence of a period of stimulation on a given day , i . e . , that it reflects a ‘rebound’ from stimulation . To test this we repeated the analysis using only the first 5 min block of each session . This block was never stimulated and any effects in this period could not be due to the effects of recent stimulation , but must rather reflect longer-term effects , at least carried over from the previous day . Indeed , we found a significant increase in speed even during the first non-stimulated block ( early vs . late sessions , p=0 . 0252 , paired t-test , SERT-Cre mice , N = 3 , Figure 7J–K ) . This suggests that the effect is present from the initial exposure of the animal to the area on a given session and does not depend on immediately preceding stimulation . Next , we considered whether the long-term effects are somehow related to an anxiolytic effects , as speeding of movement could result from an increase in center excursions , since those tend to be accompanied by a higher velocity ( Figure 5 ) . We found no increase in center occupancy after long-term stimulation ( early vs . late sessions , G1 , p=0 . 201 , paired t-test , SERT-Cre mice , N = 3; non-stimulated control G2 , p=0 . 384 , paired t-test , SERT-Cre mice , N = 3 , Figure 7L ) . Therefore , the sustained effects produced by 15 min per day of phasic DRN 5-HT activation , as the transient effects , do not appear to be secondary to anxiety-related effects measured by thigmotaxis . Finally , we tested whether session-by-session changes in the magnitude of the long-term effect were predicted by corresponding fluctuations in the short-term effect . However , our analysis failed to reveal such correlations ( see Materials and methods and Figure 7—figure supplement 1 ) , suggesting that the interaction between the two effects , to the extent that it exists , may require more sensitive behavioral assays to be resolved .
The main immediate effect of DRN 5-HT activation was to suppress locomotion . This effect was manifested first as a bias in action selection , with time locomoting and rearing reduced and time stationary enhanced ( Figure 1 ) . Non-locomotor movements such as grooming and scratching were unaffected and freezing was never observed . DRN 5-HT activation also decreased average speed during locomotion with the magnitude of the decrease positively correlated with the initial movement speed ( Figure 2 ) . These effects had rapid onset ( time constant of approximately 1 s ) and offset ( 3 s ) . Additional experiments showed that they were also strikingly context dependent . First , there was no effect on performance in the accelerating rotorod , a task that requires a kind of locomotor behavior ( Figure 3 ) . Second , there was no effect on mice crossing a brightly-lit linear track to obtain water rewards at either end ( Figure 4 ) . Therefore , suppression of locomotion does not appear to be a direct effect on the motor system , but must instead reflect modulation of factors that drive locomotion . The effects we observed recall long-standing observations that serotonin changes the effectiveness with which rewards and punishments drive behavior , enhancing the ability of organisms to tolerate delays before acting on such motivations ( Soubrié , 1986 ) . Such ‘behavioral inhibition’ has been studied extensively in tasks in which animals must refrain from responding for a delayed reward , including the five choice serial reaction time task ( Harrison et al . , 1997; Winstanley et al . , 2004b ) , go/no-go tasks ( Harrison et al . , 1999 ) and waiting tasks ( Miyazaki et al . , 2014; Fonseca et al . , 2015 ) . The effects of 5-HT appear to be largest when forces for acting and refraining from acting are balanced ( Soubrié , 1986 ) . In the open field , there are no overt rewards or punishments motivating behavior . One could infer from the dramatic effects of 5-HT in this context that a normally balanced set of covert factors is shifted toward inaction by 5-HT . Conversely , the lack of effect in linear track or rotorod could reflect the inability of 5-HT to overcome strong local drives present in these contexts ( obtaining water rewards and avoiding exposure to bright light in the linear track; avoiding falling in the rotarod ) . In the above interpretation , 5-HT would be affecting the manner in which motivating factors such as goals drive actions . Interestingly , goal-directed actions become habitual with extended practice , after which they are no longer driven by the goals , but shift to being stimulus responses , with a corresponding change in neural substrates ( Yin et al . , 2004 , 2005 ) . While 5-HT affects reward waiting times it does not affect stimulus reaction times ( Fonseca et al . , 2015 ) . Therefore , while 5-HT appears to modulate goal-directed actions ( Bailey et al . , 2016 ) it might not affect habitual actions . Although this hypothesis remains speculative , it would be consistent with the observation of DRN 5-HT activation strongly affected behavior in the open field , in which stimuli-response habits are absent , but failed to affect behavior in the rotarod and linear track , both of which might have become stimulus-response habits . Further experiments would be required to test this . The selectivity of DRN 5-HT activation effects for certain types of behavior within the open field can also be interpreted in a similar fashion . Whereas locomotion and rearing are considered ‘voluntary’ ( or type I ) behaviors , grooming and scratching are considered ‘involuntary’ ( or type II ) behaviors ( Vanderwolf , 1988 ) . 5-HT has been linked to electroencephalographic recordings associated with type I but not type II behaviors ( Vanderwolf and Baker , 1986; Vanderwolf , 1989 ) . This is consistent with our observation that type I behaviors ( i . e . locomotion ) are modulated by DRN stimulation , whereas type II behaviors ( e . g . grooming ) seem to be unaffected by it . Our data is consistent with the idea that 5-HT suppresses locomotor activity by decreasing underlying factors that promote it . These data do not reveal exactly how 5-HT achieves this , but do offer some constraints . 5-HT has been classically associated with anxiety and the thigmotaxis ( center avoidance ) in the open field has been widely used to assay anxiolytic and anxiogenic effects ( reviewed in Prut and Belzung , 2003 ) . 5-HT could increase the impact of threats that favor immobility over exploration . However , we observed no effects on thigmotaxis ( center avoidance ) ( Figure 5 ) , indicating that movement speed effects are not consequences of effects of 5-HT on anxiety but rather direct and independent effects . Our findings are consistent with a recent study showing optogenetic activation of DRN 5-HT neurons decreases locomotor activity without affecting anxiety in the elevated plus maze test ( Ohmura et al . , 2014 ) . Another recent study observed a transient increase in speed in the open field with DRN stimulation ( Warden et al . , 2012 ) . However in that study , both 5-HT and GABAergic neurons in the DRN were targeted , raising the possibility that the protocol was effectively inhibiting DRN 5-HT neurons by driving GABAergic neurons . Two recent studies using pharmacogenetic methods reported an increase in thigmotaxis in the open field following activation of 5-HT neurons ( Teissier et al . , 2015; Urban et al . , 2016 ) , and one of them ( Teissier et al . , 2015 ) also reported an overall reduction in locomotion . The additional effects observed could be due to the much slower kinetics of pharmacogenetic methods , which could result in a very different profile of 5-HT transmitter release . The anxiolytic effects of 5-HT appear to be mediated largely by 5-HT1a receptors . It is possible that Ohmura et al . and the present study did not observe these effects because they are not effectively activated by the optogenetic stimulation protocols employed . The 5-HT2a and 5-HT2c receptors have been associated with the effects of 5-HT on impulsivity ( Winstanley et al . , 2004a ) and would be candidates to underlie slowing in the open field effects as well . We found that DRN 5-HT optogenetic activation does not induce anxiety-like behaviors in the open field . However , we do not exclude the possibility that specific sub-populations of 5-HT neurons may modulate anxiety-like behaviors . Marcinkiewcz et al . ( 2016 ) showed that optogenetic activation of BNST-projecting DRN 5-HT neurons increase anxiety-like behaviors without affecting locomotion . They did not test the effects of overall DRN 5-HT stimulation . The pathway-specific effects might therefore represent an ‘off-target’ effect of this pathway ( Otchy et al . , 2015 ) that is not present when a larger portion of the DRN is stimulated , because , it is balanced by an opposite effect through a different DRN output pathway . More experiments are also necessary to identify pathway ( s ) involved in the DRN 5-HT reduction in locomotion . Activation of DRN neurons has in some reports been associated with reward or appetitive reinforcement ( Liu et al . , 2014 ) , while other reports have failed to see reinforcing effects ( Miyazaki et al . , 2014; Fonseca et al . , 2015 ) ; reviewed in Luo et al . , 2015 ) . Here , we revisited this issue using a task nearly identical to that reported previously ( Liu et al . , 2014 ) , delivering DRN 5-HT stimulation repeatedly continent on the location of the animal within a specific ROI ( Figure 6 ) . We did observe an increase in the occupancy during stimulated sessions , but this could be accounted by the decrease in speed within the ROI , consistent with inhibition of locomotion . Slowing was in turn associated with a decrease in rate of exits from the ROI . In contrast , we observed no change in the rate of entries into the ROI , i . e . , mice showed no evidence of increased tendency to approach or avoid the ROI after pairing . Furthermore , ROI occupancy returned to baseline when stimulation ceased . Thus , mice showed no evidence of learning about the ROI . The increase in ROI occupancy could thus be accounted for solely by direct locomotor effects . It is not clear whether our observations are entirely consistent with those of the previous report , in which ROI occupancy was reported to remain high after cessation of stimulation and entry and exit rates were not reported ( Liu et al . , 2014 ) . Differences between the two sets of results might instead be explained by differences in the populations of DRN neurons stimulated . Whereas we used a SERT-Cre line ( Gong et al . , 2007 ) to drive ChR2 expression in 5-HT neurons , Liu et al . used the ePet1-Cre line ( Scott et al . , 2005 ) . The Pet-1 line has a broader expression profile , including a 5-HT-negative , glutamate-positive population ( Liu et al . , 2014; Luo et al . , 2015 ) . DRN reinforcing effects appear to depend on such a glutamatergic input to the VTA ( Liu et al . , 2014; McDevitt et al . , 2014; Qi et al . , 2014 ) which appears to be absent in the more selective SERT-Cre line ( Fonseca et al . , 2015 ) . Taking our study and other open field results together with experiments using operant training ( Fonseca et al . , 2015 ) , we believe the evidence currently favors the view that activation of the DRN 5-HT system is not reinforcing . Thus , 5-HT may mirror the DA system in having an opposite direct effect on behavior ( devigorating vs . invigorating ) , but it appears to have neither the same ( Liu et al . , 2014 ) nor opposite ( Daw et al . , 2002 ) learning effects . The above observations suggest that activating DRN 5-HT neurons does not affect locomotion by producing a reward- or anxiety-like signal . There remain a variety of possible underlying signals that might be reported by DRN 5-HT neurons and that could result in inhibition of locomotion . Broadly , these could be divided into three domains . A first set of proposals , developed largely in the domain of reinforcement learning theory , posits 5-HT to affect action selection and response vigor by biasing the value of appetitive or aversive stimuli ( Dayan and Huys , 2009; Cools et al . , 2011 ) . A second set of ideas posits a signal that modulates the cost of acting ( or not-acting ) itself ( Miyazaki et al . , 2014; Fonseca et al . , 2015; Meyniel et al . , 2016 ) . A third set of ideas is related to the signaling of uncertainty , either in predicting events or the outcomes of ones own actions ( Maswood et al . , 1998; Amat et al . , 2005; Matias et al . , 2016 ) . Uncertainty-related signals may be more complex to interpret as they could bias benefits and costs toward either action or inaction depending on the shape of the value landscape or could shift the balance between different behavioral control systems ( Daw et al . , 2005 ) . There are a number of complexities in relating the firing patterns of DRN 5-HT neurons to the present results . First , other neuromodulators , particularly dopamine and norepinephrine , are also implicated in the same functions ( Yu and Dayan , 2005; Dayan and Yu , 2006 ) and 5-HT has a variety of features that suggest it acts as an opponent to dopamine ( Daw et al . , 2002; Cools et al . , 2011; Boureau and Dayan , 2011 ) . Our data are consistent with opponent effects of 5-HT and DA as modulators of action vigor , but not as reinforcement learning signals . Simultaneous or parallel recordings from multiple neuromodulator systems will be important for resolving the interplay between these systems ( Cohen et al . , 2012 , 2015; Matias et al . , 2016 ) . An additional complexity is that , due to differences in receptor kinetics , phasic firing and tonic firing of 5-HT neurons may convey different signals ( Daw et al . , 2002; Cools et al . , 2011; Cohen et al . , 2015 ) . Because of the synchronous nature of optogenetic stimulation and the rapid onset of the transient effects , we presume that our protocol mimicked primarily phasic signaling , but we cannot exclude the possibility that it also mimics tonic signals as well . This issue may underlie some of the discrepancies between different pharmacological and optogenetic results . Finally , recordings from DRN 5-HT neurons have indicated substantial diversity in firing patterns ( Nakamura et al . , 2008; Ranade and Mainen , 2009; Cohen et al . , 2015 ) . This might reflect more heterogeneous functions within the DRN that are not discriminated by global DRN stimulation . We found that repeated daily exposure to optogenetic DRN 5-HT stimulation produced an unanticipated long-term effect , progressively and persistently increasing locomotion over three weeks ( Figure 7 ) . These long-term effects were not observed in control animals that were exposed to the same tests but with stimulation withheld until after three weeks . The transient and long-term effects appeared to be opposites and interacted additively . Although further experiments will be needed to address mechanisms , this opponency suggests the possibility of a direct form of compensatory action at a physiological or biochemical level . If the effects of transient DRN 5-HT activation are interpreted as suppression of the influence of motivating factors on locomotion , as suggested above , the long-term effects could represent an enhanced influence of the same motivating factors . The ability of goals to motivate behavior is critical for healthy mental function , and its absence , apathy , is a clinically important factor in many psychiatric disorders ( Chase , 2011 ) . Although our stimulation protocol was artificial , it was carried out at a rate within the physiological range of 5-HT neurons ( Nakamura et al . , 2008; Ranade and Mainen , 2009; Cohen et al . , 2015 ) and would likely produce only a modest increase in overall DRN 5-HT activity . For example , if each optogenetic pulse caused a spike , a neuron with a 1 Hz average firing rate would experience <10% increase in total spikes over a 24 hr period . Therefore , the long-term effects might be ethologically or therapeutically relevant and long-term DRN 5-HT stimulation in the open field could be an interesting paradigm to explore in the context of animals models of affective disorders ( Markou et al . , 2013 ) . Because the persistent effects of DRN 5-HT activation accumulated gradually over many days and persisted in the absence of stimulation , we interpret them as a form of long-term plasticity . 5-HT has been described to boost plasticity in developing sensory cortices ( Gu and Singer , 1995; Maya Vetencourt et al . , 2008 , 2011; Jitsuki et al . , 2011 ) , possibly modulating the induction of LTP and LTD ( Kojic et al . , 1997; He et al . , 2015 ) . Notably , SSRIs must be administered for several weeks before anti-depressant effects can be seen , suggesting that anti-depressant effects are not a direct and immediate consequence of changes in 5-HT function but the consequence of an adaptation requiring induction of plasticity ( Branchi , 2011 ) or neurogenesis ( Miller and Hen , 2015 ) . The molecular and temporal specificity afforded by optogenetic access will facilitate the determination of the mechanism of action of the ( short and ) long-term effects of DRN 5-HT activation . For example , more refined targeting strategies , e . g . retrograde infection ( Rothermel et al . , 2013 ) or intersectional genetics ( Jensen et al . , 2008 ) , will allow the determination of the pathways involved . This may provide a new window on 5-HT-dependent plasticity .
Thirty-one adult C57BL/6 mice ( 19 SERT-Cre mice and 12 wild-type ( WT ) littermates ) were used in this study . A subset of these was used in the long-term open field experiments ( 6 SERT-Cre , 4 WT ) , in the accelerating rotarod ( 7 SERT-Cre , 5 WT ) and in the LocoMouse assay ( 7 SERT-Cre ) . All procedures were reviewed and performed in accordance with the Champalimaud Centre for the Unknown Ethics Committee guidelines , and approved by the Portuguese Veterinary General Board ( Direcção-Geral de Veterinária , approval 0421/000/000/2016 ) . The SERT protein is encoded by the Slc6a4 gene and the SERT-Cre mouse line ( Gong et al . , 2007 ) was obtained from the Mutant Mouse Regional Resource Centers ( stock number: 017260-UCD ) . Male and female mice ( 18–26 g ) were group-housed prior to surgery and individually housed post-surgery and kept under a normal 12 hr light/dark cycle ( tested at light phase ) . Mice had free access to food and water , except seven mice used in the open field test ( 4 SERT and 3 WT ) and the seven SERT-Cre mice used in the LocoMouse assay were under a mild water deprivation protocol . In the LocoMouse experiment , water availability was restricted to the behavioral sessions . Extra water was provided if needed to ensure that mice maintained no less than 85% of their original weight . The open field experiments were run in separate batches using the following order: 1 SERT-Cre , 4 SERT-Cre and 2 WT , 4 SERT-Cre and 3 WT , 3 SERT-Cre and 2 WT , 3 SERT-Cre and 2 WT . The rotarod and LocoMouse experiments were run in two batches of mice: 3 SERT-Cre and 2 WT , followed by 4 SERT-Cre and 3 WT . The surgery procedure is described in more detail at Bio-protocol ( Correia et al . , 2017 ) . For virus injection and cannula implantation , mice were first anesthetized with isoflurane ( 4% induction and 0 . 5–1% for maintenance ) and placed in a stereotaxic frame ( David Kopf Instruments , Tujunga , CA ) . Lidocaine ( 2% ) was injected subcutaneously before incising the scalp . The skull was covered with a layer of Super Bond C and B ( Morita , Kyoto , Japan ) to help stabilization of the implant . A craniotomy was drilled over lobule 4/5 of the cerebellum and a pipette filled with a viral solution ( AAV2 . 9 . EF1a . DIO . hChR2 ( H134R ) -eYFP . WPRE . hGH , 1013 GC/mL , University of Pennsylvania ) was lowered to the DRN ( Bregma −4 . 7 AP , −2 . 9 DV ) with a 32–33° angle toward the back of the animal . The viral solution ( 1 µL ) was injected using a Picospritzer II ( Parker ) or an hydraulic pump ( UMP3-1 , World Precision Instruments , Sarasota , FL ) , connected to a 5 μl Hamilton syringe ( Hamilton , Reno , NV ) , at a rate of 0 . 05–0 . 1 µL/min . An optical fiber ( 200 μm core diameter , 0 . 48 NA , 4–5 mm ) housed inside a connectorized implant ( M3 , Doric lenses , Quebec , Canada ) was lowered into the brain , through the same craniotomy as the viral injection , and positioned 200 μm above the injection point . The implant was cemented to the skull using dental acrylic ( Pi-Ku-Plast HP 36 , Bredent , Senden , Germany ) . The skin was stitched at the front and rear of the implant . Mice were monitored until recovery from the surgery and returned to their home cage . Gentamicin ( 48760 , Sigma-Aldrich , St . Louis , MO ) was topically applied around the implant . Behavioral testing started at least two weeks after virus injection to ensure good levels of expression . Previous studies using the same method reported that 94% of ChR2-YFP positive neurons were serotonergic , assessed with tryptophan hydroxylase immunohistochemistry ( Dugué et al . , 2014 ) . Light from a 473 nm laser ( LRS-0473-PFF-00800–03 , Laserglow Technologies , Toronto , Canada or DHOM-M-473–200 , UltraLasers , Inc . , Newmarket , Canada ) was controlled by an acousto-optical modulator ( AOM; MTS110-A1-VIS or MTS110-A3- VIS , AA optoelectronic , Orsay , France ) , except for the LocoMouse assay ( controlled directly with custom written software using LabView ) . The AOM controlled the laser power without any auditory noise and it was triggered by the behavioral control system ( Bcontrol ) , developed by Carlos Brody ( Princeton University ) in collaboration with Calin Culianu , Tony Zador ( Cold Spring Harbor Laboratory , Cold Spring Harbor , NY ) and Z . F . M . Light exiting the AOM was collected ( KT110/M , Thorlabs , Newton , NJ ) into an optical fiber patchcord ( 200 μm , 0 . 22 NA , Doric lenses ) , connected to a second fiber patchcord through a rotary joint ( FRJ 1 × 1 , Doric lenses ) , then to a chronically implanted optic fiber cannula through an M3 connector ( Doric lenses ) . Laser power was calibrated using a powermeter ( PM130D , Thorlabs ) before and after each animal session . The optical fiber patchord was screwed to the M3 implanted connector at the beginning of each experiment . Histological analysis were performed after photostimulation experiments to confirm viral expression of ChR2-YFP and optical fiber placement . Mice were deeply anesthetized with pentobarbital ( Eutasil , CEVA Sante Animale , Libourne , France ) and perfused transcardially with 4% paraformaldehyde ( P6148 , Sigma-Aldrich ) . The brain was removed from the skull , stored in 4% paraformaldehyde overnight and kept in cryoprotectant solution ( PBS in 30% sucrose ) for one week . Sagittal sections ( 50 μm ) were cut in a cryostat ( CM3050S , Leica , Germany ) , mounted on glass slides with mowiol mounting medium ( 81381 , Sigma-Aldrich , St . Louis , MO ) . Scanning images for YFP , RFP and transmitted light were acquired with an upright fluorescence microscope ( Axio Imager M2 , Zeiss , Oberkochen , Germany ) equipped with a digital CCD camera ( AxioCam MRm , Zeiss ) with a 5X or 10X objective . All data analysis was performed with custom-written software using MATLAB ( Mathworks , Natick , MA ) . Error bars represent standard error of the mean ( SEM ) . For the open field and ROI stimulation experiments , the animal’s path was recorded by an automated tracking system at 60 fps ( Bonsai , Lopes et al . , 2015 ) . Speed data was smoothed by applying a five-frame median filter . To compare behavior immediately before and after photostimulation we defined 'pre' and 'post' time intervals , respectively . Pre started 2 s before stimulation onset and post 1 s after onset; both bins were 2 s long . For the ethological characterization of behavior in the open field ( Figure 1E–G ) , a trained observer , blind to the experimental condition of the mice , recorded several behavioral states from the video with a custom-written software using Bonsai ( Lopes et al . , 2015 ) . The mobile states included walking ( straight locomotor activity ) , rearing ( mouse with both forepaws off the floor ) and jumping ( mouse with all four paws off the floor ) . The immobile states included resting ( non-locomotor activity that did not include any of the other states ) , digging ( mouse using the forepaws to move the bedding ) , grooming ( rapid cleaning movements of the forepaws towards the face and/or the body ) and scratching ( very rapid and repeated up and down movements of the hind paws on the side of the body , neck or face ) . For the LocoMouse assay , a detailed description of the tracker system can be found in Machado et al . ( 2015 ) . Briefly , the algorithm’s output consisted in 3D trajectories ( x , y , z ) of the four paws , the snout , and the tail ( separated into 15 points ) along the trial . All tracks were visually examined and fewer than 10% of trails were removed due to exploratory behaviors . To quantify gait parameters , the trajectories of the individual paws where divided into stride cycles that are composed by swing and stance phase . All strides were sorted into speed bins ( 5 cm/s bin width ) with a minimum of 5 strides per bin , per animal . The bin 15–20 cm/s was selected for presentation , as it included the higher number of trials , for both non-stimulated and stimulated trials . To compare pre- and post-stimulation speed intervals ( Figure 1E; Figure 2C; Figure 3D; Figure 4C–D , E–G; Figure 5B–D , H , K ) , no stimulation vs . stimulation trials ( Figure 3D; Figure 4C–G; Figure 4—figure supplement 1; Figure 5E , H , K ) and speed in early vs . late sessions ( Figure 7D , F , H , K , L ) we used a paired t test ( ttest , MATLAB ) . Comparison between independent groups of mice ( SERT-Cre vs . WT ) was done with a two-sample t test ( ttest2 , MATLAB ) . To compare delta speed across quartiles ( Figure 2K ) , areas ( Figure 5B–D ) and blocks ( Figure 5E ) we performed a two-way ANOVA ( anova2 , MATLAB ) , followed by post-hoc t-tests with Bonferroni correction ( multcompare , MATLAB ) . A two-way ANOVA was further used for comparison of between stimulated and non-stimulated trials across blocks ( Figure 5G , J ) and the pre vs . post speed intervals across sessions ( Figure 7C , E ) . To compare between stimulated and non-stimulated trials in the polar plots of the LocoMouse task ( Figure 4—figure supplement 1 ) we used a three-way ANOVA applied to a linear mixed model ( as described in Machado et al . , 2015 ) . The results were reported as conditional F tests with Satterthwaite degrees of freedom correction . A one-way ANOVA was used to compare ROI experiment parameters across days ( Figure 6D–G ) , followed by post-hoc t-tests with Bonferroni correction ( multcompare , MATLAB ) . For all the ANOVAs performed , when mice were included , they were considered as a random factor . Differences were considered significant at *p<0 . 05 , **p<0 . 01 and ***p<0 . 001 , unless when Bonferroni correction for multiple comparisons was applied . To compare stim . and non-stim . distributions ( Figure 5I ) , we applied a Kolmogorov-Smirnov test . To assess whether increased photostimulation frequency affected movement speed in a dose-dependent manner we used a linear regression analysis , fitting delta speed as a function of frequency ( Figure 2F ) . To test for long-term effects of photostimulation on speed ( Figure 7 ) we considered the speed during a time window just before photostimulation ( pre-stimulation , as defined previously ) . We regressed these values against session number and group identity according to the following equation:Speed=β0+β1⋅Session+β2⋅Group+β3⋅Session⋅Group Where Session stands for the session numberand Group is equal one for G1 mice and 0 for G2 mice . To analyze the correlation between short- and long-term effects of photostimulation on movement speed and considering that our sample size does not permit analysis at the population level , we analyzed the data within animals and across sessions ( for a total of 24 × 3 data points ) . The hypothesis being tested is that the short-term slowing effect on day n ( speedpost ( n ) - speedpre ( n ) ) would be related to the long-term speeding effect ( speedpre ( n + 1 ) - speedpre ( n ) ) . Here speedpre is the speed just before stimulation onset and therefore represents the baseline , and speedpost is the speed during photostimulation , and n and n+1 refer to consecutive sessions . We also note that a direct correlation analysis between the two quantities could be compromised due to the fact that speedpre ( n ) appears in both terms . To overcome this limitation we performed the following two analyses: | The brain controls sleep , movement and the other behaviors that an animal needs to survive . A chemical called serotonin plays an important role in controlling these behaviors as it regulates the activity of nerve cells ( known as neurons ) throughout the brain . Serotonin is produced by a specific group of neurons found in an area at the base of the brain called the raphe nuclei . From there , serotonin is released into other parts of the brain to influence different behaviors . Although drugs that target serotonin are widely used as antidepressants , how this chemical signal acts in the brain remains a mystery . This is due , in part , to it being technically challenging to carry out experiments on the serotonin-producing neurons . A technique called optogenetics uses light to selectively activate or inhibit individual cells in live animals . Here , Correia , Lottem et al . use optogenetics to activate serotonin-producing neurons in the dorsal raphe nucleus of mice . The experiments show that triggering serotonin production for a few seconds causes the mice to move around more slowly as they explore their surroundings . This short-term release of serotonin only slows the mice down if they are not already occupied with other activities , such as finding water or balancing on a moving object . These experiments suggest that serotonin decreases an individual’s motivation to move but that this can be overcome by sufficiently powerful goals . In contrast , repeatedly activating the serotonin neurons over a period of several weeks led to long-term changes of the opposite kind – the mice begin to move around more quickly . The findings of Correia , Lottem et al . have possible implications for the use of drugs that target serotonin to treat mental disorders as it suggests important links between serotonin , movement , and the ability of the brain to change how it responds to certain situations . The next steps will be to investigate how the two different effects of serotonin are connected , which areas in the brain are involved and how best to apply these findings to clinical studies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Transient inhibition and long-term facilitation of locomotion by phasic optogenetic activation of serotonin neurons |
Tumor-initiating stem cells ( SCs ) exhibit distinct patterns of transcription factors and gene expression compared to healthy counterparts . Here , we show that dramatic shifts in large open-chromatin domain ( super-enhancer ) landscapes underlie these differences and reflect tumor microenvironment . By in vivo super-enhancer and transcriptional profiling , we uncover a dynamic cancer-specific epigenetic network selectively enriched for binding motifs of a transcription factor cohort expressed in squamous cell carcinoma SCs ( SCC-SCs ) . Many of their genes , including Ets2 and Elk3 , are themselves regulated by SCC-SC super-enhancers suggesting a cooperative feed-forward loop . Malignant progression requires these genes , whose knockdown severely impairs tumor growth and prohibits progression from benign papillomas to SCCs . ETS2-deficiency disrupts the SCC-SC super-enhancer landscape and downstream cancer genes while ETS2-overactivation in epidermal-SCs induces hyperproliferation and SCC super-enhancer-associated genes Fos , Junb and Klf5 . Together , our findings unearth an essential regulatory network required for the SCC-SC chromatin landscape and unveil its importance in malignant progression .
Stem cells ( SCs ) have the capacity to self-renew and to generate and repair tissues . Advances in genome technologies provide the means to not only comprehensively analyze transcriptional profiles of SCs but also map their epigenetic landscape on a global level . Studies on cultured embryonic stem cells ( ESCs ) have shown that a special set of large open-chromatin domains , so-called super-enhancers ( SEs ) , control expression of genes important to ESC behavior ( Whyte et al . , 2013; Parker et al . , 2013 ) . Like typical enhancers ( TEs ) , SEs bind with Mediator , a complex that brings promoter and enhancer together to activate transcription . However , SEs differ from TEs by their exceptional size ( often >15kb ) and by their high density of acetylation of histone H3 at the lysine 27 position ( H3K27ac ) , which characterizes active chromatin . Studies on a variety of tissue cell lines have shown that another key feature of SEs is a high density of sequence motifs for cell stage-specific transcription factors ( TFs ) . This allows for their cooperative binding , thereby rendering SE-regulated genes particularly sensitive to the key TF cohort . Moreover , the genes encoding lineage-specific TFs often themselves harbor SEs resulting in a stable feed-forward loop to fuel and maintain the lineage ( Whyte et al . , 2013; Hnisz et al . , 2013; Lovén et al . , 2013; Chapuy et al . , 2013 ) . Mice devote much of their tissue-generating energy to making hair follicles ( HFs ) , each of which possesses a compartment of multipotent SCs that fuel HF regeneration and hair growth . Additionally , hair regeneration is governed by cyclical bouts of SC activity where HFs are either at rest ( telogen ) or actively regenerating ( anagen ) . The ability to purify large quantities of SCs directly from their native microenvironment ( ‘niche’ ) makes the HF one of the few systems where in vivo chromatin dynamics can be studied in an adult tissue SC population . Such studies on purified HF-SCs show that although their TFs are largely distinct from those of ESCs , many of the basic features of SEs are displayed by HF-SCs in their native tissue niche ( Adam et al . , 2015 ) . Studies on SC populations isolated from adult tissues have also yielded new insights into SE dynamics . Notably , most of the master regulators of HF-SCs bind within smaller chromatin domains ( <3 kb ) of SEs termed ‘epicenters’ , which are also enriched for Mediator and for H3K27ac ( Adam et al . , 2015 ) . Additionally , epicenters cloned from HF-SC SEs can drive proper developmental-specific and hair cycling-specific behavior of a GFP reporter , underscoring the functional relevance of these chromatin domains in lineage-specific gene expression . Perhaps , most intriguing is that even though HF-SCs still possess stemness outside their native niche , for example in culture or in wound-repair , they change their SE landscape dramatically . Mechanistically , this plasticity of remodeling can be traced to a dramatic reduction of HF-SC TFs , resulting in a decommissioning of most in vivo SEs with a new set of SEs gained in vitro ( Adam et al . , 2015 ) . In a related study on tissue macrophages , it was shown that when these immune cells invade different tissues and are faced with new microenvironments , they too display markedly distinct chromatin landscapes ( Gosselin et al . , 2014; Lavin et al . , 2014 ) . Together , these findings underscore the importance of the microenvironment in governing chromatin dynamics . For SCs , this feature is particularly relevant , since they reside in specific niches , and receive signals from multiple different surrounding cells to guide their behavior ( Chen et al . , 2015; Scadden , 2014; Lane et al . , 2014; Morrison and Scadden , 2014; Schofield , 1978 ) . Cancer can be viewed as a disease of lost cell identity . The cellular origin of tumor-initiating , so-called ‘cancer SCs’ may be a result of arresting SC maturation or of dedifferentiation of a progenitor lineage , endowing the incipient tumor cell with the potential to self-renew and fuel the cancer . Although largely limited to studies on self-renewing cancer cell lines in vitro and their normal whole tissue ( e . g . MCF7 breast cancer cell line and mammary epithelium ) , increasing evidence underscores the importance of SEs in cancers ( Parker et al . , 2013; Hnisz et al . , 2013; Lovén et al . , 2013; Chapuy et al . , 2013; Chipumuro et al . , 2014 ) . Recent studies suggest that the transcriptional profile and behavior of tumor-initiating SCs are highly sensitive to their tumor-microenvironment , which is markedly distinct from that of their normal counterparts ( Plaks et al . , 2015; Oshimori et al . , 2015 ) . This makes it tempting to speculate that the SE profile of tumor-initiating SCs in vivo may be substantially different from that of normal tissue SCs . If so , SE profiling could provide an avenue for identifying the special set of cancer SC TFs that drive tumor progression . This knowledge could greatly accelerate the convergence of basic science and clinical translation into therapeutics and diagnostics that cripple oncogenic behavior . Here , we specifically test this hypothesis , focusing on skin squamous cell carcinoma ( SCC ) , which is one of the most common and rapidly rising cancers world-wide ( Rogers et al . , 2010; Jemal et al . , 2010; Trakatelli et al . , 2007 ) . More broadly , SCCs can also arise in lung , breast , esophagus , cervix , and oral tissues of the head and neck , where they are associated with high risk of metastasis and mortality ( Trakatelli et al . , 2007 ) . Using SE and transcriptional profiling , we first define the SE chromatin landscape and its gene expression program within the tumor-initiating SC population of skin SCCs and show that SE-associated genes constitute a signature that specifically associates with cancer-vulnerability genes , some of which correlate with poor survival among human SCC patients . Given that HF-SCs are an established origin of these cancers ( Lapouge et al . , 2011; White et al . , 2011 ) , we conduct comparative analyses and show that SCC-SC SEs are distinct from SEs found in HF-SCs and also from rapidly proliferating short-term progeny of HF-SCs . We show that SCC-SEs are enriched for sequence motifs of a unique set of TF families , whose members are themselves regulated by SEs . Most importantly , we show by gain and loss of function and by associated chromatin landscaping , that these factors are essential for driving the chromatin dynamics that define the malignant state and govern tumor maintenance and survival .
To explore the in vivo importance of SEs in tumor-initiating cancer SCs , we first set up a reliable and reproducible SCC allograft model . The vast majority of chemically induced SCCs in mice have mutations in Hras , Kras or Rras2 ( Nassar et al . , 2015 ) , and HRasG12V alone is sufficient to induce formation of benign tumors ( papillomas ) ( Chen et al . , 2009 ) . HRasG12V in combination with loss of TGFβ receptor II ( TGFβRII ) results in invasive SCCs , which can metastasize ( Guasch et al . , 2007; Lu , 2006; Bian et al . , 2009 ) . We therefore purified primary keratinocytes from skin of newborn mice harboring a conditional Tgfbr2 allele ( Tgfbr2fl/fl ) encoding a key transmembrane receptor for TGFβ signaling , and infected them with a non-integrating adenovirus expressing a Cre-GFP recombinase . After fluorescence-activated cell sorting ( FACS ) , purified cells were then stably transduced with an integrating retrovirus expressing HRasG12V ( strategy outlined in Figure 1—figure supplement 1A ) . As expected , transduced cells showed increased Ras/MAPK signaling and abrogated TGFβ/SMAD signaling ( Figure 1—figure supplement 1B and C ) . Upon intradermal injection into Nude mice , they efficiently formed SCC tumors , typified by hyperproliferation , pyknotic nuclei , a discontinuous basement membrane and signs of invasion into the surrounding stroma ( Figure 1—figure supplement 1D ) . With this system , tumor-initiation and progression were highly reproducible . Irrespective of whether chemically or genetically induced , tumor-initiating SCs of SCCs reside at the tumor-stroma interface and are highly enriched for integrins α6 and β1 ( Oshimori et al . , 2015; Maston et al . , 2006; Dowen et al . , 2014; Lapouge et al . , 2012 ) . To profile the SEs of SCC-SCs , we therefore employed FACS to purify the GFPhighα6-integrinhighβ1-integrinhigh population from Tgfbr2-null HRasG12V-oncogenic skin SCCs ( Figure 1—figure supplement 1E ) ( Schober and Fuchs , 2011 ) . We then took advantage of the fact that when chromatin immunoprecipitation and high-throughput sequencing ( ChIP-seq ) is carried out for Mediator 1 , H3K4me1 , and H3K27ac , they generate highly overlapping patterns within SEs ( Whyte et al . , 2013; Adam et al . , 2015 ) . Using H3K27ac as our paradigm , we performed ChIP-seq and delineated the SEs of our purified SCC-SCs . Independent biological replicates showed a high degree of similarity ( Pearson correlation coefficient of genome-wide read densities r>0 . 89; representative example in Figure 1—figure supplement 2A ) , so the replicate data were subsequently combined and processed to maximize the comprehensiveness of our SE analysis . H3K27ac peaks resided within promoters ( ± 2 kb of annotated genes ) ( 43% ) and distal elements , considered enhancers ( 57% ) ( Figure 1—figure supplement 2B ) . A small fraction ( 13% ) of all enhancers ( Figure 1—figure supplement 2C ) , were in close proximity ( <12 . 5 kb apart ) and highly enriched with H3K27ac; they formed ~350 unusually large ( median size >20 kb ) distal elements , thereby fulfilling the criteria for ‘SEs' ( Whyte et al . , 2013 ) ( Figure 1A , Figure 1—figure supplement 2C–E ) . SEs were also significantly more robust than TEs in their ability to drive gene expression in SCC-SCs ( Figure 1B ) . 10 . 7554/eLife . 10870 . 003Figure 1 . Chromatin mapping reveals striking differences between normal skin SCs and SCC-SCs . ( A ) Identification of H3K27ac super-enhancers in SCC-SCs . ( B ) Differential gene expression levels driven by SCC-SC super-enhancers and typical-enhancers . p-Values are from t-test . ( C ) Venn diagram showing that super-enhancers of SCC-SCs show little overlap with HF-SCs or HF-TACs . ( D ) Examples of SCC-SC-specific super-enhancers at Myc and Cd44 loci . ( E ) Differences between HF-SC and SCC-SC super-enhancers . Note the decommissioning of HF-SC master regulators in SCC-SCs and corresponding suppression of HF-SC TF expression . ( F ) Enhancer remodeling correlates with gene expression changes . Boxplot displaying the full range of gene expression changes ( min . to max . ) . ( G ) Selected genes associated with SCC-SC super-enhancers . HF , hair follicle; SC , stem cell; SCC , squamous cell carcinoma; TF , transcription factor . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 00310 . 7554/eLife . 10870 . 004Figure 1—figure supplement 1 . Validation of the allograft tumor model . ( A ) Schematic of HRasG12V; Tgbr2-null SCC allograft model . ( B ) Immunoblot analysis showing impaired TGFβ/SMAD signaling in HRasG12V;TGFβRIIKO-GFP keratinocytes . ( C ) Immunoblot analysis showing increased Ras/MAPK/Erk activation ( P-ERK ) in in HRasG12V; Tgbr2-null keratinocytes . ( D ) Immunofluorescence microscopy validates the loss of TGFβ signaling in primary SCCs . Note that phospho-SMAD2 is activated in HRasG12V; Tgbr2fl/fl tumors , but not in HRasG12V; Tgbr2-null tumors . Scale bars , 100 μm . ( E ) FACS purification scheme for isolating SCC stem cells for ChIP-seq analysis: Representative flow cytometry profiles of cells isolated from primary HRasG12V; Tgbr2-null SCCs and fractionated based on surface α6 ( CD49f ) and β1 ( CD29 ) integrins ( high in SCC-SCs ) after selecting live , GFP-positive , transplanted tumor cells . FACS , fluorescence-activated cell sorting; SC , stem cell; SCC , squamous cell carcinoma . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 00410 . 7554/eLife . 10870 . 005Figure 1—figure supplement 2 . Super-enhancer profiling in SCC-SCs . ( A ) Representative example of H3K27ac-marked super-enhancers at the Mapk6 locus in SCC-SCs shows highly similar profiles of two independent biological replicates . ( B ) Distribution of H3K27ac occupancy at promoter and enhancers in SCC-SCs . ( C ) Distribution of typical- and super-enhancers in SCC-SCs . ( D ) Representative H3K27ac-marked typical-enhancer and super-enhancer at Xrcc6bp1 and Src loci , respectively , in SCC-SCs . ( E ) Enhancer size distribution in SCC-SCs . ( F and G ) Gene Ontology analysis of SCC-SCs super-enhancer-associated genes on molecular function and biological process . SCC-SC , squamous cell carcinoma-stem cell . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 005 When compared to the SEs of HF-SCs ( Adam et al . , 2015 ) , that is well-established precursors for skin SCCs ( Lapouge et al . , 2011; White et al . , 2011 ) , it was readily apparent that the SE landscape had been dynamically remodeled in SCC-SCs . This was not attributable merely to the difference in proliferative status , as the SE landscape of SCC-SCs was also distinct from that of rapidly proliferative , short-lived HF-SC progeny ( transit-amplifying cells , TACs ) ( Figure 1C ) . Enhancers control adjacent genes by looping to their promoters , with most of these interactions occurring within <50 kb ( Maston et al . , 2006 ) . More than 80% of SEs can be accurately assigned to their target genes by applying proximity criterion and RNA-seq expression data ( Dowen et al . , 2014 ) . Most of the remaining ambiguities arise from situations where more than one active gene resides within the vicinity of SEs for a particular cell type ( Dowen et al . , 2014 ) . These can largely be resolved by extending comparative ChIP-seq and RNA-seq analyses to multiple lineage stages for a particular cell type ( Adam et al . , 2015 ) . Therefore , after conducting RNA-seq analysis on the GFPhighα6highβ1high SCC-SC population , we assigned SE-associated genes on the basis of 1 ) their proximity to an SCC-SC SE; 2 ) their active transcription in SCC-SCs; and 3 ) their strict correlation between expression pattern and the presence of their putative SE not only in SCC-SCs but also in HF-SCs and transit-amplifying progeny . On the basis of this analysis , we readily assigned 340 genes to the 350 SEs found in SCC-SC chromatin , as 10 genes appeared to be controlled by more than one SE ( Supplementary file 1 ) . The cohort of SE-associated genes included a number of oncogenes such as Myc , and also genes associated with SCCs , for example Cd44 ( Figure 1D ) . Overall , 15% of the SCC-SEs completely lacked H3K27ac signals in HF-SCs , while 22% of the HF-SEs were silenced in SCC-SCs ( Figure 1E ) . Notably , the master transcriptional regulators of HF-SCs were largely decommissioned in SCC-SCs , while new transcriptional regulators were activated . Based on RNA-seq analysis , the genes associated with SCC-specific SEs displayed the highest increases in expression between HF and SCC-SCs , while genes associated with HF-specific SEs showed the greatest decline in expression in SCC-SCs ( Figure 1F ) . Sixty-four per cent of SCC SE-associated genes were still expressed in HF-SCs but had lost their SE and acquired a TE . In many cases , these enhancer dynamics had significant consequences , since the overall expression levels ( FPKM ) of SE-associated genes were higher than those of TE-associated genes ( Figures 1E and F ) , in agreement with previous findings for cultured ESCs ( Whyte et al . , 2013 ) . By contrast , genes which were associated with SEs both in SCC-SCs and HF-SCs were expressed at comparable levels ( Figure 1F ) . Unbiased gene ontology ( GO ) analysis categorized SE-associated genes in SCC-SCs as wound-responsive , stress-responsive , TF binding , kinase targets or actin binding , according to molecular function and biological process ( Figure 1—figure supplement 2F and G ) . Myc was particularly interesting , in that human MYC has been shown to be associated with an SE in a variety of cultured cancer cell lines ( Hnisz et al . , 2013 ) . In addition to Myc , there were a number of other established oncogenes that had SE specifically in SCC-SCs but not in HF-SCs , including Fos , Jun , Src , and Tgfa . Also on this list , there were cytokine genes Cxcl1 and Cxcl2 , as well as genes associated with cancer metabolism such as Slc2a1 and Gsr ( Figure 1G ) . Many of these genes also scored as ≥2X up-regulated in purified SCC-SCs relative to their normal counterparts in either epidermis or HF ( Schober and Fuchs , 2011; Lapouge et al . , 2012 ) . These findings underscored the importance of SE-regulated genes in cancer . Next , we sought to identify the key TFs involved in regulating the SE landscape in SCC-SCs . An unbiased motif analysis of SCC-SC SEs revealed a distinct set of putative TF binding sites that were largely non-overlapping with those found in the SEs of HF-SCs and TACs ( Adam et al . , 2015 ) ( Figure 2A ) . ETS was the most frequently found motif ( ~80% ) , with SOX and AP1 motifs found in >70% of all SCC-SEs ( Figure 2B ) . Notably , these putative binding sites occurred within epicenters , that is , small regions ( 1 . 5–3 kb ) of SE chromatin that were particularly enriched for the H3K27ac mark . Moreover , among SEs with both ETS and AP-1 sequence elements , ~40% contained the two motifs within a 100bp stretch , meeting conditions for potential cooperative binding . 10 . 7554/eLife . 10870 . 006Figure 2 . Identification of a cohort of SCC-SC specific transcriptional regulators . ( A ) Motif analysis of SCC-SC super-enhancers for putative TF binding sites . ( B ) Frequency of putative TF binding sites in SCC-SC super-enhancers . ( C ) SCC-SC TFs with potential to bind to the TF-motifs within SCC-SC super-enhancers . Genes encoding the TFs that are marked with an asterisk are SE-associated . ( D ) Immunofluorescence images showing nuclear localization of SCC-SC TFs ( red ) in allograft SCC-SC-derived tumors ( GFP ) . Scale bars , 100 μm . ( E ) Immunohistochemistry with P-ETS2 antibodies in human SCC samples . ( F and G ) High expression of mouse SCC-SC-expressed ETS family members correlate with poor survival in human SCCs . Kaplan–Meier analysis compares overall survival of TCGA head and neck SCC patients stratified according to high and low ELK3 and ETS2 expression . SC , stem cell; SCC , squamous cell carcinoma; TF , transcription factor; DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 006 Prior studies on SEs suggest that the key regulatory TFs are those whose genes are regulated by SEs ( Whyte et al . , 2013; Adam et al . , 2015 ) . Among the SE-regulated genes were three AP-1 motif genes ( Fos , Junb , Nfe2l2 ) , three KLF motif genes ( Klf5 , Sp2 , Sp3 ) , two ETS motif genes ( Elk3 , Ets2 ) , and Hoxb7 ( Figure 2C ) . SOX2 is a well-established SCC-TF ( Liu et al . , 2013; Siegle et al . , 2014; Boumahdi et al . , 2014 ) , although its gene had a TE and not an SE . When possible , immunofluorescence confirmed their expression ( Figure 2D ) . Functional analyses previously highlighted the importance of Fos , Junb , Nfe2l2 , and Sox2 in SCCs ( Oshimori et al . , 2015; Liu et al . , 2013; Siegle et al . , 2014; Boumahdi et al . , 2014; Gao et al . , 2014; Briso et al . , 2013; Ding et al . , 2013 ) . However , with the exception of a report of an elevation in Ets2 in ~75% of esophageal cancer patients ( Li , 2003 ) , an association of ETS proteins and SCCs has not been hitherto described , and functional analyses in SCCs are entirely lacking for this family . Both ELK3 and ETS2 were expressed and nuclear in our skin SCCs ( Figure 2D ) . For ETS2 , we could even confirm that this TF was threonine 72 ( T72 ) phosphorylated , that is active , in SCC-SCs ( shown ) . This modification is known to be dependent on Ras-MAPK kinase and results in an increase of its DNA-binding affinity and interaction with the histone acetylation complex CBP/p300 ( Foulds et al . , 2004 ) . Interestingly , active phosphorylated ETS2 was also readily detected in human SCCs , as judged by immunohistochemistry ( Figure 2E ) . Finally , and most importantly , when we surveyed the TCGA database and analyzed the relation between Elk3 and Ets2 expression versus median survival of SCC patients , a clear correlation was observed between the level of expression of these genes individually and poor prognosis ( Figures 2F and G ) . For patients with high Elk3 and/or Ets2 expression , the median survival was ~3–6 fold less than that of patients with lower expression levels . Based on these analyses , ELK3 and ETS2 became prime candidates to participate in autoregulation of the SCC-SC SEs . Our ChIP-seq analyses showed that genes encoding both ELK3 and ETS2 were governed by SEs . To test their functional relevance , we began by focusing on ELK3 as its expression was quantitatively suppressed in epidermal ( Epi ) progenitors and quiescent HF-SCs , and was only transcribed at low levels in proliferative progenitors during the growth phase of the hair cycle . By contrast , Elk3 was robustly transcribed in SCC-SCs , where its encoded protein ELK3 was nuclear ( Figures 1E , 3A and 3B ) . 10 . 7554/eLife . 10870 . 007Figure 3 . ELK3 is specifically induced in SCC-SCs and promotes SCC growth . ( A ) Immunofluorescence analysis showing that ELK3 is absent in normal skin epithelia but expressed in SCC-SC-derived allograft tumors . Epi-SC , epidermal basal cells; HF-SC , hair follicle stem cells; HF-TAC , short-lived , transit-amplifying cells . Scale bars , 100 μm . ( B ) Elk3 gene expression is low or absent in normal skin cells but specifically induced in SCC-SCs . Telogen ( t ) and anagen ( a ) HF-SCs; isthmus , infundibulum , and hair germ are other progenitor compartments of the HF . ( C ) ( left ) Knockdown efficiency of Elk3 shRNAs in HRasG12V; Tgfbr2-null cells as measured by quantitative RT-PCR . ( n = 3 ± SEM; *p<0 . 05 ) . ( right ) Immunoblot analyses confirming that changes in mRNA levels are reflected at the level of protein . ( D ) Immunofluorescence image of allograft tumors ( GFP ) from control and Elk3-knockdown SCC-SCs . Scale bars , 100 μm . ( E ) Changes in tumor volume of Elk3 knock-down and control allografts over time ( n = 3 ) . ( F ) Immunofluorescence images of Elk3 knock-down allograft tumor ( GFP ) showing reduced numbers of undifferentiated keratin 5 ( K5 ) cells , appearance of keratinized pearls of cells expressing differentiation marker keratin 10 ( K10 ) , and diminished signs of basement membrane breakdown ( as judged by laminin 5 ) and of invasion at the tumor-stromal interface . Asterisks mark absence of K5 in keratinized pearls . Dotted line marks tumor-stromal boundaries . Arrows denote invasive tumor cells at signs of discontinuous basement membrane . Scale bars , 100 μm . HF , hair follicle; SC , stem cell; SCC squamous cell carcinoma . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 007 To assess the physiological relevance of ELK3 in SCC tumorigenesis , we identified three distinct shRNA hairpins that knocked down Elk3 mRNA and protein levels by >70% ( Figure 3C ) . We transduced the same starting population of SCC-SCs with each of these shRNAs and with a control scrambled ( SCR ) shRNA and tested their abilities to affect tumor growth when injected intradermally into host recipient mice . Relative to scrambled ( SCR ) control shRNAs , each Elk3 shRNA efficiently knocked down the expression of ELK3 protein in the resulting tumor tissue , and exerted a potent effect , with signs of reduced allograft tumor growth appearing as early as 8–9 days after injection ( Figure 3D and E ) . Moreover , the morphologies and corresponding biomarkers of resulting cell masses were also dramatically changed in ELK3-deficient tumors . As shown in Figure 3F , SCC-SCs transduced with scrambled ( SCR ) control viruses generated tumors which were undifferentiated and almost exclusively positive for keratin-5 ( K5 ) , a marker of cells with proliferative progenitor potential ( Fuchs and Green , 1980; Blanpain and Fuchs , 2014 ) . By contrast , ELK3-deficient SCC-SCs formed tumors with pearls of keratinized K5-negative cells . These pearls were positive for keratin-10 ( K10 ) , a classic marker of terminally differentiating cells . Moreover , in contrast to classical signs of SCCs , including invasion and discontinuous basement membrane , tumors derived from ELK3-deficient SCC-SCs displayed smooth borders and quite continuous anti-laminin 5 immunolabeling at the tumor-stromal border ( Figure 3F ) . We next turned to ETS2 . Although active ( phosphorylated ) ETS2 protein was only seen in SCCs and not in normal progenitors , the Ets2 gene was more broadly and abundantly transcribed in progenitors , unlike Elk3 . Based on these data , it was equally important to examine the consequences of Ets2 loss of function on SCC tumorigenesis . To do so , we selected two powerful hairpins , which knocked down Ets2 transcription and protein production by >65% ( Figure 4A ) . Importantly , transduction of SCC-SCs with Ets2-shRNA also resulted in a loss of ETS2 protein as judged by immunolabeling of tumors derived from injecting these cells into host recipient mice ( Figure 4B ) . Of additional note , even though ETS2 was detected in the stromal cells , ETS2 protein was selectively absent in the Ets2-shRNA-transduced tumor masses . Together , these findings confirmed the efficacy of the knockdown . 10 . 7554/eLife . 10870 . 008Figure 4 . ETS2 governs SCC growth and malignancy . ( A ) ( left ) Knockdown efficiency of Ets2 shRNAs in HRasG12V; Tgfbr2-null cells as measured by quantitative RT-PCR . ( n = 3 ± SEM *p<0 . 05 ) . ( right ) Immunoblot analyses confirming that changes in mRNA levels are reflected at the level of protein . ( B ) Immunofluorescence image of allograft tumor ( GFP ) showing reduced ETS2 protein in Ets2 knock-down tumors . Scale bars , 100 μm . ( C ) Changes in tumor volume of Ets2 knock-down and control allografts over time ( n = 3 ) . ( D ) Immunofluorescence images of allograft tumors ( GFP ) of transduced SCC-SCs . Note reduction in undifferentiated K5+ cells and appearance of keratinized pearls of K10+ differentiated cells upon Ets2 knock-down . Of note , yellow arrows depict regions of discontinuous laminin5 and integrin β4 staining , indicative of a disrupted basement membrane and local tumor invasion . Scale bars , 100 μm . SC , stem cell; SCC , squamous cell carcinoma; SEM , standard error of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 008 Whether the SCC cells were transduced with Ets2 #649 or #985 , tumor growth was markedly reduced relative to control virus ( Figure 4C ) . Similar to the effects of Elk3-shRNAs , the morphologies and corresponding K5/K10 biomarkers of the resulting cell masses were also dramatically changed in ETS2-deficient tumors . Additionally , throughout much of the ETS2-deficient tissue , the β4 integrin and laminin5 patterns were uncharacteristically continuous , indicative of a well-organized underlying basement membrane . Concomitantly , very few finger-like projections into the surrounding stroma were seen with either Ets2-shRNA hairpins . Thus , while the tissue derived from Ets2 knockdown SCC cells was still disorganized , many of the classic signs of SCCs were lost without this SE-associated TF . Conversely , morphological and biochemical features of benign papillomas were enhanced in HRasG12V; Tgfbr2-null tumor masses in the absence of ETS2 ( Figure 4D ) . Overall , the impeding effects on SCC tumor growth and morphology were seen irrespective of whether we knocked down Ets2 or Elk3 . To unequivocally document the tumor-promoting effects of ETS proteins in skin , we performed rescue experiments with an shRNA-resistant Ets2-cDNA ( Figure 5—figure supplement 1A ) . When SCC cells were transduced with both the Ets2-shRNA and the Ets2-cDNA harboring a mutated sequence in the shRNA targeting site ( Ets2_mut ) and then injected subcutaneously into host recipient mice , the Ets2_mut refractory to shRNA knockdown rescued the effects of Ets2-shRNA . As shown in Figure 5A , SCC tumor growth and morphology was quantitatively restored , indicating that the effects on SCCs were directly rooted in the relative levels of ETS expression . 10 . 7554/eLife . 10870 . 009Figure 5 . ETS2 controls a transcriptional network driving SCC growth . ( A ) Rescue of SCC-SC growth by expressing an Ets2 cDNA harboring silent mutations in the Ets2-shRNA target site . ( B ) Immunofluorescence images of allograft tumors ( GFP ) derived from transduced SCC-SC cells . Note loss of SOX2 , KLF5 , and ELK3 expression in Ets2 knock-down but not control tumors . Scale bars , 100 μm . ( C ) Schematic of strategy to induce expression of a constitutively active , ETS2 ( T72D ) in normal skin . ( D ) Validation of efficient in utero transduction ( H2B-RFP ) and postnatal ETS2-T72D Doxy-induction in skin epithelium . ( E ) ETS2-T72D expression induces expansion of undifferentiated K5+ cells resulting in epidermal thickening and invagination . ( F ) Induction of constitutively active ETS2-T72D in epidermal progenitors results in marked upregulation of four additional TFs which have sequence motifs in >70% of SCC-SC super-enhancers and whose genes are themselves regulated by super-enhancers . SC , stem cell; SCC , squamous cell carcinoma . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 00910 . 7554/eLife . 10870 . 010Figure 5—figure supplement 1 . Validation of ETS2 expression constructs . ( A ) Lentiviral-based vector allowing for constitutive expression of a myc-tagged ETS2 cDNA , either WT or harboring silent mutations that abrogate shRNA binding . Immunoblot analysis to verify that both constructs express ETS2 protein at equivalent levels in HRasG12V; Tgfbr2-null keratinocytes , but that when Ets2#649 shRNA is transduced , myc-ETS2 is only suppressed in cells transduced with the WT 3’UTR and not the mutant 3’UTR version . ( B ) Vector allowing for Doxycycline-inducible expression of a myc-tagged ETS2 cDNA , either WT or harboring the phosphomimetic T72D varient . H2B-RFP , driven by a constitutively active PGK promoter , was inserted to control for lentiviral transduction . When the vector is transduced in cells or mice expressing the rtTA Doxy-inducible transactivator , Doxy will activate the Myc-tagged , ETS2-T72D protein . Immunoblot analysis to verify ETS2 protein expression in HRasG12V; Tgfbr2-null keratinocytes . ( C ) Ectopic induction of wild-type ETS2 in skin epidermis does not disrupt the morphology or function of the tissue . At E9 . 5 , litters harboring K14rtTA and control embryos were infected in utero with lentivirus harboring the wild-type ETS2 construct in ( B ) . This method allows for efficient and selective transduction of the skin epithelium by E12 . 5 . Postnatally , these pups were administered Doxycycline to activate rtTA ( if present ) and 4 weeks later , their skin was analyzed for transduction ( H2BRFP ) and epidermal architecture ( K5 and K10 ) . Sagittal sections of skin are stained with DAPI . Note no significant differences between the control and the WT ETS2 expression . ( D ) Same experiment in ( C ) but with the ETS2 ( T72D ) construct in ( B ) . Note striking difference in phenotype when this version of ETS2 , constitutively activated at its Ras/MAPK phosphorylation site , is expressed in otherwise normal skin epidermis . Immunolabeling is for β4 integrin to mark the basal surface of cells attached to an underlying basement membrane; Ki67 , to mark actively cycling cells; and CD31 , an endothelial marker for blood vessels . Quantifications of Ki67 and blood vessels are shown beneath the immunofluorescence pictures . Bars = 100 µm . ( E ) FACS profiling of the cells for CD11b , a pan marker of inflammatory immune cells . FACS , fluorescence-activated cell sorting . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 010 Previously , we showed that for HF-SCs , sustained levels of SOX9 were critical for the binding of other HF-SC TFs to SE epicenters ( Adam et al . , 2015 ) . Given the importance of ETS proteins , we wondered what the consequences of diminishing their expression might be on the SCC-SC TFs harboring binding motifs within the SCC-SEs . Interestingly , as shown in Figure 5B , loss of ETS2 abrogated expression of SOX2 , KLF5 , and ELK3 . In this regard , ETS2 was to SCC-SC SEs what SOX9 was to HF-SC SEs . If ETS factors function in SE dynamics , then elevating the levels of active ETS factors in normal skin progenitors might be expected to induce some of the phenotypic consequences of SCC progression , including hyperproliferation and the expression of other SCC-TFs or SCC genes that are regulated by SEs . To test this hypothesis , we engineered a phosphomimetic T72D version of ETS2 , thus bypassing the need for HRas/MAPK to make this modification ( Foulds et al . , 2004;O'Neill et al . , 1994; Brunner et al . , 1994; Rebay and Rubin , 1995 ) . Since ETS2 is also expressed in normal skin progenitors , we added a Myc-tag to monitor protein expression . Finally , we used a tetracycline regulatory element ( TRE ) to induce protein expression , and packaged the transgene in a lentivirus that also harbored a constitutively active H2B-RFP . When tested in transduced primary keratinocytes in vitro , expression of phosphomimetic T72D and wild-type ETS2 was markedly upregulated upon Doxycycline administration , where proteins , expressed at comparable levels , could be readily detected by ETS2 and Myc-antibodies ( Figure 5—figure supplement 1B ) . With these controls , we then proceeded to inject the high titer lentivirus into the amniotic sacs of E9 . 5 K14-rtTA mouse embryos in utero . As previously shown , this strategy selectively and efficiently transduces the single layer of surface epithelial progenitors of the embryo , particularly within the head region ( Beronja et al . , 2010 ) . Within 24 hrs , the lentiviral targeted DNA stably integrates into the host genome and can be stably propagated into the adult skin . By adding Doxycycline to the mouse chow , we could control the activation of the rtTA transactivator ( Nguyen et al . , 2006 ) , and hence the timing at which the TRE-T72D-ETS2 protein was expressed . Figure 5C illustrates this strategy . As shown in Figure 5D , Doxycycline induced overexpression of myc-tagged ETS2 transgene was confirmed . In vivo mice in which overexpression of wild-type ETS2 was induced in the skin epithelium did not exhibit overt phenotypic perturbations , and by immunofluorescence , epidermal differentiation appeared normal ( Figure 5—figure supplement 1C ) . By contrast , when the activated form ( T72D ) of ETS2 was induced , marked epidermal thickenings and invaginations were observed , accompanied by elevated immunolabeling for Ki67 , a marker of actively proliferating cells ( Figure 5E and Figure 5—figure supplement 1D ) . Additionally , immunofluorescence for the endothelial marker CD31 revealed signs of enhanced angiogenesis and FACS for CD11b suggested an increase in inflammatory cells ( Figure 5—figure supplement 1D and E ) . All these features are hallmarks of SCCs . Notably , T72D-ETS2 activated epithelium was also typified by expression of AP1 factors FOS and JUNB , as well as ELK3 ( Figure 5F ) . Intriguingly , KLF5 , whose gene is also SE-associated in SCCs , was also markedly upregulated , even though it is expressed at lower levels in normal epidermis ( Figure 5F ) . Overall , these changes were markedly distinct from Ets2 knockout mice , whose epidermal homeostasis appeared normal ( Yamamoto et al . , 1998 ) . When taken together , our findings suggest a specific role for ETS-family members in regulating skin tumor growth and malignant progression . While our findings thus far underscored the importance of SEs and ETS proteins , it was also important to test the functional relevance of SE dynamics . To do so , we first determined the transcriptional consequences of selectively inducing constitutively active ETS2 in the skin epithelium of postnatal mice . Interestingly , T72D-ETS2 not only generated SCC-like hyperproliferation and invaginations , but also induced and/or up-regulated many genes which we had found to be strongly expressed in SCCs ( Figures 6A and B ) . Among the genes upregulated by ≥2X upon ETS super-activation in normal skin progenitors ( EpiSCs ) , nearly 50% were also enhanced in SCC-SCs . Cxcl1 ( >300X ) , Cxcl2 ( >50X ) , and Elk3 ( 4X ) ranked among the most differentially expressed . 10 . 7554/eLife . 10870 . 011Figure 6 . Super-activated ETS2 drives chromatin dynamics and transcriptional changes that occur during malignant transformation . ( A ) Summary of transcriptional profiling of basal epidermal progenitors ( Epi-SC ) purified from ETS2 ( T72D ) induced or control skin . Significantly upregulated genes ( greater than twofold ) were ranked and are listed at right with fold changes . Of note , SCC-SC super-enhancer ( SE ) -associated genes are marked in red . ( B ) Venn diagram showing significant overlap between differentially regulated transcripts in T72D-ETS2 Epi-SCs and SCC-SC as compared to Epi-SC . ( C ) Venn diagram showing that SEs of SCC-SCs show high overlap with those of ETS2 ( T72D ) Epi-SC . ( D ) Heatmap showing H3K27ac ChIP-seq read densities in the SCC-SC SEs . Note that read densities of ETS2 ( T72D ) induced Epi-SCs are higher than those of control Epi-SCs . ( E ) Examples of SEs acquired in ETS2 ( T72D ) induced Epi-SC and which show significant overlap with SCC-SC SEs . Shown are Elk3 and Mapk6 loci . ( F ) Examples of SEs shared not only by SCC-SCs and ETS2 ( T72D ) -EpiSCs , but also by wild-type EpiSCs . Note that both Neat1 and Cdh1 are highly expressed in both normal and malignant skin epithelia . ( G ) qPCR fold enrichment of ETS2 and myc ChIP DNA of SCC-SC super-enhancer epicenters . Values are normalized to IgG control ( n = 3 ± SEM *p<0 . 05 ) . SC , stem cell; SCC , squamous cell carcinoma . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 011 Notably , many of the genes that were markedly up-regulated in T72D-ETS2 expressing skin epithelium also displayed SE in SCCs ( highlighted in red in Figure 6A ) . To address whether these T72D-ETS2 driven transcriptional dynamics were a reflection of SCC-like changes in the chromatin landscape , we purified α6+Sca1+ basal progenitors from the skin epidermis of our mice at 4 weeks after ETS induction , and then subjected these and control epidermal cells to ChiP-seq analysis for the SE mark H3K27ac . Remarkably , 46% of the SEs found in SCC-SCs overlapped with those of T72D-ETS overexpressing skin epidermis with a genome-wide Pearson Correlation Coefficient of 0 . 72 ( Figures 6C–E ) . A few genes , such as Cdh1 and Neat1 possessed SEs in SCC-SC , T72D-ETS2-EpiSC and EpiSCs , reflective of their sustained high expression in normal and tumor tissue ( Figure 6F ) . To test for direct binding of ETS2 to these SCC-SEs , we performed in vivo ETS2-CHIP-qPCR on induced T72D-ETS SCC-SCs . Using ChIP immunoprecipitations with antibodies against both endogenous ETS2 and also the Myc-tag , we observed clear enrichment of representative SCC-SE epicenters that harbor ETS2 motifs ( Figure 6G ) . Together , these findings provide compelling evidence that over-expressing a constitutively activated form of ETS2 bypassed the need for oncogenic Ras-driven MAP kinase activity in eliciting many of the chromatin changes associated with SCCs . To this end , we focused on the acetylation of lysine 27 of histone H3 , which renders SE chromatin mutually exclusive for Polycomb ( PcG ) -mediated repression , typified by a trimethylation mark at this same residue . In this regard , it was intriguing that a small number of genes , including Cxcl1/2 , Hmga2 , Igf2bp2 , and Vim , were PcG-repressed in normal skin progenitors ( Lien et al . , 2011 ) , but they displayed H3K27ac-associated SEs in SCC-SCs ( Figure 7A ) . Correspondingly , their expression levels were also markedly increased compared to either epidermal progenitors ( EpiSCs ) or HFSCs ( Figure 7B ) . 10 . 7554/eLife . 10870 . 012Figure 7 . Inflammatory mediators are SCC-SE-regulated and affect SCC growth . ( A ) Chromatin status at the Cxcl1/2 locus in SCC-SCs versus normal skin progenitors . Note that strong peaks of H3K27ac-associated , active chromatin are present throughout this locus in SCC-SCs , while in normal skin progenitors , the locus is heavily marked by H3K27me3 , indicative of Polycomb-mediated repression . ( B ) Gene expression changes of Cxcl1 , Cxcl2 , and Cxcr2 ( encoding the receptor for CXCL1/2 ) in SCC-SCs compared to their normal HF-SC or Epi-SC counterparts . ( C ) Immunoblot analysis shows that CXCL2 activates MAPK signaling in HRasG12V; Tgfbr2-null keratinocytes . β-actin levels are shown as controls . ( D ) Knock-down of Ets2 reduces Cxcl1/2 mRNA expression . ( E ) Knock-down efficiency of Cxcr2 shRNAs in HRasG12V; Tgbr2-null cells as measured by quantitative RT-PCR . ( n = 3 ± SEM *p<0 . 05 ) . ( F ) Immunofluorescence images of allograft tumors ( GFP ) from Cxcr2-shRNA and scrambled control-shRNA transduced SCC-SC cells . Note that CXCR2-reduction is accompanied by a marked increase in K10+ differentiated cells within the tumors . Scale bars , 100 μm . ( G ) Changes in tumor volume of Cxcr2 knock-down and control allografts over time ( n = 3 ) . ( H and I ) High CXCL8 expression ( the closest human homologue for CXCL1/2 ) correlates with shortened survival in human head & neck SCC patients . Kaplan–Meier analysis comparing overall survival ( H ) and disease-free survival ( I ) of TCGA HNSCC patients stratified according to the highest ( >5th percentile ) CXCL8 expression/amplification versus the rest ( >5th percentile ) ( please visit http://bit . ly/1Afq0Gt ) . SC , stem cell; SCC , squamous call carcinoma . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 01210 . 7554/eLife . 10870 . 013Figure 7—figure supplement 1 . SCC-SC TFs coordinately bind to Cxcl1/2 SE-epicenters . ( A ) Two replicate immunoblot analyses show that CXCL1 and CXCL2 activate MAPK signaling in HRasG12V;Tgfbr2-null keratinocytes . β-actin levels are shown as controls . Note that the response for P-AKT and P-p38 is consistently transient , reflective of feed-back regulation . ( B ) Motif analysis on Cxcl1/2 SE-epicenters reveals putative binding of a cohort of SCC-SC TFs to drive Cxcl1 and Cxlc2 expression . ( C ) Immunoblot validating the knock-down efficiency of Cxcr2 shRNAs on CXCR2 protein levels ( D ) Immunofluorescence images of Cxcr2 shRNA-transduced allograft tumor ( GFP ) showing intact and continuous laminin5+ basement membrane , reflective of reduced tumor potential . SCC-SC , squamous cell carcinoma-stem cell . DOI: http://dx . doi . org/10 . 7554/eLife . 10870 . 013 We were particularly intrigued by the Cxcl1 and Cxcl2 locus , as the encoded Cxcl1 and Cxcl2 cytokines can act in paracrine fashion to recruit neutrophils and activate angiogenesis , features which promote tumor progression ( Acharyya et al . , 2012 ) and which are characteristic of the SCC-SC niche , but not that of their normal counterparts ( Oshimori et al . , 2015 ) . Interestingly however , Cxcr2 , the gene encoding the cognate receptor for CXCL1 and CXCL2 , was also elevated in SCC-SCs relative to either HFSCs or EpiSCs ( Figure 7B ) . While induction of phospho ( active ) ERK was sustained for up to 4 hr , activation of AKTand p38 was transient indicative of feed-back regulation , which is known to occur for this pathway ( Avraham and Yarden , 2011; undefined ) . This suggested that in addition to its paracrine role , these cytokines may also serve an autocrine role in SCC behavior . In agreement with this notion , CXCL2 stimulation of HRasG12V transformed , Tgfbr2-null keratinocytes resulted in a marked activation of MAPK signaling as typified by increased ERK , AKT , and p38 phosphorylation ( Figure 7C and Figure 7—figure supplement 1A ) . Thus , CXCR2 regulation offered a good focus for our testing the physiological relevance of SE dynamics in SCC progression and the importance of changes in the niche microenvironment . Consistent with the fact that the CXCL1/2 locus gained a SE not only in CSC but also upon expression of a constitutively active ETS ( Figure 6A ) , we found ETS binding motifs within the SEs of the Cxcl1/Cxcl2 locus ( Figure 7—figure supplement 1B ) and showed by ETS ChIP-qPCR its binding ( Figure 6F ) . To verify the functional significance , we knocked down Ets2 with our two different shRNAs and examined the consequences to Cxcl1 and Cxcl2 expression . As shown in Figure 7D , a significant reduction ( ≥2X ) of Cxcl1 and Cxcl2 was seen upon transduction with Ets2 shRNA relative to the scrambled control shRNA . To further test a potential autocrine role of CXCL1 and CXCL2 in SCC progression , we knocked down Cxcr2 in HRasG12VTgfbr2-null keratinocytes and injected the scrambled shRNA-transduced and Cxcr2-shRNA transduced cells into host recipient mice to form allografts . As shown in Figure 7E and F and Figure 7—figure supplements 1C and D , knockdowns were effective in abrogating both mRNA and CXCR2 protein in vitro as well as CXCR2 protein expression in vivo . Moreover , when tumor growth was monitored , a marked reduction of tumor growth and increased differentiation of cancer cells was observed in the allografted tumors ( Figures 7F and G ) . These findings were in good agreement with studies on Cxcr2 knockout mice ( Cataisson et al . , 2009 ) , and confirmed the importance of keratinocyte-specific CXCR2 and autocrine CXCL1/2 signaling in SCCs . The closest human ortholog of the murine Cxcl1 and Cxcl2 genes is human CXCL8 , encoding interleukin 8 ( IL-8 ) ( Zlotnik et al . , 2006 ) . Upon analysis of the TCGA database , IL8 expression correlated inversely with the overall survival and disease-free state of head and neck SCC patients ( Figure 7H and I ) . Taken together , our data provided compelling evidence that the CXCL1/2-CXCR2 signaling pathway contributes to SCC maintenance and that during tumor progression , it becomes activated through dramatic remodeling of the Cxcl1/2 locus from a PcG-repressed to an ETS-regulated , SE-activated state .
SCCs are the most common cancers world-wide , occurring most frequently on the skin and affecting millions of people each year . Being visible , skin SCCs are typically treated early and mortality is low . However , SCCs can also arise in the lung , breast , esophagus , cervix , and head and neck , where they are associated with high risk of metastasis and drug resistance . Understanding the molecular complexities that underlie the tumor-initiating SCs within SCC cancers is therefore of paramount importance in developing new and improved diagnostic and therapeutic tools to establish new treatments . Like HF-SCs and EpiSCs from which SCC-SCs can derive ( Lapouge et al . , 2011; White et al . , 2011 ) , SCC-SCs are integrin-rich and reside at the interface with their surrounding stroma ( Schober and Fuchs , 2011; Lapouge et al . , 2012 ) . However , the gene expression program of these malignant SCs is markedly divergent from their normal skin counterparts ( Schober and Fuchs , 2011; Lapouge et al . , 2012 ) . Despite this knowledge , little was known about the master transcriptional regulatory network and the chromatin dynamics that underlie these dramatically different programs of gene expression in SCC-SCs versus normal SCs . We profiled the SEs of SCC-SCs and compared them to their wild-type counterparts in order to gain a better understanding of the epigenetic rewiring that might be associated with SCC transcription . Since the chromatin landscape in general , and SEs in particular , are highly sensitive to their microenvironment ( Adam et al . , 2015; Lavin et al . , 2014 ) , it was essential to carry out our experiments on SCC-SCs purified directly from the tumor , that is , in their native niche at the tumor-stroma interface . In previous studies on SCs in which chromatin landscapes regulated by master regulators have been mapped , ChIP-seq analysis of the known master TFs has preceded subsequent SE analysis . Thus , with cultured ESCs , ChIP-seq for the established pluripotency regulators SOX2 , OCT4 , NANOG , and KLF4 , was already available when ensuing SE analyses revealed that the pluripotency factors bind within these open chromatin domains ( Whyte et al . , 2013; Dowen et al . , 2014 ) . Similarly for HF-SCs that were purified directly from skin , in vivoChIP-seq analyses had been performed on NFATc1 , TCF3/4 , SOX9 , and LHX2 , before subsequent H3K27ac ChIP-seq analysis revealed that these master regulators bind within the HF-SC-specific epicenters of these large open chromatin domains ( Adam et al . , 2015 ) . For SCC-SCs , global ChIP-seq chromatin mapping in vivohas not been performed , due to the relative paucity of tumor tissue relative to normal skin . Moreover , the abundance of keratin and resilience of the epidermal plasma membrane poses technical hurdles for newer ATAC-seq methods which can be performed with fewer cell numbers and allow TF mapping for some cell types ( Buenrostro et al . , 2015 ) . Compounding the lack of knowledge on chromatin landscapes of SCC-SCs is the paucity of functional studies on putative transcriptional regulators of these cells , thus far limited to SOX2 ( Siegle et al . , 2014; Boumahdi et al . , 2014 ) . Thus , in addition to exploring the chromatin landscape of SCC-SCs , a major quest in the present study was to see whether through in vivoH3K27ac-mediated epigenetic analyses of purified SCC-SCs , we might be able to gain new insights into key transcriptional regulators that govern SCC-SC behavior in the context of their native , aberrant SC niche . To approach the problem , we identified the TFs whose expression is upregulated in SCC-SCs and which display putative TF binding sites within the epicenters of dense H3K27ac peaks of the majority of SCC-SC-specific SEs . As learned from analyses of ESCs and HF-SCs , an additional frequent , albeit not absolute , feature of SC master regulators is that their genes are often regulated by SEs themselves , thereby establishing a feed-forward loop for maintaining the SC state ( Whyte et al . , 2013; Adam et al . , 2015 ) . By SE profiling of SCC-SCs , identifying putative binding sites within SCC-SC SE epicenters and RNA-seq , we generated a list of candidates for SE-associated master regulator genes . Several lines of evidence suggest that our strategy was successful . First and foremost was the finding that SOX binding sites were present in >70% of all SCC-SC SEs , a feature predicted from SOX2’s functional role in SCCs ( Siegle et al . , 2014; Boumahdi et al . , 2014 ) . Additionally , given the broad importance of AP-1 ( FOS and JUN family ) members in cancers ( Eferl and Wagner , 2003 ) , binding sites for these factors in SCC-SC SEs were notable . Finally , although ETS proteins have not been previously implicated in SCCs , the presence of these sites in nearly 80% of SCC-SC SEs , coupled with the association of Elk3 and Ets2 with SEs , begged for functional studies . Indeed , our analyses provide compelling evidence that these factors drive hyperproliferation and SCC progression . Our added finding that high ELK3 and ETS2 expression correlates with poor prognosis in human SCCs fuels the importance of this family of proteins in these cancers . The role of ETS proteins is particularly intriguing given that their levels matter not only in tumorigenesis but also in maintaining SCC-SC master regulators . Thus when ETS2 was reduced in SCC-SCs , KLF5 and SOX2 were also absent from the resulting cellular masses , and when constitutively active ETS2 was induced in normal epidermis of juvenile mice , the epidermis acquired marked features of hyperproliferation and invasion , accompanied by upregulation/induction of FOS , JUNB and KLF5 and a shift in SE dynamics from one of normal skin progenitors to one of malignant transformation . When coupled with the fact that Fos , Junb and Klf5 are associated with SCC-SC SEs which contain AP1 , ETS and KLF binding motifs , these findings underscore the physiological relevance of these factors not only in SCC oncogenesis but also in orchestrating the chromatin dynamics of SCC-SCs . Our results are particularly intriguing in that phosphorylation of ETS2 is known to be regulated by Ras/MAPK signaling ( Foulds et al . , 2004; O'Neill et al . , 1994; Brunner et al . , 1994; Rebay and Rubin , 1995 ) . Indeed , it was recently shown that 86% of all skin SCCs induced by classical chemical carcinogenesis with 9 , 10-dimethyl-1 , 2-benzanthracene ( DMBA ) involve mutations in either HRas or KRas ( Nassar et al . , 2015 ) , and KRas is markedly upregulated even in SCCs that do not directly involve oncogenic HRas transformation ( Schober and Fuchs , 2011; Lapouge et al . , 2012 ) . While cell culture experiments have documented important roles for Ras/MAPK-activated ETS proteins in cellular transformation of NIH3T3 fibroblasts ( Foos et al . , 1998 ) , our findings now lend in vivo relevance to this connection in SCC-SCs . Given the well-established links between AP1 family members , Ras/MAPK and cancer ( Eferl and Wagner , 2003 ) , the presence of AP1 and ETS motifs in nearly 80% of SCC-SC SEs takes on all the more significance . Interestingly , many of the ETS binding motifs in SCC-SC epicenters are in close proximity with AP-1 motifs . Closely juxtapositioned ETS-AP1 motifs , essential for cooperative binding , have also been observed in a prostate cancer cell line in culture , and notably , they exist within the regulatory regions of several key oncogenes in which the Ras/MAPK pathway is activated ( Hollenhorst , 2011 ) . While the large majority of prostate cancers are adenocarcinoma and not SCCs , these parallels are intriguing and merit further investigation in the future . In addition to gaining insights into chromatin dynamics in SCCs , the SE-regulated gene list that we unearth here is rich in important oncogenes . These include not only Fos , Junb , Ets2 , and Elk3 , but also Myc , Src , Mapk6 , Map2k2 , Cd44 , and Tgfa . Also present on this list are HMGA2 , RUNX1 and FOXG1 , which are critical TFs for self-renewal of SCC-SCs and other cancer models ( Sgarra et al . , 2004 ) ( Manoranjan et al . , 2013; Scheitz et al . , 2012 ) , as well as KLF5 , which plays a role in epithelial cell hyper-proliferation under inflammatory conditions ( Sur , 2006 ) . Moreover , HIF1α is a key regulator of hypoxic response in cancer ( Wilson and Hay , 2011 ) . Overall , the list of SCC-SC SE-associated genes was rich in important cell signaling genes that promote hyper-proliferation , migration , invasion , inflammation and cancer metabolism . When taken together with our findings that the SEs regulating normal HF-SC master regulators are decommissioned in SCC-SCs , these data expose SEs as chromatin gatekeepers of the cancer SC state . In HF-SCs , the SE chromatin landscape is highly sensitive to the SC niche microenvironment ( Adam et al . , 2015 ) . In this regard , the preponderance of SCC-SC SEs associated with inflammatory genes was intriguing since inflammation in the perivascular stroma is a key feature of the SCC microenvironment ( Oshimori et al . , 2015 ) . Inflammatory players can be proangiogenic , growth-promoting or tumor suppressive . In addition , epithelial-derived factors collectively termed as the 'epimmunome' ( Swamy et al . , 2010 ) can instruct immune cells , thereby harboring the potential to impinge upon immunocompetence and tumor immunosurveillance . Our discovery that SCC cells express and secrete CXCL1 and CXCL2 , appears to be particularly relevant , because this phenomenon turned out to be hardwired through the establishment of SEs . Additionally , CXCL1/2 can signal to the immune system , solidifying their part in the epimmunome . Intriguingly , however , we found that SCC cells also express CXCR2 , the cognate receptor for CXCL1/2 , thus establishing an autocrine loop to fuel epithelial proliferation . Indeed , when we knocked down Cxcr2 , tumor growth was severely impaired . In summary , we have unearthed many new insights regarding SEs that go beyond mere cancer-specific gene expression . For the first time , we show that in vivo SEs of a cancer SC are markedly distinct from their normal counterparts , and they reflect their dramatically altered microenvironment . We show that SEs controlling the normal SC-TFs are decommissioned in cancer , and that the new SEs drive cooperative auto-regulation of novel master regulators that specify the cancer state . For SCC , this includes an ETS2/ELK3-AP1-SOX2-KLF5 network of which level of ETS family members appears to be critical in regulating SCC SE dynamics and orchestrating expression of a cohort of oncogenic , growth-promoting , and epimmune genes . Of additional relevance are the mutually exclusive H3K27 modifications of SEs and PcG-silencing that provide a powerful two-way switch for the cancer-normal SC balance .
Female CD1 mice ( Charles River , New York , NY ) were used for the purification of HF SCs . Female CD1 mice transgenic for krt14-H2B-GFP ( Tumbar , 2004 ) were used for the purification of TACs . We used Tgfbr2 floxed ( Leveen , 2002 ) mice to isolate primary keratinocytes . Nude mice were from Charles River Laboratories . For lentiviral injections , transduced mice were confirmed by genotyping with RFP primers: forward 5’ –ATCCTGTCCCCTCAGTTCCAGTAC-3’ , reverse 5’-TCCACGATGGT GTAGTCCTCGTTG-3’ . For TRE-mycETS2 or mycETS2 ( T72D ) transduced mice , positive mice were fed with doxycycline-containing chow , starting at P0 . Mice were maintained in the Association for Assessment and Accreditation of Laboratory Animal Care-accredited animal facility of The Rockefeller University ( RU ) , and procedures were performed with Institutional Animal Care and Use Committee ( IACUC ) -approved protocols ( #13622-H , #14693-H and #14765-H ) . Newborn , primary mouse epidermal keratinocytes from Tgfbr2 floxed were cultured on 3T3-S2 feeder layer in 0 . 05 mM Ca++ E-media supplemented with 15% serum ( Blanpain et al . , 2004 ) . For adenoviral infections with Ad-Cre-GFP and Ad-GFP ( 1010 pfu/ml; University of Iowa , Gene Transfer Vector Core Iowa ) , cells were plated in 6-well dishes at 200 , 000 cells/well and incubated with adenovirus at a MOI of 100 in the presence of polybrene ( 100 mg/ml ) overnight . After 2 days , infected cells were sorted to purity by FACS and expanded in culture for an additional 5 days . Of note , after 5 days cells lost their GFP expression from the transient adenoviral infection . For the following retroviral infections with MSCV-HRasV12-IRES-GFP , cells were again plated in six-well dishes at 200 , 000 cells/well and incubated with retrovirus at a MOI of about 100 in the presence of polybrene ( 100 mg/ml ) overnight . After 2 days , GFP positive cells were again sorted to purity by FACS and tested for loss of Tgfbr2 and presence of RasV12 by Western blot and RT-PCR . To generate myg-tagged ETS2 expressing SCC cells , their tumor SCsstem cells were transduced with lentivirus containing CMV-mycETS2-PGK-H2BRFP , and after 3 days , RFP expressing cells were purified by FACS , and then injected into mice to form in vivo SCCs . We analysed the publicly available data sets of the The Cancer Genome Atlas ( TCGA: http://cancergenome . nih . gov ) . The cBioPortal for Cancer Genomics developed and maintained by the Computational Biology Center at Memorial Sloan-Kettering Cancer Center was used to mine the publicly available TCGA dataset on HNSCC ( Gao et al . , 2013; Cerami et al . , 2012 ) . To retrace the exact Kaplan–Meyer analysis , please visit SurvExpress http://bioinformatica . mty . itesm . mx:8080/Biomatec/SurvivaX . jsp for the analysis of HNSCC patients stratified by the 340 SE-associated genes ( Aguirre-Gamboa et al . , 2013 ) . For allograft transplantation , 1 . 0 × 105 mouse primary tumor cells were intra-dermally injected with Matrigel ( BD , East Rutherford , NJ ) in Nude mice . Tumor size was measured every three days and calculated using the formula 4/3π × L/2 × W/2 × D/2 . Tumors were dissected from mice , and normal skin , blood vessels and connective tissue were removed . Tumor tissues were minced and treated with 0 . 25% collagenase ( Sigma , St . Louis , MO ) in HBSS ( Gibco , Carlsbad , CA ) for 20 min at 37°C with shaking . After washing with cold PBS , the tissues were treated with 0 . 25% trypsin for 10 min at 37°C , and then were washed with PBS containing 5% of fetal bovine serum ( FBS ) . The cell suspensions were filtered through 70-μm and 45-μm strainers . Isolation of HF-SCs and Epi-SCs from adult mice back skins was done as previously described ( Adam et al . , 2015 ) , Briefly , for telogen skin , subcutaneous fat was removed with a scalpel , and skins were placed dermis side down on trypsin ( Gibco ) at 37°C for 35 min . Single cell suspensions were obtained by scraping the skin gently . After washing with PBS containing 5% of FBS , cells were filtered through 70-μm and 45-μm strainers . Cell suspensions were incubated with the appropriate antibodies for 10 min on ice . The following antibodies were used for FACS: Integrin α6-PE ( 1:100 , BD Biosciences ) , Integrin β1-APC/Cy7 ( 1:200 , Biolegend , San Diego , CA ) , CD34-eFluoro660 ( 1:100 , eBiosciences , San Diego , CA ) and Sca-1-PerCP-Cy5 . 5 ( 1:100 , eBiosciences ) and CD11b ( 1:200 , eBiosciences ) . DAPI was used to exclude dead cells . Cell isolations were performed on FACSAriaII sorters running FACSDiva software ( BD Biosciences ) . Immunoprecipitations were performed on FACS-sorted populations from tumors and nano-ChIP-seq was done as described ( Adli and Bernstein , 2011 ) . For each ChIP-seq run , 2 × 107 cells were used and for nano-ChIP-seq , 1~2 × 105 cells were used . Cells were sorted by FACS . Sorted cells were cross-linked in fresh 1% ( wt/vol ) formaldehyde solution for 10 min at room temperature followed by adding one-tenth volume of 2 . 5 M Glycine to quench formaldehyde . Cells were rinsed twice with PBS , flash frozen in liquid nitrogen and stored at -80°C prior to use . Cells were resuspended and lysed in lysis buffers . To solublilize and shear cross-linked DNAs , lysates were subjected to a Bioruptor Sonicator ( UCD-200 , Diagenode , Denville , NJ ) according to a 30X regimen of 30 s sonication followed by 60 s rest . The resulting whole-cell extract was incubated overnight at 4°C with Dynabeads Protein G magnetic beads ( Life Technologies , Carlsbad , CA ) which had been pre-incubated with anti-H3K27ac ( ab4729 , abcam , Cambridge , MA ) antibody . After ChIP , samples were washed with low salt , high salt , LiCl and Tris-EDTA buffer for 10 min . Bound complexes were eluted and crosslinking was reversed by overnight incubation at 65°C . Whole cell extract DAN was also treated for crosslink reversal . ChIP DNA was prepared for sequencing by repairing sheared DNA and adding Adaptor Oligo Mix ( Illumina , San Diego , CA ) in the ligation step . A subsequent PCR step with 25 amplification cycles added the additional Solexa linker sequence to the fragments to prepare them for annealing to the Genome analyzer flow cell . After amplification , a narrow range of fragment sizes between 150–300 bp was selected by separation on a 2% agarose gel and the DNA was gel-purified and diluted to 10 nM for loading on the flow cell . Nano-ChIP-seq was carried out similarly , except that ChIP DNA was primed with Sequenase enzyme using the primer 1 that contains universal sequence , the BciVI restriction site and random 9-mer and then amplified using primer 2 that contains universal sequence ( Primer 1: 5′-GACATGTATCCGGATGTNNNNNNNNN-3’ , Primer 2: 5′-GACATGTATCCGGATGT-3′ ) . After BciVI digestion , the ChIP DNA was ligated with Illumina adapters and amplified , and fragment sizes between 200-700bp were selected by electrophoresis . The DNA was gel-purified and sequencing was performed on the Illumina HiSEq 2500 Sequencer following manufacturer protocols . ChIP-seq reads were aligned to them mouse genome ( mm9 , build 37 ) using Bowtie aligner ( Langmead et al . , 2009 ) . ChIP-seq signal tracks were presented by Integrative Genomics Viewer ( IGV ) software . H3K27ac peaks were called by the program MACS ( Zhang et al . , 2008 ) ( v1 . 4 . 2 , default parameters ) from the aligned ChIP-seq data with the input as controls . The peaks were associated to genes in the mouse RefSeq annotation; those located within 2 kb of transcription start site were called as ‘promoter’ peaks and the rest collectively considered as ‘enhancer’ peaks . The H3K27ac enhancer peaks from two biological replicates were merged and were used for the identification of SEs , using the algorithm described previously , wherein enhancer peaks were stitched together if they are located within 12 . 5 kb of each other and if they don’t have multiple active promoters in between . The stitched enhancers were then ranked according to increasing H3K27ac signal intensity ( Whyte et al . , 2013 ) . Enhancer-gene assignments were performed using the following criteria to make gene assignments: ( Whyte et al . , 2013 ) proximity of genes to the SE of cancer SCs; ( Parker et al . , 2013 ) high transcriptional activity in cancer SCs by RNA-seq; ( Hnisz et al . , 2013 ) correlation between SE and candidate gene expression in CSCs and HFSCs . GO function enrichment analyses were conducted by the software GREAT ( McLean et al . , 2010 ) using the list of SE coordinates and the default setting . For motif analysis of SEs , we searched the 1-kb sequences under the summits of H3K27ac peaks within SEs searched for enriched motifs using the software HOMER ( v4 . 6 ) with the default setting and genome as background ( Heinz et al . , 2010 ) . Chromatin immunoprecipitations were performed as described above with FACS-sorted populations from tumors , and anti-ETS2 antibody ( Santa Cruz , Dallas , TX ) , anti-myc antibody ( Abcam ) and normal rabbit IgG ( Cell Signaling , Danvers , MA ) were used . The purified ChIP DNAs were mixed with indicated primers and SYBR green PCR Master Mix ( Sigma ) , and qRT-PCR was performed on an Applied Biosystems 7900HT Fast Real-Time PCR system . The following primer sequences were used ( 5’ to 3’ ) : Cxcl1/2 SE_epicenter forward: CAACATGCCTAGCCCGTGAGTC , Cxcl1/2 SE_epicenter reverse: GTGCCCTGTTTCACAGATAGAGGC Elk3 SE_epicenter forward: CGTCCATTCTCTCCCCTTTTCTAGC , Elk3 SE_epicenter reverse: CATGATTGGCAGTGGAGTATCGAGC Ets2 SE_epicenter forward: CAATGGCTTGGAGATCCCCGAC , Ets2 SE_epicenter reverse: CAGGGTCACCAGTGAGTCACAG Fam49b SE_epicenter forward: CCACGGAACCTGAGAATGAAGCC , Fam49b SE_epicenter reverse: GCTTCAACTGACTGAACTCCCAGG Il1a/b SE_epicenter forward: , CAGAGGGTGGCACAGGATAGACAG , Il1a/b SE_epicenter reverse: CAGTGTCCTGCCCAGTCATCTG Pthlh SE_epicenter forward: , CTGAGCTACACCCTTCCACTTCAC , Pthlh SE_epicenter reverse: GTCTTCATTCCTCTGAGCCAATGTGC Negative control region forward: CATGCAAGCATCACCAACAAAGTA , Negative control region reverse: CCATGGAACTGGGACCTTCTTC FACS-isolated cells and cultured cells were directly lysed into TRIzol LS ( Invitrogen , Carlsbad , CA ) . Total RNA was purified using the Direct-zol RNA MiniPrep kit ( Zymo Research , Irvine , CA ) per manufacturer’s instructions . For RNA-Seq , quality of the RNA for sequencing was determined using an Agilent 2100 Bioanalyzer; all samples used had RNA integrity numbers >9 . Library preparation using the Illumina TrueSeq mRNA sample preparation kit was performed at the Weill Cornell Medical College Genomic Core facility ( New York , NY ) , and RNAs were single-end sequenced on Illumina HiSeq 2000 machines . Alignment of reads was done using Tophat with the mm9 build of the mouse genome . Transcript assembly and differential expression was determined using Cufflinks with Refseq mRNAs to guide assembly ( Trapnell et al . , 2010 ) . Analysis of RNA-seq data was done using the cummeRbund package in R ( Trapnell et al . , 2012 ) . For real-time qRT-PCR , equivalent amounts of RNA were reverse-transcribed by SuperScript VILO cDNA Synthesis Kit ( Life Technologies ) . cDNAs were normalized to equal amounts using primers against Ppib2 , Hprt , or Gapdh . cDNAs were mixed with indicated primers and SYBR green PCR Master Mix ( Sigma ) , and qRT-PCR was performed on an Applied Biosystems 7900HT Fast Real-Time PCR system . The following primer sequences were used ( 5’ to 3’ ) : mPpib2 forward: GTGAGCGCTTCCCAGATGAGA , mPpib2 reverse: TGCCGGAGTCGACAATGATG mHprt forward: GATCAGTCAACGGGGGACATAAA , mHprt reverse: CTTGCGCTCATCTTAGGCTTTGT mGapdh forward: GTCGTGGAGTCTACTGGTGTCTTCAC , mGapdh reverse: GTTGTCATATTTCTCGTGGTTCACACCC mEts2 forward: GACGGGGGATGGATGGGAGTTCAAG , mEts2 reverse: AGCCCAGCAAGTTCTGCAGGTCACA mElk3 forward: TCCTCACGCGGTAGAGATCAG , mElk3 reverse: GTGGAGGTACTCGTTGCGG mCxcr2 forward: TGTTCTGCTACGGGTTCACACTG , mCxcr2 reverse: GCGGCGCTCACAGGTCTC mCxcl1 forward: CCACACTCAAGAATGGTCGC , mCxcl1 reverse: TCTCCGTTACTTGGGGACAC mCxcl2 forward: CGGTCAAAAAGTTTGCCTTG , mCxcl2 reverse: TCCAGGTCAGTTAGCCTTGC For immunofluorescence microscopy of tumor sections , dissected tumors were fixed with 1% PFA in PBS for 1 hr at 4°C , washed with PBS , incubated with 30% sucrose overnight at 4°C , and embedded in OCT ( VWR , Radnor , PA ) . For telogen , anagen and phosphomimic Ets2 ( T72D ) overexpressed skin , tissues were harvested and embedded in OCT . Cryosections were cut at a thickness of 10 μm on a Leica cryostat and mounted on SuperFrost Plus slides ( VWR ) . Sections were blocked for 1 hr in blocking buffer ( 5% normal donkey serum , 1% BSA , 2% fish gelatin , 0 . 3% Triton X-100 in PBS ) . Primary antibodies were diluted in blocking buffer and incubated at 4°C overnight . The following primary antibodies were used: ETS2 ( rabbit , 1:500 , Aviva Systems Biology , San Diego , CA; or rabbit , 1:500 , Life technologies ) , Phospho-ETS2 pThr72 ( rabbit , 1:500 , Life technologies ) , ELK3 ( rabbit , 1:100 , Novus , Littleton , CO ) , KLF5 ( goat , 1:50 , R&D , Minneapolis , MN ) , SOX2 ( rabbit , 1:200 , Abcam ) , Phospho-NFkB p65 pSer276 ( rabbit , 1:100 , Cell Signaling ) , FOS ( rabbit , 1:100 , Abcam ) , Integrin β4 ( rat , 1:100 , BD Pharmingen ) , GFP ( chicken , 1:2 , 000 , Abcam ) , pSmad2 ( rabbit , 1:1 , 000 , Cell Signaling ) , CD34 ( rat , 1:100 , BD Pharmingen ) , K6 ( guinea pig , 1:5 , 000 , Fuchs laboratory ) , K5 ( guinea pig , 1:1 , 000 , Fuchs laboratory ) , K10 ( rabbit , 1:1 , 000 , Covance , Princeton , NJ ) , Laminin5 ( rabbit , 1:1 , 000 , Fuchs Laboratory ) , CXCR2 ( rabbit , 1:200 , Santa Cruz ) , Loricrin ( rabbit , 1:1000 , Fuchs Laboratory ) , CD31 ( hamster , 1:200 , Millipore , Billerica , MA ) , and Ki67 ( rabbit , 1:1000 , Novocastra , Buffalo Grove , IL ) . After washing with PBS , sections were treated for 1 hr at room temperature with secondary antibodies conjugated with Alexa 488 , RRX , or Alexa 647 ( Life Technologies ) . Slides were washed , counterstained with 4’6’-diamidino-2-phenilindole ( DAPI ) , and mounted in Prolong Gold ( Life Technologies ) . Images were acquired with an Axio Observer . Z1 epifluorescence microscope equipped with a Hamamatsu ORCA-ER camera ( Hamamatsu Photonics , Geldern , Germany ) , and with an ApoTome . 2 ( Carl Zeiss , Oberkochen , Germany ) slider that reduces the light scatter in the fluorescent samples , using 20X objective , controlled by Zen software ( Carl Zeiss ) . Z stacks were projected and RGB images were assembled using ImageJ . Panels were labelled in Adobe Illustrator CS5 . Immunohistochemistry was performed as previously described ( Bian et al . , 2009 ) . Briefly , 5-μm sections were cut , stained with H&E or processed for immunohistochemistry/immunofluorescence microscopy . For immunoperoxidase staining , paraffin-embedded sections were dehydrated and antigenic epitopes exposed using a 10-mM citrate buffer ( pH 6 . 0 ) in a pressure cooker . Sections were incubated with the following primary antibodies at 4°C overnight: rabbit anti-ETS2 ( 1:100; LifeTechnologies PA5-28053 ) and rabbit anti-phospho-ETS2 ( 1:100; LifeTechnologies 44-1105G ) . Primary antibody staining was visualized using peroxidase-conjugated anti-rabbit IgG followed by the DAB substrate kit for peroxidase visualization of secondary antibodies ( Vector Laboratories , Burlingame , CA ) . Tissue microarrays comprising healthy human skin samples , human skin SCCs as well as head and neck SCCs ( HNSCC ) were obtained from US Biomax , Rockeville , MD: SK241 , SK801b , SK811a , SK2081 , and HN803a . To generate the phosphomimic Ets2 ( T72D ) expression construct , Ets2 cDNA was PCR amplified and 72 threonine residue was mutated to glutamate by site-directed mutagenesis . Final PCR product was nserted into the LV-TRE-PGK-H2BmRFP1 construct . The resulting LV-TRE-mycEts2 ( T72D ) -PGK-H2BmRFP was used for in utero injections . Production of VSV-G pseudotyped lentivirus was performed by calcium phosphate transfection of 293FT cells ( Invitrogen ) with pLKO . 1 and helper plasmids pMD2 . G and psPAX2 ( Addgene plasmid 12259 and 12260 , Cambridge , MA ) . Viral supernatant was collected 46 hr after transfection and filtered through a 0 . 45-μm filter . For lentiviral infections in culture , cells were plated in 6-well plate at 1 . 0 × 105 cells per well and incubated with viruses in the presence of polybrene ( 20 μg/ml ) for 30 min , and then plates were spun at 1100 g for 30 min at 37°C in a Thermo IEC CL40R centrifuge . Infected cells were selected with puromycin . For in vivo lentiviral transduction , viral supernatant was filtered ( 0 . 45-μm filter ) and concentrated by ultracentrifugation . Final viral particle was resuspended in viral resuspension buffer ( 20 mM Tris pH 8 . 0 , 250 mM NaCl , 10 mM MgCl2 , 5% sorbitol ) and 1 or 0 . 5 μl was in utero injected into E9 . 5 embryos . For knockdown experiments , we used clones from the Broad Institute’s Mission TRC mouse library . We tested the knockdown efficiency of 5–10 independent shRNAs for each gene and used the following clones and target sequence: The scramble shRNA ( Sigma SHC002 , CAACAAGATGAAGAGCACCAA ) , Ets2 ( TRCN0000233985 , CATTGATAAAGAGCCGTTATA; TRCN0000042649 , CCGTCAATGTCAATTACTGTT ) , Elk3 ( TRCN0000042643 , GCTGAGATACTATTACGACAA; TRCN0000235780 , ATCAGGTTTGTGACCAATAAA; TRCN0000235783 , AGAGCGCTGAGATACTATTAC ) , Cxcr2 ( TRCN0000026605 , GCCTTGAATGCTACGGAGATT; TRCN0000026647 , CGTTACAATTACAGTGAGATA ) . Data were analyzed and statistics performed using unpaired two-tailed Student’s t-test in Prism6 ( GraphPad software , La Jolla , CA ) . Significant differences between two groups were noted by asterisks or actual p values . Quantification data were presented in mean value ± SEM or in box and whisker plots with the dimensions of the box encompassing the 25th–75th percentile , the horizontal bar representing the median , and the error bars representing minimum and maximum values . ChIPseq and RNAseq databases were deposited in GEO ( Accession #GSE72147 ) . | Many cancers contain a mixture of different types of cells . Of these , cells known as cancer stem cells can form new tumours and drive the growth and spread of the cancer around the body . A central question is how cancer stem cells differ from healthy adult stem cells . Recent evidence suggests that , in addition to having genetic mutations , cancer stem cells live in a very different environment to other cells within the tumour . This 'microenvironment'also has a major impact on how these cells behave compared to normal stem cells . Together , the genetic and environmental differences profoundly change the way genes are expressed in the cancer cells . In 2013 , a group of researchers identified regions of DNA called super-enhancers . These regions are long stretches of DNA that proteins called transcription factors can interact with to coordinate the expression of nearby genes to alter the production of certain proteins . Super-enhancers contain several transcription factor-binding sites that are close to each other with the different sites being associated with transcription factors that are only active in specific types of cells . Furthermore , super-enhancers are often self-regulatory , meaning that the binding of transcription factors to a super-enhancer can lead to an increase in the expression of the genes that encode the same transcription factors . Yang , Schramek et al . have now identified the super-enhancers in a skin cancer called squamous cell carcinoma and showed that they differ dramatically from the super-enhancers of normal skin stem cells . Their experiments show that the active super-enhancers in cancer stem cells are associated with a very different set of genes that are highly and often specifically expressed in cancer stem cells . In the cancer stem cells , a transcription factor called ETS2 binds to the super-enhancers and reprograms the expression of genes to promote the development of cancer . Yang , Schramek et al . also show that over-active ETS2 is a major driver of squamous cell carcinoma . Furthermore , ETS2 also increases the expression of genes that cause inflammation and promote the growth of cancers . Yang , Schramek et al . ’s findings reveal a new regulatory network that governs the expression of genes involved in cancer . Furthermore , the experiments show that high levels of ETS2 are linked with poor outcomes for patients with head and neck squamous cell carcinoma , which is one of the most life-threatening cancers world-wide . In the future , these findings might lead to the development of new therapies to treat these cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2015 | ETS family transcriptional regulators drive chromatin dynamics and malignancy in squamous cell carcinomas |
Multifunctional proteins are evolutionary puzzles: how do proteins evolve to satisfy multiple functional constraints ? S100A9 is one such multifunctional protein . It potently amplifies inflammation via Toll-like receptor four and is antimicrobial as part of a heterocomplex with S100A8 . These two functions are seemingly regulated by proteolysis: S100A9 is readily degraded , while S100A8/S100A9 is resistant . We take an evolutionary biochemical approach to show that S100A9 evolved both functions and lost proteolytic resistance from a weakly proinflammatory , proteolytically resistant amniote ancestor . We identify a historical substitution that has pleiotropic effects on S100A9 proinflammatory activity and proteolytic resistance but has little effect on S100A8/S100A9 antimicrobial activity . We thus propose that mammals evolved S100A8/S100A9 antimicrobial and S100A9 proinflammatory activities concomitantly with a proteolytic ‘timer’ to selectively regulate S100A9 . This highlights how the same mutation can have pleiotropic effects on one functional state of a protein but not another , thus facilitating the evolution of multifunctionality .
The innate immune system uses a small number of multifunctional proteins to respond to diverse immune challenges ( Lee et al . , 2018; Auvynet and Rosenstein , 2009; Postel et al . , 2010; Gudmundsson and Agerberth , 1999 ) . Multifunctional immune proteins are critical for pathogen defense , ( Auvynet and Rosenstein , 2009; Postel et al . , 2010; Gudmundsson and Agerberth , 1999 ) shaping host-associated microbial communities , ( Singh et al . , 2016 ) and well-regulated tissue growth ( Säemann et al . , 2009; Bals and Wilson , 2003; Leask and Abraham , 2003 ) . They also drive pathological inflammation in disease , including autoimmune disorders , cancer , and cardiovascular disease ( Ehrchen et al . , 2009; Salama et al . , 2008; Shabani et al . , 2018; Gruden et al . , 2016; Peng et al . , 2018 ) . These multifunctional proteins raise both mechanistic and evolutionary questions . How can one protein sequence satisfy the multiple constraints imposed by having multiple functions ? How can multiple functions evolve in one protein when , as a result of multifunctionality , each mutation likely has pleiotropic effects ? ( Lee et al . , 2018; Bomblies and Doebley , 2006; Stearns , 2010; Paaby and Rockman , 2013 ) . One such multifunctional protein is S100A9 ( A9 ) , a small , soluble protein found at high concentrations in the extracellular space during an inflammatory response ( Berntzen and Fagerhol , 1990 ) . It has at least two key immune functions . As a homodimer , A9 potently activates inflammation via Toll-like receptor 4 ( TLR4 ) ( Vogl et al . , 2012; Källberg et al . , 2012; Duan et al . , 2018; Shepherd et al . , 2006; Schiopu and Cotoi , 2013; Laouedj et al . , 2017; He et al . , 2016; Gao et al . , 2015; Tsai et al . , 2014; Lee et al . , 2016; Björk et al . , 2009; Kang et al . , 2015; Stríz and Trebichavský , 2004 ) . As a heterocomplex with S100A8 ( A8/A9 , also known as calprotectin ) , it is antimicrobial ( Figure 1a; Besold et al . , 2018a; Hadley et al . , 2018; Damo et al . , 2013; Clark et al . , 2016; Besold et al . , 2018b; Nakashige et al . , 2016; Hayden et al . , 2013; Nakashige et al . , 2015; Liu et al . , 2012; Brunjes Brophy et al . , 2013; Brophy et al . , 2012; Baker et al . , 2017; Gagnon et al . , 2015; Nisapakultorn et al . , 2001 ) . A9 exacerbates endotoxin-induced shock in mice ( Vogl et al . , 2007 ) . Both A9 and A8/A9 are primary biomarkers for many human inflammatory diseases ( Vogl et al . , 2014; Hara et al . , 2012; Obry et al . , 2014; Horvath et al . , 2016; Huang et al . , 2015 ) . Further , dysregulation of A9 is associated with various cancers , pulmonary disorders , and Alzheimer’s disease ( Källberg et al . , 2012; Vogl et al . , 2014; Hara et al . , 2012; Obry et al . , 2014; Horvath et al . , 2016; Huang et al . , 2015; Kim et al . , 2009; Averill et al . , 2012 ) . Understanding the mechanisms by which A9 performs its innate immune functions is critical for developing treatments for A9-mediated diseases . The mechanism of A8/A9 antimicrobial activity is well established: it sequesters a variety of transition metals through both a hexahistidine site and a His3Asp site formed at the A8/A9 heterodimer interface , thereby limiting the concentrations of essential microbial nutrients in the extracellular space ( Besold et al . , 2018a; Hadley et al . , 2018; Damo et al . , 2013; Clark et al . , 2016; Besold et al . , 2018b; Nakashige et al . , 2016; Hayden et al . , 2013; Nakashige et al . , 2015; Liu et al . , 2012; Brunjes Brophy et al . , 2013; Brophy et al . , 2012; Baker et al . , 2017; Gagnon et al . , 2015; Nisapakultorn et al . , 2001 ) . Other S100 proteins exert weaker antimicrobial activity via the His3Asp site , which has lower metal-binding affinity and binds fewer types of transition metals than the A8/A9 hexahistidine site ( Nakashige et al . , 2016; Hayden et al . , 2013; Nakashige et al . , 2015; Liu et al . , 2012; Brunjes Brophy et al . , 2013; Bozzi and Nolan , 2020; Cunden and Nolan , 2018 ) . In contrast , the proinflammatory mechanism of A9 is not well understood . A9 acts as a Damage-Associated Molecular Pattern ( DAMP ) , activating NF-κB and other cytokines through Toll-like receptor 4 ( TLR4 ) ( Vogl et al . , 2012; Källberg et al . , 2012; Duan et al . , 2018; Shepherd et al . , 2006; Schiopu and Cotoi , 2013; Laouedj et al . , 2017; He et al . , 2016; Gao et al . , 2015; Tsai et al . , 2014; Lee et al . , 2016; Björk et al . , 2009; Kang et al . , 2015; Chen et al . , 2015 ) . The interaction interface ( s ) , affinity , and stoichiometry for the A9/TLR4 interaction are not known . A small region of A9 has been suggested to form part of the A9/TLR4 binding surface , ( Vogl et al . , 2018 ) but no mutant of A9 has been identified that substantially compromises its activation of TLR4 . An additional layer of A9 immune function is that A9 and A8/A9 are thought to be regulated in the extracellular milieu by proteases . Neutrophils release multiple proteases along with A9 at sites of inflammation that regulate the inflammatory response ( Henry et al . , 2016; Stapels et al . , 2015; Kessenbrock et al . , 2011; Heutinck et al . , 2010; Janoff , 1972; Jerke et al . , 2015 ) . A9 is very susceptible to proteolytic degradation , while A8/A9 is highly resistant ( Figure 1a; Nacken and Kerkhoff , 2007; Riva et al . , 2013 ) . Proteolysis may serve to purge proinflammatory A9 from sites of inflammation and thus selectively enrich for antimicrobial A8/A9 . There may even be a direct , functional link between A9 proteolytic degradation and inflammation . Proteolytic fragments of A9 are sufficient to activate TLR4 , ( Vogl et al . , 2018 ) and proinflammatory stimuli are thought to stabilize A9 homodimers against proteolytic degradation ( Riva et al . , 2013 ) . Directly testing the relationship between A9 proteolytic susceptibility and proinflammatory activity , however , has been challenging . There is no obvious way to selectively increase the proteolytic resistance of A9 and test its effect on A9 activation of TLR4 , making it difficult to understand the relationship , if any , between these two functions . We took an evolutionary biochemical approach to mechanistically dissect the evolution of A9 innate immune functions . Using phylogenetics , ancestral sequence reconstruction ( ASR ) , and biochemical studies , we show that A9s evolved to form proteolytically resistant , antimicrobial A8/A9 complexes in early mammals . We find that A9 homodimers gained proinflammatory activity and lost proteolytic resistance in the ancestor of therian mammals from a weakly proinflammatory , proteolytically resistant amniote ancestor . We identify a pleiotropic substitution that is necessary for A9 activation of TLR4 , sufficient to increase TLR4 activation by the A9 amniote ancestor and played a role in loss of A9 proteolytic resistance . Mutating this site has minimal effect on A8/A9 antimicrobial activity or proteolytic resistance . Lastly , we show that proteolysis is not required for A9 activation of TLR4 . Taken together , this work reveals that mammals concomitantly evolved A8/A9 antimicrobial activity , A9 proinflammatory activity , and a way to selectively regulate A9 inflammation via loss of A9 proteolytic resistance . These findings provide unprecedented mechanistic and evolutionary insight into A9 function and show how a single mutation can have pleiotropic effects in one functional state of a protein while not impacting another , thus facilitating the evolution of multifunctionality .
We sought to determine when A9 evolved to form the antimicrobial A8/A9 complex . We hypothesized that A8/A9 antimicrobial activity evolved in the ancestor of therian mammals ( the shared ancestor of marsupials and placental mammals ) for several reasons . First , the broad-spectrum antimicrobial activity of human and mouse A8/A9 is well established ( Hadley et al . , 2018; Damo et al . , 2013; Clark et al . , 2016; Besold et al . , 2018b; Nakashige et al . , 2016; Hayden et al . , 2013; Nakashige et al . , 2015; Liu et al . , 2012; Brunjes Brophy et al . , 2013; Brophy et al . , 2012; Baker et al . , 2017; Gagnon et al . , 2015; Nisapakultorn et al . , 2001 ) . Second , A9 and A8 genes are only found together in therian mammals ( Figure 1b; Loes et al . , 2018 ) ; therefore the A8/A9 complex could not have arisen earlier than in the ancestor of therian mammals . Lastly , the residues composing the antimicrobial hexahistidine metal-binding site are fully conserved across therian mammals ( Figure 1—figure supplement 1 ) . To determine whether the antimicrobial A8/A9 complex arose in the ancestor of therian mammals , we compared human A8/A9 to two previously uncharacterized A8/A9 complexes . We first tested the antimicrobial activity of A8/A9 from opossum , which is one of the earliest-diverging mammals relative to humans that possesses both of the S100A8 and S100A9 genes . Opossum and human A8 and opossum and human A9 have sequence identities of approximately 50% , respectively ( Figure 1—figure supplement 1 ) . Following previous work , ( Brophy et al . , 2012 ) we produced a cysteine-free variant of the complex to avoid the use of reducing agents in the antimicrobial assay . We confirmed that cysteine-free opossum A8/A9 formed a heterotetramer ( 46 . 8 ± 0 . 7 kDa ) in the presence of calcium – like the human and mouse proteins ( Streicher et al . , 2010 ) – using size exclusion chromatography coupled with multi-angle laser light scattering ( SEC MALS , Figure 1—figure supplement 2 ) . We measured cysteine-free opossum A8/A9 antimicrobial activity against a representative gram-negative bacterium , Staphylococcus epidermidis . In studies of A8/A9 from other species , activity against S . epidermidis tracked with the broad-spectrum antimicrobial activity of the complex ( Hadley et al . , 2018 ) . We assayed activity using a previously established in vitro antimicrobial assay that monitors bacterial growth in the absence or presence of S100 proteins ( Figure 1c , Figure 1—figure supplement 3; Brophy et al . , 2012 ) . To compare the activity of different proteins , we quantified inhibition at seven hours ( Figure 1d ) . We observed a dose-dependent decrease in S . epidermidis growth in the presence of low micromolar concentrations of cysteine-free opossum A8/A9 ( Figure 1c–d , Figure 1—figure supplement 3 ) . The antimicrobial activity of opossum A8/A9 was weaker than that of human A8/A9: opossum A8/A9 delayed bacterial growth , while human A8/A9 both delayed growth and decreased bacterial carrying capacity ( Figure 1c ) . It was previously found that cysteine-free human A8/A9 was potently antimicrobial , ( Brophy et al . , 2012 ) while cysteine-free mouse A8/A9 exhibits weaker antimicrobial activity than wildtype mouse A8/A9 ( Hadley et al . , 2018 ) . To determine whether the weaker activity of opossum A8/A9 was due to the removal of cysteines , we also measured the activity of wildtype opossum A8/A9 against S . epidermidis . We found that wildtype opossum A8/A9 had higher activity than cysteine-free opossum A8/A9 over the concentration range tested ( Figure 1—figure supplement 3 ) . This suggests that the cysteines present in mouse and opossum A8/A9 play a role in their antimicrobial activity , unlike in human A8/A9 . The antimicrobial activity of the opossum A8/A9 complex thus appears to be more similar to that of mouse A8/A9 than human A8/A9 . The shared antimicrobial activity of human , mouse , and opossum A8/A9 strongly suggests that the antimicrobial A8/A9 complex evolved in the ancestor of therian mammals . To test this further , we measured the antimicrobial activity of ancestrally reconstructed therian mammalian A8/A9 ( ancA8/A9 – Figure 1b ) . We used our previously published phylogenetic tree ( Loes et al . , 2018 ) consisting of 172 S100 sequences to reconstruct therian mammalian ancestral A8 and A9 ( ancA9 and ancA8 – Figure 1 , supplementary files 1–2 ) , which were used to form the ancA8/A9 complex . AncA8 and ancA9 had average posterior probabilities of 0 . 88 and 0 . 83 , with sequence similarities to human A8 and A9 of 66% and 64% , respectively ( Figure 1—figure supplement 4 ) . Average posterior probabilities in this range have been previously described as medium confidence reconstructions , with reconstructions characterized by others having average posterior probabilities as low as 0 . 7 ( Eick et al . , 2016 ) . We confirmed that each protein was folded and had secondary structure content similar to that of human A8/A9 using far-UV circular dichroism ( CD ) spectroscopy ( Figure 1—figure supplement 5 ) . We then measured the antimicrobial activity of ancA8/A9 against S . epidermidis . We observed a potent reduction in S . epidermidis growth comparable to that of human A8/A9 ( Figure 1c–d , Figure 1—figure supplement 3 ) . To test for the robustness of this finding to phylogenetic uncertainty , we also tested the antimicrobial activity of an AltAll ( Eick et al . , 2016 ) reconstruction of ancA8/A9 against S . epidermidis ( altancA8/A9 , Figure 1c–d , Figure 1—figure supplement 4 ) . In this reconstruction , we swapped all ambiguously reconstructed amino acid positions for their second-most likely state ( see methods ) . AncA8/A9 and altancA8/A9 differ by 27 amino acids total ( 10 between ancA8 and altancA8 and 17 between ancA9 and altancA9 - Figure 1—figure supplement 4 ) . AltancA8/A9 exhibited antimicrobial activity against S . epidermidis similar to opossum A8/A9: it delayed growth but did not ultimately limit bacterial carrying capacity . While the hexahistidine site residues are conserved in ancA8/A9 and altancA8/A9 ( Figure 1—figure supplement 1 ) , it appears that a subset of the ambiguously reconstructed 27 residues are important for A8/A9 antimicrobial activity , perhaps affecting the orientation and/or affinity of the hexahistidine metal-binding site . Taken together , the antimicrobial activity of modern mammalian A8/A9 complexes ( human , mouse , and opossum ) and the antimicrobial activity of the reconstructed ancA8/A9 complex suggest that A9s evolved to form the antimicrobial A8/A9 complex in the ancestor of mammals . We next sought to determine when A9s evolved potent proinflammatory activity via activation of TLR4 . Our previous work revealed that human A9 potently activates not only human TLR4 in functional assays , but also opossum and chicken TLR4 ( Figure 2a; Loes et al . , 2018 ) . In contrast , chicken MRP126 , the sauropsid ortholog of A9s , was found to be a weak activator of all TLR4s , including chicken TLR4 . Both human and opossum A9 activate chicken TLR4 better than chicken MRP126 does . Two possibilities are consistent with these observations . Either mammalian A9s evolved enhanced proinflammatory activity from a less active amniote ancestral state , or A9s maintained a potent ancestral activity that was lost by chicken MRP126 . To differentiate between these two possibilities , we determined the ancestral proinflammatory activity of these proteins . We used ASR to reconstruct the shared amniote ancestor of A9s , A8s , A12s , and MRP126s . This group of proteins is known collectively as the ‘calgranulins’ , so we will refer to this ancestral protein as ancCG ( ancestor of calgranulins ) . We also constructed an alternate , ‘alt All’ version of this ancestor ( altancCG , supplementary file 1 ) , which differed from ancCG by eight amino acids . The average posterior probability of ancCG was 0 . 86 ( Figure 1—figure supplement 4 ) . We also expressed and purified ancA9 and altancA9 – the A9 subunits from the ancestral A8/A9 complexes described above . We confirmed that each protein was folded and had secondary structure content similar to that of modern S100s using far-UV CD spectroscopy ( Figure 1—figure supplement 5 ) . We then tested modern and ancestral S100s for activity against human TLR4 . Following previous work , ( Vogl et al . , 2007; Loes et al . , 2018; Anderson et al . , 2019; Ehrhardt et al . , 2006 ) we transiently transfected HEK293T cells with plasmids encoding TLR4 and its species-matched cofactors MD-2 and CD14 , added purified S100 proteins to the growth media , and then measured output of luciferase under control of an NF-κB promoter . Consistent with previous results , ( Loes et al . , 2018 ) we found that human A9 potently activated human TLR4 , resulting in high levels of NF-κB production ( Figure 2b , Figure 2—figure supplements 1–2 ) . Human A8/A9 and opossum A9 exhibited much weaker activity against human TLR4 . Lastly , we tested ancA9 , altancA9 , ancCG , and altancCG for activity against human TLR4 and observed weak or no activation for each ancestral protein . This result is unsurprising , as we previously found that human TLR4 is more specific than other amniote TLR4s: human TLR4 is activated much more potently by human A9 than by any other S100 protein ( Figure 2a–b; Loes et al . , 2018 ) . In contrast , TLR4s from other species ( mouse , opossum , and chicken ) appear to be more promiscuous and can be activated similarly by S100s from various species ( Figure 2a; Loes et al . , 2018 ) . This is consistent with lineage-specific coevolution between human TLR4 and human A9 – a confounding variable that makes assessment of ancestral S100 protein proinflammatory activity difficult using human TLR4 . We predicted that opossum TLR4 would be a better protein to probe ancestral S100 proinflammatory function because opossum TLR4 is broadly activated by A9s across mammals and gives little indication of lineage-specific coevolution ( Loes et al . , 2018 ) . We therefore tested the proinflammatory activity of ancA9 , ancCG , and their corresponding alternate reconstructions against opossum TLR4 . Corroborating previous results , human A9 strongly activated opossum TLR4 , while opossum A9 activity was approximately half that of human A9 ( Figure 2c , Figure 2—figure supplements 1–3 ) . AncA9 and altancA9 activated opossum TLR4 to the same extent as opossum A9 . AncCG and altancCG were the weakest activators of opossum TLR4 , with activity approximately 25% or less than that of human A9 . These findings suggest that A9s evolved enhanced proinflammatory activity early in mammals from a weakly proinflammatory amniote ancestor , while A8/A9s and chicken MRP126 maintained weak , ancestral proinflammatory activity . We next sought to determine when the differential proteolytic susceptibility of A9 and A8/A9 evolved . We used a simple in vitro assay to monitor S100 protein degradation over time in the presence of proteinase K , a potent non-specific serine protease ( Figure 3a ) . Proteinase K was chosen both because of its low specificity and to mimic other serine proteases that A9 and A8/A9 encounter when released from neutrophils during an inflammatory response ( Henry et al . , 2016; Stapels et al . , 2015; Kessenbrock et al . , 2011; Heutinck et al . , 2010; Fu et al . , 2018 ) . Proteolytic decay rates were estimated by fitting a single exponential decay function to the data ( Figure 3b , Figure 3—figure supplements 1–4 ) . Human A8/A9 has been described as extremely resistant to proteases ( Nacken and Kerkhoff , 2007 ) ; however , it has not been compared to S100 proteins besides human A8 and A9 . To establish a baseline expectation for S100 protein proteolytic resistance , we characterized the proteolytic resistance of a broad set of human S100s against proteinase K . As previously shown , ( Nacken and Kerkhoff , 2007 ) human A9 and A8 alone were rapidly proteolytically degraded , while the human A8/A9 complex exhibited strong resistance ( Figure 3c , Figure 3—figure supplements 1–2 ) . Under our conditions , the degradation rates for human A8 and A9 were approximately three orders of magnitude faster than that of the human A8/A9 heterocomplex . We then characterized closely related protein human S100A12 ( A12 ) , the chicken ortholog MRP126 , and six distantly related human S100s ( Wheeler et al . , 2016 ) . Human A12 , chicken MRP126 , and five out of six more distantly related human S100s exhibited intermediate to strong proteolytic resistance , each degrading 1–2 orders of magnitude slower than human A8 or A9 but , on average , one order of magnitude faster than human A8/A9 ( Figure 3c , Figure 3—figure supplements 1–2 ) . Notably , human A12 and chicken MRP126 formed predominantly homodimers by SEC MALS under these conditions ( Figure 1—figure supplement 2 ) , indicating that higher-order oligomerization ( >2 subunits ) isn’t required for S100 proteolytic resistance . Lastly , human A14 degraded faster than A9 or A8 . This protein is evolutionarily distant ( Wheeler et al . , 2016 ) and therefore likely reflects independent evolution of this property . Taken together , these data show that the A8/A9 complex , A9 , and A8 indeed fall at the extremes of human S100 proteolytic resistance; human A9 and A8 are among the fastest-degrading S100s tested , while human A8/A9 is one of the slowest . To test whether A9 and A8 proteolytic susceptibility and A8/A9 resistance are conserved across mammals , we characterized mouse and opossum A9 , A8 , and A8/A9 for proteolytic resistance . Mouse A9 and A8 were found to be highly proteolytically susceptible and mouse A8/A9 strongly proteolytically resistant , matching the pattern observed for their human counterparts ( Figure 3c and Figure 3—figure supplement 2 ) . Opossum A9 and A8 were also highly proteolytically susceptible , while opossum A8/A9 was resistant ( Figure 3c , Figure 3—figure supplement 2 ) . This indicates that the susceptibility of A9s and A8s and the resistance of A8/A9 complexes is conserved across mammals . When mapped onto the S100 phylogeny , the most parsimonious explanation for these data is that the shared amniote ancestor—ancCG—was proteolytically resistant ( Figure 3c ) . In this scenario , A12s , MRP126s , and A8/A9s conserved ancestral resistance , while A9s and A8s independently lost resistance early in mammals . Alternatively , ancCG could have been proteolytically susceptible . This would mean that A9s and A8s maintained an ancestral susceptibility , while MRP126s , A12s , and A8/A9s each evolved novel proteolytic resistance . To distinguish between these possibilities , we characterized ancestrally reconstructed S100s for proteolytic resistance . AncCG and altancCG exhibited extremely high proteolytic resistance ( Figure 3d , Figure 3—figure supplement 3 ) , with degradation rates 3–4 orders of magnitude slower than modern A8s or A9s and approximately one order of magnitude slower than modern A8/A9s . AncA8/A9 and altancA8/A9 also demonstrated high proteolytic resistance , with degradation rates approximately 2–3 orders of magnitude slower than A8s and A9s and comparable to modern A8/A9 complexes ( Figure 3 ) . Together , these data paint a consistent picture: the amniote ancestor of A9s , ancCG , was strongly resistant to proteolytic degradation . Modern A9s and A8s lost proteolytic resistance from an ancestrally resistant state , while modern A12s , A8/A9 complexes , and MRP126s maintained the ancestral proteolytic resistance ( Figure 3e ) . Finally , we sought to better resolve when A9s acquired proteolytic susceptibility . We hypothesized that this occurred in the ancestor of mammalian A9s before the divergence of therian mammals and marsupials . To test this hypothesis , we measured the proteolytic susceptibility of therian mammalian ancA9 and found that it degraded rapidly ( Figure 3 ) . However , its alternative reconstruction ( altancA9 ) , was slow to degrade , with a rate two orders of magnitude slower than ancA9 and comparable to other highly resistant S100s . Because the descendants of ancA9 all exhibit proteolytic susceptibility ( Figure 3 ) , the simplest explanation is that altancA9 is a low-quality reconstruction that does not capture the properties of the historical protein . Alternatively , proteolytic susceptibility could have been independently acquired along marsupial and placental mammal lineages . We found above that A9 evolved to form the antimicrobial A8/A9 complex , gained potent proinflammatory activity , and lost proteolytic resistance over the narrow evolutionary interval after the divergence of mammals and sauropsids but before the divergence of placental mammals and marsupials . We next sought to determine how A9 evolved its antimicrobial and proinflammatory activities and lost proteolytic resistance . The mechanism by which A9 evolved to form the antimicrobial A8/A9 complex is straightforward . After ancCG duplicated , additional histidines accumulated in the mammalian A8 and A9 ancestors that created the antimicrobial hexahistidine metal-binding site in the A8/A9 complex ( Figure 3e and Figure 1—figure supplement 1 ) . A8s acquired one additional histidine while retaining the three histidines present in ancCG , while A9s acquired two additional histidines via acquisition of a C-terminal extension ( Figure 1—figure supplement 1 ) . While A9s evolved five of the six histidines composing the hexahistidine metal-binding site , this was not sufficient to convey potent antimicrobial activity ( Figure 1c ) . Instead , preservation of A8/A9 heterocomplex formation resulted in proper assembly of the complete antimicrobial hexahistidine site early in mammals . The quantitative difference between ancA8/A9 and altancA8/A9 antimicrobial activity suggests that other amino acid changes tuned the antimicrobial activity of the molecule , but the core functionality is determined by whether the six histidine residues were present . This is independently supported by Brunjes Brophy et al . , who showed that mutating the two C-terminal histidines in A9 is sufficient to strongly decrease the A8/A9 complex’s antimicrobial activity ( Brunjes Brophy et al . , 2013 ) . The mechanisms by which A9s gained proinflammatory activity and lost proteolytic resistance are less obvious , particularly because the mechanism by which A9 activates TLR4 is not well understood . We reasoned that we could identify functionally important amino acid substitutions by focusing on the evolutionary interval over which these properties evolved . We therefore compared the sequences of ancCG ( weakly proinflammatory and resistant to proteolytic degradation ) and ancA9 ( potently proinflammatory and susceptible to proteolytic degradation ) . We further narrowed down sequence changes of interest by looking for residues conserved in modern A9s ( Figure 4—figure supplement 1 ) . Finally , we focused on amino acid changes in helix III of A9 , as this region is thought to be important for A9 activation of TLR4 based on in vitro binding studies and in silico docking studies ( Vogl et al . , 2018 ) . Only one historical amino acid substitution met all three criteria: position 63 ( human A9 numbering ) . This residue is a phenylalanine in both ancCG and altancCG , is conserved as a phenylalanine in 95% of modern A8s and A12s and has been substituted for a methionine or leucine ( M/L ) in 97% of A9s ( Figure 4a ) . We hypothesized that reverting this site to its amniote ancestral state—M63F—might affect A9 proinflammatory activity . We mutated this position to a phenylalanine in human A9 and opossum A9 and tested each protein for TLR4 activation . Strikingly , we found that introducing M63F into human A9 severely compromised its ability to activate human TLR4 ( Figure 4b , Figure 2—figure supplement 2 ) . This was also true for opossum A9: introduction of M63F ( human numbering ) strongly decreased opossum A9 activation of opossum TLR4 ( Figure 4c , Figure 2—figure supplement 3 ) . We next introduced the forward substitution , F63M , into ancCG and tested its proinflammatory activity against opossum TLR4 . We observed a modest increase in ancCG activity with the F63M substitution , with activity comparable to that of opossum A9 ( Figure 4c , Figure 2—figure supplement 3 ) . For most proteins we studied , the amino acid at position 63 did indeed play an important role in determining the pro-inflammatory activity of A9 . The effects of toggling position 63 between Met and Phe were not , however , universal . We introduced M63F into ancA9 and observed no change in proinflammatory activity ( Figure 4c , Figure 2—figure supplement 3 ) . Further , altancA9 has a Phe at position 63 but activates TLR4 in the assay ( Figure 2c , Figure 2—figure supplement 3 ) . Thus , while position 63 is an important contributor to activity in modern A9s , other substitutions were also important for the transition from a weakly pro-inflammatory ancestor to the modern set of potently pro-inflammatory A9s . Because A9s lost proteolytic resistance and gained proinflammatory activity over the same evolutionary time interval , we reasoned that the F63M substitution might have also played a role in A9 loss of proteolytic resistance . To test this , we characterized the proteolytic resistance of human A9 M63F and ancA9 M63F . Strikingly , reversion of this single mutation rendered both ancA9 and human A9 strongly resistant to proteolytic degradation , decreasing their respective degradation rates by 1–2 orders of magnitude and approaching the degradation rates of ancCG and various A8/A9 complexes ( Figure 4d , Figure 3—figure supplement 4 ) . To relate these findings to proteases that A9 might encounter at sites of inflammation , we also measured the proteolytic resistance of human A9 and human A9 M63F against two neutrophil-specific proteases – cathepsin G and neutrophil elastase ( Figure 4—figure supplement 2 ) . Neutrophils release these proteases along with A9 at sites of inflammation , often through Neutrophil Extracellular Traps ( NETs ) ( Henry et al . , 2016; Stapels et al . , 2015; Kessenbrock et al . , 2011; Heutinck et al . , 2010; Janoff , 1972; Jerke et al . , 2015; O'Donoghue et al . , 2013 ) . We found that M63F decreased the rate of human A9 degradation in the presence of cathepsin G and neutrophil elastase in vitro by approximately one order of magnitude , matching our results using proteinase K ( Figure 4—figure supplement 2 ) . Lastly , we tested the effect of the forward mutation – F63M – on ancCG proteolytic resistance . We observed no change in resistance for ancCG F63M , indicating that additional substitutions were required to render ancA9 proteolytically susceptible . Together these data show that a single historical reversion is sufficient to render A9s proteolytically resistant , indicating that this position played a role in the loss of A9 proteolytic resistance early in therian mammals . A primary goal of this study was to understand the role of pleiotropy in the evolution of multifunctionality . M63F clearly has pleiotropic effects on A9 , altering both its proinflammatory activity and proteolytic resistance ( Figure 4b–d ) . We next asked whether introducing M63F would pleiotropically affect the antimicrobial A8/A9 complex . Position 63 is somewhat distant from the A8/A9 interface and the antimicrobial hexahistidine site ( ~10 Å in the manganese-bound A8/A9 crystal structure ) ( Damo et al . , 2013 ) ; we therefore hypothesized that M63F should not affect A8/A9 complex formation or function . To test this hypothesis , we introduced M63F into human A8/A9 and tested it for oligomeric state , proteolytic resistance , and antimicrobial activity against S . epidermidis . As predicted , human A8/A9 M63F predominantly formed a heterotetramer in the presence of calcium by SECMALS with a molecular weight similar to that of wildtype human A8/A9 ( 48 . 7 ± 4 . 2 kDa – Figure 1—figure supplement 2 ) . We found that human A8/A9 M63F was also strongly resistant to proteolytic degradation , similar to human A8/A9 ( Figure 4d ) . Lastly , M63F had minimal impact on human A8/A9 antimicrobial activity against S . epidermidis , retaining potent antimicrobial activity ( Figure 4e ) . In contrast , neither human A9 nor human A9 M63F were antimicrobial against S . epidermidis ( Figure 4e ) . These findings suggest that this single amino acid position had important effects on the evolution of A9 activation of TLR4 and loss of proteolytic resistance without significantly impacting A8/A9 oligomeric state , proteolytic resistance , or antimicrobial activity . We next asked what effect M63F has on the biophysical properties of human A9 . Residue 63 sits in the middle of helix III of A9 , pointing inward toward helix II , and is neither a core residue nor fully surface-exposed ( Figure 5a; Itou et al . , 2002 ) . Based on the published structure of human A9 , ( Itou et al . , 2002 ) a Phe at position 63 could be plausibly tolerated without a steric clash . Using circular dichroism ( CD ) spectroscopy , we found that the bulk secondary structure content of human A9 M63F was similar to that of hA9 ( Figure 5b ) . We measured the oligomeric state of human A9 M63F by SEC MALS and found that it predominantly forms a homodimer in solution similarly to human A9 , with no detectable monomers or larger oligomers ( Figure 5c and f ) . These data together indicate that M63F does not significantly alter human A9’s secondary structure or oligomeric state . We then examined whether M63F alters the stability of human A9 . We measured equilibrium unfolding curves for human A9 and human A9 M63F using CD spectroscopy and chemical denaturation via urea . We found that M63F appears to stabilize human A9 , increasing the apparent free energy of unfolding by more than four kcal/mol and shifting the Cm by ~2M urea ( Figure 5d and f , Figure 5—figure supplement 1 ) . We also measured the unfolding kinetics of human A9 and human A9 M63F in the presence of calcium by spiking protein directly into 6M guanidinium hydrochloride ( gdn-HCl ) denaturant and monitoring its unfolding rate by CD spectroscopy . Strikingly , human A9 M63F takes several minutes to unfold under these conditions , while human A9 unfolds immediately within the dead time of the experiment ( Figure 5e–f , Figure 5—figure supplement 2 ) . We note that the folding pathway for A9 is complex and almost certainly not two-state—calcium binding , monomer folding , and dimerization all contribute—and thus we cannot reliably determine how M63F affects the stability of each of these potential folding intermediates . The large increase in apparent stability and unfolding rate suggests , however , that the mutation stabilizes some aspect of the folded structure . The work above identified a mutation that , when introduced into human A9 , increases the stability of the protein while also potently compromising its ability to activate TLR4 . The mutation is not at a surface position and is therefore not likely a direct participant in the A9/TLR4 protein/protein interface . Further , the same mutation dramatically decreases the proteolytic susceptibility of the protein . One simple way to explain these observations would be if the proteolytic susceptibility itself was the feature that evolved to allow activation of TLR4 . This would be consistent with a previous observation that proteolytic products of A9 activate TLR4 ( Vogl et al . , 2018 ) . To test whether proteolysis itself was sufficient for activity , we engineered an alternate variant of A9 that was proteolytically susceptible . We introduced the M63A mutation into human A9 , anticipating that the short alanine sidechain would not have the stabilizing effect of M63F . As expected , human A9 M63A was highly susceptible to proteolytic degradation , similar to wildtype human A9 ( Figure 6a , Figure 3—figure supplement 4 ) . We reasoned that if proteolysis is the primary determinant of A9 activation of TLR4 , then proteolytically susceptible human A9 M63A should potently activate TLR4 . Human A9 M63A , however , exhibited diminished proinflammatory activity , similar to human A9 M63F ( Figure 6b , Figure 2—figure supplement 2 ) . This indicates that the methionine at position 63 is important for A9 activation of TLR4 . Further , we quantified the amount of human A9 , human A9 M63F , and human A9 M63A before and after measuring TLR4 activity and observed no decrease in the amount of full-length protein remaining for wildtype human A9 or either mutant by western blot ( Figure 6c ) . This indicates that A9 is not digested by extracellular proteases over the course of the ex vivo assay and that proteolysis is not necessary for A9 activation of TLR4 . Although proteolysis does not appear to be a requirement for TLR4 activation , this does not rule out that proteolysis could increase A9 proinflammatory activity by releasing proinflammatory fragments of A9 . To test for this possibility , we treated human A9 with agarose-immobilized proteinase K for increasing amounts of time , removed the protease , and then measured the proinflammatory activity of A9 degradation products ( Figure 6d ) . If proteolytic products of A9 are the most proinflammatory form of the protein , we might expect to observe a spike in TLR4 activation upon A9 digestion . Instead , we observed a steady decrease in human A9 activity with increasing digestion time . This suggests that full-length human A9 is the most potent activator of TLR4 . We did observe moderate activity for proteolytic products of human A9 , as previously shown ( Vogl et al . , 2018 ) . After 30 min of digestion , no detectable full-length A9 remains by western blot ( <30 ng , Figure 6d ) , but NF-κB production is still quite high , revealing that smaller fragments of A9 are sufficient to provide some degree of activation of TLR4 . This raised the possibility that part of M63F’s deleterious effect on proinflammatory activity could be to limit the release of active proteolytic fragments of A9 . To test this , we also measured human A9 M63F activation of TLR4 after digestion for multiple hours ( Figure 6d ) . Unlike wildtype , however , fragments of human A9 M63F did not activate TLR4—even after being liberated by the protease . This strongly suggests that the historical mutation induced a change in the native structure or dynamics of the molecule to bring about increased activity , independent of its effect on proteolytic susceptibility .
Our data suggest that the proinflammatory and antimicrobial activities of A9 and the A8/A9 complex have undergone further optimization in placental mammals since these functions evolved . While the histidines composing the high-affinity metal-binding site of A8/A9 complexes are conserved , we observed differences in antimicrobial potency for different A8/A9 complexes . In particular , human A8/A9 is one of the most potently antimicrobial A8/A9 complexes characterized . This suggests that further optimization of the metal-binding site has occurred in along the human lineage within mammals . We also observed differences in activation of TLR4 by different A9s—human A9 is a potent , promiscuous activator of TLR4s from multiple species , while earlier-diverging A9s and other S100s exhibit weaker proinflammatory activity ( Loes et al . , 2018 ) . Future studies are necessary to understand how , mechanistically , later-diverging A9s and A8/A9 complexes have optimized these critical innate immune functions . While proteolysis is not required for A9 activation of TLR4 , it remains unclear why A9s lost proteolytic resistance . We suggest three possibilities . The first is that loss of proteolytic resistance in A9s was simply a byproduct of evolving proinflammatory activity . No A9 characterized in this study , with the exception of the alternate reconstruction of ancA9 , is both proteolytically resistant and potently proinflammatory . This indicates that the molecular requirements for A9 proteolytic resistance may be incompatible with those required for A9 activation of TLR4: A9s may have gained proinflammatory activity at the expense of proteolytic resistance . A second possibility is that A9 proteolytic susceptibility is being maintained to actively remove proinflammatory A9 from the cell and retain the antimicrobial A8/A9 complex . The last possibility for A9 loss of proteolytic resistance is adaptive constraint . There could be selection for some property of A9 or A8/A9 that we did not measure that is incompatible with A9 proteolytic resistance . While we cannot explicitly distinguish between each of these possibilities , the end result is that A9s lost proteolytic resistance from a resistant ancestor . As A9s activate TLR4 in the protease-rich extracellular space , the functional result of A9 loss of proteolytic resistance is that A9s evolved a proteolytic ‘timer’ concomitantly with evolving proinflammatory activity , all without affecting A8/A9 function . Our findings suggest new directions for understanding how A9 potently activates TLR4 . TLR4-driven inflammation has been the focus of intense study for over 20 years , ( Källberg et al . , 2012; Laouedj et al . , 2017; He et al . , 2016; Gao et al . , 2015; Lee et al . , 2016; Anderson et al . , 2019; Ibrahim et al . , 2013; Nagai et al . , 2002; Poltorak et al . , 1998; Prince et al . , 2011 ) and the structural basis of TLR4 activation by exogenous agonists , such as the bacterial cell wall component lipopolysaccharide ( LPS ) , is well understood ( Park et al . , 2009 ) . In contrast , little is known about how A9 activates TLR4 . We have shown here that proteolytic degradation appears dispensable for activation; however , smaller fragments of the protein are sufficient activate TLR4 ( Figure 6 ) . Given the effect of mutating position 63 on A9 proinflammatory activity , we propose that the region surrounding it—helix III—is important for activity . This is independently supported by Vogl et al . , who identified four pairs of double mutants within helix III ( amino acids 64 , 65 , 73 , and 77 ) that , when mutated to alanines in pairs , decrease A9 binding to TLR4 in vitro ( Vogl et al . , 2018 ) . Biophysical characterization of hA9 M63F ( Figure 5 ) indicates that it is more stable and unfolds more slowly , yet it maintains its bulk secondary structure and oligomeric state . The simplest explanation for these data is that M63F is affecting some functionally important dynamic process of the protein , possibly mediated by helix III , that is critical for A9 activation of TLR4 . The proteolytic susceptibility of A9s also supports this hypothesis , as proteolysis is a dynamic process that often relies on substrate flexibility and local unfolding events to proceed ( Guharoy et al . , 2016; Fontana et al . , 1997; Hubbard , 1998; Ottesen , 1967; Imoto et al . , 1986 ) . Damage-Associated Molecular Patterns ( DAMPs ) often interact with their targets via hydrophobic surfaces ( Garg et al . , 2010; Rubartelli and Lotze , 2007; Bianchi , 2007 ) ; one possibility is that A9 undergoes a local unfolding event that exposes a hydrophobic surface to interact with TLR4 . This would mean that studies of the native structure of A9 might not be sufficient to gain mechanistic understanding of how it activates TLR4 . Further work is required to understand the nature of the active functional state of A9 . Finally , our results suggest a positive role for pleiotropy in the evolution of protein function . Pleiotropy is often viewed as a constraint on evolution: as functional complexity is added to a polypeptide sequence , it becomes increasingly challenging to introduce substitutions—and new functions—without perturbing existing ones ( Lee et al . , 2018; Stearns , 2010; Paaby and Rockman , 2013; He and Zhang , 2006; Pavličev and Cheverud , 2015; Hirano , 1999 ) . Here , however , we find a single mutation that had beneficial pleiotropic effects on two important properties of A9: proinflammatory activity and proteolytic susceptibility . If A9 evolved potent proinflammatory activity without gaining susceptibility , it could potentially overstimulate inflammation simply by lingering in the extracellular milieu . Since both properties evolved at once , however , mammals evolved a proinflammatory molecule with a built-in ‘timer’: they gained a new inflammatory signal while avoiding potentially deleterious effects . This shows how pleiotropy can positively contribute to the evolution of new functions . This same mutation , in contrast , had little pleiotropic effect on another functional state of A9: the A8/A9 complex . The antimicrobial activity of the A8/A9 complex was insulated from any pleiotropic effects from the mutation because proinflammatory and antimicrobial activities were partitioned between A9 and the A8/A9 complex , respectively . A mutation arose in the A9 amino acid sequence and is thus present in both A9 and A8/A9 states , but we only observe effects on the A9 state . This shows that pleiotropic constraint can be reduced when protein functions are partitioned amongst different protein states . These findings reveal the diversity of pleiotropic roles that a single mutation can play . It further shows how the deleterious pleiotropic effects of mutations can be reduced by partitioning protein functions and properties into different functional states , thus enabling the acquisition , optimization , and expansion of new protein functions . Given the vast diversity of protein functional domains and protein-protein interactions in biology , we suspect that this is a common occurrence in the evolution of protein multifunctionality .
We reconstructed ancestral sequences using a previously published a phylogenetic tree of S100 proteins containing 172 sequences from 30 amniote taxa ( Supplementary files 1–2; Loes et al . , 2018 ) . We used PAML4 to generate maximum likelihood ancestors ( marginal probability method ) ( Yang et al . , 1995; Yang , 2007 ) using the previously-identified maximum likelihood ( ML ) substitution model ( LG+Γ8 ) ( Jones et al . , 1992 ) on the ML tree . To account for reconstruction uncertainty , we also generated ‘altAll’ versions of each ancestor ( Eick et al . , 2016 ) . We took every site in which the alternate reconstruction had a posterior probability >0 . 20 and substituted that amino acid into the maximum-likelihood ancestor . These alternate reconstructions had an average of 12 sequence differences relative to the maximum-likelihood ancestors ( Figure 1—figure supplement 4 ) . They represent a ‘worst case’ reconstruction relative to our best , maximum likelihood reconstruction . We also investigated the effect of topological uncertainty on our reconstructed ancestors . In the published phylogenetic analysis , A8s , A9s , A12s , and MRP126s all formed distinct and well-supported clades; however , the branching pattern between these four clades could not be resolved with high confidence ( Loes et al . , 2018 ) . To explore how this uncertainty altered our reconstructed ancestral proteins , we constructed all 15 possible topologies for the A8 , A9 , A12 , and MRP126 clades—i . e ( ( A8 , A9 ) , ( A12 , MRP126 ) ) , ( ( A8 , A12 ) , ( MRP126 , A9 ) ) , etc . —while maintaining species-corrected , within-clade topologies . We then optimized the tree branch lengths and substitution rates for each tree using PhyML ( Guindon et al . , 2010 ) . Finally , we used PAML to reconstruct ancA9 , ancCG , and ancA8 for all 15 possible arrangements of the MRP126 , A12 , A8 , and A9 clades . The average number of sequence differences for ancestors reconstructed using different topologies was less than or equal to the number of sequence differences between the ML and altAll reconstructions ( Figure 1—figure supplement 4 ) . Further , the sites that differed were a subset of those that differed between the ML and altAll reconstructions . Thus , the altAll reconstructions account for sequence changes due to both uncertainty given the ML tree and uncertainty due to topological uncertainty . All S100 genes in this study were purchased as synthetic constructs in pUC57 vectors from Genscript . S100 genes ( A8s , A9s , A12s , MRP126s , and ancestrally reconstructed genes ) were sub-cloned into a pETDuet-1 ( pD ) vector ( Millipore ) . A8s , A12s , MRP126s , and ancCGs were cloned into multiple cloning site #1 ( MCS1 ) of the pD vector , while A9s were cloned into MCS2 . For expression and purification of A8/A9 heterocomplexes ( A8/A9s ) , pD plasmids containing an A8 gene in MCS1 and an A9 gene in MCS2 were used as previously described ( Futami et al . , 2016 ) . Opossum A8 was sub-cloned into an MBP-LIC vector to yield a His-MBP-TEV-opA8 construct . For opossum A8/A9 , the entire His-MBP-TEV-A8 construct was then sub-cloned into MCS1 of a pD vector containing a marsupial A9 in MCS2 . Other S100s ( A1 , A5 , A7 , A11 , A14 , and P ) were previously cloned into a pET28/30 vector to yield a TEV-cleavable N-terminal His tag ( Wheeler et al . , 2016 ) . Cysteine-free versions of all S100 genes , as well as point mutants , were prepared using site-directed mutagenesis ( Agilent ) . Recombinant protein overexpression was conducted in E . coli BL21 ( DE3 ) pLysS Rosetta cells . Cultures were innoculated in luria broth overnight at 37°C , shaking at 250 rpm , in the presence of ampicillin and chloramphenicol . The following day , 10 ml of saturated culture was diluted into 1 . 5 L of media with antibiotics , grown to OD600 = 0 . 6–1 , and then induced overnight at 16°C using 1 mM IPTG . Cells were pelleted at 3 , 000 rpm for 20 min and stored at −20°C for no more than three months . Lysates were prepared by vortexing pellets ( 3–5 g ) in tris buffer ( 25 mM tris , 100 mM NaCl , pH 7 . 4 ) and incubating for 20 min at RT with DNAse I and lysozyme ( ThermoFisher Scientific ) . Lysates were sonicated and cell debris was pelleted by centrifugation at 15 , 000 rpm at 4°C for >20 min . All proteins were purified on an Äkta PrimePlus FPLC using various 5 ml HiTrap columns ( HisTrap FF ( Ni-affinity ) , Q HP ( anion exchange ) , SP FF ( cation exchange ) , and MBPTrap HP ( MBP ) - GE Health Science ) . A1 , A5 , A7 , A11 , A14 , and S100P were purified using a a TEV-cleavable His tag strategy used by our lab previously ( Lee et al . , 2018; Postel et al . , 2010; Gudmundsson and Agerberth , 1999 ) . All other S100s , except for opossum A8 and opossum A8/A9 , were purified in three steps using Ni-affinity chromatography in the presence of calcium followed by two rounds of anion exchange chromatography at different pHs . For Ni-affinity chromatography , proteins were eluted over a 50 ml gradient from 25 to 1000 mM imidazole in tris buffer . Peak elution fractions were pooled and placed in dialysis overnight at 4°C in 4 L of tris buffer ( calcium-free ) adjusted to pH 8 . Anion exchange chromatography was then performed the following day over a 50 ml gradient from 100 to 1000 mM NaCl in pH eight tris buffer . Fractions containing majority S100 were pooled and analyzed for purity on an SDS-PAGE gel . If trace contaminants remained , an additional anion exchange step was performed at pH six using the same elution strategy as for the previous anion exchange step . Opossum A8 and A8/A9 lysates were prepared as above and then flowed over a nickel column , eluting over a 50 ml gradient from 25 to 1000 mM imidazole in tris buffer . Peak elution was pooled and the MBP tag was cleaved by incubation with ~1:5 TEV protease at 4°C overnight in 4 L of tris buffer . The MBP tag was then removed by flowing the sample over an MBPTrap column , step-eluting with 10 mM maltose . Additional MBP columns were run until all MBP was removed from the purified protein , assessed by SDS-PAGE . If necessary , an additional anion exchange step at pH eight was performed to complete purification . All purified proteins were dialyzed overnight at 4°C in tris buffer + 2 g/L Chelex-100 resin ( Biorad ) , flash-frozen the following day in liquid nitrogen , and stored at −80°C . For all experiments , protein aliquots were thawed fresh from freezer stocks and were either dialyzed in the appropriate experimental buffer overnight at 4°C or exchanged 3X into experimental buffer using 3K microsep spin concentrator columns ( Pall Corporation ) . All samples were filter-sterilized using 0 . 1 µm spin filters ( EMD Millipore ) prior to measuring concentration and using in experiments . Thawed aliquots were used for no more than one week before discarding . All concentrations were measured by Bradford assay and correspond to micromolar dimeric protein . For in vitro proteolytic susceptibility experiments , proteins were dialyzed or exchanged into tris buffer + 1 mM CaCl2 . 12 . 5 µM S100 protein was treated with 5 µM monomeric Proteinase K from Tritirachium album ( Sigma Aldrich ) , cathepsin G from human neutrophils ( Athens Research ) , or neutrophil elastase from human neutrophils ( Millipore Sigma ) in thin-walled PCR tubes , which were held at a constant temperature of 25°C over the course of the experiment using a thermal cycler . Proteinase K activity was quenched at different time points by directly pipetting an aliquot of the reaction into an equal volume of 95% Laemmli SDS-PAGE loading buffer + 5% BME at 95°C in a separate thermal cycler . Time points were analyzed via SDS-PAGE , and gels were quantified by densitometry using in-house gel analysis software ( https://github . com/harmslab/gelquant , v1 . 0; copy archived at https://github . com/elifesciences-publications/gelquant; Harman , 2020 ) . An exponential decay function ( Aoe-kt ) was fit to the data to extract the decay rate , floating Ao and k . Standard deviations were calculated from fits by taking the square root of the diagonalized covariance matrix and by error propagation . Oligomeric states were measured using a superose 12 10/300 GL size exclusion column ( Amersham Biosciences ) with in-line concentration detection using refractive index ( RI ) and particle mass measured using a multiangle laser light scattering ( MALS ) instrument ( Dawn Heleos , Wyatt Technology ) . Samples were concentrated to 0 . 5–2 mg/ml in tris buffer + 0 . 5 mM CaCl2 , 0 . 1 µm sterile-filtered , and analyzed at a flow rate of 0 . 2 ml/min . Data were processed using manufacturer’s software ( Astra ) . Circular dichroism ( CD ) and chemical denaturation experiments were performed using a Jasco J-815 CD spectrometer and spectroscopy-grade guanidine hydrochloride ( gdn-HCl ) or urea . Chemical denaturation was performed using 25 µM dimeric protein in tris buffer with CaCl2 , with tris substituted for spectroscopy-grade trizma . Reversible unfolding and refolding curves were constructed by making concentrated 100 µM protein stocks in either buffer or 6M gdm or 10M urea and then preparing protein dilutions in various concentrations of gdn-HCl or urea in buffer . Samples were left to equilibrate in denaturant between three hours and overnight to allow for equilibration and were then analyzed by CD . Unfolding/refolding equilibration was confirmed by comparing unfolded vs . refolded protein at the same concentration . CD signal was quantified at 222 nm in a 1 mm cuvette using a 1 nm bandwidth , standard sensitivity , and 2 s D . I . T . HT voltage was <600 V . We fit a two-state unfolding model:bf + mfx+ ( bu + mux ) e- ∆G-mxRT1 + e- ∆G-mxRT to the data to extract thermodynamic parameters , where bf , mf , bu , and mu are the folded and unfolded baseline y-intercepts and slopes , ∆G is the unfolding free energy , m is the m-value , R = 0 . 001987 J⋅K−1mol−1 and T = 298 . 15 K . Standard deviations were calculated from fits by taking the square root of the diagonalized covariance matrix and by error propagation . Apparent unfolding kinetics studies were performed using the above conditions by spiking concentrated protein stock directly into 6M gdm and immediately monitoring CD signal at 222 nm . We purchased commercially distributed HEK293T cells from ATCC ( CRL-11268 ) . Because we are using this cell line as a host for heterologous transient transfections , the appropriate control for consistency between assays is the measurement of reporter output for a set of control plasmids and a panel of known treatments . Upon thawing each batch of cells , we run a positive control for ligand-induced response . We transfect the cells with plasmids encoding human CD14 , human MD-2 , human TLR4 , renilla luciferase behind a constitutive promoter , and firefly luciferase behind an NF-KB promoter . We then characterize the raw luciferase output for five treatments: 1 ) mock , 2 ) LPS , 3 ) LPS + polymyxin B , 4 ) S100A9 + polymyxin B , and 5 ) S100A9 + 1 . 25x polymyxin B . This has a stereotypical pattern of responses in renilla luciferase ( high for all ) and firefly luciferase ( low , high , low , high , high ) . To validate that this response is dependent on the transfected TLR4 complex as opposed to the cells themselves , we repeat the experiment but exclude the TLR4 plasmid . This should give identical renilla luciferase values but no firefly luciferase output in response to any treatment . To ensure that the cells maintain their properties between passages , we repeat the mock , LPS , and LPS + polymyxin B control on every single experimental plate . This assay has a built-in control for mycoplasma contamination: high firefly luciferase signal in the absence of added agonist . This indicates that there is another source of TLR4-induced NF-kappa B output in the cells—most plausibly , contamination . This mycoplasma sensing approach is used in the commercially available HEK-BLUE mycoplasma detection kit ( Invivogen ) . We discard any cells that exhibit high background values or reach 30 passages . The antimicrobial activity of S100s was measured against S . epidermidis using a well-established assay ( Hadley et al . , 2018; Nakashige et al . , 2016; Brunjes Brophy et al . , 2013; Brophy et al . , 2012; Cunden et al . , 2016 ) . The day before , a 5 ml starter culture of S . epidermidis in tryptic soy broth ( TSB ) was grown overnight . The next day , the culture was diluted ~1:100 in TSB and grown for approximately 2 hr to an OD600 of ~0 . 8 . Immediately prior to experiment , the S . epidermidis culture was again diluted 1:100 at a ratio of 62:38 experimental buffer ( 25 mM tris , 100 mM NaCl , 3 mM CaCl2 , pH 7 . 4 ) :TSB . S100 proteins were exchanged into experimental buffer . Each well of a sterile 96-well plate was prepared with 40 µl of S . epidermidis diluted in experimental buffer + TSB , S100 protein at the desired concentration in experimental buffer , and then filled to 200 µl , maintaining a ratio of 62:38 experimental buffer:TB . S . epidermidis growth was monitored on a plate reader , measuring OD600 every 15 min for 13 hr . Each measurement was collected in technical triplicate and background-subtracted using a blank containing experimental buffer and TSB alone . Protein samples were confirmed to lack bacterial contamination by measuring S100 protein growth in experimental buffer and TSB lacking S . epidermidis . All plasmids , cell culture conditions , and transfections for measuring the activity of S100s against TLR4s were identical to those previously described ( Nakashige et al . , 2015; Liu et al . , 2012; Loes et al . , 2018; Anderson et al . , 2019 ) . Briefly , human embryonic kidney cells ( HEK293T/17 , ATCC CRL-11268 ) were maintained up to 30 passages in Dulbecco’s Modified Eagle Media ( DMEM ) supplemented with 10% fetal bovine serum ( FBS ) at 37°C with 5% CO2 . Lipopolysaccharide E . coli K-12 LPS ( LPS - tlrl-eklps , Invivogen ) aliquots were prepared at 5 mg/ml in endotoxin-free water and stored at −20°C . Working solutions were prepared at 10 ug/ml and stored at 4°C to avoid freeze-thaw cycles . S100 proteins were prepared by exchanging into endotoxin-free PBS and incubating with an endotoxin removal column ( Thermo Fisher Scientific ) for 2 hr . S100 LPS contamination was assessed by measuring activity with and without Polymyxin B , an LPS chelating agent ( Figure 2—figure supplement 1 ) . LPS ( 200 ng per 100 µl well ) or S100 ( 0 . 8 , 0 . 4 , 2 , 4 , or 5 µM dimer ) treatments were prepared by diluting in 25:75 endotoxin-free PBS:serum-free Dulbecco’s Modified Eagle Media ( DMEM – Thermo Fisher Scientific ) . Polymyxin B ( PB , 200 µg per 100 µl well ) was added to all S100 experimental samples to limit background endotoxin contamination activity from recombinant protein preps . Cells were incubated with treatments for 3 hr prior to assaying activity . The Dual-Glo Luciferase Assay System ( Promega ) was used to assay Firefly and Renilla luciferase activity of individual wells . Each NF-κB induction value shown represents the Firefly luciferase activity divided by the Renilla luciferase activity , background-subtracted using the LPS + PB activity for each TLR4 species and normalized to the activity of LPS alone for each TLR4 species to normalize between plates . All measurements were performed using three technical replicates per plate , a minimum of three biological replicates total , and a minimum of two separate protein preps . For TLR4 activation measurements by A9 proteolytic products , 12 . 5 µM hA9 or hA9 M63F were incubated with 2 . 5 mg/ml Proteinase K immobilized to agarose at 37°C for increasing amounts of time . The reaction was quenched by spin-filtering the sample to remove Proteinase K . 2 µM A9 proteolysis treatments were then added to cells as outlined above . Western blots were performed by running an SDS-PAGE gel and transferring to a nitrocellulose membrane . Membranes were blocked using Odyssey Blocking Buffer for 1 hr , incubated with 1:1000 mouse anti-S100A9 primary antibody ( M13 clone 1CD22 , Abnova ) for 1 hr , and incubated with 1:10 , 000 IRDye Goat anti-mouse 800CW IgG ( H+L , Licor ) for 1 hr , with 3 × 5 min TBST washes in between each step . Blots were imaged using the Licor Odyssey Fc imaging system . All species cartoons were taken from the following websites: http://phylopic . org/image/c089caae-43ef-4e4e-bf26-973dd4cb65c5/ , http://phylopic . org/image/aff847b0-ecbd-4d41-98ce-665921a6d96e/ , http://phylopic . org/image/0f6af3d8-49d2-4d75-8edf-08598387afde/ , http://phylopic . org/image/dde4f926-c04c-47ef-a337-927ceb36e7ef/ . We acknowledge Sarah Werning and David Liao as authors of the opossum and mouse cartoons respectively , which were made publicly available through the creative commons attributions 3 . 0 unported license ( https://creativecommons . org/licenses/by/3 . 0/ ) . | A single protein sometimes does multiple jobs . For instance , our immune system uses a small number of multipurpose proteins to respond quickly to a large number of threats . One example is the protein S100A9 . It acts as an antimicrobial by preventing microbes from getting the nutrients they need , while also stimulating inflammation by inducing the release of molecules that recruit white blood cells . S100A9 , like all proteins , is made up of a chain of small building blocks . These building blocks interact with each other and with other molecules in the environment . The sequence of the building blocks thus determines what jobs the protein can do . Therefore , a single change to the sequence of building blocks can have a dramatic effect: one change might render the protein faulty , while another change might allow it to do a new job . Proteins face similar challenges humans do when trying to do several things at once . A person driving a car while using their phone will not do either task well . Likewise , a protein that does two jobs faces challenges a single-purpose protein does not . Harman et al . were interested in how S100A9 was able to evolve and maintain its dual functionality , despite this potential problem . They started by asking when S100A9 acquired its two purposes . They measured the antimicrobial and inflammatory activity of S100A9 proteins from humans , mice and opossums . The activities of S100A9 in these species was similar , suggesting that S100A9 acquired its different jobs in the ancestor of mammals , some 160 million years ago . Next , Harman et al . computationally reconstructed ancestral forms of S100A9 by comparing hundreds of similar proteins and building an evolutionary tree . They then measured the antimicrobial and inflammatory activity of these ancestral proteins . By comparing the last ancestor that did not have these activities to the first ancestor that did , they identified the sequence changes that gave S100A9 its dual activity . Importantly , these changes are located in separate regions of the protein , meaning they could occur independently , without affecting each other . Further , the same sequence change that converted S100A9 into an inflammatory signal also introduced a mechanism to regulate this activity . The results suggest that a small number of sequence changes – or even a single change – can make a protein more versatile . This means that evolving multipurpose proteins may not be as difficult as is often thought . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2020 | Evolution of multifunctionality through a pleiotropic substitution in the innate immune protein S100A9 |
The Notch signaling pathway consists of multiple types of receptors and ligands , whose interactions can be tuned by Fringe glycosyltransferases . A major challenge is to determine how these components control the specificity and directionality of Notch signaling in developmental contexts . Here , we analyzed same-cell ( cis ) Notch-ligand interactions for Notch1 , Dll1 , and Jag1 , and their dependence on Fringe protein expression in mammalian cells . We found that Dll1 and Jag1 can cis-inhibit Notch1 , and Fringe proteins modulate these interactions in a way that parallels their effects on trans interactions . Fringe similarly modulated Notch-ligand cis interactions during Drosophila development . Based on these and previously identified interactions , we show how the design of the Notch signaling pathway leads to a restricted repertoire of signaling states that promote heterotypic signaling between distinct cell types , providing insight into the design principles of the Notch signaling system , and the specific developmental process of Drosophila dorsal-ventral boundary formation .
The Notch signaling pathway mediates communication between adjacent cells ( Artavanis-Tsakonas et al . , 1999 ) . As the primary juxtacrine signaling pathway , Notch carries out the fine-detail work of animal development , from drawing sharp boundaries between cell populations , to laying out checkerboard-like lateral inhibition patterns across a tissue , to controlling fractal-like branching structures in vascular and lymphatic systems ( Lewis , 1998; Artavanis-Tsakonas et al . , 1999; Irvine , 1999; Bray , 2006; Phng and Gerhardt , 2009 ) . Notch signaling occurs when a DSL ( Delta/Serrate/Lag2 ) ligand binds to a Notch receptor on a neighboring cell , triggering proteolytic cleavage of the Notch receptor and endocytosis of the Notch extracellular domain into the signal-sending cell ( Fortini , 2009 ) . This mechanism releases the Notch intracellular domain ( NICD ) , which translocates to the nucleus and interacts with the CSL complex ( CBF1/Suppressor of Hairless/Lag1; also known as RBP-Jκ ) to initiate transcription of target genes ( Artavanis-Tsakonas et al . , 1999; Fortini , 2009 ) . In addition , Notch receptors and DSL ligands interact within the same cell in a process called cis-inhibition ( del Alamo et al . , 2011 ) . Overexpression and loss of function studies have revealed that DSL ligands can cis-inhibit the ability of a cell to receive a Notch signal , and that this effect depends on the interaction of the extracellular domains of the receptors and ligands in the same cell ( Jacobsen et al . , 1998; D'Souza et al . , 2008 ) . Additionally , Notch receptors can reciprocally block DSL ligands in the same cell from sending signal ( Becam et al . , 2010; del Alamo and Schweisguth , 2009 ) . Previous in vitro studies support a simple model in which ligand cis-inhibition of receptors and receptor cis-inhibition of ligands represent a single process , with ligands and receptors in the same cell forming an inactive complex that prevents them both from interacting in trans with neighboring cells ( Sprinzak et al . , 2010 ) . Recent work has suggested that these mutually inhibitory cis interactions between receptors and ligands in Notch and other signaling pathways can play a critical role in cell signaling ( Yaron and Sprinzak , 2012 ) . To illustrate , we analyze the Notch signaling state of a cell , defined by the cell's quantitative ability of a cell to send or receive signal using a given ligand . We consider a cell expressing one type of ligand and one type of Notch receptor . If the cell produces more receptor than ligand , cis interactions efficiently remove most or all ligand but leave an excess of free receptor , enabling the cell to receive , but not send , Notch signals ( Figure 1A , top left ) . On the other hand , if the cell produces more ligand than receptor , cis interactions sequester the receptor , leaving an excess of free ligand , and permitting the cell to send , but not receive , signals ( Figure 1A , top right ) . In this simple case , the relative levels of ligand and receptor expression produce a sharp threshold between sending and receiving signaling states and thereby regulate the strength and direction of signaling between neighboring cells ( Sprinzak et al . , 2010 , 2011 ) . Consistent with the ratiometric nature of this model , many Notch-dependent developmental processes are highly sensitive to changes in receptor and ligand gene dosage , and show haploinsufficient mutant phenotypes ( de Celis et al . , 1996; de Celis and Bray , 2000; Duarte et al . , 2004; Phng and Gerhardt , 2009; Sprinzak et al . , 2011 ) . 10 . 7554/eLife . 02950 . 003Figure 1 . Cis interactions between receptors and ligands lead to exclusive sending and receiving signaling states . ( A ) In the blue shaded region , receptor expression exceeds ligand expression ( as indicated schematically above plot ) , so that mutual cis interactions leave mainly free receptors , allowing the cell to receive , but not efficiently send , signals . When ligand expression exceeds Notch expression , mutual cis interactions consume most of the Notch receptors , leaving an excess of free ligand , favoring sending over receiving . ( B ) There are multiple potential ways in which Notch1 could interact in cis and trans with Jag1 and Dll1 ligands , and in which Fringe proteins could modulate these interactions . Known interactions are indicated by + and − for positive and negative regulation , respectively . Unknown ways in which Fringe proteins could modulate these interactions are indicated by question marks . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 003 With only a single type of ligand and a single type of receptor it is relatively straightforward to evaluate signaling states ( Figure 1A ) . However , in Drosophila , there are two DSL ligands , Delta and Serrate , while in mammals , there are four Notch receptors ( Notch 1–4 ) and five canonical Notch ligands , three members of the Delta family ( Dll1 , Dll3 , Dll4 ) and two members of the Serrate family ( Jag1 and Jag2 , homologues of Drosophila Serrate ) ( Bray , 2006; D'Souza et al . , 2008 ) . Each ligand–receptor pair can have a different interaction strength . For example , Dll4 interacts more strongly with Notch1 in trans than Dll1 ( Andrawes et al . , 2013 ) . Moreover , in vertebrates , Notch signaling processes typically utilize combinations of multiple receptors and ligands . For example , during angiogenesis , the sprouting of new blood vessels depends on complex spatial expression of Notch1 , Dll4 , and Jag1 ( Benedito et al . , 2009; Phng and Gerhardt , 2009 ) . In chick spinal cord development , generation of the six subtypes of sensory and motor neurons depends on distinct expression domains of Dll1 and Jag1 ( Marklund et al . , 2010 ) . In these and other examples , co-expression of multiple ligands and receptors enables a large number of possible cis and trans interactions , making it difficult to determine which cells are communicating to which other cells through which receptors and ligands . Further adding to the complexity , Fringe glycosyltransferases modulate the interaction between receptors and ligands ( Panin et al . , 1997; Moloney et al . , 2000 ) . Fringe proteins act in the Golgi to transfer N-acetylglucosamine ( GlcNAc ) to O-fucose-modified EGF repeats in the extracellular domain of Notch ( Bruckner et al . , 2000; Moloney et al . , 2000 ) . In Drosophila there is a single Fringe , while in mammals there are three homologues: Lunatic Fringe ( Lfng ) , Manic Fringe ( Mfng ) and Radical Fringe ( Rfng ) . In vitro co-culture experiments have revealed the differential effects of each Fringe on trans-activation from different DSL ligands ( Hicks et al . , 2000; Ladi et al . , 2005; Hou et al . , 2012 ) . When Lfng or Mfng is ectopically expressed in a receiver cell , trans signaling from Dll1 ligands is enhanced , while trans signaling from Jag1 ligands is decreased . The effects of Lfng and Mfng in vertebrate systems resemble the effects of Fringe in Drosophila , which strengthens Delta signaling and inhibits Serrate signaling ( Panin et al . , 1997 ) . On the other hand , Rfng increases the trans response to both ligands ( Ladi et al . , 2005 ) . Despite this work , the effects of Fringe on cis interactions , if any , remain unknown . Given the central role of cis interactions in determining signaling states , it is therefore essential to determine whether and how Fringes influence these interactions . In general , to determine the cell’s signaling state requires knowledge of ( 1 ) the levels of ligands , receptors and Fringe proteins; ( 2 ) the interaction strengths in cis and trans for each ligand–receptor pair , and ( 3 ) how the Fringe proteins act individually and in concert to modulate cis and trans interactions . Data for ( 1 ) are increasingly available in different systems , but ( 2 ) and ( 3 ) have not been measured comprehensively . Such measurements could enable prediction of the directionality and cell type specificity of signaling from expression measurements in diverse processes . To begin to obtain these measurements , we analyzed Notch-ligand cis interactions and their dependence on Fringe proteins in cell culture . We studied the Dll1-Notch1 and Jag1-Notch1 ligand–receptor pairs , as these two ligands are frequently used simultaneously for signaling in the same tissue , and because clear differences in the effects of Fringe on these ligand–receptor pairs in trans have been established ( Figure 1B ) ( Hicks et al . , 2000; Ladi et al . , 2005 ) . To confirm that our measurements were qualitatively relevant for in vivo Notch-dependent processes , we tested our findings in a series of Drosophila mutants . Together , these data support a role for Fringe in modulating cis interactions in mammalian cells and Drosophila . These data constrain the set of possible signaling states for cells expressing multiple Notch pathway components , and support the idea that Notch pathway architecture fundamentally favors heterotypic signaling between cells in distinct signaling states .
We developed a cell culture-based assay system to measure the cis interactions between different ligand–receptor pairs ( Figure 2 ) . As a base cell line , we used CHO-K1 cells that support Notch signaling , but do not endogenously express Notch receptors and ligands . We constructed a stable CHO-K1 cell line constitutively expressing a ‘diverted’ Notch1 receptor , hN1 ( ΔICD ) -Gal4esn ( Struhl and Adachi , 1998; Sprinzak et al . , 2010 ) where the intracellular domain of Notch1 is replaced with yeast Gal4 . This receptor activates a UAS-reporter gene but not endogenous Notch targets . Next , we stably integrated tetracycline-inducible Dll1 or Jag1 ligands fused to the Cerulean fluorescent protein , allowing us to control and read out ligand expression with the inducer 4-epitetracycline ( 4-epiTc ) and readout the expression level by cerulean fluorescence . We denote these cell lines ‘Notch1+Dll1’ and ‘Notch1+Jag1’ , respectively ( Figure 2A , B ) . 10 . 7554/eLife . 02950 . 004Figure 2 . The availability assay labels receptors and ligands that can participate in trans signaling . ( A and B ) Stable CHO-K1 cell lines constitutively express a Notch1-Gal4 chimeric receptor and a tetracycline-inducible Dll1 ( A ) or Jag1 ( B ) ligand fused to cerulean fluorescent protein . ( C and D ) In the receptor availability assay , soluble Dll1ext-Fc is bound to free Notch receptor on the surface of live cells . After fixation , bound Dll1ext-Fc is labeled with anti-Fc fluorescent reagents . Increasing ligand-Cerulean expression reduces receptor availability , as shown in these snapshots ( C , D , bottom panels ) . ( E and F ) The ligand availability assay works similarly , except soluble Notch1ext-Fc fragments bind free ligands on the cell surface . Increasing ligand-Cerulean expression ( E , F , middle panels ) , leads to increased ligand availability ( E , F bottom panels ) . The surface ligand availability shows high spatial correlation with the total cellular ligand staining . Note that cells were plated at high cell density for illustration purposes . For quantitative analysis , cells were dissociated and plated at low density before staining ( Figure 3—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 004 In order to analyze cis interactions in these cell lines , we induced ligand expression and measured the levels of free receptors and ligands on the cell surface . To detect Notch1 receptors , we incubated cells with saturating concentrations of a fragment of the Dll1 ligand fused to the Fc epitope ( Dll1ext-Fc , ‘Materials and methods’ and Figure 3—figure supplement 1 ) . We established that this reagent specifically labels Notch receptors that are available to participate in trans interactions with DSL ligands , but does not label Notch-ligand cis interaction complexes ( Figure 2C , D ) . Similarly , we used saturating concentrations of a Notch1 receptor fragment-Fc fusion protein ( N1ext-Fc ) to specifically bind available ligands on the cell surface ( ‘Materials and methods’ and Figure 3—figure supplement 1 ) . In both cases , we detected bound chimeric Fc ligands or receptors with a fluorescently labeled anti-Fc antibody . We used the parental Notch1 cell line to quantify receptor availability in the absence of cis-ligand , and cell lines expressing only Dll1-Cerulean or Jag1-Cerulean to quantify ligand availability in the absence of Notch1 ( Figure 3—figure supplement 1 ) . As a negative staining control , we incubated CHO-K1 cells with both the ligand and receptor availability assay reagents to establish background staining levels . We plated cells at low density to minimize trans interactions , and included only single , isolated cells in our analysis ( Figure 3—figure supplement 3 ) . Using this assay , we first verified that receptor and ligand availability correlated with receiving and sending ability , respectively . For example , siRNA knockdown of the Notch receptor decreased Notch availability ( Figure 3—figure supplement 2A ) and reduced Notch activation by Dll1 ligands adhered to cell culture plate surfaces ( Figure 3—figure supplement 2B ) . Likewise , inducing ligand expression with 4-epiTc led to both increased ligand availability ( Figure 3—figure supplement 2C ) , and increased ability to trans-activate a Notch reporter cell line in a co-culture assay ( Figure 3—figure supplement 2D ) . In principle , the C-terminal fusion of CFP to the ligand could affect its trafficking or other aspects of its regulation , such as binding to proteins that interact with its PDZ domain ( Pintar et al . , 2007 ) . We therefore compared the ability of untagged DSL ligands to cis inhibit Notch receptors with that of the CFP-tagged ligands . We transiently transfected either tagged or untagged ligands into our Notch receiver cells and measured the resulting decrease in receptor availability . We observed qualitatively similar results with untagged ligands as we did with the CFP fusions ( Figure 3—figure supplement 4 ) . Thus , although CFP fusions could affect other properties of DSL ligands , it does not appear to qualitatively affect the results shown here . Next , we compared the relative abilities of Dll1 and Jag1 to cis-inhibit Notch1 . We induced varying levels of ligand expression in the Notch1+Dll1 and Notch1+Jag1 cell lines , and analyzed the cells with the availability assay . These experiments produced three key observations ( Figure 3 ) :10 . 7554/eLife . 02950 . 005Figure 3 . Dll1 and Jag1 exhibit mutual cis-inhibition with Notch1 . ( A and B ) Single cell data show decreasing receptor availability and increasing ligand availability with increasing ligand expression . Circles denote the medians of data points in logarithmically spaced bins along the x-axis . ‘Effective total ligand’ refers to the ligand availability observed at a given ligand-CFP fluorescence value in a cell line expressing only ligand . For receptor availability data in A , n = 299 and in B , n = 352 . For the ligand availability data in A , n = 323 and in B , n = 530 . Gray bars in all panels represent background levels , defined as the 25–75 percentile range of fluorescence from parental CHO-K1 cells that do not express Notch1 or ligand-Cerulean constructs . ( C ) Transient expression of wild-type Dll1-mCherry ( n = 8817 ) , but not the Dll1-mCherry F199A mutant ( n = 14 , 292 ) , reduced Notch availability to background levels in a Notch1 cell line . Cells were analyzed by flow cytometry . Error bars in all panels denote 95% confidence interval for the bootstrapped estimate of the median . ( D ) Comparison of Notch availability in the Notch1+Dll1 and Notch1+Jag1 cell lines . Lines are fits to a model of receptor-ligand cis-interactions ( Supplementary ) . ( E ) Comparison of ligand availability in cell lines expressing Dll1 . Ligand availability a cell line expressing Dll1 only ( n = 1146 ) . Notch1 reduces ligand availability ( purple , n = 1131 ) , and this effect is rescued by siRNA against Notch1 ( orange , n = 972 ) . In the purple starred region , cells differ significantly in ligand availability between Notch1 or no target siRNA samples , while the orange star denotes regions where Dll1 and Notch1+Dll1 cells transfected with no target siRNA differ significantly . Significance was determined by applying the Wilcoxon rank sum test . Inset shows the model behavior for parameters derived from the fit in D . Knockdown of Notch was measured to be 50% . ( F ) Comparisons of ligand availability in a cell line expressing Jag1 only ( green , n = 733 ) , Notch1+Jag1 ( orange , n = 532 ) , and Notch1+Jag1 with siRNA against Notch ( purple , n = 1163 ) . Starred regions indicate significance as in E . Inset shows model behavior using parameters measured in D . Knockdown of Notch was measured to be 70% . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 00510 . 7554/eLife . 02950 . 006Figure 3—figure supplement 1 . Calibration of availability assay reagents . ( A ) The working concentration of Dll1ext-Fc was determined by incubating the parental hN1 ( ΔICD ) -Gal4esn cells with increasing concentrations of soluble Dll1ext-Fc , and staining with secondary anti-IgG reagents . Blue points are results of two replicates . Data were fit ( red line ) to Navail=aDD+K , where D is the concentration of soluble soluble Dll1ext-Fc , with a = 2 . 1 × 104 ± 0 . 17 × 104 , and K = 2 . 27 ± 0 . 6 µg/ml ( red line ) . ( B ) Similarly , the working concentration of N1ext-Fc was determined by incubating cells expressing fully induced TO-Jag1-cerulean ( orange ) and TO-Dll1-cerulean ( red ) with varying concentrations of N1ext-Fc , and staining with secondary reagents . Data were fit ( red and orange lines ) to Lavail=aLL+K . For the TO-Jag1-cerulean data , a = 2 . 1 × 104 ± 0 . 14 × 104 , and K = 0 . 14 ± 0 . 16 µg/ml ( orange line ) . For the TO-Jag1-cerulean data , a = 1 . 2 × 104 ± 0 . 11 × 104 , and K = 0 . 33 ± 0 . 26 µg/ml ( red line ) . Availability saturated at relatively low concentrations of the N1ext-Fc ( <1 µg/ml ) , while un-induced cell lines showed no availability signal . Concentrations of secondary reagents were not limiting . The working concentration of both N1ext-Fc and Dll1ext-Fc reagents was set at a saturating level ( 10 μg/ml ) based on these measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 00610 . 7554/eLife . 02950 . 007Figure 3—figure supplement 2 . Receptor and ligand availability correlate with signal receiving and sending ability , respectively . ( A ) To validate the Notch availability assay , we tested cells expressing hN1 ( ΔICD ) -Gal4esn and observed high levels of availability fluorescence , while CHO-K1 cells lacking ectopic Notch expression showed minimal availability fluorescence ( Anti-IgG Alexa 488 ) . Bars show mean Notch1 availability , and error bars show standard error of the mean ( SEM ) . ( B ) To show that receptor availability corresponds to receiving ability , we seeded a ‘receiver’ cell line expressing a diverted Notch receptor ( hN1 ( ΔICD ) -Gal4esn ) and a fluorescent reporter for Notch signaling ( UAS-H2B-citrine ) on plates coated with Dll1 ligands . We observed a strong response from the reporter ( compare ± Dplate ) . When we knocked down receptor expression using siRNA against the extracellular domain of Notch1 , receiving ability decreased , coinciding with the decrease in Notch1 availability from A . Bars show mean reporter fluorescence , error bars are SEM . ( C ) Inducible ligand cells show increased ligand availability as ligand expression is induced with 4-epiTc . The TO-Dll1-cerulean and TO-Jag1-cerulean cell lines were incubated with increasing concentrations of 4epi-Tc and stained with ligand availability reagents . Points show the mean availability at each induction level , error bars show the SEM . ( D ) Inducible ligand cells from C are able to trans-activate Notch receiver cells in a dose-dependent fashion as ligand induction is induced with 4epi-Tc . Points are the mean of the top ten percent of receiver cells' YFP fluorescence , error bars are the standard deviation of these responders . Because ligand availability also increases with induction , ligand availability and sending ability are correlated . The Jag1 trans response in the receiver cells is weaker than the Dll1 , suggesting that Jag1 is a less potent trans-activator of Notch1 than Dll1 . In the inset , the same data normalized to the maximal activation elicited by each ligand . In all panels , cells were analyzed by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 00710 . 7554/eLife . 02950 . 008Figure 3—figure supplement 3 . Availability protocol and analysis pipeline . ( A ) Test cells were induced for 48 hr with varying concentrations of 4-epiTc . On the day of the experiment , the cells are dissociated with trypsin , split at low density and allowed to reattach to the cell culture plates . Cells are blocked with 2% FBS in PBS for 30 min at 37°C and then incubated with 10 μg/ml Dll1ext-Fc ( for receptor availability ) or N1ext-Fc ( for ligand availability ) for 1 hr at 4°C . Next , cells are washed , fixed , and permeabilized . To visualize the bound reagents , we incubated the cells with anti-Fc antibody conjugated to Alexa 488 . We also added anti-CFP Alexa 594 to visualize the ligand expression . Finally , we added a blue cytoplasmic stain to identify individual cells . ( B ) After staining we imaged the cells on the microscope ( ‘Materials and methods’ ) . The images are analyzed using a custom Matlab script to identify the cells and take the mean of the total fluorescence within each cell for each fluorescence channel . At this stage we subtract a background fluorescence value from each cell , defined as the median of the background ( unsegmented ) pixels in the neighborhood of the cell . ( C ) Next , we impose a gate on the cell area to filter out doublets and segmentation errors . ( D ) Then , all cells are screened by eye so that only single , isolated cells are included in the final analysis . ( E ) Next , we normalize the total ligand ( x-axis ) to account for differences in the surface expression of each ligand . In the plot in F , the same CFP fluorescence level results in different surface availability measurements for a cell line expressing only Jag1 compared to a cell line expressing only Dll1 , with Jag1 showing higher surface expression . To adjust for this difference in efficiency of surface expression between the different ligands , we fit the total ligand vs ligand availability data from cells expressing ligand only with a linear fit . We use this fit to normalize the data for each ligand accordingly , allowing for comparison between different cell lines expressing different ligands . After this correction is applied , we refer to the total ligand as the ‘Effective total ligand’ . ( F ) The single cell data from each replicate is pooled and then divided into evenly spaced bins along the log of the x ( total ligand ) -axis and the median availability level for each bin is plotted . We use a Matlab bootstrapping method to find the 95% confidence intervals for the estimate of the median . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 00810 . 7554/eLife . 02950 . 009Figure 3—figure supplement 4 . Untagged DSL ligands show similar effects to their fluorescently tagged counterparts . ( A ) We compared the abilities of tagged and untagged DSL ligands to reduce Notch availability in cis . We transiently transfected the Dll1-cerulean plasmid ( Figure 2A ) or the Jag1-cerulean plasmid ( Figure 2B ) into the Notch-Gal4 receiver cell line . In parallel , we also transfected untagged Dll1 or Jag1 . Because we cannot visualize the expression level of untagged ligand , we co-transfected each of the four ligands with a constitutive H2B-mCherry plasmid to label transfected cells . ( B ) We transfected each of the four ligands ( co-transfected with H2B-mCherry ) in parallel into receiver cells . After 24 hr , we performed the receptor availability assay , and analyzed the fluorescence by flow cytometry . ( C ) H2B-mCherry fluorescence correlated with both Dll1-cerulean expression and Jag1-cerulean expression . ( D ) Transfection of H2B-mCherry alone into receiver cells did not reduce Notch availability . ( E ) Based on the results from C , we used H2B-mCherry expression as a proxy for ligand expression level . We found that for increasing H2B-mCherry expression levels , Notch availability decreased , both for the sample transfected with Dll1-cerulean ( blue ) , as we observed in Figure 3 , and also for the sample transfected with untagged Dll1 ( red ) . Notch availability decreased to background levels , as measured by CHO-K1 cells with availability reagents applied ( gray bar denotes 10th to 90th percentile of background cells ) . ( F ) We found that increasing H2B-mCherry levels also led to a decrease in Notch availability , both for the sample transfected with Jag1-cerulean ( blue ) , as also shown in Figure 3 , and also for the sample transfected with Jag1 ( red ) . ( G ) We then performed the experiments co-transfecting constitutive Lfng . We found that when we co-transfected Lfng with untagged Dll1 , Notch availability decreased to background levels ( magenta ) , just as it had without Lfng ( red ) . ( H ) Finally , we compared Notch availability in a sample expressing untagged Jag1 and untagged Jag1 coexpressed with Lfng . We found that Notch availability did not decrease to background when Lfng was expressed , just as we had shown for Jag1-cerulean in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 009 First , both Dll1 and Jag1 can fully cis-inhibit Notch1 receptors ( Figure 2C , D , Figure 3A , B ) . Available Notch1 staining decreased to background levels in a dose-dependent fashion with increasing expression of either Dll1 or Jag1 , indicating that both Dll1 and Jag1 ligands can fully reduce Notch availability . To confirm that this reduction in Notch availability was not an artifact of ligand overexpression , we compared wild-type Dll1 to Dll1-F199A , which contains a point mutation in the DSL domain that was previously shown to be partially deficient in trans-activation and cis-inhibition ( Cordle et al . , 2008 ) . While transiently expressed wild-type Dll1-mCherry reduced Notch availability to background levels , the Dll1-F199A ligand showed only a partial reduction in Notch availability ( Figure 3C ) . Second , Dll1 appeared to be more potent than Jag1 as a cis-inhibitor of Notch1 . We fit a simple model describing cis interactions ( see Figure 1 and ‘Materials and methods’ ) to the single cell data from each condition , and found that approximately twice as much available Jag1 on the cell surface was needed to reduce Notch availability by 50% compared to Dll1 ( Figure 3D ) . Dll1 is also more potent than Jag1 as an activator of Notch1 in trans ( Figure 3—figure supplement 2D ) , suggesting the possibility that cis and trans interaction strengths between a receptor and ligand could be correlated . Third , Notch1 reduced both Dll1 and Jag1 availability , supporting the mutual inhibition model of cis interactions for both ligands ( Sprinzak et al . , 2010 ) . In the Notch1+Jag1 and Notch1+Dll1 cell lines , we observed significantly reduced ligand availability compared to corresponding ligand-only cell lines at the same levels of total ligand expression . Moreover , this effect on available ligand could be rescued by knocking down expression of Notch1 with siRNA ( Figure 3E , F ) . Because ligand availability correlates with sending ability ( Figure 3—figure supplement 2 ) , these results support a role for Notch1 in decreasing Dll11 and Jag1 sending ability in cis , consistent with the mutual inhibition model of cis interactions . We next asked whether and how Fringe proteins modulate cis interactions . We constructed stable Notch1+Dll1 and Notch1+Jag1 cell lines constitutively expressing Lfng , Mfng , or Rfng . These constructs increased the trans response to plate-bound Dll1 ligands and decreased the response to Jag1 ( in the case of Lfng or Mfng ) , or increased the response to both ligands ( in the case of Rfng ) consistent with previous results ( Hicks et al . , 2000 ) . To analyze the effect of Fringe proteins on cis interactions , we compared Notch availability in these cell lines to parental cell lines that lack ectopic Fringe expression . For the Notch1+Dll1 cell lines expressing Lfng , Mfng , or Rfng , Notch1 availability decreased to background levels in response to increasing Dll1 expression ( Figure 4A ) . Note that absolute levels of fluorescence in the assay increase with Fringe expression , as can be observed at low total ligand levels in Figure 4A ( inset ) , because Fringes enhance binding of the Dll1-Fc detection reagent to available Notch1 . To account for this change in binding , we normalized the curves to the level of Notch availability measured when ligand levels were uninduced . After normalizing for this change , cell lines with Fringe expression appear to have stronger cis interactions ( Figure 4A ) . Consistent with this , available Dll1 ligand assays revealed reduced available Dll1 when any of the Fringes was expressed ( Figure 4B ) . Together , these results suggest that all three Fringe proteins preserve , or strengthen , Notch1-Dll1 cis interactions . 10 . 7554/eLife . 02950 . 010Figure 4 . Fringe proteins show distinct effects on Jag1-Notch1 and Dll1-Notch1 cis interactions . ( A ) Available Notch1 levels for the Notch1+Dll1 cell line without Fringe ( red ) or with Lfng ( magenta ) , Mfng ( orange ) , or Rfng ( blue ) . Lines are fits to model ( ‘Materials and methods’ ) . Addition of any of the three Fringes accelerates the drop-off of Notch1 availability . In the inset , the same data , but unnormalized , shows that addition of any of the three Fringe proteins does not prevent available Notch1 from reaching background levels . ( B ) Dll1 availability for the cell lines from A . ( C ) Similar to A , but for the Notch1+Jag1 cell lines . Addition of Lfng and Mfng prevents the depletion of Notch1 availability , while addition of Rfng accelerates depletion of Notch1 availability . In the inset , the unnormalized data shows that Lfng or Mfng , but not Rfng , can block the ability of Jag1 to reduce Notch1 availability to background levels . ( D ) Jag1 availability for the cell lines in C . In all panels , points represent medians of data points in evenly spaced bins taken along the log of the x-axis . Error bars are the 95% confidence intervals of the bootstrapped estimated of the bin medians . Solid lines are model fits to the single-cell data ( see ‘Materials and methods’ ) . Gray bars denote the 10–90th percentile fluorescence range of stained parental CHO-K1 cells that do not express Notch1 or ligands . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 010 Fringe proteins had markedly different effects on Notch1-Jag1 cis interactions . In cell lines expressing Lfng or Mfng , even saturating expression of Jag1 was not sufficient to reduce Notch1 availability to background levels , in contrast to Dll1 , indicating that Lfng and Mfng reduce cis interactions for Jag1 , but not Dll1 ( Figure 4C ) . In contrast , Rfng still allowed Notch1 availability to reach background levels at high cis-Jag1 levels . Note that here , as in Figure 4A , we observed an increased binding of the Dll1-Fc detection reagent in cell lines expressing Fringe proteins . However , when Lfng or Mfng is expressed , only Dll1 , and not Jag1 , is able to reduce Notch availability to background levels . In the ligand availability assay , we did not detect a significant increase in Jag1 availability due to Lfng or Mfng expression ( Figure 4D ) . Because the basal Jag1-Notch1 cis-inhibition strength is already weak compared to the Dll1-Notch1 cis interaction strength ( Figure 3D , F ) , the Jag1 availability assay may not be sensitive enough to detect further reductions in cis-inhibition . Together , these results suggest that Fringe expression modulates cis-inhibition between Notch1 receptor and ligands . Lfng and Mfng preserve and possibly strengthen interactions between Notch1 and Dll1 , in cis and in trans , and weaken interactions between Notch1 and Jag1 , in cis and in trans , while Rfng preserves or enhances the cis interactions with both ligands . Thus , the effects of Fringe proteins on cis interactions are in the same direction ( strengthening or weakening ) as their effects on trans interactions . Because Lfng or Mfng expression weakens Notch1-Jag1 cis interactions , a Lfng-expressing cell could maintain high expression levels of cis-Jag1 without compromising its ability to receive signals from trans Dll1 ligands through Notch1 . To test this prediction , we used a previously developed time-lapse video assay to titrate ligand levels over time in individual cells ( Figure 5A ) ( Sprinzak et al . , 2010 ) . We constructed cell lines constitutively expressing the diverted chimeric receptor , hN1ΔICD-Gal4esn , and incorporating a UAS-H2B-Citrine reporter activated by Gal4 released by activated chimeric Notch1 ( Figure 5B ) . To these cell lines we added a tetracycline-inducible Jag1-mCherry ligand . Finally , to analyze the effect of Lfng on Notch1-Jag1 cis-inhibition , we added to this parental cell line a stably integrated , constitutively expressed Lfng gene . We also constructed similar cell lines with Dll1-mCherry in place of Jag1-mCherry . 10 . 7554/eLife . 02950 . 011Figure 5 . Lfng relieves Jag1-Notch1 but not Dll1-Notch1 cis-inhibition . ( A ) Schematic design of time-lapse experiment to analyze cis-interactions . Before the video , cells are pre-induced to express high levels of ligand and then seeded on plates coated with Dll1-Fc . During the video , cell division dilutes the cis-ligand sufficiently to allow cells to respond to plate-bound ligands . ( B ) Cell lines expressing ‘diverted’ Notch1-Gal4 chimeric receptor , a fluorescent reporter for Notch signaling ( UAS-H2B-citrine ) , and inducible Jag1-mCherry ligand , with ( left ) and without Lfng ( right ) . Corresponding cell lines with the Dll1-mCherry ligand are not shown . ( C ) Typical video filmstrip . ( D ) Quantification of videos of the Notch1+Jag1 and Notch1+Dll1 cell lines , with and without Lfng . Points show the mean fluorescence of all cells in a single frame . The cell line with Lfng responds earlier than the cell line without Lfng , reflecting a weaker cis interaction between Jag1 and Notch1 . The time when the YFP slope exceeds a threshold , defined as 10% of the final YFP slope , is marked as ton . ( E ) Quantification of the ligand levels at ton for each cell line . Values are the average of two videos . Notch activity occurs even at high cis-Jag1 levels in the +Lfng , but not the −Lfng , cell line . Notch responses occurred only at low ligand levels for the Dll1-mCherry cell lines , with and without Lfng . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 011 Both parental and Lfng-expressing cell lines were seeded on plates coated with Dll1ext-Fc recombinant protein to activate Notch1 in trans , but signaling was initially blocked with the Notch signaling inhibitor DAPT . 24 hr before the start of the video ( t = −24 hr ) , we added 4-epiTc to induce cis-Jag1 production . 1 day later , at a time defined as t = 0 , we washed out 4-epiTc and DAPT , halting further cis-Jag1-mCherry production , and allowing Notch signaling to commence . We then monitored the cells for 2 days using time-lapse fluorescent microscopy . As the cells divided , cis-Jag1 levels gradually diminished in individual cells , predominantly through dilution ( Figure 5C ) . To measure the dependence of Notch activation on cis ligand levels , we used custom image analysis software to quantify fluorescence in individual cells in the video ( ‘Materials and methods’ ) . We then plotted the rate of increase of H2B-Citrine , a measure of Notch activity , against mCherry fluorescence , a measure of total ligand abundance , for each cell . In the parental cell lines Notch reporter activation was delayed by ∼24 hr , indicative of cis-inhibition ( Figure 5D , Video 1 ) . By contrast , the Lfng cell line responded earlier to the plate-bound Dll1ext-Fc despite high cis-Jag1 levels , and not significantly later than in cells lacking ligand expression altogether , indicating that Lfng reduces Notch1-Jag1 cis interaction , and thereby prevents high cis-Jag1 levels from blocking Notch activation ( Figure 5C , D , Video 2 ) . In contrast , when we performed the same experiment with cell lines expressing Dll1-mCherry , we observed no corresponding relief of cis-inhibition due to Lfng , suggesting that Lfng does not weaken Dll1-Notch1 cis interactions . These results , consistent with those from the availability assay , support the finding that Lfng inhibits Jag1-Notch1 , but not Dll1-Notch1 , cis-inhibition . 10 . 7554/eLife . 02950 . 017Video 1 . Dilution video assay with a cell line expressing Notch1+Jag1 . Reporter activation shows a delay , as cell divisions are required to dilute out cis-Jag1 . Frame rate is 10 fps , with a frame taken ever 20 min . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 01710 . 7554/eLife . 02950 . 016Video 2 . Dilution video assay with a cell line expressing Notch1+Jag1+Lfng . Reporter activation is immediate despite high cis-Jag1 levels . Frame rate is 10 fps , with a frame taken ever 20 min . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 016 To determine whether Fringe modulation of cis interactions occurs in a developmental context , we turned to Drosophila wing imaginal disc as a model system . Although the Notch pathway has fewer components in flies than vertebrates , the Drosophila wing disc provides a tractable system to examine related behaviors in a developmental context . Moreover , previous work has established both an effect of Fringe on trans interactions that is qualitatively similar to the mammalian Lfng and Mfng , as well as a clear role for cis-inhibition of Notch by its ligands Delta and Serrate ( related to mammalian Jag1 ) in this process ( Doherty et al . , 1996; Micchelli et al . , 1997; Panin et al . , 1997 ) . The adult Drosophila wing has a smooth margin lined with bristles and is patterned with five wing veins ( Figure 6A ) . Normal development of both the margin and the wing veins relies on spatially restricted Notch signaling . In the case of the wing margin , a sharp stripe of Notch signaling occurs at the boundary between dorsal wing cells that express Notch , Fringe , and the Delta and Serrate ligands , and ventral wing cells that express Notch and Delta . This Notch signaling drives the expression of the downstream targets E ( spl ) , Wingless , and Cut , that proceed to further activate the expression of ligands and to direct the patterning and development of the wing margin ( de Celis and Bray , 1997 ) . The wing veins are specified by a gradient of Delta ligand expression that drives Notch signaling in two stripes of cells that form the edges of the wing veins ( de Celis et al . , 1997; De Celis , 2003 ) . 10 . 7554/eLife . 02950 . 012Figure 6 . Delta and Serrate have distinct cis-inhibition phenotypes , and decreasing Fringe activity affects each of these phenotypes differently . ( A ) A wild-type fly wing . The five wing veins are marked . Inset is a close-up of the area marked by the rectangle . ( B ) Loss of one copy of Notch leads to mild wing margin loss and wing vein thickening ( inset ) . ( C and D ) One ( C ) or two ( D ) additional copies of Delta in the N55e11/+ background from B leads to mild enhancement of the wing margin defects and strong enhancement of vein thickening phenotypes . ( E and F ) One ( E ) or two ( F ) additional copies of Serrate in the N55e11/+ background from B results in strong enhancement of wing margin defects but only mild enhancement of vein thickening phenotypes . ( G ) Loss of one copy of fringe ( fng13/+ ) in the N55e11/+ background from B enhances wing margin loss . ( H ) Addition of one copy of Delta to the N55e11/+ fng13/+ background from G does not further enhance margin loss and seems to suppress the vein thickening phenotype ( compare to C ) . ( I ) Addition of one copy of Serrate to the N55e11/+ fng13/+ background further enhances wing margin loss in G , suggesting that Fringe usually works to counter the effects of Serrate overexpression . ( J ) Loss of one copy of fringe in a Delta overexpression background does not lead to any wing margin defects . ( K ) Loss of one copy of Fringe in a Serrate overexpression background leads to wing margin loss in some animals , suggesting that Fringe normally blocks the negative effects of Serrate on Notch signaling in animals with wild-type Notch expression levels . ( L ) Classification system used to quantify frequencies of mutant phenotypes . Class 0 denotes a wild-type fly wing morphology . Class 1 flies show mild wing margin loss adjacent to the L3 vein . Class 2 flies show more extensive margin loss extending to the L3 and L4 veins . Class 3 flies show margin loss in L3 and L4 and also in anterior regions of the wing . ( M ) Quantification of the phenotypes using the scoring system in J . At least 50 wings were scored for each genotype , except for the last two columns , for which we scored 48 and 34 wings , respectively . The most severe class 3 defects arise when Notch dosage is halved ( 1X Notch ) and Serrate dosage is doubled ( 4X Serrate ) , a consequence of Serrate cis-inhibition . Class 3 defects also arise when Notch and fringe dosages are halved , and an additional copy of Serrate is added , suggesting that Fringe normally works to block the effects of Serrate cis-inhibition ( Compare column 5 with column 9 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 01210 . 7554/eLife . 02950 . 013Figure 6—figure supplement 1 . Genomic Delta and Serrate transgenes behave similarly to endogenous copies of Notch ligands in Drosophila . ( A and B ) Schematic of genomic regions harboring Delta ( A ) or Serrate ( B ) , their neighboring genes and their corresponding rescue transgenes . ( C and D ) Addition of two extra copies of Delta ( C ) or Serrate ( D ) does not alter wing margin or wing vein formation . ( E ) Loss of one copy of Delta results in an increase in wing vein material . ( F ) This phenotype is suppressed by one copy of the Dlgt-wt transgene . ( G ) When raised at room temperature ( 22°C ) , animals harboring Dl9P in trans to the temperature-sensitive allele DlRF ( Bender et al . , 1993 ) are lethal ( not shown ) . However , one copy of the Dlgt-wt transgene rescues the lethality of the Dl9P/RF animals at 22°C . ( H ) The Dlgt-wt/+ Dl9P/RF animals raised at 30°C show a wing vein thickening phenotype similar to Dl9P/+ animals ( E ) , most likely because increasing the temperature further decreases the activity of the DlRF allele and results in the appearance of Delta haploinsufficient phenotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 013 Removing one copy of Notch results in wing vein thickening and the classic ‘notched’ wing phenotypes ( Figure 6B ) ( Mohr , 1919 ) . If Notch and its ligands were solely involved in trans-activation , one would expect the Notch haploinsufficient phenotypes to be enhanced by a simultaneous decrease in ligand expression . However , loss of one copy of Delta was reported to suppress the Notch+/− wing vein phenotype ( de Celis and Bray , 2000 ) , indicating that the haploinsufficient phenotypes observed in Notch+/− animals are caused by cis-inhibition of Notch due to an increase in the relative levels of ligands to Notch . The cell-based assays indicated that Lfng and Mfng preserve the cis-inhibition of Notch by Dll1 but weaken the cis-inhibition of Notch by Jag1 . To examine whether Drosophila Fringe affects the ability of Delta and Serrate to cis-inhibit Notch in a manner similar to Lfng and Mfng , we performed genetic interaction studies in a Notch+/− background , which seems to be more sensitive to cis-inhibition by ligands . Because the wing margin loss phenotype is only partially penetrant in Notch+/− animals , we used a classification system to quantify this phenotype in various genotypes ( Figure 6L ) : A wild-type wing is scored as 0 , a wing with mild margin loss adjacent to the L3 wing vein is scored as 1 , a wing with more extensive wing margin loss extending to L3 and L4 wing vein regions is scored as 2 , and a wing with margin loss in L3-L4 and anterior regions of the wing is scored as 3 . Based on this scoring system , only 40% of Notch+/− wings show a margin defect , all of them with a score of 1 ( Figure 6M ) . First , we sought to examine the effects of increasing ligand gene dosage on Notch haploinsufficient phenotypes . To this end , we first generated transgenic flies harboring the Delta locus ( Dlgt-wt ) or the Serrate locus ( Sergt-wt ) , in which the expression of Delta or Serrate is driven by endogenous promoter/enhancers , and showed that these transgenes behave similarly to their endogenous counterparts in flies ( Figure 6—figure supplement 1 ) . Animals with one or two copies of the Dlgt-wt transgene in a wild-type background did not exhibit any wing defects ( Figure 6—figure supplement 1C and not shown ) . Similar effects have been observed in Dp ( 3;3 ) bxd110/+ flies ( not shown ) , which harbor one copy of a homozygous lethal intrachromosomal duplication containing the Delta locus ( Vassin and Campos-Ortega , 1987 ) . However , in Notch+/− animals , one copy of the Dlgt-wt transgene resulted in a slight increase in the penetrance of class 1 wing margin defects and a moderate enhancement of the wing vein thickening ( Figure 6C , M ) . Two copies of Dlgt-wt strongly enhanced the wing vein thickening phenotype in Notch+/− animals ( Figure 6D ) . These animals also showed a moderate enhancement of the wing margin loss phenotype ( Figure 6D , M ) . These observations indicate that although Delta-Notch cis-inhibition affects both wing vein and wing margin formation , Notch is more sensitive to Delta cis-inhibition during wing vein formation . The effects of increasing the gene dosage of Serrate on the Notch haploinsufficient phenotypes were quite different from those of increasing Delta dosage . Two copies of the Sergt-wt transgene did not cause any wing abnormalities in a wild-type background ( Figure 6—figure supplement 1D ) . However , one copy of Sergt-wt significantly enhanced the wing margin loss phenotype of Notch+/− animals without affecting their wing vein phenotype ( Figure 6E ) . A second copy of Sergt-wt further enhanced the wing margin loss ( Figure 6F ) . Indeed , 44% of the Notch+/−; Sergt-wt/Sergt-wt wings showed a Class 3 wing margin loss , which was not observed in any of the other genotypes analyzed in this study so far ( Figure 6M ) . Of note , despite their severe wing margin loss , Notch+/−; Sergt-wt/Sergt-wt animals only showed a mild enhancement of the wing vein thickening compared to Notch+/− flies ( Figure 6F , compare to Figure 6B ) . These observations indicate that although both Delta and Serrate can cis-inhibit Notch during wing development , Notch is more sensitive to cis-Delta during wing vein formation and more sensitive to cis-Serrate during wing margin formation . Next we tested whether and how altering Fringe activity changed these distinct Delta and Serrate cis-inhibition phenotypes . We used fringe13 and fringeL73 strains that harbor severe loss-of-function alleles with nonsense mutations in Fringe ( Correia et al . , 2003 ) . Animals heterozygous for fringe have normal wings ( Correia et al . , 2003 ) . However , loss of one copy of fringe alters the Notch haploinsufficient wing margin and wing vein phenotypes in opposite directions: it enhances the Serrate-dependent wing margin loss but suppresses the Delta-dependent wing vein thickening ( Figure 6G , M ) . Addition of one copy of Serrate , but not one copy of Dl , enhances the wing margin loss in Notch+/−; fng+/− animals ( Figure 6H , I , M ) . Indeed , even in fng+/− animals with wild-type Notch gene dosage , addition of a copy of Serrate results in a low penetrance Class 1 wing margin loss ( Figure 6M ) and addition of two copies of Serrate enhances this phenotype by generating Class 2 wing margin loss in ∼8% of wings ( Figure 6K , M ) . Of note , adding one or two copies of Delta does not result in any wing phenotypes in a fng+/− background ( Figure 6J , M and data not shown ) . If Fringe only affects the trans activation of Notch by ligands , one would expect to see increased signaling and therefore wing margin duplication in fng+/− animals with additional copies of Serrate . However , not only do we not see an enhancement of Notch signaling in these animals , we also observe a wing margin loss which becomes more severe as we increase Serrate gene dosage . This indicates that in addition to its effect on Notch trans activation by Serrate , Fringe also regulates the cis-inhibition of Notch by Serrate . Altogether , these observations strongly support the conclusion that during Drosophila wing development , Fringe increases the sensitivity of Notch to cis-Delta but decreases its sensitivity to cis-Serrate , similar to the effect of Lfng and Mfng in the mammalian context .
The signaling states shown in Figure 7 can provide insights into developmental processes . For example , consider boundary formation in the Drosophila wing imaginal disc , a well-characterized developmental process involving both Notch ligands and Fringe ( Micchelli et al . , 1997; Panin et al . , 1997; de Celis and Bray , 1997; Irvine , 1999; Troost and Klein , 2012 ) . In the first stage of boundary formation , cells within the dorsal and ventral compartments of the wing disc signal to one another , forming a sharp stripe of Notch activation at the interface between the two cell populations . This Notch activity drives the expression of the Notch target genes E ( spl ) , Wingless , and Cut , which together refine the expression patterns of Notch ligands and regulate the spatial pattern of Notch activation in subsequent stages ( de Celis and Bray , 1997 ) . For this discussion , we focus on the initial step before these feedback mechanisms become active . Dorsal cells express Notch , Delta , Serrate , and Fringe , while ventral cells express Notch and Delta only ( Figure 8 ) . Because Fringe blocks Serrate-Notch trans signaling , dorsal cells do not signal to one another with Serrate , but can signal with Serrate to ventral cells ( Irvine , 1999 ) . However , Fringe also promotes Delta-Notch trans signaling , so dorsal cells respond to Delta expressed by ventral cells . In this way , signaling can only occur at the interface between the two cell populations ( Fortini , 2009 ) . 10 . 7554/eLife . 02950 . 015Figure 8 . A model for Notch signaling states during dorsal-ventral boundary formation in the Drosophila wing disc . ( A ) A schematic of the Drosophila wing imaginal disc during the third larval instar . During this stage of development , a sharp stripe of Notch signaling occurs at the interface of the dorsal and ventral wing compartments ( blue nuclei ) , leading to upregulation of target genes that drive further wing development . ( B ) The first step of boundary formation , with signaling states indicated in cartoons above . Initially , ventral cells express Notch and Delta . Because ventral cells can receive signal , Notch must be in excess of Delta to achieve a receiving signaling state . Dorsal cells express Serrate , Notch , Delta , and Fringe ( magenta outline denotes Fringe expression ) . Fringe promotes Delta-Notch cis interactions but weakens Serrate-Notch cis interactions , enabling dorsal cells to simultaneously receive signals from Delta while sending signals with Serrate . Thus , the first signaling step occurs from dorsal to ventral cells ( green arrow ) . We cannot rule out that there may be some low level signaling from Delta expressed in dorsal cells as suggested by Lei et al . ( 2003 ) ; however , experiments suggest that Delta in ventral cells is dispensable for proper wing development . ( C ) The second step of signaling begins with upregulation of Delta in activated ventral cells . These cells switch to a sending state , and trans activate dorsal cells . We cannot rule out the possibility that these cells send signal to their ventral cell neighbors; however , because Delta cannot efficiently send in a Fringe-negative background , and because of the existence of feedback mechanisms at this stage , Notch activation is kept to low enough levels to prevent upregulation of target genes involved in wing margin formation ( de Celis and Bray , 1997 ) . The result is the observed pattern of Notch activation at the boundary of dorsal and ventral compartments . DOI: http://dx . doi . org/10 . 7554/eLife . 02950 . 015 This picture of boundary formation appeared to challenge the notion that send and receive signaling states are exclusive ( Figure 1A ) . Indeed , elegant mosaic experiments based on mis-expression of ligands ( de Celis and Bray , 1997 ) , have suggested that both dorsal and ventral cells are capable of receiving signals , that is , all cells are in a receiving state . But in order for signaling to occur , some cells must also possess the capability to send signals . Thus , it appears that some cells should be able to send and receive signal simultaneously . Fringe modulation of cis interactions , reported above , together with a recent developmental study of this system ( Troost and Klein , 2012 ) suggest a potential resolution to this discrepancy . Troost and Klein ( 2012 ) recently examined signaling at the boundary with higher time resolution than in previous work . They found that in the early stages of wing margin formation , signaling occurred in two sequential phases: first , dorsal cells signal via Serrate to ventral cells . Subsequently , ventral cells up-regulate Delta and signal back to dorsal cells . In this model , because ventral cells are initially capable of receiving , they should have an excess of Notch over Delta , as shown in Figure 8B ( ventral ) . The data presented here suggest that in dorsal cells , Fringe should strengthen cis-inhibition between Delta and Notch , while weakening cis-inhibition between Serrate and Notch . Because dorsal cells are initially in a receiving state ( de Celis and Bray , 1997 ) , Delta expressed by these cells should not efficiently activate Notch in neighboring dorsal cells . Indeed , Delta mutant clones in the dorsal compartment do not affect adult wing margin formation ( Doherty et al . , 1996 ) , indicating that Delta-mediated signaling among dorsal cells , which occurs at most very weakly ( Lei et al . , 2003 ) , is dispensable for normal wing margin formation . The formation of a mutually inactivating , cis-inhibitory interaction between Delta and Notch could explain why dorsal cells do not strongly activate one another even though they express and are responsive to Delta ( Doherty et al . , 1996; Lei et al . , 2003 ) . At the same time , Fringe could allow Serrate and Notch to both remain available simultaneously ( Figure 8B , dorsal ) , a state resembling that in Figure 7E ( far right panel ) . Thus , in the first phase , signaling would occur only from dorsal to ventral , as observed ( Troost and Klein , 2012 ) . It is worth mentioning that the dynamic and complex expression patterns of Notch , ligands and Fringe in developmental contexts could superimpose additional layers of regulation on the effects of Fringe on cis and trans interactions between Notch and ligands . Notch activation in ventral cells in the first phase ( Figure 8B ) causes them to up-regulate Delta expression , switching them to a Delta-sending state ( de Celis and Bray , 1997 ) ( Figure 8C ) . In this second phase , ventral cells can trans-activate dorsal cells , which remain responsive to Delta . In principle , ventral sender cells could send to adjacent ventral cells; however it seems that at least in flies , Delta cannot send efficiently to cells that lack Fringe expression , and therefore in the flanking ventral cells Notch activity cannot achieve high enough levels to induce Cut and Wingless expression ( de Celis and Bray , 1997 ) . This interpretation requires Fringe modulation of cis interactions , as observed here , and implies that signaling during boundary formation is heterotypic ( Troost and Klein , 2012 ) . It will be interesting to see if other Notch-dependent boundary formation processes involve similar Notch signaling states and dynamics ( Irvine , 1999 ) . These results could help explain other puzzling observations . For example , one of the most striking vertebrate Notch phenotypes is disorganized somitogenesis in Notch pathway mutants ( Irvine , 1999 ) . In mice and chicks , Lfng is required for this process ( Lewis , 1998 ) . However , Lfng inhibits Notch1-Dll1 signaling ( Dale et al . , 2003 ) , rather than promoting signaling as expected from previous analysis of its effect on trans interactions ( Hicks et al . , 2000 ) . The ability of Lfng to strengthen Dll1-Notch1 cis interactions could explain this phenomenon , since Lfng would tend to reduce the abundance of Dll1 and Notch1 available for trans signaling interactions . Based on these results and others , different configurations of receptors and ligands , through cis interactions , could work to specify distinct signaling states in cells ( Figure 7B–E ) . These states are more complex than just ‘send’ and ‘receive’ but are still highly constrained by cis interactions and the ways they can be modulated . Understanding what signaling states are possible , how expression levels of various pathway components determine those signaling states , and how cells in different signaling states interact with one another , could provide a useful way to think about the Notch signaling system more generally and to infer the directionality and specificity of signaling in more complex contexts . However , many questions remain . We still lack systematic measurements of the interaction strengths , in cis and in trans , for the full repertoire of ligand–receptor pairs , and their quantitative dependence on Fringe expression levels . Given that multiple Fringe proteins are co-expressed in many systems , we will need to examine how Fringe proteins combine to influence Notch signaling . Fringe proteins are also known to modify the DSL ligands ( Panin et al . , 2002 ) ; however , a functional role for sugar modifications of the DSL ligands has not been forthcoming ( Muller et al . , 2014 ) . A potential effect of these modifications on cis interactions , if it exists , would also need to be accounted for in a more complete model . Finally , additional components beyond ligands , receptors , and Fringes may need to be considered in developing a more predictive view of Notch signaling . We anticipate that the experimental approaches developed here can be generalized to address these questions and provide a deeper understanding of the basic design principles of the Notch signaling system .
CHO-K1 cells were maintained as described in Sprinzak et al . ( 2010 ) . Briefly , cells were maintained in Alpha-MEM Earle's Salts media ( Irvine Scientific , Santa Ana , CA ) supplemented with 10% Tet-system approved FBS ( Clontech , Mountain View , CA ) , and L-glutamine , penicillin and streptomycin additive ( Gibco , Carlsbad , CA ) , and stored in an incubator at 37°C at 5% CO2 . Genetic constructs , including siRNA constructs , were introduced into cells using Lipofectamine 2000 reagent according to the manufacturer's protocol ( Life Technologies , Carlsbad , CA ) , or FugeneHD reagent ( Promega , Madison , WI ) . Selection was performed using 400 μg/ml Zeocin ( Life Technologies ) , 10 μg/ml Blasticidin ( InvivoGen , San Diego , CA ) , 600 μg/ml Geneticin ( Life Technologies ) , 500 μg/ml Hygromycin ( InvivoGen ) and/or 3 μg/ml Puromycin ( Life Technologies ) ( See supplementary materials for the antibiotic resistance genes used to integrate each genetic construct ) . Single clones were obtained using FACS sorting or limiting dilution . Single clones were chosen based on fluorescence or quantitative PCR for non-fluorescent constructs . The availability assay was based on the ‘binding’ assays in Shimizu et al . ( 1999 ) where soluble ligands bound to surface-expressed Notch2 receptors . Test cells were plated in 24-well plates ( BD Falcon , San Jose , CA ) at 25% confluence and treated with one of eight concentrations of 4-epiTc ranging from 0 to 200 ng/ml . In siRNA transfection experiments , silencing constructs were delivered after 24 hr of induction . After 48 hr of induction , cells from all 4-epiTc induction conditions were trypsinized , pooled into a single tube , and replated in triplicate at low ( 5–10% confluent ) density . CHO-K1 and hN1 ( ΔICD ) -Gal4esn cell lines were also plated as staining controls . After 4–6 hr , cells were blocked for 30 min at 37°C in blocking buffer ( PBS with 2% FBS and 100 μg/ml CaCl2 ) . Next , cells were incubated with 10 μg/ml soluble Mouse Recombinant Dll1ext-Fc chimera ( receptor availability ) or Mouse Recombinant Notch1ext-Fc chimera ( ligand availability ) , both from R&D Systems ( 5267-TK and 5026-DL , respectively , Minneapolis , MN ) diluted in binding buffer ( PBS with 2% Sigma bovine serum albumin and 100 μg/ml CaCl2 ) , for 1 hr at 4°C . After incubation , cells were washed three times with binding buffer and fixed with 4% methanol-free formaldehyde ( Polysciences Inc . , Warrington , PA ) . Cells were washed three times with binding buffer and permeabilized with 0 . 5% Triton X-100 ( Thermo Scientific , Waltham , MA ) and washed three more times . Next , cells were blocked with blocking buffer for 30 min at room temperature and then incubated for 1 hr at room temperature with the following fluorescent secondary reagents: 1:500 dilution of anti-mouse IgG conjugated to Alexa 488 ( Life Technologies ) to stain cell-bound recombinant protein-Fc , 1:500 dilution of anti-GFP conjugated to Alexa 594 ( Life Technologies ) to visualize the ligand-CFP expressed by the cells , and a 1:10 dilution of HCS Cell Mask Blue ( Life Technologies ) to label the cells' cytoplasm for automatic segmentation . All reagents were diluted in binding buffer . Finally , cells were washed three times with binding buffer and mounted in 70% glycerol for microscopy analysis . Images were acquired with a CoolSnap HQ2 camera on a Nikon inverted TI-E microscope using a 20× long working distance objective . Metamorph 7 . 5 ( Molecular Devices , Sunnyvale , CA ) controlled the microscope , camera , stage ( ASI instruments , Warren , MI ) and brightfield and epifluorescence shutters ( Sutter Instruments , Novato , CA ) and collected the images . Fluorescent illumination was generated by the Sola LED light source ( Lumencor , Beaverton , OR ) and filtered through the Chroma filter sets SpGold , SpRed , and SpGreen . Brightfield illumination was generated by a halogen bulb . Images were analyzed in Matlab 2012 ( Mathworks , Natick , MA ) . First , cells were segmented on their labeling with the HCS Cell Mask Blue cytoplasmic stain . Next , total fluorescence in each fluorescence channel for each cell was calculated as follows . First , the value of the background fluorescence was computed in the neighborhood of the cell by taking the median of the unsegmented pixels in the neighborhood of the cell . Next , this background value was subtracted from each pixel's fluorescence value in the cell . Finally , all of the background-subtracted pixels were averaged to give the mean fluorescence for that cell . Image analysis code is posted on GitHub ( https://github . com/llebon/image-analysis ) . After this automatic processing , manual correction of the data was performed . This included imposing a gate on the segmented cell area to filter out multiple cells and segmentation errors . Next , cells were screened by eye such that cells in physical contact with another cell were rejected , and only single , isolated cells were included in the analysis . We found that for the same measured ligand-CFP level , we obtained different levels of surface availability , suggesting the ligands reach the cell surface with different efficiencies . To account for this difference , we normalized total ligand-CFP fluorescence in the Notch1+Dll1 and Notch1+Jag1 cell lines . Total ligand was plotted as ‘Effective total ligand’ . We then plotted each cell’s effective total ligand and availability fluorescence , and grouped cells into bins logarithmically-spaced along the effective total ligand axis . We plotted the median of these bins , and used a bootstrapped estimate of the median ( MATLAB function bootci ) to find the 95% confidence intervals of the bin median . The snapshots in Figure 2 were acquired on a Leica DMIRB/E fluorescence microscope with a 20× objective using the Chroma filter sets ECFP ( 31055v2 ) and EYFP ( 41028 ) . Videos were performed as described previously in Sprinzak et al . ( 2010 ) . Cells were seeded onto glass-bottom plates ( MatTek , Ashland , MA ) coated with 5% fibronectin ( Innovative Research , Novi , MI ) in low-fluorescence imaging media , Alpha-MEM that includes 5% FBS and omits phenol red , riboflavin , folic acid , and vitamin B12 ( Life Technologies , custom made ) . Cells were maintained at 37°C and 5% CO2 in a chamber enclosing the microscope , an inverted Olympus IX81 equipped with Zero Drift Control ( ZDC ) , a 20× NA 0 . 7 objective , and an iKon-M CCD camera ( Andor , Belfast , NIR ) . All devices were controlled by Metamorph software . Videos were analyzed in Matlab . Cell nuclei in each frame were identified automatically based on the CFP nuclear fluorescence , and the total fluorescence from each channel in each cell nuclei was recorded . Background subtraction was applied to each fluorescence value . In the plots , the average fluorescence from all of the cells in the frame is plotted vs time . Video analysis code is posted on GitHub ( https://github . com/llebon/movie-analysis ) . For analysis with flow cytometry , cells were dissociated with 0 . 25% trypsin ( Life Technologies ) , diluted in FACS buffer ( 1× Hank's Balanced Salt Solution [Gibco] with 2 . 5 mg/ml BSA ) , and filtered through 40 μm strainers ( BD Falcon ) . The cell suspension was screened for single-cell forward and side scatter and fluorescence intensity on a MacsQuant VYB instrument ( Miltenyi Biotech , Bergisch Gladbach , Germany ) . Data was imported into Matlab 2012 for analysis . Analysis included imposing a gate on the forward and side-scatter area to omit dead cells and doublets and then analyzing the single-cell fluorescence intensity for each channel . Quantitative PCR was used to confirm gene expression of non-fluorescent components . RNA was isolated from samples using the Qiagen RNAeasy kit according to the manufacturer's protocol . cDNA was synthesized from 1 μg of RNA using the iScript cDNA synthesis kit ( Bio-Rad , Hercules , CA ) . For the real-time PCR reactions , 2 μl of cDNA was added to one reaction of SsoFast Probes Supermix ( Bio-Rad ) . Each reaction was performed in triplicate . In parallel , three real-time PCR reactions were performed to measure β-actin levels in the sample , allowing us to compute a delta–delta CT value for the gene of interest in our cell lines . Reactions were performed on a Bio-Rad CFX Real-Time PCR Detection System . Probe sets included the following: siRNA against Notch1-ECD was delivered to cells using Lipofectamine 2000 reagent . Three dicer-substrate ( dsiRNA from Integrated DNA Technologies , Coralville , IA ) oligonucleotide duplexes against Notch1-ECD were pooled and 20 pmol of the mix were added to each well . IDT Universal Negative Control duplex was also transfected alongside each sample as a control . Animals were grown in standard food at room temperature ( 22°C ) unless specified otherwise . The following fly strains were used in this study: ( 1 ) y w , ( 2 ) Dl9P/TM6 , Tb1 , ( 3 ) N55e11/FM7c , ( 4 ) fng13/TM3 , Sb1 , ( 5 ) fngL73/TM3 , Sb1 , ( 6 ) Dp ( 3;3 ) bxd110/TM6 , Tb1 , ( 7 ) DlRF/TM6C , Sb1 Tb1 , ( 8 ) y1 sc1 v1 P{nos-phiC31\int . NLS}X; P{CaryP}attP2 , ( Groth et al . , 2004; Bischof et al . , 2007 ) , ( 9 ) y1 M{vas-int . Dm}ZH-2A w*; PBac{y+-attP-3B}VK37 ( Bloomington Drosophila Stock Center ) , ( 10 ) FRT 82B SerRX106/TM6 , Tb1 , ( Venken et al . , 2006 ) , ( 11 ) Serrev2−11/TM3 , Sb1 Ser1 , Df ( 3R ) Ser+R82f24 Tb1/TM3 , Sb1 Ser1 , ( Fleming et al . , 1990 ) , ( 12 ) P{Dlgt-wt}attP2 , ( 13 ) PBac{Dlgt-wt}VK37 , ( 14 ) PBac{Sergt-wt}VK37 ( this study ) . To generate a Delta genomic transgene , an 83 . 5-kb fragment containing the Delta locus and its flanking regions was transferred from BACR48K23 ( BACPAC Resources Center , CHORI ) to the attB-P[acman]-ApR vector by using ‘recombineering-mediated gap repair’ ( Venken et al . , 2009 ) . To generate a Delta genomic transgene , an 83 . 5-kb fragment containing the Delta locus and its flanking regions was transferred from BACR48K23 ( BACPAC Resources Center , CHORI ) to the attB-P[acman]-ApR vector by using ‘recombineering-mediated gap repair’ ( Muller et al . , 2014 ) . First , left and right homology arms ( LA and RA , respectively ) for the region of interest flanked by appropriate enzymes were generated by PCR using BACR48K23 as template and the following primers ( restriction sites underlined ) :Dl-LA-Long-AscI-FAGGCGCGCCATGCCATGACTGTTGTACTAACATAATGADl-LA-Long-BamHI-RAACGGATCCATATACCCTAGCTGTGCGGTAGTTCATDl-RA-Long-BamHI-FTTAGGATCCGCTGCGTGTGCTCTATAGGTGGTCATTADl-RA-Long-PacI-RCGGTTAATTAATCGGGATCGGCTTGCGGATCGTCAT The LA and RA were then cloned into the AscI-PacI digested attB- P[acman]-AmpR vector via three-way ligation . The resulting construct was linearized by BamHI digestion and used to retrieve the 83 . 5-kb fragment from BACR48K23 by recombineering as described previously ( Groth et al . , 2004 ) . All exons and exon-intron boundaries were verified by sequencing . To generate a Serrate genomic transgene , a BAC clone with an 81 . 9-kb insert harboring the Serrate locus and its flanking regions was obtained from BACPAC Resources Center ( attB- P[acman]-CmR-CH321-69C08 [Venken et al . , 2009] ) . ΦC31-mediated integration ( Venken et al . , 2006; Bischof et al . , 2007 ) was used to insert the Delta and Serrate constructs into the attP2 and VK37 docking sites and to generate the P{Dlgt-wt}attP2 ( Dlgt-wt ) , PBac{Dlgt-wt}VK37 ( Dlgt-wt ) and PBac{Sergt-wt}VK37 ( Sergt-wt ) transgenes . Adult wings were separated from anesthetized flies , incubated in 100% ethanol for 3 min , dried , and mounted in the DPX medium ( Electron Microscopy Sciences , Hatfield PA ) . Wing images were obtained by using an AxioCam MRc5 camera mounted on a Zeiss Axioplan 2 microscope . In order to interpret cis interaction measurements , we built on the model of interactions among receptors and ligands from Sprinzak et al . ( 2010 ) . This model was used to generate the schematic plots in Figure 1A , as well as to fit the single-cell data in Figures 3 and 4 . The model is based on two reactions between the Notch in cell i ( Ni ) , and its interactions with ligands in the same cell ( Di ) , or a neighboring cell j ( Dj ) , Ni+ Dj⇋[NiDj]→ Si Trans-activation , with association and dissociation rates kD± and activation rate kS . Ni+ Di ⇋[NiDi]→∅ Cis-inhibition , with association and dissociation rates kC± and inactivation rate kI . Notch is created with a constant rate βN and degraded at a linear rate γNNi . Ligand is produced at a constant rate βD and degraded with a linear rate γDDi . NiDj represents the complex of a Notch receptor in cell i bound to a ligand in cell j . When i = j this terms describes a cis interaction . Because we chose cell-plating conditions to isolate cis interaction , we can ignore the trans-activation terms . These two reactions can be rewritten as a set of ordinary differential equations:d[Ni]dt= βN−γNNi− ( kC+NiDi−kc−[NiDi] ) d[Di]dt= βD−γDDi− ( kC+NiDi−kc−[NiDi] ) d[NiDi]dt= kC+NiDi− kC−[NiDi]−kI[NiDi] . We assume that the bound cis-complex achieves a quasi-steady state ( d[NiDi]dt≈0 ) . Using this assumption , we derive the following relationship:Nsteady state= βNγN1+ Dsteady statekcγN where kc−1 is defined as kC+kIkC−+kI . The physical meaning of this expression is that available Notch receptor is a decreasing function of cis-Delta concentration . When there is no cis-Delta expression , steady state receptor levels are βN/γN . However , as cis-Delta increases , the level of available Notch drops below this maximal value . The amount of Delta necessary to deplete Notch receptor to one half of its maximal concentration in the absence of cis-Delta is kcγN . | In animals , cells use a process called Notch signaling to communicate with neighboring cells . During this process , a protein known as a DSL ligand from one cell binds to a protein called a Notch receptor on a neighboring cell . This triggers a series of events in the neighboring cell that change how the genes in this cell are expressed . Notch signaling is involved in many processes including the early growth of embryos , the formation of organs and limbs , and the maintenance of stem cells throughout adult life . Enzymes called Fringe enzymes can control Notch signaling by blocking or promoting the formation of the DSL ligand-Notch receptor pairs . It is also possible for a DSL ligand and a Notch receptor from the same cell to interact . This is thought to be important because it prevents an individual cell from sending and receiving Notch signals at the same time . There are several different DSL ligands , Notch receptors and Fringe enzymes , so it is difficult to determine which configurations of receptors , ligands and Fringe enzymes can enable Notch signals to be sent or received . To address this problem , LeBon et al . investigated how Fringe enzymes acted on several different DSL-Notch receptor pairs in mammalian cells , and also in fruit flies . They focused in particular on the interactions that occurred within the same cell , as the role of Fringe enzymes in this type of interaction has not been examined previously . The experiments revealed that Fringe proteins modify specific same-cell interactions in a way that enables a cell to receive one type of Notch signal from a neighboring cell and send a different type of Notch signal to another cell at the same time . More generally , these results show how an unconventional , ‘bottom-up’ approach can reveal the design principles of cell signaling systems , and suggest that it should now be possible to use these principles to try to understand which cell types send signals to which other cell types in many different contexts . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2014 | Fringe proteins modulate Notch-ligand cis and trans interactions to specify signaling states |
Gliosis defined as reactive changes of resident glia is the primary response of the central nervous system ( CNS ) to trauma . The proliferation and fate controls of injury-reactivated glia are essential but remain largely unexplored . In zebrafish optic tectum , we found that stab injury drove a subset of radial glia ( RG ) into the cell cycle , and surprisingly , proliferative RG responding to sequential injuries of the same site were distinct but overlapping , which was in agreement with stochastic cell-cycle entry . Single-cell RNA sequencing analysis and functional assays further revealed the involvement of Notch/Delta lateral inhibition in this stochastic cell-cycle entry . Furthermore , the long-term clonal analysis showed that proliferative RG were largely gliogenic . Notch inhibition of reactive RG , not dormant and proliferative RG , resulted in an increased production of neurons , which were short-lived . Our findings gain new insights into the proliferation and fate controls of injury-reactivated CNS glia in zebrafish .
Traumatic brain injury ( TBI ) is one clinically principal type of central nervous system insults ( Burda and Sofroniew , 2014 ) . Gliosis defined as reactive changes of resident macroglia ( e . g . , mammalian astrocytes ) is a primary CNS response to TBI in mammals ( Barres , 2008; Burda and Sofroniew , 2014 ) . In mammals , gliosis undergoes three significant stages: Glial cells initially become reactive , hypertrophic , and inflammatory , with characteristic upregulation of GFAP and vimentin ( Liddelow and Barres , 2017; Zamanian et al . , 2012 ) ; subsequently , a subset of reactive glia re-enter the cell cycle and become proliferative ( Gallo and Deneen , 2014 ) ; finally , proliferative glia undergo gliogenesis , the process of glial cell production , and form structures known as glial scars ( Burda and Sofroniew , 2014 ) . Earlier studies have demonstrated both protective and detrimental roles of the gliosis in the injured CNS ( Faulkner et al . , 2004; Li et al . , 2008; Silver and Miller , 2004; Sofroniew and Vinters , 2010; Wanner et al . , 2013 ) . For instance , the blockage of initial glia reactivation worsened the injury ( Faulkner et al . , 2004; Li et al . , 2008; Wanner et al . , 2013 ) , whereas glia scars hindered neuronal regeneration ( Silver and Miller , 2004 ) . In the process , the proliferation and fate controls of injury-reactivated RG are essential but remain elusive in vivo . In contrast to the mammalian CNS , teleost fish exhibit a superior neural regeneration in response to TBI beyond embryonic development ( Baumgart et al . , 2012; Grandel et al . , 2006; Kishimoto et al . , 2012; Reimer et al . , 2008; Than-Trong and Bally-Cuif , 2015 ) . Radial glia ( RG ) , the primary form of macroglia in teleost fish , are the main cell source for injury-mediated regeneration ( Than-Trong and Bally-Cuif , 2015 ) . For instance , RG of different brain regions in zebrafish , including the telencephalon , the hypothalamus , and the spinal cord can produce newborn neurons in response to the injury ( Duncan et al . , 2016; Goldshmit et al . , 2012; Johnson et al . , 2016; Kizil et al . , 2012; Kroehne et al . , 2011; Than-Trong and Bally-Cuif , 2015 ) . Also , retinal müller glia ( MG ) can be reactivated by the injury , giving rise to newborn retinal neurons ( Goldman , 2014; Gorsuch and Hyde , 2014 ) . Molecular mechanisms underlying the proliferation and fate controls of injury-reactivated RG have been examined in zebrafish for many years ( Dias et al . , 2012; Goldman , 2014 ) . As to the proliferation control , Notch signaling is involved but is somehow context-dependent . For instance , in the zebrafish spinal cord , glial cells have low levels of Notch activity when they are in the dormant state , and enter the cell cycle by increased Notch activity after the injury ( Dias et al . , 2012 ) . In contrast , dormant RG of zebrafish telencephalon exhibit high Notch activity and become proliferative by a rapid decrease of Notch activity ( Chapouton et al . , 2010 ) . Notch signaling has also been reported to regulate fate outputs of reactivated MG in the injured zebrafish retina , that is , Notch inhibition leads to gliogenesis , whereas Notch over-activation results in the production of photoreceptor cells ( Wan et al . , 2012 ) . Zebrafish optic tectum , the higher sensory integration center , possesses a large population of RG ( Galant et al . , 2016; Ito et al . , 2010 ) . Unlike other brain regions where RG present as both dormant and proliferative forms at physiological conditions , tectal RG has been reported to be dormant , and are reactivated by injury to give rise to newborn neurons via Wnt signaling as well as Notch signaling ( Shimizu et al . , 2018; Ueda et al . , 2018 ) . Interestingly , a recent study showed that tectal RG produced a significant number of glial cells ( ~25% ) but not neurons ( Lindsey et al . , 2019 ) . It is essential to resolve this inconsistency on the fate potentials of injury-reactivated tectal RG . In this study , we set out to investigate the mechanism controlling injury responses of tectal RG in vivo . We found that stab injury drove a subset of tectal RG into the cell cycle . Surprisingly , proliferative tectal RG responding to the sequential injuries at the same injury site were distinct but overlapping . Quantitative analysis showed the probability of proliferative RG responding to both sequential stab injuries could be well explained by a model incorporating stochastic cell-cycle entry at the fixed probability of ~25% . Single-cell RNA-seq and functional analysis revealed this stochastic cell-cycle entry was dependent on Notch/Delta lateral inhibition . The clonal analysis showed that proliferative tectal RG underwent gliogenesis . Interestingly , post-injury notch inhibition drove reactive RG into the cell cycle and resulted in increased neurogenesis . Interestingly , the over-produced neurons mostly diminished by approximately 25 days post-injury ( dpi ) .
Consistent with earlier studies ( Galant et al . , 2016; Ito et al . , 2010; Jung et al . , 2012 ) , our results showed that Tg ( gfap:GFP ) ( Bernardos and Raymond , 2006 ) , Tg ( her4 . 1:dRFP ) ( Yeo et al . , 2007 ) , and the antibody against glutamine synthetase ( GS ) specifically labeled RG that line the tectal ventricle and extend their basal processes into the superficial neuropils ( Figure 1A–B3 and Figure 1—figure supplement 1A-C2 ) . To examine the dormancy of tectal RG , we quantified the proliferative RG in the optic tectum of Tg ( gfap:GFP ) fish using the antibodies against proliferating cell nuclear antigen ( PCNA ) , glutamine synthetase ( GS ) , and GFP ( Figure 1A–B3 ) . Only 1 . 4 ± 0 . 2% ( n = 5 , mean ± SEM ) of GFP+/GS+ RG at the bottom of the periventricular gray zone ( PGZ ) were PCNA+ ( Figure 1A–B3 ) . Bromodeoxyuridine ( BrdU ) is a nucleotide analog that incorporates into new synthesized DNA of dividing cells in the S phase . We noticed that 1 day’s administration of BrdU labeled very few tectal RG under physiological conditions ( Figure 1—figure supplement 1D-F1 ) . Together , these results demonstrate that zebrafish tectal RG are largely dormant under normal physiological conditions . Next , we set out to investigate whether injury induces RG proliferation . We applied a stab injury to the central-dorsal part of the optic tectum ( Figure 1C ) . At 3 dpi , immunostaining showed marked expression of PCNA in some RG underneath the injury site but not in the uninjured control hemisphere ( Figure 1D–G ) . BrdU incorporation and staining experiments showed that injury-reactivated RG were labeled by BrdU at 3 dpi , indicating that dormant RG entered S phase after injury ( Figure 1—figure supplement 2A–C ) . To confirm the proliferation of injury-reactivated RG , we took advantage of the Cre-loxP system to do the long-term clonal analysis of single tectal RG ( Kroehne et al . , 2011 ) . We used the Tg ( her4 . 1:mCherryT2ACreERT2 ) transgenic line , in which the promoter of her4 . 1 drives the expression of the mCherry fluorescent protein and CreERT2 recombinase in tectal RG ( Figure 1H ) . By crossing this line with Tg ( hsp70l:DsRed2 ( floxed ) EGFP ) ( Figure 1H ) , individual RG in the uninjured optic tectum can be genetically labeled by tamoxifen ( TAM ) administration ( at 2 to 3 months old ) and the resulting clones visualized by EGFP after heat shock at the desired time points ( Figure 1—figure supplement 2D and E ) . We combined the Cre-loxP system and ( 5-ethynyl-2′-deoxyuridine ) EdU pulses for six consecutive days after the injury to perform the clonal analysis of single RG after stab injury ( Figure 1I ) . Final clones were analyzed at desired time points after a pulse of EGFP expression by heat shock ( Figure 1I ) . At 8 dpi , we observed marked RG derived clones ( EGFP+/EdU+ , Figure 1J–K3 ) . In total 29 clones were collected ( from 11 fish ) , of which ~ 69% were 2 cell clones ( 20/29; Figures 1M–M2 , O and O1 ) , ~24% were 1 cell clones ( 7/29; Figures 1L–L2 , O and O1 ) , and the rest were 3 cell clones ( 2/29; Figure 1N–O1 ) . Taken together , dormant tectal RG are capable of proliferation after stab injury . We then examined the injury responses of RG in different geographical regions of the optic tectum . Stab injury was applied at five different regions , including the anterior-dorsal , central-dorsal ( as positive control region ) , medial-dorsal , lateral , and posterior-dorsal regions of the right hemisphere ( Figure 1—figure supplement 2F–O ) . We stained and quantified the PCNA+ RG underneath the injury sites in all regions , showing that the number of PCNA+ RG was not significantly different across regions except for the medial-dorsal region , where the RG had little proliferative capacity at 3 dpi ( Figure 1—figure supplement 2H , M and P ) . To confirm this , we performed two-sites injury on the same hemisphere of the optic tectum , the medial-dorsal region , and the central-dorsal region were injured ( Figure 1—figure supplement 2Q–S ) . Consistently , significantly more PNCA+ RG were in the central-dorsal region than in the medial-dorsal region ( Figure 1—figure supplement 2T ) . Together , our results indicate that stab injury can induce RG proliferation across distinct regions in the optic tectum except in the medial-dorsal region . To investigate the cell-cycle entry of tectal RG , we examined PCNA expression in tectal RG of Tg ( 1016tuba1α:GFP ) , a transgenic line used as the reporter for retinal MG reactivation after the injury ( Fausett and Goldman , 2006 ) . Under physiological conditions , weak GFP signals were present in tectal RG ( Figure 2A and F ) . At 1 dpi , robust GFP signals were already observed together with the upregulation of Vimentin , a hallmark of glial reactivation at the early stage of gliosis ( Figure 2—figure supplement 1A-B1 ) ( Liddelow and Barres , 2017 ) , whereas only few of RG was PCNA+ ( Figure 2B , G and Figure 2—figure supplement 1C ) , suggesting a lack of proliferation . The number of reactive RG , the ones with robust GFP signals , peaked at 3 dpi ( Figure 2C , H and K ) , while a significant number of PCNA+ RG first occurred at 2 dpi and peaked at approximately 3–4 dpi ( Figure 2K and Figure 2—figure supplement 1C ) . From 4 to 7 dpi , the number of PCNA+ RG gradually dropped back to the same level observed before the injury ( Figure 2K and Figure 2—figure supplement 1C ) . At 3 and 5 dpi , we found some robust GFP signals and PCNA+ cells at the injury site and in the region underneath ( Figure 2C and D ) . The GFP signals were likely due to hypertrophic responses of RG’s processes and other cells at the injury site ( Figure 2D ) , while PCNA+ cells likely consisted of oligodendrocytes precursor cells , recruited microglia/macrophage and other cell types ( Figure 2—figure supplement 1D-H3 ) . We further measured the spatial relation between reactive RG ( with robust GFP+ signal ) and proliferative RG ( GFP+/PCNA+ ) ( Figure 2L ) . In coronal sections , reactive RG were primarily distributed in an area with a width of 186 ± 4 μm ( n = 7 , mean ± SEM ) underneath of the injury sites ( ‘Reactive Zone’ ) while the majority of proliferative RG ( 88 ± 3% , n = 7 , mean ± SEM ) were located in an area with the width of 76 ± 5 μm ( n = 7 , mean ± SEM ) underneath of the injury site ( ‘Proliferative Zone’ ) ( Figure 2L and M ) . The small variation of the width of both zones indicated the high reproducibility of stab injury outcomes ( Figure 2M ) . Although the injury reactivated all RG underneath the injury site , only a subset of them ( ~25% , n = 8 ) became proliferative ( Figure 2H and Figure 2—figure supplement 1C ) . It raised an immediate question as to whether the proliferation of a subset of RG was due to stochastic cell-cycle entry or the presence of distinct RG subpopulations that respond differentially to the stab injury . To test this , we designed a sequential stab injury experiment . We examined the responses of reactive tectal RG to two sequential stab injuries performed at the same physical site ( Figure 2N ) . The first injury was introduced followed by BrdU pluses for six consecutive days to label proliferative RG responding to the first injury , and the second injury was introduced at 12 dpi followed by EdU pluses for six consecutive days to mark proliferative RG responding to the second injury ( Figure 2N ) . Finally , the fish were sacrificed , and coronal sections were stained for BrdU , EdU , radial glial marker BLBP , and neuronal marker HuC/D at 23 dpi ( Figure 2O–O5 ) . We found that although the number of proliferative RG induced by the first and the second injury showed no significant difference ( the first injury: 84 . 4 ± 15 . 0 cells , n = 8; the second injury: 83 ± 9 . 4 cells , n = 8; mean ± SEM; p>0 . 05; Figure 2P ) , two sets of proliferative RG were distinct with some degree of overlapping ( 23 ± 6 , n = 8 , mean ± SEM; Figure 2O–O5 and Figure 2P ) . More importantly , the proportion of overlapping RG ( those reactivated after both injuries ) was statistically indistinguishable from the multiplication of the reactivation probabilities of either injury , which suggested that individual reactive RG entered the cell cycle in the stochastic manner ( prediction: 7 . 1 ± 1 . 9% , n = 8; experiment: 6 . 8 ± 1 . 7% , n = 8; mean ± SEM; p>0 . 05; Figure 2Q ) . To further examine the molecular mechanism underlying this stochastic cell-cycle entry of injury-reactivated RG , we carried out single-cell RNA sequencing ( scRNA-seq ) analysis of tectal RG at 3 dpi , at which stage the number of proliferative tectal RG nearly reached the plateau in terms of cell number ( Figure 2K and Figure 2—figure supplement 1C ) . We dissected and dissociated the optic tecta of Tg ( gfap:GFP ) fish at 3 dpi and sorted out GFP+ RG using fluorescence-activated cell sorting ( FACS ) for further scRNA-seq on the 10x Genomics platform ( Figure 3A and Figure 3—figure supplement 1A-1A2 ) . The gene profiles of in total 2998 single cells were qualified after the initial filtering using the Seurat algorithm ( Figure 3—figure supplement 1B and C; see also details in Materials and Methods; http://satijalab . org/seurat/ ) . We performed raw cell clustering using t-stochastic neighbor embedding ( t-SNE ) analysis ( Figure 3—figure supplement 1B and C ) . According to cell-type-specific genes , we excluded the cell clusters representing non-glial cells ( Figure 3—figure supplement 1D , E ) , RG in the tectal proliferation zone ( TPZ ) , and oligodendrocytes ( Figure 3—figure supplement 2A–F; see also details in Materials and Methods ) . The remaining 1174 cells exhibited radial glial characteristics and were thus used for further analysis . They were segregated into five major cell clusters using t-SNE analysis ( Figure 3B ) . Each cell cluster had a characteristic gene expression ( Figure 3C ) . Cluster 1 cells ( RG of dormant state , dRG ) constituted the most abundant cell population with the high expression of milk-fat globule-epidermal growth factor 8a ( mfge8a , Figure 3D ) , whose analog , mfge8 , is a phagocytosis factor that maintains the pool of radial glia-like cells by controlling cellular quiescence in mice ( Zhou et al . , 2018 ) . Cluster 2 cells ( RG of reactive state ) were characterized by their up-regulation of vimentin ( vim ) , a hallmark of RG reactivation ( Figure 3E and Figure 2—figure supplement 1A-B1 ) . Proliferative RG were composed of RG of proliferative-S ( mcm2 and pcna , cluster 3; Figure 3F and G ) and proliferative-G2 states ( cdk1 and nusap1 , cluster 4; Figure 3H and I ) . Cluster 5 cells highly expressed vimentin ( vim ) and were likely to represent Vimentin+ cells from neighboring tissues under the optic tectum in the midbrain due to possible contamination during the dissection of the optic tecta ( Figure 3E and Figure 3—figure supplement 2G-2I1 ) . We did not observe such a high expression of vim in the tectal RG ( Figure 3E ) . Thus , we excluded cluster 5 cells from further analysis . Cell cycle phases analysis ( Figure 3J ) and pseudo-time analysis ( Figure 3K and Figure 3—figure supplement 2J ) were performed and suggested the temporal order of 4 remaining cell clusters , thereafter termed as the state of dormant RG ( dRG ) , the state of reactive RG ( reactive RG ) , the state of proliferative-S RG and the state of proliferative-G2 RG . Next , we looked into the expression dynamics of the genes that differentially expressed across the states . mfge8a was abundant in dormant RG ( cluster 1 ) , began to decrease in reactive RG ( cluster 2 ) and became rapidly diminished in proliferative RG ( cluster 3 and 4 ) ( Figure 3L ) . Kruppel-like transcription factor 6a ( klf6a ) , the transcription factor essential for optic axon regeneration ( Veldman et al . , 2010; Veldman et al . , 2007 ) , exhibited a peaked expression in reactive RG ( cluster 2 ) ( Figure 3M ) , and Insulinoma-associated 1a ( insm1a ) , encoding a transcriptional repressor that has been reported to be necessary for MG-based retina regeneration ( Forbes-Osborne et al . , 2013; Ramachandran et al . , 2012 ) , highly expressed in proliferative-S and -G2 RG ( cluster 3 and 4 ) ( Figure 3N ) . To verify their expression , we performed in situ hybridization . The results were consistent with our scRNA-seq data , mfge8a was down-regulated in injured-induced PCNA+ proliferative RG at 3 dpi ( Figure 3O–P1 ) , whereas klf6a and insm1a mRNA expression increased in the 2-dpi ( Figure 3Q–R1 ) and 3-dpi ( Figure 3S–T1 ) optic tecta , respectively . Interestingly , the signals of klf6a ( Figure 3Q and Q1 ) and insm1a ( Figure 3S and S1 ) were mainly distributed in the processes of RG . Notably , during the transition of reactive ( cluster 2 ) and proliferative states ( cluster 3 and 4 ) , the expression of her4 . 1 , the targeting gene of Notch signaling ( Takke et al . , 1999 ) , decreased ( Figure 4A and B ) , whereas deltaA expression increased ( Figure 4C and D ) . Further correlation analysis showed that pcna and deltaA expression were correlated , while pcna and deltaA were uncorrelated with the expression of her4 . 1 and her4 . 2 ( Figure 4E ) . Our results suggest proliferative RG with an increase of deltaA expression and a decrease of Notch activity . To visualize the Notch/Delta dynamics in vivo , we employed a reporter line Tg ( Tp1bglob:EGFP ) ( hereafter referred to as Tg ( Tp1:EGFP ) ) , in which EGFP is driven by the TP1 element , the direct target of the intracellular domain of Notch receptors ( NICD ) that is generated upon Notch activation ( Parsons et al . , 2009; Quillien et al . , 2014 ) . We performed PCNA immunostaining on the coronal sections of Tg ( Tp1:EGFP ) at 3 dpi ( Figure 5F–G3 ) . Interestingly , the results showed ~82% ( 97/119 cells , n = 6 sections ) of PCNA+ proliferative RG had no EGFP signal , indicating low Notch activity ( Figure 4H ) . Notch activity and PCNA signal were mostly exclusive ( Figure 4F1–F3 ) . Consistently , in situ hybridization of deltaA followed by immunostaining of PCNA showed ~81% ( 60/74 cells , n = 10 sections ) of PCNA+ RG expressed deltaA ( Figure 4I–4J ) . Our results suggest that Notch/Delta lateral inhibition may be at work . As Notch/Delta lateral inhibition contributes to the mosaic entry of embryonic neurogenesis of neural progenitor cells ( Cabrera , 1990; Dong et al . , 2012; Formosa-Jordan et al . , 2013; Kageyama et al . , 2008; Sato et al . , 2016; Tiedemann et al . , 2017 ) , we examined its role in the cell-cycle entry of injury-reactivated RG . We disturbed Notch/Delta lateral inhibition by blocking Notch signaling using LY411575 , a potent inhibitor of the γ-secretase complex , which acts by preventing the cleavage of NICD ( Figure 5A ) ( Geling et al . , 2002; Katz et al . , 2016 ) . Notably , 2 days’ LY411575 treatment resulted in a ~ 4 folds increase in the number of proliferative RG in the injured optic tectum ( DMSO-treated: 24 . 8 ± 3 . 7 cells , n = 6; LY411575-treated: 109 ± 16 . 9 cells , n = 5; mean ± SEM; ***p<0 . 001; Figure 5B , C , F , G , and J ) . RO4929097 , another Notch signaling inhibitor , resulted in a similar phenotype ( Figure 5—figure supplement 1A–D ) . Consistent with the findings of a recent study ( Ueda et al . , 2018 ) , Notch inhibition was also sufficient to trigger the proliferation of tectal RG even without any injury ( DMSO-treated: 6 . 3 ± 0 . 48 cells , n = 4; LY411575-treated: 26 . 7 ± 2 . 4 cells , n = 5; mean ± SEM; p>0 . 05; Figure 5D , E , H , I , J , Figure 5—figure supplement 1C and D ) , which was reminiscent of the increase of constitutively proliferative RG in the zebrafish telencephalon by Notch inhibition ( Chapouton et al . , 2010 ) . Furthermore , we took advantage of Tg ( hsp70l:gal4×UAS:NICD-Myc ) double-transgenic fish , in which a heat shock promoter drives mosaic expression of the NICD-Myc fusion protein , allowing conditional and potent over-activation of Notch signaling ( Figure 5K ) ( Scheer et al . , 2001 ) . Mis-expression of NICD significantly blocked the cell-cycle entry of tectal RG following stab injury , that is , ~94% ( 31/33 cells , n = 15 sections ) of NICD-overexpressed RG underneath the injury sites were PCNA− ( Figure 5L–N and Figure 5—figure supplement 1E-E3 ) . Torus semicircularis ( TS ) is the midbrain tissue under the PGZ of the central optic tectum , and their boundary could be unambiguously defined by DAPI staining ( Figure 1—figure supplement 1A-C2 and Figure 5—figure supplement 1F-G3 ) . We noticed that stab injury induced some cells in the TS underneath the injury site ( close to the boundary of TS and PGZ ) to become proliferative in some animals , which required further investigation ( Figure 5L–L3 ) . In sum , Notch inhibition mediates the stochastic cell-cycle entry of reactive RG after the injury . To examine the fate outputs of proliferative RG after the injury , we utilized the Cre-loxP system to perform the clonal analysis of single RG after stab injury and analyze clonal cell-type compositions by immunostaining of BLBP , a putative maker for RG , and HuC/D , a putative marker for neurons . Notably , the newborn cells were largely BLBP positive , indicative of RG identity ( Figure 6A–B3 ) . These results raised an immediate question as to whether injury-induced proliferative RG are gliogenic . To examine this at the population level , we injected wild-type fish with EdU for six consecutive days after the injury and analyzed EdU+ cells underneath the injury sites at 7 dpi combined with immunostaining for BLBP and HuC/D ( Figure 6C ) . The results showed a significant increase of newborn cells ( EdU+; uninjured: 7 . 6 ± 2 . 1 cells , n = 7; injured: 86 . 6 ± 6 . 5 cells , n = 8; mean ± SEM; ***p<0 . 001; Figure 6D–H ) . Notably , EdU+ newborn cells were largely EdU+/BLBP+ RG ( 78 . 6 ± 5 . 9 cells , 90 . 5 ± 1 . 4% of total EdU+ cells , n = 8 , mean ± SEM ) rather than EdU+/HuC/D+ newborn neurons ( 3 ± 1 . 0 cells , 3 . 3 ± 0 . 9% of total EdU+ cells , n = 7 , mean ± SEM ) in the injured hemisphere , indicating that tectal RG largely undergo gliogenesis ( Figure 6H ) . As a consequence , glial bulges formed underneath the injury sites ( Figure 6D–E3 ) . More importantly , when we analyzed EdU+ newborn cells at ~300 dpi ( Figure 6I ) , the glial bulges remained and were still largely composed of EdU+/BLBP+ RG ( 111 . 3 ± 9 . 4 cells , 93 . 6 ± 0 . 7% of total EdU+ cells , n = 3 , mean ± SEM; Figure 6J–N ) . Only a few EdU+ cells were EdU+/HuC/D+ neurons ( 2 ± 0 . 6 cells , 1 . 6 ± 0 . 4% of total EdU+ cells , n = 3; mean ± SEM , Figure 6J–N ) . EdU+/HuC/D+ neurons were found both in the deep and upper regions of the injured optic tectum ( Figure 6J–K3 and Figure 6—figure supplement 1A-B3 ) . Together , stab injury triggers the gliogenesis of tectal RG , resulting in the formation of glial bulges in the zebrafish optic tectum . After the injury , we often observed a physical wound at the injury site on the surface of the optic tectum ( 1343 ± 315 . 7 μm2 , n = 10 , mean ± SEM; Figure 6—figure supplement 1C–E ) . More strikingly , these stab wounds remained up to 300–400 dpi ( 1339 ± 768 . 6 μm2 , n = 7 , mean ± SEM; p>0 . 05; Figure 6—figure supplement 1E and F-I3 ) . These wounds were surrounded by BLBP signals but without cell nuclei , suggesting that the hypertrophic processes of RG formed a glial scar-like structure surrounding the wound , and thereby blocking the repair of the wound ( Figure 6—figure supplement 1A-B3 and F-I3 ) . Our results suggest a limited regenerative capacity of the adult zebrafish optic tectum . Down-regulation of Notch signaling is profoundly implicated in the production of neurons during embryonic CNS development ( Beatus and Lendahl , 1998; Artavanis-Tsakonas et al . , 1999 ) . We , therefore , wondered whether Notch inhibition could promote the neurogenesis of proliferative RG . As the number of proliferative RG peaked at 3–4 dpi ( Figure 2K and Figure 2—figure supplement 1C ) , we examined the fate outputs of tectal RG labeled by EdU at 1–6 dpi with Notch inhibition by LY411575 during either 1–3 dpi or 4–5 dpi ( Figure 7A ) . Interestingly , although Notch inhibition during both time windows significantly increased the number of newborn cells ( EdU+ ) compared to the control with DMSO treatment during 1–6 dpi ( DMSO ( 1–6 dpi ) : 75 . 3 ± 6 . 6 cells , n = 6; LY411575 ( 1–3 dpi ) : 215 . 4 ± 17 . 2 cells , n = 7 , ***p<0 . 001; LY411575 ( 4–5 dpi ) : 127 . 7 ± 12 . 4 cells , n = 9 , *p<0 . 05; mean ± SEM; Figure 7B–E ) , newborn neurons ( EdU+/HuC/D+ ) dramatically increased however only in the group in which Notch was inhibited during 4–5 dpi ( ( DMSO ( 1–6 dpi ) : 4 . 2 ± 1 . 2 cells , n = 6; LY411575 ( 1–3 dpi ) : 6 . 4 ± 2 . 1 cells , n = 7 , p>0 . 05; LY411575 ( 4–5 dpi ) : 14 . 2 ± 2 . 1 cells , n = 9 , **p<0 . 01; mean ± SEM; Figure 7B–D and F ) . The proportion of newborn neurons increased from 5 . 3 ± 1 . 1% ( DMSO ( 1–6 dpi ) , n = 6 , mean ± SEM ) to 11 . 9 ± 1 . 7% ( LY411575 ( 4–5 dpi ) , n = 9 , mean ± SEM; *p<0 . 05 ) ( Figure 7G ) . Note that those over-produced neurons always existed as cell clusters , suggesting that they might be clonally related ( Figure 7—figure supplement 1A–F ) . To look into this increased neurogenesis , we further examined fate outputs of tectal RG , which were labeled by EdU during either 1–3 dpi ( Figure 7H ) or 4–6 dpi ( Figure 7I ) , with the same treatment of LY411575 ( 4–5 dpi ) . The control groups were treated with 0 . 1% DMSO ( 4–5 dpi ) . Fish were sacrificed and cryosections were obtained to stain for EdU and neuronal marker HuC/D at 7 dpi . We found that newborn cells ( EdU+; DMSO-treated: 58 . 4 ± 5 . 3 cells , n = 24; LY411575-treated: 53 . 0 ± 3 . 2 , n = 35 cells; mean ± SEM; p>0 . 05 ) , newborn neurons ( EdU+/HuC/D+; DMSO-treated: 3 . 4 ± 0 . 3 cells , n = 24; LY411575-treated: 3 ± 0 . 5 , n = 35; mean ± SEM; p>0 . 05 ) derived from the proliferative RG labeled during 1–3 dpi with or without Notch inhibition ( Figure 7H ) were indistinguishable in terms of cell number ( Figure 7J , K , N and O ) . Moreover , the proportion of newborn neurons showed no significant difference in Notch inhibited fish compared to control fish ( DMSO-treated: 6 . 4 ± 0 . 6% , n = 24; LY411575-treated: 5 . 4 ± 0 . 9% , n = 35; mean ± SEM; p>0 . 05 ) , which suggested that Notch inhibition itself did not promote the neurogenesis of injury-induced proliferative RG ( Figure 7P ) . In contrast , the RG labeled during 4–6 dpi with Notch inhibition ( 4–5 dpi ) ( Figure 7I ) gave rise to much more newborn cells ( EdU+ ) than the control ( DMSO-treated: 52 . 5 ± 6 . 6 cells , n = 12; LY411575-treated: 107 ± 10 . 9 cells , n = 18; mean ± SEM; ***p<0 . 001; Figure 7L , M and N ) , which was likely due to the cell-cycle entry of reactive RG by Notch inhibition during 4–5 dpi . Interestingly , these increased newborn cells derived from RG labeled during 4–6 dpi with Notch inhibition had a much higher proportion of neurons ( DMSO-treated: 4 . 2 ± 1 . 2 cells , 6 . 8 ± 1 . 9% , n = 12; LY411575-treated: 22 . 1 ± 3 . 5 cells , 19 . 2 ± 2 . 4% , n = 18; mean ± SEM; ***p<0 . 001; Figure 7L , M , O and P ) . Together , our results suggest that post-injury Notch inhibition ( 4–5 dpi ) drives injury-induced reactive RG ( non-proliferation ) into the cell cycle , producing a significantly higher proportion of neurons compared to those derived from proliferative RG by either the injury ( 5 . 3 ± 1 . 1% , n = 6 , mean ± SEM ) or Notch inhibition ( 1 . 2 ± 0 . 6% , n = 7 , mean ± SEM ) ( Figure 7G and Q ) . We further investigated the long-term fate of these newborn neurons . We sacrificed and sectioned the DMSO-treated ( 4–5 dpi ) control fish , which were injected with EdU during 4–6 dpi , at either 7 dpi or 25 dpi ( Figure 7—figure supplement 1G ) . The coronal sections were then stained for EdU signals , neuronal marker HuC/D and RG marker BLBP ( Figure 7—figure supplement 1H–K ) . The number of newborn cells ( EdU+; 7-dpi DMSO-treated: 25 . 7 ± 2 . 5 cells , n = 10; 25-dpi DMSO-treated: 41 . 3 ± 6 . 1 cells , n = 11; mean ± SEM; *p<0 . 05 ) and newborn RG ( EdU+/BLBP+; 7-dpi DMSO-treated: 18 . 1 ± 2 . 5 cells , n = 10; 25-dpi DMSO-treated: 33 . 9 ± 5 . 8 cells , n = 11; mean ± SEM; *p<0 . 05 ) were increased slightly at 25-dpi compared to 7-dpi DMSO-treated fish ( Figure 7—figure supplement 1H , J , L and M ) , which might be due to the variability among different fish . However , the number of newborn neurons ( EdU+/HuC/D+; 7-dpi DMSO-treated: 2 . 4 ± 0 . 9 cells , n = 10; 25-dpi DMSO-treated: 2 . 1 ± 0 . 3 cells , n = 11; mean ± SEM; p>0 . 05 ) showed no significant difference ( Figure 7—figure supplement 1N ) . Interestingly , the number of newborn neurons in LY411575-treated fish decreased significantly at 25 dpi ( EdU+/HuC/D+; 7-dpi LY411575-treated: 15 . 9 ± 2 . 7 cells , n = 10; 25-dpi LY411575-treated: 2 . 4 ± 0 . 3 cells , n = 14; mean ± SEM; ***p<0 . 001 ) and became indistinguishable from the DMSO-treated fish ( p>0 . 05; Figure 7—figure supplement 1I , K and N ) . Meanwhile , the number of newborn cells ( EdU+; 7-dpi LY411575-treated: 106 . 9 ± 11 . 8 cells , n = 10; 25-dpi LY411575-treated: 112 . 3 ± 11 . 4 cells , n = 14; mean ± SEM; p>0 . 05 ) and newborn RG ( EdU+/BLBP+; 7-dpi LY411575-treated: 85 . 9 ± 9 . 6 cells , n = 10; 25-dpi LY411575-treated: 99 . 7 ± 10 . 7 cells , n = 14; mean ± SEM; p>0 . 05 ) showed no significant difference between 7-dpi and 25-dpi LY411575-treated fish ( Figure 7—figure supplement 1L and M ) . The remaining newborn neurons could survive until ~86 dpi ( Figure 7—figure supplement 1O–P ) . 76% of long-lived neurons resided in the tectum opticum ( TeO ) in 25-dpi LY411575-treated fish ( 25 dpi DMSO-treated: 83%; Figure 7—figure supplement 1Q–S ) . We measured the wound area ( 1669 ± 704 . 3 μm2 , n = 3 , mean ± SEM ) and found that the post-injury Notch inhibition did not help to complete restoration of the injured optic tectum ( Figure 7—figure supplement 1T and U ) .
How do reactive RG enter the cell cycle ? Consistent with astrogliosis in injured mammalian CNS , we find that zebrafish tectal RG are undergoing the consecutive phases of glial reactivation and glial proliferation in response to the injury ( Figure 2A–K ) ( Burda and Sofroniew , 2014 ) . After stab injury , almost all RG underneath the injury site respond to the injury by glial reactivation , characterized with the up-regulation of vimentin expression and robust GFP expression in a powerful transgenic line Tg ( 1016tuba1α:GFP ) , which has been used to monitor the injury-induced MG in zebrafish retina ( Fausett and Goldman , 2006; Wan et al . , 2012 ) . After the reactivation , ~25% of RG are entering the cell cycle , and become proliferative ( Figure 2H , L and Figure 2—figure supplement 1I-J2 ) . More interestingly , we find that the proliferative RG that respond to two sequential injuries are distinct but overlapping , which can be well explained by stochastic cell-cycle entry . It is , unexpectedly , that tectal RG underneath the injury sites randomly enter the cell cycle . How can reactive RG achieve this ? Our results show that proliferative RG have high deltaA expression while non-proliferating RG have high Notch activity , pointing to the working of Notch/Delta lateral inhibition , reminiscent of the function of Notch/Delta lateral inhibition in stochastically determining the fate of neuroblasts in Drosophila ( Sato et al . , 2016 ) . How is this Notch/Delta lateral inhibition generated after the injury ? There are at least two possibilities: 1 . The injury induces deltaA expression in some RG , leading to a decrease in Notch activity and subsequent cell-cycle entry . As a result of the lateral inhibition , an increased Notch activity keeps neighboring reactive RG in the non-proliferative state . 2 . The injury directly blocks Notch activity in some RG that then enter the cell cycle , and leave neighboring cells non-proliferative . It is certainly essential to figure out which model is actually at work . Can the optic tectum of adult zebrafish regenerate after the injury ? Recent studies come to the contradictory answers . Ohshima’s lab showed that tectal RG were capable of producing some neurons after the stab injury ( ~25% of newborn cells ) ( Shimizu et al . , 2018 ) , whereas Kaslin’s lab reported the opposite result , in which tectal RG only produce glial cells but not neurons ( Lindsey et al . , 2019 ) . Using EdU pulse-and-stain assay and Cre-loxP-based clonal analysis of tectal RG , we unambiguously demonstrate at the single-cell resolution that RG give rise to newborn glial cells ( ~91% of total newborn cells ) with only a few neurons ( ~3–5% of total newborn cells ) after the injury ( Figure 7Q ) , which is mostly consistent with the study of Kaslin’s lab . In fact , the average number of newborn neurons in the study from Ohshima’s lab ( ~4 BrdU+/HuC/D+ neurons ) was mostly similar to what we had ( ~4 EdU+/HuC/D+ neurons ) , while the number of total newborn cells in their study ( ~17 BrdU+ cells ) is much smaller than ours ( ~90 EdU+ cells ) . Thus , the conclusion of a high proportion of newborn neurons from Ohshima’s lab is perhaps due to the underestimated number of total newborn cells . Our scRNA-seq data shows that hmgb2a and hmgb2b are highly expressed in RG of proliferative state ( Figure 6—figure supplement 1J and K ) . hmgb2 is strongly associated with dormancy/activation transition of adult neural stem cells ( NSCs ) in mice ( Kimura et al . , 2018 ) , and loss of hmgb2 compromises gliogenesis and promotes neurogenesis ( Abraham et al . , 2013 ) . In situ hybridization also confirm the expression of hmgb2a in the proliferating tectal RG as the result of stab injury ( Figure 6—figure supplement 1L-M2 ) , suggesting proliferative tectal RG are likely to be gliogenic . Thus , unlike to RG of other CNS regions in zebrafish ( Dias et al . , 2012; Goldman , 2014; Kizil et al . , 2012; Kroehne et al . , 2011; Than-Trong and Bally-Cuif , 2015 ) , tectal RG resembles mammalian astrocytes in terms of injury response , undergoing the gliogenesis rather than the neurogenesis . Moreover , we also show that newborn glial cells can survive up to ~300 dpi and form a bulge structure of glial cells lining the tectal ventricle , while the wounds at the injury sites remain , which suggests that newborn glial cells are unable to migrate to the injury sites and the wounds are not restored . Notably , we observed the hypertrophic process of RG surrounding the injury sites forms a scar-like structure , which may block the restoration of the wounds ( Figure 6—figure supplement 1B-B3 and F-G3 ) . What controls fate outputs of proliferative RG ? Notch inhibition has been implicated in the production of neurons during embryonic development as well as in various injury contexts ( Dias et al . , 2012; Louvi and Artavanis-Tsakonas , 2006; Wan et al . , 2012 ) . We thus investigated whether Notch inhibition at the proper timing after the injury can switch gliogenesis to neurogenesis . Indeed , we find that post-injury Notch inhibition during 4–5 dpi but not 1–3 dpi results in a significant increase in newborn neurons ( Figure 7A–G ) . Our further analysis shows that Notch inhibition ( 4–5 dpi ) of injury-induced reactive RG labeled during 4–6 dpi results in a significant increase of neuron production ( Figure 7I , L , M , O and P ) . No significant change of neuron production from injury-induced proliferative RG labeled during 1–3 dpi is observed ( Figure 7H , J , K , O and P ) . Ueda et al . recently showed that Notch inhibition during 4–7 dpi decreases the number of newborn neurons ( Ueda et al . , 2018 ) , which is in contrast to our findings . This inconsistency is likely to be because the time window of EdU or BrdU treatment is different . Ueda and colleagues treated the injured fish with BrdU during 2–3 dpi . Instead , we treated the injured fish with EdU during 4–6 dpi in this experiment . This means the RG analyzed from two studies are different . In Ueda’s study , BrdU treatment during 2–3 dpi is most likely to label proliferative RG as a result of the cell-cycle entry of dormant RG by stab injury , whereas EdU treatment during 4–6 dpi in our analysis is most likely to label proliferative RG as a result of the cell-cycle entry of injury-induced reactive RG by Notch inhibition instead . Our findings raised an interesting hypothesis , that is , cell states of tectal RG , such as dormant state , reactive state , or proliferative state , influence their fate outcomes in the context of Notch inhibition or injury ( Figure 7Q ) . Resolving the mechanistic link of cell state and fate potentials will undoubtedly deepen the understanding of the fate control of injury-reactivated RG . Besides , long-term experiments unexpectedly showed that over-produced neurons disappeared by 25 dpi . Earlier studies also showed that manipulation of Notch signaling could drive more RG into the cell cycle in the injury context ( Dias et al . , 2012; Ueda et al . , 2018; Wan et al . , 2012 ) . However , none of them investigated the long-term survival of those newborn cells . Our observation of rapid loss of newborn cells might be due to their intrinsic property of short life or due to the lack of significant neuron loss in stab injured optic tectum . The underlying mechanisms that control the death of over-produced neurons are also appealing to be further investigated .
Zebrafish embryos , larvae , and adults were produced , grown , and maintained at 28°C according to standard protocols except during heat shock treatments . Embryos were harvested and kept in the embryo medium ( 0 . 294 g/L NaCl , 0 . 0127 g/L KCl , 0 . 0485 g/L CaCl2·2H2O , 0 . 0813 g/L MgSO4·7H2O , 0 . 3 g/L sea salt , and 2 × 10−4 g/L methylene blue ) at 28°C . Young adult zebrafish ranging in age from 2 to 4 months old were used for experiments . Approximately equal sex ratios were used for experiments . All young adult fish were fed twice daily . Published lines used in this study include: Wild type , Tg ( gfap:EGFP ) mi2001 ( ZDB-FISH-150901–29307 ) ( Bernardos and Raymond , 2006 ) , Tg ( her4 . 1:dRFP ) ( ZDB-TGCONSTRCT-070612–2 ) ( Yeo et al . , 2007 ) , Tg ( 1016tuba1α:GFP ) ( ZDB-GENO-070321–4 ) ( Fausett and Goldman , 2006 ) , Tg ( olig2:GFP ) ( ZDB-ALT-041129–8 ) ( Shin et al . , 2003 ) , Tg ( mpeg1:GFP ) ZDB-TGCONSTRCT-170801–5 ( Ellett et al . , 2011 ) , Tg ( hsp70l:gal4 ) ( ZDB-TGCONSTRCT-070117–42 ) ( Scheer et al . , 2001 ) , Tg ( UAS:NICD-Myc ) ( ZDB-TGCONSTRCT-070117–24 ) ( Scheer et al . , 2001 ) . Details of the generation of the new lines generated in this study are described below . At least three independent founders of each line were screened and checked to confirm the described expression patterns . The plasmid of pTol2-tp1bglob:EGFP ( Addgene plasmid # 73586 ) was a gift from Dr . Nathan Lawson ( UMass Medical School , Worcester , USA ) ( Quillien et al . , 2014 ) . The plasmid was co-injected with Tol2 mRNA at one-cell stage . Zebrafish embryos were grown and maintained according to standard protocols . In this line , cells with a high Notch signaling level express the fluorescent protein EGFP . The full name of this line is Tg ( Tp1bglob:EGFP ) . The plasmid was a gift from Dr . Michael Brand ( Technische Universität Dresden , Dresden , Germany ) ( Kroehne et al . , 2011 ) . The plasmid was co-injected with Tol2 mRNA at one-cell stage . Zebrafish embryos were grown and maintained according to standard protocols . In this line , radial glia express the fluorescent protein mCherry followed by a CreERT2 element . The full name of this line is Tg ( her4 . 1:mCherryT2ACreERT2 ) . The plasmid was a gift from Dr . Michael Brand ( Technische Universität Dresden , Dresden , Germany ) ( Kroehne et al . , 2011 ) . The plasmid was co-injected with Tol2 mRNA at one-cell stage . Zebrafish embryos were grown and maintained according to standard protocols . In this line , cells express DsRed2 only after heat shock . Upon CreERT2 mediated recombination , the DsRed2-floxed cassette is eliminated and cells exposed to heat shock then express the fluorescent protein EGFP . The full name of this line is Tg ( hsp70l:DsRed2 ( floxed ) EGFP ) . Fish were anesthetized using 0 . 02% MS-222 for 30 to 45 s ( s ) . Fish were placed in a piece of 0 . 02% MS-222 soaked tissue , and a set of tweezers was used to place them properly , allowing accessibility to the head . With the visual aid of a dissecting microscope , the needle ( 30 gauge , outer diameter 300 μm ) was stabbed ~400 μm deep into the optic tectum through the skull . After the injury , fish were returned back to the fish tank . The single-cell suspension of adult zebrafish optic tecta was prepared by following a published protocol ( Lopez-Ramirez et al . , 2016 ) . Large-area injuries were introduced to the central-dorsal part of the optic tecta of ~2 months old Tg ( gfap:EGFP ) fish and the fish were returned back to a fish tank under standard conditions . At three dpi , the fish were anesthetized and sacrificed . The optic tecta were dissected and dissociated by the digestion in 350 µl papain solution at 37°C for 15 min ( mins ) . During digestion , the tissues were pipetted up and down 4 × 10 times . Digestion was stopped by 1400 µl washing buffer . The cell solution was filtered with a 40 µm cell strainer ( BD Falcon ) , and then centrifuged at 200 g for 5 mins at 4°C . The supernatant was discarded and the pellets were resuspended with 1 × PBS with 0 . 04% BSA . Then fluorescence-activated cell sorting ( FACS ) was performed , collecting the cells with high EGFP fluorescence into 1 × PBS with 0 . 04% BSA . Papain solution: To prepare the papain solution , add 100 µl papain ( Worthington , LS003126 ) , 100 µl DNase ( 1% , Sigma , DN25 ) and 200 µl L-cysteine ( 12 mg/ml , Sigma , C6852 ) into 5 ml DMEM/F12 ( Invitrogen , 11330032 ) . Washing solution: To prepare the washing solution , add 65 µl glucose 45% ( Invitrogen , 04196545 SB ) , 50 µl HEPES 1M ( Sigma , H4034 ) and 500 µl FBS ( Gibco , 10270106 ) in 9 . 385 ml DPBS 1× ( Invitrogen , 14190–144 ) . All solutions were filtered through a 0 . 22 µm pore size filter ( Millipore ) to sterilize and stored at 4°C before use . To perform single-cell RNA sequencing ( scRNA-seq ) , cells after FACS were loaded onto the Chromium Single Cell Chip ( 10x Genomics , USA ) according to the manufacturer’s protocol . The scRNA-seq libraries were generated using the GemCode Single-Cell Instrument and Single Cell 3' Library and Gel Bead kit v2 Chip kit ( 10x Genomics , 120237 ) by following the manufacturer’s protocol . Library quantification and quality assessments were performed by Qubit fluorometric assay ( Invitrogen ) with dsDNA High Sensitivity Assay Kit ( AATI , DNF-474–0500 ) and the fragment analyzer with High Sensitivity Large Fragment −50 kb Analysis Kit ( AATI , DNF-464 ) . The indexed library was tested for quality , and sequenced by the Illumina NovaSeq 6000 sequencer with the S2 flow cell using paired-end 150 × 150 base pair as the sequencing mode . The sequencing depth was 60K reads per cell . Single-cell FASTQ sequencing reads ( Novogene ) were processed , and converted to digital gene expression matrices after mapping to the zebrafish genome ( Zv10 ) using the Cell Ranger Single Cell Software Suite ( v2 . 1 . 0 ) provided on 10x genomics website ( https://support . 10xgenomics . com/single-cell-gene-expression/software/pipelines/ latest/what-is-cell-ranger ) . 66 , 817 mean reads per cell and 1325 mean genes per cell were obtained . For further analysis , we used an analysis pipeline provided by Seurat R package ( http://satijalab . org/seurat/ ) . Firstly , the Seurat object was created to filter low-abundance genes , cell doublets and low-quality libraries ( with low gene numbers and high mitochondrial transcripts ) . Secondly , the filtered data were normalized and used to identify highly variable genes based on expression and dispersion . Thirdly , the data were scaled , and the unwanted sources of variation were removed . Fourthly , cell clustering analyses were performed by the t-SNE projection ( Figure 3—figure supplement 1B ) . Finally , we found out the markers for every cluster ( Figure 3—figure supplement 1C ) . Due to possible contamination during tissue dissociation and FACS , the samples were contaminated with other types of cells from the optic tecta and other neighboring tissues . Based on the markers of each cluster , these contaminated cells were identified and removed after the initial clustering . Non-glial cell clusters ( 1 , 2 , 5 , 6 , 11 , 12 , 13 , 14 ) were identified by high expression of neuronal markers such as neurod1 , elavl3 , gad1b and slc17a6b and low expression of glial markers , dormant and proliferative cell markers such as fabp7a , gfap , her4 . 1 , mfge8a and pcna ( Figure 3—figure supplement 1D and E ) . These clusters were removed and the remaining cells were used for further analysis . As many proliferative progenitors are present in the tectal proliferation zone ( TPZ ) ( Galant et al . , 2016; Ito et al . , 2010 ) , a big-area injury induced a lot of PCNA+ tectal RG at 3 dpi . We obtained two pcna+ clusters in the t-SNE plot ( cluster 1 and 2 , Figure 3—figure supplement 2A and B ) . However , based on experimental evidence: 1 . Injury caused the obvious down-regulation of her4 PCNA+ RG , whereas her4 was highly expressed in RG in TPZ ( Figure 3—figure supplement 2B–F ) ; 2 . The previous study showed progenitors in TPZ were able to generate oligodendrocytes . We did not find any new-born cell derived from injury-induced PCNA+ RG was oligodendrocyte , and olig2 expression was noticed in cluster 1 and 10 but not in cluster 2 ( Figure 3—figure supplement 2B and C ) . We identified cluster 1 as the progenitors in the TPZ , and it was removed from our data . Following these step-wise filtering processes , we obtained the purified data of each sample . To obtain the cell-cycle properties of the cells in our sample , the ‘CellCycleScoring’ function of Seurat was used . Briefly , each cell was scored based on its expression of G2/M and S phase marker genes . Then the numbers of cells in different cell cycle phases were counted and the ratios of individual cell cycle phases were calculated . The gene-gene correlation was measured according to the pairwise Pearson correlational distances . ‘bioDist’ R package was used to calculate these correlational distances . After the t-SNE cluster analysis of the single-cell data , trajectory analysis was performed to investigate the pseudo-time of four identified states by using ‘monocle’ and ‘Slingshot’ R package . LY411575 ( final concentration of 10 μM , Selleck Chemical , S2714 ) , RO4929097 ( final concentration of 50 μM , Selleck Chemicals , S1757 ) or DMSO ( Dimethyl sulfoxide , final concentration of 10 μM , Sigma , B8418 ) was applied freshly to the fish water at 28°C in the dark for desired days ( 18 hr per day ) . The LY411575 , RO4929097 , or DMSO solutions were changed twice a day . Double transgenic fish Tg ( hsp70l:gal4 ×UAS:NICD-Myc ) or wild-type fish were heat-shocked in a warm water bath at 38°C for 1 hr on three consecutive days and retrieved to their tank at 28°C . As the mosaic expression of the transgene , the same sections contained NICD-Myc-overexpressing and control cells . Myc expression was detected after heat shock . To induce CreERT2 mediated recombination , tamoxifen ( TAM , final concentration of 2 . 5–5 μM , Sigma , T5648 ) was applied to the fish water at 28°C in the dark for three days ( 12 hr per day ) . Double transgenic fish Tg ( her4 . 1:mCherryT2ACreERT2::hsp70l:DsRed2 ( floxed ) EGFP ) were heat-shocked at 38°C for 1 hr once daily on three consecutive days before sacrifice . 6 hr after the last heat shock , fish were sacrificed for analysis . Bromodeoxyuridine ( BrdU , final concentration of 10 mM , Sigma , B5002 ) was applied freshly to the fish water at 28°C in the dark for desired days ( 12 hr per day ) . Zebrafish were anesthetized , placed on a wet tissue , and injected intraperitoneally with ~5 μl 5 mM 5-ethynyl-2′-deoxyuridine ( EdU ) in 0 . 1 M sterile PBS . After injection , fish were retrieved to a fish tank and used for further experiments . To detect the EdU signal , the EdU Click-iT reaction solution ( Invitrogen , C10340 ) was prepared freshly according to the manufacturer’s protocol . Sections on slides were covered with a solution and incubated in a humid chamber at room temperature in the dark for 1 hr . After three 10 mins wash in PBS , sections were used for imaging or subsequent processing for immunohistochemistry . Brains were fixed in 4% paraformaldehyde ( PFA , Electron Microscopy Services , USA , 157–8 ) overnight , cryoprotected in 30% sucrose for 6 hr , flash-frozen and cryosectioned at a thickness of 12 μm . The fluorescent immunochemistry was performed on brain sections as described ( Tang et al . , 2017 ) . Sections were washed with 1 × PBS for 10 mins for three times and permeabilized in 1 × PBS with 0 . 5% Triton X-100 for 30 mins . After blocking with 5% BSA solution ( Sigma ) at RT for 1 hr , sections were incubated with the primary antibody at 4°C overnight . Sections were then washed with 1 × PBS and incubated with Alexa Fluor 488- , 594- , or 647-conjugated secondary antibody ( 1:1 , 000; Jackson Immuno Research Laboratories Inc ) at room temperature for 2 hr . 4' , 6-diamidino-2-phenylindole ( DAPI ) staining was performed according to the standard protocol . Slides were finally mounted using the fluorescent mounting medium ( Sigma ) . For PCNA , HuC/D staining , sections were pre-treated with Improved Citrate Antigen Retrieval Solution ( Beyotime Biotechnology , P0090 ) for 5 mins and washed by Washing Buffer ( Beyotime Biotechnology , P0106C ) for twice and 1 × PBS for once before blocking or in the citrate acid buffer ( 10 mM , 0 . 05% Tween 20 , pH 6 . 0 ) at 95°C for 30 mins . For BrdU staining , sections were treated with 2 N HCl at 37°C for 10 mins followed by neutralization with 0 . 1 M sodium borate solution at room temperature for 10 mins and washed by 1 × PBS for 10 mins for three times . Primary antibodies used in this study are listed in the Key resources table . The digoxigenin ( DIG ) -labeled her4 , mfge8a , dla , klf6a , insm1a and hmgb2a antisense probes were prepared by using the MEGAscriptTM T7 High Yield Transcription Kit ( Invitrogen , AM1334 ) and DIG RNA labeling kit ( Roche , 11277073910 ) . The cDNA of each gene was amplified by PCR using the following primers: her4 . 1-F:5'-CCCTCGAGCTGATCCTGACGGAGAACTGAACAC-3'; her4 . 1-R:5'-TAATACGACTCACTATAGTTCTAGAATAGACGAAGAGAAA ACAAACC-3'; mfge8a-F: 5'-TGCAGCCCAAACCCATGTAA-3'; mfge8a-R:5'-TAATACGACTCACTATAGGGTGAGTCGGGATTTCATGCCC-3'; klf6a-F:5'-ATGGATGTTCTACCAATGTGCAGCA-3'; klf6a-R:5'-TAATACGACTCACTATAGGGTCAGAGGTGCCTCTTCATGTGC-3'; insm1a-F:5'-ATGCCCAGAGGATTTTTAGTCAAGC-3'; insm1a-R:5'-TAATACGACTCACTATAGGGTTGTCTTCAGCAGGCTGGAC GC-3’; hmgb2a F:5'-ATGGGTAAAGATCCAAATAAGCCCAG-3'; hmgb2a-R:5'-TAATACGACTCACTATAGGGTTATTCGTCATCATCATCCTCGTCCTC-3' . The injured and uninjured zebrafish brains were fixed in 4% PFA at 4°C overnight followed by dehydration in 30% sucrose and then were cryosectioned at a thickness of 12 μm . The slices were post-fixed in 4% PFA at room temperature for 15 mins and washed with 1 × PBS at room temperature for 3 mins . To block the activity of endogenous peroxidase , all slides were treated with 0 . 1% H2O2 at room temperature for 30 mins . After being washed twice with 1 × PBS at room temperature for 3 mins , slides were treated with 10 μg/ml proteinase K ( Sigma ) diluted in TE ( 10 mM Tris-HCl , pH 8 . 0 , and 1 mM EDTA , pH 8 . 0 ) at 37°C for 8 mins , then treated with 4% PFA at room temperature for 10 mins . Subsequently , all slides were washed with 1 × PBS at RT for 3 mins , followed by the incubation in 0 . 2 M HCl at RT for 10 mins . After being washed with 1 × PBS for 5 mins , all slices were then incubated with 0 . 1 M triethanol amine-HCl ( 662 . 5 μl triethanolamine and 1 . 35 ml 1 M HCl; adding water to the final volume of 50 ml , pH 8 . 0 ) at room temperature for 1 min and in 0 . 1 M triethanol amine-HCl containing 0 . 25% acetic anhydrate at room temperature for 10 mins with gentle shaking . Slides were then washed by 1 × PBS at room temperature for 5 mins , then were dehydrated in a series of 60% , 80% , 95% , and twice in 100% ethanol at room temperature for 90 s , respectively . Slides were incubated in the hybridization buffer ( 50% formamide ( Sigma ) , 10 mM Tris-HCl , pH 8 . 0 , 200 μg/ml yeast tRNA ( Invitrogen ) , 1 × Denhart buffer , SDS , EDTA and 10% dextran sulfate ( Ambion ) containing 1 μg/ml probes at 60°C overnight . On the second day , slides were washed sequentially in 5 × SSC at 65°C for 30 mins , 2 × SSC with 50% formamide at 65°C for 30 mins , TNE buffer ( 100 ml TNE consisting of 1 ml 1 M Tris-HCl , pH 7 . 5 , 10 ml 5 M NaCl , and 0 . 2 ml 0 . 5 M EDTA ) at 37°C for 10 min and then in TNE buffer with 20 μg/ml RNaseA at 37°C for 30 mins . Slides were then incubated with 2 × SSC at 60°C for 20 mins , 0 . 2 × SSC at 60°C for 20 mins , and 0 . 1 × SSC at RT for 20 mins . Next , slides were blocked by TN buffer at room temperature for 5 mins ( 200 ml TN buffer consisting of 20 ml 1 M Tris-HCl , pH 7 . 5 , 6 ml 5 M NaCl , and 174 ml water ) followed by TNB buffer ( TN buffer + 0 . 5% blocking reagent; Roche ) at room temperature for 5 mins . Finally , slides were incubated in TNB buffer with anti–DIG-POD ( 1:500; Roche ) at 4°C overnight . On the third day , the signal was detected by the TSATM Plus Cyanine 3/Fluorescein System ( PerkinElmer , NEL753001KT ) . Images were taken using an inverted confocal microscope system ( FV1200 , Olympus ) confocal microscope using 10 × ( air , 0 . 4 NA ) , 30 × ( silicon oil , 1 . 05 NA ) , or 60 × ( silicon oil , 1 . 3 NA ) objectives . All quantification and visualization were performed with FV10-ASW 4 . 0 Viewer ( Olympus ) , and Image J . Adobe Illustrator CS6 was used to process acquired 2D Figure 3D image stacks were analyzed using Imaris software ( Bitplane ) . 12-μm-thick sections ( around the injury sites , 8–14 sections per tectum ) between the anterior optic tectum and the posterior optic tectum were used for statistical analyses . For cell counting , cryosections of the injury sites were analyzed ( every second serial section ) . Microsoft Excel was used to process the measured data . To perform the statistical analysis , p values were calculated with GraphPad Prism ( or Microsoft Excel ) . The unpaired , non-parametric Wilcoxon test was applied for comparison of two groups . The one-way ANOVA , followed by Tukey’s HSD test was applied for comparison of different groups with one treatment . The two-way ANOVA followed by Tukey’s HSD test was applied for comparison of four groups with two treatments . Error bars represent SEM . ****p<0 . 0001 , ***p<0 . 001; **p<0 . 01; *p<0 . 05; ns , p>0 . 05 . | The brain contains networks of cells known as neurons that rapidly relay information from one place to another . Other brain cells called glial cells perform several roles to support and protect the neurons including holding them in position and supplying them with oxygen and other nutrients . Damage to the brain as a result of physical injuries is one of the leading causes of death and disability in people worldwide . Brain injuries generally stimulate glial cells to enter a “reactive” state to help repair the damage . However , some glial cells may start to divide and produce more glial cells instead , leading to scar-like structures in the brain that hinder the repair process . To investigate why brain injuries trigger some glial cells to divide , Yu and He systematically examined glial cells in the part of the zebrafish brain that handles vision , known as the optic tectum . The experiments showed that a physical injury stimulated some of the glial cells to divide . Repeated injuries to the same part of the brain did not always stimulate the same glial cells to divide , suggesting that this process happens in random cells . Further experiments revealed that molecules involved in a signaling pathway known as Notch signaling were released from some brain cells and inhibited neighboring glial cells from dividing to make new glial cells . Unexpectedly , inhibiting Notch signaling after a brain injury caused some of the glial cells that were in the reactive state to divide to produce neurons instead of glial cells . Understanding how the brain responds to injury may help researchers develop new therapies that may benefit human patients in future . The next steps following on from this work will be to find out whether glial cells in humans and other mammals work in the same way as glial cells in zebrafish . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"developmental",
"biology"
] | 2019 | Stochastic cell-cycle entry and cell-state-dependent fate outputs of injury-reactivated tectal radial glia in zebrafish |
The brain possesses a remarkable capacity to compensate for changes in inputs resulting from a range of sensory impairments . Developmental studies of sound localization have shown that adaptation to asymmetric hearing loss can be achieved either by reinterpreting altered spatial cues or by relying more on those cues that remain intact . Adaptation to monaural deprivation in adulthood is also possible , but appears to lack such flexibility . Here we show , however , that appropriate behavioral training enables monaurally-deprived adult humans to exploit both of these adaptive processes . Moreover , cortical recordings in ferrets reared with asymmetric hearing loss suggest that these forms of plasticity have distinct neural substrates . An ability to adapt to asymmetric hearing loss using multiple adaptive processes is therefore shared by different species and may persist throughout the lifespan . This highlights the fundamental flexibility of neural systems , and may also point toward novel therapeutic strategies for treating sensory disorders .
A major challenge faced by the brain is to maintain stable and accurate representations of the world despite changes in sensory input . This is important because the statistical structure of sensory experience varies across different environments ( Mlynarski and Jost , 2014; Qian et al . , 2012; Seydell et al . , 2010 ) , but also because long-term changes in sensory input result from a range of sensory impairments ( Feldman and Brecht , 2005; Keating and King , 2015; Sengpiel , 2014 ) . Adaptation to altered inputs has been demonstrated in different sensory systems , particularly during development , and serves to shape neural circuits to the specific inputs experienced by the individual ( Margolis et al . , 2014; Mendonca , 2014; Schreiner and Polley , 2014; Sur et al . , 2013 ) . However , many ecologically important aspects of neural processing require the integration of multiple sensory cues , either within or across different sensory modalities ( Seilheimer et al . , 2014; Seydell et al . , 2010 ) . A specific change in sensory input may therefore have a considerable impact on some cues whilst leaving others intact . In such cases , adaptation can be achieved in two distinct ways , as demonstrated by recent studies of sound localization following monaural deprivation during infancy ( Keating et al . , 2013; 2015 ) . Monaural deprivation alters the binaural spatial cues that normally determine the perceived location of a sound in the horizontal plane ( Figure 1A ) ( Kumpik et al . , 2010; Lupo et al . , 2011 ) . Adaptation can therefore be achieved by learning the altered relationships between particular cue values and spatial locations ( Gold and Knudsen , 2000; Keating et al . , 2015; Knudsen et al . , 1984 ) , a process referred to as cue remapping . However , at least in mammals , monaural spectral cues are also available to judge sound source location in both the horizontal and vertical planes ( Carlile et al . , 2005 ) . These spectral cues arise from the acoustic properties of the head and external ears , which filter sounds in a direction-dependent way ( Figure 1B ) . Monaural deprivation has no effect on the spectral cues available to the non-deprived ear . This means it is possible to adapt by learning to rely more on these unchanged spectral cues , whilst learning to ignore the altered binaural cues ( Kacelnik et al . , 2006; Keating et al . , 2013; Kumpik et al . , 2010 ) , a form of adaptation referred to as cue reweighting . 10 . 7554/eLife . 12264 . 003Figure 1 . Effect of training on localization of broadband noise stimuli in the horizontal plane by monaurally deprived human listeners . ( A ) When one of these sounds is presented on one side of the head , it will be louder and arrive earlier at the ipsilateral ear ( blue ) , producing interaural time and level differences , which are respectively the primary cues to sound location at low and high frequencies . ( B ) Because of acoustic filtering by the head and ears , the spectrum of a sound at the tympanic membrane ( post-filtering , color ) differs from that of the original sound ( pre-filtering , black ) and varies with location ( amplitude in dB is plotted as a function of frequency; color indicates different locations ) . These spectral cues make it possible to localize sounds using a single ear , but only for sounds that have relatively flat spectra ( solid lines ) and are sufficiently broadband ( shape of spectra in narrow frequency bands varies little with location – see shaded gray region ) . When spectral features are artificially added to the pre-filtered sound source ( dotted lines ) , these added features can be misattributed to the filtering effects of the head and ears . This produces sound localization errors ( e . g . dotted green spectrum is more easily confused with solid turquoise spectrum because of additional peak at high frequencies ) . The extent of these errors allows us to infer subjects’ reliance on spectral cues . ( C , D ) Joint distribution of stimulus and response obtained from the first ( C ) and last ( D ) training session for an individual subject with an earplug in the right ear . Grayscale indicates the number of trials corresponding to each stimulus-response combination . Data are shown for trials on which flat-spectrum stimuli were used ( i . e . all spatial cues were available ) . ( E ) Sound localization performance ( % correct ) as a function of training session for the same subject . Scores for each session ( dots ) were fitted using linear regression ( lines ) to calculate slope values , which quantified the rate of adaptation . Relative to flat-spectrum stimuli ( blue ) , much less adaptation occurred with random-spectrum stimuli ( pink ) , which limit the usefulness of spectral cues to sound location ( Figure 1—figure supplement 1 ) . ( F ) Adaptation rate is shown for flat- and random-spectrum stimuli for each subject ( gray lines; n = 11 ) . Positive values indicate improvements in localization performance with training . Mean adaptation rates across subjects ( ± bootstrapped 95% confidence intervals ) are shown in blue and pink . Similar results are observed if front-back errors are excluded and changes in error magnitude are calculated ( Figure 1—figure supplement 2 ) . Dotted black lines indicate adaptation rates observed previously in humans ( Kumpik et al . , 2010; total adaptation reported divided by number of sessions , n = 8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 00310 . 7554/eLife . 12264 . 004Figure 1—figure supplement 1 . Experimental setup and stimuli . ( A ) Schematic illustrating the circular loudspeaker array used for sound localization training . Subjects sat at the centre of this array , facing in the direction indicated by the arrow . ( B ) Spectral profile for a random-spectrum stimulus ( black ) . Spectra were filtered to eliminate abrupt spectral transitions to which the auditory system is insensitive ( see Materials and methods ) . The overall amount of spectral randomization was also fixed on each trial ( SD = 10 dB ) . Although the spectrum varied considerably across trials ( many different examples are shown in gray ) , the mean spectrum was relatively flat ( red ) . ( C ) Our randomization procedure allowed us to set the amount of randomization and overall level of each stimulus , but these parameters could still vary within individual frequency bands . We can measure this for a single stimulus by dividing its spectrum ( gray ) into one-octave bands and calculating the mean ± SD amplitude values for each band ( black ) . This indicates that the level ( mean ) and amount of randomization ( SD ) of each frequency band fluctuates on each trial . ( D , E ) To determine whether these differences have an impact on sound localization , we compared the random-spectrum stimuli presented on correct ( pink ) and incorrect ( blue ) trials . Relative to mislocalized stimuli , we found that the amount of spectral randomization was smaller for correctly localized stimuli , but only at higher frequencies ( D , indicated by asterisk; significant interaction between frequency and correctness of response , P < 0 . 001 , ANOVA; P < 0 . 05 , post hoc test ) . In other words , the spectra of correctly-localized sounds tended to be relatively flat at high frequencies . No differences in sound level were observed between stimuli on correct and incorrect trials ( E ) . Data show mean ± SEM . ( F ) To understand the implications of this for sound localization , we subdivided trials into groups ( deciles ) based on the amount of spectral randomization at high frequencies ( 1st decile represents 10% of trials with the smallest amount of spectral randomization ) and quantified sound localization accuracy ( % correct ) for each group . This indicates that some of our random-spectrum stimuli were more difficult to localize than others , with performance declining as spectral randomization increased at high frequencies . It also supports the view that increasing amounts of spectral randomization progressively degrade the usefulness of spectral cues , which are most prominent at high frequencies . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 00410 . 7554/eLife . 12264 . 005Figure 1—figure supplement 2 . Effect of training on localization by human listeners of broadband stimuli using same analysis method as for narrowband stimuli in Figure 2 . ( A–D ) Joint distributions of stimulus and response obtained from the first ( A , C ) and last ( B , D ) training sessions for flat- ( A , B ) and random-spectrum ( C , D ) stimuli . Data are shown for an individual subject wearing an earplug in the left ear , with grayscale indicating the number of trials corresponding to each stimulus-response combination . Stimulus- and response-locations in the front and rear hemifields have been collapsed to provide a measure of sound localization that is insensitive to front-back errors . ( E ) Mean error magnitude plotted as a function of training session for the same subject shown in A-D . Data are plotted separately for flat- ( turquoise ) and random-spectrum ( red ) stimuli . Scores for each session ( dots ) were fitted using linear regression ( lines ) to calculate slope values , which quantified the change in error magnitude ( Δ error ) with training . Improved performance was associated with a reduction in error magnitude , producing negative values for Δ error . ( F ) Δ error for flat- and random-spectrum stimuli plotted for each subject ( gray lines; n = 11 ) . Mean values for Δ error across subjects ( ± bootstrapped 95% confidence intervals ) are shown in color . Adaptation occurs for both flat- and random-spectrum stimuli ( Δ error values are significantly <0; p<0 . 01 , bootstrap test ) , but the extent of adaptation is greater for flat-spectrum stimuli ( p<0 . 01 , bootstrap test on the within-subject differences in Δ error ) . ( G ) Bias in sound localization responses plotted as a function of training session for the subject in E . Positive values indicate that responses were biased toward the side of the open ear . Data are plotted separately for flat- ( turquoise ) and random-spectrum ( red ) stimuli . Scores for each session ( dots ) were fitted using linear regression ( lines ) to calculate slope values , which quantified the change in response bias ( Δ bias ) with training . Negative values of Δ bias indicate an adaptive shift in response bias toward the side of the plugged ear . ( H ) Δ bias for flat- and random-spectrum stimuli plotted for each subject ( gray lines; n = 11 ) . Mean values for Δ bias across subjects ( ± bootstrapped 95% confidence intervals ) are shown in color . No changes in bias were observed for either stimulus type ( Δ bias values do not deviate significantly from 0;p>0 . 05 , bootstrap test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 005 Developmental studies of sound localization plasticity following monaural deprivation have found evidence for both cue remapping ( Gold and Knudsen , 2000; Keating et al . , 2015; Knudsen et al . , 1984 ) and cue reweighting ( Keating et al . , 2013 ) , but it is not known whether these adaptive processes can occur simultaneously . Indeed , until recently , it was thought that monaural deprivation might induce different adaptive processes in different species ( Keating and King , 2013; Shamma , 2015 ) . However , whilst we now know that ferrets use both cue remapping and reweighting to adapt to monaural deprivation experienced during development ( Keating et al . , 2013; 2015 ) , it is not known whether the same neural populations are involved in each case . It is also not known whether the ability to use both adaptive processes is restricted to specific species or developmental epochs . Although the mature auditory system can adapt to monaural deprivation using cue reweighting ( Kumpik et al . , 2010 ) , conflicting evidence for cue remapping has been obtained in adult humans fitted with an earplug in one ear for several days ( Florentine , 1976; McPartland et al . , 1997 ) . To the extent that adaptive changes in binaural cue sensitivity are possible in adulthood , as suggested by other sensory manipulations ( Trapeau and Schonwiesner , 2015 ) , these may occur at the expense of cue reweighting . It is therefore unclear whether the same adult individuals can adapt to a unilateral hearing loss using multiple adaptive processes . Although numerous studies have shown that spatial hearing is more plastic early in life ( Keating and King , 2013; Knudsen et al . , 1984; Popescu and Polley , 2010 ) , behavioral training can facilitate accommodation to altered cues in adulthood ( Carlile , 2014; Carlile et al . , 2014; Kacelnik et al . , 2006; Shinn-Cunningham et al . , 1998 ) . Here , we show that adult humans are equally capable of using both adaptive processes , provided they are given appropriate training . Moreover , our results suggest that cue remapping and reweighting are neurophysiologically distinct , which we confirmed by recording from auditory cortical neurons in ferrets reared with an intermittent hearing loss in one ear .
Adult humans were trained to localize sounds from 12 loudspeakers in the horizontal plane ( Figure 1—figure supplement 1A ) whilst wearing an earplug in one ear ( ~5600 trials split into 7 sessions completed in < 3 weeks ) . In order to directly measure the efficacy of training , earplugs were worn only during training sessions . This contrasts with previous work in which adult humans received minimal training , but were required to wear earplugs for extended periods of everyday life ( Florentine , 1976; Kumpik et al . , 2010; McPartland et al . , 1997 ) . On ~50% of trials , subjects were required to localize flat-spectrum broadband noise ( 0 . 5–20 kHz ) , which provide all of the available auditory spatial cues ( Blauert , 1997 ) . With these cue-rich stimuli , trials were repeated following incorrect responses ( “correction trials” ) and subjects were given performance feedback . Across training sessions , sound localization performance ( % correct ) gradually improved ( Figure 1C–F; slope values >0; bootstrap test , p<0 . 01; Cohen’s d = 1 . 43 ) , indicating that relatively short periods of training are sufficient to drive adaptation . To determine the relative contributions of cue remapping and reweighting to these changes in localization accuracy , we measured the extent of adaptation for two additional stimulus types that restrict the availability of specific cues . For these cue-restricted stimuli , which were randomly interleaved with cue-rich stimuli , correction trials were not used and no feedback was given . The first of these additional stimulus types comprised broadband noise with a random spectral profile that varied across trials ( Figure 1—figure supplement 1B ) . These stimuli disrupt spectral localization cues because it is unclear whether specific spectral features are produced by the filtering effects of the head and ears or are instead properties of the sound itself ( Figure 1B ) ( Keating et al . , 2013 ) . Consequently , if subjects adapt to asymmetric hearing loss by giving greater weight to the spectral cues provided by the non-deprived ear , we would expect to see less improvement in sound localization performance for random-spectrum sounds than for flat-spectrum sounds . This is precisely what we found ( Figure 1E , F; random-spectrum slope values < flat-spectrum slope values; bootstrap test , p<0 . 01; Cohen’s d = 1 . 18; see also Figure 1—figure supplement 2 ) , indicating that adaptation involves learning to rely more on spectral cues . However , if adaptation were solely dependent on this type of cue reweighting , we would expect no improvement in sound localization for narrowband sounds , such as pure tones . This is because spectral cues require a comparison of sound energy at different frequencies , which is not possible for these sounds ( Figure 1B ) ( Carlile et al . , 2005 ) . Improved localization of pure tones would therefore indicate adaptive processing of binaural cues . Because interaural time differences ( ITDs ) and interaural level differences ( ILDs ) are respectively the primary cues for localizing low- ( <1 . 5 kHz ) and high-frequency ( ≥1 . 5 kHz ) tones ( Blauert , 1997 ) , we tested each of these stimuli separately . To detect changes in binaural sensitivity , and facilitate comparison with previous work ( Keating et al . , 2015; Kumpik et al . , 2010 ) , stimulus and response locations in the front and rear hemifields were collapsed . This produces a measure of performance that is insensitive to front-back errors , which reflect failures in spectral , rather than binaural , processing . We observed improvements in subjects’ ability to localize both low- and high-frequency pure tones over time , demonstrated by a decline in error magnitude ( Figure 2E , F; Δ error <0; bootstrap test , p<0 . 01 ) . The initial bias toward the side of the open ear was also reduced ( Figure 2G , H; Δ bias <0; bootstrap test , p<0 . 01; low-frequency , Cohen’s d = 0 . 7; high-frequency , Cohen’s d = 0 . 96 ) . Adaptation therefore involves a shift in the mapping of altered binaural cues onto spatial location . Together , these results show that subjects adapted to monaural deprivation using a combination of both cue remapping and cue reweighting . 10 . 7554/eLife . 12264 . 006Figure 2 . Effect of training on localization of pure tone stimuli in the horizontal plane by monaurally deprived human listeners . ( A–D ) Joint distributions of stimulus and response obtained from the first ( A , C ) and last ( B , D ) training sessions for low- ( A , B ) and high-frequency ( C , D ) tones . Data are shown for an individual subject wearing an earplug in the left ear , with grayscale indicating the number of trials corresponding to each stimulus-response combination . Because pure tones can be accurately localized only by using binaural spatial cues , which are susceptible to front-back errors , data from the front and rear hemifields have been collapsed . ( E ) Mean error magnitude plotted as a function of training session for the same subject shown in A–D . Data are plotted separately for low- ( 1 kHz , dark blue ) and high-frequency ( 8 kHz , light blue ) tones . Scores for each session ( dots ) were fitted using linear regression ( lines ) to calculate slope values , which quantified the change in error magnitude ( Δ error ) with training . Improved performance was associated with a reduction in error magnitude , producing negative values for Δ error . ( F ) Δ error for low- and high-frequency tones plotted for each subject ( gray lines; n = 11 ) . Mean values for Δ error across subjects ( ± bootstrapped 95% confidence intervals ) are shown in blue . Although there are pronounced individual differences for the adaptation observed at the two tone frequencies , almost all values are <0 , indicating that error magnitude declined over the training sessions . Dotted red line shows Δ error values that would have been observed if subjects had adapted as well as ferrets reared with a unilateral earplug ( Keating et al . 2015; total Δ error reported for ferrets was divided by the number of training sessions used in the present study , n = 7; normalization used in previous work has been removed to facilitate comparison ) . ( G ) Bias in sound localization responses plotted as a function of training session for the subject in E . Positive values indicate that responses were biased toward the side of the open ear . Data are plotted separately for low- ( 1 kHz , dark blue ) and high-frequency ( 8 kHz , light blue ) tones . Scores for each session ( dots ) were fitted using linear regression ( lines ) to calculate slope values , which quantified the change in response bias ( Δ bias ) with training . Negative values of Δ bias indicate a shift in response bias toward the side of the plugged ear . ( H ) Δ bias for low- and high-frequency tones plotted for each subject ( gray lines; n = 11 ) . Mean values for Δ bias across subjects ( ± bootstrapped 95% confidence intervals ) are shown in blue . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 006 We next considered the relationship between these two adaptive processes . Although cue remapping and cue reweighting share a similar time-course ( significant correlation between the amount of remapping and reweighting across sessions; Figure 3A , r = 0 . 81 , P = 0 . 028 ) , the overall amount of cue remapping exhibited by each subject was independent of the amount of cue reweighting ( Figure 3B , r = 0 . 03 , P = 0 . 90 ) . This inter-subject variability was not attributable to differences in the effectiveness of earplugs used ( Figure 3—figure supplement 1 ) . Instead , we found that these two adaptive processes are affected by the frequency composition of the stimulus in different ways ( Figure 3C , interaction between sound frequency and adaptation type , p = 0 . 005 , permutation test ) . As expected , cue reweighting was greater for frequencies where spectral cues are most prominent in humans ( ≥4 kHz , Figure 3C , p<0 . 05 , post-hoc test; Figure 3—figure supplement 2 ) ( Blauert , 1997; Hofman and Van Opstal , 2002 ) , whereas equal amounts of cue remapping were observed for tones above and below 4 kHz ( Figure 3C , p>0 . 05 , post-hoc test ) . 10 . 7554/eLife . 12264 . 007Figure 3 . Relationship between different adaptive processes . ( A ) Time-course of behavioral adaptation for adult humans , measured by the amount of cue reweighting ( pink ) and remapping ( blue ) . Data are normalized ( z scores ) to facilitate comparison between different adaptation measures . All data have been averaged across subjects . ( B ) Comparison between the amount of behavioral cue reweighting and remapping for individual human subjects ( black dots; n = 11 ) . Variation in the degree of adaptation across subjects was not attributable to differences in earplug effectiveness ( Figure 3—figure supplement 1 ) . ( C ) Amount of cue remapping ( blue ) and cue reweighting ( pink ) observed at frequencies above ( lighter shades ) and below ( darker shades ) 4 kHz . Greater reweighting of spectral cues ( more positive values ) is observed >4 kHz , which is where spectral cues are most prominent in humans . Frequency-specific measures of cue reweighting were determined using reverse correlation ( see Materials and methods , Figure 3—figure supplement 2 ) . ( D ) Bilateral extracellular recordings were performed in the primary auditory cortex of ferrets reared with an earplug in one ear . These data were then compared with controls to obtain measures of cue reweighting and cue remapping ( see Materials and methods ) . ( E ) Cue reweighting versus cue remapping , with each dot representing either a single neuron or small multi-unit cluster ( n = 505 ) . ( F ) Amount of cue remapping ( blue ) and cue reweighting ( pink ) observed for neurons tuned to frequencies above ( lighter shades ) or below ( darker shades ) 8 kHz . To facilitate comparison between measures of cue reweighting and remapping at different frequencies , these values were normalized separately so that they each had an overall mean of 0 and a variance of 1 . Greater reweighting of spectral cues ( more positive values ) is observed > 8 kHz , which is where spectral cues are most prominent in ferrets . Relative to humans , spectral cues in ferrets are shifted toward higher frequencies because of differences in head and external ear morphology . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 00710 . 7554/eLife . 12264 . 008Figure 3—figure supplement 1 . Variation across subjects in the degree of adaptation to acute asymmetric hearing loss is not related to differences in earplug effectiveness . ( A ) Effect of earplug on mean hearing threshold ( Δ threshold ± SD ) is plotted as a function of frequency . Positive values indicate thresholds were higher when an earplug was worn . Data from Kumpik et al . ( 2010 ) are replotted ( red ) alongside those from the present study ( black ) . For visualization purposes , symmetric displacements along the x-axis have been introduced to each dataset . ( B ) A significant correlation ( p<0 . 05 ) was observed between Δ threshold and the initial drop in sound localization performance when an earplug was worn during the first training session ( Δ performance; change in% correct relative to normal hearing conditions averaged across all stimulus types ) . In other words , initial sound localization deficits were more extensive when the earplug produced greater attenuation . Each dot represents an individual subject . ( C , D ) No obvious relationship was observed between Δ threshold and the degree of remapping ( C ) or reweighting ( D ) observed in individual subjects ( dots ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 00810 . 7554/eLife . 12264 . 009Figure 3—figure supplement 2 . Determining the behavioral importance of spectral features at different frequencies using reverse correlation . ( A ) Although the mean spectrum of random-spectrum stimuli was close to zero when averaged across all trials ( black ) , distinct spectral features emerged when averaging was restricted to trials on which subjects responded to a particular location ( gray/color ) . This provides insight into which spectral features influence sound localization behavior . To reduce the noise in this estimate , a threshold was applied ( mean ± 1 . 5 SD , dashed lines ) and this process was repeated for each response location to construct a reverse correlation map . ( B ) Reverse correlation map showing the mean stimulus spectrum associated with each response location . Color is proportional to spectral amplitude , as illustrated in A . In order to quantify the behavioral importance of spectral features in different frequency bands , we calculated the ‘feature strength’ by averaging the unsigned magnitude of these spectral features across locations . ( C ) Cue reweighting plotted as a function of frequency . Cue reweighting was estimated by calculating training-induced changes in feature strength ( i . e . feature strength values obtained in the first session were subtracted from those obtained in subsequent sessions; these differences in feature strength were then averaged ) . Positive cue reweighting values indicate an increase in feature strength , which reflects increased behavioral importance of spectral cues . Dotted line shows the upper 95% confidence interval for cue reweighting values that would be expected under the null hypothesis that cue reweighting did not occur . Values above this line ( red symbols ) indicate cue reweighting values that are significantly greater than chance . DOI: http://dx . doi . org/10 . 7554/eLife . 12264 . 009 This indicates that these adaptive processes are relatively independent of one another and suggests that they may depend on distinct neural substrates . This motivated us to revisit neurophysiological measures of cue reweighting and remapping in ferrets reared with an intermittent hearing loss in one ear ( Figure 3D ) ( Keating et al . , 2013; Keating et al . , 2015 ) . In common with our human behavioral data , we found no correlation between the degree of cue reweighting and remapping in cortical neurons recorded from ferrets raised with one ear plugged ( Figure 3E , r = 0 . 08 , p = 0 . 073 ) . The type of plasticity observed also depended on the frequency preference of the neurons ( Figure 3F , interaction between unit characteristic frequency and adaptation process , p = 0 . 012 , permutation test ) . Greater cue reweighting was found in neurons tuned to frequencies where spectral cues are most prominent in ferrets ( >8 kHz , Figure 3F , p<0 . 05 , post-hoc test; frequency tuning bandwidth at 10 dB above threshold ( µ ± SD ) = 0 . 97 ± 0 . 51 octaves ) ( Carlile and King , 1994; Keating et al . , 2013 ) , whereas equal amounts of cue remapping occurred in neurons tuned to low and high frequencies ( Figure 3F , p>0 . 05 , post-hoc test ) . Thus , different neurons can exhibit cue remapping and reweighting in a relatively independent manner .
We have shown that adult humans can adapt to asymmetric hearing loss by both learning to rely more on the unchanged spectral localization cues available and by remapping the altered binaural cues onto appropriate spatial locations . Recent work has shown that both adaptive processes occur in response to monaural deprivation during development ( Keating et al . , 2013; 2015 ) . Our results suggest that this flexibility is likely to be a general feature of neural processing that also occurs in adulthood . Moreover , we show that these two forms of adaptation emerge together and that remapping of binaural spatial cues occurs at low as well as high frequencies , indicating plasticity in the processing of both ITDs and ILDs . Although adaptive changes in sound localization have previously been observed when human subjects wear an earplug for prolonged periods of everyday life ( Kumpik et al . , 2010 ) , we found here that much shorter periods of training are sufficient to induce adaptation to an episodic hearing loss . Our results also demonstrate that subjects adapt using a combination of cue remapping and cue reweighting . In contrast , previous work has shown that cue remapping did not occur when subjects wore an earplug most of the time for several days , and were therefore able to interact with their natural environments under these hearing conditions , but received relatively little training ( Kumpik et al . , 2010 ) . This suggests that the nature of adaptation may depend on the behavioral or environmental context in which it occurs . Consequently , it should be possible to devise training protocols that would help subjects to adapt to altered auditory inputs in ways that do not ordinarily occur , or occur more slowly , during the course of everyday life . When both adaptive processes occur together , observed either behaviourally in adult humans or neurophysiologically in monaurally-deprived ferrets , there was no obvious relationship between the amount of cue remapping and reweighting . This is at least in part because the spatial cues involved differ in their frequency dependence . Whereas equal amounts of binaural cue remapping occurred at different frequencies , spanning the range where both ITDs and ILDs are available , reweighting of spectral cues was restricted to those frequencies where these cues are most prominent . This suggests that the neural substrates for cue remapping and reweighting are at least partially distinct , with separate populations of cortical neurons displaying different types of spatial plasticity depending on their frequency preferences and sensitivity to different spatial cues . It is not known , however , whether remapping and reweighting occur at different stages of the processing hierarchy . Although experience-dependent plasticity in the processing of binaural cues has been observed at multiple levels of the auditory pathway ( Keating et al . , 2015; Popescu and Polley , 2010; Seidl and Grothe , 2005 ) , the changes induced by unilateral hearing loss during development are more extensive in the cortex than in the midbrain ( Popescu and Polley , 2010 ) . Much less is known about the neural processing of spectral localization cues and how this might be affected by experience ( Carlile et al . , 2005; Keating et al . , 2013 ) . However , reweighting of these cues is likely to reflect a change in the way they are integrated with other cues , which is thought to occur in the inferior colliculus ( Chase and Young , 2005 ) . This is consistent with the finding that adaptive changes in sound localization behavior in monaurally deprived adult ferrets rely on descending projections from the cortex to the inferior colliculus ( Bajo et al . , 2010 ) . It is likely therefore that adaptive plasticity emerges via dynamic interactions between different stages of processing ( Keating and King , 2015 ) . Although we found evidence for both cue reweighting and cue remapping in our human behavioral and ferret neurophysiological data , the nature of the episodic hearing loss in each case was very different . Whereas ferrets had one ear occluded for ~80% of the time over the course of several months of development ( Keating et al . , 2013; 2015 ) , adult human subjects wore an earplug for only ~7 hr in total ( 1 hr every 1–3 days ) . It is not known whether comparable physiological changes to those observed in the ferrets are responsible for the rapid shifts in localization strategy in adult human listeners following these brief periods of acute hearing loss . Nevertheless , the close similarity in the results obtained in each species has important implications for the generality of our findings . Our results emphasize the flexibility of neural systems when changes in sensory input affect ethologically important aspects of sensory processing , such as sound localization . They also reveal individual differences in the adaptive strategy adopted ( Figure 3B ) . Further work is needed to understand the causes of these differences and to determine whether knowing how different individuals adapt to hearing loss could help tailor rehabilitation strategies . Our results also highlight the importance of training in promoting multiple adaptive processes , and this is likely to be relevant to other aspects of sensory processing ( Feldman and Brecht , 2005; Keating and King , 2015; Sengpiel , 2014 ) , particularly in situations where changes in sensory input affect some cues but not others .
All neurophysiological procedures have been previously described in detail ( Keating et al . , 2013; 2015 ) . Bilateral extracellular recordings were made under medetomidine/ketamine anaesthesia from primary auditory cortex units ( n = 505 ) in response to virtual acoustic space stimuli generated from acoustical measurements in each animal . These stimuli recreated the acoustical conditions associated with either normal hearing or an earplug in the left ear and were used to manipulate individual spatial cues independently of one another . Cue weights were determined by calculating the mutual information between neuronal responses and individual spatial cues . A weighting index was then used to calculate the weight given by each neuron to spectral cues provided by the right ear ( i . e . contralateral to the developmentally-occluded ear ) relative to all other available cues . The mapping between binaural spatial cues and neurophysiological responses was measured by determining the best ILD for each unit , which represented the ILD corresponding to the peak of the binaural interaction function ( see Keating et al . , 2015 for more details ) . Best ILDs and weighting index values were converted to z scores using the corresponding means and standard deviations of data obtained from controls . Data were normalized separately for each hemisphere and different frequency bands . Measures of cue reweighting and remapping for each unit therefore respectively reflected changes in weighting index values and best ILDs relative to those observed in controls . These values were then normalized again so that measures of reweighting and remapping had the same overall mean ( 0 ) and variance ( 1 ) prior to comparing the amount of each form of adaptation in different frequency bands . Frequency tuning was calculated using 50-ms tones ( 0 . 5–32 kHz in 0 . 25 octave steps , varying between 30 – 80 dB SPL in increments of 10 dB ) . Characteristic frequency ( CF ) and bandwidth were calculated in a manner similar to that described previously ( Bartlett et al . , 2011; Bizley et al . , 2005 ) . Briefly , firing rates were averaged across stimulus repetitions ( n = 30 ) of each combination of frequency and level . This matrix was then smoothed with a boxcar function 0 . 75 octaves wide , following which a threshold was applied that was equal to the spontaneous rate plus 20% of the maximum firing rate . CF was defined as the frequency that elicited the greatest response at threshold . Bandwidth was measured at 10 dB above threshold by first calculating the area underneath the tuning curve . We then identified a rectangle that had the same area but constrained its height to be equal to the maximum firing rate . The width of this rectangle then provided a measure of bandwidth that approximates the width at half-maximum for a Gaussian tuning curve ( Bartlett et al . , 2011 ) . Confidence intervals at the 95% level were estimated empirically for different measures using 10 , 000 bootstrapped samples , each of which was obtained by re-sampling with replacement from the original data . These samples were then used to construct bootstrapped distributions of the desired measure , from which confidence intervals were derived . A bootstrap procedure was also used to assess the significance of group differences . First , the difference between two groups was measured using an appropriate statistic ( e . g . difference in means , t-statistic , or rank-sum statistic ) . The data from different groups were then pooled and re-sampled with replacement to produce two new samples , and the difference between these samples was measured using the same statistic as before . This procedure was subsequently repeated 10 , 000 times , which provided an empirical estimate of the distribution that would be expected for the statistic of interest under the null hypothesis . This bootstrapped distribution was then used to derive a P value for the difference observed in the original sample . In all cases , two-sided tests of significance were used , with Bonferroni correction used to correct for multiple comparisons . Cohen’s d was also calculated to provide a measure of the effect size for different types of adaptation in adult humans . The significance of factor interactions was also assessed using permutation tests ( Manly , 2007 ) . This involved randomly permuting observations across different factors and calculating an F statistic for each factor and interaction ( i . e . the proportion of variance explained relative to the proportion of unexplained variance ) . This procedure was repeated many times in order to assess the percentage of repetitions that produce F values greater than those obtained for the non-permuted data . This percentage then provided an estimate of the P values associated with each effect under the null hypothesis . Precise details of the permutation procedure used have been described elsewhere ( Manly , 2007 ) . Additional comparisons between conditions were made using appropriate post-hoc tests corrected for multiple comparisons . Although bootstrap and permutation tests were used because they make fewer distributional assumptions about the data , conventional parametric and non-parametric statistical tests were also performed and produced very similar results ( not reported ) . | The brain normally compares the timing and intensity of the sounds that reach each ear to work out a sound’s origin . Hearing loss in one ear disrupts these between-ear comparisons , which causes listeners to make errors in this process . With time , however , the brain adapts to this hearing loss and once again learns to localize sounds accurately . Previous research has shown that young ferrets can adapt to hearing loss in one ear in two distinct ways . The ferrets either learn to remap the altered between-ear comparisons , caused by losing hearing in one ear , onto their new locations . Alternatively , the ferrets learn to locate sounds using only their good ear . Each strategy is suited to localizing different types of sound , but it was not known how this adaptive flexibility unfolds over time , whether it persists throughout the lifespan , or whether it is shared by other species . Now , Keating et al . show that , with some coaching , adult humans also adapt to temporary loss of hearing in one ear using the same two strategies . In the experiments , adult humans were trained to localize different kinds of sounds while wearing an earplug in one ear . These sounds were presented from 12 loudspeakers arranged in a horizontal circle around the person being tested . The experiments showed that short periods of behavioral training enable adult humans to adapt to a hearing loss in one ear and recover their ability to localize sounds . Just like the ferrets , adult humans learned to correctly associate altered between-ear comparisons with their new locations and to rely more on the cues from the unplugged ear to locate sound . Which of these adaptive strategies the participants used depended on the frequencies present in the sounds . The cells in the ear and brain that detect and make sense of sound typically respond best to a limited range of frequencies , and so this suggests that each strategy relies on a distinct set of cells . Keating et al . confirmed in ferrets that different brain cells are indeed used to bring about adaptation to hearing loss in one ear using each strategy . These insights may aid the development of new therapies to treat hearing loss . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2016 | Behavioral training promotes multiple adaptive processes following acute hearing loss |
Estrogen Receptor-alpha ( ER ) drives 75% of breast cancers . Stimulation of the ER by estra-2-diol forms a transcriptionally-active chromatin-bound complex . Previous studies reported that ER binding follows a cyclical pattern . However , most studies have been limited to individual ER target genes and without replicates . Thus , the robustness and generality of ER cycling are not well understood . We present a comprehensive genome-wide analysis of the ER after activation , based on 6 replicates at 10 time-points , using our method for precise quantification of binding , Parallel-Factor ChIP-seq . In contrast to previous studies , we identified a sustained increase in affinity , alongside a class of estra-2-diol independent binding sites . Our results are corroborated by quantitative re-analysis of multiple independent studies . Our new model reconciles the conflicting studies into the ER at the TFF1 promoter and provides a detailed understanding in the context of the ER’s role as both the driver and therapeutic target of breast cancer .
The study of the Estrogen Receptor-α ( ER ) has played a fundamental role in both our understanding of transcription factors and cancer biology . The ER is one of a family of transcription factors called nuclear receptors . Nuclear receptors are intra-cellular and , on activation by their ligand , typically undergo dimerisation and bind to specific DNA motif ( for ER: Estrogen Response Elements; EREs ) . On the chromatin , the nuclear receptor recruits a series of cofactors and promotes the basal transcription mechanism at either nearby promoters or through chromatin loops from distal enhancers . Because of the minimal nature of these systems relative to other signaling pathways , nuclear receptors have become a model system for transcription factor analysis . Simultaneously , the role of nuclear receptors as drivers in a range of hormone dependent cancers has led to focused studies in the context of the disease . Previously , it was reported that the ER and key cofactors followed a cyclical pattern in breast cancer cell lines with maximal binding at 45 min after stimulation with estra-2-diol ( Shang et al . , 2000; Métivier et al . , 2003 ) . Similar results were also reported for the AR after activation with DHT ( Kang et al . , 2002 ) and several follow-up studies exist looking at single genomic loci ( Herynk et al . , 2010; Luo et al . , 2005; Shao et al . , 2004; Burakov et al . , 2002 ) . However , subsequent genome-wide studies have provided little further detail on the specific nature of the proposed kinetics of ER binding being either limited in the number of replicates or lacking temporal resolution ( Honkela et al . , 2015; wa Maina et al . , 2014; Dzida et al . , 2017; Guertin et al . , 2014 ) . In our own network analysis ( Holding et al . , 2018 ) , we focused on 0 , 45 and 90 min and found no significant reduction in ER signal at 90 min . In the same study , quantitative proteomic analysis of ER interactions at the same time intervals by qPLEX-RIME ( Papachristou et al . , 2018 ) shows no significant difference in terms of ER interactions at 45 and 90 min . These conflicting results have so far not been resolved . Routinely used assays to measure protein binding to chromatin are based on Chromatin Immunoprecipitation ( ChIP ) . A major challenge to monitoring ER activation through ChIP is the normalization of the ChIP signal — either genome-wide with next generation sequencing or at individual loci by qPCR — as the standard protocols do not control for a significant number of confounding factors including the efficiency of the immunoprecipitation step . In the case the of the two original studies ( Shang et al . , 2000; Métivier et al . , 2003 ) , the data only provided limited controls in this regard . An alternative method that has been applied to normalize ChIP-seq data is to use the maximal read count obtained at each individual site across each time point ( Guertin et al . , 2014 ) ; however , this method is at the expense of monitoring the magnitude of ER binding and gives equal weight to low read count peaks and more robust data from stronger binding sites . In the context of these challenges , we applied two strategies to robustly and accurately monitor the process of nuclear receptor binding to chromatin on activation . The first strategy was to increase the number of replicates . We generated sample data for six independent isogenic experiments to enable better characterization of the variance within the data . This strategy provided an unprecedented level of information regarding ER activation with twice the level of replication used in previous ChIP-qPCR studies ( Métivier et al . , 2003 ) and a significant improvement on previous single replicate genome-wide studies . The second strategy was to use our recently developed method for precise quantification of binding , Parallel-Factor ChIP ( pfChIP ) ( Guertin et al . , 2018 ) , which uses an internal control for quantitative differential ChIP-seq . Combined , these two strategies enabled us to undertake the most comprehensive and precise analysis of ER activation to date .
We measured ER-binding in MCF7 cells , a widely used model system for ER biology . To maximize the reproducibility of our results , MCF7 cells were grown from ATCC stocks , keeping passaging to a minimum , and the cell line origin was confirmed by STR genotyping . Additionally , to ensure the MCF7 cell line did not show significant genetic drift during culturing within our laboratory , we applied CellStrainer ( Ben-David et al . , 2018 ) to the input data from our ChIP-seq experiments . The fraction of genome with copy-number discordance was estimated at 0 . 2787 , within the range of 0 to 0 . 3 as published by CellStrainer’s developers to ensure similar therapeutic response . Sequencing reads from the analysis of 60 pfChip-seq samples targeting ER and six input samples were demultiplexed and aligned to the Homo sapiens GRCh38 reference assembly . Visual inspection of the data using the Integrative Genomics Viewer ( IGV ) Viewer ( Robinson et al . , 2011 ) confirmed enrichment at known ER binding sites ( exemplified by TFF1 in Figure 1—figure supplement 1 ) and the presence of previously reported CTCF control peaks ( Guertin et al . , 2018 ) . From visual inspection , pfChIP-seq samples qualitatively showed minimal ER binding at 0 min while CTCF binding was constant at all time points . Peak count data from CTCF binding sites were used to normalize between conditions as these sites have previously been shown to be unchanged in response to estrogen ( Ross-Innes et al . , 2011 ) , with >70 000 binding sites discovered across all samples and >50 000 CTCF binding sites found in over 50% of samples . Analysis after normalization of the raw data showed similar levels of variability in terms of signal ( Figure 1—figure supplement 2 ) as we saw when developing the pfChIP method ( Guertin et al . , 2018 ) . The resultant normalized binding matrix of ER binding was used for all downstream analyses and is provided as Supplementary file 1 . Normalized count data for the TFF1 promoter showed that on activation with estra-2-diol the ER rapidly ( in less than 10 min ) binds the TFF1 promoter . Binding after this time point shows no significant changes ( Figure 1 ) . Analysis of the data by individual replicates ( Figure 1—figure supplement 3 ) did not demonstrate evidence of oscillatory binding in individual replicates either with a period of 90 min period or an alternative frequency . Comparison of the variance in the ER binding after induction shows that there is significantly more variance ( F-test , time points >= 10 min , p-value < 1 × 10-10 ) in the ER binding data than in CTCF binding between replicates . In contrast , pairwise F-test ( two-sided , FDR < 0 . 05 ) for ER binding at all time points showed no significant difference in the variance for any comparison . As the variance of CTCF binding in pfChIP-seq is a good estimator of the technical variance , the most likely source of increased variance in ER binding is therefore biological . These findings were validated through analysis of the RARA promoter and proximal CTCF peaks ( Figure 1—figure supplement 4 ) , which gave consistent results to those seen at the TFF1 promoter . Previously , ER binding sites were shown to reach maximum occupancy at different time points depending on genomic location , revealing a P300 squelching mechanism at early time points ( Guertin et al . , 2014 ) . Therefore , to provide a partial validation of this study , we applied the same principles of their analysis to our data , that is normalizing in the time-space setting maximum occupancy to 1 . Consistent with the previous study , the two time points with the largest numbers of sites reaching maximal occupancy in both data sets were at 10 and 40 min ( Figure 2A ) . As the remaining time points were unique to the individual data sets , these could not be directly compared . While grouping by maximum occupancy in Figure 2A was essential to highlight these features in the context of Guertin et al . 's previous study , in our analysis we found this method distorts the data and the effects that drove the appearance of blocks are not statistically significant in our dataset ( Figure 2—figure supplement 1 ) with the exception of the 0 to 10 min contrast . As far as we are aware , the loss on ER binding on activation with estra-2-diol is unprecedented , and therefore the presence of a block of ER sites with maximal binding at 0 min warranted deeper investigation . Analysis of the class average ( Figure 2—figure supplement 2A ) showed that the variance of this class is much greater than the decrease seen between 0 and 10 min . A more detailed analysis of the individual trajectories of each binding site ( Figure 2—figure supplement 2B ) showed a similar result , with the maximal binding at time zero appearing marginal and in all cases . We therefore concluded that maintaining the group of the blocks between Figure 2A and B visually overemphasized this feature within the data . pfChIP-seq allowed us to improve on the previous study by directly normalizing the data to the internal control . The resultant binding matrix provided quantification of the absolute binding affinity at each time point ( Figure 2B ) . Comparison of Figure 2A and B demonstrates the effects of different data normalization strategies . The relative normalization to maximum binding emphasizes binding maxima ( red blocks in Figure 2A ) while the absolute normalization to an internal control shows that these maxima are very shallow , barely visible in Figure 2B , and other features dominate the data . A few genes show very high levels of ER binding ( visible as thin red lines in Figure 2B ) , while most genes show intermediate levels and some very low levels ( blue lines ) . These different levels of ER binding are preserved over time , with only time point 0 showing very low levels for all genes . To elucidate potential different temporal responses to ER activation by estra-2-diol , we applied t-SNE ( Maaten and Hinton , 2008 ) , a widely used method for dimensionality reduction and data visualization ( Figure 3 ) . Each dot in the plot represents a binding site over time , that is one row in the binding matrix shown in Figure 2B . We colored each dot by the false discovery rate ( FDR; ( Benjamini and Hochberg , 1995 ) ) for the change in ER affinity between 0 to 10 min . This analysis revealed two major trajectories of binding sites in the data , one dominated by low FDR ( orange ) and one by high FDR ( blue ) . Both trajectories saw an increasing affinity in the direction of the white arrow . This pattern was stable for a wide range of perplexity , the main t-SNE parameter ( Figure 3—figure supplement 1 ) . We named the estra-2-diol responsive trajectory A , and the estra-2-diol independent trajectory B . The set of genomic sites found at the end of each trajectory were named Class A and B respectively . Motif analysis of Class A peaks demonstrated significant enrichment for the full estrogen response element ( ERE , ( Klein-Hitpass et al . , 1986 ) ) , while Class B gave enrichment for the FOXA1 binding site . Analysis of Class C ( i . e . weaker responding genes on trajectory A ) gave a partial ERE match , suggesting a greater divergence from the ERE motif and consistent with the lower levels of ER affinity found on ER activation at these sites ( Driscoll et al . , 1998 ) . Average binding profiles were computed for both Class A and Class B . Class A showed minimal binding at 0 min followed by a robust response before 10 min , the binding affinity then remained similar for the remaining time points . In contrast , Class B displayed estra-2-diol independent binding at 0 min and average ER binding affinity saw no significant changes between time points . Class C gave a similar profile to Class A ( not shown ) , but with reduced amplitude . The average amplitude of the binding from 10 to 90 min displayed a greater ER affinity for Class A then Class B . Genomic regions enrichment of annotations tool ( GREAT ) analysis ( Welch et al . , 2014 ) of Class B binding sites ( Supplementary file 2 ) identified the enrichment of six amplicons previously identified from the analysis of 191 breast tumor samples , q = 5 . 6 × 10-41 to q = 3 . 3 × 10-8 , ( Nikolsky et al . , 2008 ) and a set of genes upregulated in luminal-like breast cancer cell lines compared to the mesenchymal-like cell lines , q = 1 . 9 × 10-13 , [Charafe-Jauffret et al . , 2006] ) . Undertaking the same analysis of the ER only ChIP-seq data stream gave very similar results to that of pfChIP-seq analysis , confirming that any potential cycling is not suppressed by the method ( Figure 3—figure supplement 2 ) . As with the pfChIP-seq analysis , no clear cycling was seen for the individual replicates ( Figure 3—figure supplement 2C ) . Class A binding sites showed the strongest response to estra-2-diol , the greatest enrichment of the estrogen response element and contained the classical ER binding site at TFF1 . We therefore focused further analysis on these peaks to minimize confounding factors . A t-SNE plot of only Class A sites ( Figure 4A ) did not provide distinct clustering of points . Partial separation was seen on the basis of time point of maximal binding ( left to right ) and amplitude ( approximately top to bottom ) . As the class profiles may average out site-specific oscillatory kinetics , we undertook analysis of individual ER binding sites . Peaks were annotated on the basis of the nearest Transcription Start Sites ( TSS ) and profiles for key ER target genes TFF1 and GREB1 were generated . As previously seen in Figure 1—figure supplement 1 , ER binding at TFF1 was stable after induction . The same response was seen at the TFF1 enhancer ( dark red ) . Analysis of ER binding proximal to GREB1 again showed a robust and unidirectional response to estra-2-diol . Profiles of ER binding that showed either early or late maximal ER affinity were individually investigated . Binding near the TSS of SNX24 and ACKR3 are provided as representative examples . Given we found a robust and stable response to ER activation by estra-2-diol in contrast to the cyclical response previously described ( Shang et al . , 2000 ) , we reviewed studies that have investigated ER binding at the TFF1 promoter . Several studies either used a different promoter ( Park et al . , 2005 ) , factor ( Li et al . , 2003 ) or estra-2-diol concentration/include α-amanitin ( Métivier et al . , 2003 ) . By manually reviewing the first 1000 citations of ( Shang et al . , 2000 ) , we identified several studies ( Burakov et al . , 2002; Herynk et al . , 2010; Shao et al . , 2004 ) that undertook the same analysis in the MCF7 cell line , with the same concentration of estra-2-diol , same crosslinking time scale , and at the same promoter as the best datasets for comparison with each other and to our dataset . Since the numerical values of ER binding occupancy were not available for these studies , we read the values off the provided charts or undertook image analysis of figures ( Supplementary file 4 ) . Comparison of the data from all four studies gave little or no consistency in the temporal profile of ER , AIB1 and P300 binding at these sites ( Figure 4B ) . Quantitative image analysis presents limitations in reestablishing the exact values without the primary data; however , the primary data was not available . In lieu of this , analysis of the published data was considered adequate as the relative intensities will be preserved . Interpretation is further hindered as these studies only report a single replicate for analysis , thereby making it impossible to quantify uncertainty in the data . Therefore , there is no consistent evidence for cycling in the studies using the same conditions as the original observation .
By undertaking six biological replicates and incorporating an internal control with pfChIP , we have produced the most comprehensive analysis to date of ER binding over the first 90 min after stimulation with estra-2-diol . We found the sites at which we detected ER binding on the chromatin follows two distinct trajectories , either the rapid activation within 10 min followed by a stable response or ligand independent binding . Enrichment of the FOXA1 motif in the strongest ligand-independent/Class B sites supports our hypothesis: that these are as a result of ER interactions at these sites . Importantly , the de novo motif analysis did not find the presence of the CTCF motif , confirming that they are not an artifact of utilizing CTCF to normalize via the pfChIP-seq method . Analysis of the Class B binding sites with GREAT ( Welch et al . , 2014 ) , Supplementary file 2 , gave enrichment for 6 out of 30 ER regulated amplicons identified in a previous study of 191 breast cancer tumor samples ( Nikolsky et al . , 2008 ) . On the basis that no ERE was found at Class B sites and that the affinity of ER at these sites was less than at estra-2-diol response sites , we propose that these sites represent open regions of chromatin where ER can be recruited by other transcription factors in the absence of its own ligand . However , these interactions are weak , and very likely transient , as the average binding affinity for Class B sites is similar in level ( a normalized read count of 30–40 ) to the binding before activation at Class A binding sites , but greater than Class C binding sites ( ≈ 10 normalized reads increasing to ≈ 40 on activation ) . Ligand dependent activation of ER was seen robustly at Class A sites , but displayed no evidence of cyclical binding . We propose instead that ER activation occurs rapidly , within 10 min and binding shows no significant change after this point . The two examples we demonstrated — of increasing or decreasing ER binding after activation at the SNX24 and ACKR3 TSS ( Figure 4A ) — should be interpreted with caution as , while downstream effects are likely to modulate ER binding , searching for individual outliers results within a large data set will generate false positives . Nonetheless , the two examples imply a secondary level of modulation does occur as previously seen , but at much lower magnitude than proposed in studies focused on ER cycling . It is possible that alternative conditions may be able to induce tightly regulated cycling; however , we feel this is unlikely in terms of physiology . For example , the work of Metivier et al . makes use of α-amanitin , a RNA polymerase inhibitor . Within the cancer biology setting , these conditions have no direct interpretation . Worse , the mode of action is downstream of the ER and therefore is a confounding factor , not a clear method of synchronization . In light of our results and the lack of consistency of published results , we propose that the previously described cyclical response kinetics are likely an artefact of observing a highly variable process without replicates . With replicates , the cyclical effect is lost when averaging . Even if a cyclical response existed , our results indicate that it is not regulated tightly enough to be coherently visible across multiple replicates . The variance in ER binding may better be described by heterogeneity in the cell populations before induction and by current models regarding expression noise as an indicator for greater transcription responsiveness ( Morgan and Marioni , 2018 ) . Finally , our proposal provides , for the first time , a model that reconciles ChIP-seq data with the stochastic model of nuclear receptor binding as proposed and visualized by those undertaking single model imaging . In contrast , ER cycling has always been irreconcilable with these alternative forms of data ( Lenstra et al . , 2016 ) While we cannot discount that our cells could have specifically lost the ability to regulate ER binding in the manner previously described , we have minimized this possibility through the use of cells direct from ATCC , by confirming the cell line by STR genotype and applying the latest methods ( Ben-David et al . , 2018 ) to confirm that our cell line is genetically similar to the strains used in other labs . Nonetheless , we would welcome further replication of this study . In summary , through the use of stringent internal controls , we have reproducibly shown that estra-2-diol responsive ER binding is sustained and not cyclical , with the magnitude of the binding primarily defined by the conservation of the ERE at the binding site .
MCF7 cells ( RRID:CVCL_0031 ) were obtained from ATCC and confirmed by STR genotype before culture . For each immunoprecipitation , cells from 2 × 15 cm dishes were used . In each 15 cm plate , 2 × 106 were seeded and grown for 3 days in DMEM ( Glibco ) with 10% FBS before washing with phosphate buffered saline . Media was replaced with charcoal stripped and phenol red-free DMEM medium . Media was replaced daily for 4 days to ensure removal of estrogenic compounds . Plates were stimulated on day 5 with a final concentration of 100 nM estra-2-diol in EtOH before crosslinking at the required time . All six replicates were done on different dates and represent different passages . MCF7 cells were obtained from ATCC . The cell line was authenticated using STR profiling and are confirmed Mycoplasma free . Parallel-factor ChIP-seq was performed as previously described ( Guertin et al . , 2018 ) . CTCF antibody was D31H2 Lot:3 ( RRID:AB_2086791 , Cell Signaling ) . ER antibody was 06–965 Lot:3008172 ( Millipore ) . Reads were aligned using BWA ( Li and Durbin , 2009 ) , and ENCODE blacklist regions ( ENCODE Project Consortium et al . , 2012 ) were removed as previously described ( Carroll et al . , 2014 ) . Duplicate reads were removed and peak calling was undertaken using MACS2 ( Zhang et al . , 2008; Feng et al . , 2012 ) . ER and CTCF peaks were filtered according to the pfChIP-seq protocol ( Guertin et al . , 2018 ) , before normalization and differential binding analysis with Brundle/DiffBind ( Guertin et al . , 2018; Ross-Innes et al . , 2012 ) in R . t-SNE plots were generated with Rtsne ( Krijthe , 2015 ) . Perplexity was tested from 2 to 200 to confirm the stability of the transformation of the data into 2-dimensional space ( Figure 3—figure supplement 1 ) . Lower perplexities , 2 and 5 , gave minimal structure . For perplexities tested between 30 and 200 , two stable trajectories were seen in all cases . GREAT ( Welch et al . , 2014 ) was used to analyze Class B binding sites . Band intensities from previously published studies were measured with ImageJ ( Schneider et al . , 2012 ) . Sequencing data have been deposited in GEO under accession code GSE119057 . | Breast cancer is the most common type of cancer worldwide . The hormone estrogen drives the growth of 70% of breast cancer tumors . This form of breast cancer is called estrogen receptor positive ( ER+ ) breast cancer . In the early 2000s , several scientists found that some genes in ER+ breast cancers turn on and off in 90-minute cycles . Moreover , when the estrogen receptor binds to the DNA in the nucleus of a cell , it activates nearby genes causing the tumor cells to grow and divide . Learning more about how cancer cells respond to estrogen is very important . Many cancer drugs block estrogen to stop its tumor growth promoting effects . But the initial studies of estrogens effects were only able to look at how estrogen affected a small number of genes . Newer genome sequencing technologies allow scientists to study the effects of estrogen on more genes and provide more detailed information . Using these cutting-edge technologies , Holding et al . show that the 90-minute cycles found in the previous studies are likely artefacts of older techniques and lacking controls . The new experiments used a newer technique called parallel factor ChIP-seq to look at how all genes respond to the estrogen receptor . Then , Holding et al . reanalyzed data published in the previous studies and found that they were often contradictory and inconsistent . None of the genes – not even the ones looked at in earlier studies – were expressed in 90-minute cycles like the previous studies suggested . Instead , the expression of the genes was variable , which may make the cell even more responsive to estrogen . The previous reports of the 90-minute cycles are most likely explained by a bias of the human eye of finding patterns in a highly variable process that do not hold up to statistical analysis . Better understanding how estrogen influences genes and cell growth is essential to developing better treatments for ER+ breast cancer . This includes ruling out ideas that may be incorrect or misleading . These findings help resolve why not all studies have found estrogen receptor driven cycles of gene expression , and will provide researchers with a better foundation for future studies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"short",
"report",
"computational",
"and",
"systems",
"biology"
] | 2018 | Genome-wide Estrogen Receptor-α activation is sustained, not cyclical |
Hybrid sterility is one of the earliest postzygotic isolating mechanisms to evolve between two recently diverged species . Here we identify causes underlying hybrid infertility of two recently diverged fission yeast species Schizosaccharomyces pombe and S . kambucha , which mate to form viable hybrid diploids that efficiently complete meiosis , but generate few viable gametes . We find that chromosomal rearrangements and related recombination defects are major but not sole causes of hybrid infertility . At least three distinct meiotic drive alleles , one on each S . kambucha chromosome , independently contribute to hybrid infertility by causing nonrandom spore death . Two of these driving loci are linked by a chromosomal translocation and thus constitute a novel type of paired meiotic drive complex . Our study reveals how quickly multiple barriers to fertility can arise . In addition , it provides further support for models in which genetic conflicts , such as those caused by meiotic drive alleles , can drive speciation .
Identifying the molecular and evolutionary bases of hybrid sterility is necessary for understanding the mechanisms of speciation . Hybrid sterility is one of the earliest reproductive isolation mechanisms to evolve between two recently diverged species ( Coyne and Orr , 2004 ) , yet we are only beginning to understand the types of genetic changes that lead to hybrid infertility ( Coyne and Orr , 2004; Johnson , 2010; Presgraves , 2010 ) . Since the evolutionary forces driving genetic changes that cause infertility between species are likely also acting within species , the study of hybrid sterility also promises significant insight into mechanisms underlying infertility within species . The ( Bateson ) Dobzhansky-Muller ( BDM ) model provided a solution to the paradox of how genetic changes that lead to speciation could be tolerated by natural selection despite decreasing the fitness potential of an organism . This model proposes that hybrid sterility results from incompatibilities between genes that evolved in different populations and were therefore never tested together by natural selection ( Coyne and Orr , 2004 ) . Indeed , incompatible BDM pairs have been identified in diverse organisms that either cause hybrid sterility or reinforce species isolation ( Brideau et al . , 2006; Lee et al . , 2008; Bayes and Malik , 2009 ) . Although relatively few loci underlying hybrid incompatibilities have been identified , one theme that has emerged is that the loci are often rapidly evolving and implicated as players in ‘molecular evolutionary arms races’ . These arms races can occur between host genomes and external forces such as parasites ( Bomblies et al . , 2007 ) . Alternatively , the genetic conflicts can be between different elements within a genome , such as between selfish parasitic genes and other host genes ( Johnson , 2010; Presgraves , 2010 ) . Despite their explanatory power , DM incompatibilities are not exclusive causes of hybrid infertility . For instance , changes in ploidy are a rapid means of speciation in plants ( Otto and Whitton , 2000 ) . Defects in meiotic recombination contribute to hybrid infertility in both mouse and budding yeast hybrids ( Hunter et al . , 1996; Bhattacharyya et al . , 2013; Mihola et al . , 2009 ) . In addition , genomic rearrangements can also cause or contribute to speciation ( White , 1978; Faria and Navarro , 2010; Hoffmann and Rieseberg , 2008; Noor et al . , 2001 ) . In the classic chromosomal speciation model , chromosomal rearrangements between populations lead to infertility when heterozygous . Like DM gene incompatibilities , chromosomal rearrangements can contribute to hybrid infertility and serve as a genetic barrier between populations ( White , 1978 ) . For example , the transposition of an essential fertility gene causes male infertility in some Drosophila hybrids and chromosomal rearrangements contribute to hybrid infertility in some budding yeast hybrids ( Masly et al . , 2006; Delneri et al . , 2003 ) . How do chromosomal rearrangements become established in organisms in which they cause infertility when heterozygous ? One possibility is that a rearrangement could become fixed in a small population via genetic drift and inbreeding ( Rieseberg , 2001 ) . White proposed an alternative solution in which novel chromosomal rearrangements could increase in frequency if they were linked to meiotic drive alleles ( White , 1978 ) . These selfish genetic elements ‘cheat’ to be transmitted to more than 50% of the functional gametes of a heterozygote ( Burt and Trivers , 2006 ) . Due to their transmission advantage , meiotic drive alleles and loci linked to them can spread through a population even if they cause fertility decreases ( Crow , 1991 ) . In this way , even a chromosomal rearrangement that causes decreased fertility when heterozygous could become fixed in a population if it is linked to a strong meiotic drive allele . Because loci linked to drive alleles also benefit from the transmission advantage , linked variants that enhance drive will also be selected ( Crow , 1991 ) . Chromosome inversions that prevent recombination thereby reinforcing the linkage between drive alleles and their enhancers can also spread through a population due to enhanced drive . In this way , meiotic drive alleles can even promote the evolution of chromosomal rearrangements , in spite of their fitness costs . Consistent with this , meiotic drive loci are commonly found within inversions and have been proposed to underlie other types of dramatic karyotype evolution ( Dyer et al . , 2007; Larracuente and Presgraves , 2012; Pardo-Manuel de Villena and Sapienza , 2001; Hammer et al . , 1989 ) . However , there is currently little experimental or theoretical support for White's model; genetic conflicts and chromosomal rearrangements are still considered distinct causes of hybrid sterility and speciation . Here , we show that a combination of selfish meiotic drive loci and chromosomal rearrangements leads to near complete hybrid sterility between two recently diverged fission yeast species . This study provides support for models in which selfish genetic elements and chromosomal rearrangements are drivers of hybrid dysfunction ( Johnson , 2010; Presgraves , 2010; White , 1978 ) . Our observations in fission yeast are consistent with White's chromosomal speciation model .
S . pombe ( Sp ) and S . kambucha ( Sk ) generally exist as single-celled haploids , but cells of opposite mating types readily mate to form single-celled diploids . Fission yeasts are homothallic; they can switch mating types during mitotic growth and self-mate effectively . Although there is evidence of genetic outcrossing in yeasts closely related to Sp , the relative frequencies of outcrossing vs selfing are unknown ( Brown et al . , 2011 ) . Sk and Sp are 99 . 5% identical at the DNA sequence level ( Rhind et al . , 2011 ) . Despite their genetic similarity , Sk/Sp hybrid diploids are mostly infertile ( Singh and Klar , 2002 ) . To investigate the causes of the hybrid infertility , we generated a suite of genetic markers ( ‘Materials and methods’ ) that transformed Sk from a non-model yeast into a distinct genetically tractable model system . In addition to their use in this study of fertility , these tools will also facilitate molecular dissection of the functional consequences of evolution between fission yeast species . Sk/Sp hybrid diploids displayed no obvious mitotic defects , indicating there are no dominant lethal incompatibilities between the two species that act during mitosis ( Figure 1A , B ) . To assay hybrid fertility , we measured the viable spore yield of Sk/Sp hybrids , Sk/Sk , and Sp/Sp diploids ( Smith , 2009 ) . This assay measures the number of viable spores ( gametes ) produced per viable diploid placed on the starvation medium that induces cells to undergo meiosis . Spores are considered viable if they are able to grow into a visible colony . Because cells can undergo a few mitotic divisions prior to meiosis and not all spores can be recovered from the starvation medium , the assay is a relative , rather than an absolute , measure . Sk/Sk diploids had a slightly lower viable spore yield than Sp/Sp diploids , 3 . 6 vs 8 . 4 ( unpaired t test p=0 . 053; Figure 1C ) . However , when assayed via micromanipulation of individual spores , spores derived from Sp/Sp and Sk/Sk diploids were equally viable . We therefore conclude that the difference between Sp/Sp and Sk/Sk diploids in the viable spore yield assay is due to an extra mitosis of Sp/Sp diploids before meiotic induction , not higher viability of gametes . In sharp contrast , Sk/Sp hybrids had a viable spore yield at least 24-fold less than that of either of the pure species diploids ( t test p<0 . 01; Figure 1C ) . 10 . 7554/eLife . 02630 . 003Figure 1 . Sk/Sp hybrids are healthy but exhibit low fertility . ( A ) Sk/Sp hybrid diploids are morphologically similar to pure species diploids . ( B ) Sk/Sp hybrid diploids show no gross growth defects relative to pure species controls . ( C ) Viable spore yield tests show that Sk/Sp fertility is low relative to pure species controls ( averages of n ≥ 5 experiments , p-values obtained using t test ) . This assay does not directly measure viable spores per meiosis , so values can exceed 4 . ( D ) Sk/Sp hybrids complete both meiotic divisions with timing similar to that of pure species controls ( representative experiment of 3 , n ≥200 cells for each data point ) . ( E and F ) The asci produced by Sk/Sp hybrids contain spores that are more irregular and transparent than pure species asci . ( G ) The viable spores produced by Sk/Sp hybrids often grow into small irregularly sized and shaped colonies . ( H ) The majority of the viable spores produced by Sk/Sp hybrids are aneuploid or diploid ( p-values obtained using G-test , n >200 for each ) . These data are also shown in Figure 5A . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 003 We considered whether the low fertility of Sk/Sp hybrids reflected a defect in initiating or completing meiosis . To test this possibility , we monitored meiotic progression through a time course . Using DAPI staining of cells , we found that Sk/Sp hybrids initiated and completed both meiotic nuclear divisions with similar timing and efficiency compared to Sp/Sp and Sk/Sk diploids ( Figure 1D ) . These results suggest that spore inviability in products of hybrid meiosis is not due to an inability to enter meiosis or due to checkpoint activation preventing the completion of the meiotic divisions . However , we did find that the spores produced by Sk/Sp hybrids were more irregular and less refractile than those produced by pure species diploids ( Figure 1E , F ) . Furthermore , the viable spores produced by Sk/Sp hybrids grew into colonies that were grossly different from those of pure species controls ( Figure 1G ) . There were a few large colonies from the hybrid spores , but most were small and many had irregularly shaped edges , rather than large colonies with smooth circular perimeters like those produced by pure species spores . When the small hybrid spore colonies were streaked for single colonies on fresh Petri plates , some cells within the colony lost the slow growth phenotype and generated colonies similar in size to those of Sp or Sk haploids . These phenotypes are consistent with the small colonies being chromosome 3 aneuploids ( disomes ) ; aneuploidy for chromosome 1 or 2 is lethal in Sp ( Niwa et al . , 2006 ) . To formally test this idea , we measured the frequency of heterozygous aneuploids amongst viable spores using co-dominant heterozygous markers at allelic loci ( one on the Sk chromosome and one on the Sp chromosome ) . Spores that inherited both markers on chromosome 2 were heterozygous diploids . Spores that inherited both markers on chromosome 3 , but only one marker from chromosome 2 , were heterozygous aneuploids . Homozygous diploids and aneuploids were not detectable with the markers used here and were counted as haploids . From Sk/Sp hybrids , 33% of the viable spores were heterozygous aneuploids and 44% were heterozygous diploid , significantly more than the 2 . 2 and 3 . 5% , respectively , observed in Sk control diploids ( Figure 1H; p<0 . 01 for both; G-test ) . One possible explanation for the low fertility and the high frequency of aneuploids and diploids we observed in Sk/Sp hybrids , is that these hybrids may have defects in meiotic recombination . In many eukaryotes , meiotic recombination is essential for the production of viable gametes . This is because recombination can form crossovers , which help ensure proper disjunction of homologous chromosome pairs during the first meiotic division ( Kerr et al . , 2012 ) . When a pair of homologous chromosomes fails to form a crossover in meiosis , chromosomes segregate less faithfully and can produce aneuploid gametes . In both budding yeast and mouse hybrids , defects in meiotic recombination are major causes of hybrid infertility ( Hunter et al . , 1996; Bhattacharyya et al . , 2013; Mihola et al . , 2009 ) . These observations , combined with rapid evolution of genes that function in meiotic recombination , have led to the hypothesis that recombination defects serve as a general barrier to genetic exchange between species ( Rhind et al . , 2011; Anderson et al . , 2009; Sawyer and Malik , 2006; Myers et al . , 2010 ) . We therefore tested whether meiotic recombination is aberrant in Sp/Sk hybrids . Meiotic recombination is initiated by double-strand DNA breaks ( DSBs ) induced by a complex of proteins including the conserved Spo11 protein ( Rec12 in fission yeast ) . In many organisms , including Sp , meiotic DSBs are not randomly distributed; rather they are concentrated in regions called hotspots ( Cromie et al . , 2007 ) . DSB hotspots are thought to evolve rapidly because they can be lost due to biased gene conversion during break repair ( Boulton et al . , 1997 ) . Perhaps related to this , rapid divergence of a DSB hotspot-determining protein contributes to infertility of certain mouse hybrids and some human males ( Mihola et al . , 2009; Myers et al . , 2010; Segurel et al . , 2011; Baudat et al . , 2010; Irie et al . , 2009 ) . To test if DSB hotspot divergence contributes to Sk/Sp hybrid infertility , we mapped the locations of Sk DSB hotspots . Rec12 remains covalently linked to the DNA it cuts , so DSB sites can be mapped by identifying the DNA linked to Rec12 ( Keeney et al . , 1997 ) . To do this in Sk , we first generated a strain containing a FLAG-tagged rec12 gene and the rad50S mutation in which DSBs form normally but are not repaired ( Alani et al . , 1990; Hyppa et al . , 2008 ) . We then used chromatin immunoprecipitation on meiotic extracts to pull down Rec12-FLAG and the covalently attached DNA . We identified the bound DNA by amplification and hybridization to a microarray . Qualitatively , we found that the overall genome-wide pattern of DSB hotspots in Sk was similar to the published Sp hotspot map ( Figure 2—figure supplement 1; Cromie et al . , 2007 ) . We then compared the DSB hotspots between Sk and Sp at 286 previously identified Sp hotspots and found strong correlation between hotspot intensities in the two species ( Figure 2A; Fowler et al . , 2013 ) . This result shows that hotspots have largely not diverged between Sp and Sk , so such divergence is unlikely to contribute to hybrid infertility . 10 . 7554/eLife . 02630 . 004Figure 2 . DSB hotspot divergence and repair in Sk/Sp hybrids . ( A ) We used ChIP–chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the profile to the published DSB hotspot maps of Sp ( Fowler et al . , 2013 ) . We then compared the Rec12-enrichment in Sk at 286 defined Sp hotspots and found a strong correlation between Rec12 enrichments between the two species at these sites . ( B ) Sk/Sp cells are proficient at inducing DSBs . Ethidium bromide stained pulsed-field gel of diploids at 0 and 8 . 5 hr after inducing meiosis in liquid cultures . These diploids are rad50S mutants , so DSBs form normally but are not repaired . As DSBs are formed , the three full-sized chromosome bands disappear and the DNA runs as smaller broken fragments on the gel . ( C and D ) We find that DSBs are formed at similar locations and similar frequencies in Sk/Sp and Sk/Sk . Southern blots of pulsed-field gels to obtain a closer view of DSB formation in rad50S diploids probed to visualize two NotI restriction fragments known as NotI J [shown in ( C ) ] and NotI D [shown in ( D ) ] . Prior to DSB formation , most of the DNA runs as a single large band . After all break formation ( 8 . 5 hr ) smaller cut fragments become apparent at the same sites in Sk/Sp and Sk/Sk . ( E ) DSBs are efficiently repaired in Sk/Sp . Ethidium bromide stained pulsed-field gel of rad50+ diploids at the given times after the induction of meiosis show that DSBs do not accumulate more in Sk/Sp than the Sp/Sp control during meiotic prophase . Together with those in ( B ) these data demonstrate that Sk/Sp cells form and efficiently repair DSBs . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00410 . 7554/eLife . 02630 . 005Figure 2—figure supplement 1 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00510 . 7554/eLife . 02630 . 006Figure 2—figure supplement 2 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00610 . 7554/eLife . 02630 . 007Figure 2—figure supplement 3 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00710 . 7554/eLife . 02630 . 008Figure 2—figure supplement 4 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00810 . 7554/eLife . 02630 . 009Figure 2—figure supplement 5 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 00910 . 7554/eLife . 02630 . 010Figure 2—figure supplement 6 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01010 . 7554/eLife . 02630 . 011Figure 2—figure supplement 7 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01110 . 7554/eLife . 02630 . 012Figure 2—figure supplement 8 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01210 . 7554/eLife . 02630 . 013Figure 2—figure supplement 9 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01310 . 7554/eLife . 02630 . 014Figure 2—figure supplement 10 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01410 . 7554/eLife . 02630 . 015Figure 2—figure supplement 11 . DSB hotspots in Sk and Sp . We used ChIP-chip of Rec12-FLAG from rad50S Sk meiotic cultures to assay DSB hotspots and compared the Sk profile ( red ) to the published DSB hotspot maps of Sp ( black ) ( Fowler et al . , 2013 ) . Traces are shown in Figure 2—figure supplements 1–11 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 015 A second hypothesis is that hybrids have gross defects in forming or repairing meiotic DSBs , perhaps due to incompatibilities in the meiotic recombination machinery . To observe DSB formation , we used pulsed-field gel electrophoresis ( PFGE ) of whole chromosomes in rad50S hybrid and control diploids . We found that like those of pure species , the Sk/Sp chromosomes were effectively broken in meiosis ( Figure 2B ) . For a more precise view of DSB formation in hybrids , we assayed break formation in rad50S diploids at known meiotic DSB hotspots contained on the NotI restriction fragments designated ‘J’ ( 0 . 5 Mb ) and ‘D’ ( 1 . 1 Mb ) of the Sp genome ( Cromie et al . , 2007 ) . We found that both Sk and hybrid diploids exhibited similar amounts of breakage at the same sites , in accord with the microarray results above ( Figure 2C , D ) . To assay DSB repair , we visualized meiotic chromosomes via PFGE in rad50+ ( DSB repair proficient ) diploids . As in pure species , chromosomes of Sk/Sp diploids were largely intact throughout meiotic prophase ( Figure 2E ) . This shows that the breaks that form ( e . g . , Figure 2B ) are efficiently repaired . Together , these results suggest that DSB formation is grossly normal in Sk/Sp hybrids and that the breaks are repaired efficiently . Meiotic recombination promotes fertility but is not essential for producing viable spores in Sp ( Davis and Smith , 2003 ) . This is largely because random segregation of three chromosome pairs at the first meiotic division would frequently ( ∼25% of the time ) result in a viable chromosome complement: haploid , diploid , or chromosome 3 disome . In addition , Sp , like some other species , has a recombination-independent chromosome segregation system that is partially effective in promoting proper chromosome segregation ( Davis and Smith , 2003 ) . As expected , when we eliminated nearly all meiotic recombination by deleting rec12 ( rec12Δ ) , both Sp/Sp and Sk/Sk rec12Δ mutant diploids had viable spore yields fourfold to fivefold lower than the corresponding rec12+ diploids ( Figure 3 ) . Surprisingly , a similar decrease in fertility was not observed in rec12Δ Sk/Sp hybrid diploids: their viable spore yield was indistinguishable from that of rec12+ hybrid diploids ( Figure 3 ) . This result demonstrates that recombination does not promote Sk/Sp hybrid fertility , unlike in pure species . This could be because Sk/Sp hybrids fail to form crossovers and/or because the recombination events that occur in hybrids hurt fertility just as much as they promote it . 10 . 7554/eLife . 02630 . 016Figure 3 . Recombination does not alter Sk/Sp hybrid fertility . The average rec12+ viable spore yield of each diploid was divided by that of the corresponding rec12Δ mutant . For the pure species diploids , the viable spore yield was significantly lower in the absence of Rec12 ( n ≥ 5 experiments for each genotype; * t test p<0 . 05 ) . The viable spore yield of Sk/Sp hybrids , on the other hand , was not significantly different between rec12+ and rec12Δ Sk/Sp hybrids ( p=0 . 42 ) . This indicates that recombination likely hurts fertility just as much as it promotes fertility in Sk/Sp hybrid meiosis . These data are shown in a different format in Figure 5A . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 016 We next compared recombination frequencies in meioses from either parental pure-species or Sk/Sp hybrid diploids . We found that recombination frequencies in Sk/Sk were lower than those in Sp/Sp in at least two genetic intervals ( Figure 4—figure supplement 1; Young et al . , 2002 ) . We also observed more dramatic decreases in observed recombinant frequencies in several genetic intervals amongst viable spores produced by Sk/Sp hybrids compared to Sp/Sp ( Figure 4A , Figure 4—figure supplements 2 , 3 ) . The magnitude of the decrease was greatest in the lys1-lys7 interval where we observed >40-fold decrease in genetic distance in hybrids compared to Sp/Sp . 10 . 7554/eLife . 02630 . 017Figure 4 . Chromosome rearrangements limit the recovery of recombinant progeny in Sk/Sp hybrids . ( A ) A cartoon illustrating the fold decrease in recombinant frequencies in Sk/Sp spores compared to that in Sp/Sp spores . The detailed recombination data are in Figure 4—figure supplements 2 and 3 . The Sp karyotype is depicted with a grey/black chromosome 1 , blue chromosome 2 , and red chromosome 3 . The backwards black arrow indicates an inversion in Sp relative to the ancestral karyotype . ( B ) Pulsed-field gels separating Sp and Sk chromosomes and Southern blots of the gels probed with DNA from the indicated loci revealed a reciprocal translocation that includes several essential genes including alr2 and SPCP1E11 . 08 ( abbreviated SPCP ) . The EtBr-stained gels are on the left and the Southern blots are on the right in each pair . ( C ) A cartoon summary of the karyotype differences between Sp and Sk . The arrow indicates the location of the inversion in Sp . A few landmark loci are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01710 . 7554/eLife . 02630 . 018Figure 4—figure supplement 1 . Recombination frequencies in Sk . We crossed Sk strains to measure recombination frequencies between the indicated loci . Sk cM were calculated from the data shown using Haldane’s formula and the Sp cM were calculated using the genome average of 0 . 16 cM/kilobase ( Smith , 2009; Young et al . , 2002 ) because crossover frequencies are relatively uniform in Sp and two of the intervals are too large to be measured directly . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01810 . 7554/eLife . 02630 . 019Figure 4—figure supplement 2 . Recombination frequencies in Sk/Sp hybrids are low relative to Sp . We crossed Sk and Sp strains to measure recombination frequencies between the indicated loci . Sk/Sp cM were calculated from the data shown using Haldane's formula and the Sp cM were calculated using the genome average of 0 . 16 cM/ kilobase ( Smith , 2009; Young et al . , 2002 ) because crossover frequencies are fairly uniform in Sp and most of the intervals are too large to be measured directly . The numbers of each parental and recombinant type are listed in Figure 4—figure supplement 3 . NA , not applicable . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 01910 . 7554/eLife . 02630 . 020Figure 4—figure supplement 3 . Sk alleles are underrepresented in the progeny of Sk/Sp hybrids . This shows the numbers of each parental and recombinant class that were used to calculate recombination frequencies in Sk/Sp hybrids . These data are shown in simplified form in Figure 4—figure supplement 2 . We crossed Sk and Sp strains to measure recombination frequencies between the indicated loci . Both parental and recombinant classes illustrate the effects of meiotic drive on the inheritance of each locus . In each cross , Sk-derived alleles are overrepresented . Sk/Sp cM were calculated from the data shown using Haldane’s formula and the Sp cM were calculated using the genome average of 0 . 16 cM/ kilobase ( Smith , 2009; Young et al . , 2002 ) because crossover frequencies are relatively uniform in Sp and most of the intervals are too large to be measured directly . *In the SZY142x180 cross , not all progeny were used to calculate recombination frequencies between lys7 and other markers . This was because the cross contained both lys7− and lys1− alleles . The lys7+ lys1− and lys7− lys1− genotypes could not be distinguished because they have the same phenotype . For that reason , we considered only Lys1+ ( G418 sensitive ) progeny so that we could use the Lys phenotype to genotype lys7 . In all crosses , only spores containing one copy of the assayed chromosome were considered . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02010 . 7554/eLife . 02630 . 021Figure 4—figure supplement 4 . Sp has an inversion on chromosome 1 relative to Sk . ( A ) Scaled schematic illustrating the breakpoints of the inversion shown as dashed vertical lines at positions 2 , 683 , 632 and 4 , 911 , 515 on the Sp chromosome . Several genetic markers , the centromeres ( thick vertical lines ) , and the locations of PCR oligos ( numbered 1–4 ) used to verify the inversion are shown . ( B ) The inversion was verified using PCR: ethidium bromide stained gel of PCR products using the indicated oligos on Sk and Sp DNA templates . Oligo pairs 1+2 and 3+4 produce a band in Sp , but not Sk . Oligo pairs 1+3 and 2+4 produce a band in Sk , but not Sp . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02110 . 7554/eLife . 02630 . 022Figure 4—figure supplement 5 . Sk has a reciprocal translocation between chromosomes 2 and 3 . The translocation likely occurred as homologous recombination between repetitive elements . ( A ) Scaled schematic illustrating the location of the two full-length transposons in Sk where the translocation occured , shown as dashed vertical lines at positions 676 , 281 and 1 , 932 , 034 on the Sp chromosomes . The position on chr2 marks the start of a single LTR also found in Sp . Several genetic markers , two ( of many ) essential genes within the translocated segments , the centromeres ( thick colored vertical lines ) , and the locations of PCR oligos ( numbered 1–4 ) used to verify the translocation are shown . ( B ) The translocation junction was verified using PCR: ethidium bromide stained gel of PCR products using the indicated oligos on Sk and Sp DNA templates . Oligo pairs 1+2 and 3+4 produce a band in Sp , but not Sk . Oligo pairs 1+3 and 2+4 produce a band in Sk , but not Sp . The larger size of the Sk bands corresponds to the presence of a transposon at each location in Sk , but not in Sp . We believe the smaller bands in the Sk lanes are artefacts of PCR amplification across transposons because we see them even when we PCR across known Sp transposons . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 022 Large decreases in recombination could be explained if some DSBs are infrequently repaired as crossovers in hybrids , or if the crossovers they produce generate inviable chromosomes , for instance due to chromosomal rearrangements . To address this latter possibility , we resequenced the Sk genome ( ‘Materials and methods’ ) ( Rhind et al . , 2011 ) , which revealed that a large region of chromosome 1 ( between base pairs 2 , 683 , 632 and 4 , 911 , 515 ) is inverted in the Sp genome , relative to Sk ( Figure 4A , Figure 4—figure supplement 4 ) . This inversion has been previously described ( Brown et al . , 2011; Teresa Avelar et al . , 2013 ) and it occurred in the Sp lineage ( Brown et al . , 2011 ) . Odd numbers of crossovers within this inversion would cause lethal chromosomal rearrangements ( duplications of one arm and deletion of the other ) . This would cause spore inviability and likely explains why rec12+ Sk/Sp hybrids do not have higher fertility than rec12Δ hybrids , since odd numbers of crossovers within the inversion would likely impair viable spore recovery as much as an absence of recombination ( Figure 3 ) . The recombination pattern of markers across chromosome 1 is consistent with this interpretation . The lys1-lys7 interval , in which we observed the greatest reduction in recombinants , is located within the inversion ( Figure 4A , Figure 4—figure supplements 2 , 3 ) . The next highest reduction in recombinant frequencies was between markers flanking the inversion boundary ( Figure 4A , Figure 4—figure supplements 2 , 3 ) . In contrast , recombination frequencies outside the inversion were only ∼twofold to eightfold decreased in hybrids compared Sp/Sp diploids ( Figure 4A , Figure 4—figure supplements 2 , 3 ) . These non-inversion associated differences are not a hybrid-specific defect and may be due to different recombination frequencies in Sk and Sp , perhaps caused by different DSB frequencies or different DSB repair outcomes ( Figure 4—figure supplement 1 ) . Our analyses of recombination also revealed a surprising genetic linkage between leu1 and ade6 in Sk/Sp hybrids , despite the fact that these genes are located on chromosomes 2 and 3 , respectively , in both species ( Figure 4B , Figure 4—figure supplement 2 ) . We speculated that there was a reciprocal translocation between Sk chromosomes 2 and 3 , relative to Sp . If such a translocation included essential genes , it would render gametes with non-parental chromosome combinations of the affected arms inviable; this inviability would also create the semblance of genetic linkage between the two chromosomes . Consistent with this possibility , we found using Southern blot analyses that essential genes ( alr2 and SPCP1E11 . 08 ) had swapped chromosome locations between Sk and Sp ( Figure 4B , C; Kim et al . , 2010 ) . We partially assembled the Sk genome to map the translocation junctions to position 676 , 281 on Sp chromosome 2 and 1 , 932 , 034 on Sp chromosome 3 . We verified the translocation junctions via PCR ( Figure 4—figure supplement 5 ) . Analyses of synteny in two outgroup species ( S . octosporus and S . cryophilus; Rhind et al . , 2011 ) showed that the translocation occurred in the Sk lineage . The translocation appears to have resulted from a crossover between a Tf transposon found in Sk on chromosome 2 ( corresponding to a single Tf transposon LTR in Sp ) and a Tf transposon unique to Sk on chromosome 3 . Thus , we find that chromosomal rearrangements are likely a significant contributor to Sp/Sk hybrid infertility . However , our findings suggested that recombination defects due to chromosome rearrangements are not sufficient to explain the near complete hybrid sterility we see in Sk/Sp hybrids for several reasons . First , rec12Δ Sk/Sp hybrid diploids still have greater than six-fold lower viable spore yield than rec12Δ Sk/Sk or Sp/Sp ( pure species ) diploids ( Figure 5A ) . The chromosome 2-chromosome 3 reciprocal translocation predicts only a two-fold lower viable spore yield because only half of the gametes would inherit an incompatible chromosome combination . Second , if defects in recombination cause errors in chromosome segregation leading to the production of aneuploid and diploid gametes ( e . g . , Davis and Smith , 2003 ) , we would expect that the frequency of aneuploid and diploid spores produced by Sk/Sp hybrids in the absence of meiotic recombination ( rec12Δ ) should be similar to the frequencies observed in rec12Δ pure species controls . In contrast to this expectation , we find that the level of heterozygous aneuploids and diploids in rec12Δ hybrids is still significantly ( at least twofold ) higher than that observed in both rec12Δ Sk/Sk and rec12Δ Sp/Sp controls ( Figure 5A , B; p<0 . 01 for both; G-test ) . This result indicates that something other than recombination defects contributes to the high fraction of diploids and aneuploids , as well as the low viable spore yield . 10 . 7554/eLife . 02630 . 023Figure 5 . Increased aneuploidy amongst viable Sk/Sp gametes is recombination-independent . ( A ) We calculated both viable spore yield ( viable spores/cell ) as well as the fraction of viable spores that are aneuploid or diploid ( ‘Materials and methods’ ) . In the absence of Rec12 , the relative frequencies of aneuploids and diploids are elevated in all cases . However , there is significantly more aneuploidy and diploidy of viable spores produced by rec12Δ Sk/Sp hybrids than by rec12Δ pure species controls . This shows the phenotype is not caused solely by recombination defects . In addition , Sk/Sp diploids do not generate more aneuploids or diploids relative to the number of cells induced to undergo meiosis compared to pure-species controls . Some of these data are presented in a different format in Figures 1H and Figure 3 . ( B ) A bar graph illustrating the fraction of the viable spores produced by the indicated rec12Δ diploids that are aneuploid or diploid ( G-test , n >300 for each ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 023 There were two possibilities to explain the higher recovery of aneuploid or diploid spores generated by Sk/Sp hybrids . First is the possibility that aneuploidy or diploid spores arise more frequently in hybrid meiosis . In contrast , the second possibility is that aneuploids arise at the same frequency in hybrid meiosis but are more likely than haploids to survive hybrid meiosis; haploid gametes produced by Sk/Sp hybrids may die more frequently than aneuploids . To distinguish between these possibilities , we calculated the frequency of aneuploids and diploids produced per meiosis in hybrids and pure species controls by multiplying the viable spore yield ( viable spores produced per cell induced to undergo meiosis ) by the fraction of viable spores that were aneuploid or diploid ( i . e . , aneuploids or diploids per viable spore ) . Our analyses revealed very small differences in the proportion of aneuploids that were generated in hybrids vs pure species meiosis ( Figure 5A ) . We therefore conclude that the higher recovery of aneuploid and diploid spores from Sk/Sp hybrids is a result of lower viability of haploid spores . The lower viability of haploid spores could be partially explained by the incompatibility between Sp and Sk chromosomes 2 and 3 due to the reciprocal translocation in Sk . Aneuploids with Sk and Sp chromosome 3 ( chromosome 3 disomes ) will contain all essential genes , whereas haploids must inherit chromosomes 2 and 3 from the same species to be viable . This scenario , however , predicts only a twofold advantage of aneuploidy , yet rec12Δ Sk/Sp hybrids produce at least threefold more aneuploids than pure species rec12Δ controls . Furthermore , eliminating the chr2-chr3 incompatibility ( see below ) did not mitigate the phenotype . We therefore hypothesize that the viability benefit conferred by aneuploidy is due to disomy or heterozygosity for an allele ( s ) on chromosome 3 . This could , for example , be due to a protective effect exerted by the Sk chromosome 3 within gametes containing Sp chromosome 3 and vice versa ( ‘Discussion’ ) . One explanation for a chromosome rearrangement-independent mechanism of hybrid infertility emerged from our studies following the transmission of alleles from each species' chromosome through meiosis in Sk/Sp hybrids . To focus on cells with only one copy of a given allele , we omitted heterozygous diploids from analyses of all alleles and omitted heterozygous aneuploids when analyzing transmission of alleles on chromosome 3 . Surprisingly , we found that in all cases , the Sk allele was transmitted to more viable spores than the Sp allele ( Figure 6A , B , Figure 6—figure supplement 1 ) . This phenotype was not specific to the markers used , or which species contributed the marked ( mutant ) allele to the Sk/Sp hybrid diploid . In addition , the overrepresentation of Sk alleles from chromosomes 1 and 2 in the viable spores was observed within both haploid and aneuploid spores ( Figure 6—figure supplement 2 ) . This suggests the enhanced transmission of Sk alleles on these chromosomes is independent of the aneuploidy phenotype . 10 . 7554/eLife . 02630 . 024Figure 6 . Alleles on all three Sk chromosomes show drive ( independent of mitochondrial DNA type ) . ( A ) Sk alleles were inherited by significantly more than 50% of the viable spores produced by Sk/Sp hybrids , except ura1 and ura4 ( G-test p<0 . 01; n >100 for each ) . The markers nearest to the meiotic drive loci ( i . e . , those showing the greatest bias towards Sk inheritance ) are shown in boldface . The color scheme is the same as that in Figure 4 . The data underlying these numbers are shown in Figure 4—figure supplement 3 , and Figure 6—figure supplement 1 . ( B ) The Sk alleles of lys1 , his5 and ade6 show significant drive both in the presence and absence of recombination ( **p<0 . 01 , n >300 for lys1 and his5 , n >80 for ade6 ) . The amount of his5 drive is greater in the absence to Rec12 due to enhanced linkage with the driving locus . The data underlying this graph are shown in Figure 6—figure supplement 1 . ( C ) Incompatibilities between the Sk mitochondrial DNA and Sp nuclear genes are not responsible for the drive phenotype because we observed the same drive in rec12Δ Sk/Sp hybrids with either Sk or Sp-derived mitochondrial DNA ( **p<0 . 01 , n >200 for lys1 and his5 , n >50 for ade6 ) . The data underlying this graph are shown in Figure 6—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02410 . 7554/eLife . 02630 . 025Figure 6—figure supplement 1 . Summary of Sk/Sp hybrid and pure species diploid meiotic phenotypes and distribution of alleles in their progeny . We induced Sk/Sp hybrids as well as Sk/Sk and Sp/Sp control diploids to undergo meiosis and measured their fertility using a viable spore yield assay ( VSY ) . All diploids were heterozygous for genetic markers on chromosome 1 ( lys1 ) , chromsome 2 ( his5 ) and chromsome 3 ( ade6 ) . We genotyped the viable spores from each diploid using these markers . In the rec12+ Sk/Sp hybrids , the Sk parent contributed the his5+ , lys1− , and ade6− alleles . In cases where the diploids had co-dominant markers on chromosomes 2 and 3 , we were also able to determine the number of chromosome 3 aneuploids and diploids amongst the viable spores . Parts of these data are also presented in Figure 1H , Figure 3 , Figure 5 , Figure 6 , Figure 8 , and Figure 8—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02510 . 7554/eLife . 02630 . 026Figure 6—figure supplement 2 . Biased transmission favoring Sk alleles on chromosomes 1 and 2 is observed in aneuploid and haploid spores . The Sk/Sp rec12+ data presented in Figure 6B and Figure 6—figure supplement 1 were sorted to display aneuploid and haploid spores seperately . The diploid was heterozygous for genetic markers on chromosome 1 ( lys1 ) , chromsome 2 ( his5 ) and chromsome 3 ( ade6 ) . We genotyped the viable spores of each diploid using these markers . The Sk parent contributed the his5+ and lys1− alleles . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02610 . 7554/eLife . 02630 . 027Figure 6—figure supplement 3 . Distribution of progeny from rec12Δ Sk/Sp hybrid meiosis . rec12Δ Sk/Sp hybrid diploids were sporulated and we genotyped the viable spores using genetic markers that distinguished each chromosome ( n = 589 ) . Chromosomes 2 and 3 contained codominant markers , which allowed us to definitively determine which strains had two copies of these chromosomes . We assumed that all strains that inherited two copies of chromosome 2 and chromosme 3 were diploids , because chromosome 1 aneuploids are inviable in Sp ( Niwa et al . , 2006 ) . We assume the same is likely true in Sk and Sk/Sp hybrids because we very rarely recover marker combinations consistent with aneuploidy for chromosomes 1 or 2 from rec12Δ crosses . We assume that the few such progeny we do recover are rare recombinants ( Davis and Smith , 2003 ) . The plot shows the distribution of the viable spores that inherited each chromosome . Sp chromosomes are designated P1-P3 , Sk chromosomes are designated K1-K3 . This analysis is limited in that it assumes no recombination . However , recombination does occur at a low frequency ( Davis and Smith , 2003 ) . For example , all strains tested that have P2 K3 or K2 P3 ( incompatible combinations ) were recombinant . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 02710 . 7554/eLife . 02630 . 028Figure 6—figure supplement 4 . Meiotic drive in Sk/Sp hybrids is independent of mitochondrial DNA . We induced Sk/Sp hybrids heterozygous for genetic markers on chromosome 1 ( lys1 ) , chromosome 2 ( his5 ) and chromosome 3 ( ade6 ) containing either Sk or Sp mitochondrial ( mt ) DNA to undergo meiosis . We genotyped the viable spores using these markers and indicated the number that inherited each allele . The Sk parent contributed the his5− , lys1+ , and ade6− alleles . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 028 Since yeast meiosis is symmetric ( i . e . , all four meiotic products can be successfully packaged as gametes ) , differential transmission of alleles indicates that a nonrandom subset of the meiotic products must be rendered inviable . Given that we observe over-representation of Sk alleles in the viable spores of Sk/Sp hybrids , this implies that there must be a corresponding death of spores inheriting Sp alleles . This nonrandom death of gametes is therefore an additional cause of Sk/Sp hybrid infertility . For ease of reference , we refer to this phenotype as ‘drive’ of Sk alleles although this is not meiotic drive as originally envisioned to occur in asymmetric ( female ) meiosis ( Sandler and Novitski , 1957 ) . The term has , however , long been accepted to describe genetic elements that bias allele transmission by acting after the meiotic divisions ( Zimmering et al . , 1970 ) . Genetic markers nearer to any driving locus would be expected to show a stronger transmission bias than those that are more distantly linked . This allowed us to coarsely map the locations of the Sk driving loci using data from our hybrid recombination analyses ( Figures 6A , Figure 4—figure supplement 3 ) . On chromosome 1 , the transmission bias of Sk lys1 ( 72% ) was stronger than those of both ura1 ( 55% ) and arg3 ( 58% ) , indicating that the driving locus is closer to lys1 than to ura1 or arg3 . On chromosome 2 , the driving locus is closer to leu1 ( 94% Sk transmission bias ) than to his5 ( 67% ) , his4 , lys4 , or ade8 ( all ∼64% ) . The driving locus on chromosome 3 is closer to ade6 ( 82% ) than to ura4 ( 52% ) . The significant overrepresentation of Sk alleles on all three chromosomes was also observed in spores produced by rec12Δ Sk/Sp hybrids ( p<0 . 01 G-test; Figure 6B , Figure 6—figure supplement 1 ) . However , the transmission of alleles was altered by the absence of recombination . This was expected due to stronger linkage with the driving locus in the absence of recombination . For example , his5 is distantly linked to the chromosome 2 driving locus in rec12+ but appears more closely linked to it in the absence of recombination in rec12Δ ( Figure 6B ) . We considered the alternate possibility that the overrepresentation of Sk alleles might not reflect drive but could instead reflect genic incompatibilities . For example , if Sp gene A was incompatible with Sk gene B , but Sk gene A was compatible with both Sp and Sk alleles of gene B , Sk gene A would be over-represented in the surviving gametes . This scenario predicts that we still should expect to recover haploid spores with all three Sk chromosomes at the same frequency as haploids with all three Sp chromosomes , since pure species chromosome combinations are compatible . However , rec12Δ Sk/Sp meiosis yielded the pure Sk haploid genotype approximately five times more frequently ( 6 . 5% of viable spores ) than the pure Sp haploid genotype ( 1 . 4%; Figure 6—figure supplement 3 ) . In addition , we found that all haploid combinations of Sk and Sp chromosomes ( except those with non-parental combinations of chromosomes 2 and 3 ) are viable and have similar growth rates , suggesting that they do not suffer from a mitotic fitness loss . For example , cells that inherit Sp chromosome 1 ( P1 ) and Sk chromosomes 2 and 3 ( K2 and K3 ) grow very similarly to Sp cells . The reciprocal is also true ( Figure 7 ) . Although we cannot generate K2 P3 or P2 K3 haploids , rare recombinant strains that combine portions of K2 with portions of P3 ( and vice versa ) also grow well . Despite this , chromosome combinations containing Sp chromosomes are dramatically underrepresented amongst the viable spores of rec12Δ Sk/Sp hybrids ( Figure 6—figure supplement 3 ) . The overrepresentation of Sk alleles amongst the viable progeny therefore cannot be explained by ( BDM ) incompatibilities between nuclear Sk and Sp genes . Instead , our results are consistent with the action of meiotic drive alleles on Sk chromosomes that specifically target Sp chromosomes during or following hybrid meiosis . We have not , however , eliminated the formal possibility that genic incompatibilities exist and contribute to spore death in rec12+ crosses , which do undergo meiotic recombination potentially exposing intrachromosomal incompatibilities . 10 . 7554/eLife . 02630 . 029Figure 7 . The haploid progeny of Sk/Sp hybrids have similar growth rates . The progeny of rec12Δ Sk/Sp hybrids with the indicated chromosomes were diluted and grown on rich YEA medium . KKK indicates the Sk parental genotype , whereas PPP indicates the Sp parental genotype . The strains were genotyped using lys1 , his5 and ade6 alleles on chromosomes 1 , 2 , and 3 , respectively . Strains that inherit intact chromosomes 2 and 3 from different species are non-viable because they lack many essential genes . However , we do rarely recover viable recombinant strains that have alleles from Sk chromosome 2 and Sp chromosome 3 ( and vice versa ) . Potentially recombinant chromosomes are denoted with an * . All haploid strains recovered have growth rates similar to that of the parental species , suggesting mitotic growth defects do not underlie differential recovery of the genotypes . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 029 Mitochondrial-nuclear incompatibilities are one cause of hybrid sterility in budding yeasts ( Lee et al . , 2008 ) . Unlike in budding yeast , functional mitochondrial DNA ( mtDNA ) is essential for growth in wild-type fission yeast ( Wolf and Del Giudice , 1980 ) . Because of this , mitochondrial-nuclear gene incompatibilities could be lethal . We therefore tested the hypothesis that incompatibilities between the Sk mitochondrial genome and Sp nuclear genes could also contribute to the meiotic drive phenotype . We did this by assaying the transmission of alleles through hybrids containing either Sp or Sk mtDNA ( obtained after prolonged growth of Sk/Sp hybrid diploids ) . We genotyped the mtDNA using two distantly spaced single-nucleotide polymorphisms that alter restriction sites . We found that drive of Sk alleles was not significantly different between rec12Δ hybrid diploids with Sk or Sp mtDNA ( G-test; Figure 6C , Figure 6—figure supplement 4 ) . This indicates that Sk drive is not caused by an incompatibility between mitochondrial and nuclear genomes . Given the observed reciprocal incompatibility between Sk and Sp chromosomes 2 and 3 due to the translocation , we reasoned that the observed drive phenotypes of Sk alleles on these chromosomes could be interdependent . This scenario is easiest to understand in the rec12Δ crosses , which are not complicated by recombination . For example , if allele ( s ) on K2 drive , K3 may appear to drive just because spores that inherit K2 and P3 are inviable . The converse could also be true . This line of thinking inspired us to test whether each Sk chromosome can drive independently , in the absence of drive on the other two chromosomes . We used rec12Δ strains to eliminate recombination-associated phenotypes . To test for the ability of K1 to drive autonomously , we crossed a P1 K2 K3 strain ( obtained from a rec12Δ hybrid cross ) to a naïve K1 K2 K3 strain to generate a P1 K2 K3/K1 K2 K3 diploid . We found that even in these diploids , which were heterozygous for only chromosome 1 , the K1 chromosome showed the drive phenotype ( G-test p<0 . 01 Figure 8A , Figure 8—figure supplement 1 ) , proving that the driving allele on K1 can act even in the absence of K2 and K3 drive . 10 . 7554/eLife . 02630 . 030Figure 8 . Sk drive alleles are autonomous and contribute to hybrid infertility . Aneuploidy is largely caused by heterozygosity of Sk and Sp DNA on chromosome 3 . ( A ) Comparison of meiotic drive phenotypes between rec12Δ diploids generated by mating Sk to Sp or to haploid strains obtained from Sk/Sp hybrids . ‘R’ indicates a recombinant chromosome ( Figure 8—figure supplement 1 ) , which is compatible with all Sk chromosomes but does not contain a meiotic drive allele . All Sk chromosomes can drive autonomously ( ** indicates drive; G-test p<0 . 01 ) . However , the drive of Sk chr2 is lower in the KRK/KKK diploid than in pure PPP/KKK hybrids ( G-test; n >500 for chromosomes 1 and 2 , n >80 for chromosome 3 in each cross ) . The PPP/KKK data are also shown in Figure 6B . ( B ) Fertility defects of hybrids parallel the amount of drive observed amongst the viable spores ( see A , p-values obtained from t-tests , averages of at least five experiments are shown ) . This is consistent with drive causing spore death . ( C ) The high aneuploidy amongst the viable progeny of Sk/Sp hybrids is largely due to heterozygosity of one or more loci on chromosome 3 ( G-test , n >500 for each cross ) . The PPP/KKK and KKK/KKK viable spore yield and aneuploid data are also shown in Figure 5 . The data underlying these graphs are summarized in Figure 8—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 03010 . 7554/eLife . 02630 . 031Figure 8—figure supplement 1 . Genotype of recombinant strains used in Figure 8 . We sequenced these two genomes to determine which parent ( Sk or Sp ) contributed which components of the genome using single nucleotide polymorphisms ( SNPs ) to distinguish them . Both strains have the Sk karyotype . The black/grey indicates DNA that is located on chromosome 1 in Sp , blue indicates DNA that is located on chromosome 2 in Sp and red indicates DNA that is located on chromosome 3 in Sp . ( A ) Strain KRK has only Sk alleles on chromosomes 1 and 3 . Chromosome 2 in this strain was generated by a crossover between Sk and Sp between SNPs 2 , 113 , 023 and 2 , 115 , 195 ( which flank the mat1 locus ) . ( B ) Strain KKR has only Sk alleles on chromosome 1 . On chromosome 2 only the left-most tip has Sp alleles . This chromosome was generated by a recombination event between Sk chromosome 2 and Sp chromosome 3 . The crossover occurred between SNPs at positions 2 , 290 , 808 and 2 , 292 , 393 on Sp chromosome 3 . Chromosome 3 in this strain was generated by a recombination event between Sk chromosome 3 and Sp chromosome 3 that occurred between SNPs located at positions 1 , 558 , 635 and 1 , 559 , 674 on Sp chromosome 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 03110 . 7554/eLife . 02630 . 032Figure 8—figure supplement 2 . Summary of meiotic phenotypes for Sk/Sp hybrids and diploids with one heterozygous chromosome and the distribution of alleles in their viable progeny . Detailed data underlying the plots in Figure 8 . We induced Sk/Sp hybrids as well as diploids with only one heterozygous chromosome to undergo meiosis and measured their fertility using a viable spore yeild assay ( VSY ) . All diploids were heterozygous for genetic markers on chromosome 1 ( lys1 ) , chromosome 2 ( his5 ) and chromosome 3 ( ade6 ) . Sk chromosomes in the crosses to generate the diploids ( top ) are shown as ‘K’s whereas Sp chromosomes are shown as ‘P’s . The exact genotypes of the strains containing ‘R’ chromosomes can be found in Figure 8—figure supplement 1 . We genotyped the viable spore progeny of each diploid . The alleles shown in bold were contributed by the pure Sk ( KKK ) parent . We used co-dominant markers on chromosomes 2 and 3 to determine the number of chromosome 3 aneuploids and diploids amongst the viable spore progeny . Some of the PPPxKKK data are also shown in Figure 5 , Figure 6B , and Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02630 . 032 Testing the ability of K2 and K3 to drive autonomously was complicated by the reciprocal translocation , as we could not construct K1 P2 K3 and K1 K2 P3 strains because they would lack some essential genes . To get around this , we took advantage of the fact that rare Rec12-independent recombination can occur during mitosis or meiosis , for example at the mating type locus on chromosome 2 ( Davis and Smith , 2003; Angehrn and Gutz , 1968 ) . Because of this rare Rec12-independent recombination , we recovered some viable spores ( <1% ) from rec12Δ Sk/Sp hybrid meioses that appeared to have non-parental combinations of chromosomes 2 and 3 , based on the phenotypic markers we used . Upon further examination via PCR or whole genome sequencing , these all proved to be rare recombinants that contain the left end of chromosome 2 and the right end of chromosome 3 from the same parent . We fully sequenced two of these recombinant strains to map their recombination breakpoints ( Figure 8—figure supplement 1 ) . These two strains both contained recombinant chromosomes that have the Sk karyotype , but showed the Sp phenotype in crosses to Sk ( i . e . , they were underrepresented in the viable gametes ) . We denote these as ‘R’ ( for ‘recombinant’ ) chromosomes . By crossing strains containing these ‘R’ chromosomes to naïve Sk , we found that K2 and K3 chromosomes are also able to drive autonomously ( Figure 8A , Figure 8—figure supplement 2 ) . Specifically , in the K1 R2 K3/K1 K2 K3 diploid , only K2 showed meiotic drive . Similarly , only K3 exhibited drive in the K1 K2 R3/K1 K2 K3 diploid . Our experiments showed that both K2 and K3 have the ability to drive autonomously ( Figure 8A ) . Intruigingly , these diploids also revealed that K2 meiotic drive was less intense in the absence of linked K3 drive ( 61% instead of 93%; p<0 . 01 ) . In contrast , K3 drives more strongly in the absence of a linked driving K2 ( 91% vs 86%; p=0 . 02 ) . These changes in drive intensity could be affected by genetic interactions between chromosomes revealed by the specific genotypes of these strains ( Figure 8—figure supplement 1 ) . However , they could also be due to removing the reciprocal translocation . In pure Sk/Sp diploids , gametes that inherit Sk chromosome 3 would also have to inherit Sk chromosome 2 to survive . In this way , the driving ability of a weaker meiotic drive allele ( e . g . , on K2 ) could be augmented via pseudo-linkage to a stronger drive allele ( on K3 ) . If the drive of Sk alleles we observe in Sk/Sp hybrids is caused by nonrandom death of spores that inherit Sp alleles , then the magnitude of drive we observe should correlate with level of infertility we observe . To test this , we calculated viable spore yields for a series of diploids with varying amounts of meiotic drive ( Figure 8A ) . Consistent with our hypothesis , we observed that the fertility of diploids was correlated with the magnitude of drive . For example , the P1 K2 K3/K1 K2 K3 and the K1 R2 K3/K1 K2 K3 diploids had the lowest amount of drive and the highest viable spore yield ( Figure 8A , B , Figure 8—figure supplement 1 ) . In contrast , the K1 K2 R3/K1 K2 K3 diploid had the most drive and the lowest viable spore yield . These results are consistent with nonrandom death of gametes being directly associated with the over-representation of Sk alleles . These diploids containing one heterozygous chromosome also fortuitously allowed us to home in on the cause of the high frequency of aneuploids we observed amongst the viable spores of Sk/Sp hybrids . We predicted the phenotype was due to heterozygosity for Sk and Sp alleles on one or more chromosomes , because the phenotype is specific to Sk/Sp hybrids . Amongst the viable spores produced by both the P1 K2 K3/K1 K2 K3 and K1 R2 K3/K1 K2 K3 diploids , the frequency of aneuploid spores was similar to that of the pure species rec12Δ Sk/Sk control ( ≤30%; Figure 8C ) . This result further supports that the drive phenotype of Sk chromosomes 1 and 2 is independent of the aneuploidy phenotype of Sk/Sp hybrids , but it also shows that heterozygosity of these chromosomes does not cause the aneuploidy phenotype . Conversely , a much higher fraction ( 68% ) of the viable spores produced by the K1 K2 R3/K1 K2 K3 diploid , which is heterozygous for most of chromosome 3 , were aneuploid ( Figure 8—figure supplement 1 ) . The percentage of the viable spores from this diploid that were aneuploid was only marginally less than the value we observed in the spores produced by the completely heterozygous Sk/Sp hybrid rec12Δ diploids ( 78%; p=0 . 038; Figure 8C , Figure 8—figure supplement 1 ) . These results show that the high frequency of aneuploids we observe amongst the viable spores of Sk/Sp hybrids is largely due to heterozygosity for a locus ( or loci ) on chromosome 3 .
Our study reveals that the largest contributors to Sk/Sp hybrid infertility are two chromosomal rearrangements involving all three chromosomes . First , Sp chromosome 1 contains a large ( ∼2 . 2 Mb , or >350 cM in Sp; about half of the chromosome ) inversion relative to Sk , which retains the ancestral state ( Figure 4 ) . This inversion has been described before and is found in several fission yeast isolates in addition to the common Sp L972 lab strain used in our study ( Brown et al . , 2011; Teresa Avelar et al . , 2013 ) . This inversion contributes to hybrid infertility by limiting the possibility for interhomolog crossovers . Odd numbers of crossovers within this region would generate incomplete genomes and hence inviable gametes . There may be additional , smaller inversions present between these two species , which will be elucidated only by complete assembly of the Sk genome . Second , we discovered a reciprocal translocation that occurred in the Sk lineage in which DNA that is found on the left end of P2 is found on K3 , and DNA that is found on the right end of P3 is found on K2 ( Figure 4 ) . This translocation contributes to Sk/Sp hybrid infertility because each translocated sequence includes essential genes , rendering K2 P3 and P2 K3 chromosome combinations inviable . Although this translocation has been observed only in Sk thus far , translocations are remarkably common amongst fission yeast isolates closely related to Sp ( Brown et al . , 2011; Teresa Avelar et al . , 2013 ) . Despite very little DNA sequence divergence ( ∼0 . 5% ) and little phenotypic divergence , there have been 12 translocations identified in fission yeast isolates closely related to Sp ( Brown et al . , 2011; Rhind et al . , 2011; Teresa Avelar et al . , 2013 ) . These isolates have all been termed S . pombe for their phenotypic similarities and the low nucleotide divergence between them , despite the fact that many of these isolates might be significantly reproductively isolated from Sp , just like S . kambucha ( Brown et al . , 2011; Teresa Avelar et al . , 2013 ) . Like the Sk/Sp hybrids assayed in this work , genomic rearrangements are likely the largest contributors to hybrid infertility within other closely related fission yeasts . This high number of chromosomal rearrangements is remarkable , especially when compared to the Saccharomyces ‘sensu stricto’ group of budding yeast species , in which only 10 translocations have been identified despite 50-fold higher DNA sequence divergence ( Fischer et al . , 2000 ) . Two types of loci can achieve transmission distortion in their own favor during meiosis . The first are 'true drive' alleles in which the drive occurs due to nonrandom segregation of alleles during the meiotic divisions in asymmetric ( female ) meiosis . For example , 'knobs' on maize chromosomes can drive by preferentially segregating into the nucleus that will become the oocyte , rather than into nuclei that become polar bodies and are lost ( Dawe and Cande , 1996 ) . Similarly , centromeres in Mimulus species can drive in female meiosis ( Fishman and Saunders , 2008 ) . Alternatively , transmission distortion can be achieved not by nonrandom segregation but instead by ‘gamete killing’ in symmetric ( male ) meiosis . Gamete killer alleles segregate into gametes randomly , but the trans-acting killer alleles cause the death or malfunction of the gametes that do not inherit them . Although the mechanisms vary , this type of drive occurs in widely diverged eukaryotes . For example , male mice heterozygous for the t-haplotype produce mostly ( >95% ) functional sperm containing the t-haplotype . The sperm that do not inherit the t-haplotype are nonfunctional due to immobility ( Schimenti , 2000 ) . Similarly , in Drosophila , sperm with chromosomes carrying large amounts of the Responder DNA satellite suffer chromosome condensation delays in the presence of the Segregation Distorter haplotype and do not contribute equally to functional sperm ( Larracuente and Presgraves , 2012 ) . In Neurospora , a spore killer allele kills gametes that do not inherit it by interfering with spore development ( Hammond et al . , 2012 ) . We discovered meiotic drive loci on each of the three Sk chromosomes ( Figures 6 , 8 ) . We do not yet know the mechanism by which Sk alleles gain their transmission advantage or at what point in gametogenesis it occurs . Drive likely occurs , however , via trans-acting gamete killing , as observed in all other cases of drive in symmetric meiosis , not via nonrandom chromosome segregation . We hypothesize that a ‘poison’ produced by each driving allele during hybrid sporulation renders spores that inherit the Sp allele at that locus less viable . For instance , such a ‘poison’ may prevent hybrid spores that inherit the Sp allele from expressing an essential gene for completing sporulation or germination . Precise mapping of the driving alleles will help reveal their mechanisms of action . Genetically separating the action of each driver ( Figure 8 ) allows us to conclude that they would each meet the criteria for meiotic drive . However , simultaneous action of all three drivers exposed in the Sk/Sp cross , together with the chromosomal rearrangements , leads to a dramatic loss of viable spore yield and would curb the fixation of the driver alleles . This is why we hypothesize that each of the three drivers that favor the transmission of Sk alleles likely arose independently or sequentially along the Sk lineage . Natural selection works best when heterozygous alleles are transmitted at Mendelian ( 1:1 ) frequencies through heterozygotes ( Crow , 1991 ) . In this way , the alleles compete on an equal playing field and spread through a population by promoting fitness . By subverting Mendelian transmission , meiotic drive alleles and linked loci can spread and even drive more fit alleles to extinction . The meiotic drive we observed could therefore also be detrimental to fitness indirectly by biasing allele transmission . In this scenario , natural selection would therefore favor the evolution of drive suppressors that are unlinked to the driving loci ( Crow , 1991 ) . Meiotic drive loci that could evade such suppression and re-establish their transmission advantage would then be favored . This genetic conflict between meiotic drive loci and their suppressors would set up an evolutionary arms race where both sides must innovate to try to gain an advantage . We predict that such an arms race is occurring within fission yeasts and we are able to observe drive only in the hybrids because the drive suppressors in Sp are incompatible with the driving alleles in Sk . Under this scenario , it is completely possible that in a cross between Sp and another Sp-like isolate , the Sp drive alleles may emerge victorious . Our study of Sk/Sp hybrids also revealed a previously unanticipated feature of a meiotic drive system . The reciprocal translocation between Sk chromosomes 2 and 3 genetically links the two chromosomes . For example , leu1 and ade6 are genetically linked in Sk/Sp hybrids despite being on chromosomes 2 and 3 , respectively ( Figure 4B , Figure 4—figure supplement 2 ) . leu1 and ade6 are also both closely linked to the drive alleles on their chromosomes; thus , the translocation also links the drive alleles on separate chromosomes ( Figure 6A ) . We demonstrated that each of the drive alleles can act in the absence of drive of the other two chromosomes ( Figure 8 ) . In an exciting twist for a meiotic drive system , however , the drive of the weaker allele may be bolstered by the drive of the stronger allele . Such a two-chromosome , two-locus drive system could be a formidable selfish evolutionary force . Like an inversion , the translocation expands the region effectively linked to the driving loci , resulting in a larger proportion of the genome favoring drive enhancers and a smaller proportion favoring suppressors . In addition , if a suppressor of one of the drive alleles does arise in the population , the drive of the other allele can compensate and minimal suppression will actually occur . In other words , both drive alleles must be suppressed simultaneously to restore Mendelian allele transmission . Finding independent drivers on all three Sk chromosomes still does not explain our findings of increased frequencies of aneuploid ( and diploid ) gametes amongst the viable spores of Sk/Sp hybrids compared to the frequencies observed in pure species diploids , even in the absence of recombination ( Figure 5 ) . Our crosses ( Figure 8 , Figure 8—figure supplement 2 ) indicate that the aneuploidy phenotype is caused by Sk/Sp heterozygosity of an allele or alleles on chromosome 3 . The aneuploidy phenotype seems paradoxical given that aneuploidy is generally deleterious ( Niwa et al . , 2006 ) . Indeed , even the aneuploids produced by Sk/Sp hybrids show a slower growth phenotype ( Figure 1F ) . One hypothesis that would explain this higher recovery of aneuploid and diploid spores is that the disjunction of chromosomes ( especially the Sk and Sp third chromosomes ) is compromised during hybrid meiosis . Contrary to this , we calculate that hybrid meioses do not generate aneuploid or diploid gametes more frequently ( Figure 5A ) . Rather , we hypothesize that aneuploid and diploid spores produced by Sk/Sp hybrids are more likely than haploid spores to survive to produce a colony . Fitness advantages for aneuploids have been observed under some conditions and certain genetic backgrounds have increased aneuploidy tolerance ( Ni et al . , 2013; Pavelka et al . , 2010; Yona et al . , 2012; Torres et al . , 2010; Tange et al . , 2012 ) . For example , aneuploidy could be beneficial to fitness by increasing the dosage of certain genes ( Torres et al . , 2007 ) . However , we observed more than sevenfold more K1 K2 K3/P3 aneuploids ( 289 ) than K1 K2 K3 ( 38 ) haploids amongst the viable spores produced by rec12Δ Sk/Sp hybrids . K1 K2 K3 haploids represent the pure species genotype ( and Sk/Sk diploids have high fertility ) , so it is hard to envision how adding P3 will provide a more than a sevenfold fitness boost , especially given the growth defects of aneuploids . We favor an alternative model in which the aneuploidy and meiotic drive phenotypes are interconnected . We propose that Sk/Sp aneuploids are more resistant to poisons produced by meiotic drive alleles acting in hybrids . Our model posits that the Sp third chromosome encodes a weaker driver ( poison ) that acts against Sk chromosome 3 , whereas the stronger Sk driver ( poison ) acts against Sp chromosome 3 . In this model , only cells containing both Sk and Sp chromosome 3 are fully resistant to the effects of both drivers . Because aneuploids are genetically unstable and frequently become haploid upon mitotic growth ( Niwa et al . , 2006; Kohli et al . , 1977 ) , the enhanced viability of aneuploid ( and likely diploid ) spores must be manifest during spore formation , spore germination , or the first few mitotic divisions after germination . Testing this model and other alternatives will require identification of the loci responsible for drive . Regardless of their mechanism ( s ) of action , the meiotic drive alleles we identified contribute significantly to hybrid spore death ( Figure 8C ) . Sk alleles are preferentially found in the viable spores , whereas Sp alleles are presumably enriched amongst the dead spores . Why are rearrangements so common in fission yeasts ? Brown et al . ( 2011 ) , who first discovered the karyotype diversity in different S . pombe isolates , suggested that these organisms might live in small populations and rarely outcross . Teresa Avelar et al . then proposed that rearrangements could spread due to fitness advantages provided by the rearrangement during mitotic growth ( Teresa Avelar et al . , 2013; Colson et al . , 2004 ) . In both these scenarios , a chromosomal rearrangement could spread to fixation within a population without ever undergoing meiosis in a heterozygous state . These rearrangements would then be similar to Dobzhansky-Muller incompatibilities in that the derived karyotypes evolved in separate populations and were never tested by selection . Subsequent meioses of heterozygotes comprised of different derived karyotypes would be impaired and lead to reproductive isolation , impeding gene flow between the two populations . In theory , having a sufficiently large number of chromosomal rearrangements should suffice to ensure complete sterility . We find that chromosomal rearrangements alone are insufficient to explain near complete hybrid sterility in a case of incipient speciation in fission yeast . In addition we show that at least three drive loci exist in Sk . We suspect that meiotic drive alleles are common in fission yeasts . Meiotic drive alleles have been proposed to promote karyotype evolution in diverse eukaryotes , and it is possible that drive also played a role in the evolution of the reciprocal translocation we identified in Sk ( White , 1978; Pardo-Manuel de Villena and Sapienza , 2001; Dyer et al . , 2007 ) . As we have pointed out , it is highly unlikely that the three meiotic drive alleles and the two chromosomal rearrangements we have found in the Sk/Sp hybrids occurred simultaneously during evolution , especially given the fertility costs . It is more likely they accumulated in stepwise fashion over time . The severe fertility phenotypes we observe in the current-day hybrids therefore do not entirely recreate the exact fitness costs and benefits of each novel mutation ( rearrangements , drive ) faced during their evolutionary histories , although they serve as effective proxies . Our findings in fission yeast are highly reminiscent of emerging themes from studies of reproductive isolation mechanisms in plants , insects and even mammals , in which genetic conflicts underlie hybrid incompatibilities ( Johnson , 2010; Presgraves , 2010 ) . Here we provide additional support for such models by showing that three genetic conflicts involving selfish meiotic drive alleles directly contribute to hybrid infertility in fission yeasts . Thus , like other sexually reproducing eukaryotes , genetic conflicts could also be fundamental drivers of infertility within fission yeasts . Our study finds important contributions from both chromosomal and genic divergence in mediating hybrid sterility in fission yeast , reminiscent of the chromosomal speciation model first proposed by White ( 1978 ) .
We introduced all DNA transformation cassettes into the genome using a standard lithium acetate transformation protocol . All deletion strains were verified by PCR and phenotype analyses . All strains are listed in Supplemental file 1 . The oligos used in strain construction are listed in Supplemental file 2 . The Sk auxotrophic mutations were all complete deletions of open reading frames . For the auxotrophic mutations , we generated DNA deletion cassettes via PCR using 100 base pair ( bp ) oligos with 79 bp of homology on the 5′ ends to the target gene and 21 bp at the 3′ end homologous to DNA flanking drug resistance genes kanMX6 , natMX4 , and hphMX4 ( Goldstein and McCusker , 1999 ) . The rec12Δ::ura4+ Sk allele was constructed in a manner similar to that for the auxotrophic mutations . We used 100 bp oligos to amplify a ura4+ DNA cassette with 79 bp 5′ and 3′ tails homologous to the DNA upstream and downstream , respectively , of the rec12+ open reading frame . The transformation cassette included the entire ura4+ coding sequence plus 525 bp of upstream sequence and 772 bp of downstream sequence . To construct the natMX4::rec12::His6FLAG2 allele , we first amplified the region from 927–431 bp upstream of the rec12+ open reading frame with PCR oligos that added HindIII sites to either end of the PCR product . We then cloned the PCR product into the HindIII site of plasmid pAG25 , which contains the natMX4 gene ( Goldstein and McCusker , 1999 ) . We next amplified from pJF32 the rec12:: His6FLAG2 allele plus 385 bp of DNA upstream and 173 bp of DNA downstream of rec12 using PCR oligos that added a ClaI site to the 5′ end and an EcoRV site to the 3′ end ( Cromie et al . , 2007 ) . We then cloned that PCR product downstream of the natMX4 gene in the above ( pAG25 derivative ) plasmid to make a plasmid with 5′ homology to DNA upstream of rec12 , followed by natMX4 , then the rec12:: His6FLAG2 allele . We then cut the final plasmid with NotI to release the DNA cassette for transformation . We verified the resulting transformants using PCR and DNA sequencing . To construct the rad50S ( i . e . , natMX4::rad50-K81I ) allele , we first used PCR to amplify the DNA from 394 bp upstream of rad50 to 310 bp into the open reading frame ( Alani et al . , 1990 ) . The oligos added a SmaI site 357 bp upstream of the rad50 translational start site and added the K81I mutation ( changing the codon from AAA to ATT ) . We cloned that PCR product into a pCR2 . 1 TOPO vector ( Invitrogen , Grand Island , NY ) . We then cloned the PvuII-EcoRV fragment from pAG25 that included the natMX4 gene into the SmaI site of the above plasmid ( Goldstein and McCusker , 1999 ) . We then amplified the entire natMX4::rad50-K81I allele from the plasmid using DNA oligos that added an additional 80 bp of DNA to the end of the transformation cassette to promote targeting of the DNA to the proper site in the genome ( replacing the first part of the rad50+ CDS with the cloned rad50-K81I mutant version ) . We verified the mutant strains via PCR , sequencing and phenotypic analyses . The recombinant strains used in Figure 8 , shown as PKK ( to indicate chromosome 1 from Sp , and chromosome 2 and chromosome 3 from Sk ) , KRK ( R indicates a recombinant chromosome ) and KKR , were all isolated from the spores produced by a rec12Δ Sk/Sp diploid in which there is very little meiotic recombination . These are strains SZY147 , SZY240 and SZY302 , respectively , in Supplemental file 1 . We inferred the chromosome composition of these strains using lys1 , his5 , and ade6 alleles and their meiotic phenotypes . The PKK strain contains the Sp allele of lys1 ( on chromosome 1 ) and the Sk alleles of his5 ( on chromosome 2 ) and ade6 ( on chromosome 3 ) . The KRK strain contains the Sp allele of his5 and the Sk alleles of lys1 and ade6 . The KKR strain inherited the Sk alleles of lys1 and his5 , but the Sp allele of ade6 . However , in these last two strains , we reasoned that chromosome 2 , chromosome 3 or both must be recombinant because the strains are viable ( Sp and Sk chromosomes 2 and 3 are incompatible due to the translocation of essential genes ) . For clarity , we denote the third chromosome of KKR ( and the second chromosome of KRK ) as recombinant because they display the Sp meiotic phenotype ( transmission to less than half of the spores ) when crossed to Sk , whereas the others display the Sk meiotic phenotype ( transmission to half of the spores ) when crossed to Sk . We reasoned the KRK and KKR must have the Sk karyotype because all chromosomes were compatible with naïve Sk chromosomes . To test our assumptions about where these stains’ chromosomes were recombinant , we sequenced SZY302 and SZY240 . The contributions of both Sk and Sp DNAs to these strains are summarized in Figure 8—figure supplement 1 . The sequencing data for these hybrids are available in the NCBI sequence read archive ( accession no . PRJNA245039 ) . We used a protocol based on that of Doll et al . ( 2008 ) . We used single diploid colonies ( see below ) to start 10 ml YEL overnight cultures at 30° . The next day we diluted those cultures ( 1:100 or 1:50 ) into 100 ml PM medium and grew the cultures overnight ( ∼16 hr ) at 30° to an OD600 ∼1 . We then pelleted the cells from 50 ml of culture and resuspended them in 500 ml of PM minus nitrogen medium to induce sporulation . The cells sporulated at 30° . To assay meiotic divisions , we sampled cells at the indicated time points , fixed them in formaldehyde and DAPI stained the cells as in Malone et al . ( 2004 ) . We then visualized the cells in a Zeiss Axiovert 200 M microscope . To generate diploids , we first grew YEL cultures of each haploid parent overnight at 30° . We then mixed roughly equal volumes ( <500 μl ) of the cultures in microfuge tubes , pelleted the cells and spread them on SPA medium . We left the SPA plates at room temperature overnight ( ∼16 hr ) to allow the cells to mate . We then scooped up the mated cell mixture and spread it on minimal yeast nitrogen base ( YNB ) medium to select diploids . The haploid parents contained complementing auxotrophic markers , so the minimal medium selected diploids . After 2 days at 32° , we isolated single diploid colonies from the minimal plates and restreaked them to single colonies on fresh minimal medium . After 2 days at 32° , we used single diploid colonies to start small ( 3–4 ml ) YEL cultures that we grew overnight at 30° . The next day , we plated a small volume ( 50–100 μl ) of each culture on SPA to induce the cells to undergo meiosis at 25° . We also determined titers of each culture on YEA plates that we incubated 2 days at 32° . We used these YEA plates for two purposes: ( 1 ) we counted the colonies to assay the concentration of viable cells in the culture , and ( 2 ) we verified that the cultures contained diploids heterozygous for the correct markers via replica plating . After at least 3 days , we isolated spores from the SPA plates as in ( Smith , 2009 ) . We then plated spores on YEA medium and grew the cells at 32° for 5 days . For viable spore yield fertility assays , we counted the number of colonies on the YEA plates and compared that number to the number of cells plated on SPA ( Smith , 2009 ) . In other words , viable spore yield= ( colony forming units isolated from SPA [i . e . , viable spores] ) ( colony forming units plated on SPA ) It is important to note that the units of viable spore yields are not strictly the number of viable spores produced per diploid cell , so the values can be greater than four . We used the viable spore yield assay rather than micromanipulation of spores onto rich medium as our measure of fertility because the hybrids produced a large variety of spore morphologies that we were concerned could introduce unintended bias in spore selection . This variation was not a concern with the viable spore yield assay . We compared the Sk/Sp hybrid viable spore yields to the Sk/Sk values , rather than the Sp/Sp values , because the fold difference between Sk/Sk and the hybrids was similar when assayed via micromanipulation of random spores: ∼24-fold when assayed via micromanipulation and ∼25-fold when assayed via viable spore yield . For the meiotic drive and recombination assays , we picked single colonies to master grids on YEA , incubated them them for 2 days and replica-plated them to supplemented YNB plates and YEA+drug plates to determine genotypes . We identified heterozygous diploid spores using codominant markers on either chromosome 1 or 2 . We identified heterozygous aneuploid spores using codominant markers on chromosome 3 . For example , in our rec12Δ Sk/Sp cross ( SZY201xSZY208 ) we identified heterozygous diploids as His+ nourseothricin-resistant ( his5+/his5Δ::natMX4 ) cells . Similarly , we identified aneuploids as non-diploid Ade+ hygromycin-resistant ( ade6+/ade6Δ::hphMX4 ) cells . We were unable to identify homozygous diploids or aneuploids using our markers and thus counted them amongst the haploids . We suspect such homozygous diploids and aneuploids were rare . Recombination frequencies and genetic distances were calculated as in Smith ( 2009 ) . For the Sp genetic distances , we used the genome average of 0 . 16 cM per kb , as calculated in Young et al . ( 2002 ) . We used the Sp genome average because almost all of the genetic intervals are too large to be measured directly in Sp . For this reason , some of the actual genetic distances may be slightly larger or smaller than the values listed . Similar comparisons between the Sk/Sp hybrid and Sk/Sk recombination frequencies were not possible due to the lack of sufficient Sk genetic and sequence data . However , recombination calculations like those of Young et al . ( 2002 ) have not been carried out in Sk as this required an extensive catalog of selectable genetic markers built up by the Sp research community over many years . We used the liquid sporulation protocol described above to induce meiosis in 1000 ml of Sk/Sk rad50S , rec12-FLAG diploids . Cells were harvested 8 . 5 hr after the induction of meiosis , and DNA covalently linked to Rec12 was purified . DSBs were mapped genome-wide using a Sp tiling array as in Fowler et al . ( 2013 ) . To account for incompatible probes , we compared the Sk dye signals to previously published Sp DSB arrays ( Fowler et al . , 2013 ) and filtered probes that had spuriously low signal in the input channel ( values less than 101 . 5 ) . These probes likely represent genome positions either absent in Sk or that contain sufficient SNPs to interfere with hybridization . Hotspot intensities ( Figure 2A ) were then analyzed using 288 hotspot positions in Sp ( Fowler et al . , 2013 ) and integrating the median normalized Rec12 IP/input ratios from Sp and Sk at each hotspot . Two artificial hotspots not present in Sk but present in the Sp dataset were removed for clarity . Points on the x-axis of Figure 2A are greater than zero by definition , since the hotspots were determined in Sp . The correlation between hotspot intensities was quantified using the Pearson correlation coefficient ( r ) . Raw and processed microarray data have been deposited in the NCBI Gene Expression Omnibus ( accession no . GSE57039 ) . Data analyses used R ( http://www . r-project . org/ ) and Bioconductor ( http://www . bioconductor . org/ ) . We prepared Sk genomic DNA using the Genomic DNA Buffer Set ( Qiagen ) and Genomic-tip 20/G columns ( Qiagen ) . Illumina sequencing libraries were prepared at the Fred Hutchinson Cancer Research Center core facility using Illumina TruSeq kits , and the libraries were sequenced in 100 base pair single-end , as well as 50 base pair paired-end , reads using an Illumina HiSeq 2000 . All the sequence reads have been deposited in the NCBI sequence read archive ( accession no . PRJNA244921 ) . We assembled the single-end reads into contigs using the Velvet de novo assembler using a kmer value of 79 and default values for all other parameters ( Zerbino and Birney , 2008 ) . We used these contigs to screen out contaminant DNA sequences from the paired-end data set ( this library was contaminated with non-yeast DNA ) using Geneious software's ‘map to reference’ function using the contigs from single-end reads as the reference . We considered paired reads from the paired-end data set only when at least one of the two reads mapped to the reference . We then assembled these screened paired-end reads using Velvet default parameters except for kmer = 33 and insert size = 220 . Only contigs longer than 1000 base pairs were analyzed further . We aligned the contigs to the Sp chromosomes using BLAST . We found two contigs that failed to align with one continuous piece of Sp chromosome 1 . Instead , one end of the contigs aligned to one part of the chromosome and the other end aligned to a distant part in the opposite orientation . Both contigs had breakpoints at the same positions . We reasoned these could be due to inverted DNA in Sk relative to Sp and verified this using PCR . We first identified the translocation via Southern blot hybridization ( Figure 4 ) . An additional Southern blot revealed that the translocated DNA did not include ade7 on chromosome 2 ( data not shown ) . We reasoned that we might be able to find the translocation junctions in a more complete Sk genome assembly utilizing long PacBio sequencing reads . PacBio sequencing libraries were prepared and sequenced at the University of Washington Genome Sciences core facility . These reads have been deposited in the NCBI sequence read archive ( accession no . PRJNA244921 ) . PacBio sequencing reads were used to join the contigs from paired-end reads into larger scaffolds using PacBio's SMRT Analysis software . We used AHA for the scaffolding using default parameters . None of the scaffolds showed evidence of the translocation found via Southern blot , which led us to assume the translocation breakpoints were between two contigs that remained unjoined . We then designed PCR oligos near the ends of contigs that did not join in the regions where we predicted the translocation junction might lie ( starting near the chromosome ends ) . PCR using oligos from unjoined contigs should form a product only if the contigs were in fact connected to each other . This strategy was successful ( Figure 4—figure supplement 5 ) . Agarose DNA plugs were prepared and separated on pulsed-field gels as in Hyppa and Smith ( 2009 ) . For the Southern blot of DSB hotspots within the NotI restriction fragment D , we used a probe at the left end of the fragment known as probe c189 ( chromosome 1 base pairs 1 , 024 , 213-1 , 025 , 212 ) ( Cromie et al . , 2007 ) . For the Southern blot of the NotI restriction fragment J , we used probe c139 ( chromosome 1 1 , 025 , 344-1 , 026 , 300 ) ( Cromie et al . , 2007 ) . For the ade6 , leu1 , alr2 , and SPCP1E11 . 08 Southern blots , we used PCR products amplified either from within those genes or nearby using oligos listed in Supplementary file 2 . | It is widely thought that all of the billions of species on Earth are descended from a common ancestor . New species are created via a process called speciation , and nature employs various ‘barriers’ to keep closely related species distinct from one another . One of these barriers is called hybrid sterility . Horses and donkeys , for example , can mate to produce hybrids called mules , but mules cannot produce offspring of their own because they are infertile . Hybrid sterility can occur for a number of reasons . Mules are infertile because they inherit 32 chromosomes from their horse parent , but only 31 chromosomes from their donkey parent—and so have an odd chromosome that they cannot pair-off when they make sperm or egg cells . However , even if a hybrid inherits the same number of chromosomes from each parent , if the chromosomes from the two parents have different structures , the hybrid may still be infertile . Zanders et al . have now looked at two species of fission yeast—S . pombe and S . kambucha—that share 99 . 5% of their DNA sequence . Although hybrids of these two species inherit three chromosomes from each parent , the majority of spores ( the yeast equivalent of sperm ) that these hybrids produce fail to develop into new yeast cells . Zanders et al . identified two causes of this infertility: one of these was chromosomal rearrangement; the other was due to three different sites in the DNA of S . kambucha that interfere with the development of the spores that inherit S . pombe chromosomes . Since these two yeast species are so closely related , the findings of Zanders et al . reveal how quickly multiple barriers to fertility can arise . In addition , these findings provide further support for models in which conflicts between different genes in genomes can drive the process of speciation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2014 | Genome rearrangements and pervasive meiotic drive cause hybrid infertility in fission yeast |
Despite their potential interplay , multiple routes of many disease transmissions are often investigated separately . As a unifying framework for understanding parasite spread through interdependent transmission paths , we present the ‘ecomultiplex’ model , where the multiple transmission paths among a diverse community of interacting hosts are represented as a spatially explicit multiplex network . We adopt this framework for designing and testing potential control strategies for Trypanosoma cruzi spread in two empirical host communities . We show that the ecomultiplex model is an efficient and low data-demanding method to identify which species enhances parasite spread and should thus be a target for control strategies . We also find that the interplay between predator-prey and host-parasite interactions leads to a phenomenon of parasite amplification , in which top predators facilitate T . cruzi spread , offering a mechanistic interpretation of previous empirical findings . Our approach can provide novel insights in understanding and controlling parasite spreading in real-world complex systems .
Zoonoses are infections naturally transmitted between animals and humans , and are the most important cause of emerging and re-emerging diseases in humans ( Perkins et al . , 2005; Jones et al . , 2008; Lloyd-Smith et al . , 2009 ) . The majority of the zoonotic agents are multi-host pathogens or parasites ( Ostfeld and Holt , 2004; Alexander et al . , 2012 ) , whose various host species may differ in their contribution to parasite transmission and persistence over space and time ( Jansen et al . , 2015; Rushmore et al . , 2014 ) . This heterogeneity of host species contribution to parasite transmission is related to differences in host species’ abundance , exposure and susceptibility to infection ( Haydon et al . , 2002; Altizer et al . , 2003; Streicker et al . , 2013 ) . Further , many multi-host parasites have complex life cycles with multiple transmission modes , such as vertical , direct contact , sexual , aerosol , vector-borne and/or food-borne ( Webster et al . , 2017 ) . Among the zoonotic parasites with multiple hosts and transmission modes , Trypanosoma cruzi ( Kinetoplastida: Trypanosomatidae ) , a protozoan parasite which causes Chagas disease in humans , has a complex ecology that challenges transmission modelling and disease control ( Noireau et al . , 2009; Jansen et al . , 2015; Sosa-Estani and Segura , 2015 ) . T . cruzi has already been found in more than 100 mammalian species and its transmission may be mediated by several interdependent mechanisms ( Noireau et al . , 2009; Jansen et al . , 2015; Coura et al . , 2002 ) . For instance , T . cruzi has a contaminative route of transmission that is mediated by several invertebrate vectors ( Triatominae , eng . kissing bug ) that gets infected when blood feeding on infected hosts . Susceptible hosts can get infected after scratching and rubbing the parasite-contaminated defecation matter onto the lesion of the bite of an infected vector ( Kribs-Zaleta , 2006 ) . Transmission may also occur through a trophic route that cascades along the food-web when a susceptible predator feeds on infected vectors or preys ( Noireau et al . , 2009; Jansen et al . , 2015 ) . In general , sylvatic hosts do not suffer mortality from T . cruzi ( Kribs-Zaleta , 2010 ) but the parasite establishes a lifelong infection in almost all of them ( Teixeira et al . , 2011 ) . Chemical insecticides and housing improvement have been the main strategies for controlling Chagas disease in rural and urban areas of Latin America ( Dias and Schofield , 1999 ) . However , these strategies are proving to be inefficient in reducing transmission ( Roque et al . , 2013 ) . This is possibly related to the maintenance and transmission of parasites among local wild mammalian hosts and its association with sylvatic triatomine vectors ( Roque et al . , 2013; Roque et al . , 2008 ) . Therefore , modelling parasite transmission in a way that is explicitly considering the ecology of wildlife transmission , is fundamental to understanding and predicting outbreaks . In this work , we propose to address this challenge through the mathematical framework of multiplex networks ( De Domenico et al . , 2013; Kivela et al . , 2014; Boccaletti et al . , 2014; De Domenico et al . , 2016; Battiston et al . , 2016 ) , which have been successfully applied to epidemiology ( Lima et al . , 2015; De Domenico et al . , 2016; Sanz et al . , 2014 ) and ecology ( Sonia Kéfi et al . , 2015; Kéfi et al . , 2016; Pilosof et al . , 2017; Stella et al . , 2016 ) . Multiplex networks are multi-layer networks in which multi-relational interactions give rise to a collection of network layers so that the same node can engage in different interactions with different neighbours in each layer ( Kivela et al . , 2014; Boccaletti et al . , 2014; De Domenico et al . , 2013 ) . We study the ecology of multi-host parasite spread by multiple routes of transmission and potential control strategies by developing the ‘ecomultiplex’ framework ( short for ecological multiplex framework ) , Figure 1 . This framework is powerful in modelling complex systems such as infectious diseases or parasite transmission in wildlife . Firstly , it allows to account for multiple types of interactions giving rise to parasite transmission with similar or different time scales . Secondly , the ecomultiplex framework uses metabolic theory ( Jetz et al . , 2004 ) for estimating species frequencies , which are known to influence parasite transmission ( McCallum et al . , 2001 ) . Thirdly , by explicitly considering space , the model also allows to investigate the consequences of spatial structure on parasite transmission ( Hudson et al . , 2002 ) . The ecomultiplex framework is general in the sense that it can include many ecological interactions among a diverse set of species in a realistic ecosystem . We apply this ‘ecomultiplex’ formalism to investigating parasite spread in two vector and host communities in Brazil: Canastra ( Rocha et al . , 2013 ) and Pantanal ( Herrera et al . , 2011 ) . We exploit the theoretical framework enriched with empirical data for designing and comparing different wild host immunisation strategies based on: ( i ) taxonomic/morphological features ( e . g . immunising species belonging to the same taxonomic group or with similar body mass ) ; ( ii ) species interaction patterns ( e . g . immunising species feeding on the vector ) ; and ( iii ) species’ epidemiological role ( e . g . immunising species with higher parasite prevalence ) . Multiplex network structure proved to be an efficient measure in predicting species epidemiological role in both ecosystems . More importantly , considering together multiple transmission mechanisms allowed us to identify a parasite amplification role played by some species of top predators that would not be captured when considering the transmission mechanisms separately .
We start our analysis by investigating the layout of ecological interactions in Canastra and Pantanal ( Figure 2a , b ) obtained from animal diets and parasite infection rates ( see also Appendix 1 ) . Multiplex cartography for both Canastra and Pantanal ( Figure 2c , d ) shows that vectors are: ( i ) more connected and ( ii ) distribute their links more equally across the ecomultiplex layers than other species . Hence , vectors can get more easily infected in one layer and spread the parasite on another layer with equal likelihood . Hence , the cartography confirms that vectors facilitate parasite spread through their interactions . The local network structure around vectors in Canastra and Pantanal ( Figure 2a , b ) shows that vector colonies are in the Centre of star-like topologies on both layers . These results confirm that kissing bug vector species are pivotal for parasite spread , promoting it on both the food web and the vectorial layer . Although parasite diffusion can be hampered by removing vector colonies from the environment ( Yamagata and Nakagawa , 2006 ) , these immunisation strategies are not stable as vector reintroduction can happen shortly after elimination ( Funk et al . , 2013 ) . Hence , we focus on immunisation strategies considering vectors’ centrality in the ecomultiplex networks but immunising other species . As expected , immunising species with the highest parasite infection rate ( Hemoculture ) is the best strategy for hampering parasite spread for both Canastra and Pantanal ( Figure 3 ) . This epidemiological strategy slows down parasite spread by almost 30% in Canastra and 26% in Pantanal when the parasite spreads mainly on the food-web layer ( pv=0 . 1 ) ( Figure 3 ) . Immunising species interacting with vectors on the vectorial layer ( an ecomultiplex strategy ) also performs better than random . The difference between the epidemiological and the ecomultiplex strategies is present only at low vector frequencies ( fv=0 . 1 ) in both Canastra ( Figure 3a ) and Pantanal Figure 3c ) but vanishes when fv=0 . 25 and pv>0 . 2 ( Figure 3b , d ) . In Canastra , when 10% of the animal groups are vector colonies ( Figure 3a ) , biological immunisation strategies are equivalent to immunising species at random . The performance of biological immunisation changes dramatically when vector colonies become more frequent ( Figure 3b ) . Immunising large mammals decreases by 12% the global infection time when pv=0 . 1 , suggesting that large mammals do not facilitate parasite transmission in the model . Immunising all the Didelphidae species leads to similar results ( Figure 3b ) . Modest increases in infection time are reported for immunising Cricetidae species when pv=0 . 2 ( Figure 3b ) . Immunising species feeding on the vector ( insectivores ) is equivalent to random immunisation ( sign Test , p-values>0 . 1 ) . In Pantanal , immunising parasitised mammals , parasitised Didelphidae and species with the highest infection rates ( Hemoculture 3 ) are at least two times more effective in slowing down parasite spread compared to other strategies ( Figure 3c , d ) . Contrary to what happens in Canastra , when fv=0 . 1 and the parasite spreads mainly on the food web ( p≤0 . 2 ) , immunising parasitised Didelphidae hampers parasite diffusion more than immunising all parasitised mammals ( sign Test , p-value<0 . 01 ) ( Figure 3c ) . Immunising insectivores or large mammals is equivalent to random immunisation ( Figure 3c ) . Immunising Cricetidae species always performs worse than random immunisation ( Figure 3c , d ) . In Canastra , the strategy Hemoculture three includes also immunising one species of top predator , the Leopardus pardalis ( ocelot ) ( see Appendix 7—figure 1 ) . We compare the performances of Hemoculture three against another immunisation strategy where instead of the ocelot we immunise another top predator , the Chrysocyon brachyurus ( maned wolf ) , which had negative T . cruzi infection rate in this area ( Rocha et al . , 2013 ) . In general , top predators are related to parasite transmission control in natural environments ( Wobeser , 2013 ) so we did not expected differences between different predators . Instead , results from Figure 4 indicate a drastic increase of global infection time when a predator with positive parasite infection rate is immunised . This indicates that in Canastra the Leopardus pardalis has an amplification effect in spreading the parasite ( Figure 3b ) . This phenomenon crucially depends on the ecomultiplex structure , as discussed in the following section .
We introduce ecomultiplex networks as a powerful theoretical framework for modelling transmission of multi-host parasites by multiple routes in species-rich communities . We identify three key points related to the model . Firstly , we show that network structure offers insights on which host species facilitate parasite spread . Secondly , we show that the structure of species interactions can be as useful as epidemiological , taxonomic and morphological traits in controlling parasite transmission . Thirdly , we identify for the first time that the interdependent interactions of top predators affect their functional role in facilitating parasite transmission rather than hampering it . Ecomultiplex strategies always outperform strategies based on species taxonomic groups , which neglect species’ interactions . Further , network structure allows to design immunisation strategies performing as well as strategies considering empirical parasite infection rate , with the advantage of requiring less data . This quantitatively suggests the importance of jointly considering vector-host and predator-prey interactions for understanding T . cruzi transmission in wildlife ( Coura , 2006; Johnson et al . , 2010; Penczykowski et al . , 2016 ) . Although Pantanal and Canastra differ in species composition and their ecological interactions , immunising species exposed to contaminative infection through the vectorial layer proves to be efficient at all vectorial layer importances in both ecosystems . This underlines the importance of vectorial transmission for boosting parasite spread also in the food web . Importantly , the ecomultiplex model allows to quantitatively investigate the interplay between the multiple types of interactions leading to parasite transmission , an element conjectured being crucial for better understanding the ecology of wildlife diseases ( Lafferty et al . , 2008; Dunne et al . , 2013; Funk et al . , 2013 ) . The ecomultiplex model provides insights on how species influence parasite spread . In Pantanal , immunising only Didelphidae with positive parasitaemia ( i . e . infection rates ) slows down parasite spread more than immunising all mammals with positive parasitaemia . This finding agrees with previous studies that identify Didelphidae as important reservoirs for T . cruzi maintenance in natural ecosystems ( Herrera and Urdaneta-Morales , 1992; Noireau et al . , 2009; Coura et al . , 2002 ) . Reservoirs are a system of host populations that are able to maintain the transmission of a given parasite species in space and time ( Haydon et al . , 2002; Jansen et al . , 2015 ) . Being able to identify the functional role of species based on their topological interactions further highlights the powerfulness of the ecomultiplex framework in modelling parasite diffusion . Notice that the above results and the observed mechanism of parasite amplification are robust also to violations of the assumption of metabolic theory as they are present also in null models with animal abundance independent on body mass ( Appendix 8 ) . Model inputs for the ecomultiplex model are partly from published data ( e . g . animal diets ( Herrera et al . , 2011 ) and parasite empirical infection rates ( Herrera et al . , 2011; Rocha et al . , 2013 ) ) . In order to better calibrate and then evaluate model performance , additional observed ecological data should be used . For instance , larger samples for parasite infection rate estimation and , especially , the frequency of animal contacts would both allow for calibration of model parameters such as the SI infection probabilities , which is not feasible with the currently available ecological data . Within food webs , top predators are generally considered playing a regulating role in parasite spread by preying on infected individuals , thus eliminating infection sources for other animals ( Packer et al . , 2003; Hatcher et al . , 2006; Wobeser , 2013; Telleria and Tibayrenc , 2017 ) . Our ecomultiplex framework shows that predators can also facilitate rather than just slow down parasite spread depending on their interactions with vectors . An example is the ocelot in Canastra , which is a generalist predator that feeds on several prey species , including the vectors , and thus has an increased likelihood of becoming infected on the food web ( Figure 4b ) . Once infected , ocelots can transmit the parasite to vectors through vectorial interactions . Since vectors facilitate parasite spread , then the ocelot can indeed amplify parasite diffusion . This is true for every generalist predator getting in contact with the vectors . This phenomenon of parasite amplification , that is , increased parasite transmission mediated by top predators , emerges only when both ecomultiplex layers are considered together . Therefore , this mechanism remarks the importance of unifying ecological and epidemiological approaches for better modelling parasite transmission . Importantly , this amplification mechanism provides a theoretical explanation by which the ocelot relates with the T . cruzi spread , as found in empirical studies ( Rocha et al . , 2013; Rocha , 2006 ) . Parasite amplification by predators may also occur in other systems that show multiple transmission routes including trophic transmission , such as in the Toxoplasma gondii ( Dubey , 2004 ) and Trypanosoma evansi ( Herrera et al . , 2011 ) transmission cycles . Our theoretical model allows to design and test immunisation strategies in real-world ecosystems by relying on specific assumptions . For instance , since animal groups are embedded in space , home ranges need to be specified for them . For the sake of simplicity , in this ecological version of the model we considered only one effective average interaction radius for all species . Considering species-specific empirical data on spatial distribution represents a challenging yet interesting generalisation for future work when additional ecological data becomes available . The exposure to parasite infection in the wildlife is mediated by a network of contacts . Consequently , some species are more exposed to the parasite compared to others even when they have the same transmission probability . We consider the same parasite transmission probability across species in the Susceptible-Infected dynamics . By doing that , we give maximum importance to the structure of ecological interactions rather than to the stochasticity of the contagion process . Further , we avoid arbitrary parameter value definition . In our model , the ( i ) structure of interactions and ( ii ) the different frequencies of animal groups are analogous to considering different transmission rates . We have already showed that these two elements were sufficient for species to display different probabilities of catching the parasite in our previous work ( Stella et al . , 2016 ) . Immunisation strategies confirm this: immunising species that are more exposed to parasites leads to better immunisation performances compared to random immunisation . Considering species-dependent transmission rates as encapsulated in frequencies and links importantly reduces the number of model parameters . We assume that parasite spread is happening at much faster rates compared to other meta-population dynamics ( e . g extinction or migration ) , which are not currently considered in the model . However , including meta-population dynamics would allow to explore important research questions such as: ( i ) the interplay between predation and parasite amplification over top predators influencing parasite spread; ( ii ) the influence of migration on parasite diffusion; ( iii ) how extinction patterns influence parasite spread . Implementing the Markovian analytical approach from Gómez-Gardeñes et al . , 2018 ) in the ecomultiplex model would allow to reach even more realistic representations of real-world ecosystems .
The ‘ecomultiplex’ model describes an ecological community interacting in a spatially explicit ecosystem , see Figure 1 , using the novel framework of multiplex networks . Each layer of the ecomultiplex network represents a different route of parasite transmission: ( i ) food-web interactions and ( ii ) contaminative interactions mediated by vectors . These infection routes give rise to a multiplex network of two layers where nodes represent groups of individuals of a given species , that is animal groups . Links on the food-web layer are directed to predator species and represent predator-prey interactions . Links on the vectorial layer are undirected and represent insect vector blood meals . Distance among animal groups determines possible interactions: only groups sharing a spatial portion of their home range can interact with each other . We fixed the home range of all animal groups as a circle of radius r=0 . 03 over a unitary squared space and studied a total of N=10000 animal groups , please see Appendix 3 and[67] for more details regarding the network construction . The small value of r has been tuned in order to keep the ecomultiplex network connected ( De Domenico et al . , 2013 ) so that the parasite can infect the whole ecomultiplex network . Different community structures may affect parasite transmission dynamics . We used data from two communities that differed in species composition and interactions ( Appendix 1 ) . Predator-prey and vector-host interactions in the ecomultiplex network are based on ecological data related to T . cruzi infection in wild hosts within two different areas: Canastra , a tropical savannah in South-Eastern Brazil ( Rocha et al . , 2013 ) and Pantanal , a vast floodplain in Midwest Brazil ( Herrera et al . , 2011 ) . Both places hold a highly diverse host communities which differ in the structure of interactions , particularly in the vectorial transmission layer ( See Appendix 2 for further details ) . Trophic interactions in the food web are assigned according to literature data about animals’ diets ( AdA et al . , 2002; VdN , 2007; Cavalcanti , 2010; de Melo Amboni MP , 2007; dos Santos , 2012; Reis et al . , 2006; Rocha , 2006 ) ( Appendix 1 ) . Since kissing bugs function as a single ecological unit and previous T . cruzi epidemiological models treat the vectors as a single compartment ( Kribs-Zaleta , 2006; Kribs-Zaleta , 2010 ) , all vector species are grouped as one functional group . Species infection rate is used to estimate the contaminative interactions in the vectorial layer ( Rocha et al . , 2013; Herrera et al . , 2011 ) . Positive parasitological diagnostics for T . cruzi ( hemoculture ) are used as a proxy for connections on the vectorial layer , since only individuals with positive parasitaemia ( i . e . with high parasite loads in their blood ) are able to transmit the parasite ( Jansen et al . , 2015 ) . Body masses of host species represent averages over several available references ( Herrera et al . , 2011; Myers et al . , 2008; Reis et al . , 2006; Bonvicino et al . , 2008; Schofield , 1994 ) . Geographical proximity and biological features regulate link creation in the ecomultiplex model . Species biological features , in particular body masses , regulate the frequency of animal groups ( which are all mammals , except for vectors ) . Previous study Jetz et al . ( 2004 ) showed that the density ni−1 of individuals of the same species i having the species average body mass mi follows the metabolic scaling: ( 1 ) ni−1=β−1Ri−1mi3/4where Ri is the species-specific energy supply rate and β a constant expressing species metabolism . The above equation comes from metabolic theory and can be used for determining the scaling relationship between body mass mi and frequency fi of animal groups ( rather than animal individual ) for species i , depending on the frequency of vector colonies fv ( Appendix 3 ) : ( 2 ) fi= ( 1−fv ) mi−1/4∑j=1 mj−1/4 . As a consequence of metabolic theory , the frequencies of animal groups in our ecomultiplex model scale as a power-law of body mass with exponent −1/4 rather than −3/4 ( which is the scaling exponent for individuals rather than groups ) . We explicitly leave fv as a free parameter of the model in order to investigate the influence of the frequency of vector colonies on parasite spreading . We investigate the structure of a given ecomultiplex network through the concept of multiplex cartography ( Battiston et al . , 2014 ) , see also Appendix 5 for the definition of multiplex cartography . In our case , the cartography describes how individual groups engage into trophic interactions on the ecomultiplex structure by considering: ( i ) the total number K of trophic interactions an animal group is involved in and ( ii ) the ratio U of uniform link distribution across layers , which ranges between 0 ( all the links of a group are focused in one layer ) and 1 ( all links of a node are uniformly distributed across layers ) . The higher K , the more an animal group interacts with other groups . The higher U the more an animal group will engage in feeding and vectorial interactions with the same frequency . The multiplex cartography for Canastra and Pantanal is reported and discussed in Appendix 5 . As explained in the introduction , we focus on parasites causing lifelong infections in wild hosts . Hence , parasite spread is simulated as a Susceptible-Infected ( SI ) process ( Hastings and Gross , 2012 ) : Animal groups are susceptible or infectious . We assume that parasite transmission among animal groups happens considerably faster than both ( i ) group creation or extinction and ( ii ) parasite transmission within groups , so that fixed numbers of hosts and vectors can be considered , as in previous works ( Keeling and Rohani , 2008; Legros and Bonhoeffer , 2016 ) . At each time step , the parasite can spread from an infected group to another one along a connection either in the vectorial ( with probability pv ) or food-web ( with probability 1−pv ) layer . We consider pv as a free parameter called vectorial layer importance , that is the rate at which transmission occurs through the consumption of blood by vectors rather than predator-prey interactions . We assume that all the species have the same probability of getting infected , since the group gets infected if it interacts with an infected group . However , the transmission rate is the outcome of the probability of infection , species groups frequency in the environment and the interactions in the ecomultiplex . We characterise globally the SI dynamics by defining the infection time t∗ as the minimum time necessary for the parasite to reach its maximum spread within the networked ecosystem ( Stella et al . , 2016 ) . The infection starts from a small circle of radius 0 . 03 in the middle of the unitary space infecting all animal groups within that area . Initial conditions are randomised over different simulations . Immunisation strategies provide information on how species influence parasite spread: immunising species that facilitate parasite spreading should increase the global infection time t∗ compared to immunising random species . We focus on immunising only 10% of animal groups in ecomultiplex networks with 10000 nodes , in either high ( fv=0 . 25 ) or low vector frequency scenarios ( fv=0 . 1 ) . By immunising groups at random in ecomultiplex networks with N= 10000 nodes , we identify φ = 1000 as the minimum number of groups/nodes that have to be immunised in order to observe increases in t∗compared to random immunisation with a significance level of 5% . Immunised groups are selected according to three categories of immunisation strategies ( Table 1 ) : We define the infection time increase Δts as the normalised difference between the median infection time ts when ϕ=1000 nodes are immunised according to the strategy s and the median infection time tr when the same number of nodes is immunised uniformly at random among all animal groups , Δts=ts−trtr . Infection times are averages sampled from 500 simulated replicates . Differences are always tested at 95% confidence level . Positive increases imply that the immunisation strategy slowed down the parasite in reaching its maximum spread over the whole ecosystem more than random immunisation . Negative increases imply that random immunisation performs better than the given immunisation strategy in hampering parasite diffusion . Summing up , the ecomultiplex model adopts the following parameters ( Appendix 4 ) : The ecomultiplex model also considers the following ecological data as inputs: As outputs the model produces the dynamics of parasite spreading . The total number Ninf of infected animal groups was found to be constant across different immunisation strategies , vectorial probabilities pv and ecosystems , Ninf=8700±100 or Ninf≈ ( 97±1 ) % of susceptible hosts , in terms of model outputs we focus on the time necessary for the parasite to reach its maximum spread , that is , on the global infection time . | Many infectious diseases are contained within a species , so animals from other species are not at risk of catching them . But some diseases – known as zoonoses – can spread between animals and humans . Zoonoses are often transmitted from one host to the next by insects that feed on both animals and humans . Many well-developed mathematical models exist to understand how infectious diseases are transmitted solely among humans . But modelling how zoonoses spread among all of their hosts is much more difficult . This is because in many cases , the disease can be transmitted in multiple ways – by a contaminated food source or blood-feeding infected insects , or through both wild and domestic animals , complicating the picture further . To identify what control strategies would be more efficient for reducing the transmission of parasites that can infect multiple host species , Stella et al . created a new mathematical model called the ‘ecomultiplex framework’ . This model was used to evaluate the complex transmission of Chagas disease , a tropical disease that can be lethal . It combined both ecology ( the environment of the Chagas disease parasite ) and epidemiology ( the characteristics and progress of the disease ) to model how the parasites spread among wild animals . By simulating a real-life scenario , Stella et al . were able to identify which host species were most affected , and to test which control strategies would be the most effective in a given environment . The model also revealed that some species may reduce the transmission of the parasite , while others might amplify it , depending on how they interact with other mammals or insects . The findings will help guide the public-health management of Chagas disease to control transmission more effectively and reduce disease incidence in humans . Besides Chagas disease , many other life-threatening diseases , such as malaria , Leishmaniasis , plague and Lyme disease , are also zoonoses transmitted by multiple ways . The ecomultiplex framework could be of use to ecologists studying these diseases and developing more effective ways to control them . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"epidemiology",
"and",
"global",
"health"
] | 2018 | Ecological multiplex interactions determine the role of species for parasite spread amplification |
Heterochromatic gene silencing is an important form of gene regulation that usually requires specific histone modifications . A popular model posits that inheritance of modified histones , especially in the form of H3-H4 tetramers , underlies inheritance of heterochromatin . Because H3-H4 tetramers are randomly distributed between daughter chromatids during DNA replication , rare occurrences of asymmetric tetramer inheritance within a heterochromatic domain would have the potential to destabilize heterochromatin . This model makes a prediction that shorter heterochromatic domains would experience unbalanced tetramer inheritance more frequently , and thereby be less stable . In contrast to this prediction , we found that shortening a heterochromatic domain in Saccharomyces had no impact on the strength of silencing nor its heritability . Additionally , we found that replisome mutations that disrupt inheritance of H3-H4 tetramers had only minor effects on heterochromatin stability . These findings suggest that histones carry little or no memory of the heterochromatin state through DNA replication .
A central question in biology is how cells with identical genotypes can exhibit different , heritable phenotypes . By definition , these phenotypes are determined by information that is epigenetic , or ‘above the genome . ’ Just as genetic inheritance requires faithful replication of DNA , epigenetic inheritance requires replication of information that is transmitted to both daughter cells during division . Faithful transmission of epigenetic information is crucial for multiple heterochromatin-based processes such as X-chromosome inactivation in mammals and cold-induced gene silencing in Arabidopsis . In these cases and others , the epigenetic inheritance of heterochromatin indicates that some components of heterochromatin behave as heritable units . Surprisingly , the identity of this epigenetic information remains unclear and heavily debated . The histone subunits of nucleosomes , especially histones H3 and H4 , are marked by a variety of covalent modifications that are integral to heterochromatin function . During DNA replication , nucleosomes are partially disrupted and marked parental H3-H4 tetramers are locally inherited to daughter chromatids . As these tetramers are inherited , they are reassembled into nucleosomes that are interspersed with nucleosomes containing newly synthesized H3-H4 tetramers ( Prior et al . , 1980; Jackson , 1988; Schlissel and Rine , 2019 ) . One model for epigenetic inheritance posits that marked parental histones inherited through DNA replication recruit histone modifiers to deposit similar marks on new adjacent nucleosomes , thereby reestablishing the previous local landscape of histone modifications ( Hecht et al . , 1995; Hoppe et al . , 2002; Gaydos et al . , 2014 ) . In support of this model , the H3K27 methyltransferase PRC2 binds preferentially to H3K27me3 in vitro ( Hansen et al . , 2008 ) and some other modifying enzymes show a similar ability to bind their histone modifications ( Zhang et al . , 2008; Hecht et al . , 1995; Imai et al . , 2000 ) . If this model is correct , modified H3-H4 tetramers would constitute heritable units that drive epigenetic memory of chromatin states . Studies have come to different conclusions regarding whether histones can carry epigenetic memory . In S . pombe , localized methylation of H3K9 can silence a reporter gene , and this silenced state is heritable in the presence of the H3K9 methyltransferase Clr4p as long as the demethylase Epe1p is absent ( Audergon et al . , 2015; Ragunathan et al . , 2015 ) . These studies suggest that histone modifications can facilitate epigenetic inheritance , and caution that such a mechanism is normally obscured by H3K9 demethylation activity . Conversely , induced removal of silencer elements from silenced chromatin in S . cerevisiae causes almost all cells to lose silencing of adjacent genes after just one round of DNA replication ( Holmes and Broach , 1996 ) . Similar results are found when silencers are removed from Drosophila chromatin silenced by the Polycomb complex ( Laprell et al . , 2017 ) . These silencer-removal experiments suggest that modified histones are not sufficient to propagate the silenced chromatin state through DNA replication . The model in which histones carry epigenetic memory makes a testable prediction: since parental H3-H4 tetramers have long been thought to be randomly partitioned between daughter chromatids ( Sogo et al . , 1986; Cusick et al . , 1984 ) , rare events could occur in which most or all marked parental H3-H4 tetramers within a domain segregate asymmetrically to one daughter chromatid , causing the other to inherit primarily newly synthesized histones . A chromatin domain with an insufficient number of marked parental tetramers would be expected to experience a loss-of-chromatin-state event . In this view , a smaller chromatin domain would correspond to fewer marked nucleosomes and yield more frequent events in which parental H3-H4 tetramers segregate asymmetrically and the chromatin state is lost . This potential use of domain size for protection against epimutation is widely conjectured ( Dodd et al . , 2007; Kaufman and Rando , 2010; Moazed , 2011; Ramachandran and Henikoff , 2015 ) , and may explain why chromatin domains subject to stable epigenetic inheritance are often many kilobases long . For example , chromatin domains silenced by Polycomb Responsive Elements ( PREs ) in Drosophila usually extend beyond 10 kb ( Schwartz et al . , 2006 ) . In contrast , one study in A . thaliana found that a chromatin domain containing only three H3K27me3-marked nucleosomes is inherited more frequently than would be predicted if random segregation of tetramers caused loss events ( Yang et al . , 2017 ) . However , no study to our knowledge has systematically tested this prediction . To test directly whether inheritance of a chromatin state is affected by chromatin domain size , we focused on the heterochromatin domains at the HMR and HML loci in S . cerevisiae . These loci contain copies of mating-type genes that are silenced by the activity of Sir proteins . Specifically , the E and I silencers flanking HMR and HML are occupied by the DNA-binding proteins Rap1 , Abf1 , and ORC , that collectively recruit Sir proteins; Sir1 is present only at silencers , whereas Sir2/3/4 complexes bind to silencers and spread across the locus in a process that requires deacetylation of H4K16 ( Rusché et al . , 2002; Thurtle and Rine , 2014 ) . Notably , DNA methylation and RNA interference do not exist in S . cerevisiae . Under normal conditions , HMR and HML are constitutively silenced . Rare and transient loss-of-silencing events can be measured by a sensitive assay that uses the cre recombinase under control of the HMLα2 promoter to convert transient transcriptional events into permanent , heritable changes in fluorescence phenotypes ( Dodson and Rine , 2015 ) . In contrast , deletion of SIR1 causes genetically identical cells to be in either of two states at HMR and HML: either fully silenced or fully expressed ( Pillus and Rine , 1989; Xu et al . , 2006; Dodson and Rine , 2015 ) . These different transcriptional states are mitotically heritable and cells switch between states at a low frequency . This study addresses three questions regarding the inheritance of heterochromatin in Saccharomyces: 1 ) Does the size of a silenced domain determine the fidelity of inheritance ? 2 ) Does removal of Sir1 , a protein that facilitates recruitment of silencing machinery to silencers , uncover an effect of chromatin domain size on heritability of transcriptional states ? 3 ) Do replisome components that facilitate symmetric inheritance of parental H3-H4 tetramers also promote inheritance of transcriptional states ?
To test if nucleosome number affected the stable inheritance of a chromatin state , we used the Cre-Reported Altered States of Heterochromatin ( CRASH ) assay ( Dodson and Rine , 2015 ) ( Figure 1A ) . In this assay , cre replaces the α2 coding sequence in HMRα , and a lox cassette containing fluorescent reporters separated by loxP sites is located on a separate chromosome . Though HMRα is transcriptionally repressed , rare loss-of-silencing events cause transient expression of cre . These events lead to excision of RFP from the lox cassette , and a switch from RFP to GFP expression . Because this change is heritable , loss-of-silencing events during colony growth lead to formation of sectors of cells expressing GFP , appearing green on an otherwise red background . The number of sectors in a colony reflects the frequency at which HMRα transiently loses silencing: more sectors indicate less stable silencing . HMRα::cre contained fourteen well-positioned nucleosomes between the E and I silencers ( Figure 1—figure supplement 1 ) . To change nucleosome number within the locus , we deleted DNA corresponding to different sets of nucleosomes ( Figure 1B ) . Notably , removing DNA corresponding to different combinations of well-positioned nucleosomes allowed us to discern whether any effects on silencing stability were due to nucleosome number or to removal of specific DNA sequences . These deletions did not affect the local positions of the remaining nucleosomes as measured by MNase-Seq ( Figure 1—figure supplement 1 ) . At the limit of models by which nucleosomes transmit memory of transcriptional states , inheritance of a single parental H3-H4 tetramer to a daughter chromatid would be sufficient to template the silenced state . The expected loss-of-silencing rate would thereby reflect the frequency at which a chromatid inherits no marked parental H3-H4 tetramers due to random segregation of these tetramers during replication . For example , considering a hypothetical chromatin domain that has three nucleosomes , one would expect that a given daughter chromatid would have a one-in-eight chance of inheriting no parental tetramers during replication . Therefore , one in eight daughter cells would be expected to lose silencing . This rate would increase exponentially with shorter chromatin domains as the probability of inheriting at least one parental tetramer decreases ( Figure 1C ) . Additionally , if inheritance of two or more parental H3-H4 tetramers was necessary to template the silenced state , the expected loss-of-silencing rate would be even higher . The silencing-loss rate predicted by random segregation of H3-H4 tetramers would be approximately 0 . 006% of cell divisions for full-length HMRα::cre ( Strain N14 ) ( Figure 1D ) . Previous studies demonstrate that this strain loses silencing in approximately 0 . 1% of cell divisions ( Dodson and Rine , 2015 ) . This difference between expected and observed values could be explained by the existence of other processes besides histone inheritance that potentially destabilize silencing and thereby contribute to the overall silencing-loss rate . In contrast to the full size HMRα::cre , the smallest allele of HMRα::cre ( Strain N7 ) would be expected to lose silencing in approximately 1% of cell divisions ( Figure 1D ) . Therefore , if this model were correct , we would expect to see increased sectoring rates in strains with shorter alleles of HMRα::cre . Surprisingly , decreasing nucleosome number at HMRα::cre led to a slight decrease in silencing loss as measured by sector frequency ( Figure 1E ) . To provide an independent measurement of the silencing-loss rate , we also measured fluorescence profiles of single cells . Cells that have recently lost silencing of cre at HMRα contain both RFP and GFP due to GFP expression and the persistence of RFP prior to its degradation and dilution . Using flow cytometry to measure the frequency of cells that contain both RFP and GFP , we confirmed that nucleosome number did not strongly affect silencing-loss rates , and that reduction of nucleosomes might have a slight stabilizing effect on silencing ( Figure 1F , Figure 1—figure supplement 2 ) . Thus , the size of HMRα::cre did not dramatically influence inheritance of the silenced state , in contrast to the expectation from models in which H3-H4 tetramers carry memory of chromatin states through cell divisions . Additionally , we found that changing nucleosome number at HMLα::cre led to only a small increase in silencing loss , and that these effects were not due strictly to domain size ( Figure 1—figure supplements 3–5 ) . Since studies at HMLα are potentially complicated by its proximity to a telomere , which is also bound by Sir proteins , further studies were performed only at HMRα . The silencers flanking HMRα are bound by three different proteins that collaborate to recruit Sir proteins ( Rusche et al . , 2003 ) . One possibility for the apparent insensitivity of silencing inheritance to nucleosome number was that the constant recruitment of Sir proteins to these sites was efficient enough to mask a contribution of histone inheritance to inheritance of chromatin states . In this scenario , silencers would be capable of recruiting enough Sir proteins to keep the locus silenced during DNA replication , regardless of histone segregation patterns . Sir1 binds to silencers , and deletion of SIR1 partially disrupts silencer activity , as measured by defects in silencing establishment and silencing heritability ( Pillus and Rine , 1989; Dodson and Rine , 2015 ) . We therefore tested if parental H3-H4 tetramer inheritance contributed to transmission of the silenced state when silencer-based recruitment of Sir proteins was impaired by the sir1∆ mutation . Within individual cells in a population of sir1∆ cells , HMR is either transcriptionally silenced or fully expressed . These different states are mitotically heritable: a cell in one state usually gives rise to more cells of that state . To observe this epigenetic phenomenon , we placed the GFP coding sequence into HMRα , such that it was expressed under control of the α2 promoter . Silencing was monitored by GFP expression at the single-cell level using fluorescence microscopy and flow cytometry . In comparison to control strains in which HMRα was fully silenced ( SIR+ ) or expressed ( sir4∆ ) , HMRα was silenced in roughly 99% of sir1∆ cells and was expressed in the remaining cells ( Figure 2—figure supplements 1 and 2 ) . We also observed different epigenetic states for HMLα::RFP . We used live-cell imaging to monitor divisions of sir1∆ cells to identify cells in which silencing of HMR was lost , and other cases in which it was gained ( Figure 2A , Video 1 ) . Thus HMRα::GFP could be used to measure the efficiency of epigenetic inheritance in sir1∆ , similarly to previous studies ( Xu et al . , 2006 ) . For simplicity , we named measurements of epigenetic inheritance in sir1∆ as the FLuorescent Analysis of Metastable Expression ( FLAME ) assay , which is commonly implemented by live cell microscopy but is also adapted to flow cytometry as noted in individual experiments . To test the prediction that chromatin domain size affects silencing heritability with the FLAME assay , we removed DNA corresponding to sets of nucleosomes in the HMRα::GFP locus ( Figure 2B , Figure 2—figure supplements 3 and 4 ) . As before , models in which nucleosomes were carriers of epigenetic memory predicted that shorter chromatin domains would have a higher rate of silencing loss ( Figure 2C ) . Using time-lapse fluorescence microscopy to monitor transcriptional states in individual cells and their descendants as they divided , we found that nucleosome number did not affect the frequency of silencing loss ( Figure 2D ) . Because the expressed state is also heritable , with occasional switches to the silenced state , we also asked if the heritability of the expressed state was influenced by the number of nucleosomes in the locus . The frequency of silencing establishment was similar between strains with different numbers of nucleosomes at HMRα::GFP ( Figure 2E ) . Therefore , even in a background with defective silencer activity , chromatin-domain size did not strongly influence silencing dynamics . These findings argued against models in which parental H3-H4 tetramers and their modifications are required for the epigenetic inheritance of gene expression states in Saccharomyces . An orthogonal approach to test the role of histones in carrying epigenetic memory would be to consistently bias parental H3-H4 tetramer inheritance to one daughter chromatid , leaving the other daughter chromatid with fewer parental H3-H4 tetramers . Recent reports demonstrate conserved roles of two replisome components , Dpb3 and Mcm2 , in producing a more symmetric distribution of parental H3-H4 tetramers between the leading and lagging strands . Specifically , dpb3∆ causes biased parental H3-H4 tetramer inheritance to the lagging strand ( Yu et al . , 2018 ) and a set of point mutations in MCM2 ( mcm2-3A ) causes biased parental H3-H4 tetramer inheritance to the leading strand ( Petryk et al . , 2018; Gan et al . , 2018 ) . A complementary study found that local histone H4 inheritance in a small chromatin domain was moderately reduced in both the dpb3∆ and mcm2-3A single mutants , and severely reduced in the dpb3∆ mcm2-3A double mutant ( Schlissel and Rine , 2019 ) . Together , these studies demonstrate that Dpb3 and Mcm2 are necessary for efficient inheritance of parental H3-H4 tetramers to both daughter chromatids during DNA replication . If parental H3-H4 tetramer inheritance contributes to transmission of chromatin states , we would predict more loss-of-silencing events in strains with defects in tetramer inheritance . To test this idea , we measured silencing loss in replisome mutants using the CRASH assay ( Figure 3A ) . The dpb3∆ and mcm2-3A single mutants exhibited higher silencing-loss rates , consistent with previous studies done at HML ( Yu et al . , 2018; Gan et al . , 2018 ) , and the dpb3∆ mcm2-3A double mutant lost silencing more frequently than either single mutant . Similar results were obtained by using flow cytometry to measure silencing-loss rates ( Figure 3B ) . These data were consistent with a model in which inheritance of parental H3-H4 tetramers could contribute to inheritance of the silenced state at HMR . However , the data were also compatible with the possibility that heterochromatin assembled in such mutants was simply unstable for reasons independent of defects in its inheritance . Additionally , since previous studies did not specifically test the effects of Dpb3 and Mcm2 on histone inheritance within heterochromatin , any interpretations of silencing defects operated under the assumption that these replisome components act similarly between heterochromatin and euchromatin . It is possible that parental H3-H4 tetramer inheritance affects both transient loss-of-silencing events , as detected by the CRASH assay , and heritability of epigenetic states . Testing this possibility was important because the currently unidentified epigenetic information that determines expression states in sir1∆ is transmitted locally at HML and HMR , respectively , rather than being transmitted in trans from processes elsewhere in the cell ( Xu et al . , 2006 ) . If parental H3-H4 tetramers were the crucial local factors that transmitted this information , we would predict that disrupted tetramer inheritance would cause more loss-of-silencing events in sir1∆ . To test this possibility , we generated replisome mutant strains in combination with sir1∆ and evaluated the inheritance of transcriptional states using two different FLAME assay measurements: Fluorescence-Activated Cell Sorting ( FACS ) and live-cell microscopy . Populations of dpb3∆ , mcm2-3A , and dpb3∆ mcm2-3A mutants all showed a mix of cells that were silenced or expressed at HMRα::GFP; all three mutant strains also showed a higher frequency of expressed cells than wild type ( Figure 4—figure supplement 1 , Table 1 ) . Because silencing-loss rates and silencing-establishment rates both affect the frequency of cells in which HMR is silenced or expressed , one or both of these rates were presumably different in replisome mutants . To measure these rates , we used FACS to sort cells from each strain into two separate populations of HMR-silenced and HMR-expressed cells , and used flow cytometry to monitor the rates at which these initial sorted populations relaxed back to a mixed population of silenced and expressed cells ( Figure 4A ) . These relaxation rates , and the frequency of silenced cells at equilibrium , were products of competing silencing-loss and silencing-establishment rates . By using these relaxation rates to calculate silencing-loss rates ( Figure 4B , Figure 4—figure supplement 2 ) , we observed that dpb3∆ and mcm2-3A had higher loss-of-silencing rates than wild type ( Figure 4C ) . The dpb3∆ mcm2-3A double mutant had a higher loss rate than the single mutants . Similar loss trends were observed using time-lapse fluorescence microscopy ( Figure 4D ) , albeit with overall higher loss rates than those seen with FACS . Together , these data suggested that faithful inheritance of parental H3-H4 tetramers helped transmit the silenced state of HMR . However , we also noted that the vast majority of silenced cells still faithfully transmitted the silenced state in the replisome mutant backgrounds . We also asked if replisome mutants had differences in the frequency of silencing-establishment events . Curiously , any strain containing dpb3∆ had an increased establishment rate , whereas mcm2-3A had minimal , if any , effects on establishment rate ( Figure 4E–G ) . Additionally , any strain containing dpb3∆ showed elevated levels of HMRα::GFP expression in unsilenced cells , as measured by flow cytometry ( Figure 4—figure supplement 3 ) . Because dpb3∆ cells more readily established silencing , we inferred that the expressed state was less efficiently inherited . Therefore , Dpb3 contributed to the inheritance of the expressed state of HMR as well as to the silenced state . Though the rate of silencing loss increased in replisome mutant backgrounds , the large majority of silenced cells still faithfully transmitted the silenced state through cell divisions . Indeed , though dpb3∆ and mcm2-3A single mutants exhibit asymmetric parental H3-H4 tetramer inheritance ( Yu et al . , 2018; Petryk et al . , 2018 ) , it is likely that this asymmetry is not complete and some parental H3-H4 tetramers are still stochastically transmitted to each daughter chromatid during DNA replication . Similarly , the dpb3∆ mcm2-3A double mutant exhibits residual local inheritance of histone H4 ( Schlissel and Rine , 2019 ) . We reasoned that , if a daughter chromatid consistently inherits fewer parental H3-H4 tetramers and thereby loses the silenced state more frequently , an additional reduction in the size of a chromatin domain would cause that daughter chromatid to inherit even fewer marked parental H3-H4 tetramers and experience loss-of-silencing events even more frequently . Therefore , if parental H3-H4 tetramers carry epigenetic memory , we would expect loci with fewer nucleosomes to exhibit more loss-of-silencing events in replisome mutant backgrounds . To test this idea , we used the FLAME assay on nucleosome-number mutants in dpb3∆ and dpb3∆ mcm2-3A strains ( Figure 5A , Figure 5—figure supplement 1 ) . There was no clear correlation between silencing-loss rates and nucleosome number in these sensitized backgrounds ( Figure 5B ) . Establishment-of-silencing rates were also not strongly affected , though there was a small increase in the establishment rate with fewer nucleosomes in dpb3∆ mcm2-3A ( Figure 5C ) . Therefore , even when parental H3-H4 tetramer inheritance was disrupted and the number of parental H3-H4 tetramers available for inheritance at HMR was decreased , cells faithfully transmitted epigenetic transcriptional states .
Removal of silencers from heterochromatin via induced recombination demonstrates that silencers are necessary for maintenance of the silenced state . Specifically , induced silencer excision from HMR causes rapid loss of silencing in arrested cells ( Cheng and Gartenberg , 2000 ) . Studies at other loci in S . cerevisiae and Drosophila show that removal of silencers permits maintenance of silencing in arrested cells , but causes loss of silencing once the same cells subsequently complete one or two rounds of DNA replication ( Holmes and Broach , 1996; Laprell et al . , 2017 ) . Therefore , the presence of modified histones is not sufficient for silencing maintenance or heritability , depending on the example under consideration . Indeed , given that silencers are constantly recruiting Sir proteins to these loci , any role of H3-H4 tetramers in transmission of epigenetic information might be hard to detect . We considered the possibility that silencer activity masks an underlying contribution of H3-H4 tetramer inheritance to silencing inheritance . However , the weakened silencer activity in sir1∆ mutants did not reveal a sensitivity of silencing inheritance to the size of the silenced domain at HMR . Importantly , epigenetic states of HML and HMR in sir1∆ are a property of the locus rather than the cell , demonstrating that factors that determine these epigenetic states are inherited locally at HML and HMR respectively ( Xu et al . , 2006 ) . Similar studies of an epigenetically-inherited heterochromatin state in Arabidopsis also demonstrate that the relevant epigenetic information is carried in cis ( Berry et al . , 2015 ) . Additionally , epigenetic inheritance of transcriptional states in heterochromatin is commonly accompanied by the ability to switch stochastically between states , a feature that implies the existence of imperfectly heritable epigenetic information . Though modified H3-H4 tetramers could theoretically be cis-acting , imperfectly heritable units of information , our evidence to the contrary suggests that other cis-acting factors determine the epigenetic state of HMR in sir1∆ . Given the importance of silencers in inheritance of the silenced chromatin state , one possibility is that the silencer complex self-templates by cooperative oligomerization of silencing factors , and that stochastic changes in epigenetic states reflect the formation or dissolution of such a silencer complex . Classic studies of chromatin replication indicate that parental H3-H4 tetramers are randomly segregated between daughter chromatids during DNA replication . For example , chromatin replicated in the presence of cycloheximide , which blocks the synthesis of new histones , produces daughter chromatids with roughly half the number of nucleosomes , and these nucleosomes appear randomly dispersed along both daughter chromatids ( Sogo et al . , 1986; Cusick et al . , 1984 ) . Though our experiments built on these classic findings , it is also possible that parental H3-H4 tetramers may not be randomly segregated genome wide , or at HMR in particular . For example , it was possible that heterochromatin contained factors that facilitated alternating inheritance of tetramers between the leading and lagging strands . In this case , even if H3-H4 tetramers were to act as the sole units of epigenetic information , decreasing chromatin domain size might not affect the rate of silencing loss at HMR . If H3-H4 tetramers carry epigenetic information through DNA replication , mutations that reduce tetramer inheritance would be expected to increase the frequency of silencing loss . Studies describe roles of Dpb3 and Mcm2 in heterochromatic silencing at HML ( Yu et al . , 2018; Gan et al . , 2018 ) , and inheritance of epigenetic states at a synthetic telomere ( Iida and Araki , 2004; Foltman et al . , 2013 ) . Using the CRASH and FLAME assays , we found mild but significant increases in HMR silencing-loss rates in both dpb3∆ and mcm2-3A single mutants . Additionally , the dpb3∆ mcm2-3A double mutant exhibited higher silencing-loss rates than either of the single mutants . Together , these effects suggested that reduced tetramer inheritance caused mild defects in silencing heritability . Though deacetylated H4K16 is crucial for silencing , other modifications such as H3K56 acetylation also affect silencing ( Hyland et al . , 2005; Xu et al . , 2007 ) and reduced inheritance of these modifications may hinder their functions . Considering the variety of histone modifications that parental H3-H4 tetramers can carry through DNA replication , it was striking that cells with moderate or severe reductions in inheritance of parental H3-H4 tetramers still exhibited efficient inheritance of the silenced state . Though replisome mutants exhibit defects in parental H3-H4 tetramer inheritance , some tetramers are still transmitted to both daughter chromatids in replisome mutant backgrounds ( Yu et al . , 2018; Gan et al . , 2018; Schlissel and Rine , 2019 ) . Therefore , there are still parental tetramers that are theoretically capable of carrying epigenetic information to both daughter chromatids in the dpb3∆ , mcm2-3A , and dpb3∆ mcm2-3A mutants . Given that all replisome mutants tested showed increased silencing-loss rates , further reduction in the number of parental H3-H4 tetramers available for transmission to daughter chromatids should cause even higher rates of silencing loss . However , we saw no significant effects of HMR size on the silencing-loss rate in replisome mutant backgrounds . Therefore , cells with both reduced parental H3-H4 tetramer inheritance and a reduction in the number of tetramers available for inheritance at HMR exhibited a surprisingly robust ability to transmit the silenced state . These data strongly suggested that inheritance of parental H3-H4 tetramers has little or no impact on epigenetic inheritance of the silenced state of HMR . The expressed state of HMR in sir1∆ cells is formally an epigenetic state: it is heritable through cell divisions and can stochastically switch to the silenced state . One possibility is that the expressed state of HMR depends on the existence of heritable information , similarly to the silenced state . Histone modifications associated with active transcription can be transmitted through DNA replication ( Alabert et al . , 2015; Reverón-Gómez et al . , 2018 ) and multiple transcription factors can bind to the histone modifications they generate ( Jacobson et al . , 2000; Owen et al . , 2000 ) . Therefore , histone modifications may form positive feedback loops with both silencing machinery and transcription factors . Indeed , a model that incorporates these positive feedback loops and parental H3-H4 tetramer inheritance generates robust bistable chromatin states ( Dodd et al . , 2007 ) . This model also predicts that random segregation of parental H3-H4 tetramers would lead to loss-of-chromatin-state events , and that decreasing chromatin domain size would also decrease the heritability of both the expressed and silenced states . However , we found that shorter versions of HMR did not strongly affect inheritance of the expressed state of HMR . Alternatively , if parental H3-H4 tetramers carry memory of the expressed state , mutations that disrupt parental H3-H4 tetramer inheritance would be expected to increase the rate of silencing establishment . Curiously , dpb3∆ exhibited a ~3-fold increase in the rate of silencing establishment and mcm2-3A had no observable effect ( see Figure 4G ) . These data may suggest that parental tetramer inheritance facilitates heritability of the expressed state , though such an explanation could not account for the mcm2-3A phenotype . Alternatively , these data may suggest that inheritance of the expressed state is influenced by a function of Dpb3 that is separate from its role in tetramer inheritance . It is also important to note that dpb3∆ but not mcm2-3A led to elevated levels of GFP expression when HMRα::GFP was fully expressed . This finding is paradoxical , as one would expect elevated transcription to inhibit silencing establishment , rather than facilitate it . However , recruitment of the transcriptional activator Ppr1 to HMR causes both increased transcription in expressed cells and an increased establishment rate in sir1∆ ( Xu et al . , 2006 ) . Together , our results suggested that the fidelity of H3-H4 tetramer inheritance has minimal consequences for heritability of the silenced state and may affect heritability of the expressed state in some contexts . These findings raised doubts regarding the model in which histones are significant carriers of epigenetic memory in S . cerevisiae . As such , future studies that continue to examine histone-based memory models will be complemented by studies on other possible mechanisms of transcriptional memory .
The strains and oligonucleotides used in this study are listed in Supplementary files 1 and 2 , respectively . All strains were derived from the W303 background . CRASH assay strains , which contained HMRα , hmrα2∆::cre , ura3∆::loxP::yEmRFP:tCYC1:KanMX:loxP:yEGFP:tADH1 or hmlα2∆::cre , ura3∆::loxP::yEmRFP:tCYC1:HygMX:loxP:yEGFP:tADH1 were generated as described previously ( Dodson and Rine , 2015 ) . FLAME assay strains were generated with the following approach . To generate hmlα2∆::yEmRFP , a K . lactis URA3 swap was performed to replace the α2 coding sequence with yEmRFP coding sequence . The hmlα2∆::yEmRFP fwd/rev primers were used for integration of yEmRFP in the final step . To generate HMRα , hmrα2∆::yEGFP , a fragment spanning a portion of hmlα2∆::yEGFP was amplified using hmlα2∆::yEGFP fwd/rev primers and swapped into HMRa . To delete DNA corresponding to nucleosomes at HMRα and HMLα , CRISPR/Cas9 was employed as previously described ( Lee et al . , 2015 ) . Each deletion or repair fwd/rev primer set contained two partially overlapping primers that were amplified by PCR prior to use . The HMR-E-proximal sgRNA was used to induce Cas9 cutting between the HMR-E silencer and cre , and N14 to N12 deletion fwd/rev was used to delete DNA corresponding to two nucleosomes in this region . This sgRNA and oligo set was also used to convert sN12 to sN10 in the FLAME strain background . The HMR-I-proximal sgRNA , which cuts between the HMR-I silencer and cre , was used with N14 to N10 deletion fwd/rev ( to convert N14 to N10 , and sN12 to sN8 ) or with N14 to N9 deletion fwd/rev ( to convert N14 to N9 ) . For HMLα , the HML-E-proximal sgRNA was used to induce Cas9 cutting between the HML-I silencer and cre , and used with N22 to N19a deletion fwd/rev ( to convert N22 to N19a ) or N22 to N16a deletion fwd/rev ( to convert N22 to N16a ) . The HML-I-proximal sgRNA was used to induce Cas9 cutting between the HML-E silencer and cre , and was used with N22 to N19b deletion fwd/rev ( to convert N22 to N19b ) or N22 to N16c deletion fwd/rev ( to convert N22 to N16c ) or N22 to N13c deletion fwd/rev ( to convert N22 to N13c ) . Deletions were confirmed by junction primers and sequencing . To generate mutants with combinations of nucleosome set deletions , CRISPR/Cas9 technology was applied ( as described above ) to strains with one nucleosome set deletion already made . To generate dpb3∆ , the DPB3 sgRNA was used with Cas9 to cut within DPB3 and DPB3 deletion fwd/rev was used to delete the coding sequence . To generate mcm2-3A , the MCM2 sgRNA was used with Cas9 to cut 244 bp into the MCM2 coding sequence and mcm2-3A repair fwd/rev was used to generate the appropriate point mutations ( Y79A Y82A Y91A ) . Mutations were confirmed by sequencing . To generate colonies for analysis by the CRASH assay , RFP-expressing cells were diluted and plated at a density of ~10 cells/plate ( CSM-Trp ( Sunrise Science Products , San Diego , CA ) , 1% agar ) . After 5 days of growth , colonies were imaged using a Leica M205 FA fluorescence stereomicroscope ( Leica Camera AG , Wetzlar , Germany ) equipped with a Leica DFC3000G CCD camera , a Leica PLANAPO 0 . 63x objective , ET RFP filter ( Leica 10450224 ) , ET GFP filter ( Leica 10447408 ) , and Leica Application Suite X ( LAS X ) imaging software . At least ten colonies were imaged per genotype . Cells were grown to saturation in CSM ( Sunrise Science Products ) at 30°C overnight . These cells were then back-diluted in 5 ml CSM and grown to mid-log phase over 6 hr . 500 µl was transferred to a microfuge tube and sonicated at 20% for 15 s ( Branson Ultrasonics Digital Sonifier 100-132-888R with Sonicator Tip 101-135-066R ) ( Branson Ultrasonics , Fremont , CA ) to break up clumps of cells . 5 µl of sonicated cells were spotted onto a CSM plate ( 1% agar ) and allowed to soak into the agar . When dry , a sterile spatula was used to cut a 1 cm ×1 cm agar square surrounding the cell patch . The square was lifted out of the plate , inverted , and placed in a 35 mm glass bottom dish ( Thermo Scientific 150682 ) ( Thermo Fisher Scientific , Waltham , MA ) . Cells were imaged using a Zeiss Z1 inverted fluorescence microscope with a Prime 95B sCMOS camera ( Teledyne Photometrics , Tucson , AZ ) , Plan-Apochromat 63x/1 . 40 oil immersion objective ( Zeiss , Oberkochen , Germany ) , filters , MS-2000 XYZ automated stage ( Applied Scientific Instrumentation , Inc , Eugene , OR ) , and Micro-Manager imaging software ( Open Imaging , San Fransisco , CA ) . Given that cells were pressed between the agar and glass , the cells were all in the same focal plane and Z-stacks were not used . For time-lapse microscopy ( i . e . Figure 2D ) , samples were kept at 30°C and humidified with a P-Set 2000 Heated Incubation Insert ( PeCon , Erbach , Germany ) . Time-lapse experiments involved brightfield and fluorescence imaging of 16 different fields per sample , and images were taken every 10 min for 10 hr . Subsequent analysis of cell divisions was done in ImageJ ( NIH , Bethesda , MD ) . To measure epigenetic switching rates in the FLAME assay , cell divisions and switching events were manually counted and the counter was blind to the genotype ( single-blind study ) . This counting was performed only on cells that could be clearly distinguished from each other . If a mother and daughter cell pair switched simultaneously , we counted this as one switching event that probably appeared as two events due to the lag time in yEGFP expression or degradation . To measure fluorescence intensities per cell in the CRASH and FLAME assays , a BD LSR Fortessa cell analyzer ( BD Biosciences , San Jose , CA ) with a FITC filter ( for GFP ) and a PE-TexasRed filter ( for RFP ) was used . Subsequent analysis was performed with FlowJo software . For quantification of silencing-loss rates in the CRASH assay ( Figure 1F; Figure 3B ) , cells were first streaked out to form single colonies . Six colonies per genotype were added to CSM-Trp media ( Sunrise Science Products ) in a 96-well plate ( Corning CLS3788 ) ( Corning Inc , Corning , NY ) and grown to saturation overnight in an incubating microplate shaker ( VWR 12620–930 ) ( VWR International , Radnor , PA ) at 30°C . These samples were then back-diluted and grown to mid-log phase over 6 hr . GFP and RFP expression were then analyzed by flow cytometry ( n > 4000 cells per sample ) . Distinct populations of RFP+ GFP- ( which had not lost silencing ) , RFP+ GFP+ ( which had recently lost silencing ) , and RFP- GFP+ ( which had lost silencing less recently ) were observed . The apparent silencing-loss rate was calculated as the number of RFP+ GFP+ cells divided by the number of RFP+ GFP+ cells and RFP+ GFP- cells . Measurements from independent cultures were considered as biological replicates . For calculating the frequency of silenced and expressed cells at equilibrium in the FLAME assay , cells were first streaked out to generate single colonies . Three colonies per genotype were added to CSM media in a 96-well plate and grown to saturation overnight . These samples were then serially back-diluted in CSM media in 96-well plates and grown at 30°C . After twelve hours , the serial dilutions had a range of cell densities; the dilution that was closest to ~1 O . D . was again back-diluted in CSM media and grown at 30°C for another 12 hr . At this point , wells close to ~1 O . D . contained cells that had been growing at log-phase for approximately 24 hr . These cells were analyzed by flow cytometry . Because three populations were analyzed per genotype , the most representative profiles of silenced and expressed cells were used for figures . We considered these populations as biological replicates . To calculate GFP expression levels in expressed cells in the FLAME assay , cells were streaked out for single colonies and three colonies per genotype were grown overnight in CSM + 5 mM Nicotinamide ( NAM ) ( Sigma-Aldrich , St . Louis , MO ) . These samples were then back-diluted in CSM + 5 mM NAM and grown at 30°C for 12 hr . Samples at ~1 O . D . were analyzed by flow cytometry . For Figure 2—figure supplement 4 , the most representative profiles of the three profiles generated per strain were shown . For Figure 4—figure supplement 3 , the geometric mean intensity of GFP per cell ( excluding cells that formed a smaller , artifactual peak at a lower GFP intensity ) was calculated for each population using FlowJo software . Independent cultures were considered as biological replicates . FACS was utilized in the FLAME assay to calculate switching rates between epigenetic states in Figure 4 . To perform this experiment , cells from each genotype were serially diluted in CSM media and grown at 30°C . After 12 hr , dilutions closest to ~1 O . D . were sorted into GFP- and GFP+ populations using a BD FACSAria Fusion cell sorter ( BD Biosciences ) equipped with a FITC filter for GFP . Gates were calibrated from SIR+ ( JRY11474 ) and sir4∆ ( JRY11496 ) cells . For each sample , 150 , 000 GFP- cells were sorted into one tube and 30 , 000 GFP+ cells were sorted into another . Each sorted population was divided evenly into three populations and grown in CSM in a 96-well plate at 30°C . Serial back-dilutions were used to maintain constant log-phase growth over two days . Time-points were taken by removing a fraction of cells from each population and fixing them in a 4% paraformaldehyde solution ( 4% Paraformaldehyde , 3 . 4% Sucrose ) for 15 min at room temperature . Fixed cells were resuspended in GFP fix buffer ( 100 mM KPO4 pH 7 . 4 , 1 . 2 M Sorbitol ) and kept at 4°C . Once the experiment was complete , fixed cells from different time-points were analyzed by flow cytometry ( n > 500 cells per sample ) and FlowJo software . The percent of GFP+ cells for each sample over time is shown in Figure 4B and E . Because the initial sorting event required ~20 min per sample , the time of initial sorting ( t = 0 hr ) was different between samples; this made the time points between samples slightly staggered as seen in Figure 4B and E . Because cells were divided into subpopulations after the initial sorting , these subpopulations were considered as technical replicates . The following equations were used to model the dynamics of switching rates between epigenetic states in sir1∆ . We considered the balance of GFP+ and GFP- cells over time , and assumed that the birth and death rates of the two populations are similar . Combining the balances and introducing the ratio variable x , we can derive the following equation that describes how a population of GFP+ cells and GFP- cells would move towards equilibrium over time:1kON+kOFFdxONdt+xON=kONkON+kOFFkON is the loss rate per hour , kOFF is the establishment rate per hour , xON is the fraction of GFP+ cells at a given time , and t is time . Solving the differential equation for xON yields:xON=kONkON+kOFF ( 1−e−tkON+kOFF ) orifxON=0att=0xON=kONkON+kOFF ( 1−e−tkON+kOFF ) +e−tkON+kOFFifxON=1att=0 Therefore , the following equations were used to model switching rates between epigenetic states from data in Figure 4B and E . Sorting silenced cells ( Figure 4B ) :xON=kONkON+kOff ( 1−e−tkON+kOFF ) Sorting expressed cells ( Figure 4E ) :xON=kONkON+kOff ( 1−e−tkON+kOFF ) +e−tkON+kOFF The nls ( ) function in R was used to provide a nonlinear least squares estimate of the unknown variables kON and kOFF for each genotype , and 95% confidence intervals for estimates . With this approach , each genotype had an estimated kON and kOFF from sorting silenced cells and an estimated kON and kOFF from sorting expressed cells . Since sorting silenced cells subsequently allowed for observation of more loss-of-silencing events , the kON rates from those data were considered more accurate and used in Figure 4C . Similarly , the kOFF rates calculated from sorting expressed cells were used in Figure 4F . Because each population of sorted cells was evenly divided into three subpopulations , each genotype has three calculated values for the percent of GFP+ cells at each given time point after sorting . The nonlinear least squares estimate was made by drawing a best fit line through all data points for a given genotype , effectively combining the values of all subpopulations . The quality of the fit was calculated using the confint2 ( ) function and represented as 95% confidence intervals for kON values in Figure 4C and kOFF values in Figure 4F . An alternative approach involved drawing a best fit line for each individual subpopulation to give three kON values and three kOFF values for each genotype and averaging these values to get a single kON value and kOFF value for each genotype , with error bars representing a standard deviation . Though we also performed this latter analysis method , we favor the former analysis method because it incorporates how well the data fit the nonlinear least squares estimate . Notably , both analysis methods gave similar kON and kOFF values . The generation time of DPB3 MCM2 ( JRY11471 ) was 1 . 96 hours in CSM media at 30°C . To convert kON and kOFF as rates per hour to rates per generation , we multiplied these variables by the generation time . Similar generation times were observed for all replisome mutants . Cells were grown to saturation overnight in 5 mL CSM at 30°C . The following day , these cells were back-diluted to ~0 . 1 O . D . in 50 ml CSM and grown at 30°C for 5 hr . Cells were then centrifuged and washed twice in 500 µl SKC buffer ( 1 . 2 M Sorbitol , 100 mM KH2PO4 , 0 . 5 mM CaCl2 , 7 mM β-mercaptoethanol ) and then resuspended in 100 µl SKC buffer . Cells were incubated at 37°C for 15 min , then 30 µl of 1 mg/mL Zymolyase-100T ( MP Biomedicals , LLC , Solon , OH ) was added for a final concentration of 0 . 23 mg/ml Zymolyase-100T and incubated at 37°C for 15 min . All subsequent steps were performed on ice and subsequent centrifugations performed with an accuSpin Micro 17R ( Fischer Scientific , Hampton , NH ) . Once spheroplasting was complete , cells were spun at 3 k RPM for 3 min at 4°C . Cells were washed twice in 500 µl SPC buffer ( 1 M Sorbitol , 20 mM PIPES pH 6 . 3 , 0 . 1 mM CaCl2 , with Roche cOmplete protease inhibitors ( Sigma ) ) and spun at 2 k RPM for 3 min at 4°C between washes . Cells were resuspended in 250 µl SPC buffer , and this solution was gently mixed with 250 µl freshly prepared Ficoll buffer ( 9% Ficoll , 20 mM PIPES pH 6 . 3 , 0 . 5 mM CaCl2 ) to lyse the cell membranes . Nuclei were then pelleted by centrifugation at 10 k RPM for 20 min at 4°C . Nuclei were washed twice in 500 µl SPC and spun at 8 k RPM for 3 min at 4°C between washes . Washed nuclei were subsequently resuspended in 250 µl SPC and CaCl2 was added to a final concentration of 2 mM CaCl2 . Nuclei were incubated for 5 min at 37°C , then 20 units of Worthington MNase was added ( Worthington Biochemical Corporation , Lakewood , NJ ) . Nuclei were incubated for 15 min at 37°C . MNase activity was quenched by addition of EDTA to a final concentration of 10 mM EDTA . Nuclei were centrifuged at 3 . 7 k RPM for 5 min at 4°C . The nucleosome-containing supernatant was subsequently removed and DNA and RNA were purified using a Qiagen spin column . RNase A ( Sigma ) was added to a final concentration of 1 mg/ml RNase A and incubated for 2 hr at 37°C . DNA was then purified using a Qiagen spin column . MNase libraries were constructed with NEBnextUltra II library preparation kit ( New England Biolabs , Ipswich , MA ) and sequenced on an Illumina HiSeq4000 ( Illumina , San Diego , CA ) as 100 bp paired-end reads . Reads were mapped to the Saccharomyces cerevisiae S288C genome ( GenBank accession number GCA_000146045 . 2 ) using Bowtie2 ( Langmead and Salzberg , 2012 ) . Mapped reads between 140 bp and 180 bp in length were used in all further analysis to ensure mononucleosome resolution . The midpoint for each read was calculated and midpoints were stacked in a histogram . Finally , a 25 bp rolling mean was used to smooth out the resulting nucleosome peaks . All sequences and processed data files have been deposited in the NCBI Gene Expression Omnibus archive under accession number GSE136897 . | A crucial process in life is the ability of cells to pass on useful information to their descendants . Some of this information is encoded within molecules of DNA , including genes that contain specific coded instructions . Another layer of information helps to specify whether individual genes are switched on or off , which means cells with the same genes can perform different tasks . However , it remains unclear exactly how cells pass on this additional layer of “epigenetic” information . Inside human , yeast and other eukaryotic cells , DNA is wrapped around scaffold proteins known as histones . Cells modify histones by adding chemical tags to them , and histones within the same gene often have specific patterns of chemical tags . One popular hypothesis is that these marked histones constitute epigenetic information that may be passed on when DNA replicates before a cell divides to make two daughter cells . This model predicts that the marked histones need to be divided equally between the two sets of DNA to allow the epigenetic information to be faithfully passed on to both daughter cells . To test this prediction , Saxton and Rine studied a gene called HMR that is involved in mating in yeast . This gene is constantly silenced ( in other words , not actively providing instructions to the cell ) and contains histones with very specific patterns of chemical tags . For the experiments , Saxton and Rine made a series of mutations in the yeast that increased how often these marked histones were divided unequally when the yeast cells replicated their DNA . Unexpectedly , these mutations had little impact on the ability of the cells to pass on the silenced state of HMR to their offspring . These findings argue against the classic model that marked histones carry epigenetic information . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2019 | Epigenetic memory independent of symmetric histone inheritance |
Embryonic anterior–posterior patterning is well understood in Drosophila , which uses ‘long germ’ embryogenesis , in which all segments are patterned before cellularization . In contrast , most insects use ‘short germ’ embryogenesis , wherein only head and thorax are patterned in a syncytial environment while the remainder of the embryo is generated after cellularization . We use the wasp Nasonia ( Nv ) to address how the transition from short to long germ embryogenesis occurred . Maternal and gap gene expression in Nasonia suggest long germ embryogenesis . However , the Nasonia pair-rule genes even-skipped , odd-skipped , runt and hairy are all expressed as early blastoderm pair-rule stripes and late-forming posterior stripes . Knockdown of Nv eve , odd or h causes loss of alternate segments at the anterior and complete loss of abdominal segments . We propose that Nasonia uses a mixed mode of segmentation wherein pair-rule genes pattern the embryo in a manner resembling Drosophila at the anterior and ancestral Tribolium at the posterior .
Control of axial patterning and embryonic development is well understood in Drosophila ( reviewed in Liu and Kaufman , 2005b; Peel et al . , 2005; Rosenberg et al . , 2009; Pankratz and Jaj , 1993 ) . Extensive work has elucidated the genetic basis of establishment of the anterior–posterior ( A–P ) and dorsal–ventral ( D–V ) axes of the fly embryo . For the A–P axis , maternally loaded mRNAs generate localized signaling centers at each pole of the egg to establish morphogenetic gradients . These gradients instruct , in a concentration dependent manner , broad domains of expression of early zygotic genes , the ‘gap genes’ ( Chen et al . , 2012 ) . This is made possible in part by the syncytial environment of the early blastoderm embryo where nuclei are not bounded by membranes , allowing diffusion of morphogen transcription factors through a shared cytoplasm without the need for cell–cell signaling . In this environment , broad activation by maternal factors coupled with repressive activities by the gap genes leads to the expression of the pair-rule genes in two-segment periodicity , as pair-rule stripes . The overlapping registers of different pair-rule genes ultimately establish segment polarity through activation of the segment polarity genes , each expressed in stripes with single segmental register . This mode of development is termed ‘long germ’ embryogenesis because the embryo occupies all of the blastoderm apart from a dorsal region representing the extraembryonic ammnioserosa . A striking feature of long germ embryogenesis is that virtually all of segment patterning is completed synchronously in the syncytial environment . However , forays into other insect models have revealed that the Drosophila paradigm is an evolutionarily derived state , and that insects generally undergo a very different type of embryogenesis and segmentation ( reviewed in Liu and Kaufman , 2005b; Peel et al . , 2005; Rosenberg et al . , 2009 ) . Unlike flies , most insect embryonic primordia occupy only a small portion of the blastoderm and only few anterior segments ( head and thorax ) are patterned in a syncytial environment . The remainder of the embryo is generated after cellularization via a ‘growth zone’ , at the posterior region of the embryo . This mode is termed ‘short germ’ embryogenesis . Recently , the mechanisms governing posterior segment patterning and growth in the Tribolium embryo were characterized in elegant detail ( Choe et al . , 2006; Choe and Brown , 2009; El-Sherif et al . , 2012; Sarrazin et al . , 2012 ) : Oscillations of the pair-rule gene Tc’odd-skipped ( Tc’odd ) in the growth zone are in turn linked to a circuit of two other pair-rule genes , Tc’runt and Tc’even-skipped ( Tc’eve ) , such that each new pair of segments experiences a pulse of Tc’odd and requires both Tc’eve and Tc’runt expression in order to progress; the driver of these oscillations is still unknown . The waves of expression of Tc’odd-skipped pass through the growth zone rhythmically , generating segments and new stripes of stable expression with each periodic pulse ( Sarrazin et al . , 2012 ) . RNAi of Tc’odd , Tc’runt , or Tc’eve results in asegmental embryos , underscoring their requirement in both growth zone-derived segments and earlier blastoderm anterior segments ( Choe et al . , 2006 ) . In contrast , the pair-rule genes Tc’sloppy paired ( Tc’slp ) and Tc’paired ( Tc’prd ) appear in two-segment periodicity in head stripes and in stripes that emerge from the growth zone , and RNAi of those genes produce classical pair-rule phenotypes , in which alternating segments are lost ( Choe and Brown , 2007 ) . Live imaging revealed that formation of posterior segments results primarily from convergent extension and short-range cell movements and not strictly from cell division within the ‘growth zone’ . This mechanism appears similar to both segmentation of vertebrate presomitic mesoderm ( reviewed in [Dubrulle and Pourquie , 2004] and [Pourquie , 2011] ) and to segment formation in more basal arthropods , including the centipede Strigamia maritima ( Chipman et al . , 2004; Chipman and Akam , 2008 ) and the spider Cupiennius salei ( Stollewerk et al . , 2003 ) , suggesting it as an ancient mechanism inherited from the last common ancestor of all segmented animals ( though this interpretation is still debated; reviewed in [Davis and Patel , 1999] ) . As Drosophila is only one example of a derived long germ strategy , one outstanding question is how transitions from short germ to long germ embryogenesis occurred , such that the same set of segmentation genes possesses different functions . The careful study of additional long germ insects should shed light on what aspects of Drosophila development are essential facets of long germ embryogenesis and which aspects are more evolutionarily labile . Other model species have been studied , including long germ beetles ( e . g . , Callosobruchus order: Coleoptera; ( Patel et al . , 1994 ) ) , and several members of the order Hymenoptera , including the honeybee , Apis mellifera ( Dearden et al . , 2006; Wilson et al . , 2010; Wilson and Dearden , 2011 , 2012 ) and the jewel wasp , Nasonia vitripennis ( Nv ) ( Pultz et al . , 1999; Werren et al . , 2010 ) . However , systematic characterization of their pair-rule genes and segmentation mechanisms is still incomplete . We use the wasp Nasonia vitripennis as a model for the study of A–P patterning , as a species that appears to have evolved , independently of Drosophila , a similar mode of long germ embryogenesis . We have previously characterized the early patterns of Nasonia segmentation genes and found that maternal and gap gene expression confirms a long germ mode of embryogenesis . This conclusion was based on the existence of two polar signaling centers , each utilizing localized maternal Nv orthodenticle ( otd ) mRNA that encodes a morphogen . Nv otd acts in combination with Nv hunchback ( hb ) and localized maternal Nv giant ( gt ) at the anterior , and with localized maternal Nv caudal ( cad ) at the posterior , to specify positional identity . The domains of zygotic expression of Nv hb , Nv gt , Nv cad , Nv Krüppel ( Kr ) , Nv tailless ( tll ) , and Nv knirps ( kni ) closely resemble their Drosophila counterparts , consistent with a similar mode of blastoderm allocation ( Pultz et al . , 2005; Lynch et al . , 2006; Olesnicky et al . , 2006; Brent et al . , 2007 ) . Although these data support Drosophila-like early regulatory interactions and a long germ mode of embryogenesis , little was known about later stages of Nasonia embryonic patterning . We analyzed the expression and function of the pair-rule genes Nv eve , Nv odd , Nv runt and Nv h during embryogenesis . We found that each gene is expressed in both a canonical long-germ pair-rule stripe pattern at the anterior , as well as late-forming posterior stripes , indicating a dual mode of regulation . Strikingly , Nv eve is ultimately expressed in a total of 16 segmental stripes , of which six are derived from a single posterior stripe in the cellularized blastoderm . We also observe waves of Nv odd expression that resemble the waves of Tribolium odd expression , suggesting the residual activity of a segmentation clock in Nasonia . As in Tribolium , we found that mitoses do not occur exclusively at the site of late forming segments , but mitotic figures are not randomly distributed throughout the embryo . Instead , coordinated mitoses resembling the later mitotic domains of Drosophila ( Foe , 1989 ) appear and progress in waves from anterior to posterior , and are largely excluded from stripes of eve expression , suggesting a coordination of mitoses by segmentation genes . Using morpholinos to knock down gene function , we found that Nv eve , Nv odd and Nv h phenotypes do not affect alternating segments at the posterior , unlike what is observed in Drosophila . Instead , these ‘pair-rule’ genes are required for the formation of a continuous posterior region comprising abdominal segments A5–A10 . Phenotypes in the anterior of the embryo are gene-specific; each gene exhibits a partial pair-rule phenotype in the allelic series . We suggest that Nasonia uses ‘pair-rule’ genes to pattern the embryo in a manner that resembles both Drosophila and Tribolium . We present a model for how this mixed mode of segmentation is achieved .
The expression of eve has been studied in many insects , owing to a widely cross-reacting antibody . Its promoter has also been well studied in Drosophila and has become a classic example of modular gene control ( Patel et al . , 1992; Small et al . , 1992 , 1996 ) . We used circular RACE from total embryo RNA ( McGrath , 2011 ) to generate a fragment of approximately 1 kb corresponding to the coding region of Nv eve , including the highly conserved homeodomain ( Genbank Accession# KC168090 ) . Several minor transcript variants were captured and sequenced , but not studied further ( see ‘Materials and methods’ for GenBank accession numbers ) . The homeodomain of Nv Eve shares 81 . 7% amino acid identity with its Drosophila counterpart . We used in situ hybridization to look at the expression pattern of Nv eve during Nasonia embryogenesis ( Figure 1 ) . Nv eve expression becomes detectable as a broad early domain in the blastoderm embryo at around 3 hr after egg laying ( AEL ) ( Figure 1A ) . This domain broadens and its boundaries sharpen between 3 and 4 hr , by which time a faint posterior stripe ( hereafter referred to as ‘stripe 6’ , see below ) becomes evident ( Figure 1B , C ) . As embryogenesis progresses toward cellularization , the anterior domain splits into three distinct double-segment periodicity pair-rule stripes , stripes 1 , 2 and 3 ( Figure 1D–F ) . By cellularization , around 6 hr , a faint 4/5 stripe appears between the anterior stripes and stripe 6 , which has become more intense ( Figure 1F ) . Between 6 and 8 hr AEL , stripe 4/5 splits into distinct double-segment periodicity stripes 4 and 5 , whereas stripes 1 and 2 split into single-segment periodicity segmental stripes ( Figure 1G–I ) . In Drosophila , eve secondary stripes form de novo , between primary pair-rule stripes , in contrast to secondary paired stripes that later split from primary stripes , forming segmental stripes that affect all segments ( Macdonald et al . , 1986; Kilchherr et al . , 1986 ) . Splitting of Nasonia eve double-segment stripes into single-segment stripes may occur by a similar mechanism ( see below ) . As gastrulation progresses between 8 and 10 hr AEL , double-segment pair-rule stripes 3–5 also split to give rise to two distinct single-segment stripes each ( Figure 1J–L ) . This anterior to posterior progression of Nv eve stripes is consistent with the sequential appearance of the segment polarity genes Nv wg ( Figure 1—figure supplement 1 ) and Nv en ( Pultz et al . , 1999 ) , which are first detected around cellularization in a few anterior segments and then appear in stripes progressively , in an anterior to posterior manner . 10 . 7554/eLife . 01440 . 003Figure 1 . Summary of Nasonia eve mRNA expression . Embryos are shown with anterior left and dorsal up . Nv eve is initially expressed in a broad domain ( A and B ) , which sharpens as a posterior stripe becomes visible at around 4 hr after embryo laying ( AEL ) ( C and D ) . The broad domain retracts anteriorly and gives rise to three apparently double-segment stripes ( E and F ) . Between stripes 3 and posterior stripe 6 , an additional double stripe precursor comes up at around 6 hr AEL ( stripe 4/5; panels F and G ) and this splits to form two double-segment stripes , ‘4’ and ‘5’ as double-segment stripes 1–3 split into two single-segment stripes each between 6 and 8 hr AEL ( F–J ) . Stripes 4 and 5 also split to form single-segment stripes during early gastrulation , and stripe 6 broadens ( K and L ) , giving rise to stripes that are visibly distinct during germ band extension in non-fluorescent staining by 10–12 hr AEL ( M–R , arrowheads ) . There are a total of 16 single-segment stripes of Nv eve . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 00310 . 7554/eLife . 01440 . 004Figure 1—figure supplement 1 . Nv wingless ( Nv wg ) mRNA expression in the embryo . All embryos are shown with anterior to the left and dorsal up except as indicated . ( A ) Precellular blasoderm with no expression of Nv wg . ( B ) Cellular blastoderm embryo exhibiting head expression and one stripe of Nv wg . ( C ) Early gastrula embryo exhibiting Nv wg staining in three segmental stripes comprising thoracic segments . ( D ) Ventral view of early gastrula embryo with four segmental stripes , including one abdominal segment . ( E ) Germ band extension embryo with six segmental stripes of Nv wg expression . ( F ) Germ band extension embryo exhibiting eight segmental stripes , including three thoracic and five abdominal stripes , as well as additional head segmental stripes . ( G ) Ventral view of a fully extended germ band with the full complement of 16 segmental Nv wg stripes . ( H ) Lateral view of germ band retracting embryo with 16 segmental stripes , including three clear segmental stripes in the head . ( I ) Lateral view of dorsal closure embryo exhibiting fading segmental Nv wg staining in all segments . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 004 A remarkable feature of Nv eve expression is that the posterior stripe 6 broadens significantly at about 10 hr AEL ( Figure 1L–N ) , before generating six additional stripes with single-segment periodicity , allowing the embryo to reach 16 individual segmental stripes by the time the germ band is fully extended . This division of stripe 6 initiates with an anterior band that will give rise to four stripes ( segmental stripes 11–14; Figure 1P , Q , arrowheads; see below ) , and two later-appearing segmental stripes 15 and 16 ( Figure 1Q , arrowheads ) . The last stripe , Nv eve 16 , appears only at full germ band extension ( Figure 1R ) completing the 16 stripes observed at this stage . The Nasonia embryo has 16 segments , whereas Drosophila has only 14 ( Figure 2A ) . In Drosophila , eve and other pair-rule genes are expressed with double-segment periodicity: seven transverse ‘pair-rule’ stripes are evident as a full complement in the blastoderm embryo at cellularization . If the Nasonia embryo were patterned using the same mechanisms as Drosophila , then eight pair-rule stripes would be predicted . However , only five truly pair-rule ( double segment ) stripes are apparent at cellularization , while stripe 6 gives rise later to four , then six single-segment stripes and six segments and is therefore not pair-rule ( Figure 2E ) . This delayed sequential posterior segmentation is therefore more reminiscent of the segmentation described in short germ insects . 10 . 7554/eLife . 01440 . 005Figure 2 . Nv eve epistasis with maternal and gap genes . ( A ) Schematic representation of the germ-band-extended embryo , showing 16 single-segment stripes of Nv eve expression , and their segment counterparts in the patterned larval cuticle . Colored boxes cover the segments of the larval cuticle that are lost or fused in each RNAi background . All embryos are shown anterior left , dorsal up ( except where indicated ) . Nv eve mRNA expression is shown in each embryo ( B–G ) . Wild-type ( WT ) embryos are shown as staged controls for RNAi embryos . ( B ) WT early blastoderm embryo . ( C ) WT cellular blastoderm embryo . ( D ) WT early gastrula extension embryo . ( E ) WT germ-band-retracted embryo . ( F–H ) gt RNAi embryos stained for Nv eve mRNA expression . ( F ) Cellular blastoderm embryo with reduced Nv gt exhibits loss of anterior Nv eve stripes ( x ) . ( G ) Nv gt RNAi embryo in early germ-band-extension exhibits loss of anterior Nv eve stripes and improper splitting of Nv eve stripe 5 , as well as aberrant dorsal anterior expression of Nv eve . ( H ) Nv gt RNAi embryo at dorsal closure exhibits a stripe of Nv eve at the anterior , as well as a reduced number of posterior segmental Nv eve stripes . ( I–L ) Nv hb mutant embryos stained for Nv eve mRNA expression . ( I ) Early blastoderm Nv hb mutant embryos have a reduced central Nv eve domain ( bounded by black arrowheads ) , and an ectopic anterior Nv eve stripe ( white arrowhead ) . ( J ) Nv hb mutant cellular blastoderm embryo with a single anterior domain of Nv eve that has failed to resolve , and a single stripe 4 which exhibits delayed splitting . ( K ) Nv Hb mutant germ-band extension embryo with fused anterior domain ( line ) and 6 segmental stripes , representing derivatives of Nv eve stripes 4 and 5 and two derivatives of stripe 6; additional stripe 6 derivatives are absent ( x ) . ( L ) hb mutant dorsal closure embryo exhibiting fused anterior domain ( line ) and the same number of derivatives as in ( M ) , with more posterior segments missing ( x ) . ( M–O ) Nv cad RNAi embryos stained for Nv eve mRNA expression . ( M ) Nv cad RNAi early blastoderm with reduced central Nv eve domain that is also posteriorly shifted ( anterior boundary indicated by black arrowhead ) . ( N ) Nv cad RNAi cellular blastoderm embryo with posteriorly shifted ( arrowhead ) , reduced Nv eve central domain , whose splitting is delayed . ( O ) Nv cad RNAi early gastrula embryo with posterior shift in Nv eve expression ( black arrowhead ) . Four double-segment periodicity stripes are split into single-segment stripes and stripe 5 remains intact . ( P–S ) Nv Kr RNAi embryos stained for Nv eve mRNA expression . ( P ) Nv Kr RNAi precellular blastoderm embryo with aberrant Nv eve central domain resolution , where stripes 2–3 appear posteriorly shifted . ( Q ) Dorsolateral view of a Nv Kr RNAi embryo where stripes 2 and 3 are less refined than WT and 3 is posteriorly shifted . No stripe 4/5 expression is detected ( X ) . ( R ) Nv Kr RNAi early gastrula embryo with aberrant stripe 2 splitting and aberrant resolution of stripes 3–5 . ( S ) Moderately affected Nv Kr RNAi germ-band retraction embryo with fused segments in the middle of the embryo ( line ) . ( T–V ) Nv tll RNAi embryos stained for Nv eve mRNA expression . ( T ) Nv tll RNAi early blastoderm embryo with expanded Nv eve expression domains toward both poles ( arrowheads ) . ( U ) Nv tll RNAi precellular blastoderm embryo showing delayed resolution of Nv eve stripes 1–3 and Nv eve stripe 6 shifted to the extreme posterior pole of the embryo ( arrowhead ) . ( V ) Nv tll RNAi dorsal closure embryo showing abnormal posterior Nv eve stripe formation . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 00510 . 7554/eLife . 01440 . 006Figure 2—figure supplement 1 . Nv eve/Nv gt double FISH in the embryo . Lateral view of early Nasonia gastrula embryo double stained for Nv eve and Nv gt mRNA . ( A ) Nv eve mRNA . ( B ) Nv gt mRNA . ( C ) Merge of Nv eve and Nv gt channels . Note the position of the late posterior stripe of Nv gt relative to Nv eve stripe 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 00610 . 7554/eLife . 01440 . 007Figure 2—figure supplement 2 . ( A ) Summary of maternal and gap gene loss of function phenotypes in Nasonia . On the vertical axis , genes affected by RNAi or genetic lesion are listed , and indicated as either maternal or zygotic phenotypes . Segments lost in each background are indicated as gray bars . Segments that remain in a given background are annotated alphanumerically . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 007 To explore this apparent combination of short and long germ characters , we determined how Nv eve expression is controlled by upstream genes in the known Nasonia A–P patterning network . Early embryonic expression of Drosophila eve is controlled by maternal and gap genes , including bicoid ( bcd ) , hb , Kr , gt , kni and torso ( MJaJ & H , 1993; Small et al . , 1992 , 1996; Schroeder et al . , 2004; Small and Levine , 1991 ) , whereas later maintenance is achieved via autoregulation ( Jiang et al . , 1991 ) . Some Tribolium eve pair-rule stripes are also under the control of gap genes although some of the segments themselves are born much later than gap gene expression ( Sulston and Anderson , 1996; Cerny et al . , 2005 ) . For example , in Tc’Kr mutant embryos , segments anterior to the normal Kr expression domain ( T1–T3 ) appear wild type , but expression of both Tc’eve and Tc’en is lost in posterior segments and no segments are formed posterior to A4 ( Cerny et al . , 2005 ) . Therefore , Tribolium gap genes can affect the specification of segments that are not yet formed , presumably because of interactions with the growth zone . In other short germ insects , like Oncopeltus fasciatus , eve acts as a gap gene , regulating expression of hunchback and Krüppel ( Liu and Kaufman , 2005a ) . To determine how the known maternal and gap genes regulate Nv eve expression in the early embryo , we used parental RNAi injections in pupal Nasonia females to knockdown Nv gt , Nv Kr , Nv tll and Nv cad mRNA , as well as a null mutation in Nv hb ( Pultz et al . , 2000 , 2005 ) . As we previously reported , Nv giant knockdown results in the loss of all segments anterior to A1 and fusion of segments A6 and A7 ( Brent et al . , 2007 ) . Nv gt RNAi blastoderm embryos exhibit the loss of Nv eve double-segment stripes 1–3 ( Figure 2F ) , as well as aberrant resolution of the first splitting events of stripe 6 ( Figure 2G , arrowheads ) . Double in situ hybridization shows that , in the wild type , a late posterior stripe of Nv gt forms after Nv eve stripe 6 and appears to be within the Nv eve stripe 6 domain ( Figure 2—figure supplement 1 ) . These data suggest that a posterior Nv gt domain may partially affect stripe 6 splitting . Late Nv gt RNAi embryos exhibit strong anterior defects after dorsal closure , but the other posterior single-segment stripes of Nv eve appear unaffected ( Figure 2H ) . In Nv hb mutant ( headless ) embryos , both head and thoracic fates and abdominal fates posterior to A6 are lost ( Pultz et al . , 2005 ) . Consistent with this phenotype , Nv eve double-segment stripes 1–3 ( that give rise to head and thoracic fates ) form but never resolve ( Figure 2K , black arrowheads ) . An ectopic stripe of eve can be seen in the anterior of some embryos ( Figure 2I , white arrowhead ) , consistent with the ectopic stripe of Nv cad ( an activator of eve ) in the head of some Nv hb mutant embryos ( Olesnicky et al . , 2006 ) . Gastrula and later germ band embryos exhibit normal Nv eve double-segment stripe formation and splitting of stripes 4 and 5 , which give rise to segments A1–A4 . However , only the two anteriormost segments are formed from Nv eve stripe 6 ( 6a and 6b ) in Nv hb mutant embryos ( Figure 2K , L ) . Nv caudal ( cad ) is expressed maternally as a mRNA gradient with a localized posterior source ( Olesnicky et al . , 2006 ) . Nv cad RNAi results in loss of all segments posterior to A1 . Moderately affected Nv cad RNAi embryos exhibit a reduced early broad domain of Nv eve that is slightly shifted posteriorly ( Figure 2M ) . This domain resolves poorly , with only weak activation of anterior Nv eve stripes and no posterior abdominal expression of Nv eve ( Figure 2N–O ) . As in Drosophila , Nv Krüppel ( Kr ) is expressed in a central domain , and Nv Kr is required for formation of segments T3 to A4 ( Brent et al . , 2007 ) . In Nv Kr RNAi embryos , both anterior and posterior domains of Nv hb expression expand towards the center of the embryo ( Brent et al . , 2007 ) . Consistent with expansion of Nv hb , we observed that Nv eve stripe 2 and 3 exhibit aberrant resolution and Nv eve stripes 4 and 5 fail to resolve in embryos with knocked-down Nv Kr ( Figure 2P–S ) . Posterior segments are unaffected , as reflected by normal expression of Nv eve posterior to stripe 5 ( segment A4; Figure 2R , S ) . This phenotype is dramatically different from Tc’Kr knockdown where all posterior segments are deleted , likely because Nv Kr is expressed anterior to the growth zone while Tc’Kr abuts it . tailless mRNA is expressed in both an anterior and a posterior domain , though only posterior segments are affected by Nv tll RNAi ( Lynch et al . , 2006 ) . The most severely affected embryos are missing the six posterior abdominal segments . These embryos also exhibit an apparent slight anterior shift of the broad early domain of Nv eve expression and of stripe 6 ( Figure 2T , U ) . Stripe 6 does not appear to resolve , resulting in an enduring ring of Nv eve expression and no Nv eve single-segment stripes posterior to this ring are apparent ( Figure 2V ) . Taken together , and consistent with previously described cuticular phenotypes for maternal and gap genes in Nasonia ( Pultz et al . , 2005; Lynch et al . , 2006; Olesnicky et al . , 2006; Brent et al . , 2007; Figure 2—figure supplement 2 ) , these data show that early Nv eve expression in blastoderm embryos involves regulatory interactions reminiscent of those underlying Drosophila long germ embryogenesis . However , since severe RNAi phenotypes of several genes , such as Nv cad and Nv tll results in total loss of posterior segments , these did not provide additional information for understanding the establishment of posterior Nv eve expression . To determine whether cell division plays a role in subdivision of the Nv eve posterior domain into single-segment stripes , we used in situ hybridization to visualize Nv eve mRNA in embryos where mitotic cells were labeled with antibodies against phosphorylated histone H3 . We found that there is no cell division that is consistent with a role in pair-rule stripe splitting ( Figure 3A–A″ ) , or in stripe 6 resolution ( Figure 3B–B″ ) , suggesting that the dynamics of the Nv eve mRNA pattern mostly involves transcriptional regulation . Nevertheless , later mitoses occur in restricted spatial domains , reminiscent of later Drosophila mitotic domains within segments of the expanding germ band ( Figure 3B–D″ ) . A relationship between gap gene function and regulation of mitotic domains via regulation of string has been suggested in Drosophila ( Edgar et al . , 1994 ) but not demonstrated . Nasonia mitotic domains appear in an anterior to posterior progression , allowing for progressive expansion of segments along the A–P axis via concerted cell divisions within domains . Strikingly , mitotic figures appear to be largely excluded from Nv eve stripes in these early stages of embryogenesis ( Figure 3B″ , C″ , Figure 3—figure supplement 1 ) . Later embryos in which anterior Nv eve stripes are beginning to fade exhibit overlap of mitotic figures with weakened Nv eve stripe expression ( Figure 3D″ ) . Ultimately , embryos exhibit widespread mitotic figures that do not correspond to any apparent concerted domains or pattern , more like the pattern of mitoses described in several short germ insects ( Handel et al . , 2005; Liu and Kaufman , 2009 ) . 10 . 7554/eLife . 01440 . 008Figure 3 . Nv eve expression and cell division appear to be coordinated . Embryos co-stained for Nv eve mRNA using in situ hybridization and fluorescent detection , as well as for mitotic figures , using an antibody against phospho Histone H3 . Embryos are shown with anterior left and dorsal up , except columns B and C , which are ventral views . ( A–A″ ) An early gastrula embryo exhibiting 15 stripes of Nv eve , including five derivatives of stripe 6 ( A ) , has no evident mitotic figures in the posterior domain of Nv eve stripe 6 differentiation ( A′ ) . ( A″ ) Merge of panels A and A′ . ( B–D″ ) Timecourse series of wild-type embryos stained for Nv eve mRNA and phospho-Histone H3 . ( B–D ) . Top panels are Nv eve in situ alone , middle panels ( B′–D′ ) are phospho-Histone H3 antibody staining , and bottom panels ( B″–D″ ) are merge images of upper panels , showing localization of mitotic figures relative to Nv eve stripes . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 00810 . 7554/eLife . 01440 . 009Figure 3—figure supplement 1 . Quantification of PH3 positive cells relative to Nv eve stripes in the embryo . Embryos stained for phosphorylated histone H3 ( to label mitotic figures ) and for Nv eve mRNA ( as shown in Figure 3 ) were analyzed to quantify the number of mitotic cells occurring between stripes as compared to within eve stripes . Cell number counted is plotted on the Y axis as a function of position along the A–P axis of the embryo ( as indicated by Nv eve stripe position ) , which is indicated on the X axis . Mitotic cells that fall partially on a stripe are counted as occurring in the stripe . Each unique color indicates one embryo quantified in this manner , and embryos corresponding to those shown in Figure 3 are also labeled according to their position in the figure . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 009 The expression of Nv eve suggests a combination of long germ and short germ character . To further explore this possibility , we knocked down Nv eve gene function . Although parental RNAi in Nasonia is effective for maternal and early zygotic genes ( Lynch and Desplan , 2006; Lynch et al . , 2006; Olesnicky et al . , 2006; Brent et al . , 2007 ) , it often does not provide significant knockdown of later-acting genes . To overcome this limitation , we designed an Nv eve morpholino overlapping the translation start site , as well as one directed at the exon–intron junction in the homeobox . These two independent morpholinos are expected to disrupt Nv eve activity and indeed result in comparable phenotypes . The Nasonia larval cuticle has relatively few landmarks to allow for interpretation of segmentation phenotypes . Beyond the denticle belts present on each of the three thoracic segments and ten abdominal segments , large spiracles are found on segments T2 , A1 , A2 and A3 ( Figure 4A , yellow arrowheads ) . In the head , two bright structures indicate the positions of antennal papillae . Morpholino block of Nv eve causes a range of phenotypes ( Figure 4B–E ) , resulting in severe truncation of the embryo with loss of posterior-derived segments as well as a partial pair-rule phenotype for more anterior segments . The phenotypic series includes progressive truncation at the posterior , causing fusion of segments A9–10 in the least affected cuticles , and then A8–10 ( Figure 4C ) with segment A6 eventually lost , whereas A7 remains virtually intact . In the most severely affected embryos , the entire posterior of the embryo is truncated with A5–A10 missing . ( The approximate percentage of embryos in each phenotypic class shown in Figure 4 is indicated in the ‘Materials and methods’ . ) 10 . 7554/eLife . 01440 . 010Figure 4 . Morpholino knockdown of Nv eve , Nv hairy , and Nv odd results in embryo patterning defects . First instar larval cuticles are shown with anterior left and generally ventral denticle patterns are shown . ( A , F , K ) Wild-type larval cuticles . Yellow arrows indicate spiracles present on segments T2 , A1 , A2 and A3 . Bright anterior labral appendages are apparent at the extreme anterior of the larva . ( B–E ) Unhatched larvae from Nv eve morpholino ( MO ) -injected embryos , in order of increasing phenotype severity . Red arrowheads indicate loss of midline cuticle . Blue dot indicates head open defect . Yellow arrowheads indicate position of spiracles . ( G–J ) Unhatched larvae from Nv odd morpholino ( MO ) -injected embryos , in order of increasing phenotype severity . Yellow arrows indicate position of spiracles , red arrows indicate A3/A4 fusion . X indicates naked cuticle from segment loss . Yellow line indicates multi-segment fusion . ( L–O ) Unhatched larvae from Nv hairy morpholino ( MO ) -injected embryos , in order of increasing phenotype severity . Yellow arrowheads indicate position of spiracles . Red or orange arrowheads indicate aberrantly positioned or missing spiracles . Yellow line indicates segment fusion . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 010 At the anterior , T1 is lost with fusion of T3 and A1 ( Figure 4B ) . Segments A2 and A3 are also fused , and there is a continuous lawn of denticles from A4 to the truncated posterior . This is accompanied by a disruption of the remaining denticle belts , leaving naked cuticle along the midline ( Figure 4 , red arrows ) . In the most severely affected embryos , segments anterior to A1 are lost and are accompanied by head closure defects . This phenotype represents a partial pair-rule phenotype , accompanied by posterior truncation of the embryo . It also does not exhibit the lawn of denticles phenotype of strong eve alleles in flies ( e . g . , eveR13 [Macdonald et al . , 1986; Fujioka et al . , 1999] ) , although severely affected Nasonia embryos also exhibit cuticle defects beyond a pair-rule phenotype . Hence , these results support a mixed mode of embryogenesis in Nasonia , with characteristic features resembling both long germ and short germ insects . To further examine this possibility , we then investigated the expression patterns in Nasonia of other genes acting as pair-rule in Drosophila , and undertook functional characterization of their activity during embryonic development . In the long germ Drosophila embryo , odd is expressed with a double-segment periodicity complementary to that of eve , and its inactivation causes the absence of odd segments ( Nusslein-Volhard and Wieschaus , 1980; Coulter et al . , 1990 ) . Its critical function as a mediator of the segmentation clock in the short germ beetle was recently elegantly described ( Sarrazin et al . , 2012 ) . Tc’odd begins with blastoderm expression in double-segment periodicity stripes alternating with Tc’eve expression . Then , new double-segment stripes emanate from the growth zone to generate the entire complement of odd stripes . Secondary single-segment stripes arise later ( Sarrazin et al . , 2012 ) . There are three odd paralogs in Nasonia , as in flies , where they are named odd , bowl , and sister of bowl ( or sob; ( Hart et al . , 1996 ) ) . We used sequence alignment ( Figure 5—figure supplement 2 ) and phylogenetic analysis ( Figure 5—figure supplement 1 ) to identify the Nasonia paralog that is closest to Drosophila odd-skipped , and refer to it hereafter as Nv odd . An Nv odd cDNA fragment comprising the region encoding the conserved DNA binding domain was used as a probe for in situ hybridization ( GenBank Accession # KC142194 ) . As observed above for Nv eve , the embryonic expression of Nv odd begins as a broad early domain in syncytial blasoderm embryos ( Figure 5A ) . As this broad domain strengthens and sharpens , a ventral head patch and a posterior cap appear ( Figure 5B , C ) . The broad domain resolves into two clear apparent double-segment stripes ( Stripes 1 and 2 , Figure 5D ) . A third double-segment stripe arises from the second stripe , expanding posteriorly ( Figure 5D–F ) . At the same time , a stronger posterior domain apparently advances anteriorly . Pair-rule stripe 4 ( double-segment periodicity ) arises at the anterior of the first advancing ‘wave’ at cellularization ( Figure 5G–H ) before the posterior domain recedes again ( Figure 5I–J ) . The fifth double-segment stripe arises during a second ‘wave’ ( Figure 5K–M ) at the onset of gastrulation . A sixth stripe arises in an apparently similar manner , though it is much fainter and appears while more posterior stripes are already differentiated ( Figure 5N , arrowhead ) ; the posterior cap generates two thin pair-rule stripes ( Figure 5O ) during early germ band extension . At full germ band extension , a total of eight stripes are visible ( Figure 5P ) before these fade from anterior to posterior . This dynamic expression of Nv odd in the posterior of the embryo is reminiscent of the waves of growth zone expression of Tribolium odd , where blastoderm-derived stripes initially have double-segment periodicity and later single-segment periodicity ( Choe et al . , 2006; Sarrazin et al . , 2012 ) . 10 . 7554/eLife . 01440 . 011Figure 5 . Summary of Nv odd-skipped mRNA expression . Embryos are shown with anterior left and dorsal up , except where indicated . ( A ) Precellular blastoderm embryo showing early expression of Nv odd in a broad domain and a posterior cap with a slight clearing in between . ( B ) Precellular blastoderm embryo showing ventral head patch and darkened central broad domain and distinct posterior cap . ( C ) Precellular blastoderm embryo with sharpening pair-rule stripes and expanding posterior cap . ( D ) Precellular blastoderm embryo with dark ventral head patch and posterior cap , and expansion of expression between broad central domain and posterior domain . ( E and F ) Cellularizing blastoderm embryos with three double-segment periodicity stripes , and a continuous posterior domain of variable staining intensity . Arrowhead indicates boundary of faint expression , which prefigures position of double-segment stripe 4 . ( G ) Ventral view of cellularizing embryo with three strong double-segment stripes , and a fourth stripe forming at the anterior boundary of a more uniformly staining posterior cap ( arrowhead ) . ( H ) Cellularized blastoderm embryo with four distinct double-segment stripes and a receding posterior cap domain ( arrowhead ) . ( I ) Ventral view of cellular blastoderm showing four strong double-segment stripes and receding posterior cap ( arrowhead ) , whose anterior boundary prefigures the position of stripe 5 . ( J ) Ventrolateral view of cellular blasoderm embryo showing early appearance of stripe 5 at the anterior boundary of receding posterior domain , whose staining intensity is now less uniform . ( K ) Cellular blastoderm embryo with five double-segment stripes of expression , a strong ventral head spot , and a reduced , uniform posterior cap . ( L ) Same as K , with five equivalently strong double-segment stripes . Arrowhead indicates slightly expanded posterior cap . ( M ) Early germ-band extension embryo with five double-segment periodicity stripes and two stripes becoming evident within the posterior cap . ( N ) Slightly later embryo than M , with 2 posterior cap stripes more clearly differentiated . ( O ) Slightly later embryo than N , with anterior stripes fading and posterior segments expanding . ( P ) Dorsal view , dorsal closure embryo exhibiting eight single-segment periodicity stripes . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 01110 . 7554/eLife . 01440 . 012Figure 5—figure supplement 1 . Phylogenetic analysis of odd-skipped . Phylogenetic analysis of odd-skipped protein was performed using Maximum likelihood analysis with 1000-fold bootstrap support , using RAxML via the CIPRES portal ( Copf et al . , 2003; Nagy et al . , 1994 ) . The best scoring maximum likelihood tree is shown with bootstrap support values indicated adjacent to branch . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 01210 . 7554/eLife . 01440 . 013Figure 5—figure supplement 2 . odd-skipped protein sequence alignment . Predicted odd-skipped protein sequences from Nasonia were aligned to odd-skipped related proteins sequences from Drosophila melanogaster , Anopheles gambiae , Apis mellifera , and Tribolium castaneum using ClustalW using standard parameters . Full-length sequences were used for alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 013 Using double fluorescent in situ hybridization , we confirmed that the pair-rule stripes of Nv odd and Nv eve are indeed complementary to each other , although their mode of appearance is totally different . Nv odd double-segment stripes are posterior to , and abut each posterior single-segment stripe ( i . e . , 1b , 2b ) from each eve pair-rule doublet ( Figure 6A–C ) , that is , the even-numbered segmental stripes . Late forming Nv eve stripe 15/16 intercalates between the two Nv odd stripes 7 and 8 that derive from the cap , with Nv odd stripe 8 remaining posterior to all Nv eve stripes ( excepting the last stripe , eve 16 , which is the last to appear ) , a relationship that may have ancestral origins ( see ‘Discussion’ ) . 10 . 7554/eLife . 01440 . 014Figure 6 . Phasing of Nasonia pair-rule genes in embryos using double fluorescent in situ hybridization . ( A ) Lateral view of Nv eve expression in early gastrula embryo . ( B ) Nv odd expression alone in the same embryo . ( C ) Merge of Nv eve and Nv odd channels , illustrating their relative phasing . Nv eve mRNA is pseudo-colored pink , Nv odd is in green . Arrowheads indicate position of a posterior doublet of odd stripes . ( D ) Dorsolateral view of Nv eve in later gastrula embryo . ( E ) Nv odd expression alone in the same embryo . Arrowheads indicate position of posterior Nv odd stripes 6 , 7 and 8 . ( F ) Merge of Nv eve and Nv odd channels , illustrating their relative phasing . ( G ) Lateral view of Nv eve expression in blastoderm embryo . Arrowhead indicates position of Nv eve stripe 5 . ( H ) Nv runt expression in the same blastoderm embryo . ( I ) Merge of Nv eve ( green ) and Nv runt ( pink ) channels , indicating relative phasing . ( J ) Lateral view of Nv eve expression in germ-band-extended embryo . Numbers indicate identity of Nv eve stripe . ( K ) Nv runt expression alone in the same embryo . Arrowheads indicate position of posterior primary Nv runt stripes . ( L ) Merge of Nv eve ( green ) and Nv runt ( pink ) channels , indicating relative phasing . Note that posterior Nv runt stripes , though faint , appear to be positioned posterior to odd-numbered Nv eve segmental stripes . ( M ) Lateral view of Nv eve expression in early gastrula embryo . Line indicates broadening stripe 6 . ( N ) Nv hairy expression in the same gastrula embryo . Arrowheads indicate positions of three late forming posterior double-segment stripes . ( O ) Merge of Nv eve ( pink ) and Nv hairy ( green ) channels , indicating relative phasing . ( P ) Ventral view of gastrula embryo showing Nv eve expression alone . Arrowheads indicate positions of single-segment stripes derived from eve stripe 6 . ( Q ) Nv hairy expression alone in the same gastrula embryo . Line indicates extended anterior domain continuous with stripe 1 . ( R ) Merge of Nv eve ( green ) and Nv hairy ( pink ) channels , illustrating relative phasing . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 01410 . 7554/eLife . 01440 . 015Figure 6—figure supplement 1 . Summary of Nv runt mRNA expression . Embryos are shown anterior left and ventral up unless indicated . ( A ) Early energid stage embryo with faint staining evident in posterior . ( B ) Early precellular blastoderm embryo exhibiting stronger , mostly ubiquitous Nv runt staining biased towards the posterior . ( C ) Precellular blastoderm embryo with strong stripe of Nv runt expression within larger , gap-type domain . Expression is absent from the rest of the embryo . ( D ) Precellular blastoderm embryo exhibiting three clear pair-rule stripes of Nv runt expression . ( E ) Cellular blastoderm embryo with four pair-rule stripes of Nv runt expression . ( F ) Early gastrula embryo exhibiting three pair-rule stripes , an expanding stripe 4/5 stripe domain and faint posterior stripe 6 . ( G ) Early gastrula embryo exhibiting 6 pair-rule stripes of Nv runt expression . ( H ) Gastrula embryo exhibiting splitting of anterior stripes 1 and 2 ( and possibly 3 ) , and stronger expression of stripes 4 and 5 , with expanded interstripe distance between stripes 5 and 6 . ( I ) Dorsal view of gastrula embryo at approximately the same stage as in ( H ) . ( J ) Dorsolateral view of germ-band extension embryo with fading anterior stripes , and an additional pair-rule stripe appearing between stripes 5 and 6 ( arrowhead ) . ( K ) Germ-band extension embryo exhibiting seven pair-rule stripes of expression with segmental stripes appearing in posterior interstripes ( arrowheads ) . ( L ) Ventral view of fully extended germ band exhibiting segmental expression in all segments , in addition to continuous expression in the ventral anterior of the embryo . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 015 To examine the function of Nv odd in the embryo , we used one translation blocking and one splice blocking morpholino to knock down its expression in embryos . Inactivating Nv odd function leads to loss of the most posterior germ band-derived segments A5–A10 with additional anterior defects . The most sensitive phenotypes are the fusion of segments A3 and A4 , and loss of segment A5 ( Figure 4G , red arrowheads and x ) . More severely affected embryos exhibit loss of most segments posterior to A3/A4 and naked cuticle anterior to T2 , though larval head structures are still present ( Figure 4I–J ) . In many severely affected embryos , Nv odd knockdown causes additional loss/fusion of A1 , and of T2 . Thus , phenotypes comprise loss of segments T2 , A1 , A3 and A5 and resemble a pair-rule phenotype . A small percentage of embryos are nearly asegmental , with only small patches of denticle bands of unknown identity ( not shown ) . Therefore , as also observed for Nv eve knockdown , the segmentation defects resulting from Nv odd inactivation only corresponds to a partial Drosophila-like pair-rule phenotype , and rather , in the most severe cases , resemble the phenotype of Tc’odd loss of function . It is of note that Nv odd stripes 4 , 5 and 6 appear to emerge from waves of expression at the posterior of the embryo that likely specify segments A3–A5 , which are most sensitive to loss of Nv odd function ( Figure 5E–L , Figure 4G ) . In summary , Nv odd is expressed initially in three sequentially forming anterior double-segment periodicity stripes , which appear to have Drosophila-like pair-rule character . Three more posterior double-segment stripes ( PR stripes 4–6 ) then form sequentially , apparently as ‘waves’ of Nv odd expression , resembling the clock-driven stripes of Tc’odd . Finally , two Nv odd stripes form from a posterior cap . Nv odd knockdown affects anterior thoracic segments with a partial pair-rule phenotype; it also leads to the loss of posterior segments A5–A10 . The complementarity between Nv eve and Nv odd is suggestive of cross interaction between the two genes but it is only partial and only affects half of the Nv eve segmental stripes since Nv odd does not have single-segment periodicity stripes . We sought to determine whether the remaining single segment stripes where odd is not interdigitated with Nv eve may alternate with stripes of Nv runt , as is observed in Drosophila . We studied the expression of Nv runt throughout embryogenesis ( Figure 6—figure supplement 1 ) and then used double fluorescent in situ hybridization to visualize its register with Nv eve stripes at both early and late stages . Nv runt stripes appear cleanly in an anterior to posterior progression , with six double-segment periodicity stripes visible before cellularization; two additional double-segment stripes are added at the posterior during gastrulation . Single-segment periodicity stripes only appear much later at full germ band extension when the expression of the other pair-rule genes is already well established ( Figure 6—figure supplement 1 ) . In the early embryo , Nv runt double-segment stripes appear posterior to , and partly overlapping with , each Nv eve primary double-segment stripe ( Figure 6G–I ) . Splitting of anterior Nv eve stripes moves the posterior of each doublet ( i . e . , even-numbered Nv eve single-segment stripes ) more posteriorly beyond each Nv runt primary double-segment stripe ( Figure 6G–I ) . The appearance of Nv eve stripe 5 between Nv runt double-segment stripes 4 and 5 as they split ( Figure 6G–I , arrowhead; Figure 6—figure supplement 1F , G ) suggests that eve may help to repress Nv runt , though this remains to be tested . Late expression of Nv runt in the extending germ band is considerably weaker than that of other genes , making its detection more challenging . Still , at the posterior of the embryo , Nv runt double-segment stripes appear between Nv eve single-segment stripes arising from splitting of double-segment stripes ( Figure 6J–L ) . Altogether , our data support a model in which Nv eve single-segment periodicity stripes are established through the complementary action of Nv odd for odd numbered single-segment stripes and Nv runt for even-numbered stripes , as summarized in our model below . However , these interactions are still speculative since we have been technically unable to complete the epistasis experiments needed to test this model . A predicted role for Nv h is also described below . hairy is a primary pair-rule gene in Drosophila , but in Tribolium , its function is restricted to head segment differentiation ( Carroll et al . , 1988; Carroll and Vavra , 1989; Edgar et al . , 1989; Choe et al . , 2006 ) . There are two hairy-like genes in Nasonia , and we identified the likely hairy ( h ) ortholog using phylogenetic analysis ( Figure 7—figure supplements 1 and 2 ) . We examined expression of Nv h using a probe directed against the full-length coding region ( Genbank Accession # KC190514 ) . Nv h expression begins as a single broad anterior double segment stripe 1 that incompletely spans the dorso–ventral axis . It is soon followed by a second broad double-segment stripe 2 just anterior to the middle of the embryo ( Figure 7A ) . Double-segment stripes 3 , 4 and 5 are then added sequentially before cellularization ( Figure 7B–E ) ; a faint stripe at the extreme posterior of the embryo is also visible . By gastrulation , an anterior cap becomes more visible with continuous low expression between the anterior pole and the strong stripe 1 ( Figure 7H–J ) . As gastrulation progresses , this domain becomes stronger and more uniform , whereas double-segment stripes 6 and 7 appear sequentially ( Figure 7H–K ) . Stripe 8 broadens , becoming a posterior cap whose intensity increases during germ band extension ( Figure 7J–N ) . The anterior of the embryo exhibits diffuse staining that expands from the anterior ventral side posteriorly , until germ band retracted embryos are faintly but uniformly stained with dark segmental stripes on top ( Figure 7O ) . Nv h double-segment stripes appear cleanly , and the timing and presentation of expression of posterior stripes suggest that they may respond to waves of Nv odd . Like Nv odd and early Nv runt , Nv h does not have stripes with single-segment periodicity . 10 . 7554/eLife . 01440 . 016Figure 7 . Summary of Nv hairy mRNA expression . ( A ) Blastoderm embryo with two double-segment periodicity stripes of Nv hairy expression . Note that stripe 2 is broader and stronger than stripe 1 . ( B ) Blastoderm embryo showing four double-segment periodicity stripes of expression plus an anterior accumulation of Nv hairy transcripts ( arrowhead ) . ( C ) Dorsal view of embryo as in ( B ) , illustrating the dorsal anterior expression ( arrowhead ) that is activated in the same pattern as the anterior domain of Nv tailless ( Lynch et al . , 2006 ) . ( D ) Blastoderm embryo with strong anterior and dorsal anterior expression of Nv hairy and five pair-rule stripes . ( E ) Dorsal view of embryo as in ( D ) with increased dorsal anterior expression of Nv hairy , and the anterior spreading of expression from the anterior of double-segment pair-rule stripe 1 . ( F ) Blastoderm embryo with expanding anterior domain ( line ) , five double-segment ‘pair-rule’ stripes , and two additional stripes coming up . Note that the anterior domain between stripe 1 and the anterior pole is becoming more continuous in expression . ( G ) Dorsolateral view of embryo as in ( F ) highlighting the dorsal anterior expression . Stripe 2 is still wider than other stripes . Stripe 6 appears to be of single-segment periodicity . ( H ) Early gastrula embryo exhibiting a non-homogenous but largely continuous anterior cap of Nv hairy expression ( that includes stripe 1 ) . Four additional double-segment stripes and three single-segment stripes ( two derived from stripe 6 ) are now evident . ( I ) Dorsal view of embryo slightly older than embryo in ( H ) showing the nearly continuous head domain , and the apparent splitting of stripe 1 within that domain . Double-segment stripes are thinning . ( J ) Dorsolateral view of extending germ-band embryo . Head domain is continuous ( line ) . Stripes 1–7 have single-segment periodicity , are of non-uniform strength; stripe eight appears darker and broader . ( K ) Germ-band extending embryo with a continuous head domain ( line ) and eight discrete stripes . ( L ) Dorsolateral view of germ-band extending embryo . Stripe 8 is expanded into a wedge abutting the pole cells , and the anterior domain is expanding to include stripe 2 . ( M ) Germ-band extension embryo with expanding anterior domain , that extends to include stripe 3 ( arrowhead ) . Posterior domain is expanded . ( N ) Dorsolateral view of embryo as in ( M ) showing further expansion of posterior stripe 8 domain ( line ) . ( O ) Germ-band-retracted embryo exhibiting ubiquitous staining with striated expression evident . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 01610 . 7554/eLife . 01440 . 017Figure 7—figure supplement 1 . Phylogenetic analysis of hairy . Phylogenetic analysis of hairy-related protein sequences was performed using Maximum likelihood analysis with 1000-fold bootstrap support , using RAxML via the CIPRES portal ( Copf et al . , 2003; Nagy et al . , 1994 ) . The best scoring maximum likelihood tree is shown with bootstrap support values indicated adjacent to branch . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 01710 . 7554/eLife . 01440 . 018Figure 7—figure supplement 2 . Hairy protein sequence alignment . Predicted hairy-like protein sequences from Nasonia were aligned to hairy-like proteins from Drosophila melanogaster , Anopheles gambiae , Apis mellifera , and Tribolium castaneum using ClustalW . Full-length sequences were used for the alignment . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 018 Double fluorescent in situ hybridization with Nv eve reveals that early Nv h overlaps the anterior of Nv eve double-segment stripes ( Figure 6M–O ) . Later , these stripes appear thinner , and they overlap the anterior Nv eve single-segment odd-numbered stripes in each double-segment doublet . Nv h double-segment stripe 6 anticipates the position of the most anterior derivative of the Nv eve stripe 6 quartet ( segmental stripes 11–14 ) . Nv h stripe 7 ( double segment ) , which is thin , appears within the Nv eve early broad stripe 6 domain , coinciding with Nv eve single-segment stripe 13 ( Figure 6P–R , Figure 8 ) . A more posterior stripe , Nv h 8 , anticipates , albeit more broadly , the site of Nv eve single-segment stripe 15 ( Figure 6M–O , P–R , Figure 8 ) . Thus , Nv h and Nv eve are co-expressed at the anterior of eve pair-rule stripes and in the first of each pair of eve ( odd-numbered ) single-segment stripes , similar to the relationship described for Drosophila eve and hairy as they initiate segment polarity ( Warrior and Levine , 1990 ) . 10 . 7554/eLife . 01440 . 019Figure 8 . Summary model of pair-rule gene expression in the Nasonia embryo . ( A ) Model of register of pair-rule gene expression in the early embryo . Nv eve and Nv odd stripes are totally complementary , whereas Nv runt stripes partly overlap each of these genes at their interface . Nv hairy stripes overlap Nv eve stripes toward the anterior of each double-segment periodicity stripe . Towards the posterior of the embryo , an extended domain of low-level Nv odd expression exhibits dynamic behavior over several nuclear cycles , and stripes 4/5 of Nv eve , Nv odd , and Nv runt each differentiate during this interval . Even more posteriorly , Nv eve stripe 6 lay anterior to a continuous Nv odd cap that extends to the posterior pole of the embryo . This region is set aside for segment specification and differentiation during germ-band extension . ( B ) Model of register of pair-rule gene expression in the germ-band extension ( late ) embryo . Single-segment periodicity stripes in the germ-band-extended embryo exhibit a variation upon early gene expression patterns . Nv eve single-segment stripes are interrupted by Nv runt and then Nv odd such that Nv runt stripes follow odd-numbered Nv eve stripes , and Nv odd stripes follow even-numbered Nv eve stripes . Each of 8 Nv hairy stripes overlaps odd-numbered Nv eve stripes that derive from the anterior of Nv eve pair-rule stripes . Additional expression of several of these genes in the ventral and head domains , which appears to rely on different regulatory logic , is not shown . DOI: http://dx . doi . org/10 . 7554/eLife . 01440 . 019 We knocked down Nv hairy function using two independent morpholinos directed at two different splice junctions . Both resulted in a range of cuticle defects that indicate that Nv h is required for the formation of all posterior-derived segments and blastoderm-derived segments in the thorax and anterior abdomen ( Figure 4L–O ) . At the posterior , mildly affected cuticles exhibit fusion of segments A9–10 ( Figure 4L , N ) , along with partial loss of alternating abdominal segments posterior to A4 . In more affected cuticles , alternating segments posterior to A2 are fused ( Figure 4L–N ) , resembling a pair-rule phenotype . In severely affected embryos , all segments from A4–A10 are fused with a continuous lawn of denticles that covers the posterior of a severely reduced cuticle ( Figure 4O , yellow line ) . These phenotypes suggest a requirement for Nv hairy in specification of posterior segments . The late Nv h stripes 6–8 are positioned to affect the late forming segments as supported by double in situs ( Figure 6 ) . In spite of Nv hairy expression in the head and extreme anterior of the embryo , labral structures in Nv hairy morpholino cuticles appear to be unaffected . The expression pattern of Nv hairy is thus strikingly similar to Tc’hairy ( Sommer and Tautz , 1993; Aranda et al . , 2008 ) , though functionally quite different , since Tc’h seems to act exclusively in patterning head segments ( Choe et al . , 2006; Aranda et al . , 2008 ) . At the anterior , A2 is also nearly always affected , exhibiting loss of denticles and displacement or loss of the associated spiracle ( Figure 4L , orange arrowhead ) . Segment T1 appears to be lost and fused to T2 . Finally , more affected embryos show a loss of T3 ( Figure 4M–O ) . Therefore , segments T1 , T3 , and A2 are missing , which resembles an anterior pair-rule phenotype . In summary , Nv h expression is highly dynamic and proceeds in an anterior to posterior progression . It is distinct from Nv eve and does not exhibit single-segment periodicity stripes . At the end of embryogenesis , its expression becomes nearly ubiquitous ( Figure 7O ) . Taken together , these data support a model wherein ‘pair-rule’ genes have weak fly-type ‘pair-rule’ phenotypes in the anterior , and are required for the formation of a suite of posterior segments . Their interdigitated expression suggests extensive interactions during patterning of the posterior region after cellularization , although this has not yet been tested due to current experimental limitations . Our summary model of the phasing of ‘pair-rule’ stripes in the embryo is given in Figure 8 .
In contrast to Drosophila , whose pair-rule genes are expressed in the blastoderm in seven double-segment periodicity stripes to determine the formation of 14 segments , their orthologs in Nasonia are expressed in diverse and more intricate patterns . In no case do we observe simply eight precellularization double-segment stripes , confirming that pair ‘rule’ does not represent the regulatory dynamics of these genes across insects . We observe wave-like behavior of Nv odd stripe 4–6 , which underscores that cycling control may remain from the ancestral segmentation clock . Nv runt and , to a large degree , Nv h , also exhibit a sequential progression of sharp stripe appearance that may be responsive to the waves of Nv odd ( Figure 7 , Figure 6—figure supplement 1 ) . It is clear that Nv eve , Nv odd , and Nv h exhibit multiple modes of regulation during embryogenesis . In each case , their anterior stripes formed in the syncytial blastoderm have double-segment periodicity and arise in a manner that could be explained by the type of enhancer logic exemplified by Drosophila eve ( Small and Levine , 1991; Small et al . , 1992 , 1996; Schroeder et al . , 2004; Schroeder et al . , 2011 ) . Anterior double-segment ‘pair-rule’ stripes of Nv eve appear to be regulated by maternal and gap genes as in Drosophila , and embryos knocked down for Nv eve , Nv odd and Nv h exhibit a pair-rule phenotype in the blastoderm-derived segments , although this phenotype is most often limited . It is worth noting that the severe Nv eve anterior defects are more regional , as observed for Oncopeltus eve that behaves as a gap gene , and is attributed to the broad early domain of expression ( Liu and Kaufman , 2005a ) . Perhaps , as in Oncopeltus , Nv eve is required in combination with Nv hb or Nv gt for activation of their targets , which are in turn required for the proper formation of head and thoracic segments . Another mode must control the formation of stripes of Nv eve , Nv odd , and Nv h that arise later , in a cellular environment , from a posterior domain ( whether at the posterior pole of the embryo , as in the case of Nv odd and Nv hairy , or from a broad posterior stripe , as in the case of Nv eve ) . Knockdown of each of the three genes produces a severe posterior truncation of the embryo , deleting all six posterior segments , indicating that each gene is required for the formation of posterior-derived segments . This phenotype is unlike flies , and resembles more the short germ pair-rule gene circuit of Tribolium ( Choe et al . , 2006; Choe and Brown , 2009; Sarrazin et al . , 2012 ) . Together with co-expression data , their phenotypes suggest that each of these genes is required for refinement or maintenance of each other’s activity or expression ( Figure 6 , Figure 6—figure supplement 1 ) . We propose a model in which interactions among ‘pair-rule’ genes dominate in patterning the long germ Nasonia embryo . Unlike flies , posterior stripes of ‘pair-rule’ genes like Nv hairy and Nv runt appear sequentially . Indeed , the gene circuit involving interactions among Tc’odd , Tc’eve and Tc’runt in each round of posterior segment formation underscores the likely ancestral nature of this network , which might have been brought under the control of the gap and maternal genes in flies , and in the anterior segments of Nasonia . The ‘waves’ of Nv odd pair-rule stripe expression that give rise to blastoderm stripes 4 , 5 , and 6 suggest residual activity of a segmentation clock . The presumptive domain of six posterior segments indicated by early posterior expression of Nv eve and Nv odd may be similar to the ‘growth zone’ of short germ insects . Waves of Nv odd 4 , 5 , and 6 , and the sequential formation of Nv runt stripes both interrupt and likely pattern the continuous Nv eve stripe 6 domain ( Figure 6A–L; Figure 8 ) . Nv h expression anticipates the final position of several late forming Nv eve stripes , and in combination with the phenotype of Nv h knockdown and co-expression data , suggests that it is required for the formation of the posterior Nv eve stripes . Thus , Nasonia represents a variation on embryo allocation and patterning , but the contribution of ‘pair-rule’ gene function is enduring . The use of a clock-like mechanism is not incompatible with long germ embryogenesis , and rather , retaining this character might allow for sampling transitional states between the short- and long germ strategies , which likely occurred multiple times within holometabola . Further , it may be that the absence of significant posterior elongation is the transition state that tips the balance toward elaboration of anterior segmentation control mechanisms and loss of late forming segments . Our characterization of the Nasonia pair-rule genes illustrates one way that these two strategies can co-exist . It is also of note that although hairy-related genes are the oscillating components of vertebrate segmentation clocks ( Palmeirim et al . , 1997 ) , it is generally odd-skipped-related genes that oscillate in arthropods ( Chipman et al . , 2004; Chipman and Akam , 2008; El-Sherif et al . , 2012; Sarrazin et al . , 2012 ) . Notch-signaling has been described for its involvement in regulating hairy-related oscillations in the vertebrate clock ( e . g . , Jouve et al . , 2000 ) , and it may also be involved in the context of the arthropod segmentation clock ( Stollewerk et al . , 2003; Eriksson et al . , 2013; Kainz et al . , 2011 ) ; in all cases , the driver of the clock is yet to be elucidated ( reviewed in Pourquie , 2003 ) . eve has been suggested to have its most ancestral function as a specifier of posteriorness ( Ahringer , 1996; Brown et al . , 1997 ) . Indeed , the two mammalian eve genes are located at the most ‘posterior’ end of two of the Hox clusters ( Bastian et al . , 1992 ) , although eve is not a part of the Hox cluster in Nasonia or Tribolium or any other insects that have been studied ( Shippy et al . , 2008; Werren et al . , 2010; Munoz-Torres , 2009; Suen et al . , 2011 ) . Yet , even in Drosophila , where there is no apparent sequential segmentation , a delayed pair-rule stripe ( stripe 8 ) appears early in gastrulation ( Macdonald et al . , 1986; Frasch et al . , 1987; Kim et al . , 2000 ) . In Schistocerca , eve is expressed in a posterior mesodermal domain and no pair-rule stripes arise from this region , indicating that eve plays a role in posteriorness and not segmentation in basal insects ( Patel et al . , 1992 ) . Nasonia eve sets aside stripe 6 relatively early , at about the same time as Nv odd that is expressed even more posteriorly . This feature of Nv eve in posterior segmentation is not shared by the other pair-rule genes we studied , therefore supporting the notion of an ancestral role for eve in posteriorness in Nasonia . That its expression is complementary to that of odd in both Tribolium and Nasonia in a late-differentiating posterior region may hint at how this role in posteriorness evolved into a role in posterior growth . In non-insect arthropods , there is evidence for a role for eve in both posterior identity and segmentation . The centipede Lithobius atkinsoni expresses eve in a posterior domain and between segments ( Hughes and Kaufman , 2002 ) , and the crustacean Artemia franciscana exhibits growth zone eve expression that precedes expression in stripes in emerging segments ( Copf et al . , 2003 ) . In other basal arthropods , like spiders ( Damen et al . , 2000 ) and the centipede Strigamia maritima ( Chipman and Akam , 2008 ) , eve expression in stripes suggests that its ancestral role is segmental . The broad stripe 6 domain of eve appears to be subdivided by transcription control , likely through interactions with Nv odd and Nv runt ( Figure 6 , Figure 6—figure supplement 1 ) . Although we observed apparent mitotic domains in the early gastrula that proceed from anterior to posterior , they do not match the pattern of initial differentiation of germ band-derived segments or the splitting of anterior pair-rule stripes . Cell division patterns in mitotic domains have not been described in most insects , apart from Drosophila and the precellular blastoderm of Bombyx ( order: Lepidoptera; Nagy et al . , 1994 ) . In the short germ Tribolium and Oncopeltus embryos , cell divisions during gastrulation and elongation occur throughout the germ band , with no evidence for mitotic domains ( Brown et al . , 1994; Liu and Kaufman , 2009 ) . The relationship between Nv eve and cell division suggests coordination of cell divisions by segmentation genes , a phenomenon that has been suggested for Drosophila ( Foe , 1989; Edgar and O’Farrell , 1989; Bianchi-Frias et al . , 2004 ) . Use of coordinated mitotic domains is a strategy that seems to have evolved multiple times ( e . g . , in flies and wasps ) . We propose that the apparent coordination of mitotic domains and segmentation gene expression of both Nasonia and Drosophila development may constitute a step in the transition to long germ embryogenesis . In summary , despite obvious differences in their expression patterns , Nasonia eve , odd , and hairy function in both blastoderm- and germ band-derived segment formation . While Nasonia exhibits fly-type expression of maternal and gap genes in the precellular blastoderm , dynamic expression patterns and extensive interactions among ‘pair-rule’ genes appear to pattern a suite of late forming posterior segments . Indeed , their relative expression patterns suggest that the regulation of posterior segments may be through the type of mutual regulation described for the pair-rule gene circuit of Tribolium . This is unlike the long germ embryo of Drosophila , whose segmentation utilizes pair-rule interactions only during the late blastoderm stage . We propose that late-forming segments are set aside using remnants of ancestral control of posteriorness and the segmentation clock . Thus , Nasonia relies on a dynamic , dual mode of segmentation that has characteristics of both ancestral short germ and derived long germ embryogenesis .
Nasonia embryos were collected and fixed in 5% formaldehyde/1X PBS/Heptane for 28 min , affixed to double-sided tape , and hand peeled under 1X PBS +0 . 1% tween , as described previously ( Pultz et al . , 2005 ) , except that the embryos were collected from host-fed , mated females . The embryos were stored under methanol at −20°C between fixation and hybridization . In situ hybridizations were carried out as described previously ( Pultz et al . , 2005 ) . Briefly , the embryos stored under methanol were gradually brought up to 1X PBT and washed three times in 1x PBS +0 . 1% tween-20 ( PBT ) before a 30-min post-fix step in 5% formaldehyde/1XPBT . The embryos were then washed three times and subjected to proteinase K treatment ( final concentration of 4 μg/ml ) before three PBT washes . The embryos were blocked for 1 hr in hybridization buffer before probe preparation and addition for overnight incubation at 65°C . The next day , the embryos were washed in formamide wash buffer and then 1X MABT buffer before blocking in 2% Blocking Reagent ( BBR; Roche Applied Science , Germany ) in 1X MABT for 1 hr , then in 10% horse serum/2% BBR/1XMABT for 2 hr . The embryos were incubated overnight with primary antibody in the second blocking solution at 4°C . Anti-DIG-AP Fab fragments ( Roche Diagnostics ) were used at 1:2000 for non-fluorescent in situs . On the third day , the embryos were washed in 1X MABT for ten , 20 min washes before equilibrating the embryos in AP staining buffer and developing in AP buffer with NBT/BCIP solution ( Roche Diagnostics ) . After staining , the embryos were washed in 1× PBT three times for 5 min each before a single 25 min post-fix step in 5% formaldehyde/1XPBT . The embryos were then washed briefly and allowed to sink in 50% and then 70% glycerol/1XPBS , which were subsequently used for mounting . For fluorescent in situs , DIG probes were detected using anti-DIG-POD Fab fragments ( Roche Diagnostics ) at 1:50 dilution , followed by FastRed HNPP detection system ( Roche Diagnostics ) , according to manufacturer’s instructions . Fluorescein probes were detected using anti-Fluorescein-AP Fab fragments ( Roche Diagnostics ) at 1:500 dilution . For antibody staining of mitotic cells , we used a rabbit anti-phosphorylated histone H3 serine 10 antibody ( Millipore , Billerica , Massachusetts ) at 1:200 , and then a donkey anti-rabbit secondary conjugated to Alexa-647 ( Invitrogen , Carlsbad , California ) at 1:200 . In combination with FastRed in situ detection using anti-DIG-AP Fab fragments ( Roche Diagnostics ) at 1:500 , primary antibodies were added to blocking buffer together and incubated according to the in situ protocol , and secondary antibody detection was carried out after the FastRed staining was completed and following three 1X PBT washes . Nasonia pair-rule gene cDNAs were cloned from embryo cDNA pools generated from reverse transcription of total embryo RNA from mixed age embryos using Superscript First Strand Synthesis kit ( with Superscript II; Invitrogen ) according to manufacturer’s specifications . For cases in which long cDNA sequences could not be amplified with oligos designed according to automated genome annotation and prediction models , we used circular RACE to simultaneously amplify sequences 5′ and 3′ to smaller cloned cDNA fragments , as previously described ( McGrath , 2011 ) . Nasonia paralogs of fly pair-rule genes were identified by TBLASTN and aligned against predicted or experimentally validated ( virtually translated ) protein sequences of the same proteins from Tribolium castaneum ( Tcas ) , Anopheles gambiae ( Agam ) , Apis mellifera ( Amel ) . Protein sequences were aligned using CLUSTALW ( Larkin et al . , 2007 ) and rendered using Dendroscope ( Huson et al . , 2007 ) . Evolutionary relationships were inferred using a maximum likelihood analysis with 1000-fold bootstrap support , via RaXML hosted online at CIPRES science gateway ( http://www . phylo . org/index . php/portal/ ) ( Stamatakis et al . , 2008; Stamatakis , 2006; Miller et al . , 2010 ) . Antisense morpholinos targeting splice junctions or transcription initiation sites were designed and ordered from GeneTools , LLC ( www . gene-tools . com , Philomath , Oregon ) . Lyophilized morpholinos were resuspended in sterile nuclease-free water to a final concentration of 5 mM . For Nv odd splice block morpholino , which yielded a high percentage of dead embryos with no cuticle , injections were also carried out at 1 mM , 0 . 5 mM , and 0 . 05 mM dilutions . Nasonia embryos were collected for 35 min at 28°C and dehydrated for 30 min before injection with morpholinos ( approximate volume injected = 0 . 001 μl per embryo ) . The embryos were allowed to develop on the injection membrane at 28°C on a 1X PBS/1% agarose plate for approximately 30 hr , to ensure complete development . Unhatched larvae were peeled and transferred to a slide for cuticle preparations in Lacto–Hoyer’s media . Morpholino sequences used are as follows:Eve translation block 5′ CAAAGCTCCTCTGGAATCCTTGCAT 3′Eve E2I2 splice block: 5′ AAACGATAGTTACCTTGATGGTCGA 3′Hairy E2I2 splice block: 5′ CTGAATCTGTCAAGATACTTACGTC 3′Hairy E1I1 splice block: 5′ GAGCAAGTCGAGATACTAACCCGTCOdd splice block: 5′ AGAGAGTGTACTAAC TTGTGGTCCC 3′Odd translation block: 5′ GCTCCATCGCAAGCTGGGTAAACGT 3′ Nv odd cDNA GenBank Accession # KC142194Nv eve cDNA , isoform 1 GenBank Accession# KC168090Nv eve cDNA , isoform 2 GenBank Accession# KC168091Nv eve cDNA , isoform 3 GenBank Accession# KC168092Nv hairy cDNA GenBank Accession # KC190514 Accession numbers for sequences used in sequence alignments and trees:NvitH1: NP_001267498 XP_001601817 ( Nvit Hairy ) NvitH2: Uniprot K7J0X7_NASVI ( hairy-like/Nvit Dpn ) NvitH3: XP_001601600 . 2 GI:345484850 ( hairy-like/HES like ? ) DmelH: NP_523977 . 2 GI:24661088 ( Dmel hairy ) DmelDpn: NP_476923 . 1 GI:17136808 ( Dmel deadpan ) AmelH1: XP_001120814 . 2 GI:328784100 ( hypothetical protein ) AmelH2: XP_393948 . 3 GI:110762302 ( hairy-like ) AgambH1: XP_316733 . 3 GI:158296333 ( corrected seq; Agam hairy ) AgambH2: XP_320206 . 4 GI:158300226TcasH1: NP_001107765 . 1 GI:166796106 ( Tcas Hairy ) TcasH2: XP_967694 . 1 GI:91092620 ( Tcas similar to GA21268-PA ) TcasH3: XP_975187 . 1 GI:91083981 ( Tcas HES1 ) DmelOddsk: NP_722922 . 1 GI:24581484 ( Dmel Odd skipped ) DmelSobow: NP_476882 . 1 GI:17136746 ( Dmel Sister of odd and bowl ) DmelBowl: NP_476883 . 1 GI:17136748 ( Dmel Brother of odd with entrails limited ) NvitOddbowlA: XP_001603713 . 1 GI:156545195 ( predicted protein ) NvitOddbowlB: XP_001603827 . 2 GI:345481739 ( Nv bowel-like ) NvitOddbowlC: XP_001603660 . 1 GI:156545193 ( Nv odd-skipped like ) AmelOddbowlA: XP_001120949 . 1 GI:110762343 ( Amel odd-skipped like ) AmelOddbowlB: XP_393879 . 3 GI:110762378 ( Amel bowel-like ) AmelOddbowlC: XP_001120905 . 1 GI:110762341TcasBowl: XP_972138 . 2 GI:189240088 ( Tcas bowl-like ) TcasOdd: XP_972086 . 2 GI:189240086 ( Tcas odd-skipped ) TcasSob: XP_972035 . 1 GI:91088523 ( Tcas: predicted sister of odd and bowl ) Agam7972_PA: XP_306979 . 3 GI:118776890Agam7973_PA: XP_317495 . 3 GI:118789549Agam8222: XP_555242 . 1 GI:57914799 For Figure 4 , the phenotypic classes are approximately as follows:Eve: B 20 . 4% C 24% D 30 . 1% E 25% . Odd: G 27 . 7% H 22 . 3% I 29 . 2% J 20 . 8% . Hairy: L 13 . 4% M 25 . 0% N 36 . 6% O 25 . 0% . | Networks of genes that work together are widespread in nature . The conservation of individual genes across species and the tendency of their networks to stick together is a sign that they are working efficiently . Furthermore , it is common for existing gene networks to be adapted to perform new tasks , instead of new networks being invented every time a similar but distinct demand arises . One important question is: how can evolution use the same building blocks—such as the genes in a functioning network—in different ways to achieve new outcomes ? The gene network that sets up the ‘body plan’ of insects during development has been well studied , most deeply in the fruit fly , Drosophila . Like all insects , the body of a fruit fly is divided into three main parts—the head , the thorax and the abdomen—and each of these parts is made up of several smaller segments . There is a remarkable diversity of insect body plans in nature , and yet , these seem to arise from the same gene networks in the embryo . When a Drosophila embryo is growing into a larva , all the different body segments develop at the same time . In most other insects , however , segments of the abdomen emerge later and sequentially during the development process . The ancestors of most insects are also thought to have developed in this way , which is known as ‘short germ embryogenesis’ . So how did the so-called ‘long germ embryogenesis’ , as observed in Drosophila , evolve from the short germ embryogenesis that is observed in most other insects ? The gene network that controls development includes the ‘pair-rule genes’ that are expressed in a pattern of alternating stripes that wrap around , top to bottom , along most of the length of the embryo . These stripes mark where the edges of each body segment will eventually develop . In fruit flies , this pattern extends along the entire length of the embryo and the stripes all appear at one time . However , in the abdominal region of short germ insects , the pair-rule genes are expressed in waves that pass through the posterior region as it grows , with new segments being added one behind the other . Now , Rosenberg et al . have attempted to explain how the same genes can be used to direct the segmentation process in such different ways by studying another long germ insect species , the jewel wasp . Analysis of the expression of pair-rule genes in the jewel wasp shows that it uses a mixed strategy to control segmentation . The development of segments at the front of its body is directed in the same way as the fruit fly , with all these segments laid down together . However , the segments at the rear of the body are only patterned later , one after the other , like most other insects . The work of Rosenberg et al . suggests that the jewel wasp represents an intermediate step between ancestral insects and Drosophila in the evolution of the gene network that patterns the ‘body plan’ . Identifying and studying these intermediate forms allows us to understand the ways in which evolution can innovate by building upon what has come before . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"evolutionary",
"biology"
] | 2014 | Dual mode of embryonic development is highlighted by expression and function of Nasonia pair-rule genes |
Several different endocytic pathways have been proposed to function in mammalian cells . Clathrin-coated pits are well defined , but the identity , mechanism and function of alternative pathways have been controversial . Here we apply universal chemical labelling of plasma membrane proteins to define all primary endocytic vesicles , and labelling of specific proteins with a reducible SNAP-tag substrate . These approaches provide high temporal resolution and stringent discrimination between surface-connected and intracellular membranes . We find that at least 95% of the earliest detectable endocytic vesicles arise from clathrin-coated pits . GPI-anchored proteins , candidate cargoes for alternate pathways , are also found to enter the cell predominantly via coated pits . Experiments employing a mutated clathrin adaptor reveal distinct mechanisms for sorting into coated pits , and thereby explain differential effects on the uptake of transferrin and GPI-anchored proteins . These data call for a revision of models for the activity and diversity of endocytic pathways in mammalian cells .
Endocytosis has central roles in many cell biological processes ( Le Roy and Wrana , 2005; Doherty and McMahon , 2009 ) . Since the mid-nineties evidence has accumulated to suggest that mammalian cells utilise additional endocytic mechanisms beyond clathrin-coated pits ( Sandvig and van Deurs , 1994; Mayor and Pagano , 2007; Sandvig et al . , 2008 ) . Whilst the molecular detail of how clathrin-coated pits work is understood in ever-increasing detail ( Kirchhausen et al . , 2014 ) , similarly complete mechanistic descriptions of how endocytosis may take place outside of clathrin-coated pits are lacking . One central difficulty in defining clathrin-independent endocytic pathways has been the paucity of rigorously validated endocytic markers and pathway-specific cargoes . Dissection of clathrin-mediated endocytosis benefited greatly from signature cargoes such as the transferrin receptor , which are efficiently concentrated in the forming endocytic vesicle , and from the fact that the forming vesicle is marked specifically in space and time by transient assemblies of clathrin , adaptors and associated proteins ( Pearse , 1982; Dautry-Varsat et al . , 1983; Doxsey et al . , 1987; Ehrlich et al . , 2004 ) . By contrast , clathrin independent endocytosis has largely been defined by morphological criteria , and by the persistent uptake of cargoes that may utilise multiple pathways following perturbation of the clathrin machinery ( Saslowsky et al . , 2010; Engel et al . , 2011; Cho et al . , 2012 ) . Endocytic structures that can be defined morphologically include macropinosomes , which can readily be resolved by light microscopy ( Liberali et al . , 2008 ) , and caveolae which are distinctive in electron micrographs ( Parton and del Pozo , 2013 ) . The extent to which caveolae are involved in endocytosis is , however , controversial ( Parton and Howes , 2010 ) . Morphology also forms a large part of the definition of more recently characterised endocytic membranes termed CLICs , for clathrin-independent carriers ( Kirkham et al . , 2005; Howes et al . , 2010 ) . GPI-anchored proteins have been extensively studied as potential cargoes for clathrin-independent endocytosis ( Mayor and Pagano , 2007; Johannes and Mayor , 2010 ) . The apparent presence of these proteins in endosomes devoid of transferrin , and uptake in the presence of inhibitors of clathrin-coated pits , provides evidence for GPI-enriched endosomal compartments ( GEECs ) that are fed from the cell surface independently from coated pits ( Sabharanjak et al . , 2002; Kumari and Mayor , 2008; Bhagatji et al . , 2009 ) . It is not clear , however , that GPI-anchored proteins are ever highly concentrated in nascent endocytic vesicles in a manner analogous to transferrin receptor , so they may enter the cell via multiple mechanisms ( Mayor and Riezman , 2004; Sharma et al . , 2004; Bhagatji et al . , 2009; Johannes and Mayor , 2010 ) . The extent to which other types of cargo associated with clathrin-independent endocytosis , including glycosphingolipid-binding bacterial toxins as well as various viruses , are efficiently sorted during uptake is similarly unclear ( Romer et al . , 2007; Ewers et al . , 2010; Johannes and Mayor , 2010; Saslowsky et al . , 2010; Cho et al . , 2012; Lakshminarayan et al . , 2014 ) . In the absence of demonstrably specific cargoes , much of the literature on clathrin-independent endocytosis relies on the use of overexpressed mutant proteins to perturb clathrin function , and observation of differential effects on the uptake of transferrin and potential clathrin-independent cargoes . Dominant negative mutants used in this manner include the C-terminal clathrin-binding domain of AP180/CALM ( Ford et al . , 2001 ) , and the K44A mutant of dynamin , which renders this GTPase involved in the scission of clathrin coated pits enzymatically inactive ( van der Bliek et al . , 1993; Damke et al . , 1995 ) . Additionally , differential effects on endocytosis can be observed using overexpression of inactive forms of small GTPases such as ARF6 , ARF1 or cdc42 ( D'Souza-Schorey et al . , 1995; Palacios et al . , 2002; Sabharanjak et al . , 2002; Naslavsky et al . , 2004; Kumari and Mayor , 2008 ) . Blocking one type of endocytosis may up-regulate alternative mechanisms , over-expression of mutant proteins may induce non-physiological cellular responses , and small GTPases may , via different sets of effectors , directly or indirectly control the activity of multiple endocytic pathways . Plainly , these types of experiment need to be interpreted carefully . Ideally , different types of endocytosis would be defined by the presence of specific molecular determinants analogous to clathrin or adaptor proteins in the case of clathrin coated pits ( Kirchhausen et al . , 2014 ) . Candidates for such determinants include caveolin and cavin proteins which make caveolae ( Ludwig et al . , 2013; Parton and del Pozo , 2013 ) , flotillin proteins which define specific plasma membrane microdomains potentially involved in endocytosis ( Glebov et al . , 2006 ) , and the protein GRAF1 that may be important for the formation of CLIC/GEEC endosomes ( Lundmark et al . , 2008 ) . The precise mechanisms by which these proteins are involved in endocytosis remain to be fully understood . Both flotillins and caveolins plus cavins form protein assemblies that are stable over time , and thus can not define endocytic events temporally in the way in which coordinated assembly and disassembly of the clathrin machinery can ( Glebov et al . , 2006; Frick et al . , 2007; Taylor et al . , 2011; Gambin et al . , 2013; Ludwig et al . , 2013 ) . Thus , although several non-clathrin endocytic pathways have garnered varying degrees of supportive evidence , none has been unambiguously established in molecular and functional terms . Furthermore , the relative contributions of the multiple putative pathways to overall endocytic flux has been unclear . To resolve some of these uncertainties , ideally one would require a means to examine endocytosis in unperturbed cells in a global manner . This would allow simultaneous evaluation of a large number of cargoes and their relationship with clathrin and putative non-clathrin markers . In this study we apply a combination of new and established methods that satisfy these requirements , and thereby provide a systematic and quantitative analysis of total endocytic protein flux in cultured mammalian cells .
We sought to establish endocytic assays and protein labelling strategies to satisfy four main requirements: ( 1 ) achieve very high topological specificity in discrimination between endocytosed and extracellular protein , ( 2 ) provide a means to analyse all , or nearly all , surface proteins simultaneously , ( 3 ) provide high signal to noise , and thereby high temporal resolution for detection of primary endocytic vesicles , and ( 4 ) allow a means to follow uptake of specific cargoes . Biotinylation of extracellular free amine groups with the small , monovalent label sulfo-NHS-SS-biotin offered a way to satisfy the first three of these requirements ( Le Bivic et al . , 1990 ) . The biotin moiety can be removed from labelled proteins by reduction of the disulfide bond with membrane-impermeant sodium 2-mercaptoethanesulfonate ( MESNa ) at 4°C , so this approach provides a powerful way to detect internalisation of surface proteins ( Bretscher and Lutter , 1988; Schmid and Smythe , 1991 ) . We assessed the efficiency of MESNa treatment in removing biotin from sulfo-NHS-SS-biotin-labelled proteins , and compared this with two widely used alternatives , washing at low pH to remove antibodies bound to the outside of the cell and cleavage of extracellular GPI-anchors with PI-PLC ( Sabharanjak et al . , 2002 ) . MESNa treatment could remove over 99 . 9% of biotin from cells labelled at 4°C ( Figure 1A ) . Washes at pH3 to remove bound antibody , or PI-PLC to cleave GPI-anchors , were around two orders of magnitude less efficient , removing up to 50% and 90% of antibody bound to the GPI-anchored protein CD59 respectively ( Figure 1B ) ( Davies et al . , 1989 ) . Reducible biotin and MESNa therefore offer a highly accurate way of assaying protein internalisation . 10 . 7554/eLife . 03970 . 003Figure 1 . Experimental strategy and assay validation . ( A ) Following surface biotinylation with sulfo-NHS-SS-biotin , HeLa cell lysate was serially diluted with non-biotinylated control cell lysate , and the amount of labelling was detected using streptavidin-HRP after SDS-PAGE and blotting . Incubation with 100 mM MESNa prior to cell lysis was enough to remove more than 99 . 9% of the initial signal . ( B ) Lysate from cells labelled with anti-CD59 antibody was serially diluted with non-labelled control cell lysate , and the amount of labelling was detected using an anti-mouse HRP antibody . Acid wash removed around 50% of the initial surface signal . PI-PLC , which cleaves the GPI-anchor in CD59 , removed up to 90% of the initial signal . ( C ) Cartoon to illustrate endocytosis assay . Cell surface proteins were labelled with sulfo-NHS-SS-biotin at 4°C , then the reaction was quenched and cells were rapidly transferred to 37°C to allow endocytosis . Control cells were kept at 4°C . After defined time for endocytosis , cells were rapidly returned to 4°C , and surface-exposed biotin was removed using the membrane impermeable reducing agent MESNa . Biotin label was detected using fluorescent streptavidin , after fixation and permeabilisation . ( D ) Confocal images from control experiment to demonstrate total surface labelling with sulfo-NHS-SS-biotin , negative control , and labelling of endocytic vesicles , as illustrated in C . Bars are 20 μm . ( E ) Silver stained SDS-PAGE gel following surface biotinylation with sulfo-NHS-SS-biotin and streptavidin pull-down . Non-biotinylated cells provided a negative control . ( F ) Surface biotinylation labels the full range of plasma membrane proteins ranked according to their relative abundance in the plasma membrane . The relative abundance of plasma membrane proteins in HeLa cells was calculated based on a previously published study ( Kulak et al . , 2014 ) . These are represented graphically ranked by abundance in blue , while proteins were detected in our experiments are shown in magenta . The data are listed in Figure 1—source data 1 . ( G ) Confocal images of HeLa cells stably expressing SNAP-GPI . Cells labelled at 4°C with BG-SS-488 . Control cells were treated with MESNa to remove external fluorophore after incubation only at 4°C . Warming to 37°C for 90 s before MESNa treatment at 4°C allows identification of endocytic vesicles . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00310 . 7554/eLife . 03970 . 004Figure 1—source data 1 . Plasma membrane proteins identified by mass spectrometry . Biotinylated proteins identified in this study are listed by accession number , together with the estimated abundance takenfrom Kulak et al . , 2014 . A zero in the biotinylated proteins column indicates that we did not identify that protein . Calculation of PM abundance is described in the methods section . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00410 . 7554/eLife . 03970 . 005Figure 1—figure supplement 1 . Removal of extracellular fluorophore from BG-SS-fluorophore labelled SNAP-tag by reduction with MESNa is highly efficient . ( A ) Cells expressing SNAP-GPI were simultaneously labeled at 4°C with BG-SS-488 and SNAPsurface549 . The latter is not reducible . They were then treated with PI-PLC or PI-PLC and then MESNa . Both treatments reduced the amount of label significantly ( top row of images ) . However when remaining cell surface label was inspected by adjusting the intensity post-acquisition ( lower panels ) , differential effects in the removal of the two fluorophores were observed . PI-PLC reduced the signal of both fluorophores evenly . However , when cells were treated with PI-PLC followed by MESNa the levels of reducible BG-SS-488 become practically undetectable . ( B ) Quantification of images from the same experiment shown in A . Images used for quantification were confocal cross sections of cells . Line profiles across the plasma membrane of the cell were combined to produce average traces of fluorescence intensity . The removal of BG-SS-488 in the MESNa + PI-PLC treated cells is clearly more efficient than removal of SNAP-surface549 . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 005 To detect internalised protein in confocal images , we labelled HeLa cells with sulfo-NHS-SS-biotin , allowed internalisation for defined periods of time , removed extracellular biotin by reduction with MESNa , and used fluorescent streptavidin to detect intracellular biotin . All subsequent experiments are also in HeLa cells unless otherwise stated . Control experiments comparing incubation at 4°C with incubation for defined periods at 37°C confirmed that there was no specific signal detected unless internalisation at 37°C was allowed to take place ( Figure 1C , Figure 1D ) . Endocytic vesicles could be observed after as little as 20 s of internalisation . This is a significant improvement in time resolution , and hence in our confidence that the detected vesicles have just budded from the plasma membrane ( Sabharanjak et al . , 2002; Kirkham et al . , 2005; Glebov et al . , 2006 ) . In order to interpret uptake of proteins labelled with sulfo-NHS-SS-biotin with confidence it was necessary to confirm that the label reacts with a broad range of representative proteins . Following surface biotinylation , proteins were precipitated using streptavin-agarose beads , and eluted from the beads by reduction with 100 mM DTT . Silver staining revealed that biotinylated proteins were precipitated with high specificity ( Figure 1E ) . LC-MS identified the full list of precipitated proteins , and this list was compared with a quantitative analysis of the HeLa cell proteome ( Kulak et al . , 2014 ) . Abundant plasma membrane proteins had higher chances of being detected . Nevertheless , a wide range of rare proteins ( some with estimated copy number < 100 per cell ) was detected , assuring us that we achieved good coverage of the total plasma membrane protein complement . ( Figure 1F , Figure 1—source data 1 ) . Therefore chemical labelling with sulfo-NHS-SS-biotin does indeed provide a way to follow endocytosis of nearly all plasma membrane proteins simultaneously . Reduction of extracellular disulphide bonds with membrane impermeant reducing agents like MESNa evidently provides a good way to discriminate between intra- and extra-cellular substrates . We used the genetically encoded SNAP-tag to apply this approach to specific cargo proteins ( Gautier et al . , 2008; Cole and Donaldson , 2012; Correa , 2014 ) . The SNAP moiety was labelled with a new membrane-impermeant , reducible , fluorescent SNAP-labelling reagent , benzylguanine-SS-fluorophore [where the fluorophore can be either atto-488 ( BG-SS-488 ) or BODIPY ( BG-SS-549 ) ] . The disulfide bond linking benzylguanine to a fluorophore can be reduced with MESNa in the same way as that in sulfo-NHS-SS-biotin . Control experiments using a minimal GPI-anchored SNAP-tag construct as a model plasma membrane protein showed that BG-SS-fluorophore allows detection of internalised SNAP tag with very high efficiency and low background , and provides an improved way of discriminating between internal and external pools of tagged protein ( Figure 1G , Figure 1—figure supplement 1 ) . Global labelling of primary endocytic vesicles was achieved by incubation of cells with sulfo-NHS-SS-biotin at 4°C , warming to 37°C by rapid buffer exchange , and allowing uptake for 20 s . Endocytic vesicles detected after MESNa treatment were small puncta ( Figure 2A ) . Approximately 2% of the cells also contained distinctively larger macropinosomes over 500 nm in diameter ( Figure 2—figure supplement 1 ) . In order to determine the identity of the endocytic vesicles we carried out co-internalisation experiments with the archetypical high-affinity cargo for clathrin-coated pits , transferrin ( Figure 2A , Figure 2—figure supplement 2 ) ( Schmid and Smythe , 1991; Hansen et al . , 1992 ) . The great majority of endocytic vesicles contained transferrin , and therefore are likely to have arisen from clathrin-coated pits . Uptake of transferrin was solely due to binding to transferrin receptor , as large , fluid-filled macropinosomes were completely devoid of transferrin ( Figure 2—figure supplement 1 ) . Residual surface-bound transferrin was detected as puncta , and co-localised with clathrin , implying concentration in nascent coated pits ( Figure 2—figure supplement 2 ) . We used unbiased and semi-automated quantification , involving application of a mask derived from the image of fluorescent transferrin , to determine the proportion of biotin-positive pixels that also contained transferrin in multiple cell images ( Figure 2—figure supplement 3 ) . This revealed that over 95% of the total internalised biotin after 20 s uptake was present in transferrin-positive vesicles ( Figure 2B ) . Co-localisation remained constant when times of internalisation up to 150 s were assayed ( Figure 2B , Figure 2—figure supplement 4 ) . Additionally , after 20 s uptake over 80% of biotin-positive endocytic vesicles co-localised with clathrin in small puncta , defining them as clathrin-coated vesicles ( Figure 2C , Figure 2D , Figure 2—figure supplement 5 ) . The proportion co-localising with clathrin fell rapidly with longer periods of uptake , as one would expect due to uncoating of primary vesicles ( Figure 2D , Figure 2—figure supplement 5 ) ( Kirchhausen et al . , 2014 ) . 10 . 7554/eLife . 03970 . 006Figure 2 . Over 95% of total endocytosed protein co-localises with markers for clathrin-mediated endocytosis . ( A ) Confocal images of co-internalisation of all membrane proteins , labelled at 4°C with sulfo-NHS-SS-biotin , and transferrin-546 . Internalisation was for 20 s at 37°C . Biotin was detected with streptavidin-488 . Note that labelling with biotin and transferrin at 4°C was carried out consecutively , so transferrin was not biotinylated . The presence of external transferrin in nascent coated pits explains transferrin-positive , strepatavidin-negative puncta ( see Figure 2—figure supplement 2 ) . Zoomed in area in the lower panels is indicated with a box . Bar is 20 μm . ( B ) Quantification of proportion of internalised protein , detected as in A , that co-localises with transferrin . Bars are mean and SD , the data are from one experiment that was repeated three times with the same overall result . ( C ) Confocal images of total endocytosed membrane protein , labelled at 4°C with sulfo-NHS-SS-biotin as shown , and indirect immunofluorescence staining for clathrin heavy chain . Internalisation was for 20 s at 37°C . Biotin was detected with streptavidin-488 . ( D ) Quantification of proportion of total internalised protein that co-localises with clathrin , at the times of internalisation indicated . Bars are mean and SD , the data are from one experiment that was repeated three times with the same result . ( E ) Confocal images of co-internalisation of total membrane protein and transferrin-546 , with 90 s at 37°C for labelling and uptake as illustrated in the cartoon . Bar is 20 μm . ( F ) Quantification of proportion of internalised protein , detected as in E , that co-localises with transferrin , at the times for continuous labelling and internalisation indicated . Bars are mean and SD , the data are from one experiment that was repeated three times with the same result . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00610 . 7554/eLife . 03970 . 007Figure 2—figure supplement 1 . Macropinosomes are readily identified by labelling with sulfo-NHS-SS-biotin . Cells were labeled with sulfo-NHS-SS-biotin and transferrin-546 , allowed to endocytose at 37°C for the times indicated , MESNA-treated , fixed , and stained with streptavidin-488 . Occasionally , cells contained larger endocytic structures ( >0 . 5 µm ) that were intensely labeled with fluorescent streptavidin but did not contain transferrin , despite the presence of transferrin in the medium during incubation at 37°C . These structures correspond to macropinosomes . No clear correlation between their abundance and incubation time at 37°C was observed . Bar is 15 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00710 . 7554/eLife . 03970 . 008Figure 2—figure supplement 2 . Surface-bound transferrin is highly concentrated within clathrin-coated pits . Cells were transfected with clathrin light chain-GFP , cooled to 4°C , labelled with transferrin-546 , fixed , and imaged . Bar is 20 μm . The boxed region is shown in the lower panels . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00810 . 7554/eLife . 03970 . 009Figure 2—figure supplement 3 . Quantification of percent co-localisation . All processing was carried out in Image J . Two channel raw images were acquired by confocal microscopy . The channels were separated , subjected to Gaussian blur with σ = 0 . 7 , and then contrast adjusted using the histogram of pixel intensities as shown . In the streptavidin/biotin channel , the base of the histogram was used to set pixel intensity = 0 , maximal pixel intensity was not altered . In the transferrin channel , which is used to generate a binary mask , pixel intensity = 0 and maximal pixel intensity were both set to the base of the histogram of pixel intensities as shown . Following dilation of positive pixels in the binary mask a logical ‘AND’ operation was carried out to isolate those pixels in the streptavidin channel that also are positive in the transferrin binary mask . This image was combined with the original biotin image in a two colour overlay , and manually drawn regions of interest were used to calculate total pixel intensity in the biotin channel , and total pixel intensity in the same channel from transferrin-positive pixels . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 00910 . 7554/eLife . 03970 . 010Figure 2—figure supplement 4 . Co-localisation between internalised sulfo-NHS-SS-biotin and transferrin after labelling at 4°C and 90 s of internalisation at 37°C . Confocal images of co-internalisation of total membrane protein , labeled at 4°C with sulfo-NHS-SS-biotin , and transferrin-546 . Internalisation was for 90 s at 37°C . Biotin was detected with streptavidin-488 after MESNA treatment . Note that labelling with biotin and transferrin at 4°C was carried out consecutively , so transferrin was not biotinylated . Zoomed in area of the lower panel is indicated with a box . Bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01010 . 7554/eLife . 03970 . 011Figure 2—figure supplement 5 . Co-localisation between internalised sulfo-NHS-SS-biotin and clathrin after 20 s and 60 s of internalisation at 37°C . HeLa cells labeled at 4°C with sulfo-NHS-SS-biotin , were moved to 37°C for the indicated time-points . Surface biotin was removed by MESNA treatment , the cells were fixed and permeabilised , and then stained with streptavidin-488 , and antibodies against clathrin heavy chain by indirect immunofluorescence . Box indicates zoomed region , bars are 20 μm . 20 s image also shown in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01110 . 7554/eLife . 03970 . 012Figure 2—figure supplement 6 . Total endocytosed protein and transferrin co-localise after 90 s uptake in Cos7 and RPE cells . Confocal images of co-internalisation of total membrane protein , labeled at 4°C with sulfo-NHS-SS-biotin , and transferrin-546 . Internalisation was for 90 s at 37°C . Biotin was detected with streptavidin-488 . Note that labelling with biotin and transferrin at 4°C was carried out consecutively , so transferrin was not biotinylated . Zoomed in area of the lower panel is indicated with a box . Bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01210 . 7554/eLife . 03970 . 013Figure 2—figure supplement 7 . Absence of membrane-positive , transferrin-negative vesicles . ( A ) Cells were labelled with the membrane dye FM1-43FX and transferrin at 4°C , warmed to 37°C for 90 s , plasma membrane dye was removed by washes with ice-cold PBS , and were imaged at 4°C without fixation . Transferrin-positive puncta that do not contain FM1-43FX are likely to represent clathrin-coated pits that have not budded from the plasma membrane . Contrast levels have been set so that residual background plasma membrane staining with FM1-43FX is excluded . Zoomed in area of the lower panel is indicated with a box . Bar is 20 μm . ( B ) Quantification of the proportion of FM1-43FX signal detected in intracellular puncta that is present in transferrin-positive pixels . Note that cells where FM1-43FX clearly stained many intracellular membranes due to cell disruption were excluded from the analysis . Bars are mean , SD . Each point is one cell region . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 013 The experiments described above depended on labelling cells at 4°C before rapid warming to 37°C to permit endocytosis . This improved signal to noise ratio and thereby improved temporal resolution , but at the same time cooling and rapid warming could have an effect on the endocytic machinery ( Boucrot et al . , 2010 ) . We sought to perform the endocytic assay in unperturbed cells , under more physiological conditions , by omitting the pre-incubation step at 4°C . Cells were labelled with sulfo-NHS-SS-biotin and transferrin at 37°C ( Figure 2E ) . The earliest time after initiation of labelling at which endocytic vesicles could be reliably detected was 90 s , presumably because of the time taken for sufficient reaction of the NHS ester with primary amines at the cell surface . Consistent with the results when labelling was carried out at 4°C , nearly all of the internalised protein labelled with biotin was indeed present in transferrin-positive vesicles ( Figure 2E , Figure 2F ) . Another concern was that HeLa cells could be in some way atypical , so we repeated the same experiments in Cos7 and RPE cells . Again , there was near-complete co-localisation between endocytosed sulfo-NHS-SS-biotin and transferrin after 90 s uptake at 37°C ( Figure 2—figure supplement 6 ) . In order to verify that transferrin-negative endocytic vesicles are very rare we used a different labelling approach . We loaded HeLa cells with the amphiphilic styryl membrane dye FM1-43FX ( Diefenbach et al . , 1999 ) , and allowed co-internalisation with transferrin for 90 s at 37°C . Extraction of the dye with label-free medium reversed plasma membrane labelling and provided a means of identifying FM1-43FX-positive intracellular vesicles . Although this may not provide such stringent topological discrimination between intracellular vesicles and surface membrane as the sulfo-NHS-SS-biotin plus MESNa approach , there was a very high degree of co-localisation between FM1-43FX-positive puncta and transferrin ( Figure 2—figure supplement 7 ) . Quantification revealed that 90% of the FM1-43FX signal present in intracellular vesicles was also present in transferrin-positive pixels ( Figure 2—figure supplement 7 ) . Confocal z-stacks and volume rendering were used to allow analysis at the level of individual primary endocytic vesicles . ( Figure 3A , Figure 3—figure supplement 1A ) . Biotin-positive endocytic vesicles were identified as 3D objects using the Imaris software , and the transferrin cargo load was calculated based on the mean transferrin intensity within each vesicle ( Figure 3A , Figure 3B , Video 1 ) . In order to account for background signal generated by experimental noise or random overlap , the transferrin channel was offset by 500 nm from its correct register , and transferrin intensity in the same objects was sampled for a second time . ( Figure 3B ) . The 95th percentile of this background intensity distribution was used as a cut-off to define transferrin-negative endocytic vesicles ( Figure 3B ) . Applying this conservative criterion , after 90 s continuous labelling and uptake 96% of 2350 vesicles contained transferrin , and after labelling at 4°C and uptake for 20 s 92% of 2387 vesicles contained transferrin ( Figure 3B ) . A single Gaussian curve could describe the transferrin intensity distribution within most identified vesicles , consistent with stochastic incorporation of transferrin receptors into forming clathrin-coated pits ( Collinet et al . , 2010 ) . A small population of transferrin-negative objects could be seen as a peak at the lowest end of the intensity distribution , showing that potential transferrin-negative endocytic vesicles , where they exist , can be detected by our method ( Figure 3B ) . We observed no correlation between biotin intensity and the probability of that vesicle not containing transferrin , arguing against the possibility of a morphologically distinctive class of transferrin-negative endocytic vesicle ( Figure 3—figure supplement 1B ) . We also carried out experiments to test the possibility that the glycosphingolipid-binding B-subunit of cholera toxin ( CTB ) , which has been extensively used as a marker for clathrin-independent endocytosis ( Henley et al . , 1998; Sandvig et al . , 2004; Kirkham et al . , 2005 ) , induces the formation of transferrin-negative endocytic vesicles . Biotin-positive vesicles were identified as objects in transferrin-labeled cells as described above , with and without addition of CTB . The transferrin load in vesicles in control and CTB treated cells was the same ( Figure 3—figure supplement 2 ) . 10 . 7554/eLife . 03970 . 014Figure 3 . Over 95% of total endocytosed protein enters the cell via clathrin-coated pits . ( A ) 3D projection of cell volumes following interalisation of sulfo-NHS-SS-biotin for 90 s at 37°C . Streptavidin-488 fluorescence is shown in the left panel , vesicle objects recognised with Imaris software from the streptavidin signal are shown in the right panel . Bar 5 μm . ( B ) Analysis of the transferrin cargo load of endocytic vesicle objects identified as in A , after 20 s and 90 s of uptake as shown . Frequency distribution of mean transferrin intensity in individual vesicles is shown as the red line . Frequency distribution of transferrin intensities for the same vesicles after offsetting the transferrin channel by 500 nm provides a set of background intensities , shown as a black line and not plotted to the same y-axis scale . Cut-offs are shown as dotted lines and correspond to the 95 percentile for the offset values . The distribution of transferrin intensity , in the majority of endosomes , can be described by a Gaussian distribution ( dark grey line ) . ( C ) Internalisation of sulfo-NHS-SS-biotin and transferrin-647 , for 15 min , in cells expressing AP180C-IRES-GFP . Internalised biotin was detected by MESNa treatment and labelling with streptavidin , a wash at pH3 . 0 removed external transferrin . Transfected cells are outlined in white . Bars are 20 μm . ( D ) Internalisation of sulfo-NHS-SS-biotin and transferrin-647 , for 15 min , in cells expressing dynamin-2-K44A-dsRed . Internalised biotin was detected by MESNa treatment and labelling with streptavidin , a wash at pH3 . 0 removed external transferrin . Transfected cells are outlined in white . Bars are 20 μm . ( E ) Quantification of total protein and transferrin endocytosis in cells expressing AP180C-IRES-GFP as C . Each point is mean fluorescence intensity of one cell region , after background subtraction . Background was determined empirically from control experiments with only labelling at 4°C . Values are all normalised so mean of control = 1 . Bars mean and SD , data are all from one experiment , the experiment was repeated three times . ( F ) Quantification is for cells expressing dynamin-2-K44A-dsRed as shown in D . See E for details . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01410 . 7554/eLife . 03970 . 015Figure 3—figure supplement 1 . Correlation between streptavidin ( total endocytosed protein ) and transferrin intensities in endocytic vesicles . ( A ) Projections of confocal z-stacks showing raw fluorescence images of internalised biotin , labelled with streptavidin , and transferrin , after 90 s internalisation . The streptavidin image was used to identify endocytic vesicles using Imaris software as displayed in the lower left panel . The lower right panel displays the software-recognised vesicle objects superimposed on streptavidin and transferrin fluorescence . Bar 15 μm . ( B ) Correlation between the mean fluorescence intensity of streptavidin and transferrin present in individual endosomes . Individual endosomes have has been ranked from high to low streptavidin intensity . Red dots correspond to the transferrin intensity in the particular endosome . Dotted line represents the cut-off for transferrin-positive endosomes , based on 95th percentile of background intensities as described in the main text . Transferrin positive endosomes are within in the greyed area . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01510 . 7554/eLife . 03970 . 016Figure 3—figure supplement 2 . Effect of CTB-binding on transferrin intensities in endocytic vesicles . Analysis of the transferrin cargo load of endocytic vesicle objects identified as in Figure 3A after 90 s of uptake in control and CTB-labeled cells as shown . Frequency distribution of transferrin intensities for the same vesicles after offsetting the transferrin channel by 500 nm provides a set of background intensities , shown as a black line and not plotted to the same y-axis scale . Cut-off is shown as a dotted line and corresponds to the 95 percentile for the offset values . The proportion of vesicles that are potentially transferrin-negative ( the fraction below the dotted line ) is not altered by CTB-binding . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01610 . 7554/eLife . 03970 . 017Video 1 . Object recognition for quantification of cargo load in individual endocytic vesicles . Cells were labelled for 90 s at 37°C with sulfo-NHS-SS-biotin and transferrin-546 . After MESNA treatment internalised proteins were labelled with streptavidin-488 . 3D reconstructions were obtained from confocal z-stacks of whole cell volumes . Streptavidin channel ( green ) , transferrin channel ( red ) , an overlay of both channels , and then the same overlay with objects recognised as vesicles superimposed in blue , are shown consecutively . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 017 Both quantitative analysis of single confocal sections using pixel masks ( Figure 2 ) , and object-based quantification of all primary endocytic vesicles in 3D reconstructions ( Figure 3 ) , argue strongly that around 95% of primary endocytic vesicles are positive for the best characterised cargo of clathrin-coated pits , the transferrin receptor ( Kirchhausen et al . , 2014 ) . The simplest hypothesis arising from these observations is that essentially all proteins taken up by the cell are internalised along with transferrin via clathrin-coated pits . If this is the case , then ablation of coated pit activity should effectively block total endocytosis . Overexpression of the C-terminal clathrin-binding domain of AP180/CALM ( AP180-C ) provides an efficient means of blocking formation of coated pits ( Ford et al . , 2001 ) . In cells overexpressing AP180-C , transferrin uptake and endocytosis of total biotinylated protein were both efficiently blocked , with total protein uptake being reduced to less than 5% of control levels ( Figure 3C and Figure 3E ) . Overexpression of the K44A mutant of dynamin blocks budding of clathrin-coated pits ( van der Bliek et al . , 1993; Damke et al . , 1995 ) . In cells overexpressing dynamin 2 K44A there was efficient reduction of both transferrin and total biotinylated protein uptake to less than 5% of control levels ( Figure 3D and Figure 3F ) . Additionally , we noted that expression of very high levels of the dynamin mutant induced endocytosis of total biotinylated protein in both macropinosomes and smaller vesicular structures ( see below ) . We conclude that formation of the primary vesicles that contribute the large majority of endocytic flux in unperturbed cells is blocked by loss of clathrin or dynamin function . If plasma membrane proteins are predominantly endocytosed via clathrin coated pits then one would predict that they should accumulate in the plasma membrane when coated pits are not functional . Proteomic analysis of surface-biotinylated plasma membrane proteins offered a way to test this hypothesis directly . In order to ablate coated pit activity , cells were transfected with siRNA against the alpha adaptin subunit of the clathrin adaptor AP2 ( ‘AP2 siRNA’ ) . This blocked alpha-adaptin expression , and hence AP2 function ( Figure 4A , Figure 4B ) ( Motley et al . , 2003 , 2006 ) . AP2 siRNA blocked both transferrin uptake and uptake of total biotinylated proteins to a similar extent ( Figure 4B ) . SILAC ( stable isotopic labelling by amino-acids in culture ) followed by precipitation of surface biotinylated proteins with streptavidin-agarose provided a quantitative comparison of the relative abundance of plasma membrane proteins in control and AP2 siRNA treated cells . This revealed a clear and pronounced increase in the abundance of most proteins in AP2-siRNA cells ( Figure 4C and Figure 4—source data 1 ) . Analysis of SILAC ratios of the non-biotinylated cytosolic proteins present in the flow-through provided an internal control and did not reveal similar changes , confirming that the effects of AP2 depletion are largely restricted to the population of plasma membrane proteins ( Figure 4C ) . This is consistent with the majority of plasma membrane proteins entering the cell via clathrin-coated pits . 10 . 7554/eLife . 03970 . 018Figure 4 . Changes in plasma membrane protein composition in cells depleted of AP2 . ( A ) Western blot for the AP2 alpha subunit following non-targeting or alpha adaptin ( AP2 ) siRNA treatment . ( B ) Internalisation of sulfo-NHS-SS-biotin and transferrin-647 , for 15 min , in cells transfected with control and alpha adaptin ( AP2 ) siRNAs . Internalised biotin was detected after MESNa treatment and labelling with streptavidin , a wash at pH3 . 0 removed external transferrin . Each point is one cell region , bars are mean and SD . Background was calculated from cells labelled at 4°C , and immunofluorescence identified those cells where the siRNA efficiently reduced alpha adaptin levels . ( C ) Frequency distribution of SILAC ratios for surface biotinylated ( red line ) and non-labelled ( grey line ) proteins from control ( Heavy isotopes ) and AP2-siRNA ( Light isotopes ) transfected cells . Dotted lines represent two standard deviations on either side of the mean for the SILAC ratio distribution of non-labelled proteins . ( D ) SILAC protein ratios comparing control ( Heavy isotopes ) and AP2-siRNA ( Light isotopes ) transfected cells , plotted against summed peptide intensities . Biotinylated plasma membrane proteins isolated by precipitation with streptavidin-agarose are shown in red . Non-biotinylated proteins corresponding to intracellular proteins are shown in grey and served as an internal control . GPI-anchored proteins , are shown in green . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01810 . 7554/eLife . 03970 . 019Figure 4—source data 1 . SILAC ratios for biotinylated and non-biotinylated proteins . Proteins shaded green were classified as depleted from the the plasma membrane upon AP2 siRNA treatment , those shaded redwe classified as accumulated . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 01910 . 7554/eLife . 03970 . 020Figure 4—figure supplement 1 . Verification of changes in plasma membrane protein levels detected by SILAC . Flow cytometry was used to analyse plasma membrane abundance of transferrin receptor and CD59 . Cells were transfected with control siRNA or siRNA to knock down expression of the alpha adaptin subunit of the AP2 complex ( ‘AP2 siRNA’ ) and then labeled at 4°C with anti-CD59-AlexaFluor647 and Transferrin-546 . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 020 The comparatively small list of proteins that are depleted from the plasma membrane upon AP2 siRNA treatment is heterogeneous ( Figure 4—source data 1 ) , and includes proteins thought to be internalised via clathrin-coated pits such as ADAM 10 ( Marcello et al . , 2013 ) . The only obvious pattern within the list is the presence of most GPI-anchored proteins detected ( Figure 4D ) . SILAC ratios for GPI-anchored proteins detected in the cytosolic flow-through were all close to one , so this is likely to reflect a depletion restricted only to the plasma membrane ( Figure 4—source data 1 ) . Flow cytometry to assay changes in plasma membrane levels of individual proteins confirmed that AP2 siRNA causes an accumulation of transferrin receptor at the plasma membrane , while levels of the GPI-anchored protein CD59 are reduced ( Figure 4—figure supplement 1 ) . As GPI-anchored proteins have been studied as potential cargoes for clathrin-independent endocytic pathways , we decided to focus further on this class of protein ( Nichols et al . , 2001; Sabharanjak et al . , 2002; Mayor and Riezman , 2004 ) . We produced plasmids for expression of SNAP-tagged versions of CD59 , folate receptor and PrP , three different GPI-anchored proteins identified in our mass spectrometry data , as well a minimal SNAP-GPI construct . HeLa cells expressing each of the GPI-anchored proteins were incubated for 90 s at 37°C in the presence of BG-SS-488 to allow labelling and uptake , and extracellular fluorophore was removed by reduction with MESNa ( Figure 5A ) . Uptake of BG-SS-488 was detected only in SNAP-expressing cells ( Figure 5A ) . All four GPI-anchored proteins co-localised extensively with transferrin after internalisation ( Figure 5A ) . When quantified , around 90% of all BG-SS-488 labelled , internalised GPI-anchored proteins co-localised with transferrin ( Figure 5B ) . To assess the biological significance of the residual 10% we used precisely the same quantification method to analyse co-localisation between transferrin-alexa-647 and transferrin-alexa-546 , after mixing and adding to cells for simultaneous internalisation . Again we detected around 90% co-localisation , so it is possible that the residual 10% can be accounted for by limitations in quantification rather than a biologically significant pool of GPI-anchored protein internalised separately from transferrin ( Figure 5B ) . 10 . 7554/eLife . 03970 . 021Figure 5 . GPI-anchored proteins co-localise with transferrin in primary endocytic vesicles . ( A ) Confocal images of cells transfected with the SNAP-tagged GPI-anchored proteins indicated . Labelling with BG-SS-488 and transferrin-546 at 37°C for 90 s . External 488 fluorophore was removed by reduction with MESNa . Right hand panels are zoomed views of the regions indicated , bars are 10 μm . ( B ) Quantification of co-localisation between internalised GPI-anchored proteins revealed with BG-SS-488 and MESNa as in A , and transferrin-546 . In order to provide an empirical estimate of the sensitivity of quantification , two fluorescently labelled transferrin probes were mixed and added to the cells . Bars are mean and SD . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 02110 . 7554/eLife . 03970 . 022Figure 5—figure supplement 1 . Effect of CTB-binding on transferrin intensities in endocytic vesicles defined by uptake of GPI-linked protein . Analysis of the transferrin cargo load of endocytic vesicle objects identified as in Figure 3A after 90 s of uptake in control and CTB-labeled cells as shown . Cells were stably expressing SNAP-CD59 , endocytic vesicles were defined by BG-SS-488 labelling and MESNa treatment . Frequency distribution of transferrin intensities for the same vesicles after offsetting the transferrin channel by 500 nm provides a set of background intensities , shown as a black line and not plotted to the same y-axis scale . Cut-off is shown as a dotted line and corresponds to the 95 percentile for the offset values . The proportion of vesicles that are potentially transferrin-negative ( the fraction below the dotted line ) is not altered by CTB-binding . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 022 Several previous studies , including some from our laboratory , have suggested that GPI-anchored proteins are endocytosed in a clathrin-independent manner ( Nichols et al . , 2001; Nichols , 2002; Sabharanjak et al . , 2002; Mayor and Riezman , 2004; Kirkham et al . , 2005; Glebov et al . , 2006 ) . We sought to explain this apparent discrepancy . In several of these studies CTB has been used as a marker for clathrin-independent endocytosis . As discussed previously , it was possible that CTB induces clathrin-independent uptake of GPI-anchored proteins . Object-based analysis of the transferrin load within BG-SS-488-positive vesicles in cells expressing SNAP-CD59 argues that this is not the case ( Figure 5—figure supplement 1 ) . One type of experiment that can be interpreted as evidence for clathrin-independent endocytosis is the presence of endocytosed protein in vesicles separate from those labelled with transferrin ( Sabharanjak et al . , 2002; Glebov et al . , 2006; Bhagatji et al . , 2009 ) . We did not observe this in the experiments outlined above ( Figure 5 ) , so we chose to pursue this further by conjugating a monoclonal antibody against CD59 to both alexa-546 , and to SS-atto-488 ( Figure 6A ) . Importantly , the disulfide bond in SS-atto-488 can be reduced with MESNa as previously . This meant we could follow uptake of endogenous CD59 , and directly compare two different methods of assaying internalisation . Cells were labelled at 4°C with the doubly labelled antibody and transferrin , and then warmed to 37°C for 90 s . Subsequently cells were treated with PI-PLC to cleave GPI-anchors , and then with MESNa . The SS-atto-488 signal co-localised completely with transferrin . Alexa-546 , however , was seen in many punctate structures that lack transferrin ( Figure 6A ) . We interpret these data as strong evidence that the PI-PLC treatment did not remove all external GPI-anchored protein , while reduction with MESNa provided more stringent discrimination between internal and external pools of the antibody ( Figure 1A , Figure 1B , Figure 1—figure supplement 1 ) . Therefore , endogenous CD59 is indeed internalised in transferrin-containing vesicles , and apparent internalisation in vesicles lacking transferrin may , at least under the conditions we have employed in this study , be an artifact due to incomplete removal of surface bound antibody . 10 . 7554/eLife . 03970 . 023Figure 6 . Clathrin-dependent endocytosis of GPI-anchored proteins . ( A ) Doubly labelled anti-CD59-546-SS-488 allows comparison of MESNa reduction and PI-PLC treatment as methods for detecting internalised GPI-anchored protein . Cells were labelled at 4°C , warmed to 37°C for 90 s , and treated consecutively with MESNa and PI-PLC . Circles indicated antibody-positive puncta that appear internalised , but are demonstrated to be extracellular by the absence of MESNa-protected 488 fluorophore . Bar is 10 μm . ( B ) Internalisation of BG-SS-549 and transferrin-647 , for 15 min , in cells stably expressing SNAP-CD59 and transiently transfected with AP180C-IRES-GFP . Internalised BG-SS-549 was detected after MESNa treatment and wash at pH3 . 0 to remove external transferrin . Transfected cells are outlined in white . Bars are 10 μm . ( C ) Internalisation of BG-SS-488 and transferrin-647 , for 15 min , in cells stably expressing SNAP-CD59 and transiently transfected with dynamin-2-K44A-dsRed . Internalised BG-SS-488 was detected by MESNa treatment , a wash at pH3 . 0 removed external transferrin . Transfected cells are outlined in white . Bars are 10 μm . ( D ) Quantification of SNAP-CD59 and transferrin endocytosis in cells expressing AP180C-IRES-GFP as A . Each point is mean fluorescence intensity of one cell region , after background subtraction . Background was determined empirically from control experiments with only labelling at 4°C . Values are all normalised so mean of control = 1 . Bars mean and SEM , data are all from one experiment , the experiment was repeated three times . ( E ) Quantification is for cells expressing dynamin-2-K44A-dsRed as shown in C . See D for details . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 02310 . 7554/eLife . 03970 . 024Figure 6—figure supplement 1 . Endocytic structures induced by high dynamin-2-K44A expression . ( A ) Hela cells stably expressing SNAP-CD59 and transiently transfected with dynamin-2-K44A-dsRed were labeled with BG-SS-488 and transferrin-647 for 15 min at 37°C . Note that the cell shown has a very high level of dynamin-2-K44A expression , and this has induced abundant macropinosomes that are clearly larger than the normal endosomes in neighbouring cells . This is an extreme illustrative example . ( B ) Flow cytometry measuring internalisation of SNAP-CD59 in the population of cells transfected with dynamin-2-K44A-dsRed as in A . Note that as the expression of dynamin-2-K44A increases , SNAP-CD59 actually also goes up ( Figure 6E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 024 A second type of experiment used in many studies to provide evidence for uptake via clathrin-independent endocytosis is to perturb the function of clathrin-coated pits , and then assay continued endocytosis of a specific protein ( Doherty and McMahon , 2009; Hansen and Nichols , 2009 ) . We used overexpression of AP180-C to block formation of coated pits in a HeLa cell line stably expressing SNAP-CD59 . In cells expressing AP180-C , endocytosis of SNAP-CD59 was blocked to the same extent as endocytosis of transferrin ( Figure 6B and Figure 6D ) . Endocytosis of SNAP-CD59 was also blocked by moderate expression of dynamin-2-K44A ( Figure 6C and Figure 6E ) . In cells expressing very high levels of the mutant dynamin , abundant macropinosomes could be detected , as well as smaller endocytic structures ( Figure 6—figure supplement 1 ) ( Damke et al . , 1995 ) . These were clearly induced by high levels of dynamin-2-K44A ( Figure 6—figure supplement 1 ) . Therefore moderate dynamin-2-K4AA expression does block the physiological mechanism that leads to internalisation of CD59 in unperturbed cells . As loss of both clathrin and dynamin function blocks uptake of CD59 , the data imply that GPI-anchored proteins are likely to enter cells predominantly via clathrin-coated vesicles . We used surface protein biotinylation with sulfo-NHS-SS-biotin to compare the distribution of total endocytosed protein with markers for possible clathrin-independent endocytic pathways . We examined co-localisation with flotillin 1 ( Glebov et al . , 2006; Stuermer , 2011 ) , caveolin 1 ( Rothberg et al . , 1992; Schneider et al . , 2008; Parton and Howes , 2010 ) , GRAF1-GFP ( Lundmark et al . , 2008 ) , and ARF6-GFP ( Naslavsky et al . , 2003 , 2004 ) after 90 s and 15 min of endocytosis at 37°C ( Figure 7 , Figure 7—figure supplement 1 ) . Co-localisation was quantified using a mask generated from the streptavidin channel as explained in Figure 2—figure supplement 3 . Importantly , as the different markers label abundant structures it was possible that low levels of co-localisation would be detected due to chance overlap rather than specific labelling of the same structures . To control for this , we repeated the co-localisation quantification with the two channels in all images offset from each other by 500 nm . This provides an empirical way to estimate the degree of overlap between channels that arises by chance . 10 . 7554/eLife . 03970 . 025Figure 7 . Labelling of the total population of endocytosed proteins does not provide evidence for significant protein flux through clathrin-independent pathways . ( A , C , E , G ) Confocal images showing distribution of the marker indicated ( caveolin 1 , flotillin 1 , GRAF1 , ARF6 ) , together with total internalised protein after 90 s of endocytosis , revealed as in Figure 1D . In the case of GRAF1 and ARF6 , additional images showing co-internalised transferrin are also shown . Bars are all 10 μm . ( B , D , F , H ) Quantification of co-localisation between the markers caveolin 1 , flotillin 1 , GRAF 1 and ARF6 , and internalised protein labelled with sulfo-NHS-SS-biotin and MESNa treatment . Internalisation was for 90 s or 15 min . In order to establish empirically the degree of overlap between internalised protein and relevant marker expected by chance , quantification was carried out both with the images in the correct register , and also with one channel manually offset approximately 500 nm from the other . Bars are mean and SD , each point is one cell region . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 02510 . 7554/eLife . 03970 . 026Figure 7—figure supplement 1 . Labelling of the total population of endocytosed proteins does not provide evidence for significant protein flux through clathrin-independent endocytic pathways . ( A , B , C , D ) Confocal images showing distribution of the marker indicated ( caveolin 1 , flotillin 1 , GRAF1 , ARF6 ) , together with internalised biotinylated protein after 15 min of endocytosis . Labelling was done as in Figure 1D . This is the same experiment as shown in Figure 7 , but a longer time-point . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 026 In the case of caveolin 1 , after both 90 s and 15 min some apparent co-localisation in punctate structures could be observed , and quantification confirmed that around 2% of caveolin 1 was present specifically in biotin-positive endosomes ( Figure 7A , Figure 7B , 15 min time-point images in Figure 7—figure supplement 1 ) . In the case of flotillin 1 , although some plausible co-localisation could be observed in confocal images , quantification revealed that this may arise purely by chance as the same degree of pixel overlap was found in offset and original images ( Figure 7C , Figure 7D ) . In the case of GRAF1 , internalised biotin was detected in GRAF1-GFP-positive puncta ( Figure 7E ) . Quantification confirmed specific overlap between internalised biotin and GRAF1 ( Figure 7F ) . The fact that our previous data show that most internalised biotin/total protein co-localises with transferrin made the strong prediction that these GRAF1-positive endosomes should also contain transferrin . This was indeed the case ( Figure 7E ) . ARF6 showed a similar distribution to GRAF1 , labelling transferrin-positive endosomes ( Figure 7G , Figure 7H ) . We conclude that GRAF1-GFP and ARF6-GFP are present on transferrin-positive endocytic vesicles likely to have arisen from budding of clathrin-coated pits from the plasma membrane , that flotillin 1 is not detected on early endosomes , and that caveolin 1 is present on a very small fraction of endosomes . All of these observations are consistent with our general conclusion , that at least 95% of endocytosed protein enters the cell via clathrin-coated pits . Our data argue that GPI-anchored proteins enter the cell via coated pits , but , they behave differently from many other clathrin cargoes when AP2 expression is suppressed over 5 days . There are several factors potentially involved in this effect , such as rates of synthesis , degradation and endocytic recycling of proteins with GPI anchors . We investigated further the factor most relevant to the focus of this study , sorting during endocytosis . The absence of a cytosolic domain means that GPI-anchored proteins are fundamentally different from high-affinity cargoes such as the transferrin receptor , as they can not be recruited to the nascent pit via direct recognition of endocytic sorting motifs by adaptor proteins . It is likely that GPI-anchored proteins are partially excluded from coated pits by steric crowding effects ( Bhagatji et al . , 2009 ) . We reasoned that perturbing the recognition of endocytic sorting motifs could therefore cause differential effects on uptake of GPI-anchored proteins and transferrin receptor . As an initial test of this hypothesis , we used siRNA to partially reduce AP2 expression . AP2 recognises two major endocytic motifs ( Edeling et al . , 2006; Motley et al . , 2006; Kelly et al . , 2008; Kirchhausen et al . , 2014 ) . We measured transferrin and SNAP-CD59 uptake by flow cytometry , 3 days after siRNA transfection . Cells still endocytosed significant amounts of both cargoes ( Figure 8—figure supplement 1 ) . We compared the ratio between the two cargoes , on a cell by cell basis , for the populations of control and AP2 siRNA transfected cells ( Figure 8A ) . AP2 siRNA caused a more pronounced reduction of transferrin than CD59 uptake . Confocal imaging confirmed the flow cytometry results . In cells where AP2 levels were reduced , the reduction in transferrin uptake appeared more severe than reduction in SNAP-CD59 uptake ( Figure 8B ) . Using these imaging data , we correlated the amount of AP2 remaining in each cell with the total fluorescence intensity of internalised transferrin and SNAP-CD59 ( Figure 8C ) . The uptake of both cargoes was efficiently blocked when AP2 levels became close to zero , but transferrin uptake was clearly more sensitive to reduction of AP2 levels than uptake of SNAP-CD59 . ( Figure 8C; see also Figure 4B ) . 10 . 7554/eLife . 03970 . 027Figure 8 . Differential effects on uptake of transferrin receptor and GPI-anchored proteins via coated pits . ( A ) Frequency distribution of the ratio between internalised transferrin and internalised SNAP-CD59 in individual HeLa cells , determined by flow cytometry as in Figure 8—figure supplement 1 . ( B ) AP-2 knockdown can affect endocytosis of clathrin coated vesicle cargo proteins differentially . HeLa cells stably expressing SNAP-CD59 were transfected with AP2 siRNA and assayed at various timepoints up to 72 hr after transfection . Uptake of BG-SS-488 and transferrin-546 was for 15 min . Cells were MESNa treated , acid washed , fixed and stained with anti-alpha-adaptin . Top row; AP2 levels appear normal . Second row; AP2 levels are intermediate , transferrin uptake is blocked while SNAP-CD59 uptake is less severely inhibited . Third row; AP2 levels are very low and uptake of both SNAP-CD59 and transferrin is blocked . ( C ) Correlation of the amount of internalised transferrin and SNAP-CD59 with the amount of AP-2 present in each cell from the above experiment . Data were fit to a simple one-phase association . Shaded area around the curve fitted corresponds to 95% CI . ( D ) HeLa cells stably expressing SNAP-CD59 were transfected with µ2-IRES-GFP or µ2 ( F174S/D176A ) -IRES-GFP . After 4 days , cells were incubated at 37°C for 150 s with transferrin-647 and BG-SS-546 . The white lines outline transfected cells . The blue box highlights a region shown in E . Bars 20 μm . ( E ) Co-localisation between internalised SNAP-CD59 and transferrin in a cell expressing µ2 ( F174S/D176A ) -IRES-GFP . ( F ) Frequency distribution of CD59 cargo load within individual vesicles . Cells were labelled as in D . Vesicles were identified as objects in 3D reconstructions from confocal images with Imaris software using the transferrin signal . ( G ) As F , but displaying transferrin cargo load in the same population of vesicles . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 02710 . 7554/eLife . 03970 . 028Figure 8—figure supplement 1 . Reduction of AP-2 ( alpha adaptin ) levels affects the amount of uptake of both transferrin and SNAP-CD59 , 50 hr after siRNA transfection . Cells stably expressing SNAP-CD59 were incubated at 37°C for 15 min with BG-SS-549 and transferrin-647 . Surface label was removed with MESNa and acid wash and then the cells were analysed by flow cytometry . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 02810 . 7554/eLife . 03970 . 029Figure 8—figure supplement 2 . Incorporation of mutant μ2 subunits into AP2 complexes . Cells were transfected with plasmids expressing the μ2 YXXΦ-binding mutant-myc as shown . Cell lysates were subjected to immunoprecipitation with anti-alpha adaptin , or as a negative control anti-GFP antibodies . Immunoprecipitates were analysed by Western blotting with anti-alpha adaptin and anti-myc antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 03970 . 029 Specific amino acid changes in the YXXΦ binding site of the µ2 subunit of AP2 abolish binding of YXXΦ-containing cargoes like transferrin receptor ( Honing et al . , 2005; Jackson et al . , 2010 ) . Overexpression of this mutant form of μ2 for 4 days resulted in incorporation into endogenous AP2 complexes , and thereby provided a tool to further investigate sorting of transferrin and GPI-anchored proteins into coated pits ( Figure 8—figure supplement 2 ) . Overexpression of µ2 ( F174S/D176A ) , but not wild-type μ2 , caused a dramatic differential effect on the uptake of the two cargoes after 90 s of labelling and internalisation . As predicted , transferrin uptake was reduced ( Honing et al . , 2005 , Motley et al . , 2006 ) . Uptake of SNAP-CD59 , however , was significantly increased ( Figure 8D ) . Importantly , this divergence did not arise from separate endocytic structures , as internalised SNAP-CD59 still co-localised well with the residual internalised transferrin ( Figure 8E ) . Moreover , when 3D object-based analysis was used to identify transferrin-positive vesicles , and to quantify the amount of SNAP-CD59 present in these vesicles , a clear increase in the SNAP-CD59 load in mutant-expressing cells was observed ( Figure 8F ) . This analysis also confirmed the reduced transferrin load in endocytic vesicles in mutant-expressing cells ( Figure 8G ) . These results are in agreement with our initial hypothesis of distinctive sorting of GPI-anchored proteins into clathrin coated pits . They argue that steric exclusion of GPI-anchored proteins from the nascent pit becomes less acute when recruitment of high-affinity cargoes by AP2 , or potentially other adaptors is abolished ( Bhagatji et al . , 2009 ) . More generally , the data imply that rates of endocytosis of high affinity cargoes may be more susceptible to a range of perturbations of coated pit function than cargoes like GPI-anchored proteins , which have a lower affinity for the nascent pit .
An imaging approach to studying endocytosis in unperturbed cells , where all surface proteins are labelled , is intrinsically easier to interpret than experiments where mutant proteins , siRNA , or chemical inhibitors cause loss of function , and potentially indirect or off-target effects , over days . The use of small , monovalent reducible chemical labels provides high temporal resolution , signal-to-noise ratio and efficiency of topological discrimination between intracellular and extracellular protein . This has allowed us to follow endocytosis of effectively all proteins , and thereby to observe all endocytic intermediates formed by the cell at specific time-points , starting with as little as 20 s of internalisation . Our results are at some variance with previous studies from our own and other laboratories , and with our preconceptions before being confronted with the data ( Hansen and Nichols , 2009 ) . Our data support the conclusions that the predominant pathway for endocytosis in mammalian cells , accounting for at least 95% of total protein endocytic flux , is uptake via the clathrin-coated pit . Other mechanisms are not likely to make a significant contribution . Primary endocytic vesicles involved in clathrin-independent endocytosis should not contain significant amounts of transferrin receptor , and while we can detect a class of such vesicles they account for less than 5% of total endocytosis . Further data , including perturbations of clathrin-coated pit function , assay of the endocytosis of GPI-anchored proteins , and co-localisation experiments looking at candidate markers for clathrin-independent endocytic pathways are all consistent with this conclusion . A recent systematic analysis of multiple assays for endocytosis using siRNA screens also provides indirect support , in that it reveals extensive functional and regulatory links between uptake of cargoes previously thought to enter the cell via different mechanisms ( Liberali et al . , 2014 ) . In addition , this study conspicuously does not detect new linked sets of proteins likely to act together in a concerted fashion to make novel types of endocytic vesicle . Some cargoes for clathrin coated pits have been reported to be present in sub-populations of early endocytic vesicles , most likely through mechanisms related to cargo-dependent recruitment of specific adaptors ( Lakadamyali et al . , 2006; Mettlen et al . , 2010; Henry et al . , 2012 ) . However , there is good evidence that transferrin receptor does not behave in this way ( Lakadamyali et al . , 2006; Kirchhausen et al . , 2014 ) , and our analysis of transferrin load in primary endocytic vesicles confirms this . In the light of these observations , it is difficult to reconcile our data with reports of abundant clathrin-independent endocytic carriers or CLICs . We note that the ultrastructure of CLICs , although clearly different from clathrin-generated vesicles , is not dissimilar to that of the early endosomes that receive cargoes taken up via coated pits ( Harding et al . , 1983; Willingham et al . , 1984; Gruenberg et al . , 1989; Hansen et al . , 1991; Ullrich et al . , 1996 ) . Much of the literature on clathrin-independent endocytosis is based on perturbations with differential effects on the uptake of transferrin and putative clathrin-independent cargoes ( Puri et al . , 2001; Sabharanjak et al . , 2002; Sandvig et al . , 2008; Doherty and McMahon , 2009; Hansen and Nichols , 2009; Lakshminarayan et al . , 2014 ) . Our data suggest a possible re-interpretation of this type of experiment . We demonstrate that GPI-anchored proteins enter mammalian cells via clathrin coated pits , and that this is fully compatible with the differential effects on endocytic uptake between transferrin receptor and GPI-anchored proteins reported here and observed in previous studies ( Nichols et al . , 2001; Sabharanjak et al . , 2002; Naslavsky et al . , 2004; Kumari and Mayor , 2008 ) . The coated pit is a crowded environment , and GPI-anchored proteins will have to compete for a place inside the nascent pit with cargo proteins possessing high affinity sorting signals recognised directly by adaptor proteins ( Kirchhausen et al . , 2014 ) . When the protein levels of adaptors such as AP2 are reduced , or when the ability to bind and recruit certain cargoes is lost , it is likely that the pit becomes less crowded , thereby becomes more populated with low affinity or passive cargoes like GPI-anchored proteins , and yet retains some ability to bud . This leads to the differential effects on transferrin and CD59 uptake that we observed experimentally . Divergent effects of coated pit perturbation on cargoes which all demonstrably enter the cell via coated pits , means that such effects need no longer be interpreted as evidence for clathrin-independent endocytosis . Our data place important constraints on models of endocytic flux . Additional endocytic mechanisms , like the still poorly understood set of protein machinery responsible for macropinocytosis , and the budding of caveolae , contribute very little to total protein uptake but can be detected by our assays . Such clathrin-independent pathways could still be functionally important when uptake of specific cargoes , and the variety of cell types and functions in vivo are considered ( Commisso et al . , 2013; Watanabe et al . , 2013 ) . Nevertheless , we conclude that clathrin-independent pathways do not make a significant contribution to total endocytic flux in cultured cells .
HeLa , RPE and Cos7 cells where cultured at 37°C , 10% CO2 , in DMEM supplemented with 10% FBS . When required , cells were transfected with FugeneHD ( Promega , Madison , WI ) , 14–20 hr before imaging . CD59 and FOLR1 were cloned from human cDNA . PrP was a gift from Manu Hegde . The N-terminal tagging of SNAP-CD59 , SNAP-FOLR1 and SNAP-PrP was obtained by inserting the SNAPtag domain following the N-terminal signal peptide of each protein . The minimal SNAP-GPI was constructed by inserting a SNAPtag domain between the N-terminal signal peptide and the ω-2 site of human CD59 . To obtain AP180C-IRES-GFP , AP180-C ( A gift from H McMahon ) was inserted into the multiple cloning site of IRES2-AcGFP1 , using the XhoI and SacII sites . DynII ( aa ) K44A-dsRed was a gift from M Frick . GRAF1-GFP was a gift from H McMahon . AP2 µ2 and µ2 ( F174S/D176A ) ( Gifts from M Robinson ) were cloned in IRES2-AcGFP1 , using the XhoI and SacII sites . A typographic error for the YXXΦ binding mutant is present in previous publications ( Honing et al . , 2005; Edeling et al . , 2006; Motley et al . , 2006 ) . The mutant used and characterised in those studies was as described here ( David Owen , personal communication ) . Following transfection , G418 ( 400 µg/ml ) was used to select cells stably expressing SNAP-CD59 . Resistant cells were sorted twice by flow cytometry to ensure retention of the construct and uniform expression levels . antiCD59-546-ss488 was generated by conjugating anti-CD59 ( MEM43 ) ( Abcam , United Kingdom ) simultaneously to NHS-ss-ATTO488 ( ATTO-TEC , Germany ) and NHS-AlexaFluor546 ( Molecular Probes , Waltham , MA ) . Following conjugation the antibody was separated from unreacted reagents using size-exclusion chromatography . Antibodies and reagents were obtained from the following sources; AntiCD59-AlexaFluor647 ( MEM43 ) ( AbD Serotec , United Kingdom ) , mouse monoclonal ( AP . 6 ) against alpha adaptin used for immunofluorescence and immunoprecipitation ( Abcam ) , mouse monoclonal against alpha adaptin used for western blots ( BD , Franklin Lakes , NJ ) , rabbit polyclonal anti-caveolin 1 ( BD ) , mouse monoclonal anti-flotillin 2 ( BD ) , rabbit polyclonal against clathrin heavy chain ( Abcam ) , Streptavidin-HRP ( Pierce , Rockford , IL ) , PI-PLC from Bacillus cereus ( Life Technologies , Waltham , MA ) , Streptavidin-488 , -546 or -647 ( Molecular Probes ) , Transferrin-546 or -647 ( Molecular Probes ) , Cholera toxin subunit B ( CTB ) -647 ( Molecular Probes ) , FM1-43FX ( Molecular Probes ) , SNAP-surface 549 ( NEB ) , BG-SS-488 and BG-SS-549 were kindly provided by our collaborators in NEB . Typically , non-targeting siRNA ( Dharmacon , Lafayette , CO ) or alpha-adaptin siRNA ( Dharmacon ) were delivered to the cells at a final concentration of 100 nM , using oligofectamine ( Invitrogen ) . Transfections took place on days 1 and 3 , while assays were carried out on day 5 . For partial depletion of AP-2 , one round of siRNA transfection took place and assays were performed at different timepoints up to 72 hr later . The siRNA targeting the alpha-subunit of AP-2 , has been described previously ( Robinson et al . , 2010 ) [5ʹ-GAG CAU GUG CAC GCU GGC CAdT dT-3ʹ] . AP2 complexes were immunoprecipitated with an anti-alpha adaptin antibody ( AP . 6 ) and protein G-sepharose after lysis with immunoprecipitation buffer ( 25 mM Tris–HCl pH 7 . 4 , 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 and 5% glycerol ) . To test for incorporation of the overexpressed mutant subunit into endogenous AP2 complexes , HeLa cells were transfected with μ2 ( F174S/D176A ) -IRES-GFP and maintained in culture for the indicated periods . HeLa cells were cultured for 7 days in R0K0 or R10K8 DMEM ( Dundee Cell Products , United Kingdom ) supplemented with dialysed fetal bovine serum ( MWCO 10 kDa–Dundee Cell Products ) . Following surface biotinylation , cells were lysed in 1% Triton X-100 , 1% Octyl glucoside ( Sigma , United Kingdom ) in TBSE buffer ( 50 mM Tris pH 7 . 4 , 150 mM NaCl , 5 mM EDTA ) in the presence of protease inhibitors ( Roche ) . The lysates were left to rotate in the coldroom for 30 min , and then spun at 20 . 000 rcf for 20 min . The supernatant was transferred to a clean eppendorf tube and incubated for 1 hr with high capacity streptavidin-agarose resin ( Pierce ) . Every sample was then transferred to a chromatography column ( Bio-Rad ) and washed with 25 ml 1%Triton in TBSE . To elute biotinylated proteins the resin was incubated for 5 min with 100 mM DTT in TBS ( 50 mM Tris pH 7 . 4 , 150 mM NaCl ) . SDS-PAGE gels were stained with Sypro Ruby ( Lonza , Switzerland ) or silver stain ( Pierce ) . Peptide identification from each sample was done using LTQ Orbitrap XL ( Thermo Scientific , Waltham , MA ) . Calculation of SILAC ratios and further data analysis were performed using MaxQuant ( Cox and Mann , 2008 ) and Prism ( GraphPad , San Diego , CA ) respectively . The AP2 siRNA SILAC experiment was repeated three times , data shown are from one experiment . The same overall trend in terms of accumulation of most plasma membrane proteins in the AP2 siRNA treated cells and depletion of GPI-anchored proteins , were observed in all three experiments . A recently published estimate for protein copy numbers in HeLa cells ( Kulak et al . , 2014 ) was correlated with a list of human plasma membrane proteins [GO:0005886] . Plasma membrane abundance ( PMA ) for a protein x was calculated as shown;PMA ( x ) =copy number of protein xsum copy number of all plasma membrane proteins×100 We then compared this list of plasma membrane proteins with the list of biotinylated surface proteins we detected by mass spectrometry ( Figure 1—source data 1 ) . To suspend endocytosis , cells were moved to a coldroom ( 4°C ) and washed twice with ice-cold PBS pH 7 . 9 . Primary amines at the cell surface were labelled with 0 . 2 mg/ml sulfo-NHS-SS biotin in PBS pH 7 . 9 . After 20 min , remaining sulfo-NHS-SS biotin was quenched with 50 mM Tris pH 8 . 0 in PBS , and the cells were washed two more times with PBS . Where required a further 10 min incubation at 4°C with PBS +12 . 5 µg/ml transferrin was carried out . To allow endocytosis , pre-warmed medium was added to the cells and the cultures were incubated at 37°C for various time points . Fluorescent transferrin , if required , was present in the prewarmed medium at 12 . 5 µg/ml . At the same time , positive control cells to estimate the amount of surface labelling , and negative control cells to estimate the efficiency of reduction would remain in the coldroom . Following internalisation , cells were chilled with pre-cooled PBS to immediately arrest endocytosis and then washed with ice-cold MESNa buffer without BSA in the coldroom . Remaining surface exposed label was removed by incubating the cells 2 × 10 min in cold 100 mM MESNa in MESNa buffer ( MESNa buffer: 50 mM Tris , 100 mM NaCl , 1 mM EDTA , 0 . 2wt/vol BSA , pH 8 . 6 ) Finally , transferrin , if used , was removed from the plasma membrane by acid wash at 4°C ( 2 × 2 min with 150 mM Glycine , pH 3 . 0 ) . The above protocol was modified for labelling at 37°C , with no labelling step at 4°C preceding the timeframe for endocytosis . Sulfo-NHS-SS Biotin was dissolved in pre-warmed HBSS ( HBSS + calcium , magnesium , glucose—Gibco ) at a final concentration of 0 . 2 mg/ml . If required fluorescent transferrin was added , at a final concentration of 12 . 5 µg/ml , before addition to the cells . After incubation at 37°C for the indicated time cells were rapidly chilled using pre-cooled PBS at 4°C , and moved to the coldroom to remove surface exposed biotin or transferrin as described above . BG-SS-fluorophore and fluorescent transferrin were diluted in prewarmed DMEM supplemented with 10%FBS ( Final concentrations; BG-SS-Fluorophore 2 . 5 nmol/ml , transferrin 12 . 5 µg/ml ) . After incubation at 37°C for the indicated time , cells were rapidly chilled using pre-cooled PBS at 4°C , and moved to the coldroom to arrest endocytosis and remove surface exposed label or transferrin as described previously . The protocol used has been previously described ( Diefenbach et al . , 1999 ) . Briefly , cells were labelled for 1 min at 4°C with 10 µM FM1-43FX ( Invitrogen ) in HBSS . Following internalisation at 37°C , excess FM1-43FX was removed from the plasma membrane by incubating the cells with PBS without FM1-43FX for 20 min at 4°C . Cells were imaged at 4°C without fixation . Temperature was maintained by addition of frozen , crushed DMEM to the cell chambers . GPI-anchored proteins were removed from the plasma membrane by incubating the cells at 4°C with PI-PLC ( 10 U/ml ) in PBS for 20 min . For flow cytometry , endocytic assays were performed as described above . Cells were then trypsinised , resuspended in cold 0 . 2% FBS in PBS and analysed using BD LSRFortessa . For microscopy , imaging dishes ( CG 1 . 5–Miltenyi Biotec ) were coated with fibronectin ( 5 µg/ml in PBS ) ( Sigma Aldrich ) overnight . The following day cells were seeded at a density of 6 × 103 cells/cm2 . After the internalisation assays , cells were fixed and permeabilised with either 4% paraformaldehyde in PBS followed by 0 . 1% Saponin or with methanol at −20°C ( required for staining with anti-flotillin antibodies ) . Images were acquired on a Zeiss 780 confocal microscope using a 63× 1 . 40NA Plan-Apochromat objective ( Zeiss ) and a GaAsP detector . Images were processed in Image J . All images were subjected to noise reduction with Gaussian blur σ = 0 . 7 , and contrast settings were adjusted for optimal visualisation of colour overlays . The black level in all images was set using levels empirically determined from negative control experiments to exclude cellular auto-fluorescence and other non-specific sources of signal . Quantification of co-localisation is explained in Figure 2—figure supplement 3 , and the relevant legend . In brief , one channel of a 2-colour image was converted into a binary mask that could be used to isolate those pixels from the second channel that are positive in the mask . Quantification of mean fluorescence intensity was carried out by simply drawing cell outlines as regions of interest in Image J . Importantly , empirically determined background from negative control experiments in which cells were incubated with labels only at 4°C was subtracted from all values . As there is inherent variability in both signal and background from cell to cell , this resulted in negative values in some instances . High resolution confocal stacks ( 150 nm interval ) were used for volume rendering and object identification with Imaris ( Bitplane ) . Primary endocytic vesicles were detected in the streptavidin channel using a region growing algorithm that used intensity parameters ( mean , center , maximal , standard deviation ) as detection criteria . To estimate and account for noise and random overlap we offset the transferrin channel by 20 voxel in x ( ∼500 nm ) . We then compared the updated statistics for individual endosomes with the values for the same endosomes when the two channels were correctly aligned . | Cells are enclosed by a ‘plasma membrane’ that allows nutrients and certain small molecules to move in and out of cells . Larger molecules—such as proteins—are carried into cells through a process known as endocytosis , where part of the plasma membrane engulfs the molecule and transports it through the cell inside a bubble-like compartment called a vesicle . There may be several different ways by which endocytosis can occur . The most common method involves a protein known as clathrin , which coats part of the plasma membrane on the side facing the inside of the cell . This causes the membrane to deform into a pit . The pit grows around , and eventually completely surrounds , the molecule to be transported , at which point the clathrin-coated membrane pinches off from the rest of the plasma membrane to form a vesicle . Other forms of endocytosis do not need clathrin to form vesicles , and so are collectively known as clathrin-independent endocytosis . However , the details of how these other types of endocytosis work and how important they are for moving molecules into the cell remain unclear . This is partly because it is difficult to identify particular types of endocytosis . Previous attempts to do this have involved trying to identify molecules that are specifically and solely associated with that type of endocytosis , and using these to track the vesicle . However , few—if any—such molecules are known for clathrin-independent methods of endocytosis . Another approach is to inhibit the formation of clathrin-coated pits and study those molecules that are still taken into cells . The problem here is that incomplete inhibition can make interpreting the results difficult . Furthermore , complete inhibition of an important process like clathrin-dependent endocytosis is likely to have severe effects on many other aspects of cell function . Bitsikas et al . have developed a new method that allows a vesicle to be identified—regardless of how it forms—in cells that have not been treated with inhibitors . This method involves labelling proteins in the plasma membrane with a chemical that allows them to be traced , and so shows when they are included in vesicle membranes . Importantly , this new method can provide very accurate information as to whether or not proteins have been included in vesicles , and this may provide advantages over previous approaches . Bitsikas et al . selected a group of proteins that are thought to only enter cells in a clathrin-independent manner , but unexpectedly found that these proteins predominantly enter cells through clathrin-coated vesicles . Further analysis revealed that approximately 95% of all molecules that enter cells by endocytosis are taken up via clathrin-coated endocytosis . Therefore , clathrin-independent endocytosis does not make a significant contribution to the transport of large molecules into cells . These results are at odds with current thinking in the field . Future work could reveal whether the techniques applied by Bitsikas et al . detect more active clathrin-independent endocytosis in special situations , for example during cell migration , or in specific cell types . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2014 | Clathrin-independent pathways do not contribute significantly to endocytic flux |
Hypoxia Inducible transcription Factors ( HIFs ) are principally regulated by the 2-oxoglutarate and Iron ( II ) prolyl hydroxylase ( PHD ) enzymes , which hydroxylate the HIFα subunit , facilitating its proteasome-mediated degradation . Observations that HIFα hydroxylation can be impaired even when oxygen is sufficient emphasise the importance of understanding the complex nature of PHD regulation . Here , we use an unbiased genome-wide genetic screen in near-haploid human cells to uncover cellular processes that regulate HIF1α . We identify that genetic disruption of the Vacuolar H+ ATPase ( V-ATPase ) , the key proton pump for endo-lysosomal acidification , and two previously uncharacterised V-ATPase assembly factors , TMEM199 and CCDC115 , stabilise HIF1α in aerobic conditions . Rather than preventing the lysosomal degradation of HIF1α , disrupting the V-ATPase results in intracellular iron depletion , thereby impairing PHD activity and leading to HIF activation . Iron supplementation directly restores PHD catalytic activity following V-ATPase inhibition , revealing important links between the V-ATPase , iron metabolism and HIFs .
HIFs are major transcriptional regulators of cellular responses to oxygen availability , promoting several metabolic adaptations to ensure cell survival . In aerobic conditions , the HIFα subunit is constitutively expressed but rapidly degraded by the proteasome , in a process requiring two post-translational modifications: ( i ) prolyl hydroxylation of the HIFα oxygen dependent degradation ( ODD ) domain by prolyl hydroxylases ( PHDs ) ( Bruick and McKnight , 2001; Epstein et al . , 2001 ) , and ( ii ) subsequent ubiquitination by the von-hippel lindau ( VHL ) E3 ligase ( Maxwell et al . , 1999 ) . Prolyl hydroxylation of HIFα acts as the recruitment signal for VHL , which rapidly ubiquitinates the ODD domain facilitating proteasomal degradation . Indeed , HIF1α ( the ubiquitously expressed HIFα isoform ) is a very short-lived protein ( Berra et al . , 2001 ) , and the efficiency of VHL in promoting proteasomal degradation has led to the recent development of small molecules that hijack the VHL complex to selectively destroy target proteins as a potential therapeutic tool ( Bondeson et al . , 2015 ) . Despite this clear role for proteasomal degradation of HIF , it has been reported that lysosomal inhibitors can lead to stabilisation of the HIFα subunit in both normal oxygen levels and in hypoxia . Moreover , this stabilisation can lead to a functional HIF response ( Lim et al . , 2006 ) , and upregulation of target genes to promote glucose metabolism and angiogenesis ( Hubbi et al . , 2013 ) . Initial observations regarding lysosomal degradation and HIFs arose from studies using Bafilomycin A ( BafA ) to chemically inhibit the vacuolar H+ ATPase ( V-ATPase ) , the main complex responsible for acidification of endosomal and lysosomal compartments . BafA treatment stabilised HIF1α and prevented its degradation ( Lim et al . , 2006 ) . Others report similar findings , with several proposed mechanisms to explain the stabilisation of HIF1 upon BafA treatment , including chaperone-mediated autophagy ( CMA ) ( Bremm et al . , 2014; Ferreira et al . , 2015; Hubbi et al . , 2014 , 2013; Selfridge et al . , 2016 ) , mitochondrial uncoupling ( Zhdanov et al . , 2012 ) and binding of the V-ATPase to VHL ( Lim et al . , 2007 ) . However , the relative importance of these mechanisms compared to the canonical degradation of HIF1α by prolyl hydroxylation and VHL mediated proteasomal degradation was not clear . We recently developed a forward genetic screen in near-haploid KBM7 cells to identify genes that regulate HIF1α in aerobic conditions ( Burr et al . , 2016 ) . Here , we used this screen to focus on cellular pathways enriched for gene-trapping insertions , and find that mutations in several V-ATPase subunits result in increased HIF1α levels . In addition , we identify two uncharacterised V-ATPase accessory proteins , TMEM199 and CCDC115 , which we show are required for V-ATPase function and form the mammalian orthologue of the yeast Vma12p-Vma22p V-ATPase assembly complex . Although the V-ATPase is required for lysosomal degradation ( Maxson and Grinstein , 2014 ) , we find that the mechanism for HIF1α stabilisation following V-ATPase inhibition is through intracellular iron depletion , leading to decreased PHD activity . Iron supplementation to V-ATPase depleted cells directly restores PHD hydroxylation of HIF1α in cellular assays and in vitro . These findings support a novel role for the V-ATPase and its assembly factors in regulating HIF1α levels through the control of intracellular iron levels .
We developed a forward genetic screening approach to identify genes involved in the regulation of HIF1α under aerobic conditions using near haploid human KBM7 cells expressing a HIF1α-specific fluorescent reporter ( HIF1α-GFPODD ) ( Burr et al . , 2016 ) . Briefly , this screen involved randomly mutagenising a clonal population of KBM7 HIF1α-GFPODD reporter cells with a gene-trapping retrovirus , fluorescence activated cell sorting ( FACS ) to enrich for rare mutations that resulted in increased GFP expression , and mapping the insertion sites in these GFPHIGH cells with Illumina HiSeq . This approach successfully identified genes involved in the canonical proteasomal degradation of HIF1α ( PHD2 and VHL ) as well as genes involved in mitochondrial regulation of HIF1α ( oxoglutarate dehydrogenase and lipoic acid synthase ) ( Figure 1A ) ( Burr et al . , 2016 ) . In addition , we noted that several other genes were enriched for trapping insertions compared to the control library , although their significance levels were lower ( Figure 1A ) . Therefore , we ranked genes most enriched for trapping insertions according to biological process and molecular function ( Figure 1B , Supplementary file 1 ) . As expected , the HIF pathway was ranked highly , due to the enrichment of trapping insertions in genes such as PHD2 , VHL and CUL2 . However , the top ranked biological process was transferrin transport and V-ATPase function ( Figure 1B ) , principally relating to mutagenesis of genes encoding five V-ATPase subunits: ATP6AP1 , ATP6V1A , ATP6V1G1 , ATP6V0A2 and ATP6V0D1 ( Figure 1A ) . Subsequent analysis of the location of V-ATPase gene-trap inserts confirmed that these genes were enriched for mutations in a trapping orientation ( Figure 1C ) , consistent with mutations resulting in deletion phenotypes . 10 . 7554/eLife . 22693 . 003Figure 1 . Depletion or inhibition of the V-ATPase stabilises HIF1α in aerobic conditions . ( A ) Bubble plot depicting genes enriched in the forward genetic screen . Bubbles represent the genes enriched in the GFPHIGH population compared to unmutagenised KBM7 cells expressing the HIF1α-GFPODD reporter . Proteins involved in V-ATPase assembly and function ( green ) , canonical HIF1α regulation ( purple ) , and the oxoglutarate dehydrogenase complex ( blue ) are highlighted , with the number of independent gene trap insertions indicated ( brackets ) . ( B ) Pathway analysis of enriched genes in the KBM7 forward genetic screen . The top 114 genes enriched for multiple independent gene-trapping integrations in the GFPHIGH population compared to unmutagenised KBM7 cells expressing the HIF1α-GFPODD reporter were analysed by gene ontology clustering for pathways significantly targeted in the screen . An individual gene enrichment p value < 0 . 1 was used as a threshold value for genes to be included in the pathway analysis . ( C ) Schematic of enriched gene trap insertion sites in the 5 V-ATPase subunits ( ATP6V0D1 , ATP6V1G1 , ATP6AP1 , ATP6V1A , ATP6V0A2 ) identified in the forward genetic screen . ( Red = sense insertions , Blue = antisense insertions ) . ( D , E ) Validation of the V-ATPase subunits identified in the screen using CRISPR-Cas9 targeted gene editing in HIF1α-GFPODD reporter ( D ) and wildtype ( E ) HeLa cells . Cells were simultaneously transduced with Cas9 and sgRNAs to ATP6V0D1 , ATP6V1G1 , ATP6AP1 , ATP6V1A1 , or ATP6V0A2 . GFP levels were assessed by flow cytometry after 10 days ( % GFPHIGH cells indicated ) ( D ) , and HIF1α levels measured by immunoblot ( E ) . PHD2 and β2m were used as positive and negative controls respectively . ( F , G ) Chemical perturbation of V-ATPase function . Wildtype and HIF1α-GFPODD HeLa cells were cultured in the presence of BafA ( 10 nM or 100 nM ) for 24 hr and HIF1α levels measured by GFP fluorescence ( F ) or immunoblot ( G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 003 We validated whether disrupting the V-ATPase genes stabilised HIF1α using CRISPR ( Clustered regularly-interspaced short palindromic repeats ) -Cas9 gene-editing in wildtype HeLa cells and those expressing the HIF1α-GFPODD reporter ( Figure 1D , E ) . Several single guide RNAs ( sgRNA ) were designed for each V-ATPase subunit , and transduced along with Cas9 into the HIF1α-GFPODD reporter or wildtype HeLa cells . GFP or endogenous HIF1α levels were measured by flow cytometry or immunoblot at least ten days after transduction . Depletion of all of the V-ATPase subunits identified in the screen increased GFP levels in the sgRNA-targeted cells compared with the wildtype cells and control sgRNA to MHC Class I ( β2 microglobulin , β2m ) ( Figure 1D ) , but to a lesser extent than depletion of the main PHD enzyme for HIF1α , PHD2 ( Figure 1D ) . Depletion of the V-ATPase subunits also increased endogenous HIF1α , depending on the efficiency of the sgRNA ( Figure 1E ) . Furthermore , chemical inhibition of V-ATPase activity with the inhibitor Bafilomycin A ( BafA ) activated the GFP reporter , increasing HIF1α levels in aerobic conditions ( Figure 1F , G ) without affecting HIF1α mRNA expression ( see Figure 9J ) . In addition to the V-ATPase subunits detected in the screen , we identified that TMEM199 was significantly enriched for gene-trapping insertions ( Figure 2A ) . TMEM199 is a putative transmembrane protein with homology ( 24% sequence identity ) to the yeast V-ATPase assembly protein Vma12p ( also known as Vph2p ) ( Hirata et al . , 1993 ) ( Figure 2B ) . Depletion of TMEM199 resulted in accumulation of the HIF1α reporter ( Figure 2C ) , similarly to levels observed for depletion of the V-ATPase subunits ( Figure 1D ) . FACS of the GFPLOW and GFPHIGH TMEM199 sgRNA-targeted cells followed by immunoblot showed that endogenous HIF1α only accumulated in the GFPHIGH population ( Figure 2C , D ) . Moreover , when we overexpressed a CRISPR resistant TMEM199 in these TMEM199 sgRNA-targeted cells , the accumulation of HIF1α was reversed ( Figure 2E ) . We also isolated TMEM199 knockout ( KO ) clones , and while most TMEM199 knockouts were lethal after three weeks , a small number of clones that showed undetectable levels of endogenous TMEM199 by immunoblot had elevated HIF1α levels despite several passages ( Figure 2F ) . Reconstituting these clonal cells with overexpressed TMEM199 decreased HIF1α levels ( Figure 2F ) , further confirming the effect of TMEM199 depletion on HIF1α accumulation . 10 . 7554/eLife . 22693 . 004Figure 2 . TMEM199 and CCDC115 are the human orthologues of the yeast Vma12p-Vma22p V-ATPase assembly complex . ( A ) Enriched gene trap insertion sites in TMEM199 identified in the forward genetic screen . ( Red = sense insertions , Blue = antisense insertions ) . ( B ) Schematic for TMEM199 ( left ) and Vma12p ( right ) membrane topology . TMEM199 and Vma12p demonstrate 23 . 89% sequence identity ( Clustal Omega tool ( EMBL-EBI ) ) . ( C , D ) HIF1α-GFPODD reporter cells transduced with Cas9/TMEM199 sgRNA were sorted into GFPLOW ( Lo ) and GFPHIGH ( Hi ) populations by FACS ( C ) , lysed , and immunoblotted for endogenous HIF1α and TMEM199 ( D ) . PHD2 and β2m were used as positive and negative controls respectively , and β-actin served as a loading control . ( E , F ) TMEM199 reconstitution decreases HIF1α levels in TMEM199 deficient cells . TMEM199 KO clones were isolated following lentiviral transduction with sgRNA to TMEM199/Cas9 and serial dilution . Null clones were identified by immunoblot . A CRISPR resistant TMEM199 was overexpressed by lentiviral transduction in mixed populations of TMEM199 deficient cells ( E ) or clonal cells ( F ) . HIF1α and TMEM199 levels were measured by immunoblot , and short and long exposures of TMEM199 levels are shown ( E ) . ( G ) Co-immunoprecipitation coupled mass spectrometry . Wildtype HeLa cells and TMEM199 null cells were lysed in 1% NP-40 and immunoprecipitated for TMEM199 for 3 hr . Samples were validated by immunoblotting and submitted for mass spectrometry analysis . Proteins immunoprecipitated in wildtype HeLa compared to TMEM199 KO cells with a unique peptide count >2 are shown . ( H ) PyMOL structural alignment of CCDC115 ( pink ) and Vma22p ( green ) based on Phyre2 predictions . ( I ) Immunoprecipitation of FLAG-CCDC115 with endogenous TMEM199 in wildtype ( + ) or TMEM199 deficient ( - ) HeLa cells . An unrelated FLAG tagged protein ( FLAG-Ct ) was used as a control . The lysate inputs and immunoprecipitated samples are shown . *non-specific band . ( J , K ) HIF1α-GFPODD reporter cells were depleted of CCDC115 by transduction with Cas9 and sgRNA . After 12 days , cells were sorted into GFPLOW ( Lo , grey box , left ) and GFPHIGH ( Hi , grey box , right ) populations by FACS ( J ) , and immunoblotted for endogenous HIF1α ( K ) . β-actin served as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 004 Human TMEM199 mutations have been recently associated with glycosylation defects ( Jansen et al . , 2016c ) , but the role of TMEM199 in V-ATPase function was not known . To determine if TMEM199 was involved in the V-ATPase complex , we immunoprecipitated endogenous TMEM199 from wildtype HeLa cells and analysed the associated proteins by mass spectrometry , using the TMEM199 KO HeLa clones as a control ( Figure 2G ) . Six proteins were identified as associating with TMEM199 with high confidence compared to the TMEM199 KO cells ( Figure 2G ) . Of these , two were V-ATPase subunits , ATP6V0D1 and ATP6V0A2 , which were also identified in the genetic screen ( Figure 1A , C ) , and have been shown to associate with Vma12p in yeast ( Graham et al . , 1998 ) . A further protein , CCDC115 ( coiled-coil domain-containing protein 115 ) , was of particular interest as a 3-dimensional structural prediction analysis ( Kelley et al . , 2015 ) showed high structural homology to the yeast V-ATPase assembly factor Vma22p ( Figure 2H ) known to bind Vma12p ( Graham et al . , 1998 ) , and human mutations in CCDC115 have been recently reported to show glycosylation defects ( Jansen et al . , 2016b ) similarly to the TMEM199 mutations . Immunoprecipitation of endogenous TMEM199 with FLAG-tagged CCDC115 confirmed the interaction identified by mass spectrometry ( Figure 2I ) , and depleting CCDC115 from HeLa HIF1α-GFPODD cells stabilised the fluorescent reporter and endogenous HIF1α similarly to TMEM199 deletion ( Figure 2J , K ) . Thus , it was likely that TMEM199 and CCDC115 were mammalian homologues of the yeast V-ATPase assembly proteins . Yeast Vma12p-Vma22p localise to the endoplasmic reticulum ( ER ) , where they are thought to be involved in V-ATPase assembly ( Graham et al . , 1998 ) . Using cell fractionation experiments we observed that TMEM199 was only present in the total membrane pool and not in the soluble fraction ( Figure 3A ) . Furthermore , immunofluorescence microscopy showed endogenous TMEM199 localised predominantly to the ER , rather than endosomal or lysosomal compartments ( Figure 3B , C ) . While it was not possible to visualise endogenous CCDC115 by immunoblot or fluorescence microscopy , we observed co-localisation of endogenous TMEM199 with HA-CCDC115 at the ER , although CCDC115 is mostly cytosolic ( Figure 3D , E ) . These results suggest that TMEM199 and CCDC115 form a complex at the ER , analogous to the yeast Vma12p-Vma22p V-ATPase assembly proteins ( Graham et al . , 1998 ) . 10 . 7554/eLife . 22693 . 005Figure 3 . TMEM199 and CCDC115 localise to the ER . ( A ) HeLa cells were homogenised and separated into membrane and cytosolic fractions by ultra-centrifugation . The samples were analysed by immunoblotting for TMEM199 . Calnexin was used as a loading control for membrane compartments , whilst tubulin was used as a control for cytosolic fractions . ( B , C ) Representative immunocytochemical staining for endogenous TMEM199 ( red ) with the ER marker KDEL , the golgi apparatus marker TGN46 , early endosome marker EEA1 , late endosome marker M6PR and lysosomal marker LAMP-1 ( all in green ) ( B ) . Scale bar represents 5 µm . Quantification of colocalisation for TMEM199 and the respective organelle markers using Pearson’s Correlation Coefficient ( C ) n ≥ 16 cells . ( D , E ) Confocal immunofluorescence microscopy of HeLa cells transduced with HA-CCDC115 ( green ) and endogenous TMEM199 ( red , top ) or KDEL ( red , bottom ) ( D ) . Scale bar represents 10 μm . Quantification of colocalisation for CCDC115 with TMEM199 or KDEL using Pearson’s Correlation Coefficient ( E ) n ≥ 50 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 005 As a major function of the V-ATPase is facilitating the lysosomal degradation of proteins by acidification of endosomal-lysosomal compartments , we examined whether TMEM199 or CCDC115 were required for the degradation of a known lysosomal substrate , epidermal growth factor receptor ( EGFR ) . We sorted for TMEM199 or CCDC115 KOs in HIF1α-GFPODD reporter Hela cells by FACS , and measured EGFR degradation following EGF stimulation using BafA as a control ( Figure 4A–C ) . EGF treatment stimulated the degradation of EGFR at 45 min in the control cells , with near complete loss of the receptor by 3 hr , consistent with prior reports ( Mizuno et al . , 2005 ) ( Figure 4A–C ) . This degradation was prevented with BafA treatment ( Figure 4A ) , confirming the role of the V-ATPase in the degradation of EGFR . However , EGF had no effect on the levels of EGFR in the TMEM199 or CCDC15 null cells , and no degradation was detected over 3 hr ( Figure 4B , C ) . Indeed , TMEM199 and CCDC115 deficient cells prevented EGFR degradation similarly to BafA treatment ( Figure 4A ) . 10 . 7554/eLife . 22693 . 006Figure 4 . TMEM199 and CCDC115 and are required for lysosomal degradation of EGFR and MHC Class I . ( A ) EGFR degradation assay for wildtype and BafA treated cells . HeLa cells were cultured in the presence or absence of 10 nM BafA for 24 hr . Cells were stimulated with EGF and lysed at the indicated times . Lysates were subjected to SDS-PAGE and immunoblotted for EGFR . β-actin was used as a loading control . ( B , C ) EGFR degradation assay for TMEM119 and CCDC115 deficient cells . HIF1α-GFPODD cells were transduced with Cas9 and sgRNA to TMEM199 ( B ) or CCDC115 ( C ) . After 14 days , cells were sorted into TMEM199 or CCDC115 sufficient ( +/+ , GFPLOW ) , and TMEM199 or CCDC115 null ( −/− , GFPHIGH ) populations as described . Cells were then cultured for 24 hr before stimulation with EGF ( 100 ng/ml ) , harvested at indicated times and immunoblotted for EGFR . ( D ) MHC Class I degradation in HeLa cells expressing K3 . HeLa-K3 cells were transduced with Cas9 and sgRNA to TMEM199 , CCDC115 , ATP6V1A1 or ATP6V0D1 . After 14 days , cell surface MHC Class I levels were measured by flow cytometry ( mAb W6/32 ) . Wildtype HeLa cells were used as a control for total MHC Class I . Percentages of cells with MHC Class I at the cell surface are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 006 The requirement for TMEM199 and CCDC115 in lysosomal degradation was not limited to EGFR , as we observed impaired degradation of another lysosomal substrate , Major Histocompatibility Complex ( MHC ) Class I . The degradation of cell surface MHC Class I molecules by the Kaposi Sarcoma Herpesvirus ( KSHV ) gene product K3 is a well characterised pathway dependent on ubiquitination and lysosomal degradation ( Coscoy and Ganem , 2000; Hewitt et al . , 2002; Ishido et al . , 2000 ) . Therefore , we used HeLa cells expressing K3 , which have low resting levels of cell surface MHC Class I ( Figure 4D ) , and transduced these cells with sgRNA targeting TMEM199 , CCDC115 or core V-ATPase subunits ( ATP6V1A1 , ATP6V0D1 ) . Mixed populations of TMEM199 or CCDC115 deficient cells partially rescued MHC Class I at the cell surface ( Figure 4D ) . Similar findings were observed with depletion of the core V-ATPase subunits , consistent with a role for TMEM199 and CCDC115 in V-ATPase facilitated lysosomal degradation ( Figure 4D ) . As the V-ATPase is also required to acidify endosomal compartments , we explored the role of TMEM199 and CCDC115 in endosomal acidification using a fluorescent pH sensitive transferrin receptor ( Tfnr-phl ) ( Merrifield et al . , 2005 ) . This construct encodes a super-ecliptic GFP phlourin attached to the extracellular domain of the receptor , which is quenched on transition from pH 7 to 5 ( Merrifield et al . , 2005 ) ( Figure 5—figure supplement 1A ) , and therefore not visible in acidified compartments . To examine how endosomal pH ( e . g . Tfnr-phl fluorescence ) was associated with HIF1α stabilisation , we substituted GFP in our HIF1α reporter construct to generate HIF1α-mCherryODD reporter cells . Live cell microscopy of these cells expressing Tfnr-phl showed GFP fluorescence only at the cell surface ( Figure 5A , B ) , as the transferrin receptor typically recycles between the plasma membrane and recycling vesicles ( Maxfield and McGraw , 2004 ) , and is quenched in the acidic endosomal compartments . Confocal fluorescence microscopy of fixed Tfnr-phl expressing cells ( i . e . no longer pH sensitive ) confirmed that the receptor was still present in endosomal compartments ( Figure 5—figure supplement 1B ) . However , live cell microscopy of HIF1α-mCherryODD reporter cells treated with BafA revealed Tfnr-phl within intracellular vesicles , confirming that V-ATPase inhibition prevented acidification of endosomal compartments ( Figure 5A , B ) . Indeed , the localisation of Tfnr-phl following BafA treatment was similar to the confocal fluorescence microscopy of fixed Tfnr-phl expressing cells ( Figure 5—figure supplement 1B ) . Moreover , mCherry fluorescence was only observed in the BafA treated cells ( Figure 5A ) , confirming that V-ATPase inhibition led to decreased endosomal acidification and HIF1α stabilisation . 10 . 7554/eLife . 22693 . 007Figure 5 . TMEM199 and CCDC115 are required for acidification of endosomal compartments . ( A , B ) Live cell confocal microscopy of HIF1α-mCherryODD reporter cells transfected with the pH sensitive Tfnr-phl . HIF1α-mCherryODD reporter cells were transfected with Tfnr-phl and treated with or without 10 nM BafA for 24 hr . Lower ( LM ) and Higher ( HM ) magnifications of representative BafA treated cells are shown ( A ) . Quantification of intracellular Tfnr-phl/total Tfnr-phl fluorescence in BafA treated cells compared to no treatment ( ≥9 cells ) ( B ) . Additional control experiments are shown in Figure 5—figure supplement 1 . ( C , D ) Live cell confocal microscopy of Tnfr-phl fluorescence in HIF1α-mCherryODD reporter cells depleted for TMEM199 , CCDC115 or core V-ATPase subunits . TMEM199 , CCDC115 , ATP6V1A1 or ATP6V0D1 were depleted by sgRNA as described . After 10–12 days the cells were transfected with Tfnr-phl and live cell fluorescence measured after a further 24 to 48 hr ( C ) . Quantification of intracellular Tfnr-phl/total Tfnr-phl fluorescence is shown ( ≥12 cells ) ( D ) . Representative wide field images are shown in Figure 5—figure supplement 1C . Scale bars represent 10 µm ( A ) or 5 µm ( C ) . Values are mean±SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 00710 . 7554/eLife . 22693 . 008Figure 5—figure supplement 1 . TMEM199 and CCDC115 are required for acidification of endosomal compartments . ( A ) pH clamping of HeLa cells expressing Tfnr-phl . HeLa cells were transfected with Tfnr-phl . After 48 hr the cells were clamped at the indicated pH and fluorescence measured by live cell confocal microscopy . Differential interference contrast ( DIC ) microscopy confirmed the presence of intact cells at pH 5 and 6 . ( B ) Tfnr-phl localisation in fixed cells . HIF1α-mCherryODD reporter cells were transfected with Tfnr-phl as described . After 24 hr the cells were plated on cover slips and treated with or without 10 nM BafA for a further 24 hr . Cells were fixed ( 4% paraformaldehyde ( PFA ) ) prior to confocal microscopy . ( C ) Representative wide field images of Tfnr-phl fluorescence in HIF1α-mCherryODD reporter cells depleted for TMEM199 , CCDC115 or core V-ATPase subunits . Scale bars represent 20 µm ( A ) , 5 µm ( B ) , and 10 µm ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 008 We next examined the effect of TMEM199 or CCDC115 depletion on Tfnr-phl intracellular fluorescence in HIF1α-mCherryODD HeLa reporter cells ( Figure 5C ) , using mCherry accumulation to identify cells where V-ATPase formation was disrupted . Mixed KO populations of the core V-ATPase subunits ATP6V1A1 or ATP6V0D1 were used as a control . Live cell imaging of transiently transfected Tfnr-phl in mixed KO populations of TMEM199 , CCDC115 , or the core V-ATPase subunits showed that Tfnr-phl intracellular fluorescence was almost entirely observed in cells that stabilised the HIF1α-mCherryODD reporter ( Figure 5C , D and Figure 5—figure supplement 1C ) . Thus , TMEM199 and CCDC115 depletion prevents acidification of endosomes , similarly to BafA treatment or depletion of core V-ATPase subunits . The identification of V-ATPase components as regulators of HIF1α levels by a genetic approach was unexpected , given that depletion of PHD2 and VHL , the principal genes involved in the canonical pathway for HIF1α proteasomal degradation , are sufficient for HIF1 activation . We therefore measured the levels of HIF1α in HeLa cells treated with different V-ATPase inhibitors in comparison to the proteasome inhibitor , MG132 ( Figure 6A ) . Surprisingly , we observed similar levels of HIF1α accumulation following treatment with either V-ATPase inhibitors or MG132 . Furthermore , BafA treatment increased HIF1α levels after just 4 hr ( Figure 6B ) . Rather than observing HIF1α accumulation within endosomal-lysosomal compartments , as would be expected following disruption of the V-ATPase , HIF1α was only observed in the nuclei of cells treated with BafA ( Figure 6C ) or depleted of TMEM199 ( Figure 6D ) . This nuclear increase in HIF1α was sufficient to activate HIF responsive genes , demonstrated by an increase in cell surface and total Carbonic Anhydrase 9 ( CA9 ) when cells were depleted of TMEM199 , CCDC115 or core V-ATPase subunits ( Figure 6E , F ) . V-ATPase inhibition or depletion also stabilised HIF2α in HeLa cells and the HIF target gene Heme oxygenase 1 ( HO-1 ) ( Bertout et al . , 2009 ) ( Figure 6G ) . 10 . 7554/eLife . 22693 . 009Figure 6 . Disrupting the V-ATPase activates HIF1 and HIF2 . ( A ) Immunoblot of HIF1α levels in response to the proteasome inhibitor MG132 , the V-ATPase inhibitor BafA , the lysosomotropic agent Chloroquine , and the oxidative metabolism inhibitor NH4Cl . ( B ) Immunoblot of HIF1α levels in HeLa cells in response to BafA treatment at 0 . 5 , 1 , 2 , 4 , 8 and 24 hr . ( C ) Confocal immunofluorescence microscopy of WT ( top ) and BafA ( bottom ) treated HeLa cells stained for endogenous HIF1α . Cells were treated in the presence or absence of 10 nM BafA for 24 hr before immunofluorescence staining with HIF1α ( white ) . Cells were mounted using DAPI ( blue ) and visualised by confocal microscopy . Scale bar represents 10 μm . ( D ) Immunocytochemical staining to examine HIF1α stabilisation in TMEM199 depleted HIF1α-GFPODD reporter cells . HIF1α-GFPODD reporter cells were depleted of TMEM199 using CRISPR-Cas9 genetics and stained for HIF1α ( white ) , TMEM199 ( red ) and DAPI ( blue ) . Scale bar represents 20 μm . ( E–G ) Levels of HIF1α or HIF2α and their target genes in cells depleted of V-ATPase subunits . HIF1α-GFPODD reporter cells were transduced with sgRNA to the indicated V-ATPase subunits as described . After 14 days , cell surface CA9 was measured by flow cytometry ( E ) . Levels of HIF1α , HIF2α and their targets CA9 and HO-1 were measured by immunoblot ( F , G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 009 Prior studies suggest that CMA may contribute to the degradation of HIF1α ( Hubbi et al . , 2013 ) . To determine if CMA or macroautophagy played a significant role in the regulation of HIF1α in aerobic conditions , we examined the effect of depletion of HSC70 , LAMP2A and ATG16 on HIF1α levels . Depletion of the CMA mediators , HSC70 and LAMP2A , or loss of the macroautophagy protein , ATG16 , had no effect on HIF1α stabilisation in aerobic conditions ( Figure 7A–C ) . Moreover , BafA treatment still stabilised HIF1α even when LAMP2A , HSC70 or ATG16 were depleted ( Figure 7A , C , D ) . We therefore sought alternative explanations to account for the accumulation of HIF1α following V-ATPase inhibition , and focused on the principle mechanism for regulating HIF1α stability , prolyl-hydroxylation . 10 . 7554/eLife . 22693 . 010Figure 7 . V-ATPase depletion or inhibition stabilises HIF1α in a non-prolyl hydroxylated form . ( A ) HIF1α stabilisation in ATG16 null HeLa cells . HeLa cells and ATG16 null cells were treated with increasing concentrations of BafA ( 10 nM and 100 nM ) before immunoblotting for HIF1α . ( B ) HIF1α levels following depletion of HSC70 and LAMP2A in aerobic conditions . HSC70 and LAMP2A depleted cells were generated using CRISPR-Cas9 gene editing with three individual sgRNAs ( g1 , g2 , g3 ) . HIF1α , LAMP2A and HSC70 levels were visualised by immunoblot . Untreated ( Ct ) and BafA treated HeLa cells were used as controls . ( C ) HIF1α levels following siRNA-mediated depletion of HSC70 . HeLa cells were transfected with siRNA to HSC70 or an siRNA control ( Ct ) , and HIF1α or HSC70 levels measured by immunoblot after 96 hr . Cells were treated with or without 10 nM BafA for 24 hr prior to lysis . ( D ) LAMP2A deficient HeLa cells were treated with or without 10 nM BafA for 24 hr . Three different sgRNAs were used ( g1 , g2 , g3 ) . ( E , F ) Immunoblot of total HIF1α and the prolyl hydroxylated form in response to MG132 , DMOG , BafA and Chloroquine ( E ) . Quantification of immunoblots represented using ImageJ analysis ( F ) ( n = 3 ) . ( G , H ) In vitro prolyl hydroxylation of the HIF1αODD protein following incubation with lysates from WT , BafA and DMOG treated HeLa cells . The levels of hydroxylated HIF1α were measured using a prolyl hydroxy-HIF1α specific antibody ( G ) . Quantification of the in vitro hydroxylation assay using ImageJ analysis ( H ) ( n = 3 ) . Values are mean±SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 010 We measured the levels of total and prolyl-hydroxylated HIF1α in HeLa cells treated with several V-ATPase inhibitors , the proteasome inhibitor MG132 , or the PHD inhibitor DMOG ( dimethyloxaloylglycine ) using a HIF1α prolyl hydroxyl-specific antibody ( Figure 7E , F ) . While all inhibitors increased total HIF1α levels , only MG132 resulted in the accumulation of prolyl-hydroxylated HIF1α ( Figure 7E , F ) . Conversely , BafA and chloroquine stabilised HIF1α in a non-hydroxylated form , similarly to DMOG treatment ( Figure 7E , F ) . To directly measure if PHD activity was reduced following V-ATPase inhibition , we used an in vitro assay of HIF1α prolyl-hydroxylation ( Burr et al . , 2016 ) , which allows measurements for PHD activity from cell lysates without the addition of excess cofactors ( Figure 7G , H ) . Lysates from wildtype HeLa cells or cells treated with BafA for 24 hr were incubated with a purified His-tagged HIF1αODD protein for 15 min , and hydroxylation measured using the hydroxyprolyl-specific antibody . While HIF1αODD was rapidly hydroxylated in the wildtype HeLa lysate , BafA treatment markedly reduced hydroxylation ( Figure 7G , H ) . Thus , rather than preventing the lysosomal degradation of HIF1α , the V-ATPase inhibition stabilised HIF1α by decreasing PHD enzymatic activity . PHDs are 2-oxoglutarate ( 2-OG ) dependent dioxygenases , which aside from molecular oxygen , require Fe ( II ) and 2-OG as cofactors for catalytic activity . As hydroxylation of HIF1α was impaired by V-ATPase inhibition in aerobic conditions , it was possible that V-ATPase activity altered the level of PHD cofactors . We focused on iron as: ( i ) iron chelators and iron metabolism can alter PHD activity ( Wang and Semenza , 1993 ) , and ( ii ) the V-ATPase is implicated in iron homeostasis via clathrin-mediated endocytosis of transferrin ( Kozik et al . , 2013 ) , the conversion of ferric to ferrous iron within endosomes ( Dautry-Varsat et al . , 1983; Straud et al . , 2010 ) , and the release of iron from ferritin stores ( Mancias et al . , 2014 ) . To examine if V-ATPase inhibition resulted in cytosolic iron depletion , we treated the cells with BafA and measured the levels of IRP2 ( also known as IREB2 ) ; a sensitive maker of intracellular free iron that is rapidly ubiquitinated and undergoes proteasomal degradation in iron-replete cells , but accumulates in iron deficient conditions ( Iwai et al . , 1995; Salahudeen et al . , 2009; Vashisht et al . , 2009 ) . The iron chelator , desferrioxamine ( DFO ) , which has an established role in inhibiting PHD activity and activating the hypoxia pathway , was used as a control for intracellular iron depletion , ( Jaakkola et al . , 2001; Wang and Semenza , 1993 ) . BafA treatment stabilised IRP2 similarly to cells treated with DFO ( Figure 8A ) , consistent with V-ATPase inhibition leading to depletion of intracellular iron . DFO and particularly BafA treatment also increased NCOA4 , the main cargo receptor for the autophagic degradation of ferritin ( ferritinophagy ) and mobilisation of intracellular iron stores ( Mancias et al . , 2014 ) . However , while NCOA4 promoted ferritinophagy in DFO treated cells , ferritin levels were unchanged in BafA treated cells , consistent with a complete block in autophagy when the V-ATPase was inhibited ( Figure 8A ) . Similar increases in IRP2 and NCOA4 were observed in FACS GFPHIGH populations of TMEM199 , CCDC115 , ATP6V0D1 or ATP6V1A1 deficient HIF1α-GFPODD reporter cells ( Figure 8B ) . Together , these findings demonstrate that disrupting V-ATPase integrity and assembly results in intracellular iron depletion . 10 . 7554/eLife . 22693 . 011Figure 8 . Disrupting V-ATPase activity decreases intracellular iron levels . ( A , B ) V-ATPase inhibition leads to intracellular iron depletion . ( A ) HeLa cells were treated with BafA ( 10 nM or 100 nM ) or 100 µM DFO for 24 hr . HIF1α , IRP2 , NCOA4 and ferritin ( ferritin heavy chain 1 , FTH1 ) levels were measured by immunoblot . ( B ) HIF1α-GFPODD reporter cells transduced with Cas9 and sgRNA targeting V-ATPase components ( TMEM199 , CCDC115 , ATP6V0D1 and ATP6V1A1 ) were sorted into GFPLOW ( Lo ) and GFPHIGH ( Hi ) populations as described . The lysates were immunoblotted for HIF1α , IRP2 , or NCOA4 . β-actin served as a loading control . ( C ) Iron chelation prevents HIF1α hydroxylation . In vitro prolyl hydroxylation of the HIF1αODD protein following incubation with lysates from WT or DFO treated lysates ( 100 µM for 24 hr ) as previously described . DMOG served as a control for PHD inhibition . ( D–F ) In vitro hydroxylation of PHD activity in DFO or BafA treated lysates supplemented with ferrous iron . Lysates from control , DFO ( D ) or BafA ( E ) treated HeLa cells were extracted as previously described , incubated with the HIF1αODD protein , and supplemented with increasing concentrations of iron chloride ( FeCl2 , Fe ( II ) ) . Prolyl-hydroxylated HIF1αODD levels were visualised by immunoblot and quantified by densitometry for the BafA treated lysate ( F ) ( n = 3 ) . Values are mean±SEM . *p<0 . 05 , **p<0 . 01 Fe ( II ) compared to no treatment in BafA treated cells . NS=not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 011 We next asked if replenishing ferrous iron levels directly restored HIF1α prolyl hydroxylation . We first validated that ferrous iron restored in vitro prolyl hydroxylation of the HIF1αODD in DFO treated lysates ( Figure 8C , D ) . We then examined the effect of ferrous iron treatment on PHD activity in extracts from BafA treated HeLa cells ( Figure 8E , F ) . The addition of 1 µM Fe ( II ) completely restored prolyl hydroxylation of the HIF1αODD ( Figure 8E , F ) . Indeed , the levels of HIF1α hydroxylation were slightly higher than observed in wildtype cells , although this increase was not significant . To explore if iron supplementation was sufficient to restore HIF1α turnover in cells where the V-ATPase had been inhibited , BafA treated HIF1α-GFPODD reporter or wildtype HeLa cells were supplemented with or without 50 µM iron ( Fe ( III ) ) citrate for 24 hr , and HIF1α levels measured by flow cytometry and immuoblot ( Figure 9A , B ) . While BafA treatment stabilised the GFP reporter and endogenous HIF1α , this was completely prevented by the addition of iron to the media . Similar findings were observed in cells treated with DFO and iron citrate ( Figure 9C , D ) , although the concentration of iron needed to restore HIF1α degradation was higher ( 200 µM Fe ( III ) ) . Iron supplementation also restored HIF1α turnover in BafA treated primary human dermal fibroblasts , HEK293ET cells and RCC10 cells that had been reconstituted with VHL ( Figure 9E–G ) . Furthermore , iron citrate also prevented the downstream activation of several HIF1 target genes ( GLUT1 , VEGF and carbonic anhydrase 9 ) following BafA treatment without affecting HIF1α mRNA levels ( Figure 9J ) . Consistent with iron supplementation only affecting HIF1α levels following V-ATPase inhibition , iron citrate did not affect HIF1α stabilisation following proteasome inhibition or in VHL null RCC10 cells ( Figure 9G , H ) . Iron citrate had no direct effect on lysosomal degradation , as EGFR turnover was still impaired in BafA treated cells following iron supplementation ( Figure 9I ) . 10 . 7554/eLife . 22693 . 012Figure 9 . Iron supplementation restores HIF1 activity to basal levels following V-ATPase inhibition in cell lines and primary cells . ( A–D ) Iron reconstitution in BafA or DFO treated HeLa cells . ( A , C ) HIF1α-GFPODD reporter cells were treated with BafA ( 10 nM or 100 nM ) , or 100 µM DFO for 24 hr with 50 µM iron citrate ( Fe ( III ) ) ( red ) , 200 µM Fe ( III ) ( blue ) or no iron ( green ) , and GFP levels analysed by flow cytometry . ( B , D ) Wildtype HeLa cells were treated with or without BafA ( 10 nM or 100 nM ) or DFO ( 100 µM ) and Fe ( III ) ( 50–200 µM ) as described , and endogenous HIF1α levels were measured by immunoblot . ( E–G ) HEK293T cells ( E ) , human dermal fibroblasts ( F ) and RCC10 VHL null and VHL reconstituted cells ( G ) were treated with BafA ( 10 nM ) with or without the addition of 50 µM iron citrate ( Fe ( III ) ) . HIF1α levels were visualised by immunoblot . β-actin served as a loading control . ( H ) HeLa cells were treated with 20 µM MG132 for 2 hr or 10 nM BafA for 24 hr with or without the addition of 50 µM iron citrate . ( I ) EGFR degradation assay for BafA treated cells following iron treatment . HeLa cells were cultured with 10 nM BafA for 24 hr , with or without 50 µM iron citrate ( Fe ( III ) ) , and stimulated with EGF as previously described . EGFR , NCOA4 , and ferritin ( FTH1 ) levels were visualised by immunoblot . β-actin was used as a loading control . ( J ) RT-qPCR analysis of HIF1α and its target genes in response to BafA and iron citrate treatment ( n ≥ 2 ) . ( K–M ) Populations of mixed CRISPR KO cells for ATP6V1A1 ( K ) , TMEM199 ( L ) and CCDC115 ( M ) were treated with 50 µM iron citrate for 24 hr and HIF1α levels measured by immunoblot . Values are mean±SEM . *p<0 . 05 , **p<0 . 01 . NS = not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 012 Although iron citrate restored HIF1α to basal levels following BafA treatment , it was unclear whether HIF1α stabilisation following genetic disruption of the V-ATPase could be prevented by iron supplementation . Therefore , iron citrate was supplemented to the media of mixed KO populations of individual V-ATPase subunits ( TMEM199 , CCDC115 or ATP6V1A1 ) and HIF1α levels measured by immunoblot ( Figure 9K–M ) . Iron treatment reversed the HIF1α stabilisation in the sgRNA-targeted cells ( Figure 9K–M ) , confirming that iron supplementation was sufficient to restore HIF1α degradation following genetic disruption of the V-ATPase . While it was evident that V-ATPase inhibition resulted in iron depletion , we wanted to explore the relative importance of disrupting iron uptake compared to mobilisation of iron stores . Moreover , the contribution of transferrin independent iron uptake to the intracellular pool was unclear ( Liuzzi et al . , 2006; Oudit et al . , 2006 ) . We examined if treating cells with ferric iron prior to BafA treatment prevented HIF1α accumulation and conversely , also measured if iron supplementation after BafA treatment affected HIF1α levels ( Figure 10A ) . Pre-conditioning the media with iron citrate did not prevent the accumulation of HIF1α following BafA treatment ( Figure 10A ) . However , iron supplementation at the time of BafA treatment or after 12 hr did decrease HIF1α levels ( Figure 10A ) , suggesting that iron supplementation restores HIF1α turnover following disruption of the V-ATPase by a transferrin independent mechanism . 10 . 7554/eLife . 22693 . 013Figure 10 . Disrupting transferrin uptake leads to iron-dependent HIF1 activation . ( A ) The effect of iron treatment on HIF1α stabilisation in the presence or absence of BafA . HeLa cells were treated with 50 µM iron citrate ( red ) or 10 nM BafA ( blue ) for the indicated times and HIF1α levels measured by immunoblot . β-actin served as a loading control . ( B–E ) The effect of NCOA4 or IRP2 depletion on HIF1α levels and intracellular iron levels . HIF1α-GFPODD reporter HeLa cells were transduced with Cas9 and sgRNAs to NCOA4 ( B , C ) or IRP2 ( D , E ) and HIF1α levels measured by GFP accumulation ( B , D ) top ) and immunoblot after 8 ( IRP2 ) or 12 days ( NCOA4 ) ( C , E ) . Cell surface transferrin levels were measured by flow cytometry ( B , D ) bottom ) . IRP2 and ferritin ( FTH1 ) levels were visualised by immunoblot ( C , E ) . An sgRNA to PHD2 was used as a control . ( F ) HIF1 activation following depletion of the transferrin receptor . HIF1α-GFPODD reporter HeLa cells were transduced with Cas9 and sgRNAs to the transferrin receptor . HIF1 activation was measured by GFP accumulation and cell surface expression of CA9 ( top ) . Depletion of the transferrin receptor was measured by flow cytometry ( bottom ) ( G , H ) Iron reconstitution restores HIF1α turnover in transferrin receptor deficient cells . HIF1α-GFPODD reporter HeLa cells were transduced with Cas9 and sgRNAs to the transferrin receptor as described , and 50 µM ferrous citrate added to media for 24 hr . HIF1α levels were measured by flow cytometry for GFP fluorescence ( G ) or immunoblot ( H ) . IRP2 , NCOA4 and FTH1 were visualised to measure intracellular iron . β-actin served as a loading control . Additional experimental examples with alternative sgRNAs to NCOA4 , IRP2 or the transferrin receptor are shown in Figure 10—figure supplement 1 . TFR=transferrin receptor . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 01310 . 7554/eLife . 22693 . 014Figure 10—figure supplement 1 . Disrupting transferrin uptake leads to iron-dependent HIF1 activation . ( A ) Flow cytometry of mixed populations of NCOA4 sgRNA-targeted HIF1α-GFPODD reporter HeLa cells 8 days to 12 days post transduction . Three different sgRNA were used . ( B ) HIF1α-GFPODD reporter and transferrin receptor levels using sgRNA2 to the transferrin receptor . ( C ) HIF1α-GFPODD reporter and transferrin receptor levels using two different sgRNA to IRP2 . ( D ) HIF1α levels following depletion of IRP2 at different time points post transduction . Mixed populations of IRP2 sgRNA-targeted HIF1α-GFPODD reporter HeLa cells 10 or 12 days post transduction using three different sgRNA . ( E ) HIF1α-GFPODD reporter levels in mixed populations of transferrin deficient cells ( using sgRNA2 to the transferrin receptor ) with or without iron citrate treatment . TFR=transferrin receptor . DOI: http://dx . doi . org/10 . 7554/eLife . 22693 . 014 To further explore the relative contributions of iron uptake and ferritin turnover to intracellular iron homeostasis , we depleted cells of transferrin receptor , IRP2 or NCOA4 by sgRNA , and used HIF1α stability as a sensitive functional measure of iron levels . Mixed KO populations of NCOA4 cells had no effect on HIF1α levels ( Figure 10B , C – Figure 10—figure supplement 1A ) . Interestingly , NCOA4 depletion did increase IRP2 levels after 12 days , suggesting that IRP2 may increase iron flux through the endosomal pathway , but cell surface transferrin receptor and total ferritin levels did not change ( Figure 10B , C ) Depletion of either IPR2 or the transferrin receptor stabilised the HIF1α-GFPODD reporter and endogenous HIF1α , although this was most marked in the transferrin receptor null cells ( Figure 10D–F , H – Figure 10—figure supplement 1B–D ) . IRP2 depletion also did not alter cell surface transferrin receptor levels ( Figure 10D ) , consistent with IRP2 regulating iron flux rather than just transferrin receptor expression . Thus , ferritinophagy does not seem to contribute significantly to cytosolic Fe ( II ) levels when the transferrin pathway is functional . Lastly , we examined if iron supplementation restored HIF1α turnover in transferrin receptor deficient cells . Mixed transferrin receptor KO populations were treated with iron citrate for 24 hr and HIF1α levels measured . Iron treatment decreased HIF1α levels and increased ferritin stores without altering NCOA4 levels ( Figure 10G , H – Figure 10—figure supplement 1E ) . Thus , when the transferrin pathway is impaired , either by depletion of the transferrin receptor or disrupting V-ATPase activity , increasing the availability of extracellular Fe ( III ) is sufficient to restore PHD activity and HIF1α degradation .
The use of a forward genetic approach to examine cellular processes that regulate HIF1α revealed that disruption of the V-ATPase complex stabilised HIF1α in aerobic conditions . HIF1α accumulation following inhibition of V-ATPase activity was unexpected , given the clear role for the proteasome in degrading ubiquitinated HIF1α in normal oxygen tensions . Rather than V-ATPase inhibition directly preventing HIF1α degradation , we find that HIF1α is stabilised due to depletion of intracellular iron , which results from decreased transferrin uptake and reduced conversion to ferrous iron within endosomal compartments . These findings may have diverse physiological implications , particularly as germline mutations in V-ATPase subunits are associated with several human diseases . Homozygous mutations in a V0 subunit isoform ( ATP6V0A3 ) cause infantile osteopetrosis ( failure of bone resorption ) ( Kornak et al . , 2000 ) , while mutations in kidney specific isoforms ( ATP6V0A4 ) cause distal tubular renal acidosis ( Karet et al . , 1999; Smith et al . , 2000 ) . Furthermore , several cancer cell types express V-ATPase complexes at the plasma membrane ( Capecci and Forgac , 2013 ) , possibly as a mechanism for regulating cytosolic pH , which appear to render tumour cells more susceptible to cell death when the V-ATPase is inhibited ( Perut et al . , 2014 ) . Given the diverse role of the V-ATPase , it will be of future interest to examine whether known germline or somatic human mutations in V-ATPase subunits impact on the HIF pathway in animal models . While HIF1α degradation by proteasome independent mechanisms , such as CMA , has been reported ( Bremm et al . , 2014; Ferreira et al . , 2013; Hubbi et al . , 2014 , 2013; Selfridge et al . , 2016 ) , we did not observe any effect on HIF1α levels in HeLa cells when key mediators of CMA were depleted ( HSC70 and LAMP2A ) . The ability of iron treatment to completely restore HIF1α turnover when the V-ATPase is inhibited , without altering the lysosomal degradation of EGFR or the autophagy cargo receptor NCOA4 , also argues against a role for CMA in HIF1α regulation . It remains possible that non-proteasomal degradation of HIF1α may occur in certain cell types and under particular conditions . The requirement for the AAA ATPase P97 in HIF1α degradation suggests that it may be incorporated into larger complexes that require unfolding prior to proteolysis ( Alexandru et al . , 2008 ) . Nevertheless , we find that the major consequence of V-ATPase inhibition is stabilisation of HIF1α by preventing prolyl hydroxylation . Our findings are consistent with prior studies showing impaired uptake of transferrin following V-ATPase inhibition ( Kozik et al . , 2013; Straud et al . , 2010 ) , and confirm the importance of V-ATPase activity in regulating iron through clathrin-mediated endocytosis ( Kozik et al . , 2013 ) , endosomal acidification ( Eckenroth et al . , 2011; Ohgami et al . , 2005 ) and ferritinophagy ( Mancias et al . , 2014 ) . Directly measuring the levels of free intracellular iron within the cytosol is challenging , as reagents typically used for these assays rely on indirect enzymatic assays , and most Fe ( II ) is rapidly bound to enzymes or incorporated into the biosynthesis of iron-sulphur complexes . However , IRP2 is a sensitive marker of intracellular iron availability ( Iwai et al . , 1995; Salahudeen et al . , 2009; Vashisht et al . , 2009 ) , and its induction following V-ATPase inhibition is consistent with depletion of ferrous iron . Furthermore , our in vitro hydroxylation assay clearly shows that supplementation with Fe ( II ) can completely restore prolyl hydroxylation of HIF1α , confirming that V-ATPase inhibition reduces the available pool of intracellular Fe ( II ) . The V-ATPase is not only implicated in iron uptake and conversion to Fe ( II ) , but also required for the release of iron from ferritin stores via ferritinophagy . Using HIF1α stabilisation as a sensitive marker for intracellular iron , we find that inhibition or disruption of the V-ATPase both alters iron uptake and degradation of ferritin . However , HIF1α was only stabilised in the transferrin receptor deficient cells , implying that increased ferritin turnover is not sufficient to compensate for prolonged loss of transferrin-mediated iron uptake . Conversely , while loss of NCOA4 activated IRP2 , there was still sufficient intracellular iron for PHDs to function and HIF1α was not stabilised . These findings are consistent with decreased transferrin uptake and conversion of Fe ( III ) to Fe ( II ) being the predominant mechanisms for reduced PHD activity when the V-ATPase is inhibited , and highlight the relative importance of iron uptake versus release from intracellular stores . Interestingly , Fe ( III ) supplementation to the media was sufficient to restore HIF1α turnover when the V-ATPase was inhibited or following depletion of the transferrin receptor . Thus , transferrin-independent uptake of iron can occur when there is sufficient extracellular iron availability , presumably by transporters such as Zip14 , which mediate non-transferrin bound iron uptake ( Liuzzi et al . , 2006 ) . It will be of interest to explore the biological implications of transferrin-independent mechanisms and the role of ferritinophagy in future studies . Moreover , it is plausible that HIF activation and V-ATPase activity serve as a feedback mechanism to control intracellular iron , particularly as HIFs promote genes involved in iron metabolism ( Pantopoulos et al . , 2012; Peyssonnaux et al . , 2008; Simpson and McKie , 2015 ) . The identification of TMEM199 and CCDC115 as V-ATPase assembly proteins has interesting implications for our understanding of how the human V-ATPase complex forms . In yeast , Vma12p and Vma22p dimerise and are thought to promote the assembly of the membrane embedded complex ( V0 ) , which can subsequently associate with the peripheral complex ( V1 ) ( Graham et al . , 1998; Hill and Stevens , 1995; Hirata et al . , 1993; Jansen et al . , 2016c ) . Similarly to the yeast Vma12 studies , we find that TMEM199 is predominantly localised to the endoplasmic reticulum , implying that it is likely to be involved in V-ATPase assembly , rather than forming part of the mature complex . Human germline mutations in TMEM199 and CCDC115 cause disorders with a liver storage disease phenotype ( Jansen et al . , 2016b , 2016c ) and it remains to be determined how impaired glycosylation relates to V-ATPase function , or if these defects relate to other roles of TMEM199 and CCDC115 . However , it is of interest that all described TMEM199 and CCDC115 human mutations have impaired glycosylation of transferrin , suggesting a role for the V-ATPase associated factors in iron metabolism ( Jansen et al . , 2016b , 2016c ) . ATP6AP1 ( also known as Ac45 ) was also identified in our forward genetic screen and has recently been identified as the human orthologue of the yeast Voa1 V-ATPase assembly factor ( Jansen et al . , 2016a ) . Human mutations in ATP6AP1 have a distinct phenotype to that of TMEM199 and CCDC115 mutations , resulting in immunodeficiency as well as abnormal glycosylation ( Jansen et al . , 2016a ) . Furthermore , human mutations in another V-ATPase assembly factor , VMA21 , cause an autophagic myopathy ( Ramachandran et al . , 2013 ) . The explanation for the diverse nature of human diseases caused by mutations in V-ATPase assembly factors is unclear . Moreover , while yeast studies support the role of Vma12p , Vma21p and Vma22p in V-ATPase assembly , this has not been studied in mammalian cells , where they may serve additional regulatory functions on V-ATPase activity . Addressing the pathological role of these human V-ATPase mutations on iron metabolism and HIFs in animal models will be important in future studies . Nevertheless , our findings show that V-ATPase activity can alter HIF signalling through regulating intracellular iron .
HeLa , HEK293T and RCC10 cells were cultured and maintained at 37°C with 5% ( v/v ) CO2 in Dulbecco’s modified Eagle Medium supplemented with 10% ( v/v ) Fetal Calf Serum ( HyClone ) and 100IU/ml Penicillin G and 100 µg/ml Streptomycin . Primary dermal fibroblasts ( Lonza ) were cultured under the same conditions but with 20% ( v/v ) FCS . ATG16 null HeLa cells were a kind gift from David Rubinsztein ( University of Cambridge ) . HeLa and HEK293 cells were originally a gift from Paul Lehner ( University of Cambridge ) . RCC10 cells were a gift from Patrick Maxwell ( University of Cambridge ) . HeLa , HEK293 and RCC10 cells were authenticated by STR profiling ( Eurofins Genomics ) . Human primary dermal fibroblasts were authenticated by Lonza . All cells were confirmed to be mycoplasma negative ( Lonza MycoAlert ) . Primary antibodies were prepared for immunoblotting as follows: TMEM199 ( Atlas , HPA027051 , 1:2000 ) , LAMP2A ( Abcam , ab18528 , 1:1000 ) , HSC70 ( Abcam , ab19136 , 1:1000 ) , EGFR ( Santa Cruz , Sc-03 , 1:500 ) , HIF1α ( BD Transduction Laboratories , 610959 , 1:1000 ) , β-actin ( Sigma , A228 , 1:30000 ) , Calnexin ( Abcam , ab75801 , 1:1000 ) , Tubulin ( eBioscience 14–4502 , 1:1000 ) , ATG16L ( MBL , PMO40 , 1:1000 , Gift from Rubinzstein lab , CIMR ) , M2-FLAG ( Sigma , F3165 , 1:5000 ) , CCDC115 ( Atlas , HPA034598 , 1:1000 ) , PHD2 ( Novus bioscience , NB100-137 , 1:5000 ) , Hydroxy-HIF-1α ( Cell signaling technology , 3434 , 1:2000 ) , β2M ( Dako , A00072 , gift from Paul Lehner , CIMR , 1:10000 ) , NCOA4 ( ARA70 , Bethyl Laboratories , A302-272A-T , 1:2000 ) , IRP2 ( IREB2 , Cell Signalling Technology , 37135 , 1:1000 ) , Ferritin heavy chain ( FTH1 , Cell Signalling Technology , 3998 , 1:1000 ) , Carbonic anhydrase 9 ( CA9 , clone M75 , gift from Egbert Oosterwijk , Radboud , 1:5000 ) . Immunofluorescence antibodies: LAMP1 ( Santa Cruz , SC-20011 , 1:100 ) , TGN46 ( AbD Serotec , AHP500GT , 1:200 ) , KDEL ( gift from Geoff Butcher , Babraham Institute , 1:400 ) , EEA1 ( gift from Evan Reid , 1:250 ) , M6PR ( gift from Evan Reid , 1:500 ) , HIF1α ( 1:100 ) , TMEM199 ( 1:50 ) , HA-11 ( Ms Covance , MMS-101P , 1:100 ) , rabbit polyclonal to HA-11 ( gift from Paul Lehner , 1:100 ) . Antibodies for cell-surface staining: MHC Class I ( W6/32 , gift from Paul Lehner , CIMR , 1:100 ) , Transferrin receptor ( TFRC , CD71 , BD Pharmigen , 555534 , 1:500 ) , CA9 ( 1:500 ) . The following reagents were used: MG132 ( Sigma-Aldrich , 20 µM ) , Bafilomycin A1 ( Alfa Aesar , J61835 , 10 nM-100nM ) , Ammonium Chloride ( Sigma-Aldrich , A9434 , 10 mM ) , DMOG ( Sigma-Aldrich , D3695 , 0 . 5 mM ) , Chloroquine ( Sigma-Aldrich , C6628 , 50 µM ) , DFO ( Sigma-Aldrich , D9533 , 100 µM ) . Puromycin , Hygromycin , and Blasticidin were purchased from Cambridge Bioscience ( all used at 10 µg/ml ) . ProLong Gold Antifade Reagent with DAPI ( 8961 , Cell signaling technology , 8961 ) . The following plasmids were used . LentiCRISPR v2 ( sgRNA/Cas9 , F . Zhang Addgene #52961 ) , pHRSIN-pSFFV-FLAG-MPP8-pPGK-Blasto ( Gift from Paul Lehner ) , TMEM199 Image Clone ( Source Bioscience ) . CCDC115 Image Clone ( Source Bioscience ) , pHRSIN-pSFFV-HA-Ube2J2 -pPGK-Puro ( gift from Paul Lehner ) . pMD . G ( Lentiviral VSVG ) , pMD . GagPol ( lentiviral Gag/Pol ) . Lentiviral plasmids used the pHRSIN backbone ( Demaison et al . , 2002 ) . The HIF1α-GFPODD reporter was generated as described ( Burr et al . , 2016 ) . The HIF1α-mCherryODD reporter was generated by excising the GFP and subcloning mCherry CL1 using the BamHI and NotI restriction sites . Tfnr-phl was a gift from Christien Merrifield ( Merrifield et al . , 2005 ) . The KBM7 forward genetic screen was carried out as previously described ( Burr et al . , 2016 ) . Bioinformatic pathway analyses were performed by taking genes enriched for trapping insertions with an adjusted Fisher Exact test p-value<0 . 1 and running an 'express analysis' using Metascape 1 . 0 ( Tripathi et al . , 2015 ) . All statistically enriched terms were identified ( accumulative hypergeometric p-values and enrichment factors were calculated and used for filtering ) , and were hierarchically clustered into a tree based on Kappa-statistical similarities among their gene memberships . Then 0 . 3 kappa score was applied as the threshold to cast the tree into term clusters . We selected a term with the best p-value as the representative term for each cluster and show the p-values of all clusters in a bar graph . Structures of CCDC115 and Vma22p were predicted using the Protein Homology/Analogy Recognition Engine v2 . 0 ( Phyre2 ) ( Kelley et al . , 2015 ) . These were then imported into the PyMOL Molecular Graphics System ( Schrödinger LLC ) . Images were rendered using the ‘cartoon’ function and aligned using the ‘align’ function . Lentivirus was prepared by transfection using Mirus Trans-IT−293 Transfection Reagent in Hek293ET cells . Cells were triple transfected with pMD . G ( VSVG envelope ) , the suitable lentiviral transgene vector and pCMVR8 . 91 ( gag/pol ) at a ratio 2:3:4 . Transfection was performed in six well plates at 70–80% confluency . The resultant viral supernatant was harvested at 48 hr , filtered through a 0 . 45 µm filter and stored at −80°C . To achieve stable integration by transduction , cells were seeded to 24 well plates in 500 µl DMEM . Viral supernatant ( 500 µl ) was added to each well and plates spun at 1800 rpm for 1 hr at 37°C . Plates were incubated post spin and selected for antibiotic resistance after 48 hr . Cells were harvested , washed in PBS before fixing in 1% paraformaldehyde , PBS and analysed on a FACSCalibur ( BD ) . For FACS , cells were harvested , washed in 10 mM Hepes PBS and re-suspended in sorting medium ( 10 mM Hepes , 2% FCS , PBS ) . Cell suspensions were filtered prior to sorting in a High speed Influx Cell Sorter ( BD Biosciences ) . For cell surface staining , cells were washed in PBS and incubated with the appropriate primary antibody for 30 min at 4°C . After a further PBS wash the cells were incubated with secondary antibody for 30 min at 4°C . A final PBS wash was performed and cells were fixed in PBS with 1% paraformaldehyde ( PFA ) and analysed as described . Gene-specific CRISPR sgRNA oligonucleotide sequences were selected using the GeCKO v2 library . Sense and antisense sgRNAs oligonucleotides were designed with 5’ CACC and 3’ CAAA overhangs respectively . The subsequent sgRNAs were cloned into the LentiCRISPRv2 backbone ( Sanjana et al . , 2014 ) or pKLV-U6gRNA ( BbsI ) -PGKpuro2ABFP . The following sgRNAs were used: TMEM199TATGGCGTCCTCTTTGCTTGATP6V0D1TCGATGACTGACACCGTCAGATP6AP1GCTGACTGCATACCAGTCGAATP6V1A1GTAACTTACATTGCTCCACCATP6V0A2GCGACACTCACGTCTCGGAACATP6V1G1GTGAAAACAGGAAAGAACCGB2MGGCCGAGATGTCTCGCTCCGPHD2ATGCCGTGCTTGTTCATGCACCDC115GGGGGCTCACCTGCTTCGCGHSC70 sgRNA1ACCATAGAAGACACCTCCTCHSC70 sgRNA2CTAGACTGTTACCAATGCTGHSC70 sgRNA 3GACAGATGCCAAACGTCTGATLAMP2A sgRNA1ACCAGAACGAGCCCTGAGCCLAMP2A sgRNA2TCCGGGCTCAGGGCTCGTTCLAMP2A sgRNA3CAAGAACATCCCAGTAGTGTNCOA4 sgRNA1GGTATGGCTGTATGAACAGGNCOA4 sgRNA2CAATCTCCACACCTTTGGGCNCOA4 sgRNA3TAGCTGTCCCTTTCAGCGAAIRP2 sgRNA1AATGCACCAAATCCTGGAGGIRP2 sgRNA2TGAGCCATTCCAGTTCCAGGIRP2 sgRNA3GCATAAGCTACCACTAAGGGTransferrin Receptor sgRNA1AAAGTCTCCAGCACTCCAACTransferrin Receptor sgRNA2GCTCTGGAGATTGTCTGGAC 100 nM HSC70 siRNA ( Dharmacon Smartpool , L-017609-00-0005 ) or MISSION siRNA Universal Negative Control ( control ) was transfected into 3 × 105 HeLa cells using Oligofectamine Transfection Reagent ( Invitrogen ) according to the manufacturer’s protocol . Cells were harvested after 96 hr for further analysis by immunoblot . Cells were cultured on glass coverslips , washed in PBS and fixed in 4% ( w/v ) PFA , PBS at room temperature for 20 min . Cells were then blocked and permeabilised in 3% BSA , 0 . 3% TritonX-100 , PBS . For observing endosomal compartments , cells were permeabilised in 3% BSA , 0 . 05% Saponin , PBS . Coverslips were incubated with primary antibody for 1 hr , washed in PBS and fluorophore-conjugated secondary antibodies applied for 30 min . Coverslips were mounted to microscope slides using ProLong Gold antifade with DAPI . Imaging was performed on a Zeiss LSM880 inverted confocal microscope . Cytoplasmic colocalisation analyses of stains were performed using CellProfiler image analysis software ( Carpenter et al . , 2006 ) . The analysis was performed blinded . Nuclear and cell boundaries were identified manually , with the area between the nuclear and cell peripheries classified as the cytoplasm . The ‘Measure Correlation’ module was used to calculate the Pearson’s Correlation Coefficient between the pixel intensities for the described stains . For the Tfnr-phl experiments , wildtype or HIF1α-mCherryODD reporter HeLa cells were transfected with TfR-pHluorin using TransIT-HeLaMONSTER Transfection kit ( Mirus Bio LLC ) according to manufacturer’s protocol . To perform fixed cell immunofluorescence , cells were cultured on glass coverslips , washed in PBS and fixed in 4% ( w/v ) PFA , PBS as previously described . Cells were permeabilised in PBS with 3% BSA , 0 . 05% Saponin , and coverslips mounted to microscope slides using ProLong Gold antifade with DAPI . Imaging was performed on a Zeiss LSM880 confocal microscope . Live cell imaging was performed on a Zeiss LSM780 inverted confocal microscope equipped with a 63x objective . To examine Tfnr-phl localization cells were treated with or without 10 nM BafA for 24 hr prior to imaging . pH clamping was performed during live cell imaging by incubating cells with 25 mM sodium acetate buffer ( pH 5 ) , 25 mM MES buffer ( pH 6 ) and 25 mM Hepes buffer ( pH 7 ) containing 5 mM NaCl , 1 mM CaCl2 , 115 mM KCl , 1 . 2 mM MgSO4 , 10 mM glucose , 10 µM nigericin and 10 µM monensin for 5 min at 37°C as previously described ( Bright et al . , 2016 ) . Fluorescence intensity measurements of Tfnr-phl were performed using CellProfiler image analysis software ( Carpenter et al . , 2006 ) . The analysis was performed blinded . Cell surface and cytoplasmic boundaries were identified manually and the ratio between cytoplasmic fluorescent intensity and whole cell fluorescence was calculated . Cells were lysed in SDS lysis buffer ( 1% SDS , 50 mM Tris pH7 . 4 , 150 mM NaCl , 10% glycerol and 5µl/ml benzonase nuclease ) for 10 min on ice , heated at 90°C for 5 min and centrifuged at 14 , 000 rpm for 10 min . Proteins were separated by SDS-PAGE , transferred to methanol activated Immobilon-P 0 . 45 µm PVDF membrane , probed with appropriate primary and secondary antibodies , and developed using SuperSignal West Pico or Dura Chemiluminescent Substrates ( Thermo Scientific ) . HeLa cells ( 3 × 106 ) were harvested , washed and resuspended in 1 ml break buffer ( 20 mM Hepes pH7 . 4 , 1 mM EDTA , 0 . 5 mM MgCl2 , 0 . 13 M sucrose , 50 mM NaCl , 1 mM PMSF supplemented with Roche complete EDTA free protease inhibitors ) , and passed through an equilibrated ball bearing homogenizer at 1 µm diameter for 20 passes . The homogenized lysate was collected and centrifuged at 3000 rpm for 10 min to pellet nuclei . The post nuclear supernatant was ultracentrifuged at 50 , 000 rpm for 1 hr to pellet intracellular membranes before resuspension in SDS loading buffer for separation by SDS-PAGE and immunoblot . This assay was performed as previously described ( Almeida et al . , 2006 ) , with some modifications . HeLa cells were cultured in the presence or absence of BafA on six well plates prior to serum starvation for 90 min in DMEM supplemented with 2% FCS and cyclohexamide ( 100 µg/ml ) . Following starvation , cells remained in 2% FCS and were stimulated with 100 ng/ml EGF ( Cabiochem ) . The reaction proceeded for 0 , 10 , 45 , 90 and 180 min until quenched on ice . The cells were washed in ice cold PBS and lysed on ice in SDS lysis buffer . EGFR expression levels were probed by immunoblotting . HeLa cells ( 5 × 106 ) were lysed in 1% NP-40 , TBS supplemented with 1x Roche cOmplete EDTA-free protease inhibitor cocktail for 30 min at 4°C before centrifugation at 14 , 000 rpm for 10 min . The supernatants were then diluted to 0 . 1% detergent and pre-cleared with SP sepharose Fast-Flow ( GE Healthcare ) for 1 hr . The pre-cleared supernatant was incubated with 40 μl anti-FLAG M2 magnetic beads ( M8823 , Sigma Aldrich ) for 3 hr at 4°C . The resins were then washed and the bound proteins eluted with 100 μg/ml FLAG peptide ( 30 min ) . The eluted proteins were then separated by SDS-PAGE and immunoblotted as described . For mass spectrometry analysis of TMEM199 associated proteins , wildtype or TMEM199 KO HeLa cells ( 1 × 108 ) were immunoprecipitated with the TMEM199 antibody conjugated to Dynabeads Protein G for 3 hr at 4°C . The resins were then washed and sample eluted with SDS lysis buffer at 70°C for 10 min . Samples were resolved a short distance into an SDS-polyacrylamide gel , the lanes excised and subjected to in-gel tryptic digestion . The resulting peptides were analysed using a Q Exactive ( Thermo Scientific ) coupled to an RSLC3000nano UPLC ( Thermo Scientific ) . Files were searched against a Uniprot human database ( downloaded 09/06/14 , 20 , 264 entries ) using Mascot with peptide and protein validation performed in Scaffold . Total cellular RNA was isolated and purified using the Qiagen RNeasy Plus Mini Kit ( Qiagen , UK ) according to manufacturer’s protocol followed by reverse transcription using SuperRT ( HT Biotechnology Ltd ) . PCR reactions ( 15 µl ) were prepared using SYBR Green PCR Master Mix ( Applied Biosystems ) with 125 ng starting cDNA template . The reaction proceeded in an ABI 7900 HT Real-Time PCR system ( Applied Biosystems ) and the resultant Ct values were normalized to housekeeping genes ( GAPDH and RPS2 ) . The following primers were used: GAPDH F:ATGGGGAAGGTGAAGGTCG R: CTCCACGACGTACTCAGCG HIF1α F: CCAGTTACGTTCCTTCGATCAGT R: TTTGAGGACTTGCGCTTTCA GLUT1 F: TGGCATGGCGGGTTGT R: CCAGGGTAGCTGCTCCAGC VEGF F: TGCCAAGTGGTCCCAG R: GTGAGGTTTGATCCGC Prolyl hydroxylation of the HIF1αODD protein was performed as described in Burr et al . ( 2016 ) . Briefly , the hydroxylation assay was performed by incubating 10 µM HIF1αODD with 50 µl HeLa cell extract for 15 min at 37°C . The reaction was stopped by addition of SDS loading buffer , and the proteins separated by SDS-PAGE . Hydroxylation was measured using the HIF prolyl hydroxylation specific antibody . Measurements of HIF1α hydroxylation following the addition of Fe ( II ) were performed similarly , except the lysate was pre-incubated with iron chloride for 10 min at 4°C before the addition of the HIF1αODD protein . Data were expressed as mean ± s . e . m . and P values were calculated using two-tailed Student’s t-test for pairwise comparisons , unless otherwise stated . The cytofluorometric colocalisation studies were analysed as described and performed blinded . No statistical method or power analysis was used to predetermine sample size . | Most organisms have developed strategies to survive in low oxygen environments . Central to this response are proteins called Hypoxia Inducible Factors ( HIFs ) , which activate genes involved in energy production and blood vessel growth when oxygen is scarce . When plenty of oxygen is present , HIFs are rapidly broken down . This is important because HIFs have also been linked to the growth and spread of cancers . Oxygen sensing enzymes , termed prolyl hydroxylases , play a principal role in controlling the break down of HIFs when oxygen is abundant . However , the activity of these prolyl hydroxylases can be reduced by changes in the nutrient or iron levels present in the cell . This raises questions about how other cell mechanisms help to control HIF levels . By using a technique called an unbiased forward genetic screen to study human cells , Miles , Burr et al . set out to identify the cellular pathways that regulate HIF levels when oxygen is still abundant . Disrupting a pump called the V-ATPase – which normally helps to break down unwanted proteins by acidifying the cellular compartments where they are destroyed – stabilised HIFs . Moreover , Miles , Burr et al . identified two previously uncharacterised genes that are required for the V-ATPase to work correctly . While the V-ATPase is typically associated with the destruction of proteins , a different , unexpected aspect of its activity is responsible for stabilising HIFs . Blocking activity of the V-ATPase reduces levels of iron inside the cell . This inhibits the activity of the prolyl hydroxylases , resulting in HIFs being activated . Overall , the findings presented by Miles , Burr et al . show key links between oxygen sensing , the use of iron and the V-ATPase . Further work is now needed to investigate how V-ATPase activity affects levels of HIFs found inside cells during diseases such as cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2017 | The vacuolar-ATPase complex and assembly factors, TMEM199 and CCDC115, control HIF1α prolyl hydroxylation by regulating cellular iron levels |
We collated contact tracing data from COVID-19 clusters in Singapore and Tianjin , China and estimated the extent of pre-symptomatic transmission by estimating incubation periods and serial intervals . The mean incubation periods accounting for intermediate cases were 4 . 91 days ( 95%CI 4 . 35 , 5 . 69 ) and 7 . 54 ( 95%CI 6 . 76 , 8 . 56 ) days for Singapore and Tianjin , respectively . The mean serial interval was 4 . 17 ( 95%CI 2 . 44 , 5 . 89 ) and 4 . 31 ( 95%CI 2 . 91 , 5 . 72 ) days ( Singapore , Tianjin ) . The serial intervals are shorter than incubation periods , suggesting that pre-symptomatic transmission may occur in a large proportion of transmission events ( 0 . 4–0 . 5 in Singapore and 0 . 6–0 . 8 in Tianjin , in our analysis with intermediate cases , and more without intermediates ) . Given the evidence for pre-symptomatic transmission , it is vital that even individuals who appear healthy abide by public health measures to control COVID-19 .
The novel coronavirus disease , COVID-19 , was first identified in Wuhan , Hubei Province , China in December 2019 ( Li et al . , 2020b; Huang et al . , 2020 ) . The virus causing the disease was soon named severe acute respiratory syndrome coronavirus 2 ( SARS-CoV-2 ) ( Hui et al . , 2020 ) and quickly spread to other regions of China and then across the globe , causing a pandemic with over 5 million cases and 300 , 000 deaths at the time of writing ( Johns Hopkins University , 2020 ) . In Tianjin , a metropolis located at the north of China , the first case was confirmed on January 21 , 2020 ( Tianjin Health Commission , 2020 ) . Two days later , the first case was confirmed in Singapore ( Ministry of Health Singapore , 2020 ) , a city country in Southeast Asia . As of February 28 , 2020 , 93 and 135 cases had been confirmed in Singapore and Tianjin ( Ministry of Health Singapore , 2020; Tianjin Health Commission , 2020 ) . The first Singapore COVID-19 case was confirmed as an individual who had travelled to Singapore from Wuhan . Many of the initial cases were imported from Wuhan , with later cases being caused by local transmission . Singaporean officials worked to identify potential contacts of confirmed cases; close contacts were monitored and quarantined for 14 days from their last exposure to the patient , and other low-risk contacts were put under active surveillance and contacted daily to monitor their health status . These early outbreaks continue to provide the opportunity to estimate key parameters to understand COVID-19 transmission dynamics . We screened publicly available data to identify datasets for two COVID-19 clusters that could be used to estimate transmission dynamics . In both Singapore and Tianjin , the COVID-19 outbreak occurred within a relatively closed system where immediate public health responses were implemented , contacts were identified and quarantined , and key infection dates were tracked and updated daily . With its experiences in control of the SARS outbreak , the Singaporean government had been adopting a case-by-case control policy from January 2 , 2020 . Only close contacts of a confirmed case were monitored and quarantined for 14 days . In Tianjin , a number of COVID-19 cases were traced to a department store , where numerous customers and sales associates were likely infected . Additional customers who had potential contact were asked to come forward through state news and social media , as well as asked if they had visited the department store at various checkpoints in the city . All individuals identified as having visited the store in late January were quarantined and sections of the Baodi District where the store is located were sealed and put under security patrol . We estimate the serial interval and incubation period of COVID-19 from clusters of cases in Singapore and Tianjin . The serial interval is defined as the length of time between symptom onset in a primary case ( infector ) and symptom onset in a secondary case ( infectee ) , whereas the incubation period is defined as the length of time between an infectee’s exposure to a virus and their symptom onset . Both are important parameters that are widely used in modeling in infectious disease , as they impact model dynamics and hence fits of models to data . While the pandemic has progressed far beyond these early outbreaks , it remains the case that mathematical modelling , using parameters derived from estimates like these , is widely used in forecasting and policy . The serial interval and incubation period distributions , in particular , can be used to identify the extent of pre-symptomatic transmission ( i . e . viral transmission from an individual that occurs prior to symptom onset ) . There is evidence that pre-symptomatic transmission accounts for a considerable portion of COVID-19 spread ( Arons et al . , 2020; Baggett et al . , 2020; Li et al . , 2020a ) and it is important to determine the degree to which this is occurring ( Peak et al . , 2020 ) . Early COVID-19 estimates borrowed parameters from SARS ( Wu et al . , 2020; Jiang et al . , 2020; Abbott et al . , 2020 ) , but more recent estimates have been made using information from early clusters of COVID-19 cases , primarily in Wuhan . Depending on the population used , estimates for incubation periods have ranged from 3 . 6 to 6 . 4 days and serial intervals have ranged from 4 . 0 to 7 . 5 days ( Li et al . , 2020b; Ki and Task Force for 2019-nCoV , 2020; Backer et al . , 2019; Linton et al . , 2020; Nishiura et al . , 2020 ) ; however , it is crucial that the estimates of incubation period and serial interval are based on the same outbreak , and are compared to those obtained from outbreaks in other populations . Distinct outbreak clusters are ideal for understanding how COVID-19 can spread through a population with no prior exposure to the virus . Here , we estimate the portion of transmission that is pre-symptomatic based on estimates of the incubation period and serial interval . We estimate both quantities under two frameworks: first , we use samples as directly as is feasible from the data , for example assuming that the health authorities’ epidemiological inferences regarding who exposed whom and who was exposed at which times are correct . Second , we use estimation methods that allow for unknown intermediate cases , such that the presumed exposure and infection events may not be complete . We also separate the analysis of incubation period according to earlier and later phases of the outbreaks , since measures were introduced during the time frame of the data .
Figures 1 and 2 show the daily counts , putative origin of the exposure and individual time courses for the Singapore and Tianjin data . In the Singapore dataset , new hospitalization and discharge cases were documented daily from January 23 to February 26 , 2020 . 66 . 7% ( 62/93 ) of the confirmed cases recovered and were discharged from the hospital by the end of the study period ( Figure 1 ( a ) ) . The disease progression timeline of the 93 documented cases in Figure 1 ( c ) indicates that symptom onset occurred 1 . 71 ± 3 . 01 ( mean ± SD ) days after the end of possible viral exposure window and cases were confirmed 7 . 43 ± 5 . 28 days after symptom onset . The mean length of hospital stay was 13 . 3 ± 6 . 01 days before individuals recovered and were discharged . In the Tianjin dataset , new confirmed cases were documented daily from January 21 to February 22 , 2020 . 48 . 1% ( 65/135 ) recovered and 2 . 2% ( 3/135 ) had died by the end of the study period ( Figure 2 ( a ) ) . The timeline of the 135 cases is shown in Figure 2 ( c ) . Symptom onset occurred 4 . 98 ± 4 . 83 ( mean ± SD ) days after the end of the possible viral exposure window . Cases were confirmed 5 . 23 ± 4 . 15 days after symptom onset . The duration of hospital stay of the Tianjin cases is unknown as the discharge date of each case was not available . In both datasets , daily counts decline over time , which is likely a combination of delays to symptom onset and between symptom onset and reporting , combined with the effects of strong social distancing and contact tracing . In the Singapore dataset , we find that the median incubation period in our direct analysis ( without accounting for intermediate cases ) is 5 . 32 days with the gamma distribution; shape 3 . 05 ( 95%CI 2 . 0 , 3 . 84 ) ; and scale 1 . 95 ( 1 . 23 , 2 . 34 ) . The mean incubation period is 5 . 99 ( 95%CI 4 . 97 , 7 . 14 ) days . In Tianjin , we find a median 8 . 06 days; shape 4 . 74 ( 3 . 35 , 5 . 72 ) ; scale 1 . 83 ( 1 . 29 , 2 . 04 ) . The mean is 8 . 68 ( 7 . 72 , 9 . 7 ) days . These results are summarised in Table 1 , and we also fitted Weibull and log normal distributions; see Appendix 1—table 1 . These are consistent with , or slightly longer than , previous estimates , see Appendix 1—table 5 for comparison . In Singapore , these estimates are based on a combination of cases for whom last possible exposure is given by travel , and later cases ( for whom the presumed infector was used ) . In Tianjin , social distancing measures were implemented during the outbreak . We find that the estimated incubation period is different , particularly in Tianjin , for cases with symptom onset on or prior to January 31st: see Figure 3 and Figure 4 . The estimated median incubation period for pre-Feb one cases in Tianjin is 6 . 48 days; the q= ( 0 . 025 , 0 . 975 ) quantiles are ( 2 . 5 , 13 . 3 ) days . In contrast , post-Jan 31 the median is 12 . 13 days with q= ( 0 . 025 , 0 . 975 ) quantiles ( 7 . 3 , 18 . 7 ) days . The means are 6 . 88 ( 5 . 97 , 7 . 87 ) days for early cases and 12 . 4 ( 11 . 1 , 13 . 7 ) days for later cases . Social distancing seems unlikely to change the natural course of infection , but these results might be explained if exposure occurred during group quarantine or otherwise later than the last time individuals thought they could have been exposed . Pre-symptomatic transmission would enable this , if an individual was thought to have been exposed before group quarantine , but in actuality was exposed during quarantine by a pre-symptomatic individual . The time interval in the data would then not be a sample of the incubation period , instead it would be a sample of one or more generation times plus an incubation period . In Singapore , we find the same effect , although much less pronounced . The estimated median incubation time is 5 . 26 , with ( 0 . 025 , 0 . 975 ) quantiles of ( 1 . 30 , 13 . 8 ) days for early cases ( also defined as cases with symptom onset on or prior to January 31st ) and 5 . 35 ( quantiles ( 1 . 22 , 14 . 6 ) ) days for late-arising cases . The means are 5 . 91 ( 4 . 50 , 7 . 64 ) days for early cases and 6 . 06 ( 4 . 70 , 7 . 67 ) days for later cases . Fits of gamma and log-normal distributions are similar; see Appendix 1—table 2 . Changes in perception of exposure times after control measures were introduced ( i . e . people may assume that they must have been exposed prior to control measures ) , together with pre-symptomatic transmission , could result in missing intermediate transmission events and hence lengthened incubation period estimates . This in part motivates our analysis with intermediate cases . Our estimates of the incubation period with intermediates are similar , under the assumption that intermediates are relatively rare . Results are shown in Figure 5 and Table 1 . We find that the median of the bootstrapped mean incubation periods for Singapore with a low ( 0 . 05 per day ) rate of unknown intermediates is 4 . 91 days ( 4 . 35 , 5 . 69 95% bootstrap CI ) , compared to a generation time of 3 . 71 ( 2 . 36 , 4 . 91 ) days . The Tianjin bootstrapped mean incubation period is 7 . 54 ( 6 . 76 , 8 . 56 95% CI ) days and the generation time is only 2 . 82 ( 1 . 83 , 3 . 52 ) days . The estimates are lower when the assumed probability of unknown intermediates is higher . Indeed , if intermediates were present between assumed exposure and onset , naturally the generation time would be shorter than if they were not . The mean generation times are consistently shorter than the mean incubation periods , indicating that infection can occur prior to symptom onset . The difference is particularly pronounced in Tianjin , where long intervals were observed . However , this approach makes a number of assumptions and is limited by the fact that if we do not know the true infectors then we are also unlikely to know the true exposure . The data we have is well suited to this method in the sense that there were particular events where exposure is thought to have occurred , and so we can account for intermediates in the manner we have done , but we do not have information for the alternative scenario in which the true exposures were prior to those given in the data . This could happen if , for example , individuals were exposed before attending an event or before known contact , and developed symptoms well after it . Exposure would thus be wrongly attributed to the event or contact . We have accommodated this with uncertainty in the exposure intervals , in particular not insisting that individuals who are likely to be the index case for a cluster ( e . g . who developed symptoms on the same day as an event ) must have been exposed then , but instead allowing the possibility that they were exposed earlier . Figure 6 represents the empirical serial intervals between all potential transmission case-pairs as noted in the data and represented in Figure 7 , split into groups based on date of first symptom onset for each case-pair . The empirical mean serial intervals shorten in the ’late’ group in both Singapore and Tianjin; however , the empirically derived 95% confidence intervals overlap ( Singapore early 4 . 44 ( -2 . 81 , 11 . 7 ) vs . late 3 . 18 ( -1 . 52 , 7 . 88 ) ; Tianjin early 5 . 48 ( -0 . 968 , 11 . 9 ) vs . late 4 . 18 ( -2 . 33 , 10 . 7 ) ) . Negative lower bounds are due to the high standard deviation . Shortening serial intervals are expected as increased quarantine measures are enacted during the course of an outbreak and can be an indication of improved control through successful contact tracing , as seen in SARS ( Lipsitch et al . , 2003 ) . Our results suggest that serial intervals shortened as the outbreak progressed in both clusters , but they could also be due to right truncation . Accounting for this , we found that the mean serial intervals were 4 and 5 days ( Singapore , Tianjin ) ; a Cox regression found no significant difference between the early and late groups’ serial intervals . This estimate is made directly from case pairs in the data without accounting for intermediate infectors and co-primary infection , as in the ICC analysis . Table 1 shows our ICC estimates of the mean and standard deviation for the serial intervals , with comparison to other analyses and assumptions in Appendix 1—table 5 . The ICC method finds the mean serial interval to be 4 . 17 ( 2 . 44 , 5 . 89 95% bootstrap CI ) days ( 0 . 882 bootstrap standard deviation ) for Singapore and 4 . 31 ( 2 . 91 , 5 . 72 ) days ( 0 . 716 bootstrap sd ) for Tianjin , using the first four cases in each cluster . This is consistent with the results with right truncation . We estimated incubation periods and serial intervals with and without accounting for intermediate unknown cases . To estimate the portion of transmission that occurs before symptom onset , we compare the ‘direct’ ( no intermediate ) estimates of each , and the ‘indirect’ ( accounting for intermediates ) estimates of each . We estimate consistently shorter serial intervals than incubation period , suggesting that there is pre-symptomatic transmission . We took the covariation of incubation periods and serial intervals ( and of generation times and incubation periods ) into account by sampling the intervals jointly before estimating the fraction of the relevant differences that are negative . Even accounting for correlation , the estimated fraction of pre-symptomatic transmission for Singapore is 0 . 74 ( regardless of early/late split ) and for Tianjin is 0 . 72 , 0 . 96 , 0 . 81 ( early , late , all ) , based on the direct estimates of the incubation periods and serial intervals ( see also Figure 8 ) . When we use the incubation period estimates that account for intermediates , the portions pre-symptomatic transmission are 0 . 53 in Singapore and 0 . 79 in Tianjin , when the assumed ‘rate of appearance’ of intermediates r is 0 . 05 ( i . e . when we assume a relatively low rate of unknown intermediates ) . If this rate r is increased , the portion of pre-symptomatic transmission decreases , but even for r=0 . 2 we estimate the pre-symptomatic portion to be 0 . 38 in Singapore and 0 . 64 in Tianjin . These results were obtained under an estimated correlation between the incubation period and serial interval of 0 . 289 in Tianjin . If instead the correlation were 0 . 1 , the portion of pre-symptomatic transmission in Tianjin under r = 0 . 05 , 0 . 1 , 0 . 15 and 0 . 2 , respectively , is estimated as 0 . 783 , 0 . 725 , 0 . 663 and 0 . 62 . With correlation 0 . 8 , the equivalent portions are 0 . 849 , 0 . 781 , 0 . 704 and 0 . 660 . We therefore find that the degree of positive correlation does not greatly impact our estimates of pre-symptomatic transmission . We retain high estimates of the fraction pre-symptomatic in Tianjin , due to the long apparent incubation periods . It seems likely that these are an artifact of either pre-symptomatic transmission during quarantine/lockdown , or of other assumptions made about exposures in the creation of the original dataset . We conclude that overall for this data and under reasonable assumptions , we see evidence of at least 65% of transmission occurring before symptom onset . In our direct analysis , we estimate that infection occurred on average 1 . 99 and 3 . 68 days before symptom onset of the infector ( Singapore , Tianjin ) . Because the incubation period is different for early- and late-occurring cases in our data , on average transmission for early-occurring cases is 1 . 91 and 2 . 06 days before symptom onset ( Singapore , Tianjin ) and 1 . 88 , 7 . 4 days before ( Singapore , Tianjin ) for late-occurring cases . Taking a low rate ( r=0 . 05 ) of potential unknown intermediate cases into account , the mean difference reduces to 0 . 77 and 3 . 23 days ( Singapore , Tianjin ) , though we still estimate a significant portion of pre-symptomatic transmission ( 0 . 53 , 0 . 79 ) , as above . Overall , serial intervals are robustly shorter than incubation periods in our analyses ( Table 1 ) . These estimates are strengthened by the fact that we have estimated both incubation period and serial interval in the same populations and by the fact that we obtain the same result in two distinct datasets . In both sets of estimates , samples of the incubation period minus serial interval are negative with probability 0 . 38 or higher ( Singapore ) and 0 . 64 or higher ( Tianjin ) , and these lower bounds require a high rate of unknown intermediates early in the outbreak . This indicates that a substantial portion of transmission may occur before symptom onset ( see Appendix 1 and Figure 8 ) , consistent with the clinical observations reported by Rothe et al . , 2020 and ( Bai et al . , 2020 ) . Shorter serial intervals yield lower reproduction number estimates . For example , if the epidemic grows at a rate of 0 . 15 ( doubling time of 4 . 6 days [Jung et al . , 2020] , scenario 1 ) , an estimated reproduction number using the mean of the bootstrapped estimates is R=1 . 76 ( 1 . 30 , 2 . 17 ) with a serial interval of 4 . 17 days ( Singapore ) and R=1 . 95 ( 1 . 72 , 2 . 47 ) with a serial interval of 4 . 3 days ( Tianjin ) . In contrast , if a longer serial interval ( 7 . 5 days [Jung et al . , 2020; Li et al . , 2020b] ) is used , the estimate is R=3 . 05 . This is based on the relationship between R0 , serial interval , and growth rate , and is a simple estimate that does not take into account a complex and variable natural history of infection ( Wallinga and Lipsitch , 2007 ) . It serves primarily to illustrate how our estimated serial intervals impact R in simple models for COVID-19 dynamics .
Here , we use transmission clusters in two locations where cases have reported links , exposure and symptom onset times to estimate both the incubation period and serial interval of COVID-19 . We make these datasets available in a convenient spreadsheet form; they were available publicly but the Singapore dataset was presented in free text updates and the Tianjin cluster was described on multiple sites and in graphical form , in Chinese . We anticipate that the datasets themselves will remain useful for understanding COVID-19’s early spread in these well-documented outbreaks . The incubation period and serial interval are key parameters for transmission modeling and for informing public health interventions; modeling remains one of the primary policy aids in use in planning local and global COVID-19 responses . Serial intervals , together with R0 , control the shape and distribution of the epidemic curve ( Anderson et al . , 2004 ) . They influence the disease’s incidence and prevalence , how quickly an epidemic grows , and how quickly intervention methods need to be implemented by public health officials to control the disease ( Anderson et al . , 2004; Fraser et al . , 2004 ) . In particular , the portion of transmission events that occur before symptom onset is a central quantity for infection control ( Fraser et al . , 2004 ) , and will impact the efficacy of contact tracing and case finding efforts ( Peak et al . , 2020 ) . Singapore and Tianjin officials both reacted quickly when COVID-19 cases appeared and started implementing contact tracing and containment measures; however , there was a dramatic difference in the severity of the measures taken . The first case was identified in Singapore on Jan 23 , 2020 and in Tianjin on Jan 21 . By Feb 9 , Singapore had identified 989 close contacts and implemented a travel advisory to defer all travel to Hubei Province and all non-essential travel to Mainland China , asked travellers to monitor their health closely for 2 weeks upon return to Singapore , and asked the public to adopt precautions including avoiding close contact with people who are unwell , practicing good hygiene and hand washing , and wearing a mask if they had respiratory symptoms ( Ministry of Health Singapore , 2020 ) . Comparatively , by February 9 in Tianjin , 11 , 700 contacts were under observations and the Baodi district of almost 1 million people was placed under lockdown with restrictions including: one person per household could leave every 2 days to purchase basic needs , public gatherings were banned , no one could leave their homes between 10PM and 6AM without an exemption , entrances to Tianjin were put under control , and all the buses linking nearby provinces and cities were halted ( www . chinadaily . com ) . While Singapore contained the virus spread relatively well until mid-March , they reached 500 confirmed cases on March 23 , 1000 cases on April 1 , 10 , 000 cases on April 22 , and 25 , 000 cases on May 13 ( Ministry of Health Singapore , 2020 ) ; Tianjin province began to flatten their epidemic curve by mid-to-late-February and had plateaued at 192 confirmed cases as of May 19 ( github . com/CSSEGISandData/COVID-19 ) . In Singapore and Tianjin we estimated relatively short serial intervals . Of particular note , early estimates of R0 for COVID-19 used the SARS serial interval of 8 . 4 days ( Abbott et al . , 2020; Majumder and Mandl , 2020; Wu et al . , 2020 ) . Our serial interval findings from two populations mirror those of Zhao et al . , 2020 and ( Nishiura et al . , 2020 ) , who estimated a serial interval of 4 . 4 and 4 . 0 days . Du et al . , 2020 obtain a similar estimate for the serial interval ( 3 . 96 days with 95% CI: 3 . 53–4 . 39 ) but with standard deviation 4 . 75 days , based on 468 cases in 18 provinces . Furthermore , we estimate the serial interval to be shorter than the incubation period in both clusters , which suggests pre-symptomatic transmission . This indicates that spread of SARS-CoV-2 is likely to be difficult to stop by isolation of detected cases alone . However , shorter serial intervals also lead to lower estimates of R0 , and our serial intervals support R0 values just below 2; if correct this means that half of the transmissions need to be prevented to contain outbreaks . We stratified the incubation period analysis for Tianjin by time of symptom onset ( pre- or post- Jan 31 , 2020; motivated by quarantine/social distancing measures ) and found that the apparent incubation period was longer for those with post-quarantine symptom onset . The reason for this is unclear , but one possible explanation is that there were ( unknown , therefore unreported ) exposures during the quarantine period . If people are quarantined in groups of ( presumed ) uninfected cases , pre-symptomatic transmission in quarantine would result in true exposure times that are more recent than reported last possible exposure times . Although it may seem contradictory that for example Singapore’s efforts were able to keep the epidemic under control using mainly case-based controls if pre-symptomatic transmission is common , it remains the case that detailed contact tracing combined with case finding may be key to limiting both symptomatic and pre-symptomatic spread . In Singapore , symptom-free close contacts of known cases were quarantined preemptively for 14 days , and other less high-risk contacts were placed under phone surveillance ( Lee et al . , 2020 ) . In addition , if case finding is able to prevent a large portion of symptomatic transmission , it seems logical that the remaining observed transmission may be pre-symptomatic . The large extent of pre-symptomatic spread that is occurring , however , may be one reason that the spread of COVID-19 in Singapore was ultimately only delayed and not prevented . There are several limitations to this work . First , the times of exposure and the presumed infectors are uncertain , and the incubation period is variable . We have not incorporated uncertainty in the dates of symptom onset . We have used the mixture model approach for serial intervals to avoid assuming that the presumed infector is always the true infector , but the mixture does not capture all possible transmission configurations . Our R0 estimates are simple , based on a doubling time of 4 . 6 days , and could be refined with more sophisticated modeling in combination with case count data . We have not adjusted for truncation ( e . g . shorter serial intervals are likely to be observed first ) or the growth curve of the epidemic . However , the serial interval estimates are consistent between the two datasets , are robust to the parameter choices , and are consistently shorter than the estimated incubation times . We estimated both the incubation period and the serial interval in Singapore and Tianjin COVID-19 clusters . Our results suggest that there is substantial transmission prior to onset of symptoms , as the serial interval is shorter than incubation period by 2–4 days . We find differences in estimated incubation period between early and later cases; this may be due to pre-symptomatic transmission or differences in reporting and/or in perceived exposure as the outbreak progressed , in the context of social distancing measures . Evidence of transmission from apparently healthy individuals makes broad-scale social distancing measures particularly important in controlling the spread of the disease .
All datasets and R code are available on GitHub ( github . com/carolinecolijn/ClustersCOVID19; Tindale et al . , 2020; copy archived at https://github . com/elifesciences-publications/ClustersCOVID19 ) . Singapore data was obtained from the Ministry of Health Singapore , 2020 online press releases . The Singapore dataset comprised 93 confirmed cases from the date of the initial case on January 23 , 2020 until February 26 , 2020 . Tianjin data was obtained from the Tianjin Health Commission , 2020 online press releases . The Tianjin dataset comprises 135 cases confirmed from January 21 to February 22 , 2020 . The symptom onsets were available on the official website for all but a few patients who had not had symptoms before being diagnosed at a quarantine center . Both datasets contained mainly information on exposure times , contacts among cases , time of symptom onset ( See Appendix 1 for column descriptions and data processing ) . All statistical analyses were performed using R ( R Development Core Team , 2013 ) . | The first cases of COVID-19 were identified in Wuhan , a city in Central China , in December 2019 . The virus quickly spread within the country and then across the globe . By the third week in January , the first cases were confirmed in Tianjin , a city in Northern China , and in Singapore , a city country in Southeast Asia . By late February , Tianjin had 135 cases and Singapore had 93 cases . In both cities , public health officials immediately began identifying and quarantining the contacts of infected people . The information collected in Tianjin and Singapore about COVID-19 is very useful for scientists . It makes it possible to determine the disease’s incubation period , which is how long it takes to develop symptoms after virus exposure . It can also show how many days pass between an infected person developing symptoms and a person they infect developing symptoms . This period is called the serial interval . Scientists use this information to determine whether individuals infect others before showing symptoms themselves and how often this occurs . Using data from Tianjin and Singapore , Tindale , Stockdale et al . now estimate the incubation period for COVID-19 is between five and eight days and the serial interval is about four days . About 40% to 80% of the novel coronavirus transmission occurs two to four days before an infected person has symptoms . This transmission from apparently healthy individuals means that staying home when symptomatic is not enough to control the spread of COVID-19 . Instead , broad-scale social distancing measures are necessary . Understanding how COVID-19 spreads can help public health officials determine how to best contain the virus and stop the outbreak . The new data suggest that public health measures aimed at preventing asymptomatic transmission are essential . This means that even people who appear healthy need to comply with preventive measures like mask use and social distancing . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2020 | Evidence for transmission of COVID-19 prior to symptom onset |
Life history theory posits that the sequence and timing of events in an organism's lifespan are fine-tuned by evolution to maximize the production of viable offspring . In a virus , a life history strategy is largely manifested in its replication mode . Here , we develop a stochastic mathematical model to infer the replication mode shaping the structure and mutation distribution of a poliovirus population in an intact single infected cell . We measure production of RNA and poliovirus particles through the infection cycle , and use these data to infer the parameters of our model . We find that on average the viral progeny produced from each cell are approximately five generations removed from the infecting virus . Multiple generations within a single cell infection provide opportunities for significant accumulation of mutations per viral genome and for intracellular selection .
RNA viruses are excellent models for evolution . They replicate quickly and have extremely high mutation rates , orders of magnitude greater than those of most DNA-based life forms ( Drake , 1993 ) . While this combination of traits creates the potential for rapid adaptation , it necessitates a life history strategy that balances the need for explosive , exponential growth with the requirement to maintain genomic integrity . The life history strategies of viruses are largely reflected by their mode of intracellular replication . Two classic replication modes have been described for single-stranded RNA viruses: the ‘stamping machine’ mode ( Stent , 1963 ) and the ‘geometric replication’ mode ( Luria , 1951 ) . In the stamping machine mode ( SM ) , templates made from the original infecting genomes are used for the production of all progeny genomes . In the geometric replication mode ( GR ) , newly made progeny genomes are used to create further templates for additional rounds of replication within a single cellular infection cycle ( Figure 1 ) . Progeny produced from stamping machine replication are all a single generation away from the parental strand whereas progeny generated from geometric growth represent a distribution of generations from the parental strand , often resulting in a fractional mean number of generations ( see Figure 1 ) . The iterative nature of GR creates branched genealogies that allow for expansive exploration of sequence space and results in a mutation distribution that differs from the SM mode ( Luria , 1951 ) . Recent studies with population-genetic models ( Draghi et al . , 2010 ) and RNA enzyme populations ( Hayden et al . , 2011 ) have shown that differences in the distribution of mutants can significantly impact the adaptability of a population . Recent studies with poliovirus ( PV ) have also demonstrated that mutational differences within a population can have dramatic effects on pathogenicity ( Pfeiffer and Kirkegaard , 2005; Vignuzzi et al . , 2006 ) as well as fitness , virulence , and robustness ( Lauring et al . , 2012 ) . 10 . 7554/eLife . 03753 . 003Figure 1 . Illustrations of the genealogies of different replication modes . Red dots indicate positive-sense strands . Blue dots indicate negative-sense templates . Stamping machine ( SM ) progeny are one generation from the initial infecting genome ( left ) . In an example of geometric replication ( GR ) , progeny are an average of 2 . 33 generations from the initial infecting genome ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 003 Poliovirus' simple genomic architecture and medical importance have made it one of the most extensively studied viruses ( Racaniello , 2006 ) . However , despite decades of mechanistic studies and recent revelations of the importance of population structures , the replication mode and resulting mutation distribution have yet to be determined . PV therefore proves an excellent candidate for the rigorous construction of a computational model of virus replication to predict population structure and mutation distribution . A major feature of PV intracellular dynamics is that the genome participates in multiple reactions: translation , replication , and encapsidation . Its 7 . 5 kb genome contains a single open reading frame , which encodes 7 nonstructural proteins and 4 capsid proteins . Translation produces a single polyprotein , which is cleaved into individual functional viral proteins . Replication of the positive-sense genome by the virus-encoded RNA-dependent RNA polymerase produces a negative-sense strand , which is used as a template for further genome synthesis . Evidence suggests that the initial , infecting positive-sense genomes must be translated before they can replicate ( Novak and Kirkegaard , 1994 ) . The switch from translation to replication appears to be dependent on the concentration of a viral protein product , 3CD , which stimulates a transition from a linear , translating RNA to a noncovalently associated ‘circular’ RNA competent for replication ( Gamarnik and Andino , 1998 , 2000; Herold and Andino , 2001 ) . Encapsidation is thought to result from protein–protein associations of capsid pentamers with the RNA replication machinery and protein–RNA association of capsid pentamers with viral RNA ( Pfister et al . , 1992; Nugent and Kirkegaard , 1995; Liu et al . , 2010 ) . Actively replicating genomes are preferentially encapsidated and packaging is biased to exclude negative-sense strands , although the mechanism of this is not understood ( Nugent et al . , 1999 ) . Although multiple ribosomes can translate a genome at the same time and multiple viral polymerases can replicate a genome at the same time , the processes are mutually exclusive ( Gamarnik and Andino , 1998 ) . Similarly , neither translation nor replication can occur after a genome is packaged into a virion . Several studies have demonstrated that PV genomes are often localized to the cytosolic surfaces of the endoplasmic reticulum , Golgi bodies , lysosomes , or vesicles derived from these ( Schlegel et al . , 1996; Bolten et al . , 1998; Cui et al . , 2005; Egger and Bienz , 2005; den Boon et al . , 2010 ) . Replication complexes are thought to form on these membranes in cis , resulting in a close association of translation products and positive-sense genomes ( Novak and Kirkegaard , 1994; Egger et al . , 2000 ) . Compartmentalization of replication complexes likely accounts for the observation that many functions of nonstructural proteins cannot be complemented in trans ( Novak and Kirkegaard , 1994; Ansardi et al . , 1996 ) . Only capsid proteins , 3CD , and 3D have been demonstrated to trans-complement ( Novak and Kirkegaard , 1994; Nugent et al . , 1999; Oh et al . , 2009 ) . Taken together , these studies suggest that the essential transitions—from translation to replication , and from replication to encapsidation—are largely localized and influenced by the dynamics of the molecules in each compartment . In recent years , modeling approaches have begun to examine the trade-offs that come with having a genome that is a template for both replication and translation ( Krakauer and Komarova , 2003; Regoes et al . , 2005; Sardanyés et al . , 2009; Thébaud et al . , 2010; Martinez et al . , 2011 ) . These studies have raised mechanistic and evolutionary questions about the life cycle of single-stranded , positive-sense RNA viruses , but most have not produced models that can be directly compared to data . Several previous models are deterministic in nature ( Krakauer and Komarova , 2003; Regoes et al . , 2005; Martinez et al . , 2011 ) and assume a well-mixed , spatially uniform cellular environment ( Krakauer and Komarova , 2003; Regoes et al . , 2005; Sardanyés et al . , 2009; Thébaud et al . , 2010; Martinez et al . , 2011 ) . Experimental evidence suggests that each of these assumptions is problematic and do not reflect the biological constraints and properties of viral replication . The small numbers of the critical molecules that initiate an infection suggest that a stochastic model would more accurately describe early reactions and could make distinct predictions from previous deterministic approaches ( Srivastavawz et al . , 2002 ) . Often infections begin with relatively few virions that release their genomes into the cell and continue with the translation of these few initial genomes . Random variation in the switch from translation to replication is amplified by the subsequent exponential phase of the infection , and this amplification is likely to bias the mean dynamics of a set of infections . Indeed , recent single-cell studies demonstrated the significant impact of stochastic effects on poliovirus infections ( Schulte and Andino , 2014 ) . Here , we have developed a stochastic simulation model in which we compartmentalize reactions in an effort to accurately describe intracellular dynamics in both space and time . Additionally , rather than fixing each parameter on an estimated value , an approach used by previous models , we use an Approximate Bayesian Computation approach to fit our parameters from temporal quantitative data . We find that by combining stochasticity and spatial structure , our model reflects and describes the population dynamics and structure of the viral population during an infection cycle more accurately than previous models . Fitting our model to RNA abundances over time , we find that poliovirus follows the geometric replication mode: multiple iterative generations of genomic replication produce progeny virus . Posterior parameter fits indicate that progeny of a single cellular infection are approximately five generations away from the initial , infecting genomes . This replication mode produces populations with expansive , branched genealogies , creating the dramatic potential for the exploration of sequence space , as well as creating the potential for intracellular selection among related mutant genomes .
We used temporal , quantitative RT-PCR data of both positive-sense genomes and negative-sense strands to estimate the free parameters in our model . The role of each parameter in poliovirus replication and in the mathematics of our model are diagrammed in Figure 2 and described in detail in the ‘Materials and methods’ . We chose to use measurements of positive- and negative-sense RNA at multiple time points for three multiplicities of infection ( 1 , 10 , and 100 ) , as well as measurements of virion numbers at multiple times for MOI 10; this amounted to 27 measured means , with three data points for each mean . Strand-specific qRT-PCR was performed to quantify positive-sense and negative-sense poliovirus RNA against in vitro transcribed standard RNAs of each sense ( Burrill et al . , 2013 ) . Along with cell counts , this allowed for temporal measurements of the average positive-sense and negative-sense poliovirus RNA copies per cell . Negative-sense RNA was not detectable until 2 hr post infection for MOIs 10 and 100 and 3 hr post infection for MOI 1 . Positive-sense RNA was clearly quantifiable for all time points at the MOI 10 and 100 but did not rise above background levels until 3 hr post infection for MOI 1 . Using a newly developed virion immunoprecipitation assay ( Burrill et al . , 2013 ) , we observed de novo virion assembly between 2 hr and 3 hr post infection . Along with total positive-sense RNA measurements from this time course , we obtained a percentage of genomes encapsidated in quadruplicate at 3 hr , 4 hr , and 5 hr post infection . Figure 3 illustrates this data alongside projections from inferred parameters from the second round of SMC ( see Figure 3—source data 1 ) . 10 . 7554/eLife . 03753 . 004Figure 2 . The replication cycle of poliovirus as represented in our model . Numbered steps correspond to sections and equations in the ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 00410 . 7554/eLife . 03753 . 005Figure 3 . Projected mean abundances of positive-sense RNA ( solid line simulations vs filled circle experimental measurements ) , negative-sense RNA ( dashed line simulations vs hollow circle experimental measurements ) and virions ( orange dotted line simulations vs star experimental measurements; measured only for MOI = 10 ) . Each row represents a different example parameter set ( see ‘Results’ ) ; each line is the mean of 20 individual cell simulations , and the means of five sets of 20 replicate simulations are plotted in each panel . Parameter values are given in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 00510 . 7554/eLife . 03753 . 006Figure 3—source data 1 . ‘Best’ parameter set used in Figure 3—figure supplement 1 , and Figure 4—figure supplement 4 . Representative parameter sets ( sets 1–5 ) used in Figure 3 . Note that natural log values are provided for all parameters except cstay . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 00610 . 7554/eLife . 03753 . 007Figure 3—figure supplement 1 . Distribution of virions in 10 , 000 replicates for simulations with ( points ) and without ( line ) a deterministic threshold for waiting times ( see ‘Materials and methods’ ) . Parameter values are the ‘best’ set given in Figure 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 007 The relatively high number of data dimensions , combined with the computationally intensive and highly stochastic nature of our simulations , made a traditional maximum likelihood approach impractical . Instead , we turned to Approximate Bayesian Computation , using as our summary statistic the sum of the squared deviations of the average simulated RNA concentrations ( and average fraction of virions for MOI 10 ) from their corresponding empirical means . This algorithm produces progressively more accurate estimates of each parameter over several rounds; Figure 4—figure supplement 1 illustrates that , for most parameters , round one restricts the credible range of each parameter in comparison to the flat prior and round two leads to further focusing . The data appear to be uninformative for at least one parameter , cpack; a second parameter , commax , appears to be poorly constrained by the comparison to MOI = 10 in round one , but somewhat constrained by the broader measurement against all three MOIs in round two . Round two also appears to significantly move the mode of two other parameters , ccom and c3A . Figure 4—figure supplement 1 indicates that ABC inference informed the values of nine of our ten parameters , but these marginal parameter distributions alone do not capture correlations between parameter values . Figure 4—figure supplement 2 shows evidence of significant correlations , and Figure 4—figure supplement 3 shows that parameter sets drawn from the marginal distributions in Figure 4—figure supplement 1 ( i . e . , uncorrelated parameter values ) do a poor job of matching the data . While not unexpected , these significant correlations require that we work directly with the sampled parameter sets arising from our inference process , which is the approach we take below . Each parameter in the posterior is supported over a significant range of possibilities . This remaining uncertainty reflects two factors: the data may be insufficient to determine each parameter , and the inference process may not have fully exploited the inferential power of the data . We took several approaches to quantify the sufficiency of the data and the effectiveness of the inference process . First , we measured the mean error of parameter sets when compared to the data for each multiplicity of infection independently; we asked if performance at one MOI predicted performance at the other two . If so , the dimensionality of our data would be effectively lower than we had initially assumed . Surprisingly , pairwise correlations between mean error at one MOI and another were very weak: Spearman's rho is 0 . 031 for MOIs 1 and 10 , −0 . 092 for MOIs 1 and 100 , and 0 . 129 for MOIs 10 and 100 . This suggests that measurements at each MOI are contributing distinct information to our inference process . Second , we determined the sensitivity of our measure of model ‘fit’ to variation in each of the parameters . This analysis , described in detail in the ‘Materials and methods’ , showed that the data significantly informed the values of eight of ten of the parameters ( Figure 4—figure supplement 4 ) . We also performed a separate validation analysis which attempted to infer the replication phenotype , g¯ , from mock data simulated from parameter sets drawn from our prior distribution . As described more fully in the ‘Materials and methods’ , this exercise confirms that the data and method are adequate to infer the trait of interest , albeit with some degree of inaccuracy ( Figure 4—figure supplement 5 ) . Finally , we examine the fit between the data and the mean dynamics of inferred parameter sets . Figure 3 shows that the inferred parameter sets generally capture the information in the RNA and virion data , although some parameter sets deviate consistently from the data for some values . Variability among replicate sets of twenty single-cell simulations is substantial , correlated across a time series , and greatest for the smallest MOI . Further inferential effort could improve either the accuracy of the mean predicted dynamics or the precision of replicate simulation dynamics , though Figure 3 suggests that such improvements could only be modest . This variability is expected due to the stochastic nature of the simulations , and it may better reflect the biological noise of the infection ( Schulte and Andino , 2014 ) . Figure 4 shows the inferred posterior distribution of g¯ , the mean number of generations for a packaged virion based on two rounds of inference with measured RNA and virion abundances . This distribution is plotted for MOI = 10; the predicted values at MOI = 1 and MOI = 100 are very similar and highly correlated ( weighted means: MOI = 1 , 4 . 96; MOI = 10 , 5 . 06; MOI = 100 , 4 . 85; Spearman's rho ( unweighted ) : MOI 1 and 10 , 0 . 92; MOI 1 and 100 , 0 . 85; MOI 10 and 100 , 0 . 96 ) . While this distribution does show substantial variance , it is strongly inconsistent with a ‘stamping machine’ mode of replication , which would have a g¯ near one . 10 . 7554/eLife . 03753 . 008Figure 4 . Left: posterior distribution of the mean number of generations of replication ( g¯ ) . Right: distribution reweighted by the fit of predicted fractions of translating positive-sense RNA to empirical measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 00810 . 7554/eLife . 03753 . 009Figure 4—figure supplement 1 . Prior and posterior distributions after each of two rounds of inference by Approximate Bayesian Computation . Empirical posteriors for rounds one and two are based on 1000 and 1135 points , respectively , grouped into 15 bins . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 00910 . 7554/eLife . 03753 . 010Figure 4—figure supplement 2 . Correlations between parameters in the round two posterior . Spearman rank correlations with magnitudes below 0 . 1 were ignored; those with magnitudes between 0 . 1 and 0 . 25 are noted as ‘+’ or ‘−’ , those with magnitudes between 0 . 25 and 0 . 5 are noted as ‘++’ or ‘−−’ , and those with magnitudes above 0 . 5 are noted as ‘+++’ or ‘−−−’ . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 01010 . 7554/eLife . 03753 . 011Figure 4—figure supplement 3 . Log of total error for inferred , weighted parameter sets in round two ( solid ) vs 1000 sets assembled from parameter values drawn independently from the weighted posterior ( dotted ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 01110 . 7554/eLife . 03753 . 012Figure 4—figure supplement 4 . Goodness-of-fit ( 1/[1 + mean error] ) of highly replicated simulations for MOI = 10 and the ‘best’ inferred parameter set . Each parameter value was varied independently for 1000 sets of 20 single-cell replicates for each MOI . Orange lines represent the base value of each parameter . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 01210 . 7554/eLife . 03753 . 013Figure 4—figure supplement 5 . Inference results from three validation experiments . ( A ) Three parameter sets , spanning a range of values of g¯ , were selected from the prior distribution . The parameter sets produced substantial differences in mean levels of positive sense ( solid ) and negative sense ( dashed ) RNA over time . ( B ) Distribution of g¯in the prior , round one posterior , and round two posterior for the ‘low’ parameter set ( g¯ = 3 . 20 ) . ( C ) Distribution of g¯in the prior , round one posterior , and round two posterior for the ‘middle’ parameter set ( g¯ = 5 . 97 ) . ( D ) Distribution of g¯in the prior and round one posterior ( round two was not performed due to the high degree of convergence in round one ) for the ‘high’ parameter set ( g¯ = 8 . 58 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 01310 . 7554/eLife . 03753 . 014Figure 4—figure supplement 6 . Histograms of the projected fraction of positive-sense RNA undergoing translation for the mean simulated dynamics of each parameter set , compared to empirical measurements ( orange dots ) . To better visualize the diversity of the predictions made from the inferred parameter sets , these histograms are not corrected by the importance weights . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 01410 . 7554/eLife . 03753 . 015Figure 4—figure supplement 7 . Summed squared error ( SSE ) of fraction of translating positive-sense RNA for all 1135 parameter sets , plotted against g¯ at MOI = 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 015 To explore the robustness of this inference , we compared the predicted dynamics of the model to an additional type of data: the fraction of positive-sense RNA molecules translating at each time point . We fractionated infected cell lysates and quantified positive-sense RNAs in monosome and polysome fractions relative to total positive-sense RNA copies . These data render a percentage of genomes associated with translation machinery and provide an additional set of data to evaluate the parameter sets produced by SMC . When measured at an MOI of 10 at 1 , 2 , 3 , 4 , and 5 hr post infection , the majority of positive-sense RNAs appeared to be associating with translation machinery , consistently averaging near 85% . Many of the inferred parameter sets are consistent with the measured values but a substantial fraction is clearly inconsistent ( Figure 4—figure supplement 6 ) . The summed squared error of the translating fractions is also correlated with g¯ ( Figure 4—figure supplement 7 ) . To estimate how these new data inform our prediction of g¯ , we calculated a weighting factor based on the relative rank of the summed squared error of translating fractions , such that the parameter set with the best fit was assigned a weight of 1 , the next a weight of 1134/1135 , etc . Reweighting the distribution of g¯ by this additional factor produced the distribution shown in Figure 4; the mean g¯ shifts from 5 . 06 to 4 . 78 . We simulated mutation and selection during infections to understand how replication dynamics shape the distribution of mutation frequencies among virions . To illustrate how mutant frequencies depended on g¯ , we chose two parameter sets with values of g¯ at the low and high end of the range supported by the posteriors in Figure 4 and included the ‘best’ parameter set as a representative of the more common values of g¯ . Mutation frequencies for these parameter sets ( ‘best’ , ‘low’ , and ‘high’—see Figure 3—source data 1 ) are plotted in Figure 5A for a range of mutants that have a diminished rate of replication relative to the wild type . We chose to model this particular type of deficiency because we expected that replication deficits would directly affect the growth and packaging of the mutants . We observed that deficits in a different trait , the rate of complex formation , were effectively invisible to intracellular selection ( the frequency of a mutation with an 80% reduction in complex formation was estimated to be reduced by 0 . 6–4 . 6% compared to a neutral mutation in the ‘best’ parameter set ) ; we expect that mutations in traits like the rate of translation would also be complemented by the wild-type phenotype and so would not experience significant selection during the infection in which they arose . 10 . 7554/eLife . 03753 . 016Figure 5 . Left: mean mutation frequencies for three parameter sets ( ‘low’ , g¯ = 3 . 94; ‘best’ , g¯ = 4 . 65; ‘high’ , g¯ = 5 . 76 ) . Mutation rate is 2 × 10−5 per replication event; ‘relative replication rate’ reflects the reduced probability of a mutant template to replicate , relative to an unmutated strand . Grey lines indicate the expected mean for each parameter set with no selection ( deficit of zero ) ; the black line shows the mutation rate in one replication step , and therefore the expected frequency when mutants cannot replicate . Bars indicate 95% confidence intervals . Right: distributions of g of progeny from single cell infections for three parameter sets ( ‘low’ , g¯ = 3 . 94; ‘best’ , g¯ = 4 . 65; ‘high’ , g¯ = 5 . 76 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03753 . 016 Several distinct features of mutation in this model are evident from Figure 5A . Mutation frequency does not decrease linearly as intracellular selection approaches its maximal value; the curve results from the fact that mutant genomes that are immediately packaged are not subject to selection , while the contribution of rare , early mutants to average mutation frequency may be reduced by multiple rounds of intracellular selection . Knowing g¯ and the mutation rate allows us to directly calculate the fate of neutral or very unfit mutations , but estimating the frequency of mutations of intermediate fitness requires additional simulations using our model . A third feature is the sizable confidence intervals relative to the number of infections sampled ( 10 million for each point ) . This high variability reflects the large contribution of very rare mutations that arise early in an infection and can contribute 1000s of mutant virions , especially when selection is weak . The effect of these rare , early mutations in the overall mutant distribution can be seen as a departure from a Poisson process . To remove a potential confounding variation in burst size , we compare the distribution of mutations from infections within 10% of the median burst size and calculate a Poisson expectation for a median-sized burst with the same expected frequency . For the ‘best’ parameter set , median infections produced many more bursts with no copies of a given mutation ( 79 . 4% vs 22 . 5% for the Poisson ) , but also many more bursts with five or more copies of the mutant ( 8% vs 1 . 82% for the Poisson; n = 51 , 365 ) . The distribution of the number of generations between progeny virions and initial infecting genomes is displayed for three parameter sets ( ‘best’ , ‘low’ , and ‘high’—see Figure 3—source data 1 ) in Figure 5B . Only a very small percentage of progeny are produced via a single genomic replication cycle . Although all three parameter sets have means close to five generations , the distributions show a portion of the progeny virions representing up to 10 generations between the infecting genotypes and packaged virions within a single cellular infection .
The intracellular replication mode of a virus strongly influences the frequency and distribution of mutations among progeny , which shape the long-term behavior of an infecting population ( Vignuzzi et al . , 2006; Lauring et al . , 2012 ) . Due to the complex nature of intracellular dynamics , assessing the mode of replication of viruses is a difficult task ( but see Chao et al . , 2002 ) . Here , we built on decades of mechanistic studies and recent modeling efforts to construct a stochastic computational model coupled with new Bayesian inference methods . We combined these mathematical and computational techniques with accurate temporal data to produce a detailed picture of viral infection . We found that positive- and negative-sense RNA measurements made across multiple MOIs , along with quantitative data on virion packaging , are sufficient to infer that poliovirus replication occurs in several layers of intermediate replication , in contrast to the oft-assumed ‘stamping machine’ model . The implications of the inferred geometric replication mode are as follows: ( 1 ) error rates per-replication are considerably lower than measured rates from full-replication-cycle in vivo studies , ( 2 ) for a given viral polymerase error rate , mutation will progressively accumulate in both genome and anti-genome RNAs , which should result in a more accentuated departure from the master sequence , allowing a better exploration of the available sequence space during a single infection cycle , and ( 3 ) there exists a significant potential for intracellular selection and competition among related genomes , even in infections initiated by only a single genome . Accurate estimates of viral mutation rates are essential for studying viral evolution and have crucial practical applications in drug and vaccine design . While estimates of mutation rates exist for nearly two dozen viruses , estimates of replication modes exist for only a few ( Sanjuan et al . , 2010 ) . Calculating per-replication event mutation rates from observed mutant frequencies is not possible , or even meaningful , without knowledge of the replication mode . Thus , estimates of poliovirus per-replication event mutation rates can vary over 10-fold depending on the assumed replication mode ( Drake , 1993; Sanjuan et al . , 2010 ) . By inferring the mode of replication , we have been able to link estimates of per-replication event mutation rates to published mutant frequencies . The most extensive poliovirus mutant frequency data set estimated an average mutant frequency of 2 × 10−4 ( Acevedo et al . , 2014 ) . Using our inferred value of approximately five intracellular generations , we calculate a per-replication event mutation rate of 2 × 10−4/5 × 2 = 2 × 10−5 , which is in agreement with the average estimates of poliovirus mutation rates calculated in vivo from lethal mutation frequencies ( Acevedo et al . , 2014 ) . Rates of specific types of mutations , such as transversions and transitions , could each be inferred from their mutation frequencies by the same approach . Our inference of five intracellular generations is also in line with previous inferences of replication mode using the Luria-Delbruck fluctuation test null-class method ( Sanjuan et al . , 2010 ) . However , our results highlight some limitations for inferring mutation rates from frequencies: intracellular selection may strongly affect mutation frequencies , and the strong stochastic nature of virus replication appears to deeply modulate minor allele distribution , which in turn will result in imprecise estimates of the expected frequency . In particular , assuming that mutation frequency can be modeled as a Poisson process will lead to inappropriate confidence in measured frequencies . As a consequence , multiple empirical mutation frequencies measurements will be required to obtain a more precise determination of true mutation frequencies . The branched genealogy inferred in our study implies the potential for significant amounts of intracellular complementation , selection , and competition between mutant genomes , even in infections initiated by a single genome ( Novak and Kirkegaard , 1994; Turner and Chao , 1999; Vignuzzi et al . , 2006 ) . Figure 5A demonstrates the extent to which the frequency of a mutation can be skewed by negative selection during the course of an infection . On the other hand , a mutational event that occurred early in replication and conveyed an intracellular replication advantage could potentially give rise to hundreds or thousands of descendant virions in a single generation . If the mutation distribution data in Figure 5B were displayed as a tree ( as in Figure 1 ) , it would contain over 7000 terminal nodes , too many to resolve in a figure . Hence , the apparent potential for mutant interactions is vast . These results suggest that the evolutionary fate of mutations may depend strongly on their intracellular competitive ability , even when multiplicities of infection are low . Additionally , studies that rely on bottlenecks to reduce selection in viral mutation studies ( e . g . , de la Peña et al . , 2000 ) may be allowing more selection than anticipated . Future population dynamics studies should consider the implications of the intracellular expansion of mutant phenotypes . Virus infections are normally depicted as deterministic processes that follow a stereotypical path from infection to progeny production and death of the infected cell . However , experimental data show that some infected cells produce few progeny while others produce large populations of progeny ( Schulte and Andino , 2014 ) . These observations support the notion that stochasticity is an important factor shaping the outcome of infection . By combining accurate experimental measurements with a stochastic model of viral replication , we have obtained a realistic description of how the molecular events driving the life cycle of the virus govern the outcome of infection in each cell . A significant benefit of computational modeling is that the information learned in the empirical process of the development of a model can yield important insights in the biologic processes under study . For example , our initial attempts to fit temporal strand measurement data were unable to match the sharp transition to exponential growth seen in the data . Only after removing the requirement for positive-sense genomes to be translated before becoming replication-competent was our model flexible enough to rapidly create templates for exponential replication . While Novak and Kirkegaard ( 1994 ) demonstrate a requirement of the initial , infecting genomes to be translated before replication can occur , their data did not implicate that all genomes produced at any time during infection must be translated before replicating . Our study suggests that newly synthesized positive-sense genomes may or may not disperse to nucleate new replication complexes within a single cellular infection , allowing us to model intracellular dynamics in a novel way by permitting a portion of newly made positive-sense strands to immediately act as templates for replication without the requirement of translation . Our model succeeds in describing many experimentally observed features of viral replication and is an excellent staging point for future and more accurate models of viral replication and evolution . With the realistic benefits of stochasticity , compartmentalized reactions , and parameters inferred from quantitative , temporal data , it acts as a baseline intracellular viral replication algorithm . More quantitative data , including data on the formation and number of replication compartments , would further inform the model . Potential additions of intracellular selection , complementation , and recombination parameters would allow population evolution studies to explore intracellular dynamics with more precision than previous approaches . The ultimate goal is to generate a comprehensive model incorporating mechanistic replication dynamics learned from virology with selection and complementation dynamics learned from population genetics . This tool could be very powerful for informing future therapeutic and preventative strategies . | Viruses with genetic information made up of molecules of RNA can multiply quickly , but not very accurately . This means that many errors , or mutations , occur when the RNA is copied to create new viruses . The advantage of this rapid , but mistake-filled , RNA replication process is that some of the mutations will be beneficial to the virus . This allows viruses to rapidly evolve , for example , to develop resistance against drugs . The poliovirus is an RNA virus that can cause paralysis and death in humans . To prevent such infections , scientists have extensively studied the poliovirus and have developed effective vaccines against it that have eliminated the virus from all but a few countries . Because so much is known about the poliovirus and because it has a very simple structure , scientists continue to use the poliovirus as a model to study virus behavior . One unknown aspect of the poliovirus' behavior is how it replicates after invading a cell . Are all of its RNA copies made from the original viral RNA that first infected the cell , in what is known as a ‘stamping machine’ model ? Or do the new copies of the RNA also get copied themselves in a ‘geometric replication mode’ that increases the likelihood of mutations and enables the virus to evolve more rapidly ? Viral RNA molecules are copied by one of the virus's own proteins and so before the viral RNA can be replicated , it must first be translated to form viral proteins . When and where replication begins depends on the concentration of translated proteins around the RNA and so replication tends to begin in particular areas of the cell at different times . Schulte , Draghi et al . used mathematical modeling to create computer simulations of the number of polioviruses in a cell that take into account these time and space constraints . By including random elements in the model , the simulated behavior more accurately follows experimentally recorded data than previously used models . The results of the model led Schulte , Draghi et al . to conclude that the poliovirus replicates by the ‘geometric mode’; as new copies of the poliovirus RNA are made , each copy goes on to make more copies . This means that in a single infected cell there are multiple generations of RNA , and each generation may undergo distinct mutations that are passed on to the next set of RNA copies . In fact , Schulte , Draghi et al . found that the average virus released from an infected cell is the great-great-great-granddaughter of the original virus that infected the cell . With so many different generations of virus coexisting in a cell , there are a lot of opportunities for new genetic combinations to occur and for viruses to evolve new abilities . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | Experimentally guided models reveal replication principles that shape the mutation distribution of RNA viruses |
Mammalian articular cartilage is an avascular tissue with poor capacity for spontaneous repair . Here , we show that embryonic development of cartilage in the skate ( Leucoraja erinacea ) mirrors that of mammals , with developing chondrocytes co-expressing genes encoding the transcription factors Sox5 , Sox6 and Sox9 . However , in skate , transcriptional features of developing cartilage persist into adulthood , both in peripheral chondrocytes and in cells of the fibrous perichondrium that ensheaths the skeleton . Using pulse-chase label retention experiments and multiplexed in situ hybridization , we identify a population of cycling Sox5/6/9+ perichondral progenitor cells that generate new cartilage during adult growth , and we show that persistence of chondrogenesis in adult skates correlates with ability to spontaneously repair cartilage injuries . Skates therefore offer a unique model for adult chondrogenesis and cartilage repair and may serve as inspiration for novel cell-based therapies for skeletal pathologies , such as osteoarthritis .
Hyaline cartilage is a skeletal tissue that consists of a single cell type ( the chondrocyte ) embedded within a homogeneous , collagenous extracellular matrix ( reviewed in Gillis , 2018 ) . In mammals , hyaline cartilage is predominantly an embryonic tissue , making up the anlage of the axial ( chondrocranial , vertebral and rib ) and appendicular ( limb ) endoskeleton . The vast majority of mammalian hyaline cartilage is replaced by bone during the process of endochondral ossification , with cartilage persisting temporarily in epiphyseal growth plates , and permanently at relatively few sites within the adult skeleton ( e . g . in joints , as articular cartilage – Decker , 2017 ) . In juvenile mammals , growth of articular cartilage occurs by cellular rearrangement and increases in chondrocyte volume ( Decker et al . , 2017 ) , and by appositional recruitment of new chondrocytes from a superficial population of slow-cycling chondroprogenitor cells ( Hayes et al . , 2001; Dowthwaite et al . , 2004; Karlsson et al . , 2009; Williams et al . , 2010; Kozhemyakina et al . , 2015 ) . However , evidence for the presence of chondroprogenitor cells in adult articular cartilage is scant , and this – combined with the avascular nature of the tissue – may account for why mammalian articular cartilage cannot heal spontaneously following injury ( Hunziker , 1999 ) . Chondrichthyans ( cartilaginous fishes – sharks , skates , rays and holocephalans ) , on the other hand , possess an endoskeleton that is composed largely of hyaline cartilage , and that remains cartilaginous throughout life . Though chondrichthyans reinforce their endoskeleton with a superficial layer of calcified cartilage ( in the form of small mineralized plates called ‘tesserae’ – Dean and Summers , 2006 ) , the core of their endoskeletal elements persists as hyaline cartilage and does not undergo endochondral ossification . Like many fishes ( and unlike mammals ) , chondrichthyans also exhibit an indeterminate type of growth , with a continued ( albeit slow ) increase in size through adulthood ( Dutta , 1994; McDowall , 1994; Frisk and Miller , 2006 ) . It therefore stands to reason that , in chondrichthyans , skeletal tissues may possess a persistent pool of chondroprogenitor cells to facilitate continued growth of their cartilaginous endoskeleton throughout adulthood , and that such cells ( if present ) could also impart the endoskeleton with an ability to undergo spontaneous repair following injury . However , basic mechanisms of hyaline cartilage development , growth and repair in chondrichthyans remain largely unexplored . Here , we characterize the development and growth of the cartilaginous endoskeleton of a chondrichthyan , the little skate ( Leucoraja erinacea ) , from embryonic development to adulthood . We demonstrate conservation of fundamental cellular and molecular characteristics of cartilage development between chondrichthyans and mammals , and we identify unique features of adult skate cartilage that contribute to its continued growth through adulthood . We further show that skates can repair surgically induced partial-thickness cartilage injuries , highlighting this system as a unique animal model for adult chondrogenesis and spontaneous hyaline cartilage repair .
The pectoral fin endoskeleton of jawed vertebrates consisted ancestrally of three basal cartilages – from anterior to posterior , the propterygium , mesopterygium and metapterygium – and a series of articulating distal radials ( Davis et al . , 2004 ) . Among extant jawed vertebrates , this ancestral ‘tribasal’ condition has been retained in the pectoral fins of chondrichthyans and non-teleost ray-finned fishes ( e . g . in sturgeon , gar and bichir ) , but has been reduced in tetrapods and teleosts , to include only derivatives of the metapterygial and pro-/mesopterygial components , respectively ( Davis , 2013 ) . Our study focused on the metapterygium of the skate ( Figure 1 ) , as this element is relatively large , reliably identifiable across all embryonic and post-embryonic stages and easily accessible for surgical manipulation . In mammals , early cartilage development is marked by the accumulation of preskeletal mesenchyme into a ‘condensation’ at the site of future chondrogenesis ( Hall and Miyake , 2000 ) . Cells within this condensation begin to secrete cartilage extracellular matrix ( ECM ) components and undergo overt differentiation into chondrocytes . To investigate the early development of cartilage in the skate , we prepared a histological series of skate metapterygia from embryonic stage ( S ) 30 through to hatching and used a modified Masson’s trichrome stain to visualize condensation , differentiation and ECM secretion . At S30 , the presumptive metapterygium exists as condensed mesenchyme , with pericellular Light Green staining ( which appears blue ) indicating onset of ECM secretion by cells within the condensation ( Figure 2a ) . From S31-33 ( Figure 2b–d ) , the metapterygium differentiates into cartilage and grows , with cells in the centre of the element adopting differentiated chondrocyte morphology ( i . e . cells residing within ECM cavities or ‘lacunae’ ) , but cells in the periphery maintaining a less differentiated appearance . By hatching ( Figure 2e ) , the cartilage of the metapterygium is surrounded by a distinct fibrous perichondrium ( Figure 2ei ) , beneath which we begin to observe modifications of the ECM at sites of superficial calcification ( i . e . developing tesserae – Figure 2eii ) . By hatching , cells throughout the cartilage have adopted differentiated chondrocyte morphology , and are embedded within extensive collagenous ECM ( Figure 2f ) . Mammalian hyaline cartilage ECM is composed largely of fibrils of type II collagen , which entrap aggregates of the hydrated proteoglycan aggrecan ( Eyre , 2002; Kiani et al . , 2002 ) . Col2a1 and Agc ( the genes encoding type II collagen and aggrecan , respectively ) , in turn , are directly transcriptionally regulated in chondrocytes by the SRY-box transcription factors Sox9 , Sox5 , and Sox6 ( Bell et al . , 1997; Lefebvre et al . , 1998; Lefebvre et al . , 2001 ) . To test for conservation of these gene expression features in chondrocytes of the skate metapterygium , we characterized the co-expression of genes encoding cartilage ECM components and upstream transcriptional regulators in situ . We first cloned fragments of skate Cola2a1 ( Figure 3—figure supplement 1 ) and Agc ( Figure 3—figure supplement 2 ) and tested for their expression in the S32 metapterygium by chromogenic mRNA in situ hybridization . We found that both Col2a1 ( Figure 3a ) and Agc ( Figure 3b ) are expressed in chondrocytes throughout the skate metapterygium , reflecting shared ECM properties between skate and mammalian hyaline cartilage . To test for conservation of the regulatory relationship between Sox5 , Sox6 and Sox9 , we used multiplexed fluorescent in situ hybridization by chain reaction ( HCR ) to test for co-expression of these genes ( Figure 3—figure supplements 3–4 ) in metapterygium chondrocytes . We observed co-expression of Col2a1 and Sox9 in chondrocytes throughout the metapterygium ( Figure 3c–d ) , as well as co-expression of Col2a1 , Sox5 and Sox6 ( Figure 3e–f ) , indicating likely conservation of regulation of genes encoding cartilage ECM components by SoxE- and SoxD-class transcription factors in skate cartilage . To characterize patterns of cell proliferation within the growing metapterygium , we conducted a label retention experiment in skate hatchlings . Incorporation and detection of thymidine analogues , such as 5-ethynyl-2'-deoxyuridine ( EdU ) , provides a sensitive readout of DNA synthesis and , by extension , cell proliferation ( Salic and Mitchison , 2008 ) . Briefly , hatchling skates were given a single intraperitoneal microinjection of EdU , and were then harvested at 1- , 5- , 10- and 40 days post-injection ( hereafter referred to as 1- , 5- , 10- and 40 day chase , respectively ) , to test for label retention within the metapterygium . In animals analyzed at 1 day chase , EdU+ chondrocytes were recovered throughout the cartilage of the metapterygium , though with a concentration of EdU+ cells around the periphery of the element ( Figure 4a ) , pointing to the continued proliferation of differentiated chondrocytes at hatching . We also observed EdU+ cells within the perichondrium of the metapterygium at 1 day chase ( Figure 4b ) . These label-retaining cells exhibited a distinct , flattened nuclear morphology ( relative to adjacent chondrocytes ) , and were recovered in increasingly greater numbers within the perichondrium at 5- ( Figure 4c ) and 10 day chase ( Figure 4d ) . By 40 days chase ( Figure 4d ) , we observed a marked decrease in the number of EdU+ perichondral cells . This pattern of label retention ( Figure 4f ) is suggestive of an expanding or self-renewing cell population within the perichondrium , with a greater number of label-retaining cells arising through proliferation of EdU+ progenitors but reduction in the number of label-retaining cells with eventual dilution of EdU to undetectable levels . In animals analyzed at 10- ( Figure 4g ) and 40 days chase ( Figure 4h ) , we observed numerous clusters of EdU+ chondrocytes in cartilage immediately adjacent to EdU+ perichondral cells . As cells of the inner perichondrium are known to give rise to new chondrocytes in the cartilaginous anlage of the chick limb skeleton ( Scott-Savage and Hall , 1979 ) , we speculate that these subperichondral EdU+ chondrocytes are the progeny of label-retaining perichondral cells , and that growth of the hatchling metapterygium occurs both through proliferation of differentiated chondrocytes , and by recruitment of new chondrocytes from progenitor cells within the perichondrium . Using mRNA in situ hybridization by HCR , we investigated gene expression at the cartilage-perichondrium interface within the metapterygium of skate hatchlings . As in S32 embryos , we noted co-expression of Col2a1 and Sox9 ( Figure 4i ) and Col2a1 , Sox5 and Sox6 ( Figure 4j ) in chondrocytes of the metapterygium . However , we also observed cells within the perichondrium that expressed Sox9 , Sox5 and Sox6 but not Col2a1 ( Figure 4i , j ) . These cells were invariably located within the innermost region of the perichondrium , at the cartilage-perichondral interface , and exhibited the flattened nuclear morphology of the EdU+ perichondral cells described above . Taken together , our label retention and gene expression data point to the perichondrium as a source of cartilage progenitor cells in skate hatchlings . We next examined histological features of the adult skate metapterygium , by vibrotomy ( Figure 5a ) and histochemical staining of paraffin sections with a modified Masson’s trichrome stain ( Figure 5b ) . In transverse sections through the metapterygium , we noted that the core of the element has the glassy appearance of hyaline cartilage , while the surface of the cartilage is covered by a rind of calcified tesserae ( Figure 5a–b ) . Cells in the core of the metapterygium exhibit typical hyaline chondrocyte morphology ( Figure 5c ) , are embedded in an ECM that stains with Light Green , and express high levels of type II collagen ( Figure 5d ) . A higher magnification view of the tesserae reveals that these sit beneath a well-developed , fibrous perichondrium , which stains variably red and blue/green with trichrome staining ( Figure 5e ) . We intermittently see a thin layer of unmineralized cartilage between the surface of the tesserae and the perichondrium , likely corresponding with the ‘supratesseral’ cartilage that has been previously reported in the stingray , Urobatis halleri ( Seidel et al . , 2017 ) . Closer examination of the intertesseral joint region ( i . e . the zone of cartilage between adjacent tesserae – Figure 5f ) reveals cells with typical chondrocyte morphology within ECM up to the boundary between hyaline cartilage and the perichondrium , as well as a distinct population of flattened , spindle-shaped cells at the boundary between cartilage and perichondrium ( Figure 5fi ) , hereafter referred to as ‘inner perichondral cells’ . Interestingly , the adult skate metapterygium is permeated by a series of canals , which originate at the surface of the cartilage , and extend toward the core of the element ( Figure 5a–b ) . These cartilage canals occur throughout the metapterygium , originate within the perichondrium and enter the cartilage between tesserae ( i . e . through the intertesseral joint region ) , and are not lined by an epithelium ( Figure 5g ) . Cartilage canals contain an abundance of cells , including some red blood cells ( Figure 5h ) , but predominantly cells with connective tissue/mesenchymal morphology – many of which appear to be invading from the canal into surrounding cartilage ECM ( Figure 5i ) . Immunostaining for type II collagen reveals that cartilage canals are zones of active ECM synthesis , with high levels of type II collagen being secreted by cells at the periphery of the canals ( Figure 5j ) . To test for cells that are actively expressing cartilage ECM products in the adult skeleton , we analyzed expression of Col2a1 by mRNA in situ hybridization on sections of adult metapterygium . High levels of Col2a1 transcription were detected around the periphery of the cartilage – that is in the intertesseral joint region and in cartilage adjacent to tesserae – and also in the thin layer of supratesseral cartilage that sits between the tesserae and the perichondrium ( Figure 6a ) . In situ hybridization by HCR revealed that both supratesseral and peripheral chondrocytes co-expressed Col2a1 and Sox9 ( Figure 6b–c ) , as well as Col2a1 , Sox5 , Sox6 ( Figure 6d–e ) , pointing to retention of transcriptional features of developing cartilage around the periphery of the adult metapterygium . Interestingly , as in hatchling skates , we also observed co-expression of Sox5 , Sox6 and Sox9 ( but not Col2a1 ) in the flattened cells of the inner perichondrium ( Figure 6b , d ) . Given the indeterminate growth of cartilaginous fishes , we speculated that the transcriptional signature of embryonic cartilage in the periphery of the adult skate metapterygium could reflect recently born chondrocytes contributing to ECM expansion , while the presence of this signature in the inner perichondrium could reflect progenitors of new chondrocytes . We therefore sought to test for the presence and fate of cycling cells in the adult skate metapterygium using a label retention experiment . Due to the relatively slow growth rate of cartilaginous fishes , we reasoned that a pulse-chase label retention experiment could be used not only to localize cell proliferation within the metapterygium , but also to lineage trace label-retaining cells to test for contributions to hyaline cartilage . Briefly , eight adult female skates were given three intraperitoneal injections of EdU , 48 hr apart , and two animals were then euthanized , fixed and processed for EdU detection 3 days , 1 month , 2 months and 5 . 5 months following the final IP EdU injection ( hereafter referred to as 3 day , 1 month , 2 month and 5 . 5 month chase , respectively ) . EdU detection was performed on transverse paraffin sections though the metapterygium , with EdU+ cells scored according to their location in the outer perichondrium , inner perichondrium , cartilage canals or cartilage ( i . e . chondrocytes ) . 30–40 sections were analyzed from each animal , but quantification was performed by counting the sum total and tissue localization of EdU+ cells in five adjacent sections through the metapterygium ( as indicated in Figure 1i ) after pulse + 3 days , 1 month , 2 month and 5 . 5 month chase ( Table 1 ) . After a 3 day chase , EdU+ cells were recovered almost exclusively in the perichondrium , with most appearing as cells with rounded nuclei in the outer perichondrium ( Figure 7a ) , and relatively few as flattened cells of the inner perichondrium ( Figure 7b ) . No EdU+ chondrocytes were detected after a 3 day chase . After 1 month and 2 month chases , we continued to detect EdU+ cells within the outer and inner perichondrium , and we also observe EdU+ cells within the cartilage canals that originate in the perichondrium and permeate the core of the metapterygium ( Figure 7c–d ) . After a 5 . 5 month chase , we detected an abundance of EdU+ cells in both the outer and inner perichondrium ( Figure 7e ) , as well as abundant EdU+ chondrocytes in the peripheral hyaline cartilage of the intertesseral joint region ( Figure 7f ) and in cartilage subjacent to the tesserae ( Figure 7g ) , as well as relatively few EdU+ chondrocytes within the mineralized matrix of the tesserae ( Table 1 ) . EdU+ cells are present in greater abundance within cartilage canals after a 5 . 5 month chase ( Figure 7h ) and given the relative dearth of EdU+ cells in cartilage canals after the 3 day , 1 month and 2 month chases , we speculate that these EdU+ cartilage canal cells are of perichondral origin . After the 5 . 5 month chase , we also observed EdU+ cells that appeared to be invading hyaline cartilage from the blind ends of cartilage canals ( Figure 7i ) , and in one individual we observed three instances of EdU+ chondrocytes immediately adjacent to the blind ends of cartilage canals ( Figure 7j ) . Taken together , our label retention and gene expression data point to the morphologically distinct Sox5/Sox6/Sox9 + cells of the inner perichondrium as adult cartilage progenitor cells , with a capacity to give rise to new chondrocytes both in the periphery ( i . e . appositional growth ) , and also in the core of the metapterygium , via cartilage canals ( i . e . interstitial growth ) . Additionally , while the number of EdU+ cells is highly variable between individuals , the general trend of a greater number of label-retaining perichondral cells after the 2 month and 5 . 5 month chases – and , more specifically , a greater number of label-retaining cells within the inner perichondrium – points to the likely self-renewal of perichondral cells , perhaps with a progressive sequence of differentiation from outer perichondral cell to inner perichondral cell , and eventually to chondrocyte – either in the periphery , or deeper in the core ( via cartilage canals ) . Mammalian articular hyaline cartilage is unable to spontaneously heal following injury . Rather , articular cartilage injuries tend to infill with fibrocartilage – a subtype of cartilage that exhibits large bundles of collagen fibres , and with ECM composed substantially of type I collagen ( Eyre and Wu , 1983; Benjamin and Ralphs , 2004 ) . Fibrocartilage is mechanically inferior to hyaline cartilage at the articular surfaces of synovial joins , and its formation within articular cartilage lesions can result in the onset of degenerative osteoarthritis . We sought to test whether the persistence of adult chondrogenesis in the skate metapterygium – and the presence of cartilage progenitor cells in the perichondrium – correlated with an ability to spontaneously repair injured hyaline cartilage . We conducted a surgical cartilage injury experiment , in which a metapterygium cartilage biopsy was performed in 26 adult skates using a 4 mm biopsy punch ( producing a cartilage void of ~1/4 to 1/3 diameter of the metapterygium – Figure 8a ) . Two animals were euthanized one-week post-operation , and at monthly intervals for the following year , and processed histologically to assess the extent of repair . In animals assessed at 1 and 2 months post-operation ( mpo ) ( n = 4 ) , cartilage injuries were in-filled with a fibrous connective tissue ( Figure 8b , bi ) , but by 3mpo ( n = 2 ) , this connective tissue began to differentiate into cartilage ( i . e . with typical hyaline chondrocytes , albeit at a much higher density than in the adjacent , uninjured cartilage – Figure 8c , ci ) . In animals assessed at 4-10mpo ( n = 14 ) , tissue within the injury sites showed varying degrees of progressive differentiation into cartilage , starting from the interface between the injury site and adjacent cartilage and progressing toward the surface of the metapterygium ( Figure 8—figure supplement 1 ) , and by 11-12mpo ( n = 4 ) , injury sites were completely or near-completely filled with repair cartilage ( Figure 8d , Figure 8—figure supplement 2 , 3 ) . Chondrocytes within the repair cartilage remained at much higher density than in the adjacent cartilage , and the surface of the injury site remained irregular , with some superficial red staining of the ECM ( Figure 8di ) . This could reflect the re-appearance of tissue with a perichondral-like nature , or a step toward re-establishment of tesserae at the surface of the metapterygium , as these tissues stained variably red-blue and red , respectively , with modified Masson’s trichrome stain . However , the vast majority of repair tissue resembled typical hyaline cartilage , with a Light Green-stained ECM that integrates with adjacent uninjured cartilage and no evidence of ECM fibre bundles typical of fibrocartilage ( Figure 8dii ) . We tested whether the ECM of repair cartilage was composed of type II collagen ( as in typical hyaline cartilage ) or a mixture of types I and II collagen ( as in fibrocartilage ) using immunofluorescence . We observed strong , positive staining for type II collagen throughout the ECM of repair cartilage , as well as pericellular staining for type II collagen in adjacent uninjured hyaline cartilage ( Figure 8e – Figure 8—figure supplement 4 ) . Conversely , we observed no positive staining for type I collagen in repair or uninjured cartilage ( including cartilage canals ) , despite strong positive staining in adjacent skeletal muscle fibres ( Figure 8—figure supplement 4 ) . These findings suggest that adult skate repair cartilage produces ECM similar to that of adjacent hyaline cartilage and is unlike the fibrocartilaginous repair tissue that typically fills mammalian chondral defects . Interestingly , in two animals collected between 4-10mpo , our biopsy had been unsuccessful , with the biopsy punch perforating the surface of the metapterygium but failing to remove a wedge of cartilage . In both cases , a large mass of ectopic cartilage formed above the tesserae on the surface of the metapterygium , but beneath the fibrous perichondrium ( Figure 8—figure supplement 5 ) . This suggests that mechanical perturbation of the perichondrium may act as an inductive cue for onset of a chondrogenic injury response , and is consistent with the perichondrium as a source of new cartilage not only during normal adult growth , but also following injury .
The endoskeleton of bony fishes ( including tetrapods ) forms largely through a process of endochondral ossification . In endochondral ossification , pre-skeletal mesenchyme aggregates to form a condensation at the site of skeletogenesis , and cells within this condensation undergo progressive differentiation , starting from the centre , into chondrocytes and eventually enlarged , hypertrophic chondrocytes ( Hall , 2005; Karsenty et al . , 2009; Long and Ornitz , 2013 ) . Chondrocyte and hypertrophic chondrocyte fates are determined and characterized by the expression of genes encoding distinct sets of transcription factors and ECM components , with the transcription factors Sox5 , Sox6 and Sox9 regulating the expression of Col2a1 and Agc1 in chondrocytes ( Bell et al . , 1997; Lefebvre et al . , 1998; Lefebvre et al . , 2001; Smits et al . , 2001; Akiyama et al . , 2002; Han and Lefebvre , 2008 ) , and the transcription factor Runx2 regulating the expression of Col10a1 ( the gene encoding non-fibrillar type X collagen ) in hypertrophic chondrocytes ( Linsenmayer et al . , 1991; Takeda et al . , 2001; Zheng et al . , 2009; Simões et al . , 2006; Higashikawa et al . , 2009; Ding et al . , 2012 ) . Hypertrophic cartilage is ultimately invaded by vasculature , and is replaced by bone , with bone-forming cells ( osteoblasts ) arising both from the perichondrium/periosteum , and through transdifferentiation of hypertrophic chondrocytes ( Colnot et al . , 2004; Roach , 1992; Roach et al . , 1995; Zhou et al . , 2014; Yang et al . , 2014; Park et al . , 2015; Hu et al . , 2017 ) . Within growing endochondral bones , non-hypertrophic cartilage persists in the growth plate , where chondrocytes continue to proliferate and contribute to lengthening of an element , and at points of endoskeletal articulation . Upon cessation of growth , growth plate cartilage will hypertrophy and ossify , with non-hypertrophic cartilage persisting only at articular surfaces . In skate , onset of endoskeletal development is marked by the appearance of mesenchymal condensations at sites of skeletogenesis , and cells within condensations differentiate into chondrocytes progressively , from the centre of the condensation to the periphery . This differentiation of condensed mesenchyme is accompanied by co-expression in chondrocytes of Sox5 , Sox6 , Sox9 , Col2a1 and Agc1 , pointing to conservation of the regulatory interaction between SoxD- and E-class transcription factors and the genes encoding type II collagen and aggrecan in cartilaginous and bony fishes , and to the broad developmental and biochemical comparability of cartilage between these major vertebrate lineages . Unlike in bony fishes , however , chondrocytes in the skate metapterygium do not undergo hypertrophy , but rather remain terminally differentiated in a non-hypertrophic state . While cartilaginous fishes , strictly speaking , lack bone , they nevertheless possess the vast majority of transcription factors and ECM components required to make bone ( Venkatesh et al . , 2014 ) , and there are instances of mineralization within the skeleton of cartilaginous fishes that share molecular properties with bone – for example expression of Col10a1 , Col1a1 and SPARC in the mineralized areolar tissue of the vertebral column , and immunolocalization of types I and X collagen to tesserae and pre-mineralized supratesseral cartilage ( Egerbacher et al . , 2006; Enault et al . , 2015; Criswell et al . , 2017; Seidel et al . , 2017; Debiais-Thibaud et al . , 2019 ) . These observations , combined with palaeontological evidence for the presence of bone along the gnathostome stem ( Donoghue et al . , 2006; Charest et al . , 2018 ) , are consistent with bone as an ancestral feature of jawed vertebrates , the loss of this tissue in extant cartilaginous fishes , and the independent re-deployment of deeply conserved mechanisms of cartilage mineralization at sites such as tesserae and the axial column . There has been relatively little work on postembryonic growth of hyaline cartilage in cartilaginous fishes , perhaps owing to the general difficulties of locating and maintaining a suitable range of subadult life stages . Growth of the tessellated calcified cartilage of cartilaginous fishes has been well documented , particularly in the stingray Urobatis halleri , where it has been shown that tesserae continue to grow throughout life , and that this growth likely occurs by accretion , with continuous mineralization of a thin layer of ‘supratesseral’ cartilage that sits between the tesserae and the perichondrium , and of ‘subtesseral’ cartilage beneath the tesserae ( Dean et al . , 2009; Seidel et al . , 2016; Seidel et al . , 2017 ) . We have discovered a population of label-retaining cartilage progenitor cells ( characterized by co-expression of Sox5 , Sox6 and Sox9 ) in the inner perichondrium of adult skates , and we have traced their chondrocyte progeny to the peripheral unmineralized cartilage and ( occasionally ) mineralized tesserae of the metapterygium . Based on our observation of a similar cell type in the metapterygium perichondrium of hatchling skates , these observations point to a general mechanism of appositional post-embryonic cartilage growth , wherein progenitor cells of perichondral origin give rise to new chondrocytes in the periphery of the metapterygium – some of which will become incorporated into growing tesserae , while others ( e . g . in the intertesseral joint regions ) will persist as unmineralized hyaline cartilage ( Figure 9 ) . In addition to appositional growth of cartilage in the metapterygium , we also find some evidence of interstitial growth , by the addition of new chondrocytes to the unmineralized cartilaginous core of the metapterygium . In our pulse-chase label retention experiments , EdU+ cells are only recovered in the perichondrium after a pulse + 3 day chase ( with the exception of a single label-retaining cell in a cartilage canal ) , with no label-retention in differentiated chondrocytes . However , we observe a striking increase in label-retaining cells within cartilage canals and core cartilage after pulse + 1–5 . 5 month chases , including a small number of chondrocytes in the centre of the metapterygium cartilage core , immediately adjacent to cartilage canals . Given the initial distribution of label retaining cells after the 3 day chase , we speculate that these core chondrocytes derive from label-retaining progenitors in the perichondrium and were transported to the cartilaginous core of the metapterygium via cartilage canals ( Figure 9 ) . However , skate cartilage canals also contain red blood cells ( Figure 4h ) , which suggests that these structures may serve to vascularise the adult cartilaginous endoskeleton . If this is the case , then such vasculature could serve to transport cartilage progenitor or mesenchymal stem cell-like cells , from niches elsewhere in body , to the core cartilage of the metapterygium . Cartilage canals that extend from the perichondrium into the core of endoskeletal elements have been shown to occur in the vertebrae , jaws and pectoral girdles of sharks and rays ( Hoenig and Walsh , 1982; Dean et al . , 2010 ) . Similar canals have been described from the cartilage of embryonic tetrapods , though in tetrapods these are transient structures that function in mediating the replacement of cartilage by bone during endochondral ossification ( Blumer et al . , 2004; Blumer et al . , 2008 ) . Conversely , cartilage canals in cartilaginous fishes persist within the adult skeleton , have been described as containing blood vessels and lymph-like and other amorphous materials , as well as immature chondrocytes , and have been speculated to function in the nourishment and maintenance of cartilage in the adult endoskeleton ( Hoenig and Walsh , 1982; Dean et al . , 2010 ) . The cartilage canals that we have described in the metapterygium of the skate closely resemble those described previously in other cartilaginous fishes , and findings from our label retention experiments are consistent with a function for these canals in transporting material ( including pre-chondrocytes ) to the core cartilage of the metapterygium . It therefore seems likely that cartilage canals do , indeed , function in the maintenance , nourishment , and , most likely , interstitial growth of cartilage in the endoskeleton of cartilaginous fishes . We have found that the persistence of cartilage progenitor cells and chondrogenesis in the adult skate skeleton correlates with an ability to spontaneously repair injured cartilage – albeit with a tissue containing a significantly higher density of chondrocytes relative to normal adult cartilage . A comparative analysis of the mechanical properties of normal and repair cartilage remains to be conducted , though the ECM of skate repair cartilage is composed of type II collagen and appears to integrate seamlessly with adjacent tissue , and cells within repair cartilage appear indistinguishable from adjacent chondrocytes . While it is not currently possible to precisely trace the cell lineage of repair cartilage within our injury paradigm , the demonstrated chondrogenic potential of the adult skate perichondrium during normal growth , as well as our observation that mechanical perturbation of the surface of the metapterygium is sufficient to induce a large mass of ectopic cartilage beneath the perichondrium , points to the perichondrium as the most likely source of repair cartilage following injury . Our findings stand in contrast with previous work in the dogfish ( Scyliorhinus spp . ) , which determined that the cartilaginous skeleton of sharks could not heal following injury ( Ashhurst , 2004 ) . This was based on a design in which cartilaginous fin radials were bisected and monitored for repair over 26 weeks . It was observed that cut surfaces of dogfish fin rays were initially capped by a fibrous tissue , with subsequent appearance near the injury site of a disorganized , cartilage-like tissue that failed to integrate with existing ray cartilage or to unite the bisected elements . Importantly , the cartilage injury in that study ( i . e . complete bisection ) was severe , and may have posed an insurmountable barrier to repair , even for a tissue with local progenitors and repair potential ( i . e . it is possible that the repair response that we report in skate requires some scaffold of normal cartilage as a foundation for repair ) . Additionally , fin rays are relatively small , and therefore may exhibit different growth properties relative to larger elements of the endoskeleton . Additional studies of repair potential across the range of skeletal elements and tissue types in cartilaginous fishes are needed to determine whether cartilage repair is a general feature of the skeleton , or a unique property of specific skeletal elements . Osteoarthritis ( OA ) is a debilitating deterioration of joint cartilage with symptoms ranging from stiffness and joint pain to complete immobility . OA can severely impact quality of life , and has an extremely high economic burden , and so there is great interest in identifying novel therapeutic strategies to promote joint cartilage repair . Joint cartilage repair still poses a substantial clinical challenge , owing to the avascular and aneural nature of articular cartilage , and therefore its limited capacity to initiate spontaneous repair . Recently , focus has shifted from surgical approaches ( e . g . microfracture and autologous chondrocyte implantation – Rodrigo et al . , 1994; Brittberg et al . , 1994 ) to stem cell-based therapies for cartilage defects – namely the application of patient-derived mesenchymal stem cells ( MSCs ) or induced pluripotent stem cells ( iPSCs ) as sources of repair tissue for damaged cartilage ( Bernhard and Vunjak-Novakovic , 2016; Murphy et al . , 2017; Harrell et al . , 2019 ) . While these approaches hold great promise , some challenges nevertheless remain . Derivation of persistent cartilage from MSCs is challenging , as chondrogenically differentiated MSCs will often continue on a path toward hypertrophy and ultimately ossification ( Pelttari et al . , 2006; Steinert et al . , 2007 ) , while the use of incompletely differentiated iPSCs can result in heterogeneous repair tissue , and may come with a risk of teratoma formation ( Heng et al . , 2004; Saito et al . , 2015 ) . The unique endoskeletal growth and repair properties of cartilaginous fishes may offer a powerful model to inform novel cell-based strategies for mammalian cartilage repair . It remains to be determined whether the adult skate perichondrium is a homogeneous cell population with equivalent chondrogenic potential throughout , or a heterogeneous tissue containing a specialized subpopulation of chondroprogenitors . However , the ability of this tissue to give rise to cartilage as a terminal product – rather than cartilage as an intermediate step toward endochondral bone – is significant and could be exploited to further our understanding of the molecular basis of stable and reliable adult chondrogenesis in vitro and in vivo . Further characterization of the transcriptional fingerprint of cell types within the skate perichondrium , and the gene regulatory basis skate chondrocytes differentiation during growth and repair , could provide a rich source of information on how MSCs and/or iPSCs could be edited or manipulated to enhance their efficacy in mammalian articular cartilage repair .
Leucoraja erinacea adults and embryos were maintained in large rectangular tanks or seatables , respectively , in flow-through natural seawater at a constant temperature of 15°C at the Marine Resources Center of the Marine Biological Laboratory in Woods Hole , MA , U . S . A . Adult skates were fed on a diet of squid and capelin . Skate embryos were staged according to the lesser-spotted dogfish ( Scyliorhinus canicula ) staging table of Ballard et al . ( 1993 ) and the winter skate ( Leucoraja ocellata ) staging table of Maxwell et al . ( 2008 ) . Prior to fixation , all skate embryos and adults were euthanized with an overdose of ethyl 3-aminobenzoate methanesulfonate salt ( MS-222 - Sigma ) in seawater ( 1 g/L MS-222 buffered with 2 g/L sodium bicarbonate ) . Animals were kept in a euthanasia bath until the cessation of gill pumping and heartbeat , and for adults , decapitation was used as a secondary method of euthanasia . All skate embryos were fixed overnight in 4% paraformaldehyde ( Electron Microscopy Science ) in 1X phosphate buffered saline ( PBS – ThermoFisher ) , rinsed 3 × 10 min in 1X PBS , dehydrated stepwise into 100% methanol , and stored in methanol at −20°C prior to analysis . Pieces of dissected adult skate cartilage to be used for histochemical staining , immunofluorescence or EdU detection were fixed in 4% paraformaldehyde in filtered seawater for 48 hr , rinsed into filtered seawater containing ~0 . 5% paraformaldehyde and stored at 4°C . Pieces of dissected adult skate cartilage to be used for mRNA in situ hybridization were fixed for 48 hr in 4% paraformaldehyde in 1X PBS , then rinsed 3 × 10 min in 1X PBS , dehydrated stepwise into 100% methanol , and stored in methanol at −20°C . All work involving skate embryos and adults was conducted in strict accordance with protocols approved by the Marine Biological Laboratory Institutional Animal Care and Use Committee . For paraffin embedding of embryonic tissue , specimens were cleared 3 × 20 min in Histosol ( National Diagnostics ) at room temperature , transitioned through 2 × 30 min steps in 1:1 Histosol:molten paraffin in a standard wax oven at 60°C , then left in molten paraffin ( RA Lamb Wax – Fisher Scientific ) at 60°C overnight . The following day , specimens were moved through four changes of molten paraffin ( each >1 hr ) before positioning and embedding in a Peel-A-Way embedding mold ( Sigma ) . For paraffin embedding of adult skate tissue , samples were rinsed several times in water , and then demineralised in a 10% ( w/v ) solution of ethylenediaminetetraacetic acid ( EDTA ) in 0 . 1M Tris pH 7 . 2 for 20–25 days on a rocking platform at 4°C . Upon completion of demineralisation , samples were washed several times with water , and then infiltrated with paraffin under vacuum in a tissue processor ( with stepwise dehydration into 100% ethanol , 3 × 1 hr washes in xylene and 4 × 45 min washes in molten paraffin ) before embedding in a Peel-A-Way embedding mold . All blocks were left to set overnight at room temperature before sectioning at 8 μm on a Leica RM2125 rotary microtome . Sections were mounted on Superfrost plus slides ( VWR ) and stained with a modified Masson’s trichrome stain , according to the protocol of Witten and Hall ( 2003 ) . All histochemical staining was carried out at least in triplicate ( i . e . on three separate stage-matched individuals ) . Slides to be used for immunofluorescence were dewaxed in histosol and rehydrated through a descending ethanol series into 1X PBS + 0 . 1% Triton X-100 ( PBT ) . For enzymatic antigen retrieval , slides were incubated in 147 u/mL hyaluronidase ( Sigma ) in PBS pH 6 . 7 for 1 hr at 37°C followed by 0 . 1% ( w/v ) pepsin ( Sigma ) in 0 . 01N HCl for 30 min at 37°C for anti-Col2a1 or 0 . 1% ( w/v ) pepsin in 0 . 5M acetic acid for 2 hr at 37°C for anti-Col1a1 , according to the protocol of Egerbacher et al . ( 2006 ) . Slides were then rinsed 3 × 10 min in PBT , blocked for 30 min in 10% sheep serum and incubated in primary antibody ( under a parafilm coverslip ) in a humidified chamber overnight at 4°C . The following day , slides were rinsed 3 × 5 min in 1X PBT , and then incubated in secondary antibody ( under a parafilm coverslip ) in a humidified chamber overnight at 4°C . Slides were then rinsed 3 × 10 min in PBT and coverslipped with Fluoromount G containing DAPI ( Southern Biotech ) . Primary and secondary antibodies were diluted in 10% sheep serum in PBT to the following concentrations: anti-COL2A1 ( II . II6B3 , Developmental Studies Hybridoma Bank , University of Iowa; 1:20 ) , anti-COL1A1 ( LF-68 , Kerafast; 1:100 ) , AF568-conjugated goat-anti-rabbit IgG ( A11011 , Invitrogen; 1:500 ) and AF488-conjugated goat-anti-mouse IgG ( A11001 , Invitrogen; 1:500 ) . All immunostaining was carried out in triplicate ( on three separate stage-matched individuals ) , and negative controls were conducted by following the same staining protocol but in the absence of primary antibody . 5-ethynyl-2'-deoxyuridine ( EdU – ThermoFisher Scientific ) retention experiments were conducted to label the nuclei of cell that have undergone S-phase DNA synthesis and their progeny ( Salic and Mitchison , 2008 ) . For skate hatchling EdU retention experiments , hatchlings were anaesthetized in MS-222 in seawater ( 150 mg/L MS-222 buffered with 300 mg/L sodium bicarbonate ) and given a single intraperitoneal ( IP ) microinjection of 2 μL of a 5 mM EdU solution in 1X PBS with a Picospritzer pressure injector . Animals were then recovered in seawater and reared in a flow-through seatable for 1 , 5 , 10 or 40 days , at which point animals were euthanized and fixed as described above . For adult skate EdU pulse-chase experiments , eight adults weighing between 750–850 g were given three 5 mL IP injections of 22 mM EdU in 1X PBS ( a dose of 27 . 7 mg EdU/injection ) , with 48 hr between each injection . Animals were anaesthetized in MS-222 in aerated seawater ( 150 mg/L MS-222 buffered with 300 mg/L sodium bicarbonate ) prior to injection , and after injection , animals were recovered in aerated seawater before being returned to their tank . Two animals were euthanized , dissected and fixed as described above 3 days , 1 month , 2 months and 5 . 5 months after the final EdU injection . EdU detection was carried out on 8 μm paraffin sections using the Click-iT EdU Cell Proliferation Kit ( ThermoFisher Scientific ) according to the manufacturer’s instructions . After detection , slides were coverslipped with Fluoromount G containing DAPI , imaged , then de-coverslipped in water and stained with modified Masson’s trichrome . Given the large size of the metapterygium and the relative sparsity of certain label-retaining cell types ( e . g . chondrocytes in adult sections ) , cell counts were performed across five successive sections through the region of the metapterygium indicated in Figure 1 for adult skates , and three successive sections through the equivalent region of the metapterygium in hatchlings . For counts of EdU+ perichondral cells in skate hatchlings , the sum total of EdU+ cells for each individual were plotted against chace time in MS Excel , and a polynomial trend line was added . For paraffin embedding and sectioning of adult tissues for mRNA in situ hybridization , tissues were rehydrated into diethyl pyrocarbonate ( DEPC ) -treated water and demineralised for 24 hr in Morse Solution ( 5 g sodium citrate dihydrate , 12 . 5 mL formic acid and 37 . 5 mL DEPC water ) . Demineralised tissues were then dehydrated stepwise into 100% ethanol before infiltration , paraffin embedding and sectioning as described above . Chromogenic mRNA in situ hybridization on paraffin sections for Leucoraja erinacea Col2a1 ( GenBank MT254563 ) and Agc ( GenBank MT254564 ) was carried out according to the protocol of O'Neill et al . ( 2007 ) , with modifications according to Gillis et al . ( 2012 ) . mRNA in situ hybridization by chain reaction ( HCR ) was carried out according to the protocol of Choi et al . ( 2018 ) with the following modifications: slides were pre-hybridized for 30 min at 37°C; for the hybridization step , 0 . 8 μL of each 1 μM probe stock was used per 100 μL of hybridization buffer; and hairpins were used at 4 μL of hairpin stock per 100 μL of amplification buffer . Probe sets for Leucoraja erinacea Col2a1 ( GenBank MT254563 ) , Sox9 ( GenBank MT254560 ) , Sox5 ( GenBank MT254561 ) and Sox6 ( GenBank MT254562 ) and hairpins were purchased from Molecular Instruments . Molecular Instrument probe lot numbers are as follows: Col2a1 ( PRB574 ) , Sox9 ( PRB571 ) , Sox5 ( PRB572 ) and Sox6 ( PRB573 ) . All mRNA in situ hybridization slides were coverslipped with Fluoromount G containing DAPI . Orthology of the Leucoraja erinacea Col2a1 , Agc , Sox9 and Sox6 sequences used for probe design was confirmed by phylogenetic analysis . Full or partial coding sequences were translated using ORFfinder ( NCBI ) , and multiple sequence alignments with aggrecan , clade A collagen , SoxD and SoxE protein family members were constructed using Clustal Omega ( Sievers and Higgins , 2018 ) . The alignments were trimmed with TrimAl ( Capella-Gutiérrez et al . , 2009 ) and subsequently used to infer evolutionary relationships with maximum likelihood method in IQ-TREE v1 . 6 . 12 ( Nguyen et al . , 2015 ) . ModelFinder ( Kalyaanamoorthy et al . , 2017 ) implemented in IQ-TREE was used to find the best-fit substitution model based on the Bayesian information criterion ( BIC ) for each protein alignment . The branch supports for ML analyses were obtained using the ultrafast bootstrap ( UBS ) ( Minh et al . , 2013 ) with 1000 replicates . Phylogenetic trees ( Figure 3—figure supplement 1–4 ) were prepared using iTOL v5 ( Letunic and Bork , 2019 ) and bootstrap values below or equal to 75% are shown . The following amino acid sequences were retrieved from GenBank for inclusion in our phylogenetic analyses: Mouse ( Mus musculus ) Sox5 , XP_006506994; human ( Homo sapiens ) Sox5 , NP_008871; chick ( Gallus gallus ) Sox5 , XP_015145677; rat ( Rattus norvegicus ) Sox5 , XP_006237666; zebrafish ( Danio rerio ) Sox5 , XP_021330769; cow ( Bos taurus ) Sox5 , XP_005207008; frog ( Xenopus tropicalis ) Sox5 , XP_031753364; dog ( Canis lupus familiaris ) Sox5 , XP_022267096; elephant fish ( Callorhinchus milii ) Sox5 , XP_007895487; mouse ( Mus musculus ) Sox6 , XP_006507558; human ( Homo sapiens ) Sox6 , NP_001354802; rat ( Rattus norvegicus ) Sox6 , XP_006230149; zebrafish ( Danio rerio ) Sox6 , NP_001116481; chick ( Gallus gallus ) Sox6 , XP_025006442; frog ( Xenopus tropicalis ) Sox6 , XP_031755685; cow ( Bos taurus ) Sox6 , XP_024831142; dog ( Canis lupus familiaris ) Sox6 , XP_022263574; elephant fish ( Callorhinchus milii ) Sox6 , XP_007885710; mouse ( Mus musculus ) Sox8 , NP_035577; human ( Homo sapiens ) Sox8 , NP_055402; rat ( Rattus norvegicus ) Sox8 , NP_001100459; chick ( Gallus gallus ) Sox8 , NP_990062; cow ( Bos taurus ) Sox8 , XP_002698019; frog ( Xenopus tropicalis ) Sox8 , XP_002932315; dog ( Canis lupus familiaris ) Sox8 , XP_022275986; horse ( Equus caballus ) Sox8 , XP_005599176; elephant fish ( Callorhinchus milii ) Sox8 , XP_007901694; Mouse ( Mus musculus ) Sox9 , NP_035578; human ( Homo sapiens ) Sox9 , NP_000337; chick ( Gallus gallus ) Sox9 , NP_989612; rat ( Rattus norvegicus ) Sox9 , NP_536328; zebrafish ( Danio rerio ) Sox9 , NP_571718; cow ( Bos taurus ) Sox9 , XP_024836864; frog ( Xenopus tropicalis ) Sox9 , NP_001016853; dog ( Canis lupus familiaris ) Sox9 , NP_001002978; horse ( Equus caballus ) Sox9 , XP_023507898; mouse ( Mus musculus ) Sox10 , NP_035567; human ( Homo sapiens ) Sox10 , NP_008872; chick ( Gallus gallus ) Sox9 , XP_015139949; rat ( Rattus norvegicus ) Sox10 , NP_062066; zebrafish ( Danio rerio ) Sox10 , NP_571950; cow ( Bos taurus ) Sox10 , NP_001180176; frog ( Xenopus tropicalis ) Sox10 , NP_001093691; dog ( Canis lupus familiaris ) Sox10 , XP_538379; horse ( Equus caballus ) Sox10 , XP_023487097; fruit fly ( Drosophila melanogaster ) Sox100B , NP_651839; fruit fly ( Drosophila melanogaster ) Sox102F , NP_001014695; mouse ( Mus musculus domesticus ) aggrecan , AAC37670; human ( Homo sapiens ) aggrecan , AAH36445; cow ( Bos taurus ) aggrecan , AAP44494; rat ( Rattus norvegicus ) aggrecan , AAA21000; zebrafish ( Danio rerio ) aggrecan , XP_021326217; elephant fish ( Callorhinchus milii ) aggrecan , XP_007906559; frog ( Xenopus tropicalis ) aggrecan , XP_017948155; horse ( Equus caballus ) aggrecan , XP_005602856; mouse ( Mus musculus domesticus ) Col2α1 , NP_112440; rat ( Rattus norvegicus ) Col2α1 , XP_006242370; human ( Homo sapiens ) Col2α1 , NP_001835; cow ( Bos taurus ) Col2α1 , NP_001001135; chick ( Gallus gallus ) Col2α1 , XP_025001042; zebrafish ( Danio rerio ) Col2α1a , NM_131292; dog ( Canis lupus familiaris ) Col2α1 , NP_001006952; horse Col2α1 ( Equus caballus ) , XP_005611139; frog ( Xenopus tropicalis ) Col2α1 , NP_989220; elephant fish ( Callorhinchus milii ) Col2α1 , XP_007908719; sea lamprey ( Petromyzon marinus ) Col2α1a , ABB53637; sea lamprey ( Petromyzon marinus ) Col2α1b , ABB53638; mouse ( Mus musculus domesticus ) Col1α1 , CAI25880; human ( Homo sapiens ) Col1α1 , BAD92834; zebrafish ( Danio rerio ) Col1α1 , AAH63249; dog ( Canis lupus familiaris ) Col1α1 , NP_001003090; cow ( Bos taurus ) Col1α1 , AAI05185; frog ( Lithobates catesbeianus ) Col1α1 , BAA29028; mouse ( Mus musculus domesticus ) Col1α2 , NP_031769; human ( Homo sapiens ) Col1α2 , AAH42586; chick ( Gallus gallus ) Col1α2 , XP_418665; dog ( Canis lupus familiaris ) Col1α2 , NP_001003187; frog ( Xenopus laevis ) Col1α2 , AAH49287; zebrafish ( Danio rerio ) Col1α1 , NP_892013; human ( Homo sapiens ) Col3α1 , AAL13167; dog ( Canis lupus familiaris ) Col3α1 , XP_851009; frog ( Xenopus laevis ) Col3α1 , AAH60753; cow ( Bos taurus ) Col3α1 , NP_001070299; mouse ( Mus musculus domesticus ) Col5α2 , NP_031763; human ( Homo sapiens ) Col5α2 , NP_000384; chick ( Gallus gallus ) Col5α2 , XP_015144688; dog ( Canis lupus familiaris ) Col5α2 , XP_535998; cow ( Bos taurus ) Col5α2 , XP_581318; rat ( Rattus norvegicus ) Col5α2 , NP_445940; sea urchin ( Strongylocentrotus purpuratus ) ColP2α , NP_999675; tunicate ( Ciona intestinalis ) fCol1 , XP_026690723; acorn worm ( Saccoglossus kowalevskii ) fibrillar collagen , ABB83364 . The sea lamprey aggrecan-like protein sequence ( ENSPMAP00000001826 ) was retrieved from P . marinus Ensembl genome assembly ( Pmarinus_7 . 0 ) based on BLAST searches for sequence conservation . Similarly , tunicate clade A fibrillar collagen gene ( ci0100150759 ) was retrieved from JGI C . intestinalis genome assembly ( C . intestinalis V2 . 0 ) . The Leucoraja erinacea Sox5 sequence corresponds with the 3’ untranslated region ( UTR ) of the Sox5 transcript , and so orthology could not be confirmed as described above . However , this sequence showed significant homology with the 3’ UTR of predicted Sox5 transcripts from other chondrichthyan species ( Rhincodon typus , Amblyraja radiata and Callorhinchus milii ) by BLAST . For cartilage injury experiments , adult skates ranging in weight from 500 to 750 g were anaesthetized in MS-222 in aerated seawater ( 150 mg/L MS-222 buffered with 300 mg/L sodium bicarbonate ) until the animals failed to respond to noxious stimulus ( e . g . a pinch with forceps ) . Anaesthetized animals were given a pre-operative analgesic ( 0 . 25 mg butorphanol by intramuscular injection ) and then moved from the anesthesia bath to an operating table , where their gills were perfused with anesthetic seawater for the duration of the procedure ( ~5 min ) . A small ( ~2 cm ) surgical incision was made through the dorsal surface of the fin , ~3 cm from the base of the metapterygium , and a wedge of cartilage was removed from the metapterygium using a 4 mm biopsy punch . Following biopsy , the incision was sutured , animals were given a postoperative dose of antibiotic ( 15 mg ceftazidime by intramuscular injection ) and then recovered in aerated seawater until fully awake before returning to their holding tank . No animals died as a result of the procedure . Two animals were collected one-week post-operation , and then at monthly intervals for the following year . All animals were euthanized , dissected , fixed and processed for histological analysis as described above . Tissue samples from adult skates collected one-week post-biopsy were imaged by X-ray microtomography ( microCT ) at the Cambridge Biotomography Centre ( Department of Zoology , University of Cambridge ) . Samples were scanned using a Nikon XTH225 ST scanner , at 100kV and a 120microamps beam current . All microscopy was performed with a Zeiss Axioscope A1 and Zen software . All images were processed and plates prepared using Adobe Photoshop CC and Adobe Illustrator CC . | For our joints to move around freely , they are lubricated with cartilage . In growing mammals , this tissue is continuously made by the body . But , by adulthood , this cartilage will have been almost entirely replaced by bone . It is also difficult for adult bodies to replenish what cartilage does remain – such as that in the joints . When growing new cartilage , the body uses so-called progenitor cells , which have the ability to turn into different cell types . Progenitor cells are recruited to the joints , where they transform into cartilage cells called chondrocytes , which generate new cartilage . But adults lack these progenitor cells , leaving them unfit to heal damaged cartilage after injury or diseases like osteoarthritis . In contrast , certain groups of fishes , such as skates , sharks and rays , produce cartilage throughout their life — indeed their whole skeleton is made of cartilage . So , what is the difference between these cartilaginous fishes and mammals ? Why can they generate cartilage throughout their lives , while humans are unable to ? And does this mean that these adult fish are better at healing injured cartilage ? Marconi et al . used skates ( Leucoraja erinacea ) to study how cartilage develops , grows and heals in a cartilaginous fish . Progenitor cells were found in a layer that wraps around the cartilage skeleton ( called the perichondrium ) . These cells were also shown to activate genes that control cartilage development . By labelling these progenitor cells , their presence and movements could be tracked around the fish . Marconi et al . found progenitor cells in adult skates that were able to generate chondrocytes . Skates were also shown to spontaneously repair damaged cartilage in experiments where cartilage was injured . Marconi et al . have identified the skate as a new animal model for studying cartilage growth and repair . Studying the mechanisms that skate progenitor cells use for generating cartilage could lead to improvements in current therapies used for repairing cartilage in the joints . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"developmental",
"biology"
] | 2020 | Adult chondrogenesis and spontaneous cartilage repair in the skate, Leucoraja erinacea |
The organization of DNA into chromatin is dynamic; nucleosomes are frequently displaced to facilitate the ability of regulatory proteins to access specific DNA elements . To gain insight into nucleosome dynamics , and to follow how dynamics change during differentiation , we used a technique called time-ChIP to quantitatively assess histone H3 . 3 turnover genome-wide during differentiation of mouse ESCs . We found that , without prior assumptions , high turnover could be used to identify regions involved in gene regulation . High turnover was seen at enhancers , as observed previously , with particularly high turnover at super-enhancers . In contrast , regions associated with the repressive Polycomb-Group showed low turnover in ESCs . Turnover correlated with DNA accessibility . Upon differentiation , numerous changes in H3 . 3 turnover rates were observed , the majority of which occurred at enhancers . Thus , time-ChIP measurement of histone turnover shows that active enhancers are unusually dynamic in ESCs and changes in highly dynamic nucleosomes predominate at enhancers during differentiation .
The organization of the genome into chromatin acts as a mechanism for regulating access to the information encoded in DNA . The simplest repeating unit of chromatin is the nucleosome consisting of a histone octamer , made up of two copies each of H2A , H2B , H3 and H4 , around which approximately 147 bp of DNA is wrapped . Variants of the histone proteins exist that have distinct incorporation dynamics and functions . These include a number of histone H2A variants such as H2A . Z , H2A . X and macroH2A and histone H3 variants such as H3 . 3 and CENP-A ( Maze et al . , 2014; Skene and Henikoff , 2013 ) . As nucleosomes can act as barriers to reading the DNA sequence , they must be disrupted in order for many processes requiring DNA access to occur . Transcription and the binding of proteins to regulatory regions are known to be dependent upon increased accessibility to nucleosomal DNA . Therefore , assessment of nucleosome turnover can provide important information concerning which genomic regions are involved in regulation and the extent to which nucleosome dynamics contribute to their function . Nucleosome dynamics is a generally underexplored property of chromatin . Previous studies examining nucleosome turnover genome-wide have made use of two main strategies – induction of epitope-tagged histone transgenes ( Kraushaar et al . , 2013; Yildirim et al . , 2014 ) and metabolic labeling of histone proteins ( Deal et al . , 2010 ) . These have yielded a number of insights into the genome-wide pattern of nucleosome turnover in Drosophila and mammalian systems . Such insights include high nucleosome turnover at gene bodies , regulatory elements and replication origins in Drosophila cells ( Deal et al . , 2010 ) and , in mouse cells , rapid deposition of H3 . 3-containing nucleosomes at enhancers and promoters associated with active histone modification marks and the histone variant H2A . Z ( Kraushaar et al . , 2013 ) . However , each set of techniques has limitations . For example , systems based on inducible histones have a lag time after induction before histones are synthesized and incorporated into chromatin . In addition , they mainly look at where newly synthesized histones are deposited rather than disruption of existing histones . A TET-OFF system examining dissociation of tagged histones has recently been described although a significant proportion of tagged histone is observed up to 12 hr after addition of doxycycline , limiting the sensitivity of this technique ( Ha et al . , 2014 ) . Metabolic labeling methods such as 'CATCH-IT' , on the other hand , are not specific for particular histone variants and have limited resolution at the level of individual genes . CATCH-IT analyzes incorporation rates rather than existing histones and cannot be employed for long chase times . To circumvent these limitations , we have employed a newly developed method called time-ChIP ( Gómez-Rodríguez et al . , submitted ) to assess histone turnover genome-wide during differentiation of mouse embryonic stem cells ( ESCs ) to neural stem cells ( NSCs ) . Time-ChIP is based on SNAP technology ( Bodor et al . , 2012; Keppler et al . , 2003 ) , where histones are tagged with the suicide enzyme SNAP . Pulse labeling with a SNAP-tag specific biotin moiety leads to covalent biotinylation of the SNAP-tagged histone pool . Histone dynamics can then be determined by following the loss of biotin-labeled histones over time as nucleosomes are disrupted and replaced with nucleosomes comprised of newly synthesized unlabeled histones . Sequencing of the DNA associated with these biotinylated histones allowed us to quantify relative histone turnover genome-wide . Previous studies using SNAP-tagged histones suggest that they are recognized by the correct chaperone proteins and incorporated into chromatin at the appropriate cell cycle stage ( Bodor et al . , 2013; Dunleavy et al . , 2009; Foltz et al . , 2009; Jansen et al . , 2007; Ray-Gallet et al . , 2011 ) . Time-ChIP allows labeling of existing , incorporated histones , is specific for particular histone variants and allows us to study histone dynamics in vivo . We focused on measuring turnover of the histone variant H3 . 3 . In contrast to canonical H3 , histone H3 . 3 is deposited independently of replication at places where nucleosomes are disrupted such as promoters , the bodies of active genes , and regulatory elements ( Goldberg et al . , 2010; Schwartz and Ahmad , 2005 ) . Therefore examining this variant is likely to yield insights into the dynamics at regions important for gene regulation . To examine H3 . 3 turnover at developmental loci and regulatory elements , and understand how turnover changes during cell specification , we applied time-ChIP to neural differentiation of mouse ESCs . We found that time-ChIP could be used a priori to identify regions important for gene regulation; specifically , enhancer regions stood out for their rapid turnover . In ESCs , consistent with previous work , we found high H3 . 3 turnover at active enhancers ( Ha et al . , 2014; Kraushaar et al . , 2013 ) with even higher rates of turnover observed at super-enhancers . We also observed high turnover of histone H3 . 1 at these regions . Regions bound by Polycomb-group proteins , which are involved in developmental gene repression , showed H3 . 3 enrichment in ESCs but low turnover compared to active regions . Upon differentiation to NSCs , many changes in H3 . 3 turnover were observed , with ESC and NSC enhancers overrepresented in the set of regions showing changes . We corroborated our time-ChIP findings by correlating H3 . 3 turnover with DNA accessibility measured by micrococcal nuclease ( MNase ) titration ( Mieczkowski et al . , 2016 ) and found good correlation at enhancers and other active regions . Our findings demonstrate the utility of time-ChIP for profiling H3 . 3 turnover during differentiation , for determining relative rates of turnover at regulatory elements , and for characterizing known and potentially novel regulatory regions .
We used time-ChIP to assess histone H3 . 3 turnover genome-wide in ESCs in an unbiased manner . To do this , we used a SNAP-tagged histone H3 . 3 which can be labeled with fluorescent or biotin-containing substrates in cell culture ( Bodor et al . , 2013; Ray-Gallet et al . , 2011 ) . We generated ESC lines expressing SNAP-tagged H3 . 3 in single copy from a defined locus . To achieve this we used recombination-mediated cassette exchange to insert H3 . 3 C-terminally tagged with SNAP and HA-tags into A2lox-Cre ESCs which allow for dox inducible transgene expression from the hprt locus ( Iacovino et al . , 2011 ) ( Figure 1—figure supplement 1 ) . Western blotting showed that H3 . 3_SNAP was expressed at low levels relative to total endogenous H3 ( Figure 1A ) and undetectable when blotting was performed using an H3 . 3-specific antibody ( Figure 1—figure supplement 1 ) . This was desirable , as we wanted H3 . 3_SNAP to act as a tracer to report the turnover of all H3 . 3 rather than overwhelm the cells with high levels of transgenic histone . We also generated a line expressing low levels of SNAP-tagged histone H3 . 1 ( Figure 1—figure supplement 2 ) . Post-translational modifications were detected on the SNAP-tagged histones ( Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 15316 . 003Figure 1 . Time-ChIP assay in H3 . 3_SNAP ESCs . ( A ) Western blot for total H3 in H3 . 3_SNAP ESCs after 48 hr dox induction of transgene . ( B ) Outline of time-ChIP protocol . ( C ) Quantification of DNA recovered from time-ChIP experiments . The amount of DNA relative to the 0 hr sample ( expressed as % ) was averaged over two H3 . 3 time-ChIP replicates . Error bars are ± standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 00310 . 7554/eLife . 15316 . 004Figure 1—figure supplement 1 . Time-ChIP assay in H3 . 3_SNAP ESCs . ( A ) Generation of H3 . 3_SNAP ESCs . The p2lox H3 . 3_SNAP plasmid was introduced into A2lox . Cre ESCs resulting in H3 . 3_SNAP ESCs expressing H3 . 3_SNAP_HA under dox inducible control from the hprt locus along with a neomycin resistance gene . These cells contain rtTA expressed from the Rosa26 locus to mediate dox inducible expression . ( B ) Western blot for H3 . 3 in H3 . 3_SNAP ESCs . For dox induced samples ( 'dox' ) , increasing amounts of protein were loaded , for the uninduced sample ( 'un' ) , only the highest protein concentration was loaded . Non-specific bands present in both the uninduced and dox induced samples are marked as 'NS' . Endogenous H3 serves as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 00410 . 7554/eLife . 15316 . 005Figure 1—figure supplement 2 . Post-translational modification of SNAP-tagged histones . ( A ) Western blot for H3K4me3 H3 . 3_SNAP ESCs . Uninduced sample ( 'un' ) does not have any H3 . 3_SNAP , upon dox induction ( 'dox' ) H3K4me3 modified H3 . 3_SNAP is observed . Non-specific bands present in both the uninduced and dox induced samples are marked as 'NS' . Histone H3 serves as a loading control . ( B ) Western blot for H3K27me3 in H3 . 1_SNAP and H3 . 3_SNAP ESCs . Uninduced sample ( 'un' ) does not have any H3 . 1_SNAP or H3 . 3_SNAP , upon dox induction ( 'dox' ) H3K27me3 modified H3 . 1_SNAP is observed . Non-specific bands present in both the uninduced and dox induced samples are marked as 'NS' . Histone H3 serves as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 00510 . 7554/eLife . 15316 . 006Figure 1—figure supplement 3 . Spike-in control . Amount of DNA recovered relative to 0 hr time point ( red bars ) and proportion of sequencing reads mapping to the mouse genome relative to 0 hr time point based on the number of reads recovered from the spike-in control ( blue bars ) . See 'Materials and methods' for more details . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 006 ESCs expressing SNAP-tagged H3 . 3 were used to perform time-ChIP and assess genome-wide H3 . 3 turnover at high resolution . After inducing tagged H3 . 3 for 48 hr , H3 . 3_SNAP was biotin labeled in intact cells by adding the cell permeable molecule chloropyrimidine biotin ( CP-biotin ) ( Correa et al . , 2013 ) to the media for 40 min followed by a wash step to remove unbound substrate . A portion of the labeled cells was harvested immediately representing the 0 hr time point and the remaining cells were re-plated for harvest at 3 hr , 6 hr and 12 hr post-labeling . Nuclei were isolated from these cells , chromatin liberated by MNase digestion and biotin labeled H3 . 3 recovered by streptavidin pull-down . DNA was purified from the recovered nucleosomes , quantified and prepared for sequencing ( Figure 1B ) . As expected , decreasing amounts of DNA were recovered over time as biotin labeled H3 . 3_SNAP was replaced by newly synthesized H3 ( Figure 1C ) . Approximately the same number of reads was sequenced from each time point . A spike-in control was added when preparing the sequencing libraries to verify that the proportion of reads resulting from pull-down of SNAP-tagged histones was consistent with expectation for each time point . Later time points in each experiment had fewer fragments associated with tagged histones compared to spike-in fragments and thus were consistent with the expected overall H3 . 3 turnover rate in the genome ( Figure 1—figure supplement 3 and Materials and methods ) . To validate that the SNAP tagged histone showed appropriate localization , we analyzed the 0 hr sample . As expected , H3 . 3 was enriched at enhancers and the promoters and bodies of active genes . H3 . 3 also showed high enrichment at transcription termination sites ( TTSs ) ( Figure 2—figure supplement 1 ) . To ensure that the biotin labeling of H3 . 3 was representative of all tagged H3 . 3 , we compared the enrichment profiles for the H3 . 3 0 hr sample to those generated by performing HA ChIP for transgenic H3 . 3_SNAP_HA and the two correlated well ( Figure 2—figure supplement 1 ) . To further validate our approach we examined the transcription start sites ( TSSs ) of active genes as these are elements where one would expect high H3 . 3 turnover . We plotted H3 . 3 time-ChIP data for the 0 hr , 3 hr , 6 hr and 12 hr time points over the transcription start sites ( TSSs ) of active genes ( Figure 2A ) . As MNase was used to digest the chromatin , a characteristic MNase digestion pattern was observed consisting of a nucleosome 'depleted' region at the TSS and a well-positioned +1 nucleosome . As time progressed , labeled H3 . 3 was lost from these TSSs at an appreciable rate , particularly upstream of the TSS , as indicated by the fact that the profiles for each time point were well spaced apart . In contrast , at the TSSs of silent genes , H3 . 3 was less enriched and the profiles for each time point were closer together indicating slower H3 . 3 turnover compared to active promoters ( Figure 2A ) . Fast H3 . 3 turnover at the TSSs of active genes was evident at the level of individual loci as well as when average behavior was observed ( Figure 2B ) . H3 . 3 enrichment and rapid turnover was also observed at the 3’ ends of expressed genes ( Figure 2B and C ) . These observations are consistent with the displacement of nucleosomes by the transcription machinery and the enrichment of the replacement H3 variant , H3 . 3 , at sites of active transcription as reported previously ( Goldberg et al . , 2010 ) . The timeframe for turnover of H3 . 3_SNAP is concordant with previous estimates based on incorporation of H3 . 3 ( Kraushaar et al . , 2013 ) . 10 . 7554/eLife . 15316 . 007Figure 2 . Measuring H3 . 3 turnover in ESCs . ( A ) Average H3 . 3 profiles for 0 hr , 3 hr , 6 hr and 12 hr time points over the transcription start sites ( TSSs ) of expressed and silent genes in ESCs . ( B ) H3 . 3 in 1kb bins profiles for 0 hr , 3 hr , 6 hr and 12 hr time points at the actively transcribed Nanog locus . ( C ) Average H3 . 3 profiles over the transcription termination sites ( TTSs ) of genes expressed in ESCs . ( D ) Calculating turnover index ( TI ) in 1 kb bins across the genome . H3 . 3 signal for each bin was plotted for every time point , these were fitted to a linear regression model and the slope of this line was weighted and multiplied by minus 1 to get TI . ( E ) Average TI over the TSSs of expressed genes . For ( A–C ) results from one experiment representative of three biological replicates is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 00710 . 7554/eLife . 15316 . 008Figure 2—figure supplement 1 . H3 . 3 enrichment in ESCs and TI correlation between replicates . ( A ) H3 . 3 enrichment in ESCs at annotated genomic elements – promoters , 5’ end of genes , gene bodies , TTSs and enhancers . Genes are grouped by ESC expression level and ESC enhancers are from Whyte et al . ( 2013 ) . ( B ) Scatter plot of H3 . 3 enrichment calculated in 1 kb bins for ESC H3 . 3_HA ChIP and H3 . 3 0 hr time-ChIP samples . R value is from Pearson’s correlation . ( C ) Scatter plots comparing TI calculated for the three ESC time-ChIP replicates . R value is from Pearson’s correlation . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 008 Our data allows assessment of the relative H3 . 3 turnover between different genomic sites . Regions of high turnover will have fewer reads in later time points than regions with slower turnover when compared to the number of reads seen for these regions at 0 hr . To examine how turnover compared between different regions , it was necessary to develop a method to quantify turnover rate . To do this we used data from the 0 hr , 3 hr , 6 hr and 12 hr time points to calculate a 'turnover index' . The genome was divided into 1 kb bins and , for each bin , signal from each time point was fitted to a simple linear regression model . The slope of the linear regression was weighted to account for significance of the data fit by the linear model and multiplied by minus 1 ( Figure 2D and Materials and methods ) . The resulting value was taken as the turnover index ( TI ) such that regions with fast H3 . 3 turnover had higher positive TI values . Indeed , at the TSSs of active genes , TI was highest upstream of the TSS reflecting what we observed when individual time points were examined ( Figure 2E ) . We calculated TI for three individual biological replicates and these samples correlated well with each other having R values of 0 . 67–0 . 82 ( Figure 2—figure supplement 1 ) . Therefore , for subsequent analyses , we used the mean TI of the three replicates as the TI measurement . This TI reflects relative H3 . 3 turnover between different genomic sites and , by its nature , is internally comparable across the genome . Having established time-ChIP as a method for assessing H3 . 3 turnover and shown that it offers reproducible measurements genome-wide we then went on to examine H3 . 3 turnover in ESCs in more depth . We were interested in how turnover relates to regulation , so , without prior assumptions , we used TI to pick out the 1000 1 kb bins with the highest H3 . 3 turnover in ESCs ( False Discovery Rate ( FDR ) = 0 . 003 ) ( Supplementary file 1 ) . The majority of these ( approximately 77% ) were located in at least one annotated regulatory region such as a promoter , gene body , TTS or enhancer . High TI regions were overrepresented , as compared to the occurrence of the features in the genome , at promoters , the 5’ end of genes , TTSs , and enhancers ( Figure 3A ) . In particular , high H3 . 3 enrichment and TI occurred at TTSs , often at those close to other transcription units , ( Figure 3B ) and at enhancers and super-enhancers ( Figure 3C ) . We examined the high TI regions that did not correspond to annotated genomic features in more detail . Some of these overlapped with a larger UCSC gene set , leaving 146 high TI regions in truly unannotated locations . It is likely that a portion of these represents unidentified regulatory regions or regions with regulatory functions in differentiated cells ( see example in Figure 3D ) . Indeed , 95 of these 146 unannotated regions were located within 50 kb of genes robustly expressed ( FPKM > 9 . 5 ) in ESCs ( p-value = 0 . 001 ) suggesting potential regulatory roles for them in ESCs . Thus , without prior assumptions , time-ChIP identifies known regulatory , and also potential novel regulatory regions . We conclude that time-ChIP can be used to identify regions important for gene regulation . 10 . 7554/eLife . 15316 . 009Figure 3 . Regions with high TI in ESCs have regulatory functions . ( A ) Representation of annotated genomic features in the 1000 highest TI regions in ESCs relative to their representation in the genome as a whole . Overrepresentation was calculated by taking the log2 ratio of the proportion of the high TI regions overlapping a particular feature divided by the proportion of whole genome overlapping the same feature . ( B ) The region 5’ of the Fgf18 gene and 3’ of the neighboring Npm1 gene has high TI in ESCs . Signal for each time point is shown for one replicate ( top 4 tracks – green ) while TI averaged for 3 replicates is shown ( heatmap ) . RNA-seq in ESCs is shown in blue while the highest TI regions are marked with red bars . ( C ) Super-enhancers in the NR_037986 locus are amongst the most highly turned over regions in ESCs . Super-enhancer constituents are as defined by ( Whyte et al . , 2013 ) and shown as blue bars . ( D ) Unannotated regions are also present in the set of highest TI regions . The red bar marks a high TI region downstream of the Gulp1 gene on chr1 . ( E ) ESC TI distribution at annotated genomic elements – promoters , 5’ end of genes , gene bodies , TTSs and enhancers . Genes are grouped by ESC expression level and ESC enhancers are from Whyte et al . ( 2013 ) . ( F ) Average TI over conventional ESC enhancers ( purple ) and super-enhancer constituent enhancers ( pink ) as described in Whyte et al . ( 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 00910 . 7554/eLife . 15316 . 010Figure 3—figure supplement 1 . H3 . 3 turnover at enhancers and super-enhancers calculated using a 300 bp bin size . Average H3 . 3 TI over conventional ESC enhancers ( 'Enhancers' – purple ) and super-enhancer constituent enhancers ( 'Super-enhancers' – pink ) where TI was calculated using a 300 bp bin size . The mean TI for three biological replicates is shown . Enhancers and super-enhancers are as described in Whyte et al . ( 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01010 . 7554/eLife . 15316 . 011Figure 3—figure supplement 2 . H3 . 1 turnover at ESC enhancers and super-enhancers . Average H3 . 1 TI over conventional ESC enhancers ( 'Enhancers' – purple ) , super-enhancer constituent enhancers ( 'Super-enhancers' – pink ) and random genomic regions with a similar size distribution to enhancers ( 'Random' – teal ) . Enhancers and super-enhancers are as described in Whyte et al . 2013 ) . ( A ) H3 . 1 biological replicate 1 ( B ) H3 . 1 biological replicate 2DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 011 If TI changes with the nature of a regulatory region , we anticipated that TI would be high in regions of high expression when these were examined over the entire genome . TI in fact did correlate with gene activity and was higher over the promoters , bodies and TTSs of highly expressed genes compared to genes with little or no expression ( Figure 3E ) . ESC enhancers showed particularly high TI values . This was consistent with their overrepresentation in the set of regions with the highest TI in ESCs , and further indicated that enhancers are especially dynamic . We examined enhancers in more detail , and found that TI distinguishes isolated enhancers from the grouped enhancers that modulate expression of genes involved in cell identity; super-enhancers or locus control regions ( Magram et al . , 1985; Whyte et al . , 2013 ) . Plotting average TI profiles over conventional ESC enhancers and ESC super-enhancer constituents revealed higher H3 . 3 TI over the constituent enhancers that make-up super-enhancers domains compared to conventional enhancers despite the fact that these regions are similar in size ( Figure 3F ) . When TI was calculated using a smaller 300 bp bin size , exceptionally high H3 . 3 turnover was still evident at super-enhancers ( Figure 3—figure supplement 1 ) . We also examined turnover of histone H3 . 1 at ESC enhancers and super-enhancers by performing H3 . 1 time-ChIP in the same manner as was done for H3 . 3 . Although the depth of sequencing required to interrogate H3 . 1 limits the resolution of H3 . 1 time-ChIP , we clearly saw elevated TI over enhancers compared to randomized regions in two seperate biological replicates ( Figure 3—figure supplement 2 ) . In addition , super-enhancer constituents showed higher H3 . 1 turnover than conventional enhancers , consistent with the results for H3 . 3 . The high H3 . 1 and H3 . 3 TI observed at super-enhancers suggests that nucleosomes are more frequently disrupted at these regions regardless of which H3 variant they contain . We conclude that TI offers a metric for identifying enhancers and super-enhancers , and that super-enhancers have an unusually dynamic nucleosome population . Having observed increased TI associated with activation , we were interested in examining regions associated with repression . Heterochromatin , the most extensively repressed segment of the genome , does not have H3 . 3 ( Figure 4—figure supplement 1 ) , indicating a lack of turnover and causing us to be unable to measure H3 . 3 turnover at these regions . However Polycomb-Group ( PcG ) regulated genes are bound by H3 . 3 in ESCs ( Figure 4—figure supplement 1 ) , although H3 . 3 levels at these genes were lower than at enhancers and CpG islands ( Figure 4—figure supplement 1 ) . PcG proteins mediate repression , at least in part , through chromatin compaction ( Simon and Kingston , 2013 ) . One outstanding question is whether this compaction could lead to stabilization of nucleosomes and reduced nucleosome turnover at PcG-bound genes . Measurement of TI at PcG-bound regions revealed slower turnover of H3 . 3 compared to all CpG islands , the majority of which are associated with active chromatin ( Figure 4A ) while PcG-bound TSSs had a TI intermediate between the TSSs of active and silent genes ( Figure 4B ) . The classical PcG-target , the Hoxb locus shows low levels of H3 . 3 throughout the region with average TI values of 0 . 03 ( Figure 4C ) , compared to high levels of H3 . 3 enrichment and an average TI of 0 . 14 in the active Hnrnpa3 locus ( Figure 4D ) . This shows that , in ESCs , PcG-bound regions have decreased TI compared to active regions . 10 . 7554/eLife . 15316 . 012Figure 4 . H3 . 3 turnover at PcG targets in ESCs . ( A ) TI averaged over PcG-bound genomic regions ( turquoise ) and all CpG islands ( red ) . ( B ) TI averaged over the TSSs of active genes ( red ) , PcG targets ( turquoise ) and silent genes ( blue ) in ESCs . ( C ) The PcG-regulated Hoxb locus shows enrichment for H3 . 3 ( top 4 tracks show H3 . 3 at each time point ) but low TI ( heatmap ) . The grey tracks show ChIP-seq for PcG protein Ring1b and the blue tracks RNA-seq . ( D ) The non-PcG-bound active gene Hnrnpa3 shows greater H3 . 3 enrichment and higher TI than Hoxb . TI scale is the same as in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01210 . 7554/eLife . 15316 . 013Figure 4—figure supplement 1 . H3 . 3 enrichment at PcG targets in ESCs . ( A ) H3 . 3 enrichment in ESCs at PcG targets ( grey ) compared to H3K9me3 regions ( green ) . ( B ) H3 . 3 enrichment in ESCs at PcG targets ( blue ) , CpG islands ( CGI – red ) , enhancers ( purple ) and randomized PcG regions ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 013 Overall , these data indicate that histone turnover can relate to known regulatory features . Regions involved in activation have high TI , those involved in repression have low TI . ESCs have been proposed to have a more plastic chromatin structure compared with more differentiated cells ( Bernstein et al . , 2006; Meshorer et al . , 2006 ) . Therefore it was possible that the high turnover we had observed at enhancers and super-enhancers was due to ESC-specific plasticity . We therefore determined whether turnover remained high at regulatory regions and enhancers following cellular differentiation . We generated neural stem cells ( NSCs ) expressing H3 . 3_SNAP by injecting H3 . 3_SNAP ESCs into blastocysts , harvesting E13 . 5 embryos and deriving NSCs from embryonic brain . NSC lines were screened to identify transgenic lines ( Figure 5—figure supplement 1 ) and these lines were drug-selected to obtain pure populations of cells carrying the H3 . 3_SNAP transgene . H3 . 3_SNAP NSCs expressed neuronal markers ( Figure 5—figure supplement 1 ) and , upon dox induction , expressed H3 . 3_SNAP at similar levels to the ESC line ( Figure 5—figure supplement 2 ) . To characterize turnover in neural stem cells , time-ChIP was performed with cells harvested at 0 hr , 3 hr , 6 hr and 12 hr post-labeling . As expected , genes that were activated during neural differentiation and lost PcG binding showed a dramatic increase in H3 . 3 enrichment and TI ( Figure 5A and B ) . Regions still bound by PcG in NSCs were no longer enriched in H3 . 3 as they were in ESCs ( Figure 5C and D ) . This suggests nucleosome stabilization at these regions upon differentiation . ESC-specific enhancers showed reduced H3 . 3 and TI in NSCs compared to ESCs but still showed high H3 . 3 turnover compared to the rest of the genome ( Figure 5E and Figure 5—figure supplement 2 ) . 10 . 7554/eLife . 15316 . 014Figure 5 . H3 . 3 turnover and distribution changes during neural differentiation of ESCs . ( A ) Average TI profile around TSSs of genes that lose PcG binding and increase gene expression upon differentiation to NSCs . ( B ) The region containing the Olig1 and Olig2 genes , which are activated during neural differentiation , shows a dramatic increase in H3 . 3 turnover when ESCs are differentiated to NSCs . The top tracks represent H3 . 3 ( green ) and TI ( heatmap ) in ESCs while the bottom tracks show H3 . 3 ( purple ) and TI in NSCs . The grey tracks show ChIP-seq for PcG protein Ring1b and the blue tracks RNA-seq . ( C ) Average H3 . 3 enrichment ( 0 hr sample ) over PcG-bound regions in ESCs and NSCs . ( D ) Hoxb remains bound by PcG ( Ring1b ) in NSCs but H3 . 3 enrichment is lost . ( E ) Average TI at ESC-specific enhancers in ESCs ( cyan ) and NSCs ( purple ) . ( F ) NSC TI distribution at annotated genomic elements – promoters , 5’ end of genes , gene bodies , TTSs and enhancers . Genes are grouped by NSC expression level and putative NS enhancers were defined using H3K27ac ChIP-seq data from Encode ( E14 . 5 embryonic brain - ENCFF001XZR ) . ( G ) Representation of annotated genomic features in the 1000 highest TI regions in NSCs relative to their representation in the genome as a whole . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01410 . 7554/eLife . 15316 . 015Figure 5—figure supplement 1 . Generation of H3 . 3_SNAP NSCs . ( A ) Screening for chimeric NSC lines expressing H3 . 3_SNAP based on puromycin resistance . Red bars indicate cell lines that scored as highly chimeric . NSC line '3_1' was used for time-ChIP studies . ( B ) RT-qPCR showing expression of NSC markers ( Sox1 , Sox2 , Olig2 and Slc1a3 ) but not the pluripotency marker Oct4 in H3 . 3_SNAP NSCs . RT-qPCR in ESCs is shown for comparision . Expression is relative to the housekeeping gene Eef1A1 . ( C ) Immunofluorescence for NSC markers Sox2 ( red ) and Nestin ( green ) in H3 . 3_SNAP NSCs . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01510 . 7554/eLife . 15316 . 016Figure 5—figure supplement 2 . Measuring turnover in H3 . 3_SNAP NSCs . ( A ) Western blot showing expression of H3 . 3_SNAP upon dox induction in ESCs compared to NSCs . For dox induced samples ( 'dox' ) , increasing amounts of protein were loaded , for uninduced samples ( 'un' ) only the highest protein concentration was loaded . H3 . 3_SNAP is detected using an anti-HA antibody while endogenous H3 and TBP serve as loading controls . ( B ) H3 . 3 enrichment at ESC enhancers is reduced in NSCs ( purple ) compared to ESCs ( cyan ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 016 Similar to what we observed in ESCs , the promoters , gene bodies and TTSs of highly expressed genes show high TI in NSCs as did putative NSC enhancers ( identified using H3K27ac ChIP-seq data from embryonic brain ) . Average TI values were lower for putative NSC enhancers than those observed for active enhancers in ESCs ( Figure 5F ) . This might reflect inclusion of enhancers expressed in embryonic brain that are not expressed in NSCs . When the 1000 regions showing the highest TI in NSCs ( FDR = 0 . 002 ) were examined ( Supplementary file 2 ) , TTSs and enhancers were the most overrepresented genomic elements , similar to what was observed for the highest TI regions in ESCs . However , gene bodies were overrepresented in the NSC set but not in the set of top TI regions in ESCs ( Figure 5G ) . This might reflect either unusually low turnover at gene bodies in ES cells , or high turnover at these sites in NSCs . We conclude that H3 . 3 turnover is broadly similar in ESCs and NSCs with the exception of PcG-bound regions , which seem to stabilize during differentiation as evidenced by a loss of H3 . 3 . Differences in H3 . 3 TI between pluripotent ESCs and lineage committed NSCs might identify regulatory regions important for differentiation . To explore this , we identified regions where TI changed upon differentiation . We compared TI between ESCs and NSCs for each 1 kb bin in the genome . Bins changing turnover were defined as those with a difference in TI between ESCs and NSCs in the top 2 . 5% with a p-value of ≤0 . 01 . Using these criteria , 34 , 070 1 kb bins were identified which changed TI during differentiation ( Figure 6A and Supplementary file 3 ) with more bins showing increased TI in NSCs compared to ESCs ( 19 , 870 bins compared to 14 , 200 ) . Strikingly , regions showing higher TI in ESCs were enriched for the presence of an ESC enhancer ( Figure 6B ) and , similarly , regions with increased TI in NSCs were enriched for NSC enhancers ( Figure 6C ) . Thus , enhancer regions showed high H3 . 3 turnover in a manner that correlated with cell type-specific enhancer activity . Interestingly , NSC enhancers were also somewhat overrepresented in regions with higher turnover in ESCs ( Figure 6B ) . This could reflect roles for these enhancers in both ESCs and NSCs , or could reflect a ‘poised’ state for these enhancers ( Ferrari et al . , 2014; Rada-Iglesias et al . , 2011 ) . Regions with higher turnover in ESCs were also enriched for promoters and the 5’ end of genes while rshowed a greater enrichment of gene bodies . In both sets , TTSs were overrepresented ( Figure 6B and Figure 6C ) . 10 . 7554/eLife . 15316 . 017Figure 6 . Changes in TI occurring during neural differentiation often occur at enhancers . ( A ) Volcano plot showing differences in TI between ESCs and NSCs . Regions with a TI difference in the top 2 . 5% with a p-value≤0 . 01 were selected as changing during differentiation ( red dots ) . ( B ) Representation of annotated genomic features in regions showing higher TI in ESCs compared to NSCs ( selected as described in A ) . The numbers to the right of the plot indicate the number of regions corresponding to each type of genomic feature . ( C ) Representation of annotated genomic features in regions showing higher TI in NSCs compared to ESCs ( selected as described in A ) . ( D ) An unnanotated region on chromosome 3 , which shows an increase in H3 . 3 and TI upon differentiation of ESCs to NSCs ( asterisk ) and an expression decrease in the nearby Stmn2 gene . The heatmap shows average TI while differences in TI are represented by blue bars . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 017 We observed TI differences in unannotated regions; 8732 of 34 , 070 differences , when regions overlapping UCSC genes were excluded . Such regions could represent unidentified elements with roles in cell identity and differentiation . This is supported by the fact that 1608 of these regions are located within 50 kb of a gene changing expression during differentiation ( example in Figure 6D ) . When we examined only the 1000 largest turnover changes in unannotated regions , 212 regions were close to genes changing expression . In both instances the intersection between unannotated differences and these genes was statistically significant ( p-value = 0 . 001 ) , consistent with the possibility that these are previously unannotated regulatory regions . We conclude that H3 . 3 turnover is regulated during differentiation and that the biggest changes in histone dynamics are found prominently in enhancer regions . Time-ChIP may be useful as a tool for picking out novel enhancers or other regulatory elements involved in cell fate changes . One attractive hypothesis is that histone turnover is directly related to accessibility of DNA sequences; regions of rapid turnover are expected to be more accessible . We wished to test this hypothesis by measuring the accessibility of DNA to enzymatic access genome-wide . DNase hypersensitivity , or the conceptually related technique ATAC-seq , detects regions of open chromatin ( Buenrostro et al . , 2013; John et al . , 2013 ) . We examined H3 . 3 turnover over previously annotated DNase hypersensitive sites in mouse ESCs ( Vierstra et al . , 2014 ) and found that TI was indeed high over these sites in ESCs ( Figure 7—figure supplement 1 ) . TI over ESC DNase sensitive regions was reduced in NSCs consistent with regulatory changes during differentiation ( Figure 7—figure supplement 1 ) . To expand upon this analysis , we used MNase , which is capable of measuring accessibility more broadly than these aforementioned technologies . It is established that nucleosomes have differential accessibility to MNase , with some nucleosomes liberated at low MNase and some requiring high MNase ( Knight et al . , 2014; Xi et al . , 2011 ) . We have recently developed a metric that measures how nucleosomes respond to a titration of MNase . Nucleosomes that show decreased signal as MNase concentration decreases are deemed inaccessible ( more MNase is needed to liberate them ) , while nucleosomes that show increased signal as MNase concentration decreases are deemed accessible ( they are released at low MNase concentration and their associated DNA is over-digested at high concentrations; [Mieczkowski et al . , 2016] ) We therefore expanded upon the relationship between TI and DNase hypersensitity to examine whether this new parameter and H3 . 3 turnover were correlated . ESC and NSC chromatin was digested with a range of concentrations of MNase – 1U , 4U , 16U and 64U . We then calculated the ‘MNase accessibility’ ( MACC ) metric that reflects the change in signal at a particular locus in response to decreasing MNase concentration ( Figure 7—figure supplement 1 ) ( Mieczkowski et al . , 2016 ) . The metric is designed so that positive MACC values reflect regions that are easily digested by MNase and therefore are classed as 'accessible' . MNase titration was performed on two biological replicates of H3 . 3_SNAP ESC and NSC chromatin ( 'a' replicates ) . The same assay was also carried out on different ESC and NSC lines ( 'b' replicates ) thus increasing the number of replicates we could assess and providing greater confidence in our findings . We examined annotated genomic elements to see if turnover index correlated with DNA accessibility as assessed by MACC . At the TSSs of genes that change expression upon neural differentiation , TI and MACC were positively correlated in a cell-type specific manner ( Figure 7—figure supplement 2 ) . Increased turnover and accessibility at genes becoming activated during neural differentiation was evident at neuronal genes such as Olig1 ( Figure 7—figure supplement 2 ) . In ESCs , at PcG-bound regions which are expected to be repressed , we observed slower turnover and also saw less accessibility ( lower MACC ) than enhancers and the TSSs of expressed genes ( Figure 7A ) . These results are consistent with transcriptional repression at PcG-bound regions , although the presence of histone H3 . 3 at these regions implies that some replication-independent histone turnover occurs to deposit this histone . 10 . 7554/eLife . 15316 . 018Figure 7 . H3 . 3 turnover correlates with DNA accessibility at enhancers . ( A ) Boxplot of average ESC TI and MACC at ESC enhancers , the TSSs of expressed genes , and PcG-bound regions . ( B ) Pearson’s correlation between DNA accessibility as measured by MNase titration ( MACC ) and TI at ESC enhancers . For MACC four replicates are shown for each cell type , two for ESC_ and NSC_H3 . 3_SNAP lines ( 'a' replicates ) and two for separate ESC and NSC lines ( 'b' replicates ) . For TI three biological replicates are shown . ( C ) Boxplot of average TI and MACC values at ESC enhancers . ( D ) The Oct4/Pou5f1 super-enhancer shows high MACC ( green ) and TI ( heatmap ) in ESCs but not in NSCs . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01810 . 7554/eLife . 15316 . 019Figure 7—figure supplement 1 . Measuring DNA accessibility using MNase titration . ( A ) Boxplot of average TI over DNase hypersensitive regions in E14 ESCs ( Encode accession ENCSR000CMW ) . TI index was plotted over called DNase peaks ( E14-DS18505 . peaks . fdr0 . 01 ) . ( B ) MNase digestion patterns generated by digestion of ESC and NSC chromatin with varying concentrations of enzyme ( 1U , 4U , 16U , 64U ) . ( C ) Sequencing data from each of the above titration points was used to generate an accessibility score ( MACC ) by fitting these to a linear model . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 01910 . 7554/eLife . 15316 . 020Figure 7—figure supplement 2 . Comparing turnover to DNA accessibility measured by MNase titration . ( A ) Pearson’s correlation between DNA accessibility as measured by MNase titration ( MACC ) and TI at the TSSs of genes differentially expressed in ESCs and NSCs . For MACC four replicates are shown for each cell type , two for ESC_ and NSC_H3 . 3_SNAP lines ( 'a' replicates ) and two for separate ESC and NSC lines ( 'b' replicates ) . Three biological replicates are shown for TI . ( B ) Olig1 is activated during neural differentiation and shows high MACC ( green ) and TI ( heatmap ) in NSCs but not ESCs . RNA-seq is shown in blue . ( C ) Pearson’s correlation between DNA accessibility as measured by MNase titration ( MACC ) and TI at ESC super-enhancers . Details as described in C . ( D ) Boxplot of average TI and MACC values at ESC super-enhancers . DOI: http://dx . doi . org/10 . 7554/eLife . 15316 . 020 We then examined DNA accessibility at ESC enhancers and super-enhancers as these regions show particularly high H3 . 3 turnover . We found that , in ESCs , TI and MACC correlated particularly well at these regions ( Figure 7B and Figure 7—figure supplement 2 ) . When the magnitude of TI and MACC at ESC enhancers was assessed , both had values well above the genome average in ESCs but not in NSCs , consistent with their regulatory role in pluripotent cells ( Figure 7C and Figure 7—figure supplement 2 ) . As well as being evident on a global scale , the relationship between TI and MACC at enhancers was also apparent at the level of individual loci . For example , the Oct4/Pou5f1 super-enhancer shows high TI and high MACC in ESCs while both TI and MACC are decreased upon differentiation to NSCs ( Figure 7D ) . These findings indicate that , at enhancers , increased turnover of nucleosomes might account for increased accessibility to the DNA sequence . This is consistent with increased transcription factor binding at these sites .
Our unbiased assessment of H3 . 3 turnover identified enhancers as having highly dynamic nucleosomes . Notably , the clustered enhancers that make up super-enhancer domains were more rapidly turned over than conventional ESC enhancers . Measurement of H3 . 1 turnover showed the same phenomenon . These observations are consistent with the notion that nucleosomes must be disrupted in order to allow transcription factor binding and are consistent with increased DNA accessibility at enhancers . Notably , our data suggest that nucleosomes are frequently disrupted at enhancers regardless of which H3 variant they contain . Nucleosomes containing H3 . 3 predominate at these regions due to the high levels of nucleosome disruption but both H3 . 1 and H3 . 3 show rapid turnover . A variety of mechanisms might be used to generate high turnover in enhancers . Pioneer transcription factors such as Oct4 and Sox2 occupy these sets of ESC enhancers and super-enhancers and can bind partial motifs on the nucleosome surface ( Soufi et al . , 2015 ) . These factors might contribute to nucleosome turnover , either directly through their binding or by recruiting remodeling complexes that facilitate turnover . Transcription through enhancers might also contribute ( Kim et al . , 2010 ) . It is intriguing that super-enhancers show significantly higher turnover than enhancer elements . This could reflect a synergistic interaction between adjacent enhancer modules , or might indicate that the individual enhancer modules within super-enhancers each have intrinsically high turnover . In either case , the ability of turnover to identify these regions is striking and is likely to be related to their potency . High histone turnover at enhancers was not limited to ESCs , but was also seen when we looked at lineage-committed NSCs , indicating that this property is seen both in pluripotent and in differentiated cell lineages . There were significant changes in where high turnover occurred in NSCs due to the activation of a distinct set of enhancers . In broad classifications , regions with rapid turnover were in general similar between ESCs and NSCs , except that gene bodies were overrepresented in high turnover regions in the differentiated cells but not in ESCs ( Figures 3 and 5 ) . Increased turnover in gene bodies might reflect widespread transcriptional activation and consequent histone turnover in differentiated cells . Alternatively , it might be that pluripotent cells have a plastic chromatin structure at the promoters of genes poised for activation in future lineages , increasing turnover relative to gene bodies at those locations . Many unannotated regions showed differentiation-associated turnover changes , approximately one-quarter of turnover changes occurred outside of genes and enhancers . Many of these regions are likely to represent genomic elements with hitherto unknown roles in cell fate specification . In support of this , over 20% of the unannotated regions with the largest changes in turnover were located close to genes that changed expression during neural differentiation . Nucleosome turnover might be related to accessibility of the underlying sequence as regions with rapid turnover are likely to have DNA sequences that are exposed for protein binding more frequently . To begin to analyze whether this is true , we measured accessibility genome wide using MNase , which is able to probe accessibility everywhere in the genome , and compared the resultant accessibility metric to turnover index . At ESC enhancers , both TI and MACC had high values and the two metrics were highly correlated . Similarly at the TSSs of expressed genes , TI and MACC had higher than average values while at repressed PcG targets both turnover and accessibility were reduced . These data thus cross validate these two distinct metrics and support the hypothesis that regions of high turnover correspond to regions of high DNA accessibility . Our work and previous work are consistent in showing high incorporation rates for histone H3 . 3 and high turnover for that histone in enhancers and other regulatory regions ( Ha et al . , 2014; Kraushaar et al . , 2013 ) . We have extended these observations by also showing high turnover of H3 . 1 at enhancer regions and in demonstrating more rapid turnover of both H3 . 3 and H3 . 1 at super-enhancers compared to conventional enhancers . Notably , we also observed high H3 . 3 turnover around the transcription termination sites of expressed genes with our sets containing the ( 1000 ) highest TI regions showing a large overrepresentation of TTSs . RNA Polymerase II occupancy or transcription itself could contribute to nucleosome displacement at these regions . Indeed , enrichment of RNA Pol II and nascent transcription at the 3’ end of genes has been reported to be widespread ( Anamika et al . , 2012; Core et al . , 2008; Rahl et al . , 2010 ) . Time-ChIP allows examination of relative histone turnover in a genome-wide manner at high resolution , with changes in turnover evident at individual gene loci and regulatory elements . Labeling of SNAP-tagged histones for time-ChIP takes approximately one hour meaning that very rapid turnover might be missed and consequently turnover underestimated for some regions . Nevertheless , labeling time is relatively short compared to the cell cycle time of ESCs ( 12 hr ) and NSCs ( 18 hr ) and most H3 . 3 enriched loci are still detectable at the first time point as evidenced by the strong correlation between ChIP for H3 . 3_SNAP_HA and the H3 . 3 0 hr time-ChIP sample ( Figure 2—figure supplement 1 ) . We are therefore confident that we accurately quantify relative turnover for the vast majority of H3 . 3-enriched regions . In this study , H3 . 3 turnover was assessed because of its enrichment at regions important for gene regulation . In theory , however , time-ChIP can be applied to study the dynamics of any core histone or histone variant . We performed time-ChIP for H3 . 1 but a comprehensive assessment of its turnover , and that of other broadly distributed histones , is currently limited by the depth of sequencing required . Measuring nucleosome dynamics provides a distinct measure of chromatin characteristics that can help identify and functionally characterize regulatory elements . | In animal , plant and other eukaryotic cells , DNA wraps around histone proteins to form structures called nucleosomes . This compacts long strands of DNA to fit them inside a cell . However , nucleosomes also act as barriers that can prevent access to the DNA . This affects the activity , or “expression” , of genes because gene expression requires proteins called transcription factors to bind to specific DNA regions . Therefore , nucleosomes must be disrupted or removed in order to access their DNA and allow their genes to be expressed . Transcription factors can bind to DNA sequences called enhancers to activate nearby genes . Groups of enhancers , called super-enhancers , also exist to further bolster the activity of certain genes , particularly those involved in determining cell identity . Recent work has shown that nucleosomes are frequently lost and then replaced by new ones ( in a process referred to as turnover ) in DNA regions that include enhancers . Measuring the rate of turnover of nucleosomes can thus provide information about which DNA regions regulate gene expression . Embryonic stem cells can transform or “differentiate” into any type of cell in the body . During this transformation process , different genes are switched on or off in the cell in order to give it a new identity . It is not known how nucleosome turnover changes when this happens . Deaton et al . have now developed a new method called time-ChIP that can measure the rate of nucleosome turnover across the entire DNA of a cell . Using this technique to analyze mouse embryonic stem cells revealed that nucleosome turnover occurs rapidly at enhancers . Furthermore , nucleosomes at super-enhancers are particularly dynamic and turn over more quickly than in any other DNA region . Deaton et al . next analyzed how turnover changes after the mouse embryonic stem cells have developed into neural stem cells . This revealed that the regions of DNA where high turnover occurs change as the cells differentiate , in part because this transformation activates a different set of enhancers . However , the most rapid turnover still takes place at enhancers . Overall , these observations suggest that the high rate of nucleosome turnover at enhancers makes DNA accessible to transcription factors . The next step is to use the new time-ChIP method to study how nucleosome turnover changes during the processes that pattern gene expression as an animal develops from an embryo . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2016 | Enhancer regions show high histone H3.3 turnover that changes during differentiation |
Mutating RNA virus genomes to alter codon pair ( CP ) frequencies and reduce translation efficiency has been advocated as a method to generate safe , attenuated virus vaccines . However , selection for disfavoured CPs leads to unintended increases in CpG and UpA dinucleotide frequencies that also attenuate replication . We designed and phenotypically characterised mutants of the picornavirus , echovirus 7 , in which these parameters were independently varied to determine which most influenced virus replication . CpG and UpA dinucleotide frequencies primarily influenced virus replication ability while no fitness differences were observed between mutants with different CP usage where dinucleotide frequencies were kept constant . Contrastingly , translation efficiency was unaffected by either CP usage or dinucleotide frequencies . This mechanistic insight is critical for future rational design of live virus vaccines and their safety evaluation; attenuation is mediated through enhanced innate immune responses to viruses with elevated CpG/UpA dinucleotide frequencies rather the viruses themselves being intrinsically defective .
Protein encoding regions of all organisms , eukaryotic , bacterial and viral , are subject to a number of functional constraints in addition to coding capacity , many of which contribute to regulation of translation . These include the widely reported biases in the relative frequencies of codons encoding the same amino acid ( Bennetzen and Hall , 1982; Sharp et al . , 2005; Wu et al . , 2010 ) which in some organisms represents optimisation of the coding sequence for specific tRNAs , elongation rates and translation accuracy ( reviewed in Gingold and Pilpel , 2011 ) . There are , in addition , consistent under- and over-representations of codon pairs ( CPs ) in all organisms ( Yarus and Folley , 1985; Gutman and Hatfield , 1989; Boycheva et al . , 2003; Moura et al . , 2005; Tats et al . , 2008 ) that have been proposed to influence gene expression through alterations in translation efficiency . Because of its potential effect on gene expression , altering CP frequencies towards those that are disfavoured in their hosts has recently been advocated as a novel strategy to reduce RNA virus replication ( Coleman et al . , 2008; Wimmer et al . , 2009; Mueller et al . , 2010; Martrus et al . , 2013; Yang et al . , 2013; Le Nouen et al . , 2014; Ni et al . , 2014 ) . This procedure potentially provides the means to produce a new generation of safer , non-reverting , live attenuated vaccines . Classically , virus genomes have been empirically attenuated by serial passage in tissue-culture leading to the accumulation of mutations . This lengthy , stochastic process produced attenuated virus vaccine strains which have produced major effects on human ( eg . , poliovirus—Sabin vaccines ) and animal health ( eg . , eradication of Rinderpest using the Plowright vaccine ) . However , reversion to virulence by back-mutation of characteristically a small number of key , attenuating , mutations is a well-known problem . Novel strategies by which synonymous coding changes are introduced to modify codon usage ( Mueller et al . , 2010 ) ( Martrus et al . , 2013; Yang et al . , 2013; Le Nouen et al . , 2014; Ni et al . , 2014 ) have the advantage that the resulting virus attenuation is dependent on a large number of mutations each of which only slightly reduce replicative fitness , but taken together produce significant attenuation with greatly enhanced genetic stability . As one of the first examples , Coleman et al . ( 2008 ) generated synthetic poliovirus capsid gene sequences containing codon pairs that were specifically disfavoured in human coding sequences . These CP de-optimised sequences were inserted into an infectious cDNA clone of poliovirus . Virus generated from these mutants showed a remarkably attenuated replication phenotype attributed by the authors to impaired translation efficiency . Codon pair de-optimisation ( CPD ) has since been developed as a strategy for the production of a wide range of other live attenuated virus vaccines including influenza A virus ( IAV ) , porcine reproductive and respiratory syndrome virus ( PRRSV ) , human immunodeficiency virus type 1 ( HIV-1 ) and respiratory syncytial virus ( Mueller et al . , 2010; Martrus et al . , 2013; Yang et al . , 2013; Le Nouen et al . , 2014; Ni et al . , 2014 ) . While altering translation efficiency through manipulation of codon or codon pair usage may attenuate virus replication , other virus compositional features may additionally contribute to replication phenotypes . One prominent compositional abnormality among RNA and small DNA viruses infecting mammals and plants is the marked suppression of the frequencies of CpG and UpA dinucleotides ( Karlin et al . , 1994; Rima and McFerran , 1997; Simmonds et al . , 2013 ) . The functional basis for this suppression was recently demonstrated by the marked attenuating effect of artificially increasing the numbers of CpG and UpA dinucleotides in the genome of echovirus 7 ( E7; ( Atkinson et al . , 2014 ) ) . We and others ( Burns et al . , 2009 ) have speculated that effects of CpG/UpA frequencies on virus replication may indeed account for , at least in part , the attenuating effect of selecting disfavoured codon pairs in CPD mutant of poliovirus and other candidate attenuated vaccines . Supporting this conjecture , regression analysis of the effects of numerous compositional variables in a range of codon- and codon pair deoptimised mutants on poliovirus replication demonstrated the primary effect of CpG and UpA frequencies on replication ability rather than alterations in codon or codon pair usage ( Burns et al . , 2009 ) . In the current study , we have used a variety of bioinformatic analyses to investigate the relationship between dinucleotide frequencies and codon pair usage . We have subsequently designed and assessed the replication phenotypes and fitness of mutants of E7 constructed in such a way that allows effects of codon pair and dinucleotide frequency alterations to be separately altered . The findings demonstrate the primary influence of CpG and UpA frequencies on virus replication that was independent of codon pair usage and translation efficiency .
Coding regions of poliovirus , IAV , PRRSV and HIV-1 have all been subjected to CP de-optimisation and effects on virus replication quantified ( Coleman et al . , 2008; Mueller et al . , 2010; Martrus et al . , 2013; Yang et al . , 2013; Ni et al . , 2014 ) . Despite their diversity of replication and translation mechanisms , each showed a similar relationship between the extent of CPD and reduction in virus replication ability ( Table 1 ) . Typically , 10-fold or greater attenuation in cell culture required >12–15% replacement of WT genome with CPD sequences . It is notable , however , that for each virus , CPD invariably increased frequencies of CpG and UpA dinucleotides ( Table 1 ) , typically from 0 . 4–0 . 6 to 1 . 4–1 . 6 ( CpG ) and from 0 . 5–0 . 8 to 1 . 1–1 . 4 ( UpA ) in the mutated regions . 10 . 7554/eLife . 04531 . 003Table 1 . Relationship between codon pair de-optimisation , CpG and UpA frequencies and virus fitness reductionDOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 003WTCPDVirusGeneProp'nCP biasCpGUpACP biasCpGUpAReplication ReductionRefPoliovirusPV-XCapsid14 . 8%−0 . 030 . 520 . 75−0 . 461 . 341 . 25×25Coleman et al . , 2008PV-XYCapsid25 . 9%−0 . 030 . 540 . 75−0 . 461 . 311 . 27×400Influenza A virus*HAMinSegs . 411 . 4%0 . 020 . 430 . 64−0 . 421 . 651 . 11×3 . 5Mueller et al . , 2010HA/NPMinSegs . 4 , 521 . 3%0 . 020 . 440 . 55−0 . 421 . 561 . 14×14PR83FSegs . 1 , 4 , 529 . 1%0 . 010 . 430 . 53−0 . 411 . 551 . 07×35HIV-1Agag4 . 6%0 . 030 . 471 . 04−0 . 431 . 431 . 25×7Martrus et al . , 2013Bgag4 . 7%0 . 0800 . 91−0 . 371 . 221 . 15×3Cgag4 . 8%0 . 030 . 311 . 00−0 . 381 . 501 . 09× 8Dgag2 . 1%−0 . 0200 . 49−0 . 421 . 470 . 99×1 . 5PRRSVSAVE5gp52 . 6%†−0 . 060 . 630 . 73−0 . 381 . 371 . 14×4‡Ni et al . , 20142*Codon pair minimised sequences of IAV were not provided in ( Coleman et al . , 2008 ) and for the purposes of comparison these have been reconstructed in SSE . Note that the CP scores described in Table 1 of that paper ( −0 . 386 , −0 . 420 and −0 . 421 for PB1 , HA and NP respectively ) are not minimum scores; these are in fact −0 . 533 , −0 . 585 and −0 . 602 . Therefore , for the purposes of comparison , CP score minimisation in the current study was targeted to the former values . Although the sequences generated by SSE were not identical to those obtained previously , they would demonstrate a similar distortion of dinucleotide frequencies to those used in the previous study ( Coleman et al . , 2008 ) . †Mutated region only ( positions 147–542 in gp5 ) . ‡Data from replication assay in PAM cells . This linkage can be accounted for at least in part by the association between CP choice and the identity of the dinucleotide between the third and first ( 3–1 ) codon positions . We analysed these parameters in coding regions of a curated dataset of over 35 , 170 human mRNA sequences . The representation of each of the 3904 codon pairs found in coding sequences ( ie . , 61 × 64 ) was calculated , taking into account both the nucleotide composition of the sequences and the amino acid usage as previously described ( Gutman and Hatfield , 1989 ) . Relative under- and over-representation of each was indicated in a heat map , with values ranging from −0 . 222 to +0 . 271 ( mean 0 . 072 ) . Values were plotted on x- and y-axes using values that reflected base identities at each of six positions in the codon pair ( Figure 1A ) . Most of the 256 CPs with CpG at the 3–1 position ( sixth main column ) were markedly under-represented . There was further influence of codon pair position 6 ( CP score was more suppressed if A or U ) but with generally minimal and inconsistent influences of nucleotide identities at other codon positions ( Bennetzen and Hall , 1982; Buchan et al . , 2006; Atkinson et al . , 2014 ) . In the overall distribution of human codon pair representations , codon pairs containing CpG across the codon boundary distributed towards the negative tail of the distribution of CP score ( Figure 1B ) and accounted for almost all those with scores below −1 . 25 . 10 . 7554/eLife . 04531 . 004Figure 1 . ( A ) CP usage in human coding sequences arranged in a 64 x 64 grid . CP frequencies relative to those expected from nucleotide and amino acid frequencies ( CP bias ) are colour coded in a heat map . The primary division on the x-axis is by identity of the 3–1 dinucleotide as annotated . Within these , further divisions within each of the 16 columns show the identity of the nucleotide at position 2 ( A , C , G or U ) . The y-axis records nucleotides at positions 4 ( 4 main divisions on the y-axis ) , 1 ( 4 subdivisions of position 4 ) and 6 ( 4 subdivisions of position 1 ) . Positions of unused codon pairs containing a 5′ stop codon ( translated as *|x ) are shaded in grey . CP usage heat maps for A . thaliana , C . elegans and E . coli coding sequences are shown in Figure 1—figure supplement 2A–2C . ( B ) Distribution of codon pair bias scores in human coding sequences; separate labelling of the 64 codon pairs with CpG ( red ) or UpA ( blue ) across the codon junction ( 3–1 ) demonstrates their consistent under-representation based on their component nucleotide and amino acid frequencies . The distribution of codon pair scores for A . thaliana , C . elegans and E . coli are shown in Figure 1—figure supplement 3A–3C . Correlations between codon pair scores between human coding sequences and those of A . thaliana , C . elegans and E . coli are shown in Figure 1—figure supplement 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 00410 . 7554/eLife . 04531 . 005Figure 1—figure supplement 1 . Distribution of relative synonymous codon usage values for degenerate codons in the human genome ( stop codons were excluded ) . Codons with CpG and UpA at the 1–2 or 2–3 codon position are shaded as indicated in the key ( data derived from http://bioinformatics . weizmann . ac . il/databases/codon ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 00510 . 7554/eLife . 04531 . 006Figure 1—figure supplement 2 . CP scores of codon pairs of ( A ) A . thaliana , ( B ) C . elegans and ( C ) E . coli ORFeomes . The primary division on the x-axis is by identity of the 3–1 dinucleotide ( labelled on y-axis ) , divisions within each column show the identity of codon position 2 . The y-axis records codon positions 5 ( 1 cycle ) , 1 ( 4 cycles ) and 6 ( 16 cycles ) . Positions of codon pairs translated as *|x are shaded grey . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 00610 . 7554/eLife . 04531 . 007Figure 1—figure supplement 3 . Distribution of codon pair scores for other organisms- ( A ) A thaliana , ( B ) C . elegans and ( C ) E . coli , with separate representation of codon pairs with CpG and UpA across the codon junction . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 00710 . 7554/eLife . 04531 . 008Figure 1—figure supplement 4 . Correlation between representations of human codon pairs ( x-axis ) with those of other organisms- ( A ) A thaliana , ( B ) C . elegans and ( C ) E . coli ( y-axis ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 008 CPs with UpA at the 3–1 position were also specifically under-represented in human mRNA sequences ( Figure 1A , B ) , consistent with global under-representation of this dinucleotide in coding sequences from eukaryotes ( Beutler et al . , 1989; Duan and Antezana , 2003 ) . The dataset additionally demonstrated over-representation of CpA and UpG dinucleotides at the 3–1 position; these are typically created by the ( methylation-associated ) C->T transition upstream of G ( fifth and 14th main columns in Figure 1A ) and of CpC and CpU ( Simmen , 2008 ) . However , with a few exceptions , such as the prominent over-representation of GCG|GCG and CCG|CCG , other codon pairs showed infrequent or minor differences in representation . The avoidance of CpG and UpA in human mRNA sequences at the 3–1 position was further manifested at other three codon position ( Bennetzen and Hall , 1982; Atkinson et al . , 2014 ) ; among the 61 degenerate codons , those containing CpG or UpA at these positions showed lower relative synonymous codon usage than those containing other dinucleotides ( Figure 1—figure supplement 1 ) . Avoidance of codon pairs with CpG at the 3–1 position was also observed in the plant genome of A . thaliana that also possesses a methylation-dependent suppression of CpG dinucleotides ( Figure 1—figure supplements 2A and 3 ) . Codon pair usage of human and plant coding sequences was indeed significantly correlated ( R2 = 0 . 146; Figure 1—figure supplement 4 ) . In contrast to plant coding sequences , no equivalent avoidance of CpG-containing codon pairs was observed in organisms with non-methylated genomes ( Caenorhabditis elegans and Escherichia coli; Figure 1—figure supplements 2B , C , 3 , 4 ) . The close association between CP usage and the identity of dinucleotides at codon boundaries immediately complicates any observational assessment of the potentially separate contributions of CP bias and CpG/UpA dinucleotide frequencies on virus replication . On the one hand , it could be hypothesised that the suppression of CpG and UpA at position 3–1 in mammalian codon pairs was a simple consequence of avoiding disfavoured codon pairs . Conversely , it could be conceptualised that codon pair choice is driven in part through avoidance of specific dinucleotides . To resolve this functionally , we compared replication dynamics and relative fitness of native E7 with a series of novel mutants of E7 in which dinucleotide frequencies and CP usage were independently manipulated ( Figure 2; Table 2 ) . To achieve this , a mutational program was developed ( Sequence Mutate in the SSE package ( Simmonds , 2012 ) that allowed synonymous changes to be introduced into a coding sequence to achieve a pre-specified CP score target while under constraints such as retaining CpG and UpA dinucleotide frequencies and mononucleotide composition . 10 . 7554/eLife . 04531 . 009Figure 2 . Codon pair scores and numbers of CpG and UpA dinucleotides in native ( WT ) and mutated regions of E7 . Mean CP scores for Regions 1 and 2 combined are shown on the x-axis; the total numbers of CpG and UpA dinucleotides in each sequence are shown on the y-axis . The histogram shows CP scores for the 35 , 170 human mRNA sequences >200 bases in length ( mean 0 . 072; standard deviation ±0 . 031 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 00910 . 7554/eLife . 04531 . 010Table 2 . Composition and codon usage of E7 wt and mutant insert sequencesDOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 010RegionSequence ( Symbol ) G+C contentCpG Total*O/E ratio† , ‡UpA Total*O/E ratio† , ‡Codon UsageCAI¶ENcCP Bias1Native ( WT ) 47 . 6%51 ( − ) 0 . 73062 ( − ) 0 . 7420 . 68556 . 5−0 . 043Permuted ( P ) 47 . 6%51 ( 0 ) 0 . 7302 ( 0 ) 0 . 7420 . 69455 . 8−0 . 025CpG/UpAL ( cu ) 47 . 5%0 ( −51 ) 019 ( −43 ) 0 . 2270 . 68643 . 50 . 087Max-U50 . 1%47 ( −4 ) 0 . 61043 ( −19 ) 0 . 5730 . 70849 . 60 . 328Min_E47 . 5%51 ( 0 ) 0 . 73662 ( 0 ) 0 . 7350 . 74854 . 3−0 . 131Min_U47 . 5%69 ( +18 ) 0 . 99276 ( +14 ) 0 . 9390 . 70958 . 3−0 . 134Min_H49 . 8%106 ( +55 ) 1 . 40079 ( +17 ) 0 . 9810 . 69649 . 2−0 . 1302Native ( WT ) 47 . 1%18 ( − ) 0 . 32048 ( − ) 0 . 6950 . 74353 . 20 . 015Permuted ( P ) 47 . 6%18 ( 0 ) 0 . 32048 ( 0 ) 0 . 6950 . 73949 . 00 . 013CpG/UpAL ( cu ) 48 . 5%0 ( −18 ) 048 ( 0 ) 0 . 2140 . 73947 . 20 . 118Max-U46 . 3%24 ( +6 ) 0 . 44043 ( −3 ) 0 . 6010 . 75046 . 10 . 311Min-E45 . 7%18 ( 0 ) 0 . 34348 ( 0 ) 0 . 6570 . 78553 . 3−0 . 091Min-U47 . 4%37 ( +19 ) 0 . 64950 ( +2 ) 0 . 7380 . 76757 . 6−0 . 083Min-H47 . 8%68 ( +50 ) 1 . 17265 ( +15 ) 0 . 9700 . 71549 . 7−0 . 085*Total number of CpG and UpA dinucleotides in sequence . Changes in numbers between mutated and original WT sequence are indicated in parentheses . †Ratio of observed dinucleotide frequency ( O ) to that expected based on mononucleotide composition ( E ) that is , f ( CpG ) /f ( C ) × f ( G ) . ‡Values deliberately changed are shown in red ( maximised ) and blue ( minimised ) . ¶Calculated from http://genomes . urv . es/CAIcal/ ( Puigbo et al . , 2008 ) . The mutant , Min-E was constructed from two genome regions , together comprising 31% of the E7 genome , in which the coding sequence possessed the minimum possible CP score ( −0 . 111 ) while retaining identical CpG and UpA frequencies as WT virus ( CP score: −0 . 014; CpG: 0 . 525; UpA: 0 . 718; Figure 2 , Table 2 ) . Inserts with the same CP frequencies as Min-E but without dinucleotide frequency constraints ( Min-U; CpG: 0 . 82; UpA: 0 . 95 ) or where CpG and UpA frequencies were maximised ( Min-H; CpG: 1 . 3; UpA: 0 . 98 ) were generated similarly . The three mutants provided the opportunity to investigate effects of dinucleotide frequency differences of viral fitness without the compounding effect of CP bias . It was similarly possible to compare fitness of the mutant , Max-U , with a maximised CP score ( 0 . 320 ) but with similar CpG and UpA frequencies to WT with the previously described mutant , cu|cu , with minimised CpG and UpA frequencies ( 0 and 0 . 22 respectively ) but a CP score marginally greater that WT ( 0 . 11; Figure 2 ) . P|P was the permuted mutant control with randomised codon order but identical coding and dinucleotide frequencies to WT sequence . If CP usage solely determined virus replication ability , the seven mutants would be expected to display the following fitness ranking: Max-U > cu|cu > ( WT = P|P ) > ( Min-H = Min-U = Min-E ) . Conversely , if virus fitness was determined by CpG and UpA dinucleotide frequencies , the following ranking would be expected: cu|cu > Max-U > ( WT = P|P = Min-E ) > Min-U > Min-H . These predictions were determined by generation and infectivity measurement of virus stocks corresponding to the seven mutants and comparing their relative fitness in competition and replication assays ( Figures 3 , 4 ) . 10 . 7554/eLife . 04531 . 011Figure 3 . Replication of WT and mutants of E7 with altered CP and dinucleotide frequencies . Bars are shaded diagrammatically based on their relative CpG/UpA composition . RD cells were infected with E7 WT , at an MOI of 0 . 03 and infectious titres quantified at 8 , 18 and 30 hr time points post inoculation ( p . i . ) by TCID50 determination . Results are the mean of three biological replicates; error bars show standard errors of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 01110 . 7554/eLife . 04531 . 012Figure 4 . RD cells were co-infected with pairs of WT ( W|W ) and E7 mutants at equal MOI and the supernatant serially passaged through cells after development of CPE . RNA was isolated and the composition of each virus determined through selective restriction digests using enzymes listed in Table 3 . ( A ) Examples of three competition assays showing cleavage patterns of individual viruses ( lanes 1 , 2 ) , the starting inoculum ( lane 3 ) and two biological replicates after 10 ( panels 1 , 2 ) or 5 ( panel 3 ) passages in lanes 4 and 5 . Results from the other competition assays are shown in Figure 4—figure supplement 1 . ( B ) Summary of pairwise fitness comparisons of viruses with outcomes for the viruses listed in columns at passages 5 and 10 indicated by colour shading . For example , Min-E and WT showed equal fitness ( yellow shading ) and cu|cu outcompeted WT by passage 5 ( red ) and Max-U by passage 10 . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 01210 . 7554/eLife . 04531 . 013Figure 4—figure supplement 1 . Competition assays between E7 mutants showing competing variants ( lanes 1 and 2 ) andout at indicated passage number ( lane 3 ) for each . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 013 Full length RNA transcribed from each E7 mutant cDNA constructs all generated infectious virus after transfection into RD cells . Stocks of virus were generated from WT and each mutant and infectivity quantified by quantal limiting dilution . To investigate replication kinetics , RD cells were infected with WT and each mutant at an MOI of 0 . 03 in triplicate and infectivity of supernatants measured at 8 , 18 and 30 hr ( Figure 3; ) . During the exponential period of replication ( 8 and 8 hr ) , Min-U and Min-H mutants showed 1 and >2 log reductions in virus replication respectively compared to WT E7 . Contrastingly , the CpG/UpA-minimised mutant , cu|cu replicated to approximately 1 log higher levels that WT . Significantly for the analysis of effects of CP and dinucleotide frequencies on replication , virus titres obtained from mutant with identical ( Min-E , CDLR ) or similar ( Max ) CpG/UpA frequencies to WT were highly similar at both time points . At the last timepoint ( 30 hr ) , RD cells infected with WT , CDLR , Min-E , Max and cu|cu were entirely destroyed or almost entirely destroyed ( Max-U ) and showed similar residual infectivities , while those infected with Min-H showed an incomplete cytopthic effect . Competition assays were used as a more stringent measure of fitness differences in mutants with different codon pair biases . Equal MOIs of WT and mutants were co-inoculated onto RD cells and serially passaged up to ten times . Population compositions were determined by amplification of sequences across modified regions and cleavage with restriction enzymes that differentiated WT mutant sequences from each other ( Figure 4A , B; Table 3 ) . In the examples of competition assays ( Figure 4A ) , Max-U showed similar fitness to WT but a greater population representation at passage 10 . cu|cu completely out-competed Max-U by passage 10 while in the final example , WT and Min-E showed equal fitness at passage 5 and at passage 10 ( Figure 4B ) . A total of 12 pairwise comparisons were made and outcomes in terms of population representation recorded at passage 10 ( Figure 4B; see Key ) . The results are internally consistent and with their replication kinetics ( Figure 3 ) and indicate the following fitness ranking:cu | cu>Max-U> ( WT=P | P=Min-E ) >Min-U>Min-H10 . 7554/eLife . 04531 . 014Table 3 . Enzymes used in selective digests for competition ASSAYSDOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 014Virus 1Virus 2RegionEnzymeTargetW|WP|P1SpeIPermutedW|WMax-U1SacIMaxW|WMin-E1NcoIWTW|WMin-U1NcoIWTW|WMin-H1EcoRVWTW|Wcu|cu1EcoRVWTP|Pcu|cu1SpeIPermutedMax-UP|P1SpeIPermutedMax-Ucu|cu1SacIMaxMin-EMin-U1ClaIMin-UMin-EMin-H1EcoRVMin-EMin-UMin-H1ClaIMin-U Using the Spearman rank correlation test , fitness ranking was significantly associated with CpG and UpA frequencies in the insert region ( p < 0 . 001 ) but showed no association with CP frequencies and other measures of codon usage that potentially influence translation rates , codon adaptation index ( CAI ) and effective number of codons ( ENc ) ( Table 4 ) . Consistently , these results demonstrate that when altered independently from CP bias , only dinucleotide frequencies were associated with replication fitness . 10 . 7554/eLife . 04531 . 015Table 4 . Correlation between fitness ranking and sequence compositionDOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 015VariableSpearman Rp value†CpG/UpA*1 . 0<0 . 001CP bias−0 . 700 . 1 ( n . s . ‡ ) CAI−0 . 334>0 . 5 ( n . s . ) ENc0 . 593>0 . 5 ( n . s . ) G + C content0 . 075>0 . 5 ( n . s . ) Translation efficiency−0 . 074>0 . 5 ( n . s . ) *Number of CpG and UpA dinucleotides in insert region . †From values tabulated in ( Ramsey , 1989 ) . ‡n . s . : not significant . The maxim that any effects of CP frequencies on replication are mediated through its influence on translation efficiency was investigated for the mutants constructed in the study . Translation assays were evaluated in vitro to avoid effects mediated through stress response-related RNA recognition mechanisms that restrict E7 translation and subsequent replication immediately after entry ( Atkinson et al . , 2014 ) . Viral RNA transcripts from E7 WT and mutant cDNA clones were used to program rabbit reticulocyte lysates in the presence of [35S]-methionine . Electrophoresis of reactions after 3 hr showed translation of several bands representing cleaved and partially cleaved E7 proteins ( Figure 5; Figure 5—figure supplement 1 ) . Translation efficiencies of each of the mutant E7 transcripts were comparable to WT RNA; what variability there was between mutants ( Figure 5—figure supplement 1 ) did not correlate with replication fitness ( R = −0 . 075; p > 0 . 5; Table 4 ) . This indicates that , at least in a whole genome context , alteration of either CP or dinucleotide frequencies had no significant effect on viral polyprotein translation and therefore cannot be attributed to the marked differences in replication phenotypes observed . 10 . 7554/eLife . 04531 . 016Figure 5 . Translation of RNA templates generated from WT and mutant E7 cDNAs in a rabbit reticulocyte cell free assay . Assignments of bands to E7 proteins were based on molecular weights on SDS-PAGE . A comparison of densitometry values for viral proteins is shown in Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 01610 . 7554/eLife . 04531 . 017Figure 5—figure supplement 1 . Translation efficiencies estimated by densitometry of band intensities of viral proteins translated in a rabbit reticulocyte cell free assay . Translation efficiencies of mutant E7 cDNAs were quantified relative to expression from the WT template . Bars show mean values for seven viral proteins; error bars show standard errors of the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 017
Understanding what limits the replication of viruses with altered CP and dinucleotide frequencies is critical in the evaluation of their broader safety as attenuated virus vaccines . The proposed mechanism in which alterations in CP bias alter translation efficiency and it is this that inhibits virus replication introduces a conceptual model in which it is the virus that is intrinsically defective . With the large number of mutations required for reversion , such viruses should be stably attenuated in whatever context they are used . However , as we have now shown , the replication defect of CPD viruses is actually mediated through alterations in dinucleotide frequencies in the genome that influence their recognition by the cell . In this alternative paradigm , viruses with elevated frequencies of CpG and UpA are not intrinsically defective but they are more readily recognised by the cell and prevented from initiating replication . Their attenuation is therefore dependent on the efficacy of the host innate immune response . The cellular mechanisms responsible for differential recognition and response to RNA sequences with different dinucleotide composition are currently unknown . In our previous study , we obtained evidence that replication inhibition of high CpG/UpA mutants of E7 occurred shortly after cell entry and was not mediated though conventional pattern recognition receptors ( Atkinson et al . , 2014 ) . In that study , we additionally demonstrated that it was additionally not the result of differences between high and low CpG/UpA viruses in their sensitivity to the cellular interferon response . We did observe , however , that the attenuated phenotype of high mutants could be entirely reversed by the kinase inhibitor , C16 , a finding that suggests that recognition may occur through an as yet uncharacterised PKR-related component of the stress response pathway in the cell . Both the adaptive and innate arms of the human immune system are highly polymorphic with remarkable variability in function and expression of many key components of recognition or effector proteins mediating antiviral responses ( Thomas et al . , 2009; Everitt et al . , 2012; Hambleton et al . , 2013; Pothlichet and Quintana-Murci , 2013 ) . Although uncharacterised mechanistically , there is clearly a potential danger that pathways that restrict the replication of high CpG/UpA RNA viruses may be similarly variable in the efficacy in humans and in veterinary species with different genetic backgrounds . The attenuation of live vaccines and safety margins established for their large scale use may be similarly variable; investigation of population differences in innate cellular responses to viruses of different dinucleotide compositions is essential in the evaluation of the safety of this new generation of high CpG/UpA live attenuated vaccines .
RNA transcripts of the pT7:E7 infectious cDNA clone of the isolate Wallace ( accession number AF465516 ) were used to generate E7 viral stocks . E7 was propagated in rhabdomyosarcoma ( RD ) cells using Dulbecco modified Eagle medium ( DMEM ) with 10% foetal calf serum ( FCS ) , penicillin ( 100 U/ml ) and streptomycin ( 100 µg/ml ) . All cells were maintained at 37°C with 5% CO2 . The two regions in the E7 genome used previously to investigate effects of dinucleotide frequencies on virus replication ( Atkinson et al . , 2014 ) were used in the current study ( Region 1: 1878–3119 and 5403–6462 ) . Previously characterised mutants comprised the CpG/UpA-low mutant cu|cu with all CpG dinucleotides and as many UpA dinucleotides possible eliminated and the permuted mutant P|P in which codon order was permuted while retaining protein coding and native dinucleotide frequencies . Further mutants ( Max-U , Min-E , Min-U , Min-H ) are described in the main text; sequences listed in Supplementary file 1 . Manipulation of dinucleotide frequencies and codon pair scores in coding sequences was performed using the program Sequence Mutate in version 1 . 2 of the SSE package ( Simmonds , 2012 ) . Reference datasets of human , A . thaliana , C . elegans and E . coli messenger RNA sequences were obtained from the Refseq database ( http://www . ncbi . nlm . nih . gov/nuccore ) . Codon pair usage tables were generated from coding regions of each mRNA sequence datasets from the four organisms using the program Composition Scan in the SSE package ( Simmonds , 2012 ) . Codon pair tables generated by SSE were used to calculate CP frequencies for all mRNA sequences with coding regions >200 bases in length from each organism . These comprised 35 , 770 human mRNA sequences , 32 , 768 from A . thaliana , 24 , 093 from C . elegans and 4316 from E . coli . The codon pair table generated by SSE from our dataset of human mRNA sequences was used in preference to that previously described ( Table S2 in reference Coleman et al . , 2008 ) because of the larger number of human mRNA sequences now available . The previously published dataset as presented additionally unaccountably omitted a large number of codon pairs ( 3586 were listed instead of the 3904 expected—61 × 64 ) . There was however a good correlation between CP frequencies between the two datasets ( Figure 6 ) . 10 . 7554/eLife . 04531 . 018Figure 6 . Comparison of codon pair scores generated by SSE using a dataset of 35 , 770 human mRNA sequences ( y-axis ) with those used in a previous analysis ( Coleman et al . , 2008 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04531 . 018 The codon adaptation index for human codon usage was calculated through the website http://genomes . urv . es/CAIcal/ ( Puigbo et al . , 2008 ) . The effective number of codons ( Enc; Wright , 1990 ) and CP usage ( Buchan et al . , 2006; Coleman et al . , 2008 ) were calculated using the program Composition Scan in SSE . Mutant E7 constructs with altered CP frequencies were generated from custom synthesised DNA sequences ( Eurofins Genomics , Ebersberg , Germany ) . Mutant clones were constructed as previously described ( Atkinson et al . , 2014 ) . All clones were sequenced over the insert regions prior to further applications . Infectious virus from each cDNA clone was recovered by transfection of RNA transcripts produced from plasmids linearised using NotI using a Riboprobe System-T7 in vitro transcription kit ( Promega Ltd . Southampton , UK ) . 100 ng of RNA was transfected into RD cells using Lipofectamine 2000 ( Invitrogen , Life Technologies Ltd . , Paisley , UK ) according to the manufacturer's instructions . The resulting cell lysates were used to generate passage 1 stocks by re-infecting RD cells . Viral titres were determined by TCID50 titration in RD cells . Multi-step growth curves for each virus were generated by infecting RD cells in triplicate in 24-well plates at an MOI of 0 . 03 as previously described ( Atkinson et al . , 2014 ) . Supernatant collected at time points ( 8 , 18 and 30 hr post-infection ) were assayed for infectivity by quantal dilution . Competition assays were performed as previously described . Briefly , equal titres of virus pairs ( combined MOI = 0 . 01 ) were applied simultaneously to RD cells in 25 cm2 bottles . Following the development of CPE , supernatant was collected and 300 µl applied to fresh RD cells . This was continued for up to 10 passages . The results of the competition assays were determined by restriction enzyme digestion of the amplicon amplified from Region 1 by combined reverse transcription—PCR ( Atkinson et al . , 2014 ) . Restriction enzymes used to differentiate each mutant pair are listed in Table 3 . RNAs were produced by in vitro T7 transcription ( Riboprobe System T7 , Promega ) of the various cDNA plasmids , each linearised with NotI ( Promega ) . Transcript RNAs were used to program nuclease-treated rabbit reticulocyte lysates ( Promega ) supplemented with HeLa cell S10 cytoplasmic extracts ( Dundee Cell Products , Dundee , UK ) . Reactions were set-up as follows; 7 μl rabbit reticulocyte lysate , transcript RNA ( 0 . 25–2 μg ) , 0 . 5 μl 1 mM amino acid mix ( minus methionine ) , 0 . 5 μl [35S]-methionine ( 1200 Ci/mmol ) , 10 U RNasin Ribonuclease Inhibitor and 2 . 25 μl HeLa cell extract in a total volume of 12 . 5 μl . Reactions were incubated at 30°C for 3 hr and analysed by SDS-PAGE ( 4–20% Tris-Glycine , Expedeon Ltd . Cambridge , UK ) . Gels were exposed to film ( Thermo Scientific , Basingstoke , UK ) for 1–4 days at −70°C . To determine the relative density of the protein bands , densitometry was carried out on the scanned gel image using ImageJ 1 . 48 software ( http://imagej . nih . gov/ij ) . | Viruses cause a number of diseases in humans , such as measles or polio , that can be prevented by vaccines . Some vaccines contain whole viruses that have been weakened or modified so that they do not cause illness . In response to infection with these weakened viruses , the immune system creates cells that can ‘recognise’ , and protect against , the disease-causing forms of the virus if these are encountered later . Directly altering the genetic material of viruses has been suggested as a safe way of weakening them for use in vaccines . Viruses store their genetic material in strands of either DNA or RNA , which are each made up of building blocks called nucleotides . Each nucleotide is commonly represented by a single letter—for RNA , these are A , C , G and U . These letters are read in blocks of three , called codons , when the RNA sequence is translated to make proteins . Some codons perform the same tasks . However , in many organisms , a ‘codon bias’ exists , where one codon is used more often than the others that could perform the same role . Certain codons also tend to be found next to each other while some codon pair combinations are avoided; this is called ‘codon pair bias’ . Separately , some individual nucleotides are more likely to occur in certain pairs than others . The nucleotide pairs CG and UA , for example , generally occur less often than expected in the RNA of viruses , as well as in the genomes of the animals and plants they infect . Currently , much research is focusing on developing weakened viruses for vaccines by introducing unfavourable host codon pairs into viral RNA , as this has been suggested to reduce how efficiently virus RNA sequences are translated into proteins . However , Tulloch et al . noticed that introducing these unfavourable codon pairs also increases the number of CG or UA nucleotide pairs present in the RNA of the virus . It is therefore unclear whether increasing the number of unfavoured codon pairs or unfavoured nucleotide pairs produces the weakening effect . Tulloch et al . took a virus that infects the human gut , called echovirus 7 , and created two types of mutant virus . One set of viruses had altered codon pair frequencies but the overall number of CG or UA nucleotide pairs was kept constant . In the other set , codon pair frequencies were kept the same but CG and UA frequencies were changed . Tulloch et al . found that altering codon pair frequencies did not affect virus replication or fitness . In contrast , increasing CG or UA nucleotide frequencies weakened the virus . Other work by the researchers involved in this study has shown that this impaired replication arises through the viruses being more readily targeted by the immune response of the invaded cell; the viruses themselves are not replication-defective . Engineering viruses with precisely tuned replication properties is a breakthrough in the quest for safer virus vaccines , and provides useful tools for investigating virus–host interactions . The dramatic effect on viral replication of altering nucleotide pair frequencies has revealed the existence of new pathways that defend cells against virus infections , knowledge that could be exploited to rationally design vaccines . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2014 | RNA virus attenuation by codon pair deoptimisation is an artefact of increases in CpG/UpA dinucleotide frequencies |
In Gram-negative bacteria , lipid asymmetry is critical for the function of the outer membrane ( OM ) as a selective permeability barrier , but how it is established and maintained is poorly understood . Here , we characterize a non-canonical ATP-binding cassette ( ABC ) transporter in Escherichia coli that provides energy for maintaining OM lipid asymmetry via the transport of aberrantly localized phospholipids ( PLs ) from the OM to the inner membrane ( IM ) . We establish that the transporter comprises canonical components , MlaF and MlaE , and auxiliary proteins , MlaD and MlaB , of previously unknown functions . We further demonstrate that MlaD forms extremely stable hexamers within the complex , functions in substrate binding with strong affinity for PLs , and modulates ATP hydrolytic activity . In addition , MlaB plays critical roles in both the assembly and activity of the transporter . Our work provides mechanistic insights into how the MlaFEDB complex participates in ensuring active retrograde PL transport to maintain OM lipid asymmetry .
The cell envelope of Gram-negative bacteria such as Escherichia coli is composed of two lipid bilayers termed the inner and outer membranes . The presence of the outer membrane ( OM ) makes Gram-negative bacteria generally resistant to external insults , including antibiotics and detergents , and allows these bacteria to survive in harsh environments ( Nikaido , 2003 ) . Unlike the inner membrane ( IM ) , which is a phospholipid ( PL ) bilayer , the OM contains lipopolysaccharides ( LPS ) and PLs in the outer and inner leaflets , respectively ( Kamio and Nikaido , 1976 ) . This unique lipid asymmetry , characterized by a tightly-packed LPS outer leaflet , renders the OM impermeable to a wide range of compounds , including hydrophobic molecules ( Nikaido , 2003 ) . During growth or in the event of stress , PLs may appear in the outer leaflet of the OM ( Jia et al . , 2004; Wu et al . , 2006; Dalebroux et al . , 2014 ) . This disrupts the LPS layer and reduces the barrier function of the OM . The cell has developed mechanisms to cope with such perturbations on lipid asymmetry . Two OM β-barrel enzymes degrade PLs that have accumulated in the outer leaflet of the OM; OmpLA cleaves both acyl chains from the glycerol backbone of PLs ( Dekker , 2000 ) while PagP transfers one acyl chain from PLs to LPS ( Bishop , 2005 ) or phosphatidylglycerol ( PG ) ( Dalebroux et al . , 2014 ) . In addition , the OmpC-Mla system is thought to maintain OM lipid asymmetry by removing PLs from the OM and transporting them back to the IM ( Malinverni and Silhavy , 2009; Chong et al . , 2015 ) . Osmoporin OmpC forms a complex with the OM lipoprotein MlaA that likely removes PLs from the outer leaflet of the OM ( Chong et al . , 2015 ) . The periplasmic chaperone MlaC is believed to transport these extracted PLs across the aqueous periplasm , and hand them over to a putative ATP-binding cassette ( ABC ) family transporter , MlaFEDB , at the IM ( Malinverni and Silhavy , 2009 ) . Whether these PLs get inserted into the outer leaflet or transported back to the inner leaflet of the IM is not known . The exact composition and the roles of the respective components of the ABC transporter complex have not been elucidated . MlaE and MlaF constitute the core components of the ABC transporter and are predicted to form the transmembrane domains ( TMDs ) and nucleotide-binding domains ( NBDs ) , respectively ( Malinverni and Silhavy , 2009 ) . In addition to MlaC , this system is proposed to contain a second substrate-binding protein ( SBP ) , a single-pass membrane protein MlaD , that may be associated with the IM complex . Consistent with this idea , MlaD , along with MlaE and MlaF , are conserved across many species of Gram-negative bacteria and can also be found in actinomycetes ( Casali and Riley , 2007 ) and in plants ( Benning , 2009 ) . The Mce4 pathway in Mycobacterium tuberculosis is important for cholesterol uptake ( Pandey and Sassetti , 2008 ) while the TGD system in the chloroplasts of Arabidopsis thaliana functions to transport phosphatidic acid ( PA ) from the plastid OM to the IM ( Benning , 2009 ) , providing support for the proposed role of the OmpC-Mla system in lipid transport . In E . coli and a few other Gram-negative bacteria , a small cytoplasmic protein MlaB is also predicted to be part of the ABC transporter ( Malinverni and Silhavy , 2009; Casali and Riley , 2007 ) . MlaB contains a Sulfate Transporter and Anti-Sigma factor antagonist ( STAS ) domain that is believed to have a general nucleoside triphosphate ( NTP ) binding function ( Aravind and Koonin , 2000 ) . Its role in the IM complex is unclear . Here , we characterize the putative MlaFEDB complex in E . coli , and elucidate critical roles for auxiliary proteins , MlaD and MlaB , in the assembly and activity of the complex . We show that MlaF , MlaE , MlaD and MlaB interact specifically with each other to form a stable complex . Within this complex , MlaD forms an SDS-resistant hexamer via its soluble domain , and we demonstrate that this domain binds PLs . We further show that MlaD depresses ATP hydrolytic activity upon association with the MlaFEB sub-complex . Finally , we establish that MlaB is necessary for both the assembly and activity of the ABC transporter , likely by modulating MlaF structure and stability . Our work provides novel mechanistic insights into how the MlaFEDB complex functions to maintain lipid asymmetry in the OM .
To determine whether MlaF , MlaE , MlaD and MlaB interact with each other , we performed affinity purification experiments using wild-type ( WT ) cells exogenously expressing N-terminally His-tagged MlaE protein ( His-MlaE ) . His-MlaE is expressed from a “leaky” expression plasmid that has been shown to yield low cellular levels of other proteins ( Wu et al . , 2006; Chong et al . , 2015 ) . This construct is fully functional , as it is able to restore SDS/EDTA resistance in the ΔmlaE mutant strain ( Figure 1—figure supplement 1 ) . Two unique protein bands at ~29 kDa and ~19 kDa co-purified with His-MlaE ( Figure 1A ) . These bands represent MlaF and MlaD , respectively , as they are no longer co-purified in the corresponding ΔmlaF and ΔmlaD mutant strains . Even though His-MlaE enabled enrichment of MlaF and MlaD , we are unable to detect the His-tagged protein in these samples , which have been heated prior to analysis . As there is a tendency for hydrophobic membrane proteins to aggregate when heated , we wondered if this was true for His-MlaE . Indeed , we are able to detect His-MlaE on immunoblots when the samples are not heated ( Figure 1A , right panel ) . His-MlaE migrates anomalously as a diffuse band around 24 kDa ( expected 28 kDa ) , likely because it is still partially folded in the presence of SDS , a phenomenon commonly observed for multi-pass membrane proteins ( Rath et al . , 2009 ) . Remarkably , MlaD also migrates differently , now as a high molecular weight species , when the samples are not heated , suggesting that it may be forming oligomers . To examine if the interactions between MlaF , MlaE and MlaD were specific , we performed reciprocal affinity purification experiments using WT cells expressing C-terminally His-tagged MlaF ( MlaF-His ) or MlaD ( MlaD-His ) proteins , which we show are functional ( Figure 1—figure supplement 1 ) . Even though we could not probe for MlaE due to the lack of appropriate antibodies , we demonstrate that MlaF-His and MlaD-His are able to pull down MlaD and MlaF , respectively , likely via interactions with MlaE ( Figure 1—figure supplement 2A ) . Taken together , these results establish that MlaF , MlaE and MlaD interact specifically in a complex . 10 . 7554/eLife . 19042 . 003Figure 1 . MlaF , MlaE , MlaD and MlaB form a stable complex . ( A ) Co-TALON affinity purification using WT and indicated mutant strains harboring empty vector ( pET23/42 ) or pET23/42His-mlaE ( pHis-mlaE ) . Samples ( heated or non-heated ) were subjected to SDS-PAGE ( 12% Tris . HCl gel ) , and visualized by silver staining and immunoblot analyses using antibodies against the pentahistidine tag . ( B ) SEC profile of MlaF ( His-E ) DB complex purified from cells over-expressing MlaF ( His-E ) DCB . The peak fraction ( heated or non-heated ) was subjected to SDS-PAGE ( 4–20% Tris . HCl gel ) followed by Coomassie Blue ( CB ) staining and immunoblot analysis . His-MlaE can only be detected on immunoblots when samples are not heated . Under the same conditions , MlaD migrates as a high molecular weight species . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 00310 . 7554/eLife . 19042 . 004Figure 1—figure supplement 1 . His-tagged Mla proteins are able to rescue SDS/EDTA sensitivity in the respective mla mutant strains . Serial dilutions of cultures of wild-type ( WT ) , ∆mlaF , ∆mlaE , ∆mlaD and ∆mlaB strains harboring pET23/42 empty vector , pET23/42mlaF-His ( pmlaF-His ) , pET23/42His-mlaE ( pHis-mlaE ) , pET23/42mlaD-His ( pmlaD-His ) or pET23/42His-mlaB ( pHis-mlaB ) , respectively , were spotted on LB agar plates containing 200 μg/mL ampicillin , supplemented with or without 0 . 50% SDS and 0 . 60 mM EDTA as indicated , and incubated overnight at 37°C . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 00410 . 7554/eLife . 19042 . 005Figure 1—figure supplement 2 . MlaF , MlaE , MlaD and MlaB form a stable complex . Co-TALON affinity purification experiments using wild-type ( WT ) and indicated mutant strains harboring ( A ) pmlaF-His , pmlaD-His , or ( B ) pHis-mlaB . Samples were heated and subjected to SDS-PAGE ( 12% Tris . HCl gel ) , and visualized by silver staining and immunoblot analyses using antibodies against the pentahistidine tag . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 00510 . 7554/eLife . 19042 . 006Figure 1—figure supplement 3 . MlaF , MlaD and MlaB co-purify with His-tagged MlaE following overexpression and affinity purification . Tandem MS/MS sequencing confirmed the identities of protein bands corresponding to MlaF , MlaD and MlaB in a preparation of the purified MlaFEDB complex . The top three most abundant proteins in each band are shown . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 00610 . 7554/eLife . 19042 . 007Figure 1—figure supplement 4 . SEC-MALS analysis of the MlaF ( His-E ) DB complex . Molecular mass: 256 kDa ( predicted , MlaF2E2D6B2 ) , 285 ( ± 0 . 5% ) kDa ( observed ) . Molecular masses of DDM fraction in complex and free micelles are 35 ( ± 4 . 4% ) kDa and 65 ( ± 2 . 1% ) kDa , respectively . Numbers stated after ± show statistical consistency of analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 007 We were not able to detect MlaB in our pulldown experiments , suggesting either that it is not part of the complex or it may have eluded detection due to its small size ( ~10 kDa ) . To determine whether MlaF , MlaE and MlaD form a stable complex with MlaB , we over-expressed the mlaFEDCB operon engineered to encode His-MlaE , and purified the complex on cobalt affinity resin followed by size exclusion chromatography ( SEC ) . The proteins elute as a single peak , indicating the formation of a stable complex ( Figure 1B ) . SDS-PAGE analysis of the complex show three bands at ~29 kDa ( MlaF ) , ~20 kDa ( MlaD ) and ~10 kDa ( MlaB ) when the sample is heated . The identities of these bands are confirmed by mass spectrometry ( MS ) ( Figure 1—figure supplement 3 ) . In the non-heated sample , we are also able to detect His-MlaE ( ~24 kDa ) and the oligomeric form of MlaD ( ~120 kDa ) ( Figure 1B ) . Consistent with MlaB being a part of the complex , we are able to show that functional His-tagged MlaB can pull down the other components in WT cells ( Figure 1—figure supplements 1 and 2B ) . Even though we also expressed MlaC , it was not co-purified with the rest of the complex . This suggests that interactions between MlaC and the IM complex may be weak and/or transient , in line with the idea that it is a periplasmic binding protein ( Malinverni and Silhavy , 2009 ) . We conclude that MlaF , MlaE , MlaD and MlaB form a stable complex . MlaD appears to form an oligomeric structure stable to SDS when purified as part of the IM complex . Full length MlaD comprises an N-terminal transmembrane helix and a large C-terminal periplasmic substrate-binding domain ( Figure 2A ) ( UniProtKB P64604 ) ( Magrane and the UniProt consortium , 2011 ) . To determine whether this oligomeric state is dependent on the transmembrane helix or its association with the complex , we over-expressed and purified the substrate-binding domain ( or soluble domain ) of MlaD ( sMlaD-His ) alone for characterization . We found that sMlaD-His does not exist in the monomeric state , as judged by its SEC profile ( Figure 2B ) . Multi-angle light scattering ( MALS ) analysis revealed that the absolute molar mass of purified sMlaD-His is ~110 . 7 kDa , establishing that these are in fact hexamers ( Figure 2C ) . Furthermore , the experimental molar mass of the MlaFEDB complex ( ~285 kDa by SEC-MALS ) is consistent with the presence of six copies of MlaD ( Figure 1—figure supplement 4 ) . Remarkably , hexamers formed by sMlaD-His alone are also resistant to SDS denaturation ( Figure 2B ) . These results indicate that oligomerization and extreme stability are unique properties of the soluble domain . 10 . 7554/eLife . 19042 . 008Figure 2 . MlaD forms SDS-resistant hexamers via its soluble domain . ( A ) Domain organization of MlaD . ( B ) SEC profile of purified soluble domain of MlaD ( sMlaD-His ) . Elution volumes of standard globular proteins ( aldolase 158 kDa , conalbumin 75 kDa and ovalbumin 44 kDa ) are indicated . The peak fraction ( heated or non-heated ) was subjected to SDS-PAGE ( 4–20% Tris . HCl gel ) followed by CB staining . Positions of relevant molecular weight markers are indicated in kDa . ( C ) SEC-MALS analysis of sMlaD-His . Hexamer molecular mass: 107 kDa ( predicted ) , 110 . 7 ( ± 0 . 5% ) kDa ( observed ) . Numbers stated after ± show statistical consistency of the analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 008 The TGD pathway in A . thaliana has been proposed to transport PA between the inner and outer membranes of chloroplasts ( Benning , 2009 ) . In this system , TGD2 is the direct homolog of MlaD and was found to bind PA ( Awai et al . , 2006; Roston et al . , 2011 ) . We therefore hypothesized that MlaD also binds PLs as part of its role in maintaining lipid asymmetry in the OM . As a first step , we examined whether our purified preparations of sMlaD-His contained bound PLs . We extracted potentially bound PLs from purified sMlaD-His and analyzed these samples using thin layer chromatography ( TLC ) and 31P nuclear magnetic resonance ( NMR ) spectroscopy . We use delipidated LolB ( dLolB-His ) , the OM receptor for OM-targeted lipoproteins , as a specificity control . LolB is not expected to bind PLs even though it contains a hydrophobic cavity that allows it to chaperone the triacyl modification found on lipoproteins ( Takeda et al . , 2003 ) . We detected both phosphatidylethanolamine ( PE ) and PG bound to sMlaD-His ( Figure 3A and B ) but not to dLolB-His ( Figure 3A ) , indicating that MlaD has affinity for PLs . Since sMlaD-His no longer contains the transmembrane helix and has little tendency to associate with membranes , we believe that bound PLs in sMlaD-His are directly reflective of its proposed role in PL transport . Comparable amounts of PE and PG ( ~1:1 ) co-purified with sMlaD-His despite PE being the predominant lipid in E . coli cellular extracts ( ~72% PE , 17% PG , 11% cardiolipin ( CL ) ) ( Lugtenberg and Peters , 1976 ) ( Figure 3B ) . These results suggest that MlaD may have a preference for binding PG . 10 . 7554/eLife . 19042 . 009Figure 3 . sMlaD co-purifies with endogenous PLs . ( A ) TLC analysis of PLs extracted from BL21 ( λDE3 ) cells , purified dLolB-His and sMlaD-His . ( B ) 31P NMR analysis of PL extracts from BL21 ( λDE3 ) cells and purified sMlaD-His in 5% Triton X-100 . Compositions of bound PLs were obtained via integration of peak areas , and normalized to the number of phosphorus atoms per PL molecule ( i . e . one for PE/PG and two for CL ) . Unknown peaks that cannot be assigned to any PL species in E . coli ( see Figure 3—figure supplement 1 ) are annotated with asterisks ( * ) . ( C ) Positive mode , non-denaturing electrospray ionization ( ESI ) mass spectrum of sMlaD-His . Under native conditions ( 20 mM ammonium acetate , pH 6 . 9 ) , sMlaD hexamers with charge states centred around +24 could be detected . After deconvolution , the molecular weight of the native hexamers was ~110 kDa , indicating the presence of at least four bound PLs ( i . e . P6L4 , assuming an average mass of 750 Da per PL ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 00910 . 7554/eLife . 19042 . 010Figure 3—figure supplement 1 . 31P NMR analysis of E . coli PLs . 31P chemical shifts ( ppm ) of phosphatidic acid ( 14:0 ) ( PA ) , phosphatidylserine ( PS ) , cardiolipin ( CL ) , phosphatidylglycerol ( PG ) , and phosphatidylethanolamine ( PE ) are given in the table . PS were extracted from E . coli EH150 strain , which accumulates PS when grown at 42°C ( Hawrot et al . , 1975 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 01010 . 7554/eLife . 19042 . 011Figure 3—figure supplement 2 . MS analyses of sMlaD-His . ( A ) Positive-mode ESI-MS spectrum of sMlaD-His under denaturing conditions ( 50% acetonitrile 0 . 2% formic acid ) , revealing monomeric sMlaD-His at various charge states . Inset: Deconvoluted spectrum of monomeric sMlaD-His ( experimental 17828 . 32 Da , theoretical 17827 . 77 Da ) . ( B ) Positive mode ESI-MS spectrum of sMlaD-His under non-denaturing conditions ( 20 mM ammonium acetate , pH 6 . 9 ) with increasing collision energies ( 0 V ( black ) , 80 V ( green ) , 100 V ( pink ) and 150 V ( blue ) ) , progressively revealing hexameric sMlaD-His with four to zero bound PL molecules . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 011 To further characterize the interaction between MlaD and PLs , we analyzed purified sMlaD-His using MS . In denaturing MS , the monomeric form of sMlaD-His ( deconvoluted mass 17 , 828 Da ) is the only molecular species observed ( Figure 3—figure supplement 2A ) . In native MS , we detected the hexameric form of sMlaD-His with bound ligands at various charge states ( Figure 3C ) . Deconvolution of the native MS spectrum indicates the molecular weight of the native hexamers to be ~110 kDa , suggesting the presence of at least four bound PL molecules ( ~750 Da each , in agreement with the average mass of PE and PG ) and therefore , strong protein-PL interactions . By increasing the collision energy in the mass spectrometer collision cell , we are then able to destabilize the native structure , revealing the presence of hexamers binding three , two , one or zero PL molecules ( Figure 3—figure supplement 2B ) . This suggests that the presence of ligands is not strictly required for the formation of sMlaD hexamers . Taken together , our data establish that hexameric sMlaD is a PL-binding complex with at least four binding sites . MlaE and MlaF constitute TMDs and NBDs of a canonical ABC transporter , respectively ( Malinverni and Silhavy , 2009 ) . MlaD binds PLs , consistent with its proposed role as a periplasmic binding protein; however , it is not clear if MlaD has other roles in the activity of the ABC transporter . The function of MlaB is also unknown . To characterize the importance of MlaD and MlaB in the complex , we attempted to over-express and purify sub-complexes containing His-MlaE , including MlaFE , MlaFED and MlaFEB ( Figure 4 ) . We were successful in obtaining purified MlaFEB ( Figure 4C ) , but not MlaFE and MlaFED ( Figure 4A and B ) . Despite reasonable expression of MlaF ( Figure 4—figure supplement 1 ) , it is not co-purified with His-MlaE in preparations of MlaFE and MlaFED ( Figure 4A and B ) . Therefore , MlaF does not interact with MlaE unless MlaB is present . To exclude artifacts due to over-expression , we also performed affinity purification using His-MlaE expressed at low levels in cells lacking MlaB . Unlike in WT cells , MlaF is not co-purified with His-MlaE in the absence of MlaB ( Figure 4D ) . Similarly , we show that MlaF-His could not pull down MlaD ( through MlaE ) without MlaB ( Figure 4—figure supplement 2 ) . In fact , the affinity-enriched levels of MlaF-His appear to be reduced in the absence of MlaB , suggesting that MlaB modulates the stability of MlaF . Our results demonstrate that MlaB plays a key role in the assembly of the complex . 10 . 7554/eLife . 19042 . 012Figure 4 . MlaB is required for the stability and/or assembly of the canonical ABC transporter . SEC profiles of ( A ) His-MlaE purified from cells over-expressing MlaF ( His-E ) , ( B ) ( His-MlaE ) D purified from cells over-expressing MlaF ( His-E ) D , and ( C ) MlaF ( His-MlaE ) B purified from cells over-expressing MlaF ( His-E ) B . The respective peak fractions ( non-heated ) were subjected to SDS-PAGE ( 4–20% Tris . HCl gel ) followed by CB staining . ( D ) Co-TALON affinity purification using WT and ∆mlaB strains harboring empty vector ( pET23/42 ) or pET23/42His-mlaE ( pHis-mlaE ) . Samples ( heated or non-heated ) were subjected to SDS-PAGE ( 12% Tris . HCl gel ) , and visualized by silver staining and immunoblot analyses using antibodies against the pentahistidine tag . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 01210 . 7554/eLife . 19042 . 013Figure 4—figure supplement 1 . MlaF is produced at high levels in strains over-expressing full and sub-complexes of the IM ABC transporter . SDS-PAGE ( 12% Tris . HCl gel ) analysis of cell lysates isolated from strains over-expressing MlaF ( His-E ) , MlaF ( His-E ) D , MlaF ( His-E ) B and MlaF ( His-E ) DCB from indicated vectors , with or without IPTG induction . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 01310 . 7554/eLife . 19042 . 014Figure 4—figure supplement 2 . MlaD is not co-purified with MlaF-His in the absence of MlaB . Co-TALON affinity purification using wild-type ( WT ) , ∆mlaB and ∆mlaE strains harboring an empty pET23/42 vector or pmlaF-His . Samples were heated and subjected to SDS-PAGE ( 12% Tris . HCl gel ) , and visualized by silver staining and immunoblot analyses using antibodies against the pentahistidine tag . Positions of relevant molecular weight markers are indicated in kDa . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 014 We next examined whether MlaB plays additional roles in the activity of the ABC transporter . To test this , we constructed a variant with a mutation in the STAS domain of MlaB at position 52 , which in two other STAS domain proteins have been shown to be important for function ( Aravind and Koonin , 2000; Diederich et al . , 1994; Rouached et al . , 2005 ) . This mutation does not affect the assembly of the MlaFEB complex ( Figure 5A ) , allowing us to compare the steady-state ATP hydrolysis rates of purified MlaFEBWT and MlaFEBT52A complexes . We demonstrate that MlaFEBWT exhibits high intrinsic ATPase activity in detergent micelles ( kcat = 1 . 8 ± 0 . 5 μmol ATP s-1/μmol complex ) ( Figure 5A ) , comparable to known ABC importers ( Reich-Slotky et al . , 2000; Tal et al . , 2013 ) . In addition , ATP binding is cooperative ( Hill coefficient = 1 . 5 ± 0 . 5 ) . Remarkably , no activity is detected with the MlaFEBT52A complex , similar to a variant of the complex containing a predicted non-functional mutation in the MlaF Walker A motif ( MlaFK47REB ) ( Walker et al . , 1982 ) . Accordingly , the mlaBT52A or mlaFK47R alleles do not complement SDS-EDTA sensitivity observed in ΔmlaB and ΔmlaF strains , respectively ( Figure 5—figure supplement 1 ) . These results imply that MlaB assumes critical role ( s ) during catalysis , and validate this STAS domain protein as an essential and functional component of the IM ABC transporter . 10 . 7554/eLife . 19042 . 015Figure 5 . MlaD and MlaB modulate ATP hydrolytic activity of the IM ABC transporter . ( A ) Enzyme-coupled ATPase assays of indicated complexes ( 0 . 1 μM ) performed in detergent micelles ( 0 . 05% DDM ) . Average ATP hydrolysis rates ( obtained from triplicate experiments , see Figure 5—source data 1 ) were plotted against ATP concentrations , and fitted to an expanded Michaelis-Menten equation that includes a term for Hill coefficient ( n ) ; MlaFEDB ( kcat = Vmax/[complex] = 0 . 6 ± 0 . 3 μmol ATP s-1/μmol complex , Km = 181 . 1 ± 203 . 6 μM , n = 1 . 0 ± 0 . 6 ) and MlaFEB ( kcat = 1 . 8 ± 0 . 5 μmol ATP s-1/μmol complex , Km = 161 . 4 ± 75 . 57 μM , n = 1 . 5 ± 0 . 5 ) . SDS-PAGE analysis of the complexes ( non-heated ) used for these assays is shown on the right . Error bars in the graph and numbers stated after ± are standard deviations of triplicate data . ( B ) A proposed model for how the MlaFEDB complex functions to drive PL transport from the OM to the IM . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 01510 . 7554/eLife . 19042 . 016Figure 5—source data 1 . Source data for ATPase assay . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 01610 . 7554/eLife . 19042 . 017Figure 5—figure supplement 1 . mlaFK47R and mlaBT52A are non-functional alleles . Serial dilutions of cultures of wild-type ( WT ) and ∆mlaF strains harboring pET22/42 empty vector , pET22/42mlaF-His or pET22/42mlaFK47R-His , or WT and ∆mlaB strains harboring pCDF empty vector , pCDFmlaB or pCDFmlaBT52A were spotted on LB agar plates containing appropriate antibiotics , supplemented with or without 0 . 50% SDS , 0 . 65/0 . 70 mM EDTA as indicated , and incubated overnight at 37°C . The strains used here do not express T7 polymerase . IPTG ( 1 mM ) was added to allow de-repression and thus low-level transcription ( by endogenous polymerases ) from the T7-lacO promoter in these plasmids . DOI: http://dx . doi . org/10 . 7554/eLife . 19042 . 017 SBPs are also known to modulate the activities of their associated ABC transporters ( Davidson et al . , 2008 ) , so we asked whether the presence of MlaD affected the activity of the MlaFEB complex . We show that the association of MlaD with MlaFEB does not significantly affect the affinity of the complex for ATP ( Km = 181 . 1 ± 203 . 6 μM for MlaFEDB vs 161 . 4 ± 75 . 57 μM for MlaFEB ) ( Figure 5A ) ; however , its presence in the MlaFEDB complex reduces ATP hydrolytic activity by ~3-fold ( kcat = 0 . 6 ± 0 . 3 μmol ATP s-1/μmol complex ) and affects cooperativity ( Hill coefficient = 1 . 0 ± 0 . 6 ) . Given that the MlaED complex can be stably isolated ( Figures 1A and 4B ) , and that MlaD does not co-purify with MlaF-His in the absence of MlaE ( Figure 4—figure supplement 2 ) , it is likely that MlaD interacts with the ABC transporter solely via MlaE . Our findings suggest that such an interaction may cause structural changes in MlaF ( the NBDs ) and/or directly influence how the two NBDs come together to perform catalysis .
The OmpC-Mla system is believed to maintain OM lipid asymmetry by removing PLs from the outer leaflet of the OM ( OmpC-MlaA complex ) , and shuttling these PLs across the periplasm ( MlaC ) to the IM , where they get handed off to a putative ABC transporter ( Malinverni and Silhavy , 2009; Chong et al . , 2015 ) . In this study , we have characterized the structure and function of the IM transporter . We have shown that MlaF , MlaE , MlaD and MlaB constitute and form a stable complex . Within this complex , we have demonstrated that MlaD forms SDS-resistant hexamers , and co-purifies with PLs . We have also revealed that MlaD modulates the ATPase activity of the complex via its interaction with MlaE . Finally , we have established multiple roles for MlaB in both the assembly and activity of the ABC transporter , and have identified a single mutation in MlaB that separated these two functions . Substrate import via ABC transporters typically involves delivery of the substrate from a periplasmic SBP to the IM complex , where ATP binding and hydrolysis are coupled to substrate translocation across the membrane ( Davidson et al . , 2008 ) . For a few well-characterized transporters , ATP hydrolytic activity is intrinsically low and can only be activated upon the binding of the corresponding SBPs ( Chen et al . , 2001; Liu et al . , 1997 ) , an arrangement believed to reduce the futile hydrolysis of ATP in the absence of substrate ( Davidson et al . , 2008 ) . In the OmpC-Mla system , a second SBP ( MlaD ) with unknown function is associated with the IM ABC transporter ( MlaFEB ) , in addition to the periplasmic SBP ( MlaC ) ( Malinverni and Silhavy , 2009 ) . We have shown that MlaD does indeed bind PLs ( Figure 3 ) , the proposed substrates for this system , consistent with its role as an SBP . In addition , we have demonstrated that MlaFEB possesses high intrinsic ATPase activity that is attenuated by MlaD association ( Figure 5A ) . Interactions with MlaD may stabilize conformations that are inactive in ATP hydrolysis . Furthermore , because our MlaFEDB preparations are solubilized in detergent micelles , which can disrupt lipid binding , it may be possible that MlaD does not have PLs bound . While this idea remains to be tested , the complex may be unstimulated without bound substrates , resulting in low ATPase activity ( Figure 5B ) . Our results suggest a function for MlaD in ensuring that the full complex conserves ATP , and presumably retains the capacity to be activated only when MlaC binds and/or delivers the PL substrate . How the OmpC-Mla pathway mediates retrograde PL transport from the OM back to the IM ( against a concentration gradient ) is not clear . One can posit that ATP hydrolysis by the MlaFEDB complex in the IM is directly coupled to the removal of PLs from the outer leaflet of the OM . Such a scenario would likely require physical interactions between the IM and OM complexes , as is the case for the LPS transport ( Lpt ) machinery ( Chng et al . , 2010a; Freinkman et al . , 2012 ) . However , the facts that the MlaC homolog from Ralstonia solenacerum can be purified and crystallized as a soluble complex with PE ( Protein Data Bank ID: 2QGU; DOI: 10 . 2210/pdb2qgu/pdb ) , and that we could not detect strong interactions between MlaC and the OmpC-MlaA ( Chong et al . , 2015 ) or MlaFEDB complexes , indicate that Mla-mediated PL transport across the periplasm may proceed instead via a soluble intermediate . In this alternative scenario , we posit that PL transfer from the OmpC-MlaA complex to MlaC , and then to MlaD in the IM complex , is largely driven by affinity ( Figure 5B ) . We have demonstrated that MlaD forms homo-hexamers ( Figure 2 ) , suggesting a maximum of six PL binding sites within the MlaFEDB complex . At least four of these binding sites have reasonably high affinity for PLs , thereby facilitating co-purification with MlaD and detection in native MS analysis ( Figure 3 ) . MlaD may have higher affinity for PL substrates than MlaC , driving overall PL transfer ( Figure 5B ) . In the MlaFEDB complex , the energy derived from ATP hydrolysis may then be utilized to release bound PLs into the membrane , perhaps by inducing conformational changes that alter PL binding affinities within the MlaD oligomer . The fate of these PL substrates , whether being translocated across the IM or simply released into the outer leaflet of the membrane , remains an intriguing question . We believe that the hexameric architecture of MlaD , which contains the MCE domain ( Figure 2A ) , is conserved in homologs of the Mla system found in plants and actinomycetes . In the chloroplasts of A . thaliana , the TGD pathway has a similar function in PL transport from the OM back to the IM ( Benning , 2009 ) . Here , the MlaD-like protein , TGD2 , has been proposed to exist as an oligomer ( Roston et al . , 2011 , 2012 ) . In M . tuberculosis , there are four Mce systems , of which Mce4 is essential for cholesterol uptake and catabolism ( Pandey and Sassetti , 2008 ) , while others may have functions in transporting other lipids ( Forrellad et al . , 2014 ) . These systems are expressed from operons each comprising genes encoding two MlaE-like permeases and six MlaD-like ( MCE ) proteins ( Casali and Riley , 2007 ) . The organization of these operons suggests the assembly of ABC transporters containing a hetero-hexameric arrangement of MCE domains . This hexameric MCE structure may thus be a general architecture for binding and transporting various lipid substrates . MlaD appears to be specific for PG over PE , the major PLs in E . coli ( Figure 3 ) ; however , we do not know if other minor anionic PLs , such as PA and PS , are also substrates . The observed preference for PG may reflect a higher affinity of MlaD for this lipid simply during purification conditions . Alternatively , it may indicate specificity for PG during PL transport . Since the OmpC-Mla system is proposed to remove PLs from the outer leaflet of the OM so as to maintain lipid asymmetry , this finding suggests that either PG has a higher tendency to appear on the cell surface during normal growth , or removal of outer leaflet PG is more important to ensure OM stability . Consistent with these ideas , the OM enzyme PagP has been evolved to acylate outer leaflet PG ( in addition to LPS ) , but not PE , as a mechanism to fortify the OM barrier ( Dalebroux et al . , 2014 ) . It has also been reported that the OM has lower PG content than the IM in E . coli ( Lugtenberg and Peters , 1976 ) and Salmonella Typhimurium ( Osborn et al . , 1972 ) . While these observations are still controversial , an active PG-specific retrograde transport system could have utility in establishing the apparent differences in membrane PL compositions . The OmpC-Mla system is unique in that the IM ABC transporter contains a cytoplasmic STAS domain auxiliary protein , MlaB , whose function is not known ( Malinverni and Silhavy , 2009 ) . We have shown that MlaB is critical for the assembly of the ABC transporter . In the absence of MlaB , the canonical ABC transporter complex ( MlaFE ) could not be isolated ( Figure 4A ) , owing to the decreased stability of MlaF and/or weakened interactions between MlaE and MlaF . We have also established a separate role for MlaB in the activity of the complex; a single T52A mutation in the STAS domain protein does not affect complex assembly but abolishes ATPase activity ( Figure 5A ) . Corroborating these conclusions , we note that STAS domains have similarly been reported to be critical for the assembly and activity of non-ABC family transporters , including the Sultr1;2 sulfate transporter in A . thaliana ( Shibagaki and Grossman , 2006 ) . In fact , MlaB-like and MlaE-like domains can be found in the same polypeptides in some organisms ( Pfam accession number: PF13466 ) ( Finn et al . , 2016 ) , further highlighting the importance of MlaB in these transporters . MlaB may regulate ATPase activity via novel mechanisms . The T52A mutation removes a conserved threonine residue in MlaB that is mapped to a serine in the STAS domain anti-sigma factor antagonist , SpoIIAA , in Bacillus subtilis ( Aravind and Koonin , 2000 ) . This residue can be phosphorylated by its target anti-sigma factor SpoIIAB , which disrupts the ability of SpoIIAA to bind SpoIIAB , presumably due to a conformational change ( Diederich et al . , 1994 ) . By analogy , we hypothesize that MlaB may become phosphorylated in cells . Due to its influence on MlaF structure and stability , any conformational changes associated with this putative phosphorylation event might indirectly regulate MlaF and affect its function in ATP hydrolysis . Intriguingly , SpoIIAA has also been shown to exhibit NTP binding and hydrolytic activity ( Najafi et al . , 1996 ) . This raises an alternative possibility that MlaB may be directly involved in ATP hydrolysis in conjunction with MlaF , perhaps with T52 serving as an alternative nucleophile . While such mechanistic involvement has not been reported , direct participation by an auxiliary protein in catalysis would necessitate a shift in our understanding of how ABC transporters can operate .
All bacterial strains used in this study are summarized in Supplementary file 1 . Strains MC4100 [F- araD139 Δ ( argF-lac ) U169 rpsL150 relA1 flbB5301 ptsF25 deoC1 ptsF25 thi] ( Casadaban , 1976 ) and BL21 ( λDE3 ) [F– ompT gal dcm lon hsdSB ( rB–mB– ) λ ( DE3 [lacI lacUV5-T7 gene 1 ind1 sam7 nin5] ) [malB+]K-12 ( λS ) ] ( Novagen ) were used for in vivo and protein over-expression experiments , respectively . Deletion mutants were generated by P1 transduction using BW25113 [F- ∆ ( araD-araB ) 567 ∆lacZ4787::rrnB-3 λ- rph-1 ∆ ( rhaD-rhaB ) 568 hsdR514] mutant strains from the Keio collection ( Baba et al . , 2006 ) . MlaB deletion mutation was constructed using recombineering as previously described ( Datsenko and Wanner , 2000; Malinverni and Silhavy , 2009 ) . Primers MlaB-cam-N5 and MlaB-cam-C3 ( Supplementary file 3 ) with homologies to regions upstream and downstream of mlaB , were used to amplify the CamR cassette from pKD3 by PCR . The resulting PCR product was used to replace the mlaB gene by electroporation into BW25113 expressing λRed recombinase ( pKD46 ) ( Datsenko and Wanner , 2000 ) . The mlaB::cam allele was subsequently introduced into the wild-type ( WT ) strain by P1 transduction . Luria-Bertani ( LB ) broth and agar were prepared as previously described ( Silhavy et al . , 1984 ) . Unless otherwise noted , ampicillin ( Amp ) ( Sigma-Aldrich , MO , USA ) was used at a concentration of 200 μg/mL , chloroamphenicol ( Cam ) ( Alfa Aesar , Heysham , UK ) at 15 μg/mL , kanamycin ( Kan ) ( Sigma-Aldrich ) at 25 μg/mL , streptomycin ( Sm ) ( Sigma-Aldrich ) at 50 μg/mL and spectinomycin ( Spec ) ( Sigma-Aldrich ) at 20 μg/mL . All plasmids are listed in Supplementary file 2 . To construct most plasmids , the desired gene or DNA fragment was amplified by PCR using appropriate DNA template and primers listed in Supplementary file 3 . The amplified DNA fragment was digested with appropriate restriction enzymes ( New England Biolabs , MA , USA ) and ligated into the same restriction sites of the desired vector . NovaBlue cells were transformed with the ligation product and selected on LB plates containing appropriate antibiotics . MlaF and MlaB site substitution mutants were constructed by site-directed mutagenesis using the parent plasmids and primers listed in Supplementary file 3 . The plasmid pET23/42 ( N ) was used to construct pET23/42His-mlaE and pET23/42His-mlaB . To construct pET23/42 ( N ) , which contains a His-tag at the N-terminus , the first cloning site of pETDuet-1 ( Novagen ) was cloned and inserted into the pET23/42 plasmid ( Wu et al . , 2006 ) . To construct pET22/42mlaF ( His-E ) DCB , mlaFEDCB was first amplified and inserted into NdeI/AvrII sites of the pET22/42 ( Chng et al . , 2010b ) . The coding sequence for the linker region-His6 was subsequently inserted into pET22/42mlaFEDCB via two rounds of site-directed mutagenesis/insertion using 1-Linker-NHis-MlaE-N FWD/1-Linker-NHis-MlaE-C REV and 2-Linker-NHis-MlaE-N FWD/2-Linker-NHis-MlaE-C REV primer pairs . All constructs were verified by DNA sequencing ( Axil Scientific , Singapore ) . WT or mutant strains used in these experiments express the His-tagged Mla proteins from the pET23/42 plasmid under the control of the T7 promoter , which is transcribed inefficiently by endogenous non-T7 RNA polymerases . We believe that these proteins are produced at near physiological or low levels ( as opposed to being over-expressed ) because we have previously shown that leaky expression of MlaA from this plasmid is in fact lower than what is produced from the endogenous locus ( Chong et al . , 2015 ) . For each strain , a 3-L culture ( inoculated from an overnight culture at 1:100 dilution ) was grown in LB broth at 37°C until OD600 of ~0 . 5–0 . 7 . Equal amounts of cells for each strain ( normalized by OD600 ) were harvested by centrifugation at 4 , 700 x g for 20 min . Cells were resuspended in 25 mL TBS containing 1 mM PMSF ( Calbiochem , Darmstadt , Germany ) , 50 μg/mL DNase I ( Sigma-Aldrich ) and 100 μg/mL lysozyme ( Calbiochem ) . Cells were passed once through a high pressure French Press ( French Press G-M , Glen Mills , NJ , USA ) homogenizer at 20 , 000 psi . Cell debris was removed by centrifugation at 4 , 700 x g for 10 min at 4°C . Subsequently , the supernatant was subjected to ultra-centrifugation ( Model Optima L-100K , Beckman Coulter , CA , USA ) at 145 , 000 x g for 1 hr at 4°C to separate membrane and soluble fractions . The membrane pellet fraction was extracted ( 5 mL of 50 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 5 mM MgCl2 , 10% glycerol , 1% n-dodecyl β-D-maltoside ( DDM ) ( Merck Millipore , Italy ) for 2 hr ( with rocking at 4°C ) and was subjected to second round ultra-centrifugation at 145 , 000 x g for 1 hr at 4°C . The supernatant was then incubated with 0 . 25 mL TALON cobalt resin ( Takara Bio Inc , Japan ) , which had been equilibrated with 5 mL of wash buffer ( 50 mM Tris . HCl pH 8 . 0 , 300 mM NaCl , 10% glycerol , 0 . 05% DDM , 10 mM imidazole ) , by rocking for 1 hr on ice . The mixture was later loaded onto a column and allowed to drain by gravity . The filtrate was passed through the resin again , drained and the column was washed with 5 x 2 mL of wash buffer and finally eluted with 1 mL of elution buffer ( 50 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 10% glycerol , 0 . 05% DDM , 50 mM imidazole ) . The eluate was concentrated in an Amicon Ultra 10 kDa cut-off ultra-filtration device ( Merck Millipore , Ireland ) by centrifugation at 3 , 800 x g to ~100 μL . The concentrated sample was mixed with equal amounts of 2X Laemmli reducing buffer , either kept at room temperature ( - heat ) or boiled at 100°C for 10 min ( + heat ) , and subjected to SDS-PAGE and immunoblot analyses . MlaF ( His-E ) DB , MlaF ( His-E ) , MlaF ( His-E ) D were over-expressed and purified from BL21 ( λDE3 ) cells harboring pET22/42mlaF ( His-E ) CDB , pET22/42mlaF ( His-E ) and pET22/42mlaF ( His-E ) D . In order to optimize amounts of MlaB during MlaF ( His-E ) DB purification , a second over-expression vector pCDFmlaB was introduced into BL21 ( λDE3 ) pET22/42mlaF ( His-E ) CDB . To over-express MlaF ( His-E ) B , pCDFmlaB together with pET22/42mlaF ( His-E ) were introduced into BL21 ( λDE3 ) cells . A 30-mL culture was grown from a single colony in LB broth supplemented with 200 μg/mL Amp and 50 μg/mL Sm ( when necessary ) at 37°C until OD600 ~ 0 . 6 . The cell culture was then used to inoculate a 3-L culture and grown at the same temperature until OD600 ~ 0 . 6 . For MlaF ( His-E ) DB and MlaF ( His-E ) B mutant complexes , 1 mM IPTG ( Axil Scientific , Singapore ) was added and the culture was grown at 37°C for another 3 hr . For other membrane sub-complexes , 0 . 1 mM IPTG was added and the culture was grown at 18°C for another 20 hr . Cells were pelleted by centrifugation at 4700 x g for 20 min and then resuspended in 25 mL TBS containing 1 mM PMSF ( Calbiochem ) , 50 μg/mL DNase I ( Sigma-Aldrich ) and 100 μg/mL lysozyme ( Calbiochem ) . Cells were passed once through a high pressure French Press ( French Press G-M , Glen Mills ) homogenizer at 20 , 000 psi . Cell debris was removed by centrifugation at 4700 x g for 10 min at 4°C . Subsequently , supernatant was subjected to ultra-centrifugation ( Model Optima L-100K , Beckman Coulter ) at 145 , 000 x g for 1 hr at 4°C to separate membrane and soluble fractions . The membrane pellet fraction was extracted ( 20 mL of 20 mM Tris . HCl pH 8 . 0 , 300 mM NaCl , 5 mM MgCl2 , 10% glycerol , 1% n-dodecyl β-D-maltoside ( DDM ) ( Merck Millipore ) and subjected to second round ultra-centrifugation at 145 , 000 x g for 1 hr at 4°C . The supernatant was incubated with 1 mL TALON cobalt resin ( Takara Bio Inc ) , which had been equilibrated with 20 mL of wash buffer ( 20 mM Tris . HCl pH 8 . 0 , 300 mM NaCl , 5 mM MgCl2 , 10% glycerol , 0 . 05% DDM , 20 mM imidazole ) , by rocking for 1 hr on ice . The mixture was later loaded onto a column and allowed to drain by gravity . The filtrate was passed through the resin again , drained and the column was washed with 8 x 10 mL of wash buffer and eluted with 8 mL of elution buffer ( 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 5 mM MgCl2 , 10% glycerol , 0 . 05% DDM , 50 mM imidazole for full complex and 200 mM imidazole for sub-complexes ) . The eluate was concentrated in an Amicon Ultra 10 kDa cut-off ultra-filtration device ( Merck Millipore ) by centrifugation at 3 , 800 x g to ~500 μL . Proteins were further purified by SEC system ( AKTA , GE Healthcare , UK ) at 4°C on a prepacked Superdex 200 increase 10/300 GL column , using 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 5 mM MgCl2 , 10% glycerol , 0 . 05% DDM as the eluent . For MlaF ( His-E ) B complexes , two columns were connected in series to allow better peak separation . sMlaD-His and dLolB-His were over-expressed and purified from BL21 ( λDE3 ) harboring pET22/42smlaD-His and pET22blolB-His ( Chng et al . , 2010b ) , respectively . A 5-mL culture was grown from a single colony in LB broth supplemented with 200 μg/mL Amp at 37°C until OD600∼ 0 . 6 . The cell culture was then used to inoculate into 500 mL LB broth and grown at the same temperature until OD600∼ 0 . 6–0 . 7 . At this time , 1 mM and 0 . 1 mM IPTG ( Axil Scientific ) were added to cultures and grown at 37°C for 3 hr and at 18°C for 20 hr to induce sMlaD-His and dLolB-His over-expression , respectively . Cells over-expressing sMlaD-His were pelleted by centrifugation at 4700 x g for 20 min and then resuspended in 20 mL TBS pH 6 . 25 ( 20 mM Tris . HCl pH 6 . 25 , 300 mM NaCl ) whereas cells over-expressing dLolB-His were resuspended in 20 mL TBS pH 8 . 0 ( 20 mM Tris . HCl pH 8 . 0 , 300 mM NaCl ) , both buffers containing 5 mM imidazole . We found that purification of sMlaD-His at ∼ pH 6 gave more homogenous preparations . Buffers were supplemented with 1 mM PMSF ( Calbiochem ) , 50 μg/mL DNase I ( Sigma-Aldrich ) and 100 μg/mL lysozyme ( Calbiochem ) . Resuspended cells were passed twice through a high pressure French Press ( French Press G-M , Glen Mills ) homogenizer at 20 , 000 psi . Cell debris was removed by centrifugation at 4700 x g for 10 min at 4°C . Subsequently , the supernatant was subjected to ultra-centrifugation ( Model Optima L-100K , Beckman Coulter ) at 145 , 000 x g for 1 hr at 4°C to separate membrane and soluble fractions . The soluble fraction was incubated with 2 mL TALON cobalt resin ( Takara Bio Inc ) which had been equilibrated with 5 mL of TBS pH 6 . 25 ( for sMlaD-His ) and TBS pH 8 . 0 ( for dLolB-His ) both containing 20 mM imidazole , and incubated by rocking for 1 hr on ice . The resin mixtures were later loaded onto gravity column . The filtrates were collected and the columns were washed with 4 x 20 mL TBS pH 6 . 25 for sMlaD-His and TBS pH 8 . 0 for dLolB-His , both containing 20 mM imidazole . sMlaD-His and dLolB-His proteins were eluted from columns with 8 mL of 20 mM Tris . HCl pH 6 . 25 , 150 mM NaCl , 200 mM imidazole and 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 200 mM imidazole , respectively . The eluates were concentrated in an Amicon Ultra 10 kDa cut-off ultra-filtration device ( Merck Millipore ) by centrifugation at 3 , 800 x g to ~500 μL . Proteins were further purified by SEC system ( AKTA , GE Healthcare ) at 4°C on a prepacked Superdex 200 10/300 GL column , using 20 mM Tris . HCl pH 6 . 25 , 150 mM NaCl ( for sMlaD-His ) and 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl ( for dLolB-His ) as the eluents . Prior to each SEC-MALS analysis , a preparative SEC was performed for bovine serum albumin ( BSA ) ( Sigma-Aldrich ) to separate monodisperse monomeric peak and to use as a quality control for the MALS detectors . In each experiment , monomeric BSA was injected before the protein of interest and the settings ( calibration constant for TREOS detector , Wyatt Technology ) that gave the well-characterized molar mass of BSA ( 66 . 4 kDa ) were used for the molar mass calculation of the protein of interest . For sMlaD-His , Superdex 200 Increase 10/300 GL column was equilibrated with 20 mM Tris . HCl pH 6 . 25 , 150 mM NaCl . The injected protein concentration was kept at 5 mg/mL . Light scattering and refractive index data were collected online using miniDAWN TREOS ( Wyatt Technology , CA , USA ) and Optilab T-rEX ( Wyatt Technology , CA , USA ) , respectively , and analyzed by ASTRA 6 . 1 . 5 . 22 software ( Wyatt Technology ) . The dn/dc value was set to 0 . 185 mL/g for the soluble protein ( Slotboom et al . , 2008 ) . For MlaF ( His-E ) DB , SEC column equilibration and BSA calibration were performed in 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 5 mM MgCl2 , 10% glycerol and 0 . 05% DDM . 1 mg/mL complex was used and data collected as above . To calculate non-proteinaceous part of the complex , we used the protein-conjugate analysis in ASTRA software . In this analysis , we used dn/dc value of 0 . 143 mL/g and 0 . 185 mL/g for DDM and protein complex , respectively ( Slotboom et al . , 2008 ) . For BSA , UV extinction coefficient of 0 . 66 mL/ ( mg . cm ) was used . For the MlaF ( His-E ) DB complex , that was calculated to be 0 . 84 mL/ ( mg . cm ) , based on its predicted stoichiometric ratio MlaF2 ( His6-E ) 2D6B2 . To analyze endogenously bound PLs , lipids from purified sMlaD-His , dLolB-His ( Chng et al . , 2010b ) , BL21 ( λDE3 ) and EH150 cell pellet were extracted according to Bligh-Dyer method ( Bligh and Dyer , 1959 ) . For BL21 ( λDE3 ) lipid extraction , an overnight BL21 ( λDE3 ) culture was grown in 5 mL LB broth at 37°C . 1 . 5 mL of the culture was pelleted down and resuspended in 100 μL deionized water . For the lipid extraction from temperature-sensitive E . coli EH150 strain ( Hawrot and Kennedy , 1975 ) , a 3-mL culture ( inoculated from an overnight culture at 1:100 dilution ) was grown first at 30°C until OD600 of ~0 . 5–0 . 7 and subsequently at 42°C for another 4 hr . 1 . 5 mL culture was taken , pelleted down and resuspended in 100 μL deionized water . 2 . 5 mg of purified proteins ( sMlaD-His and dLolB-His ) were used for PL extraction . Purified protein solutions , BL21 ( λDE3 ) and EH150 pellet resuspensions were mixed with 3 . 75 volumes of chloroform:methanol ( 1:2 vol/vol ) . The mixtures were vortexed and sonicated sequentially for 30 s for three times . The mixtures were centrifuged at 21 , 000 x g for 5 min and the supernatants were recovered . Appropriate volumes of chloroform and TBS ( for protein samples ) or deionized water ( for cell pellet ) were added to the supernatants to form a two phase mixture chloroform:methanol:water ( 2:2:1 . 8 ) . The mixtures were then centrifuged at 4000 x g for 5 min to separate organic and aqueous phases . Organic phase was gently removed to another vial and the organic solvent was evaporated under N2 gas . Remaining dried lipids were dissolved in 50 μL chloroform:methanol ( 4:1 vol/vol ) and 10 μL spotted onto a TLC Silica gel 60 F254 plate ( Merck ) . The plate was developed in chloroform:methanol:water ( 65:25:4 ) solvent system , left to dry at room temperature and then stained with iodine vapor for 5 min . Visualization was done with G:Box Chemi-XX 6 ( Genesys version 1 . 4 . 3 . 0 , Syngene ) . To find out the selectivity of sMlaD-His for different PL head groups , we performed 31P NMR analysis with Bruker AV500 ( 500 MHz ) instrument ( MA , USA ) . Commercial E . coli phosphatidylethanolamine ( PE ) , phosphatidylglycerol ( PG ) and cardiolipin ( CL ) ( Avanti Polar Lipids , AL , USA ) were used as standards . Synthetic 1 , 2-dimyristoyl-sn-glycero-3-phosphate ( sodium salt ) ( 14:0 ) ( Avanti Polar Lipids ) was used as phosphatidic acid ( PA ) standard . Phosphatidylserine ( PS ) was extracted from E . coli EH150 strain by accumulating PS at 42°C ( Hawrot and Kennedy , 1975 ) . To achieve resolved 31P signals , phospholipid-Triton X-100 micelles were prepared with some modifications ( London and Feigenson , 1979; Sotirhos et al . , 1986 ) . PE , PG , CL and PA were first solubilized in chloroform:methanol ( 4:1 ) at 3 mM , 2 mM , 4 mM and 1 mM final concentrations , respectively . Lipids from sMlaD-His ( from 10 mg of purified protein ) , BL21 ( λDE3 ) and EH150 strains were extracted as described above ( Bligh and Dyer , 1959 ) . Organic solvents from samples were evaporated under N2 gas and remaining dried lipids were resuspended in 1 mL water containing 5% Triton X-100 ( w/v ) and 10% D2O ( v/v ) , and subsequently sonicated for 1 hr at room temperature . NMR analysis was performed at room temperature and acquisition times for all samples were 10 hr ~ 40 , 000 number of scans ( NS ) , except PA sample for 3 hr . Fourier transformation of FID files and integration of phosphate peak areas were done in MestReNova 9 . 0 software . Purified protein samples were desalted with Amicon Ultra-0 . 5 mL 10 kDa cut-off centrifugal filters ( Merck Millipore ) to exchange the initial buffer with 20 mM ammonium acetate pH 6 . 8 and to eliminate eventually co-purified and loosely bound contaminating small molecules . Non-denaturing analyses were performed in positive ion mode . The sample ( 6 μL , 20 μM ) was directly injected at a flow of 2 μL/min into a quadrupole time-of-flight ( QTOF ) 6550 Agilent mass spectrometer equipped with an Agilent 1200 HPLC system ( CA , USA ) . The nanoelectrospray voltage was set to 1500 V ( Vcap ) , temperature 280oC , drying gas flow 11 L/min and fragmentor 175 V . The samples were injected in 50 mM ammonium acetate pH 6 . 8 and spectra acquired at a scan rate of 1 spectra/sec . Data were collected over a mass range of 1000–8000 m/z . Denaturing mass spectrometry , to confirm the theoretical mass of the monomeric purified protein , was performed injecting the samples in positive ion mode using conditions that could break the protein-ligand complex ( protein diluted 1:10 in 0 . 2% formic acid in 50% acetonitrile ) . Data were acquired between 100 and 8000 m/z ( positive mode ) to monitor the protein conformational states . The sample was analyzed with the same mass spectrometer connected with the Agilent 1200 series ChipLC system and using a C8 chip ( Zorbax 300SB , Agilent ) . The following solvents were used for the protein analysis on the C8 reversed phase HPLC: 0 . 2% formic acid in water ( solvent A ) , 0 . 2% formic acid in acetonitrile ( solvent B ) . The 6550 QTOF electrospray voltage was set to 1600 V ( Vcap ) , temperature 200°C , drying gas 14 L/min , fragmentor voltage 175 V . The denatured protein analysis data were collected in MS only mode and deconvoluted using Bioconfirm software ( Agilent ) and the Maximum Entropy deconvolution algorithm . ATP hydrolytic activity was determined using an NADH enzyme-linked assay ( Nørby , 1988 ) adapted for a microplate reader ( Kiianitsa et al . , 2003 ) . 50-μL reactions contained assay buffer ( 20 mM Tris . HCl pH 8 . 0 , 150 mM NaCl , 5 mM MgCl2 , 10% glycerol , 0 . 05% DDM ) with 200 μM NADH ( Sigma-Aldrich ) , 20 U/mL lactic dehydrogenase ( Sigma-Aldrich ) , 100 U/mL pyruvate kinase ( Sigma-Aldrich ) , 0 . 5 mM phosphoenolpyruvate ( Alfa Aesar ) and different ATP ( Sigma-Aldrich ) concentrations . The assays were performed at room temperature and fluorescence emission at 340 nm was measured using a SPECTRAmax 250 microplate spectrophotometer equipped with SOFTmax PRO software ( Molecular Devices , CA , USA ) . Readings were taken in ~9 s intervals . The rate of decrease of NADH fluorescence ( due to oxidation ) was calculated from a linear fit to each 10 min time course and converted to ATP hydrolysis rates with a standard curve obtained using known ADP concentrations ( Figure 5—source data 1 ) . Samples were performed in technical triplicates and data were fit to the built-in Hill equation in Origin 9 . OM sensitivity against SDS/EDTA was judged by colony-forming-unit ( CFU ) analyses on LB agar plates containing indicated concentrations of SDS/EDTA . Briefly , 5-mL cultures were grown ( inoculated with overnight cultures at 1:100 dilution ) in LB broth at 37°C until OD600 reached ~0 . 5–0 . 7 . Cells were normalized according to OD600 , first diluted to OD600 = 0 . 1 ( ~108 cells ) , and then serially diluted ( ten-fold ) in LB using 96-well microtiter plates . 2 μL of the diluted cultures were manually spotted onto the plates , dried , and incubated overnight at 37°C . Plate images were visualized by G:Box Chemi-XT4 ( Genesys version 1 . 4 . 3 . 0 , Syngene ) . All samples subjected to SDS-PAGE were mixed with equal amounts of 2X Laemmli reducing buffer . The samples were subsequently either kept at room temperature ( - heat ) or subjected to boiling at 100°C for 10 min ( + heat ) . Equal volumes of the samples were loaded onto the gels . Unless otherwise stated , SDS-PAGE was performed according to Laemmli using the 4–20% Tris . HCl gradient gels ( Laemmli , 1970 ) . After SDS-PAGE , gels were visualized by either Coomassie Blue staining ( Sigma-Aldrich ) , silver staining ( Silver Quest , Invitrogen ) or subjected to immunoblot analysis . Immunoblot analysis was performed by transferring protein bands from the gels onto polyvinylidene fluoride ( PVDF ) membranes ( Immun-Blot 0 . 2 μm , Bio-Rad , CA , USA ) using semi-dry electroblotting system ( Trans-Blot Turbo Transfer System , Bio-Rad ) . Membranes were blocked by 1X casein blocking buffer ( Sigma-Aldrich ) . α-His antibody ( pentahistidine ) conjugated to the horseradish peroxidase ( HRP ) ( Qiagen , Hilden , Germany ) was used at a dilution of 1:5 , 000 . Luminata Forte Western HRP Substrate ( Merck Millipore ) was used to develop the membranes and chemiluminescence signals were visualized by G:Box Chemi-XT4 ( Genesys version 1 . 4 . 3 . 0 , Syngene ) . | Escherichia coli are bacteria that can cause vomiting and diarrhoea in humans and other mammals . Each E . coli cell is surrounded by two membranes , which are each made of two layers of fat molecules known as lipids . The outer membrane prevents the entry of toxic compounds and allows E . coli to withstand damaging agents from outside the cell , such as antibiotics . The outer membrane’s ability to act as an effective barrier depends on an asymmetric , or uneven , distribution of lipid molecules across its two layers . The inside layer is dominated by phospholipids , whereas the outside layer is comprised mainly of lipids with attached sugars . The distribution of the two lipid types is maintained by a molecular machine with components that can be found in both the inner and outer membranes . This machine is thought to remove phospholipids from the outside layer of the outer membrane and transport them back to the inner membrane . A group ( or”complex” ) of proteins known as MIaFEDB operates as a part of this machine at the inner membrane . MlaFEDB is believed to use energy derived from the breakdown of a molecule called ATP to help ensure that phospholipids removed from the outside layer of the outer membrane are reinserted into the inner membrane . It was proposed that the complex contains four proteins , but it was not clear exactly how these components are arranged . Now , Thong et al . reveal how MlaFEDB is organized and characterize the roles of the individual protein components . The experiments confirm that the MlaFEDB complex is made up of four proteins , including two core components and two support proteins ( called MlaB and MlaD ) . There are six copies of MlaD in the complex . In addition , MlaD has a strong affinity for phospholipids and plays a role in controlling the rate at which energy is harnessed through the breakdown of ATP . Further experiments show that the other support protein MlaB is necessary for both the proper assembly and activity of the complex , likely through its interaction with one of the core components . The next step following on from this work is to directly observe MlaFEDB in action to find out how it uses energy to insert lipids into the inner membrane . In the long term , more information about the structure of the complex would be needed to further understand how it works at the molecular level . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Defining key roles for auxiliary proteins in an ABC transporter that maintains bacterial outer membrane lipid asymmetry |
A single , low dose of the NMDA receptor antagonist ketamine produces rapid antidepressant actions in treatment-resistant depressed patients . Understanding the cellular mechanisms underlying this will lead to new therapies for treating major depression . NMDARs are heteromultimeric complexes formed through association of two GluN1 and two GluN2 subunits . We show that in vivo deletion of GluN2B , only from principal cortical neurons , mimics and occludes ketamine's actions on depression-like behavior and excitatory synaptic transmission . Furthermore , ketamine-induced increases in mTOR activation and synaptic protein synthesis were mimicked and occluded in 2BΔCtx mice . We show here that cortical GluN2B-containing NMDARs are uniquely activated by ambient glutamate to regulate levels of excitatory synaptic transmission . Together these data predict a novel cellular mechanism that explains ketamine's rapid antidepressant actions . In this model , basal glutamatergic neurotransmission sensed by cortical GluN2B-containing NMDARs regulates excitatory synaptic strength in PFC determining basal levels of depression-like behavior .
A single , sub-anesthetic dose of the n-methyl d-aspartate ( NMDA ) receptor antagonist ketamine produces anti-depressant effects in treatment-resistant depressed patients ( Berman et al . , 2000; Zarate et al . , 2006 ) . Understanding the cellular signaling mechanisms underlying this effect will lead to new therapeutic strategies , with the potential for enhancing beneficial actions and minimizing side effects associated with treatment regimes . Ketamine evokes a rapid increase in protein synthesis and enhances excitatory synaptic transmission in cortical neurons ( Li et al . , 2010; Autry et al . , 2011 ) . However , it's unclear how suppression of NMDA receptor ( NMDAR ) signaling promotes protein synthesis . The cortical NMDAR complex is heteromultimeric , containing two GluN1 and two GluN2 subunits , the latter of which are encoded by four genes ( GluN2A-D ) ( Monyer et al . , 1992 ) . Cortical NMDARs are dominated by GluN2A and GluN2B subunits . We recently demonstrated that GluN2B-containing NMDARs act in a unique manner , distinct from GluN2A , to directly suppress mammalian target of rapamycin ( mTOR ) signaling and repress protein synthesis ( Wang et al . , 2011a ) . Consistent with a role for GluN2B , selective antagonists of GluN2B-containing NMDARs are effective in producing rapid changes in behavior in both clinical patient populations and rodent models of depression ( Li et al . , 2010 ) ( Maeng et al . , 2008; Preskorn et al . , 2008; Li et al . , 2011 ) . However , it is unknown how selective antagonism of GluN2B-containing receptors produces similar effects as antagonizing NMDARs using non-subunit selective antagonists . We hypothesized that ambient glutamate tonically activates GluN2B-containing NMDARs to basally , and directly , suppress protein synthesis in principal cortical neurons and that antagonism of this action , either by GluN2B-selective or pan-NMDAR antagonists , would initiate the rapid antidepressant effects by increasing protein synthesis and enhancing excitatory synaptic transmission in prefrontal cortex ( PFC ) . This hypothesis predicts that genetic deletion of GluN2B selectively from principal cortical neurons should mimic and occlude the actions of ketamine on depression-like behaviors and excitatory synaptic transmission . To test this , we generated animals with selective genetic knockout of GluN2B in principal cortical neurons ( 2BΔCtx ) by crossing mice with a conditional GluN2B KO allele ( Brigman et al . , 2010 ) and mice expressing Cre-recombinase ( Cre ) under control of the NEX promoter ( Goebbels et al . , 2006 ) . We then sequentially measured behavior , excitatory cortical synapse physiology , and synaptic protein expression following single dose ketamine injection compared to saline-injected control animals . We show here that genetic deletion of GluN2B from principal cortical neurons both mimics and occludes the effects of ketamine in suppression of depression-like behavior and increased frequency of individual excitatory synaptic events onto layer II/III pyramidal neurons in PFC . We also show that mTOR is present in synaptic protein fractions of cortical lysates and ketamine induces a rapid , yet transient , increase in mTOR phosphorylation , which is occluded in 2BΔCtx animals . Cortical GluN2B removal also eliminated susceptibility to chronic corticosterone exposure . Furthermore , GluN2B-containing receptors can be uniquely activated by ambient glutamate , supporting a model whereby GluN2B maintains tonic suppression of protein synthesis in principal cortical neurons . In support of this , we show that modulation of glutamate transporter function , in vivo , bidirectionally regulates excitatory synaptic transmission and that enhancing glutamate transporter function suppresses depression-like behavior while increasing excitatory synaptic drive in PFC . In summary , our data suggest a novel mechanistic model for the antidepressant actions of ketamine that involves tonic activation of GluN2B-containing NMDARs in helping set basal levels of despair through regulation of protein synthesis and excitatory synaptic drive in PFC .
To test the importance of cortical GluN2B-containing NMDARs in regulating despair-like behavior and excitatory synaptic transmission , we generated cortex- and principal neuron-specific GluN2B knockout animals ( 2BΔCtx ) by crossing mice carrying a Lox-P flanked GluN2B allele ( Brigman et al . , 2010 ) with animals containing a Cre-recombinase ( Cre ) cassette expressed in principal neurons of the neocortex: NEXCre ( Goebbels et al . , 2006 ) ( Figure 1 ) . We first confirmed this genetic technique resulted in the removal of GluN2B protein by PCR and western blot analyses . PCR analysis of genomic DNA isolated from tail tissue confirmed the presence of both the NEXCre and GluN2B-floxed alleles in 2BΔCtx mice ( Figure 1A ) . For all experiments involving 2BΔCtx mice , experimental animals ( NEXCre/+ : GluN2Bflox/flox ) were compared to littermate controls ( either NEX+/+ : GluN2Bflox/flox or NEX+/+ : GluN2Bflox/+ ) . In contrast to brainstem lysates , cortical lysates from 2BΔCtx animals at P10 showed significant decrease in GluN2B expression compared to protein samples from controls ( Figure 1B ) . GluN2B protein levels were also significantly reduced at P50–P70 and were not accompanied by any statistically significant change in expression of either GluN1 or GluN2A ( Figure 1B ) . Residual GluN2B protein is due to the expression in non-principal neurons including inhibitory interneurons . 10 . 7554/eLife . 03581 . 003Figure 1 . Genetic knockout of GluN2B from principal cortical neurons in vivo . ( A ) Conditional ‘floxed’ GluN2B knockout mice were crossed with NEX-Cre animals to ablate GluN2B from principal cortical neurons in vivo ( 2BΔCtx ) . Genotyping strategy and data , generated using tail tissue DNA samples , are shown . ( B ) Western blots , normalized to actin , and quantification for cortical and brainstem lysates from animals at P10 demonstrate cortex-restricted suppression of GluN2B protein in 2BΔCtx mice ( PFC = prefrontal cortex ) . Control lysates from P0 cortices of global GluN2B KO and WT animals run on the same blots served as positive and negative controls . Western blots and quantification of protein expression in control and 2BΔCtx cortex in vivo from P50–P70 animals shows selective decrease in GluN2B and no change in GluN1 and GluN2A . ( C ) Example images from NEX-Cre expressing DsRed flox-stop-flox GFP reporter mouse brain slices . Top , coronal section showing restricted cortical expression pattern of NEX-Cre ( GFP+ = Cre+ ) . Bottom , parasagittal slice showing strong Cre expression in both dorsal and ventral hippocampus and quantification , v = ventral , d = dorsal , scale bars = 500 µm . ( D ) Example traces recorded at +50 and −65 mV overlaid show the current response of layer II/III pyramidal neurons in control and 2BΔCtx slices from P18 to P21 animals in response to intracortical stimulation . Traces at +50 mV from control and 2BΔCtx slices demonstrate loss of ifenprodil sensitivity and faster decay kinetics of NMDAR-mediated current in 2BΔCtx neurons . ( E ) Combined analysis revealed a significant increase in the AMPAR to NMDAR-mediated current ratio in 2BΔCtx neurons , no change in overall peak current at +50 mV , a significant increase in the ifenprodil insensitive current at +50 mV , and a significant decrease in decay tau of the NMDAR-mediate current . Data values are means ± sem . *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 t test with respect to control; n . c . = no significant change . DOI: http://dx . doi . org/10 . 7554/eLife . 03581 . 003 As expected , crossing NEX-cre mice with a dsRed flox-stop-flox GFP reporter line ( JAX stock #008705 ) resulted in strong GFP signal in cortical brain regions while subcortical , cerebellar , and brain stem structures expressed RFP due to the absence of cre expression in these structures . In PFC , 80 . 9 ± 1 . 3% of neurons were GFP positive . We also noted that cre expressed strongly in both ventral and dorsal aspects of the hippocampus , in contrast to the cortex-restricted CaMKII promoter-driven cre mouse line , which lacks expression in the ventral hippocampus ( Figure 1C ) ( Tsien et al . , 1996; Brigman et al . , 2010 ) . Consistent with previous reports characterizing this cre encoding animal ( Goebbels et al . , 2006 ) , we observed GFP expression in these reporter mice in pyramidal neurons in the neocortex . To confirm that this genetic manipulation resulted in loss of GluN2B-containing NMDAR-mediated current at cortical synapses , we applied whole-cell voltage-clamp recordings in acute brain slices . Intra-columnar electrical stimulation in slices from PFC of control animals evoked strong synaptic currents in layer II/III pyramidal neurons ( Figure 1D , E ) . By voltage clamping at −65 mV , to maximize the Mg2+ block of NMDARs and thereby isolate AMPAR-mediated current , we observed rapid onset , inward synaptic current in both genotypes confirming the presence of functional AMPAR-containing excitatory synapses . At depolarized potentials ( +50 mV ) these evoked synaptic responses are a mix of non-rectifying AMPAR-mediated current and slower decaying NMDAR-mediated current . Measuring the relative current carried by AMPARs to NMDARs , we observed a significant increase in this ratio in 2BΔCtx animals , indicating either a decrease in NMDAR-mediated signaling , an increase in AMPAR-mediated signaling , or both ( Figure 1E ) . We did not observe any differences in the decay kinetics of these AMPAR-mediated EPSCs ( tau for Cont = 10 . 2 ± 0 . 4 ms vs 2BΔCtx = 11 . 4 ± 3 . 4 ms , n . s . ) nor in integrated charge transfer at −70 mV ( Cont = 1 . 33 ± 0 . 67 pC vs 2BΔCtx = 1 . 77 ± 0 . 53 pC , n . s . ) . Importantly , as predicted by the loss of GluN2B expression , sensitivity of the NMDAR-mediated current recorded at +50 mV to the GluN2B-selective antagonist ifenprodil ( 3 µM ) , which was approximately 40% in control slices , was completely lost in 2BΔCtx animals ( Figure 1D , E ) . Furthermore , loss in ifenprodil sensitivity was accompanied by a significant decrease in the decay tau of the receptor complex , consistent with the fact that GluN2A-containing NMDARs undergo more rapid receptor deactivation ( Flint et al . , 1997 ) . Together with the genetic and biochemical data , these results confirmed that GluN2B-containing NMDARs are absent in pyramidal PFC neurons in 2BΔCtx mice . If GluN2B-mediated signaling , specifically in principal cortical neurons , is directly involved in regulating levels of despair-like behavior in response to NMDAR antagonism we would expect the 2BΔCtx genetic mutation to both mimic and occlude the actions of ketamine . To test this , we injected 2BΔCtx experimental and control animals with either ketamine or saline and then measured behavior and synaptic physiology in the same animals , 30 min and 25 to 30 hr later , respectively ( Figure 2A ) . Consistent with our hypothesis , 2BΔCtx animals exhibited a dramatic decrease in despair-like behavior when compared to littermate control animals , as measured in two behavioral tests with strong predictive ability for antidepressant effectiveness; the forced swim test ( FST ) and tail suspension test ( TST ) ( Castagné et al . , 2011 ) . In fact , the decrease in immobility scores in the FST test was so strong it precluded testing of whether or not the actions of ketamine injection were occluded in 2BΔCtx animals . Suppression of immobility scores in the TST by ketamine was occluded , as predicted , in 2BΔCtx mice and could be recapitulated by injection of the GluN2B-selective antagonist Ro 25–6981 ( Ro ) ( Figure 2B ) . We next tested the sensitivity of these animals to chronic exposure to corticosterone ( 25 µg/ml ) . Surprisingly , while control animals exhibited a predicted increase in immobility in the TST following chronic corticosterone treatment ( 20 days of exposure ) , we saw no change in immobility scores in corticosterone-treated littermate 2BΔCtx animals ( Figure 2B ) . These data show that 2BΔCtx animals are less susceptible to stress-associated changes in depression-like behavior . 10 . 7554/eLife . 03581 . 004Figure 2 . Decreased despair-like behavior and occlusion of ketamine's actions in 2BΔCtx animals . ( A ) Experimental timeline: male animals between P50 and P70 were subjected to i . p . ketamine injection ( ket ) or saline control injection ( sal ) . 30 min following injection , animals were analyzed in the tail suspension test ( TST ) . At 25–30 hr post-injection , animals were subjected to electrophysiological analysis ( see Figure 3 ) . ( B ) A significant decrease in immobility scores was measured in 2BΔCtx animals in the TST and this was mimicked by treatment with the GluN2B-containing NMDAR selective antagonist Ro 25–6981 . The decrease in immobility scores in the 2BΔCtx animals occluded effect of ketamine injection seen in littermate control animals . 2BΔCtx animals were also insensitive to chronic corticosterone treatment ( Cort ) , while this same treatment increased immobility times in TST in control , corticosterone-treated animals . ( C ) 2BΔCtx exhibited a strong anxiolytic behavioral phenotype compared to control animals as measured in the elevated plus maze ( EPM ) . ( D ) Measuring total distance traveled in the open field test ( OFT ) showed a strong and significant effect of ketamine in both genotypes , suggesting that the decreased immobility time in 2BΔCtx animals was not simply due to hyperlocomotion . ( E ) Food restriction was also used in control animals to demonstrate that the decreased immobility times in 2BΔCtx animals was not a consequence of decreased body mass . Data values are means ± sem . *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 t test with respect to control; n . c . = no significant change . DOI: http://dx . doi . org/10 . 7554/eLife . 03581 . 004 In control animals changes in immobility measured in the TST were observed as early as 30 min after ketamine injection , consistent with previous reports ( Li et al . , 2010; Autry et al . , 2011 ) . 2BΔCtx animals also exhibited a strong and significant change in behavior in the elevated plus maze ( EPM ) , which is less sensitive to changes in locomotor activity . Relative to control animals , which prefer the apparent safety of the closed arms of the maze , 2BΔCtx animals actually showed a reversal in this behavior , spending a significantly higher percentage of their time in the open arms of the maze , consistent with a strong reduction in anxiety in these mice ( Figure 2C ) . Furthermore , subjecting animals to a chronic variable stress ( CVS ) paradigm resulted in a significant decrease in open arm time in control animals ( Cont = 32 . 4 ± 10% vs Cont + CVS = 7 . 2 ± 1 . 7% p < 0 . 05 ) but no change in 2BΔCtx mice ( 2BΔCtx = 86 . 6 ± 6 . 7% vs 2BΔCtx + CVS = 69 . 3 ± 14 . 5% n . s . ) . Examining these mice in the open field test ( OFT ) , we noted that changes in despair-like behavior in 2BΔCtx animals could not be accounted for by increased locomotor activity . Specifically , although total baseline locomotor activity is strongly increased in these 2BΔCtx animals , ketamine injection at this age caused a significant suppression of locomotion in both control and 2BΔCtx animals , indicating a dissociation of acute NMDAR suppression from locomotor activity , and consistent with previous data ( Figure 2D ) ( Akillioglu et al . , 2012 ) . At this same time point immobility scores actually decreased ( Figure 2B ) . Additionally , in food-restricted animals matched in age and mass to 2BΔCtx animals , we observed no significant change in immobility scores showing that the smaller body mass of these animals could not account for their decreased immobility times ( Figure 2E ) . We examined reward-based behavior in the sucrose preference test in these animals , interestingly we measured no significant difference in the ratio of 10% sucrose water intake to regular water by 2BΔCtx mice compared to controls ( Cont = 80 . 8 ± 4 . 7% vs 2BΔCtx = 70 . 5 ± 7 . 9% n . s . ) . The inability to detect a difference may reflect the high basal level of sucrose consumption in control animals or could reflect a dissociation of the domains of the behavioral phenotype in 2BΔCtx animals . It should be noted that cre is expressed strongly in the olfactory bulb of these conditional knockout animals , which could affect feeding behavior . Consistent with this idea , and their decreased body mass , food-deprived 2BΔCtx mice showed significantly higher latency to eat in a hidden food pellet task compared to littermate food-deprived control animals ( Cont = 96 . 1 ± 19 . 4 s vs 2BΔCtx = 309 ± 74 . 4 s , p < 0 . 01 ) . Our observation that 2BΔCtx mice had decreased motivation to feed based upon the hidden food pellet task as well as exhibiting longer latency to feed in non-anxiogenic , home-cage conditions ( data not shown ) , in addition to their potentially disrupted olfactory sensitivity , complicated our ability to interpret data from a novelty suppressed feeding task ( NSFT ) . In the NSFT , in a novel cage testing environment , we observed a strong trend to decreased latency to feed in control animals in response to ketamine injection ( 24 hr post-injection ) as expected ( Cont = 207 . 4 ± 70 . 6 s vs Cont + Ket = 58 . 7 ± 29 . 4 s , p = 0 . 17 ) . Yet under basal conditions 2BΔCtx animals actually exhibited increased latency to feed compared to littermate control animals ( Cont = 40 . 6 ± 7 . 1 s vs 149 . 1 ± 45 . 3 s , p = 0 . 06 ) . Thus , we did not further test the effect of ketamine on these animals . These data generated from FST , TST , and EPM , both under basal conditions and in response to corticosterone treatment or chronic variable stress , strongly support the idea that GluN2B-mediated signaling , in cortical pyramidal neurons , is directly involved in setting basal levels of despair-like behavior in mice . Furthermore , loss of effectiveness of ketamine in 2BΔCtx animals indicates that GluN2B-mediated signaling might be required for ketamine's antidepressant actions . We wondered whether or not increased synaptic activity , seen in response to ketamine injection , might be due to an increase in the number of individual excitatory synapses or modulation of the strength of existing synapses . We hypothesized that any changes observed in excitatory transmission in response to ketamine injection should be mimicked and occluded in 2BΔCtx animals , if they are in fact causally related to the behavioral phenotype . We focused on the PFC due to previous observations showing antidepressant-like actions of ketamine are antagonized by local infusion of rapamycin in PFC ( Li et al . , 2010 ) , as well as the recognized importance of this structure in major depressive disorder ( MDD ) ( Price and Drevets , 2012 ) , and in response to behavioral challenge ( Warden et al . , 2012 ) . Cortical excitatory synaptic transmission was examined by preparing acute brain slices from PFC 24 hr after vehicle or ketamine injection and applying whole-cell voltage clamp recordings at a holding potential of −65 mV while perfusing with TTX and picrotoxin to isolate miniature AMPAR-mediated excitatory synaptic currents ( mEPSCs ) . These spontaneous synaptic currents result from individual neurotransmitter vesicle fusion events that are action potential independent , and therefore , allow assessment of increases or decreases in synaptic weight ( amplitude ) or changes in synapse number or probability of presynaptic transmitter release ( frequency ) under different experimental conditions . Our measurements showed that a single dose of ketamine increased excitatory synaptic neurotransmission onto layer II/III pyramidal neurons of control animals measured 24 hr after injection and this effect was mimicked and occluded in 2BΔCtx animals . Increased excitatory neurotransmission was measured as an increase in the frequency ( decrease in inter-event interval ) of mEPSCs in ketamine-injected animals relative to saline-injected controls ( Figure 3A–C ) . Importantly , this ketamine-driven increase in mEPSC frequency was occluded in 2BΔCtx animals , which exhibited an increase in average baseline frequency of mEPSC events ( Figure 3A–C ) . Behaviorally naïve animals also exhibited strong increase in mEPSC events 24 hr after ketamine injection , showing that this was not an effect of the behavioral testing ( Figure 3C ) . In further support of a role for GluN2B signaling in ketamine's actions the selective antagonist Ro drove a similar increase in mEPSC events that correlated with decreased immobility in TST ( Figure 3C c . f . 2B ) . The tight correlation between mEPSC frequency and immobility times in the TST strongly supports our hypothesis that excitatory synapse number in PFC contributes to setting basal levels of despair-like behavior . This prompted us to test whether or not this correlation could also be observed in corticosterone-exposed animals . Consistent with our hypothesis , we recorded a strong decrease in frequency of mEPSCs onto layer II/III pyramidal neurons in corticosterone-treated animals , compared to littermate vehicle exposed animals and this effect was absent in 2BΔCtx animals , where baseline frequency of synaptic events was similar to ketamine-injected control animals ( Figure 3C ) . Interestingly , on average , amplitude of mEPSC events in response to corticosterone treatment in either genotype were unchanged ( Cont = 10 . 74 ± 0 . 66 pA; Cont + Cort = 9 . 58 ± 0 . 67 pA n . s . ; 2BΔCtx = 10 . 50 ± 0 . 57 pA; 2BΔCtx + Cort = 10 . 46 ± 0 . 52 pA n . s . ) . 10 . 7554/eLife . 03581 . 005Figure 3 . Increased excitatory synaptic transmission in prefrontal cortex following ketamine injection is occluded in 2BΔCtx animals . ( A ) mEPSC recordings from prefrontal layer II/III cortical pyramidal neurons 24 hr after a single injection of saline ( upper ) or ketamine ( lower ) in control and 2BΔCtx mice . ( B ) Cumulative histograms showing the strong increase in frequency ( decreased inter-event interval ) ( left ) of mEPSC events in control animals by ketamine is occluded in 2BΔCtx littermates . ( C ) Quantification and statistical IEI results are shown for data in ( B ) . The increase in mEPSC frequency was not due to subjecting the animals to the behavioral testing as ketamine significantly suppressed IEIs in behaviorally naïve control animals . Additionally , 2BΔCtx animals showed no change in frequency of events measured after chronic corticosterone treatment , while this caused a strong decrease in event frequency in control animals . ( D ) The increase in event number in 2BΔCtx prefrontal cortical neurons was not correlated with a change in the paired pulse ratio measured by evoking synaptic responses at this synapse ( intracolumnar stimulation indicated by the arrowheads ) . Example traces of evoked responses recorded by whole-cell voltage clamp at 100 ms inter stimulus interval are shown . Quantification across a number of inter-stimulus intervals is presented . ( E ) Ranked mEPSC plot showing the disproportionate increase in small amplitude events ( arrow ) in control ketamine-injected animals compared to saline injected controls . Data values are means ± sem . *p < 0 . 05; **p < 0 . 01; t test with respect to control; n . c . = no significant change . DOI: http://dx . doi . org/10 . 7554/eLife . 03581 . 005 Increased mEPSC frequency could be due to an increase in synapse number or increase in probability of presynaptic neurotransmitter release . Surprisingly , the dramatic increase in mEPSC event frequency in 2BΔCtx animals was independent of a change in paired pulse ratio at intra-columnar II/III synapses ( Figure 3D ) . While we did not test PPR at all synapses , and therefore , cannot rule out changes in presynaptic release at other contacts , our data are consistent with an increase in functional synapse number in line with previous recordings from GluN2B null neurons in hippocampal slices ( Gray et al . , 2011 ) . The increase in synaptic event frequency in vivo was accompanied by a decrease in average mEPSC amplitude ( Cont = 11 . 42 ± 0 . 97 pA vs Cont + Ket = 7 . 79 ± 0 . 98 pA p < 0 . 05 ) . We infer this could either be due to a homeostatic down-regulation of synaptic strength in response to an increase in synapse number , or could be a reflection of insertion of young , immature synapses , which have low AMPAR content ( Harris and Stevens , 1989; Zito et al . , 2009 ) . In support of this latter explanation , we saw a disproportionate increase in small amplitude events when comparing saline and ketamine injected event amplitude distributions in control animals ( Figure 3E ) . Together , these data strongly support our hypothesis that ketamine suppresses GluN2B function in pyramidal cortical neurons leading to an increase in functional excitatory synaptic number in the PFC and a decrease in despair-like behavior . How are changes in synaptic physiology and behavior related to changes in protein synthesis and how does ketamine injection activate translational machinery in these cortical neurons in relation to GluN2B function ? Previous results demonstrated a strong increase in cortical protein levels in response to ketamine however protein synthesis rates were not directly measured and the subunit specificity of NMDAR antagonism was not examined ( Li et al . , 2010; Autry et al . , 2011 ) . Our hypothesis predicts that loss of tonic suppression of the translational machinery by GluN2B-containing NMDARs would increase protein synthesis rates in GluN2B null neurons . To determine whether or not GluN2B-containing NMDARs act to directly suppress protein synthesis in pyramidal cortical neurons , we prepared cultures from WT and global GluN2B KO ( Kutsuwada et al . , 1996 ) cortices and analyzed rates of translation using fluorescence-based non-canonical amino acid tagging ( FUNCAT ) ( Dieterich et al . , 2010 ) . Incorporation of the non-canonical amino acid AHA , which substitutes for methionine in nascent peptides during translation , was used as a measure of protein synthesis rates . After a controlled , 6-hr incubation in AHA-containing methionine-free media , neurons were fixed and AHA was fluorescently labeled . Anti-MAP2 immunostaining was used to delineate dendritic segments for analysis ( Figure 4A ) . We observed a dramatic increase in the rate of protein synthesis in GluN2B KO neurons compared to controls , consistent with our hypothesis ( Figure 4B ) . However , at these ages ( 11–14 DIV ) in rodent cortical cultures both GluN2A and GluN2B protein is expressed ( Li et al . , 1998; Hall et al . , 2007 ) . We therefore wanted to know if the increase in protein synthesis in GluN2B KO cultures was due to loss of GluN2B specifically , or reduction in NMDAR-mediated current generally . To test this , we generated neuronal cultures from cortices of animals in which GluN2B had been genetically replaced by GluN2A ( 2B→2A ) ( Wang et al . , 2011a ) . In homozygous 2B→2A cultures , genetic replacement of GluN2B by GluN2A was not sufficient to restore tonic suppression of translation , confirming that this is a unique function of GluN2B-containing NMDARs ( Figure 4B ) . These data provide direct evidence that GluN2B acts in a unique manner , under conditions of basal activity , to tonically suppress protein synthesis in cortical neurons in vitro . However , if ketamine antagonizes this cellular action , we should be able to increase protein synthesis rates by applying ketamine to control cultures . Indeed , ketamine caused a significant increase in AHA signal in dendrites of control neurons compared with non-treated cells and this increase was mimicked in 2BΔCtx neurons in a manner that was sensitive to rapamycin treatment ( Figure 4C ) . The inability of ketamine to further increase protein synthesis rates in 2BΔCtx neurons is highly suggestive that GluN2B-containing receptors are required for the effects of ketamine ( Figure 4C ) . Next , we asked what cellular mechanisms in vivo were responsible for these changes in protein synthesis in response to ketamine , and whether or not these changes could be mimicked by GluN2B loss of function , in vivo , in 2BΔCtx animals . 10 . 7554/eLife . 03581 . 006Figure 4 . Changes in protein synthesis rates , synaptic protein expression , and phosphorylation in response to ketamine injection are occluded in GluN2B null neurons . ( A ) FUNCAT was used to measure rates of protein synthesis in cortical neurons . AHA signal intensity in MAP2-stained dendrites reveals relative levels of new protein synthesized over a 6-hr period . Examples of 14 DIV cortical neurons from GluN2BKO and 2B→2A as well as WT control neurons treated with the protein synthesis inhibitor anisomycin ( Aniso ) or transcription inhibitor actinomycin-D ( ActD ) . ( B ) Combined data show significant increase in protein synthesis rates in GluN2BKO and 2B→2A neurons and suppression of basal levels by anisomycin . ( C ) Combined data showing the significant increase in protein synthesis rates evoked by ketamine , occlusion of this increase in 2BΔCtx and sensitivity of the AHA signal to rapamycin in 2BΔCtx neurons . ( D–H ) Cortical synaptoneurosomes from animals following saline or ketamine injection . Western blot analysis showed the presence of mTOR in synaptoneurosomes as well as basal levels of phosphorylated protein ( p-mTOR ) and expression of BDNF , Rheb , and the synaptic proteins , SAP-102 , GluA1 , and GluA2 . ( D ) Expression levels , and phosphorylation status of mTOR , following ketamine injection ( 1 , 3 , and 6 hr post injection ) compared to saline-injected controls . ( E ) p-mTOR measured relative to total mTOR levels demonstrating time-dependent changes in phosphorylation status in response to ketamine . ( F ) Synaptic protein expression is shown in relation to levels of actin at 1 , 3 , or 6 hr post ketamine normalized to saline-injected controls . Increased levels of GluA1 , SAP102 , and BDNF as well as decreased rheb expression seen at 3 and 6 hr post injection . ( G ) Example of western blots at 6 hr post-injection for both control and 2BΔCtx genotypes . ( H ) Quantification of the data showed significant increases in synaptic proteins in response to ketamine ( k ) , consistent with ( A–C ) , but also revealing occlusion of these increases in SAP-102 , GluA1 and p-mTOR in 2BΔCtx animals . Data values are means ± sem . **p < 0 . 01; ***p < 0 . 001 t test with respect to WT; n . c . = no significant change . DOI: http://dx . doi . org/10 . 7554/eLife . 03581 . 006 To examine the mechanisms through which ketamine regulates the translational machinery in vivo and determine its regulation by GluN2B , we prepared synaptoneurosomal fractions from both control and 2BΔCtx animals at increasing times after injection and performed quantitative western blot analysis . Western blot analysis revealed mTOR expression in these biochemical synaptic fractions , predicting a close association between mTOR and synaptic sites ( Scheetz et al . , 2000; Li et al . , 2010 ) ( Figure 4D ) . In response to ketamine injection , we observed a rapid yet transient increase in phosphorylated mTOR ( p-mTOR ) , relative to total mTOR ( Figure 4D , E ) . Interestingly , the late-phase decrease in mTOR activation we observed coincided with decreased expression of the upstream mTOR activator Rheb , suggesting a possible negative-feedback mechanism ( Figure 4D , F ) . Expression of a number of synaptic proteins in these synaptic fractions was increased , including brain-derived neurotrophic factor ( BDNF ) , synapse associated protein 102 ( SAP-102 ) , and the AMPAR subunit GluA1 ( Figure 4F ) . This is consistent with the increase in excitatory synapse number predicted by our mEPSC analysis . BDNF has been shown to regulate increases in synapse number and synapse unsilencing ( Itami et al . , 2003; Shen and Cowan , 2010 ) and a rapid increase in the number of immature synaptic connections is consistent with the increased expression of SAP-102 and GluA1 protein , which are associated with developmentally young synapses ( Pellegrini-Giampietro et al . , 1992; Martin et al . , 1998; Sans et al . , 2001; van Zundert et al . , 2004 ) . Because 2BΔCtx both mimics and occludes the behavioral actions of ketamine , alterations in protein expression that are causally associated with these effects should be mimicked and occluded by GluN2B loss of function . In 2BΔCtx synaptoneurosomes , we observed increased baseline levels of BDNF , SAP-102 , GluA1 , and p-mTOR and increased basal levels of expression of these proteins ( and phosphorylation of mTOR ) occluded any further increase in response to ketamine injection in 2BΔCtx animals ( Figure 4G , H ) . In addition to p-mTOR , we also observed strong increases in p-P70S6K in both 2BΔCtx PFC lysates and 2A→2B lysates from P50–P70 animals compared to controls ( Cont = 100 ± 13 . 3% vs 2BΔCtx = 217 ± 11 . 4% p < 0 . 01; WT = 100 ± 34% vs 2B→2A = 318 ± 55% p < 0 . 05 ) . These data strongly suggest that these increases in protein expression are driven by mTOR activation and causally related to the actions of ketamine and that these levels are regulated by GluN2B-containing NMDARs . The above data support a predicted requirement for GluN2B-mediated signaling in the actions of ketamine , yet an important question remains , how does a non-selective NMDAR antagonist like ketamine produce a seemingly selective effect on GluN2B-mediated signaling ? Despite the fact that GluN2B-selective antagonists have been shown to have antidepressant-like actions in both clinical and preclinical studies , the exact role of GluN2B-containing NMDARs has remained unclear . From our FUNCAT data , we infer that GluN2B-containing NMDARs function under non-stimulated conditions to tonically suppress protein synthesis in a manner that is not rescued by replacement with GluN2A . From these observations , we hypothesized that GluN2B-containing NMDARs might be selectively activated under basal , non-stimulated , conditions . If this receptor pool is uniquely tonically active , this would explain why either GluN2B-selective or non-selective pan-NMDAR antagonists have the same effect of promoting protein synthesis . Consistent with this idea , spontaneous , non-synchronized transmitter release has been shown to be sufficient in vitro to limit protein synthesis in cortical neurons including through suppression of mTOR ( Sutton et al . , 2007; Wang et al . , 2011a ) . However , in addition to synaptically released glutamate , there is strong evidence supporting the presence of persistent , low-level ambient glutamate in vivo and in brain slices ( Meldrum , 2000 ) . Interestingly , changes that are consistent with increased ambient glutamate , including glial retraction and decreased expression of glutamate transporters , have been demonstrated in human postmortem studies of depressed patients and observed in animal models of depression ( Moghaddam et al . , 1994; Boudaba et al . , 2003; Rajkowska and Miguel-Hidalgo , 2007; Banasr and Duman , 2008; Zink et al . , 2010 ) . We therefore wondered if ambient glutamate could activate a tonic NMDAR-mediated current in cortical neurons and whether or not this current is mediated selectively by GluN2B-containing NMDARs ? To examine this , we first perfused WT-cultured cortical neurons with TTX , to block synchronized transmitter release , and DNQX and picrotoxin to block AMPAR and GABAAR currents , respectively , then washed in 0Mg2+ ACSF . This revealed a tonic current evident as a reliable increase in the signal noise of the holding current ( root mean square signal—RMS ) at Vhold −65 mV . The increase in tonic current was suppressed by APV , demonstrating its dependence upon NMDAR activation , and by ifenprodil , supporting the prediction that GluN2B-containing NMDARs are selectively activated by ambient glutamate ( Figure 5A ) . To further verify the subunit specificity of this NMDAR-mediated response , we next examined its presence in our genetically modified neurons . Strikingly , in a manner consistent with selective activation of GluN2B-containing NMDARs , the ambient glutamate-activated tonic current was absent in GluN2B null cultures , as well as in cells where NMDARs levels have been restored but GluN2B was genetically replaced by GluN2A ( Figure 5A ) . This tonic current could also be evoked in acute cortical slices and was completely antagonized by ketamine application or by the GluN2B selective antagonist ifenprodil but was not suppressed by Zn2+ , which is a strong and selective GluN2A antagonist at low concentrations ( 250 nM ) ( Figure 5B ) . We also saw that the tonic current activated by ambient glutamate in 0Mg2+ was lost in 2BΔCtx slices ( Figure 5B ) . Finally , we generated mosaic GluN2B knockout conditions in PFC by in vivo injection of a GFP-Cre expressing AAV virus under the CaMKII promoter into homozygous conditional GluN2B KO animals and recorded tonic current in acute brain slices . In these GluN2B null neurons , we also noted that 0Mg2+ conditions generated no increase in RMS signal yet GluN2A-containing receptors could be activated under these conditions as perfusion of these slices with 25 µM NMDA in 0Mg2+ saline , resulted in a strong increase in RMS signal ( baseline ACSF = 1 . 29 ± 0 . 18 , n = 7; 0Mg2+ = 1 . 25 ± 0 . 15 , n = 11; 0Mg2+ + 25 µM NMDA = 2 . 88 ± 0 . 60 , n = 5; p < 0 . 001 c . f . baseline ) . Furthermore , the average holding current increased to 352 ± 103 . 9 pA ( n = 3 ) after 100 s of continued slow perfusion with NMDA . These data confirm that ambient glutamate is sufficient ( both in acute brain slices and primary neuronal cultures ) to activate a uniquely sensitive population of GluN2B-containing NMDARs on cortical neurons that can be antagonized by ketamine . 10 . 7554/eLife . 03581 . 007Figure 5 . Activation of GluN2B-containing NMDARs by ambient glutamate regulates mEPSC frequency and depression-like behavior . Whole-cell recordings of cortical pyramidal neurons ( Vhold −65 mV ) under control conditions ( ACSF ) and in 0Mg2+ ACSF were used to test the subunit contribution of the NMDA receptor pool activated by ambient glutamate . ( A ) Perfusion of 0Mg2+ ACSF uncovered a tonic current in cultured cortical neurons ( 14 DIV ) evidenced by an increase in baseline current noise that could be reversed by the NMDAR antagonist APV . This tonic NMDAR-mediated current was completely absent in GluN2B null neurons , including those in which GluN2B had been genetically replaced by GluN2A ( GluN2B→2A ) . ( B ) Ambient glutamate also evoked a tonic current in acute brains slices that was suppressed by ketamine and was absent in 2BΔCtx slices ( P21 ) . ( C–D ) Manipulation of tonic activation of GluN2B-containing NMDARs alters mEPSC frequency in cortical cultures and pyramidal neurons in vivo , as well as expression of depression-like behavior . ( C ) Acute treatment of cultures with the glutamate transporter antagonist dl-TBOA enhanced the evoked current , while NDGA , which enhances glutamate transporter function , decreased this current . ( D ) Elevating ambient glutamate by blocking transporters with dl-TBOA , or decreasing it by enhancing glutamate transporter function with NDGA , or upregulating glutamate transporter expression using ceftriaxone bidirectionally regulated mEPSC frequency in cultured cortical neurons . ( E ) Enhancing glutamate transporter expression by i . p . injection of ceftriaxone resulted in a decrease in mEPSC IEI , and significant antidepressant-like action in control mice . ( F ) Injections of NDGA resulted in a significant decrease in mEPSC IEI in pyramidal neurons and decreased immobility in the TST . The effect of NDGA was mimicked and occluded on both measures in 2BΔCtx mice . Data points are mean ± sem . p-values are *<0 . 05 , **<0 . 01 , ***<0 . 001 , and n values are shown for each experiment . n . c . = no significant change . DOI: http://dx . doi . org/10 . 7554/eLife . 03581 . 007 Next , we used excitatory amino acid transporter ( EAAT ) inhibitors and enhancers to manipulate the tonic current and test its ability to regulate mEPSC frequency . As shown in Figure 5C , pre-treatment with the glutamate transporter antagonist dl-TBOA resulted in a significant increase in the tonic current evoked in 0Mg2+ , while the EAAT enhancer nordihydroguaiaretic acid ( NDGA ) ( Boston-Howes et al . , 2008 ) significantly suppressed the strength of this GluN2B-mediated tonic current . Because our in vivo data predict an inverse relationship between GluN2B activation and synapse number , we then determined whether or not manipulation of this current in cultured neurons alters mEPSC frequency . We pretreated cultures with dl-TBOA , to increase ambient glutamate concentration , or with NDGA or ceftriaxone ( 24 hr and 1 week , respectively ) , to enhance EAAT function and decrease ambient glutamate ( Rothstein et al . , 2005; Rasmussen et al . , 2011 ) . Unfortunately , chronic NDGA application turned out to be lethal to the cultured cells so only ceftriaxone was testable in the chronic experiments . In line with our prediction , the frequency of mEPSC events measured in pyramidal cortical neurons increased in response to 24 hr NDGA treatment and decreased dramatically in response to antagonism of glutamate re-uptake by dl-TBOA ( Figure 5D ) . At this timepoint , however , ceftriaxone caused no significant increase in mEPSC frequency , perhaps owing to the mechanism of its action , which involves increasing translation and membrane insertion of EAATs , rather than direct enhancement of existing transporters . However , after 1 week of treatment ceftriaxone-treated cells showed dramatically increased mEPSC frequency . Cells treated with TBOA , which were otherwise healthy , were silenced after 7 days of treatment ( Figure 5D ) . These data show that regulation of tonic glutamate levels can affect synapse function in cortical networks and that modulation of extracellular glutamate uniquely activates GluN2B-containing NMDARs to regulate excitatory synaptic drive in cortical neurons . Based upon our in vitro results , we therefore hypothesized that enhancing glutamate transporter function should suppress depression-like behavior and increase mEPSC frequency , in vivo . To test this , we injected animals with either NDGA ( acute ) or Ceftriaxone ( chronic ) to increase EAAT function or expression , respectively ( Rothstein et al . , 2005; Mineur et al . , 2007; Boston-Howes et al . , 2008 ) . Supporting our model , we observed decreased depression-like behavior in TST in these experimental animals compared to vehicle injected control animals , as well as a corresponding increase in mEPSC frequency in PFC ( Figure 5E , F ) . Taken together these data strongly supporting a role for tonic activation of GluN2B-containing NMDARs in setting basal levels of despair-like behavior .
Based upon these data , we propose that tonic activation of GluN2B-containing NMDARs suppresses protein synthesis in cortical pyramidal neurons and loss of this results in increased protein synthesis , an increase in the number of excitatory inputs onto pyramidal neurons in PFC , and a decrease in despair evidenced , in mice , by increased motivation in TST and decreased anxious behavior in EPM . Our data suggest this involves regulation of mTOR downstream of GluN2B . Consistent with our hypothesis that these receptors are involved in the rapid actions of ketamine , restricted genetic removal of GluN2B from cortical pyramidal neurons decreased basal despair-like behavior , occluded the actions of ketamine , and suppressed the increase in depression-like behavior associated with chronic exposure to corticosterone . These studies extend our understanding of the mechanisms underlying the rapid anti-depressant actions of ketamine while providing a novel genetic animal model for depression studies . Importantly , they provide a novel mechanistic framework for interpreting the relationships between ambient glutamate , basal GluN2B-signaling , excitatory cortical neurotransmission , and behavioral despair . Conventional antidepressant treatments are effective in approximately one-third of major depressed patients and in the responding population standard selective serotonin reuptake inhibitor ( SSRI ) therapy exhibits a pronounced delay to efficacy ( Trivedi et al . , 2006 ) . This observation led to the ‘initiation and adaptation’ hypothesis of depression treatment ( Hyman and Nestler , 1996 ) . According to this theory , treatment initiates an adaptive response leading to the delayed , yet consequential , effects of pharmacological intervention . This may involve alterations in NMDAR function . Consistent with a role for NMDARs in MDD , inescapable stress has been shown to lead to disruptions in NMDAR-dependent synaptic plasticity concurrent with induction of behavioral depression-like behavior ( Trullas and Skolnick , 1990 ) . Moreover , SSRI treatment can lead to an adaptation of the NMDAR-mediated response ( Skolnick et al . , 1996 ) , decreases in expression of NMDAR subunits ( Nowak et al . , 1996 ) and expression of NMDAR encoding mRNAs ( Boyer et al . , 1998 ) . The mechanism through which NMDAR antagonism exerts rapid antidepressant actions has therefore become an important direction of study ( Duman et al . , 2012 ) . A critical question is how does suppression of NMDAR function promote protein synthesis to enhance synaptic function ? Previous explanations have proposed circuit-level hypotheses in which NMDAR antagonism preferentially acts upon inhibitory GABAergic neurons leading to an indirect increase in firing in principal neurons , release of BDNF , and increased synaptic strength ( Stefani and Moghaddam , 2005; Duman et al . , 2012 ) . Our data suggest that ketamine acts directly on principal cortical neurons to remove basal suppression of mTOR-mediated protein synthesis by antagonism of GluN2B-containing NMDARs . In support of this , our data reveal strong suppression of depression-like behavior in mice lacking GluN2B specifically in pyramidal cortical neurons . These two mechanistic explanations are not mutually exclusive and could actually work together: initial increases in protein synthesis due to suppression of NMDAR function activates mTOR and promotes an increase in synapse number in PFC while increased network activity , increased glutamate release , or repeated activation of these contacts by spontaneous activity , might be required for stabilization ( Moghaddam et al . , 1997; McKinney et al . , 1999 ) . Consistent with this model , it is known that antagonism of AMPARs , which would prevent stabilization of new contacts , blocks the antidepressant actions of ketamine ( Maeng et al . , 2008; Koike et al . , 2011 ) . We also have preliminary data to support this idea as we find that the AMPAR antagonist DNQX does not block the rapid increase in FUNCAT signal ( i . e . , increased protein synthesis ) after ketamine exposure but does block the increase in spine density ( i . e . , stability of contacts ) in , in vitro cortical cultures , 24 hr after exposure to ketamine . Also in line with this mTOR based model is the understanding that GSK-3 inhibition promotes ketamine's actions via augmentation of mTORC1 activation , while constitutively active GSK-3 mice are insensitive to ketamine's antidepressant-like actions ( Li et al . , 2002; Beurel et al . , 2011; Liu et al . , 2013 ) . It will be interesting in future experiments , therefore , to test the relative roles of these pathways in regulating excitatory synapse number in PFC and behavior while exploring potential links between GluN2B and GSK-3-mediated signaling . In terms of additional downstream effectors , we have also shown recently that protein synthesis can be suppressed by SynGAP signaling via inhibition of mTOR and that this is unique to GluN2B-containing NMDARs ( Wang et al . , 2013 ) . Examining the time-course of mTOR activation in synaptic fractions of cortical lysates following in vivo ketamine injection , we observed that the initial increase in mTOR activation is transient . Interestingly , we also observed a decrease in the mTOR upstream activator Rheb in ketamine-treated cortex 3–6 hr after ketamine injection . Rheb is an immediate early gene and a strong activator of the mTOR kinase ( Yamagata et al . , 1994; Garami et al . , 2003; Laplante and Sabatini , 2012 ) . This suggests that Rheb itself may be regulated at a translational level , which is supported by the observation that Rheb mRNA is present in forebrain neuronal dendrites ( Cajigas et al . , 2012 ) . Additional studies will be required to determine the full complement of proteins whose expression is altered in response to ketamine . Our data show increased levels of SAP-102 and GluR1 following ketamine injection , suggesting a rapid increase in new , excitatory synapses in layer II/III of prefrontal cortex since these proteins are associated with young synapses ( Pellegrini-Giampietro et al . , 1992; Martin et al . , 1998; Sans et al . , 2001; van Zundert et al . , 2004 ) . BDNF signaling and mTOR activation enhance synaptogenesis and promote synapse unsilencing in cortex ( Itami et al . , 2003; Luikart and Parada , 2006; Hoeffer and Klann , 2010; Shen and Cowan , 2010 ) . In addition , BDNF can enhance pre-synaptic function ( Henry et al . , 2012 ) . Our biochemical data show mTOR is localized close to synapses , as evident by its presence in our synaptoneurosome preparations . This is in line with previous reports ( Scheetz et al . , 2000; Li et al . , 2010 ) and implies that protein synthesis , in response to ketamine , might be occurring locally in dendritic compartments . Local protein synthesis maintains proper levels of synaptic strength in cortical and hippocampal neurons and this regulation has been implicated in regimes of homeostatic synaptic plasticity ( Ju et al . , 2004; Sutton et al . , 2006; Aoto et al . , 2008; Wang et al . , 2011a ) . In light of this , other cellular regulators of dendritic protein synthesis and homeostatic synaptic plasticity need to be examined in relation to ketamine's actions , including mRNA binding proteins FMRP ( Ashley et al . , 1993 ) , CPEB ( Richter , 2007 ) , and translational regulators retinoic acid ( Wang et al . , 2011b ) , and eIF4E ( Gkogkas et al . , 2012 ) . Interestingly , GluN2B knockout induced by CaMKII-promoter-driven Cre expression results in a weaker phenotype than the NEX-Cre phenotype reported here ( Von Engelhardt et al . , 2008; Kiselycznyk et al . , 2011 ) . This phenotypic discrepancy , if not due to sex , age , or mouse strain differences , is most likely due to differences in the spatial expression pattern of Cre-recombinase protein . While NEX-Cre results in developmental excision of GluN2B and CaMKII-promoter driven Cre does not ( Tsien et al . , 1996; Goebbels et al . , 2006 ) , this is unlikely to be solely responsible for the different phenotypes since the antidepressant actions of ketamine are evoked by acute treatment ( Preskorn et al . , 2008 ) . This suggests that the specificity of the genetic knockout in cortical structures is responsible for this discrepancy . Indeed , cortical lysates from CaMKII-driven Cre animals contain more residual GluN2B protein , supporting the interpretation of a wider distribution of Cre expression in 2BΔCtx animals ( Brigman et al . , 2010 ) . Another notable difference is that Cre-driven gene excision is restricted to the dorsal hippocampus in the CaMKII animal ( Tsien et al . , 1996 ) . Suppression of GluN2B function in the ventral hippocampus of Nex-Cre based 2BΔCtx animals ( Figure 1C ) likely contributes to the behavioral phenotype we observe as activity in ventral dentate strongly regulates anxiety ( Kheirbek et al . , 2013 ) . Future experiments will require more precise targeting of specific brain regions and cell types to determine their role in aspects of the GluN2B KO phenotype . Of course , potential developmental changes induced by GluN2B knockout should not be ruled out especially in light of the importance of circuit maturation changes in vulnerability to mood disorders ( Stefani and Moghaddam , 2005; Ansorge et al . , 2007 ) . The absence of a positive effect on sucrose intake in the SPT in the 2BΔCtx animals is also notable , as this hedonic behavior is considered a domain of the depression-like phenotype in preclinical models . Interestingly , recent elegant experiments have shown a critical and specific role for synaptic transmission in the nucleus accumbens in sucrose intake in SPT in mice ( Lim et al . , 2012 ) . The absence of any Cre expression in this nucleus in the 2BΔCtx mice ( data not shown ) may explain this observation while also supporting the idea that these phenotypic domains can be separated in preclinical models of depression . There is strong evidence supporting the presence of persistent , low-level ambient glutamate both in vivo and in brain slices ( Meldrum , 2000 ) and NMDARs can be activated by ambient glutamate in hippocampal neurons ( Sah et al . , 1989; Papouin et al . , 2012 ) . Our data show that a tonic NMDAR-mediated current can be evoked in cortical neurons both in culture and acute brain slices . Strikingly , this current is absent in GluN2B null cortical neurons and in neurons in which GluN2B has been genetically replaced with GluN2A ( 2B→2A ) . This shows that endogenous GluN2B receptors are more sensitive to ambient glutamate than those containing GluN2A . Indeed , GluN2B-containing NMDARs have higher sensitivity to agonist , a shifted Mg2+ sensitivity ( Mori and Mishina , 1995 ) and higher concentrations of NMDA and d-serine are required to evoke equal current in neurons expressing only GluN2A-containing NMDARs ( Wang et al . , 2011a ) . Furthermore , it is still unclear exactly how endogenous proteins regulate NMDAR function in situ , compared to heterologous expression studies . Levels of ambient , extracellular glutamate are tightly regulated by glutamate transporters especially EAAT2 ( GLT-1 ) , which is highly expressed on glial cells ( Rothstein et al . , 1996; Guo et al . , 2003 ) . While the absolute concentration and role ambient glutamate is highly debated , our data show that basal activation of GluN2B signaling is functionally relevant and can tonically suppress protein synthesis in these neurons ( Figure 4 ) . Furthermore , in support of a role for tonic NMDAR signaling in regulating excitatory synapse function , we show that mEPSC frequency changes in a predicted manner when glutamate transporter function is chronically altered under standard Mg2+ in cortical cultures maintained at 37°C ( Figure 5 ) . Thus , we propose that ambient glutamate suppresses protein synthesis in a GluN2B-dependent manner in order to maintain synapse number in cortical neurons and thereby contribute to setting levels of despair-like behavior . This model is consistent with the strong association between decreased glial/glutamate transporter function and depression . This includes evidence that EAAT expression is decreased by exposure to learned-helplessness in rats ( Zink et al . , 2010 ) , a decrease in glial density observed in the PFC of chronically stressed animals ( Banasr and Duman , 2008 ) , decreased glial density in postmortem assessment of PFC in depressed subjects ( Rajkowska and Miguel-Hidalgo , 2007 ) , glial retraction from synapses induced by stress ( Boudaba et al . , 2003 ) , and induction of depression-like behavior through chemical ablation of glial cells in PFC in mouse ( Banasr and Duman , 2008 ) . The strongest support for this hypothesis comes from experiments in which transporter function is manipulated in vivo . As our data show , acute increase in EAAT function ( NDGA ) and increasing EAAT expression ( ceftriaxone ) in vivo result in decreased immobility in TST and a corresponding increase in frequency of mEPSCs in layer II/III pyramidal neurons of the PFC ( Figure 5E , F ) . NDGA and ceftriaxone have multiple functional effects; however , since they have no published common targets ( other than in enhancing EAAT2-mediated glutamate uptake ) , the most likely explanation we have at this time is that they are both acting through a common pathway to enhance tonic GluN2B activation and cause the resultant downstream effects we observe in behavior . Future studies need to focus on determining the location of the GluN2B-containing receptor pool activated by ambient glutamate and the in vivo conditions under which they are tonically activated , especially under pathophysiological conditions . Since this receptor pool is uniquely tied to regulation of protein synthesis , it could prove to be an important target for antidepressant therapy and other disorders associated with dysregulation of protein synthesis . Recent reports have shown that NMDARs are activated by distinct co-agonists: d-serine for GluN2A and glycine for GluN2B-containing receptors ( Papouin et al . , 2012 ) . While our data show that GluN2B-containing NMDARs are selectively activated by ambient glutamate , they do not distinguish between potential synaptic and extra-synaptic locations of this receptor pool , in addition , we did not examine the role of triheteromeric NMDARs in regulation of protein synthesis and maintenance of excitatory synaptic strength . Regardless of their subcellular location , however , our data strongly implicate a role for NMDARs containing GluN2B in the rapid antidepressant actions of ketamine , via their ability to directly suppress mTOR signaling and limit protein synthesis in principal cortical neurons .
GluN2B principal neuron KO mice were obtained by crossing GluN2B ‘floxed’ mice ( Brigman et al . , 2010 ) and NEX-Cre mice ( Goebbels et al . , 2006 ) . For genotyping GluN2B floxed allele , primers AGG GTT TTA CAT ACC CCA GGC TGC and AGA GGA TCT ACC AGT AAC ATG C were used to produce a 412-bp WT fragment and 326-bp fragment from the floxed allele . NEX-Cre genotyping involved three primers: NEX . 148 s ( GAG TCC TGG AAT CAG TCT TTT TC ) , NEX . as ( AGA ATG TGG AGT AGG GTG AC ) , and Cre . a ( CCG CAT AAC CAG TGA AAC AG ) , producing a 770-bp fragment from WT NEX allele , 525-bp fragment from the NEX-Cre allele . GluN2Bflox/+ : NEXCre/+ mice were crossed to obtain GluN2Bflox/flox : NEXCre/+ mice , which we refer to as GluN2BΔCtx ( or 2BΔCtx ) . NEX+/+ : GluN2Bflox/+ or NEX+/+ : Glun2Bflox/flox mice served as controls . Mice were housed on a 12/12 hr light–dark cycle , at ∼20°C and ∼55% humidity . Regular rodent chow and tap water were available ad libitum . All pups , including controls , were fed daily with Ensure starting at postnatal day 18 ( P18 ) until P25 to increase survival rates of the mutants . Mice were weaned at P30 by gender . Juveniles were group-housed to not >5 animals/cage . Behavioral experiments were performed on male mice aged P50–P70 and these same animals were used to generate all of the data in Figure 3 . Sex was not determined for embryos used to generate cell cultures; however , all of the remaining data were generated from male mice . All protocols were approved by the Tulane University IACUC . Mice were brought to the test room and habituated for at least 1 hr before testing . Videos were taken using a Canon PowerShot A2200 camera . TST was performed as previously described ( El Yacoubi et al . , 2003 ) . Mice were suspended using adhesive tape and video recorded for 6 min . Two mice were tested simultaneously with an opaque screen separating them . Videos were scored offline for immobility blind to genotypes and treatments . Forced swim test was carried out as described ( Porsolt et al . , 1977; El Yacoubi et al . , 2003 ) . Mice were introduced to a cylinder filled with room temperature water . The cylinder was 20 cm in diameter and filled about 12 cm deep to prevent mice using their tails to support themselves . Two mice were tested simultaneously with a screen to separate the cylinders . Videos were recorded from a top-mounted camera for 6 min . After the testing , mice were rubbed dry and placed under an infrared lamp for about 20 min . Immobility in the last 4 min of 6 min test session was scored blind to genotype and treatment . Open field test was conducted using Accuscan open field system . Mice were introduced into the arena at a fixed corner , and they were able to explore the arena freely . Ketamine and saline were administered via i . p . injection and ketamine concentration used was 50 mg/kg , a dose we empirically found to be necessary to elicit anti-despair-like behavior in our control animals . Consistent with other results ( Lindholm et al . , 2012 ) , our data showed that the effectiveness of ketamine on despair-like behavior on P50–P70 mice required higher doses than reported for rats and adult mice at older ages . Corticosterone ( Sigma–Aldrich , St . Louis , MO ) was dissolved in ethanol at 5 mg/ml and further diluted in drinking water to 25 µg/ml . Water containing corticosterone or ethanol vehicle ( 0 . 5% , Veh ) replaced animals' regular drinking water and was available at all times in the home cage , providing the only source of hydration . All food provided was soaked in the same corticosterone containing water . Exposure began between P30 and P40 . After continuous treatment for 20 days , mice were assessed in TST . The same mice were then used for mEPSC recordings . For in vivo injections , ceftriaxone was dissolved in saline and mice were injected for 20 days at 200 mg/kg as per previous reports ( Mineur et al . , 2007 ) and tested in the TST on day 21 which fell between P50 and P70 consistent with the ketamine experiments . NDGA was dissolved in ethanol to 100 mg/ml and then further diluted to 10 mg/ml in olive oil . Olive oil containing 10% ethanol was used as the vehicle control . P50–P70 mice were injected twice 12 hr apart with 0 . 1 ml of either NDGA or vehicle . 12 hr after the second injection animals were tested in the TST . In chronic variable stress paradigm P30–P50 group-housed male mice were stressed in a pseudo-random order for 3 weeks using the following: cold room ( 2 hr at 4°C ) ; shaker , and no bedding ( 1 hr at 50–60 rmp followed by overnight in clean cage with no bedding ) ; warm swim and single housing ( 30°C for 20 min followed by overnight in single-housed cage ) ; cold swim ( 20°C for 10 min ) ; wet bedding ( 2 hr with wet bedding ) . All tissue culture methods have been previously published ( Hall et al . , 2007; Wang et al . , 2011a ) . To validate glutamate-reuptake modulating drugs' ability to alter levels of ambient glutamate , we pretreated cultured cortical cells ( 15–18 DIV ) for 1 hr with 20 µM dl-TBOA ( to suppress EAAT-mediated reuptake ) or 4 µM NDGA ( to enhance EAAT-mediated reuptake—Santa Cruz Biotechnology , Dallas , TX ) . To chronically alter the tonic current , we pre-treated cells for 7 days before recording . Under these conditions , NDGA caused widespread cell death , and so for these recordings , we substituted a different enhancer of EAAT function , ceftriaxone ( 100 µM–Sigma ) . For synaptoneurosome preparations ( generated at P50–P70 ) , we followed the methods of Li et al . ( 2010 ) . Epifluorescence-based imaging was performed per standard protocols , previously described in the reference Hall et al . ( 2007 ) and FUNCAT method was as per that described in the reference Dieterich et al . ( 2010 ) . Stereotaxic bilateral injections into the medial prefrontal cortex were performed on P40- to P50-day-old , homozygous , loxP-based conditional allele containing mice with 400 nl of AAV-CaMKIIa-GFP-Cre virus ( 8 × 1012 genome copy/ml ) . Mice were sacrificed for electrophysiological recordings in acute brain slices 10–14 days later , as detailed below . For acute slice recordings juvenile ( P15–P21 , for ifenprodil sensitivity assessment ) and adult ( P50–P70 , for all other experiments ) control and 2BΔCtx mice were anesthetized with isoflurane and decapitated . Brains were removed and immediately placed into ice-cold ACSF containing high Mg2+ ( 8 mM ) and low Ca2+ ( 0 . 25 mM ) to promote slice health . 350-μm thick coronal slices from the PFC , defined as caudal to the olfactory bulb and rostral to the commissure of the corpus callosum were obtained using a Leica VT1200 vibratome . Slices were transferred to a holding chamber where they were incubated in bicarbonate buffered ACSF at room temperature for at least 45 min before transferring to a recording chamber for whole-cell voltage clamp recording . ACSF solutions were bubbled with 95%O2/5%CO2 at all times to maintain consistent oxygenation and pH . Synaptic activity was recorded from acute brain slices while perfused at room temperature in ACSF . Borosilicate glass pipettes were filled with a cesium-substituted intracellular solution as previously described ( Hall et al . , 2007 ) . Pipette resistances ranged 4–7 MΩ . Series access resistance ranged from 7 to 15 MΩ and was monitored for consistency . Recordings were discarded if leak current rose above 300 pA . In slice recordings , layer II/III pyramidal neurons were targeted for recording . Recorded cells were confined to medial cortex in the more caudal slices ( to target PrL and IL cortex ) and were medial and dorsal in more rostral slices , defined by appearance of forceps minor corpus callosum , observable in live differential interference contrast images . Synaptic responses were evoked using concentric bipolar-stimulating electrodes placed approximately halfway across the cortical layers in coronal orientation to evoke stimulation of intracolumnar axons . Current decay tau values were determined using a single exponential decay function in IgorPro: y = y0 + a ( − ( x − x0/tau ) ) and fitting between the current peak and Δ50 ms . | Depression is the leading cause of disability worldwide , with hundreds of millions of people living with the condition . The ‘gold standard’ for depression treatment involves a combination of psychotherapy and medication . Unfortunately , current antidepressant medications do not help everyone , waiting lists for psychotherapy are often long , and both normally take a number of weeks of regular treatment before they begin to have an effect . As patients are often at a high risk of suicide , it is crucial that treatments that act more quickly , and that are safe and effective , are developed . One substance that may fulfill these requirements is a drug called ketamine . Studies have shown that depression symptoms can be reduced within hours by a single low dose of ketamine , and this effect on mood can last for more than a week . However , progress has been hindered by a lack of knowledge about what ketamine actually does inside the brain . Neurons communicate with one another by releasing chemicals known as neurotransmitters , which transfer information by binding to receptor proteins on the surface of other neurons . Drugs such as ketamine also bind to these receptors . Ketamine works by blocking a specific receptor called the n-methyl d-aspartate ( NMDA ) receptor , but how this produces antidepressant effects is not fully understood . The NMDA receptor is actually formed from a combination of individual protein subunits , including one called GluN2B . Now Miller , Yang et al . have created mice that lack receptors containing these GluN2B subunits in neurons in their neocortex , including the prefrontal cortex , a brain region involved in complex mental processes such as decision-making . This allowed Miller , Yang et al . to discover that when the neurotransmitter glutamate binds to GluN2B-containing NMDA receptors , it limits the production of certain proteins that make it easier for signals to be transmitted between neurons . Suppressing the synthesis of these proteins too much may cause depressive effects by reducing communication between the neurons in the prefrontal cortex . Both mice lacking GluN2B-containing receptors in their cortical neurons and normal mice treated with ketamine showed a reduced amount of depressive-like behavior . This evidence supports Miller , Yang et al . 's theory that by blocking these NMDA receptors , ketamine restricts their activation . This restores normal levels of protein synthesis , improves communication between neurons in the cortex , and reduces depression . Understanding how ketamine works to alleviate depression is an important step towards developing it into a safe and effective treatment . Further research is also required to determine the conditions that cause overactivation of the GluN2B-containing NMDA receptors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2014 | GluN2B-containing NMDA receptors regulate depression-like behavior and are critical for the rapid antidepressant actions of ketamine |
In Saccharomyces cerevisiae and in humans , the telomerase RNA subunit is bound by Ku , a ring-shaped protein heterodimer best known for its function in DNA repair . Ku binding to yeast telomerase RNA promotes telomere lengthening and telomerase recruitment to telomeres , but how this is achieved remains unknown . Using telomere-length analysis and chromatin immunoprecipitation , we show that Sir4 – a previously identified Ku-binding protein that is a component of telomeric silent chromatin – is required for Ku-mediated telomere lengthening and telomerase recruitment . We also find that specifically tethering Sir4 directly to Ku-binding-defective telomerase RNA restores otherwise-shortened telomeres to wild-type length . These findings suggest that Sir4 is the telomere-bound target of Ku-mediated telomerase recruitment and provide one mechanism for how the Sir4-competing Rif1 and Rif2 proteins negatively regulate telomere length in yeast .
The ends of linear eukaryotic chromosomes are protected by telomeres , which contain DNA repeats that buffer against shortening caused by the end-replication problem . In most eukaryotes , telomeres are lengthened by the enzyme telomerase ( Greider and Blackburn , 1985 ) . Telomerase is a multi-subunit ribonucleoprotein ( RNP ) complex , containing the telomerase reverse transcriptase ( TERT ) and the telomerase RNA , which contains the template for telomeric repeat synthesis by TERT ( Shippen-Lentz and Blackburn , 1990 ) . The telomerase RNA is more than just a template , however , as it has conserved structures in its core that are required for catalytic activity ( Bhattacharyya and Blackburn , 1994; Tzfati et al . , 2000; Lin et al . , 2004; Mefford et al . , 2013; Niederer and Zappulla , 2015 ) and serves as a scaffold for assembling the telomerase RNP holoenzyme ( Zappulla and Cech , 2004; Zappulla et al . , 2011; Lebo and Zappulla , 2012; Lebo et al . , 2015 ) . Telomeric DNA and neighboring subtelomeric regions are often packaged into heterochromatin . In the yeast Saccharomyces cerevisiae , telomeric silent chromatin is largely composed of the histone deacetylase Sir2 and structural components , Sir3 and Sir4 . Complexes of these three proteins are recruited to telomeric DNA in part by the DNA-binding protein Rap1 , which interacts with Sir3 and Sir4 ( Moretti et al . , 1994 ) . Sir2/3/4 complexes then associate with hypoacetylated H3 and H4 tails from telomeric into subtelomeric regions ( Luo et al . , 2002 ) and can cause silencing of telomere-proximal genes ( Gottschling et al . , 1990 ) . In addition to the Sir2/3/4 complex , telomeric silencing also requires the Ku heterodimer ( Boulton and Jackson , 1998 ) , a highly conserved DNA end-binding complex of the proteins Ku70 and Ku80 ( Yku70 and Yku80 in yeast ) well known for its function in non-homologous end-joining . Ku binds telomeres ( Martin et al . , 1999 ) and has been found to interact with Sir4 in two-hybrid screens as well as by co-immunoprecipitation ( Tsukamoto et al . , 1997; Roy et al . , 2004 ) . Ku also protects the telomeric 5′ end from resection ( Gravel et al . , 1998; Polotnianka et al . , 1998; Bonetti et al . , 2010 ) and is a subunit of the telomerase holoenzyme . Ku binds telomerase RNA both in yeast ( Peterson et al . , 2001; Stellwagen et al . , 2003; Dalby et al . , 2013 ) and in humans ( Ting et al . , 2005 ) . In S . cerevisiae , Ku binds to the tip of a 74-nt hairpin in the 1157-nt telomerase RNA , TLC1 ( Figure 1 ) ( Peterson et al . , 2001; Stellwagen et al . , 2003; Dandjinou et al . , 2004; Zappulla and Cech , 2004; Dalby et al . , 2013 ) . In cells where the tip of this hairpin is deleted ( tlc1Δ48 ) , telomeres shorten by ∼70 base pairs ( Peterson et al . , 2001; Stellwagen et al . , 2003; Zappulla et al . , 2011 ) . This defect can be mostly rescued by inserting a Ku-binding hairpin at other locations within the mutant tlc1Δ48 RNA , whereas inserting additional Ku-binding hairpins into wild-type TLC1 causes progressive telomere hyper-elongation ( Zappulla et al . , 2011 ) . Lack of Ku binding to TLC1 has also been reported to reduce nuclear localization of TLC1 ( Gallardo et al . , 2008 ) and recruitment of telomerase to telomeres ( Fisher et al . , 2004 ) . Also important for telomerase recruitment to telomeres is the protein Est1 . Est1 was the first telomerase subunit identified ( Lundblad and Szostak , 1989 ) and is required for recruiting telomerase to telomeres through an interaction with the single-stranded telomeric DNA-binding protein Cdc13 ( Evans and Lundblad , 1999; Qi and Zakian , 2000 ) . In contrast to Est1 , the mechanism by which Ku recruits telomerase to telomeres has yet to be elucidated . 10 . 7554/eLife . 07750 . 003Figure 1 . Secondary structure model of yeast telomerase RNA . The 48 nucleotides deleted in the tlc1Δ48 allele are highlighted in red . The 74-nucleotide hairpin shown in the orange box was inserted at positions 446 and 1029 ( indicated by the orange arrows ) to create TLC1 ( Ku ) 3 . The TLC1 secondary structure shown is based on previously published models of the core ( Niederer and Zappulla , 2015 ) and of the arms ( Dandjinou et al . , 2004; Zappulla and Cech , 2004 ) , while the Ku crystal structure shown is that of the human Ku70/80 complex ( Walker et al . , 2001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 003 An initial working model for Ku-mediated telomerase recruitment to telomeres was that TLC1-bound Ku simply recruits telomerase to telomeres by also binding telomeric DNA ( Peterson et al . , 2001; Fisher and Zakian , 2005 ) . However , this model has been largely discounted by in vitro binding experiments showing that purified Ku cannot bind DNA and RNA concurrently ( Pfingsten et al . , 2012 ) . It therefore seemed likely to us that Ku recruits telomerase to telomeres by interacting with a telomere-associated protein . Such a protein must bind Ku and associate with telomeres . Knowing that Ku also plays a role in the formation of telomeric silent chromatin , we chose to investigate its binding partner in this process , the protein Sir4 , as a possible candidate . Sir4 associates with telomeres , and as mentioned above , an interaction between Sir4 and Ku has been reported previously ( Tsukamoto et al . , 1997; Roy et al . , 2004 ) . Additionally , sir4Δ cells have been shown to have shortened telomeres ( Palladino et al . , 1993; Askree et al . , 2004; Gatbonton et al . , 2006 ) , although the cause of this phenotype has remained apparently unexplored . Here , we provide genetic evidence suggesting that SIR4 and TLC1-bound Ku promote telomere lengthening through the same pathway and that SIR4 is required for Ku-mediated telomere lengthening . In contrast , the negative regulators of telomerase , Rif1 and Rif2 , which compete with Sir3 and Sir4 for binding to Rap1 ( Moretti et al . , 1994; Wotton and Shore , 1997 ) , appear to inhibit Ku-mediated telomere lengthening . By measuring telomerase recruitment to telomeres by chromatin immunoprecipitation ( ChIP ) , we find that a TLC1 RNA containing three Ku-binding sites , TLC1 ( Ku ) 3 , causes increased telomerase recruitment in wild-type cells . Furthermore , sir4Δ cells display a defect in telomerase recruitment indistinguishable from that of tlc1Δ48 cells , even when expressing TLC1 ( Ku ) 3 . Finally , we show that tethering Sir4 directly to tlcΔ48 RNA restores telomeres to wild-type length , while tethering Sir3 to tlc1Δ48 does not . Together , these results suggest that Ku recruits telomerase to telomeres through its interaction with Sir4 and that this recruitment pathway is counterbalanced by Rif1 and Rif2 .
Although the exact mechanism of Ku-mediated telomerase recruitment remains unclear , a simple model is that Ku recruits telomerase to telomeres by binding a telomere-bound protein . The telomeric silent chromatin protein Sir4 is an attractive candidate for playing this role , since it has been shown to bind Ku and because sir4Δ cells have telomeres 50–150 bp shorter than wild type ( Palladino et al . , 1993; Gatbonton et al . , 2006 ) , a phenotype similar to the ∼70-bp reduction seen in tlc1Δ48 cells ( Peterson et al . , 2001; Stellwagen et al . , 2003; Zappulla et al . , 2011 ) . As a first test of the hypothesis that SIR4 is involved in Ku's function as a telomerase subunit , we accurately measured the length of telomeres in sir4Δ cells and tlc1Δ48 cells , as well as sir4Δ tlc1Δ48 double mutants . We found that telomeres in tlc1Δ48 cells were 85 ± 23 bp shorter than wild type , while those in sir4Δ cells were 53 ± 13 bp shorter than wild type ( Figure 2A , Table 1 ) . When these two mutations were combined to make a double-mutant strain , telomeres were 71 ± 26 bp shorter than wild type , a telomere length defect very similar to that of the tlc1Δ48 single-mutant ( p = 0 . 31 ) . This genetic epistasis suggests that SIR4 promotes telomere lengthening in the same pathway as TLC1-bound Ku . 10 . 7554/eLife . 07750 . 004Figure 2 . SIR4 , Ku , and the Ku-binding site in TLC1 are in the same telomere-lengthening pathway . ( A ) Deleting SIR4 in tlc1Δ48 cells does not cause further telomere shortening . A tlc1Δ pTLC1-URA3 strain and an isogenic sir4Δ strain were transformed with CEN plasmids expressing either TLC1 or tlc1Δ48 , and then the pTLC1-URA3 cover plasmid was shuffled out . The cells were serially re-streaked five times , and genomic DNA was isolated and analyzed by Southern blotting . The Southern blot was probed for telomeric sequence and for a 1621-bp non-telomeric XhoI restriction fragment from chromosome IV ( ‘non-telomeric control’ ) used as a relative-mobility control . Pairs of lanes represent independent transformants . Changes in telomere length were quantitated using the Y′ telomere bands as described in ‘Materials and methods’ . Telomere lengths calculated from the two sets of replicates shown were averaged with telomere lengths from four other sets of replicate samples from similar experiments to give the numbers shown , ± standard deviation . The numbers shown here are the same as those in Table 1 . ( B ) Deleting SIR4 in yku80Δ cells does not cause further telomere shortening . A SIR4/sir4Δ YKU80/yku80Δ diploid strain was sporulated , and tetrads were dissected . The haploid spores of a tetratype tetrad were serially re-streaked three times on plates to equilibrate telomere length before Southern blot analysis . The pairs of lanes on the blot shown are different colonies from streak-outs of the haploid spores . Telomere lengths calculated from the two sets of replicates shown were averaged with telomere lengths from a third set of replicate samples to give the numbers shown , ± the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 00410 . 7554/eLife . 07750 . 005Table 1 . Average Yʹ telomere length in sirΔ cells containing TLC1 , tlc1Δ48 , or TLC1 ( Ku ) 3DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 005SIR GenotypeTLC1 GenotypeTLC1tlc1Δ48TLC1 ( Ku ) 3SIR0−86 ± 23†Dysregulated*sir4Δ−53 ± 13†−71 ± 26†−148 ± 36‡sir2Δ§−41 ± 16−50 ± 74−71 ± 26sir3Δ§−51 ± 20−84 ± 14−123 ± 2The weighted-average mobility of the Yʹ telomeric restriction fragments was calculated as described in the ‘Materials and methods’ . The numbers shown are averages of multiple biological-replicate samples ± standard deviation . *Yʹ telomere length was not quantified in this condition because signal from Yʹ telomere restriction fragments overlapped with that from the non-telomeric control fragment . †n = 6 . ‡n = 4 . §n = 2 . To test if this result is , in fact , indicative of related function in telomere-length maintenance between TLC1 , Ku , and SIR4 , and not simply between TLC1 and SIR4 , we performed a similar genetic epistasis experiment with sir4Δ and yku80Δ mutants . Similar to what has been reported previously , we observed that yku80Δ cells supported telomeres 220 ± 14 bp shorter than wild type ( Figure 2B ) ( Gravel et al . , 1998; Askree et al . , 2004; Gatbonton et al . , 2006 ) . However , while deleting SIR4 resulted in a ∼70-bp telomere-length defect in a wild-type background , it appeared to have little effect on telomere length in a yku80Δ background; telomeres in sir4Δ yku80Δ cells were 193 ± 19 bp shorter than wild type . None of the strains in these experiments senesced ( data not shown ) , and telomeres have been reported to shorten by as much as ∼260 bp in other mutants without causing senescence ( Lebo and Zappulla , 2012 ) . Thus , the lack of further telomere shortening in the sir4Δ yku80Δ double-mutant strain relative to yku80Δ is not explained by telomeres already being the shortest-possible length supporting cell growth . These findings suggest that Ku is involved in the same telomere length-maintenance pathway as SIR4 . Inserting an extra Ku-binding hairpin into TLC1 causes progressive telomere hyper-lengthening ( Zappulla et al . , 2011 ) . Furthermore , we have generated a telomerase RNA , TLC1 ( Ku ) 3 , that contains extra Ku-binding hairpins inserted at positions 446 and 1029 . This TLC1 ( Ku ) 3 telomerase RNA accumulates to essentially the same level ( 93 ± 9% ) as wild-type TLC1 ( Figure 3—figure supplement 2 ) . If SIR4 is required for Ku's function in maintaining telomere length as a telomerase subunit , deleting SIR4 should prevent TLC1 alleles with extra Ku-binding hairpins from causing telomere hyper-lengthening . We passaged TLC1 ( Ku ) 3 cells in liquid culture and assessed telomere length over time . TLC1 ( Ku ) 3 caused progressive telomere hyper-lengthening over the course of passaging in addition to some telomere shortening ( Figure 3A ) , similar to TLC1 RNAs with two Ku-binding sites ( Zappulla et al . , 2011 ) . We also probed the Southern blot from Figure 3A for Yʹ telomeric restriction fragments and determined that telomeres in TLC1 ( Ku ) 3 cells range from ∼70 bp shorter than wild type to ∼1000 bp longer after 220 generations , continuing to progressively elongate at a rate of ∼5 bp/generation ( Figure 3—figure supplement 3 ) . This increasingly heterogeneous distribution of telomere lengths in TLC1 ( Ku ) 3 cells could be due to diverse telomere lengths in the population of cells or an abnormality of telomeric DNA structure affecting how it migrates on gels ( e . g . , extremely long single-stranded tails ) . To differentiate between these possibilities , we plated the liquid culture-passaged cells for single colonies and found that telomeres from these clonal isolates were subsets of the heterogeneous liquid-cultured population ( Figure 3—figure supplement 1 ) , a behavior of telomeres that has been reported previously ( Shampay and Blackburn , 1988; Levy and Blackburn , 2004 ) . These results show that the wide variety in the relative mobility of telomeric restriction fragments in the gel is due to a broad distribution of telomere lengths from the population of cells . 10 . 7554/eLife . 07750 . 006Figure 3 . TLC1-bound Ku requires SIR4 to promote telomere lengthening . ( A ) TLC1 ( Ku ) 3 , a TLC1 RNA containing two extra Ku-binding sites , causes both telomere hyper-lengthening and shortening . This experiment was performed as described in Figure 2A , except a tlc1Δ pTLC1-LYS2 rad52Δ strain was used . Additionally , instead of passaging cells on plates , single colonies were inoculated to liquid cultures , which were then serially passaged and harvested at various points throughout the passaging process . ( B ) TLC1 ( Ku ) 3 does not cause telomere hyper-lengthening in sir4Δ cells . This experiment was performed as described in Figure 2A , but the liquid culture passaging method described in Figure 3A was used instead of re-streaking single colonies on plates . ( C ) TLC1 ( Ku ) 3 does not cause telomere hyper-lengthening in sir2Δ or sir3Δ cells , and tlc1Δ48 causes greater telomere shortening in rif1Δ and rif2Δ cells than in wild-type cells . This experiment was performed as described in Figure 3B except that cells were passaged to ∼250 generations by re-streaking on plates rather than passaging in liquid cultures . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 00610 . 7554/eLife . 07750 . 007Figure 3—figure supplement 1 . Three Ku-binding sites in yeast telomerase RNA increase telomere-length heterogeneity . Cells were initially serially passaged in liquid culture as described in Figure 3A and then ∼25 generations before the end of passaging , liquid cultures were plated for single colonies . Genomic DNA was isolated from both the liquid-passaged cultures ( ‘Liq . ’ ) and from cells cultured from the ( clonal ) colonies from solid medium ( ‘Colonies’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 00710 . 7554/eLife . 07750 . 008Figure 3—figure supplement 2 . TLC1 RNA abundance is largely unaffected in TLC1 ( Ku ) 3 cells and is not decreased in sir4Δ cells . Total RNA was isolated from the cells used in the experiment described in Figure 3B and subjected to Northern blot analysis . The pairs of lanes on the Northern blot represent two independent sets of biological replicates . The blot was probed for TLC1 and for the U1 snRNA . Telomerase RNA abundance was normalized to U1 and is expressed relative to the SIR4 TLC1 condition . The values shown are averages of these two replicates and another set biological replicates from a separate Northern blot , ± the standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 00810 . 7554/eLife . 07750 . 009Figure 3—figure supplement 3 . TLC1 ( Ku ) 3 causes Y′-telomere shortening and hyper-lengthening , while deletion of RIF1 or RIF2 causes Y′-telomere hyper-lengthening . The blots from Figure 3A ( A ) and Figure 3C ( B ) were re-probed with a Y′ probe and re-imaged . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 009 Next , we tested if telomere hyper-lengthening caused by TLC1 ( Ku ) 3 is dependent on SIR4 . Whereas TLC1 ( Ku ) 3 caused a combination of telomere hyper-lengthening and shortening in a wild-type SIR4 strain , it did not cause any hyper-elongation in sir4Δ cells ( Figure 3B ) . The average telomeres supported by TLC1 ( Ku ) 3 in a sir4Δ strain were 148 ± 36 bp shorter than those in wild-type TLC1 SIR4 cells ( Table 1 ) . This inability of TLC1 ( Ku ) 3 to cause telomere hyper-lengthening without SIR4 provides further evidence that Sir4 is required for telomerase RNA-bound Ku to promote telomere lengthening in yeast . We also tested whether the other two members of the Sir2/3/4 complex , Sir2 and Sir3 , were required for Ku-mediated telomere lengthening . We observed similar results in sir2Δ and sir3Δ cells , which , like sir4Δ cells , completely lack telomeric silencing , although telomeres supported by TLC1 ( Ku ) 3 in these backgrounds were not quite as short as those supported by TLC1 ( Ku ) 3 in a sir4Δ background ( Figure 3C , Table 1 ) . Of the three members of the Sir2/3/4 complex , only Sir4 has been identified as a binding partner for Ku by screening a two-hybrid library ( Tsukamoto et al . , 1997; Roy et al . , 2004 ) . Deleting SIR2 or SIR3 likely affects Ku-mediated telomere lengthening indirectly by substantially , but not completely , removing Sir4 from telomeres ( Hoppe et al . , 2002 and see Figure 6 ) . In addition to promoting telomere lengthening , Ku binding to TLC1 is also known to increase telomerase RNA abundance ( Mozdy et al . , 2008; Zappulla et al . , 2011 ) . To test if the telomere-length phenotypes we observed are a function of RNA abundance , we assessed telomerase RNA levels by Northern blotting . We found that deleting the 48-nt Ku-binding site in TLC1 ( tlc1Δ48 ) reduced RNA abundance to 61 ± 22% the level of wild type , similar to what has been reported ( Figure 3—figure supplement 2 ) ( Zappulla et al . , 2011 ) . In contrast to TLC1 , RNAs with two Ku-binding sites , which exhibit a ∼20% increase in telomerase RNA abundance ( Zappulla et al . , 2011 ) , TLC1 ( Ku ) 3 showed little change relative to wild type ( 93 ± 9% ) . Although sir4Δ cells display a telomere-length defect very similar to tlc1Δ48 cells , telomerase RNA abundance did not decrease in sir4Δ cells relative to wild type; in fact , it increased ∼twofold in sir4Δ cells , and ∼1 . 5-fold in sir4Δ tlc1Δ48 cells , and remained near wild-type levels in sir4Δ TLC1 ( Ku ) 3 cells . These results suggest that the telomere-length phenotypes shown in Figure 3B are not caused by decreased telomerase RNA abundance . In the process of identifying proteins involved in Ku-mediated telomere lengthening , we tested the effects of tlc1Δ48 and TLC1 ( Ku ) 3 on telomere length in cells lacking Rif1 or Rif2 , negative regulators of telomerase that bind to the same region of Rap1 as Sir3 and Sir4 ( Hardy et al . , 1992; Wotton and Shore , 1997; Teng et al . , 2000 ) . As shown previously , we found that both rif1Δ and rif2Δ cells have hyper-elongated telomeres ( Figure 3C and Figure 3—figure supplement 3 ) . Notably , we found that tlc1Δ48 caused telomeres to shorten by ∼500 bp in rif1Δ and rif2Δ cells . This is a substantially greater effect than the ∼70-bp decrease caused by tlc1Δ48 in a wild-type background , and it suggests that Rif1 and Rif2 inhibit Ku-mediated telomere lengthening . When we introduced TLC1 ( Ku ) 3 into rif1∆ or rif2∆ cells , some telomeres became further hyper-elongated and others became shorter , suggesting that TLC1-bound Ku does not require Rif1 or Rif2 to promote telomere lengthening , in contrast to the requirement we identified for Sir4 as well as Sir2 and Sir3 shown in Figure 3C . A binding interaction between Ku and Sir4 has been reported previously through yeast two-hybrid forward-genetic screens and by co-immunoprecipitation from yeast cell extracts ( Tsukamoto et al . , 1997; Roy et al . , 2004 ) . Using yeast two-hybrid , the N-terminus and C-terminus of Sir4 have been shown to interact with Yku80 and Yku70 , respectively , and two different regions of Yku80 have been shown to be important for binding Sir4 ( Tsukamoto et al . , 1997; Roy et al . , 2004; Ribes-Zamora et al . , 2007 ) . However , these studies do not rule out the possibility that Ku and Sir4 could be interacting indirectly , bridged by another yeast protein . Using purified yeast Ku heterodimer provided by the Cech lab ( Pfingsten et al . , 2012; Dalby et al . , 2013 ) , we tested for the Ku-Sir4 interaction in vitro . [35S]-Sir4 was synthesized by using a rabbit reticulocyte lysate transcription/translation system ( RRL ) spiked with 35S-methionine . Prior to Sir4 protein synthesis , purified Ku heterodimer ( Yku80-Myc•Yku70 ) was also added to the lysate . After Sir4 protein synthesis , Ku was then immunoprecipitated by anti-myc affinity pull-down . The input , unbound supernatant , and bound fraction were resolved on a gel and subjected to autoradiography . As shown in Figure 4 , when Ku heterodimer was omitted from this procedure , a trace amount of radioactive Sir4 was recovered in the bound fraction , indicating a small amount of non-specific Sir4 binding the beads , and when Sir4 template DNA was omitted , no bands were detected . However , when both Ku and Sir4 template DNA were present in the RRL , ∼fivefold more radioactive Sir4 was recovered in the bound fraction than in the no-Ku control , providing evidence for a direct interaction between the Ku heterodimer and Sir4 . To test if this protein–protein interaction was specific , we repeated this experiment with Ku heterodimer that had been boiled before being added to the RRL . In this condition , only trace amounts of Sir4 were recovered in the bound fraction , similar to what was observed when Ku was omitted altogether . These results indicate that Sir4 participates in a specific protein–protein interaction with the Ku heterodimer , as has been suggested by previous in vivo experiments . 10 . 7554/eLife . 07750 . 010Figure 4 . Purified Ku binds Sir4 in vitro . 35S-methionine-labeled Sir4 was synthesized in vitro in a rabbit reticulocyte lysate transcription/translation system ( RRL ) to which purified Ku heterodimer , bearing a 2myc epitope on the C-terminus of Yku80 , was added . After Sir4 synthesis , the RRL was subjected to anti-myc immunoprecipitation . The input , unbound , and bound fractions were run on an SDS polyacrylamide gel , which was imaged by autoradiography . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 010 Having shown that SIR4 is required for TLC1-bound Ku to promote telomere lengthening , we next tested if SIR4 is also required for Ku-mediated telomerase recruitment . We assessed the level of telomerase recruitment to telomeres in wild-type and sir4Δ cells expressing either TLC1 , tlc1Δ48 , or TLC1 ( Ku ) 3 by performing ChIP on myc-tagged telomerase catalytic subunit , TERT ( Est2 ) . Quantitative real-time PCR ( qPCR ) was then used to determine the enrichment of a telomere-proximal locus relative to a telomere-distal control locus ( Sabourin et al . , 2007 ) . We assayed for telomerase recruitment at telomeres VI-R and XV-L and observed highly similar results for both of these chromosome ends ( Figure 5A , B ) . We observed that TERT enrichment at these telomeres in tlc1Δ48 cells was reduced to 15% of wild type , similar to what has been reported previously ( Fisher et al . , 2004 ) . In contrast , there was a ∼10-fold increase in TERT at telomeres in TLC1 ( Ku ) 3 cells relative to wild type . However , in a sir4Δ background , enrichment of TERT at telomeres was decreased relative to wild type , regardless of which TLC1 allele was expressed . The level of TERT at telomeres in sir4Δ TLC1 and sir4Δ tlc1Δ48 cells was reduced to 15% of wild type , and this is indistinguishable from what was observed in SIR4 tlc1Δ48 cells ( p = 0 . 80 and p = 0 . 91 , respectively ) . In sir4Δ TLC1 ( Ku ) 3 cells , telomeric TERT enrichment was decreased to 35% the level of wild type , which is also similar to our observations in SIR4 tlc1Δ48 cells ( p = 0 . 11 ) . These telomerase recruitment results provide molecular evidence that SIR4 is required for Ku-mediated telomerase recruitment to telomeres . 10 . 7554/eLife . 07750 . 011Figure 5 . Ku-mediated telomerase recruitment to telomeres requires SIR4 . ( A , B ) In strains similar to those used in Figure 3B , TERT ( Est2 ) was expressed from its endogenous locus bearing a C-terminal 13myc tag , separated by an 8-glycine linker . TLC1 , TLC1Δ48 , and TLC1 ( Ku ) 3 were expressed as in Figure 3B , but cells were not passaged after loss of the pTLC1-URA3 cover plasmid . Cells were crosslinked and subjected to chromatin immunoprecipitation ( ChIP ) using the myc epitopes on TERT , as described ( Fisher et al . , 2004 ) . Telomeric enrichment was measured using quantitative real-time PCR ( qPCR ) amplicons close to telomere VI-R ( A ) and telomere XV-L ( B ) . An amplicon at the ARO1 locus was used as a non-telomeric control locus . The thick horizontal lines on the graphs represent averages of three to five independent biological replicates , which themselves are indicated by black dots . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 011 We also performed ChIP for TERT in sir3∆ and sir2∆ backgrounds . Compared to wild-type cells , TERT enrichment at telomeres was reduced to 16% in sir3Δ TLC1 cells , to 12% in sir3Δ tlc1Δ48 cells , and to 39% in sir3Δ TLC1 ( Ku ) 3 cells ( Figure 5A ) , similar to our observations in sir4Δ cells described above . TERT enrichment at telomeres was also decreased in a sir2Δ background relative to wild type but not as extensively as in a sir3Δ or sir4Δ background . In sir2Δ cells expressing TLC1 , tlc1Δ48 , or TLC1 ( Ku ) 3 , TERT enrichment at telomeres was reduced to 44% , 21% , or 86% of wild type , respectively . These data suggest that Sir3 , and to a lesser degree Sir2 , is also important for Ku-mediated telomerase recruitment to telomeres in addition to Sir4 . We have shown that Sir2 and Sir3 are important for Ku-mediated telomere lengthening and telomerase recruitment , but , unlike Sir4 , neither Sir2 nor Sir3 has been shown to bind Ku . The simplest explanation is that Sir2 and Sir3 affect Ku-mediated telomerase recruitment indirectly through altering Sir4 association with telomeres , particularly since the amount of telomere-bound Sir4 has been shown to decrease greatly , but not completely , in the absence of Sir2 or Sir3 ( Hoppe et al . , 2002 ) . We have also shown that the proteins Rif1 and Rif2 function to inhibit Ku-mediated telomere lengthening . Because Rif1 and Rif2 compete with Sir3 and Sir4 for binding to Rap1 ( Moretti et al . , 1994; Wotton and Shore , 1997 ) , this inhibition of Ku-mediated telomere lengthening could be explained by there being more Sir4 bound to telomeres in the absence of Rif1 or Rif2 . To test this hypothesis , we performed ChIP on myc-tagged Sir4 in rif1Δ and rif2Δ cells as well as in sir2Δ and sir3Δ cells and used real-time quantitative PCR to measure fold telomeric enrichment . Similarly to what has been shown previously , we observed that Sir4 telomeric enrichment was decreased in sir2Δ and sir3Δ cells , to 26% and 17% of wild-type levels , respectively ( Figure 6 ) . In contrast , Sir4 telomeric enrichment was increased ∼2 . 5-fold in rif1Δ cells and ∼1 . 5-fold rif2Δ cells . These results suggest that Sir2 , Sir3 , Rif1 , and Rif2 affect Ku-mediated telomerase recruitment and telomere lengthening by affecting the amount of Sir4 bound to telomeres . 10 . 7554/eLife . 07750 . 012Figure 6 . Sir4 binding to telomeres is decreased in sir2Δ and sir3Δ cells and increased in rif1Δ and rif2Δ cells . Sir4 bearing a C-terminal 13myc tag on an 8-glycine linker was expressed from its endogenous chromosomal gene locus . Cells were crosslinked and subjected to ChIP using the myc epitopes . Telomere VI-R enrichment was measured using real-time quantitative PCR as in Figure 5A . The thick horizontal lines on the graph represent averages of three independent biological replicates indicated by black dots . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 012 Although disruption of the Est1-Cdc13 telomerase recruitment pathway results in an ever-shortening telomere phenotype , this can be rescued by tethering Cdc13 to telomerase through a protein fusion with TERT , effectively bypassing the need for Est1 in telomerase recruitment ( Evans and Lundblad , 1999 ) . Similarly , if Sir4 is in fact Ku's binding partner in telomerase recruitment , tethering Sir4 directly to telomerase RNA could rescue the short-telomere phenotype of tlc1Δ48 cells . To test this , we tagged the sole chromosomal copy of SIR4 with sequence encoding two tandem copies of the MS2 coat protein ( MS2CP ) and expressed either TLC1 or tlc1Δ48 with MS2 RNA hairpins , which have been employed previously to target proteins to telomerase RNA in yeast ( Gallardo et al . , 2011; Lebo et al . , 2015 ) . We passaged these cells for approximately 125 generations and then assessed telomere length . As shown in Figure 7 , tethering Sir4 to wild-type TLC1 resulted in telomeres ∼30-bp longer than the wild-type no-tag control ( compare lanes 22 and 23 to 2 and 3 ) , and tethering Sir4 to tlc1Δ48 resulted in approximately wild-type length telomeres ( lanes 24 and 25 ) . To test the specificity of telomere length in tlc1∆48 cells being rescued by tethering Sir4 to the RNA , we also performed the same tethering experiment with Sir3 , a telomeric silencing protein which has not been shown to bind Ku . When Sir3 was tethered to tlc1Δ48 ( lanes 16 and 17 ) , telomeres remained 107-bp shorter than the wild-type no-tag control . The MS2CP tags on Sir3 caused a small amount of telomere shortening ( lanes 10 and 11 ) , and tethering Sir3 to wild-type TLC1 ( lanes 14 and 15 ) restored telomeres to approximately wild-type length , but , again , this was dependent on the 48-nt TLC1-binding site for Ku ( lanes 16 and 17 ) . In summary , the finding that specifically tethering Sir4 to Ku-binding-defective tlc1Δ48 RNA restores wild-type length telomeres provides direct support for Sir4 being the telomere-associated factor required for Ku-mediated telomerase recruitment . 10 . 7554/eLife . 07750 . 013Figure 7 . Tethering Sir4 to the tlc1Δ48 RNA restores telomeres to wild-type length . Using the same tlc1Δ pTLC1-URA3 strain background from Figure 2A , Sir3 and Sir4 were expressed from their endogenous loci bearing C-terminal ( MS2CP ) 2 tags , separated by an 8-glycine linker . These strains were transformed with CEN plasmids containing either TLC1 , tlc1Δ48 , TLC1 ( MS2 ) 10 , or tlc1Δ48 ( MS2 ) 10 . Cells were then cured of the pTLC1-URA3 cover plasmid and passaged as in Figure 2A . Each pair of lanes represents two independent biological replicates , and the relative telomere-length values are averages of the two replicates . In the no MS2 coat protein ( MS2CP ) tag and Sir4- ( MS2CP ) 2 conditions , values from a third set of replicates were included in the average , allowing for standard deviation to be calculated . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 013
Telomerase faces formidable challenges in binding and extending telomeres in the nucleus . Obstacles include the enzyme's extremely low concentration ( Mozdy and Cech , 2006 ) , the short period of time when telomerase has to act at the end of S phase ( Diede and Gottschling , 1999 ) , and the fact that chromosome ends are likely difficult to access due to telomeric heterochromatin ( Guertin and Lis , 2013 ) . Considering these impediments , it becomes clearer why telomerase would have multiple recruitment pathways assisting in providing enzyme access to telomeres . Furthermore , carefully regulating telomerase activity is critical – it is upregulated in 90% of cancers ( Shay and Bacchetti , 1997 ) and reduced in telomere syndromes ( Armanios and Blackburn , 2012 ) – and multiple pathways provide opportunities for layers of regulatory control . In S . cerevisiae , the primary telomerase recruitment pathway is essential and requires the Est1 telomerase subunit binding to the telomere-specific DNA-binding protein Cdc13 ( Evans and Lundblad , 1999 ) . In humans , it has been shown recently that telomere-bound Pot1•Tpp1 recruits telomerase through the Tpp1 ‘TEL patch’ binding TERT ( Zhong et al . , 2012; Nandakumar and Cech , 2013; Schmidt et al . , 2014 ) . A second recruitment pathway in yeast has been identified that requires the Ku telomerase subunit and its binding to TLC1 ( Peterson et al . , 2001; Stellwagen et al . , 2003; Fisher et al . , 2004 ) . In this study , we provide evidence that Ku recruits telomerase to telomeres in yeast through its binding to the telomeric transcriptional silencing protein Sir4 . There is substantial evidence for the existence of the Ku-mediated telomerase recruitment pathway in S . cerevisiae . Ku binding to TLC1 promotes telomere lengthening ( Peterson et al . , 2001; Stellwagen et al . , 2003; Zappulla et al . , 2011 ) and recruitment of the telomerase catalytic protein subunit to telomeres as assessed by ChIP ( Fisher et al . , 2004 ) . Our current findings support this Ku-mediated recruitment pathway and provide the first evidence that it is achieved by binding the telomeric silencing protein Sir4 . First , we report genetic epistasis between SIR4 and the Ku-binding site in TLC1 with respect to telomere-length maintenance . Second , we show that telomere hyper-elongation caused by a TLC1 RNA containing three Ku-binding sites , TLC1 ( Ku ) 3 , is dependent on SIR4 . Third , using ChIP , we report that deleting SIR4 causes low levels of telomerase catalytic subunit at telomeres and that this low degree of recruitment is indistinguishable from what is observed in tlc1Δ48 cells . Furthermore , TLC1 with two extra Ku-binding sites causes a 10-fold increase in TERT enrichment at telomeres , and this increase in telomerase recruitment requires SIR4 . Finally , tethering Sir4 directly to a TLC1 RNA lacking its Ku-binding site restores telomeres to wild-type length , whereas tethering Sir3 to Ku-binding-defective TLC1 RNA does not . The fact that the loss of the telomere length-promoting function of the Ku-binding site in TLC1 can be rescued by directly tethering Sir4 to TLC1 provides strong evidence that Sir4 protein directly participates in Ku-mediated telomerase recruitment and telomere extension . Based on the results presented here , we propose that Ku mediates telomerase recruitment to telomeres by binding to Sir4 ( Figure 8 ) . Since Ku-mediated telomerase recruitment occurs via Sir4 associating with telomeric DNA-bound Rap1 , regulation of Sir4 association with Rap1 , in turn , can control the ability of the Ku-Sir4 recruitment pathway to assist in telomere lengthening . As shown in Figure 6 , Sir4 association with telomeres is increased in the absence of Rif1 or Rif2 , suggesting that Rif1 and Rif2 inhibit Sir4 binding to telomeres . The Rif1 and Rif2 proteins have been proposed to represent a ‘counting mechanism’ for telomere-length homeostasis in which decreased binding of the Rif proteins – negative regulators of telomerase – leads to increased telomerase recruitment at short telomeres ( Marcand et al . , 1997; Levy and Blackburn , 2004; Teixeira et al . , 2004; Bianchi and Shore , 2007 ) . Our model provides a parsimonious explanation for one way in which Rif1 and Rif2 regulate telomere length; that is , as competitive inhibitors of Sir4 binding to Rap1 , and therefore , the Sir4-Ku telomerase recruitment pathway . The model that the Ku-Sir4 recruitment pathway is subject to Rif protein-mediated negative regulation is supported by our findings and reports in the literature . First , we showed that TLC1 ( Ku ) 3 , a telomerase RNA with three Ku-binding sites , causes many telomeres to become hyper-elongated , which is similar to telomere hyper-elongation exhibited by rif1Δ and rif2Δ mutants . Second , deleting the Ku-binding site in TLC1 greatly reduces the telomere hyper-lengthening observed in rif1Δ and rif2Δ cells ( see Figure 3C ) . Similarly , it has been shown that the hyper-lengthening of telomeres in rif2Δ as well as rif1Δ rif2Δ mutants is greatly reduced when combined with a yku70Δ mutation ( Mishra and Shore , 1999 ) . Regulation of the Ku-Sir4 pathway by Rif1 and Rif2 likely modulates , in turn , the essential Est1-Cdc13 telomerase-recruitment pathway . Ku-mediated telomerase recruitment has been proposed to promote function of the Est1-Cdc13 pathway by increasing Est1 association with telomeres ( Fisher et al . , 2004; Williams et al . , 2014 ) , a mechanism that is complementary to our model for Ku-mediated telomerase recruitment . 10 . 7554/eLife . 07750 . 014Figure 8 . Model for Ku-Sir4 telomerase recruitment to telomeres and its role in telomere-length regulation in Saccharomyces cerevisiae . Telomerase has previously been shown to extend a telomere infrequently and shorter telomeres are preferentially extendable . We propose here that Ku recruits telomerase to telomeres by binding Sir4 . Since it has been shown that Rif1 and Rif2 compete with Sir4 and Sir3 for binding telomere-bound Rap1 , the Ku-Sir4 telomerase recruitment pathway is inhibited by Rif1 and 2 , providing a simple mechanistic explanation for one way in which Rif proteins function to inhibit telomerase action at telomeres . ( A ) Ku recruitment of telomerase via Sir4 is inhibited by Rif1/2 competition for Rap1 binding with Sir4 . In situations where Ku-Sir4-mediated telomerase recruitment does not occur , Est1-Cdc13-mediated telomerase recruitment can still happen , although with low efficiently . ( B ) When telomerase is recruited to a telomere through the Ku-Sir4 pathway , subsequent Est1-Cdc13-mediated recruitment to the end of the telomere becomes more efficient , resulting in increased telomerase extension of telomeres . The counterbalancing of Ku-Sir4 telomerase recruitment and Rif1/Rif2 occlusion of Sir4 binding to Rap1 may represent a system for maintaining telomere-length homeostasis in yeast . DOI: http://dx . doi . org/10 . 7554/eLife . 07750 . 014 Sir4-mediated telomerase recruitment via Ku suggests that there is a relationship between ( semi-stable ) telomeric silencing and telomerase recruitment . Such a relationship has already been suggested by the fact that Rif1 and Rif2 proteins compete with silencing proteins Sir3 and Sir4 for association with telomere-bound Rap1 and inhibit telomerase acting at longer telomeres . Accordingly , we propose that the Ku-Sir4 recruitment pathway tends to occur at shorter telomeres and is part of the negative-feedback loop regulating telomere length homeostasis . This is supported by telomerase preferentially acting at shortened telomeres ( Teixeira et al . , 2004 ) , which tend to have weaker silencing than longer telomeres and have fewer Rif proteins ( Kyrion et al . , 1993; Park and Lustig , 2000 ) . Our results and previous studies show that when telomeric transcriptional silencing is absent due to sir2∆ or sir3∆ mutation , a reduced but detectable amount of Sir4 is found at telomeres by ChIP ( Hoppe et al . , 2002 ) . Thus , in wild-type cells , it may be that Sir4 binds to Rap1 to recruit telomerase via its Ku subunit without having established telomeric silent chromatin at the end . In summary , it seems most likely that short , non-silenced chromosome ends are the ones targeted for extension by the Sir4-Ku telomerase recruitment pathway and that this is an important part of the negative-feedback loop that maintains telomere-length homeostasis . Although the Ku-Sir4 recruitment mechanism we propose is inhibited by the important negative regulators of telomerase Rif1 and Rif2 , it is clear that the Ku-Sir4 pathway normally has a more modest role in telomere-length maintenance than the Est1-Cdc13 pathway . Disrupting the Ku-Sir4 pathway results in short but stable telomeres , whereas , in contrast , loss of Est1-Cdc13 recruitment causes complete loss of telomeres and cellular senescence . The typically smaller magnitude of the effects of the Ku-Sir4 pathway makes its disruption more likely to have been missed in prior studies . Furthermore , it has been difficult to separate the roles of Sir proteins in telomerase recruitment from competing negative roles of Rif1 and Rif2 . The C-terminal domain of the Rap1 protein that binds telomeric dsDNA repeats is bound by Rif1 and Rif2 and recruits Sir3 and Sir4 to telomeres . Deleting the C-terminal domain of Rap1 therefore not only disrupts Sir-dependent silencing at telomeres but also abolishes Rif-protein inhibition of telomerase , and consequently , the net result is that telomeres become extremely long in a mutant lacking the Rap1 C-terminus ( Kyrion et al . , 1992 ) . However , our model cannot fully explain the long-telomere phenotype of this mutant . In the absence of Rif1 and Rif2 binding at telomeres in rap1∆c cells , telomeres will only become hyper-elongated if a process that the Rif proteins inhibit persists . Because mutants lacking the Rap1 C-terminus have long telomeres despite having lost Sir4 binding at telomeres ( at least via Rap1 ) , the Ku-Sir4-mediated recruitment pathway we propose is not the only one inhibited by Rif1 and Rif2 . There are , however , a few noteworthy results in the literature suggesting that telomeric silent chromatin factors , particularly Rap1 , contribute to telomere-length maintenance . First , deleting SIR3 or SIR4 causes telomere-length maintenance defects ( Palladino et al . , 1993; Askree et al . , 2004; Gatbonton et al . , 2006 ) . Second , the Rap1 M763A mutation , which abolishes the Rap1-Sir3 interaction and slightly impairs telomeric silencing , has been shown to cause shortened telomeres ( Feeser and Wolberger , 2008 ) . Lastly , Rap1 binding near a telomeric seed sequence has been shown to promote telomerase-dependent de novo telomere formation , although this activity was reportedly SIR4-independent ( Ray and Runge , 1998 ) . In summary , we have shown that telomerase RNA-bound Ku recruits telomerase to telomeres by binding the telomeric silent chromatin protein Sir4 . The Ku-Sir4 pathway is inhibited by the telomerase regulators Rif1 and Rif2 and likely promotes telomerase recruitment through the essential Est1-Cdc13 recruitment pathway . Thus , this pathway represents an important mechanism by which telomerase is regulated to maintain telomere length in S . cerevisiae and it may be generally conserved in many other species . For instance , both telomeric silent chromatin and the Ku-telomerase RNA interaction are present in humans ( Baur et al . , 2001; Ting et al . , 2005 ) . Although humans lack an obvious Sir4 homolog , the protein HP1α has been implicated in human telomeric silencing ( Koering et al . , 2002; Arnoult et al . , 2012 ) and has been shown to bind Ku70 ( Song et al . , 2001 ) , so it will be interesting to learn if these interactions also comprise a telomerase-recruitment pathway .
Lists of the yeast strains and plasmids used can be found in Supplementary file 1 . Experiments described in Figures 2A , 3 , 5 , 7 are all based on tlc1∆ complementation assays reported previously ( Lebo et al . , 2015 ) . Plasmid pRS414-based constructs containing TLC1 alleles were transformed into a tlc1Δ strain harboring a pTLC1-LYS2-CEN or pTLC1-URA3-CEN ‘cover’ plasmid . The TLC1-containing cover plasmid was then shuffled out by plating transformants on medium containing α-aminoadipate to select for LYS− cells that lost the pTLC1-LYS2-CEN cover plasmid or medium with 5-fluoroorotic acid to select for URA− cells that lost the pTLC1-URA3-CEN cover plasmid . In Figure 5 , cells were streaked once to solid minimal medium lacking tryptophan before being grown for ChIP . In all other cases , cells were passaged by one of two methods after cover plasmid loss . In Figures 2A , 7 , 3C , cells were passaged by serially re-streaking single colonies on solid minimal medium lacking tryptophan . When using this passaging technique , generation time was estimated as 25 generations per re-streak ( including the streak used to shuffle out the cover plasmid ) . For Figure 3A , B , and Figure 3—figure supplements 1 , 2 , cells were passaged differently . First , after the streak used to shuffle-out the cover plasmid , cells were re-streaked once to solid minimal medium lacking tryptophan . Colonies from this minus-tryptophan ( −TRP ) plate were then used to inoculate 20-ml −TRP liquid cultures and culturing was performed at 30°C for ∼24 hr . Cells were then back-diluted by a factor of 210 into 20-ml cultures of fresh medium , which were grown for another ∼24 hr before being passaged again . Cultures reached the same approximate density each day as measured spectrophotometrically ( 600-nm light ) . In these experiments , generation time was approximated as 50 generations ( 25 generations for the colony forming after streaking to solid medium when shuffling out the cover plasmid plus 25 more for the −TRP medium growth ) plus 10 generations for each day of passaging in liquid cultures . In Figure 2B , a SIR4/sir4Δ YKU80/yku80Δ diploid was sporulated , and tetrads were dissected to isolate tetratype spores . The spores from this ascus were then re-streaked three successive times on rich YPD medium before telomere length was assessed . Southern blotting was performed as described previously ( Zappulla et al . , 2005 , 2011 ) . Briefly , cells were pelleted either directly from liquid cultures used for passaging or from cultures grown from serial re-streaking plates . Genomic DNA was isolated from these cells ( Gentra Puregene system from Qiagen , Hilden , Germany ) , and roughly equal amounts of genomic DNA were digested with XhoI . Digested genomic DNA samples were resolved on a 1 . 1% agarose gel . The DNA was then transferred to Hybond-N+ Nylon membrane ( GE , Little Chalfont , United Kingdom ) , which was probed for telomeric sequence and for a 1627-bp , non-telomeric XhoI restriction fragment from within chromosome IV and then imaged using phosphor screens and a Typhoon 9410 Variable Mode Imager ( GE ) ( Friedman and Cech , 1999 ) . Average Y′ telomere length was calculated using the weighted average mobility method as previously described ( Zappulla et al . , 2011 ) . In Figure 3—figure supplement 3 , Southern blots were probed for Yʹ sequence . Yʹ probe was made by first performing PCR with the following primers using genomic DNA as template DNA: 5ʹ-TGTTGTCTCTTACCCGGATGTTCAACC-3ʹ , 5ʹ-AAAGTTGGAGTTTTTCAGCGTTTGCG-3ʹ . The DNA amplified in this reaction was in turn used as template for making the radiolabeled Yʹ probe . Northern blotting was performed as previously described ( Zappulla et al . , 2005 ) . Briefly , cells were harvested in the same manner as those used for Southern blots , and total RNA was isolated using the hot-phenol method ( Kohrer and Domdey , 1991 ) . 10–15 μg of RNA from each sample was boiled and then resolved by urea-PAGE . The RNA was transferred to Hybond-N+ Nylon membrane ( GE ) , which was then UV-crosslinked and probed for TLC1 and U1 sequences . Due to the low abundance of TLC1 RNA relative to U1 , blots were probed with 100-fold fewer counts of U1 probe than TLC1 probe . Blots were then imaged using phosphor screens and a Typhoon 9410 Variable Mode Imager ( GE ) . ChIP was performed similarly to that described ( Fisher et al . , 2004 ) . Briefly , cells were grown to saturation in 10-ml cultures of liquid minimal medium , back-diluted into 60 ml cultures , and then grown to an OD600 of 0 . 5–0 . 8 . 50 ml of cells were crosslinked with formaldehyde , pelleted , rinsed in lysis buffer , and then re-suspended in lysis buffer . Cells were flash-frozen , thawed , and then lysed using sterile glass beads . Lysates were then sonicated to shear crosslinked chromatin . Anti-myc immunoprecipitation was carried out using mouse anti-myc monoclonal antibodies ( Clontech , Mountain View , CA , United States ) and Protein G Dynabeads ( Life Technologies Oslo , Norway ) . After immunoprecipitation , formaldehyde crosslinks were reversed , and DNA was purified using reagents from the Qiagen PCR Purification Kit . Fold telomeric enrichment in ChIP DNA samples was quantified by qPCR using iQ SYBR Green Supermix and a CFX96 Real–Time Cycler ( Bio-Rad Hercules , CA , United States ) . The primer sets used at telomere VI-R , telomere XV-L , and the ARO1 locus were the same as those described previously ( Sabourin et al . , 2007; McGee et al . , 2010 ) . For a given sample of DNA obtained from ChIP , qPCR reactions for each primer set were performed in technical duplicate or triplicate , and the CT values were averaged together . Using these averages , fold telomeric enrichment was then calculated as 2^[ ( CT ( ARO IP ) − CT ( ARO Input ) ) − ( CT ( TEL IP ) − CT ( TEL Input ) ) ] . Additionally , each time qPCR was performed , the efficiency of amplification was calculated for each primer set being used . From a sample of ChIP input DNA , a series of 10-fold dilutions were made and used as template DNA for qPCR reactions . For these reactions , −log ( dilution factor ) was plotted against CT value , and a line of best fit was found for the graph . Using the slope of this line , percent amplification efficiency was calculated as 100*[10^ ( −1/slope ) − 1] . If amplification efficiency was between 70% and 95% , average CT values were corrected using the slope and Y-intercept values from the line of best fit: Relative amount = 10^[ ( AvgCT − intercept ) /slope] . Then , fold telomeric enrichment was instead calculated as ( RelAmtTEL IP/RelAmtTEL Input ) / ( RelAmtARO IP/RelAmtARO Input ) . Fold telomeric enrichment values in Figure 5 are expressed relative to wild type ( SIR4 TLC1 ) . First , ∼16 pmol of purified , myc-tagged yeast Ku heterodimer ( Pfingsten et al . , 2012; Dalby et al . , 2013 ) was added to the RRL transcription and translation system ( TNT Quick Coupled , Promega Madison , WI , United States ) . In the ‘boiled’ condition in Figure 4 , the Ku heterodimer was heated at 95°C for 5 min before being added to the RRL . Sir4 synthesis was then initiated by adding 1 μg of SIR4 template DNA ( plasmid pDZ930 ) and 35S-L-methionine to the RRL , and the reaction was incubated at 30°C for approximately 90 min . 5 μl of mouse anti-myc monoclonal antibodies ( Clontech , used at a 1:400 dilution in TBST ) was added to the reaction , which was then incubated at 4°C for 1 hr . 40 μl of Protein G Dynabeads ( Life Technologies ) was prepared for each RRL reaction by first pipetting off the storage buffer , rinsing once in 1 ml of ‘standard’ Ku-Sir4 buffer ( 25 mM HEPES pH 7 . 5 , 100 mM NaCl , 1 mM DTT , 10% glycerol , 1 mM EDTA , 0 . 1% IGEPAL ) , and then re-suspended in 40 μl of ‘standard’ Ku-Sir4 buffer per RRL reaction . Before adding beads to the RRL reactions , a 2 μl aliquot was taken from the RRL and set aside to be used as the input sample for the protein gel . 40 μl of prepared beads was added to each RRL reaction , and the reactions were left to rotate at 4°C overnight . The next morning , the beads were pulled down with a magnet , and a 2 μl aliquot of the supernatant was set aside to be used as the unbound sample for the protein gel . The remaining supernatant was discarded , and the beads were washed twice with 500 μl of ‘stringent’ Ku-Sir4 buffer ( same as ‘standard’ Ku-Sir4 buffer but with 200 mM NaCl and 0 . 2% IGEPAL ) . Beads were then re-suspended in 120 μl TE +1% SDS and heated at 95°C for 5 min . The beads were pulled down with a magnet , and the entire supernatant was saved as the bound fraction . 2 μl of 2× protein sample buffer was added to the input and unbound aliquots , which were then heated at 95°C for 5 min . These samples along with a bound sample ( 10 μl of the bound fraction plus 10 μl of 2× protein sample buffer ) were resolved by SDS-PAGE on a 7 . 5% polyacrylamide gel . The resulting gel was imaged using phosphor screens and a Typhoon 9410 Variable Mode Imager . It should be noted that the ‘standard’ and ‘stringent’ Ku-Sir4 buffers were designed based off of the buffers used in co-immunoprecipitation experiments described previously ( Roy et al . , 2004 ) . | Inside a cell's nucleus , DNA is packaged into structures called chromosomes . The ends of every chromosome are capped by repeating sequences of DNA known as telomeres , which protect the chromosomes from damage . Every time a cell divides , the telomeres shorten . If telomere length falls below a critical level , the cell can die or enter a state in which it can no longer divide . During cell division , an enzyme called telomerase normally restores telomeres to their original length . Telomerase is made up of several proteins and an RNA molecule . In yeast and humans , a protein called Ku is one part of the telomerase enzyme . Ku binds to the RNA subunit of telomerase and helps the enzyme find and interact with the telomeres . Previous research has shown that Ku is unable to work alone to recruit telomerase to the chromosome . A protein called Sir4 binds to telomeres and cells lacking it have short telomeres , but the reason behind this was not known . Hass and Zappulla confirmed previous reports that Ku binds to Sir4 using a biochemical approach . Additional experiments provided genetic evidence that this binding interaction is important for telomerase to lengthen telomeres appropriately . Cells in which the RNA subunit of telomerase is unable to bind effectively to Ku have short telomeres . Hass and Zappulla directly tethered Sir4 to this defective RNA and found this restored the shortened telomeres to a normal length , indicating that Sir4 normally binds Ku to recruit telomerase . Discovering this mode of recruitment also helps to explain how two other telomeric proteins ( Rif1 and 2 ) limit telomere lengthening; they compete with Ku-Sir4 recruitment to form a length-regulating system . Taken together , Hass and Zappulla's results provide strong evidence that Sir4 cooperates with Ku to control the lengthening of chromosome ends . Future research will hopefully reveal the precise space and time requirements for this telomerase-controlling system in yeast . Additionally , because Ku has been reported to be a subunit of human telomerase , future studies could also explore whether human cells use a similar strategy . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2015 | The Ku subunit of telomerase binds Sir4 to recruit telomerase to lengthen telomeres in S. cerevisiae |
Many intercellular signals are synthesised as transmembrane precursors that are released by proteolytic cleavage ( ‘shedding’ ) from the cell surface . ADAM17 , a membrane-tethered metalloprotease , is the primary shedding enzyme responsible for the release of the inflammatory cytokine TNFα and several EGF receptor ligands . ADAM17 exists in complex with the rhomboid-like iRhom proteins , which act as cofactors that regulate ADAM17 substrate shedding . Here we report that the poorly characterised FERM domain-containing protein FRMD8 is a new component of the iRhom2/ADAM17 sheddase complex . FRMD8 binds to the cytoplasmic N-terminus of iRhoms and is necessary to stabilise iRhoms and ADAM17 at the cell surface . In the absence of FRMD8 , iRhom2 and ADAM17 are degraded via the endolysosomal pathway , resulting in the reduction of ADAM17-mediated shedding . We have confirmed the pathophysiological significance of FRMD8 in iPSC-derived human macrophages and mouse tissues , thus demonstrating its role in the regulated release of multiple cytokine and growth factor signals .
The cell surface protease ADAM17 ( also called TACE ) mediates the release of many important signalling molecules by ‘shedding’ their extracellular ligand domains from transmembrane precursors . A prominent example is the role of ADAM17 in releasing tumour necrosis factor alpha ( TNFα ) ( Black et al . , 1997; Moss et al . , 1997 ) , a primary cytokine involved in the inflammatory responses to infection and tissue damage ( Kalliolias and Ivashkiv , 2016 ) . In addition , ADAM17 is the principal sheddase of the epidermal growth factor ( EGF ) receptor ligands amphiregulin ( AREG ) , transforming growth factor alpha ( TGFα ) , heparin-binding EGF ( HB-EGF ) , epigen , and epiregulin ( Sahin et al . , 2004; Sahin and Blobel , 2007 ) . The control of ADAM17 activity has therefore been the focus of much fundamental and pharmaceutical research ( reviewed in [Rose-John , 2013; Zunke and Rose-John , 2017] ) . We and others have previously reported that the rhomboid-like iRhom proteins have a specific and extensive regulatory relationship with ADAM17 , to the extent that iRhoms can effectively be considered as regulatory subunits of the protease ( Grieve et al . , 2017 ) . iRhoms are members of a wider family of evolutionarily related multi-pass membrane proteins , called the rhomboid-like superfamily ( Freeman , 2014 ) . The family is named after the rhomboids , intramembrane serine proteases that cleave substrate transmembrane domains ( TMDs ) , but many members , including iRhoms , have lost protease activity during evolution . iRhom1 and its paralogue iRhom2 ( encoded by the genes RHBDF1 and RHBDF2 ) show redundancy in regulating ADAM17 maturation , but differ in their tissue expression ( Christova et al . , 2013 ) . Many cell types express both iRhoms , so the loss of one can be compensated by the other ( Christova et al . , 2013; Li et al . , 2015 ) . Macrophages are , however , an exception: iRhom1 is not expressed , so iRhom2 alone regulates ADAM17 and therefore TNFα inflammatory signalling in macrophages ( Adrain et al . , 2012; McIlwain et al . , 2012; Issuree et al . , 2013 ) . iRhoms control ADAM17 activity in multiple ways . First , they bind to the catalytically immature pro-form of ADAM17 ( proADAM17 ) in the endoplasmic reticulum ( ER ) , and are required for its trafficking from the ER to the Golgi apparatus ( Adrain et al . , 2012; McIlwain et al . , 2012 ) . Once proADAM17 reaches the Golgi , it is matured by the removal of its inhibitory pro-domain by pro-protein convertases ( Schlöndorff et al . , 2000; Endres et al . , 2003 ) and is further trafficked to the plasma membrane . iRhoms have further regulatory functions beyond this step of ADAM17 maturation . Still bound to each other , iRhom2 prevents the lysosomal degradation of ADAM17 ( Grieve et al . , 2017 ) . Later , iRhom2 controls the activation of ADAM17: the phosphorylation of the iRhom2 cytoplasmic tail promotes the recruitment of 14-3-3 proteins , which promote the shedding activity of ADAM17 , thereby releasing TNFα from the cell surface in response to inflammatory triggers ( Grieve et al . , 2017; Cavadas et al . , 2017 ) . Finally , iRhoms are also reported to contribute to ADAM17 substrate specificity ( Maretzky et al . , 2013 ) . This intimate regulatory role of iRhoms make them essential players in ADAM17-mediated signalling and thus new targets for manipulating inflammatory signalling . The significance of this potential is underlined by the fact that anti-TNFα therapies , used to treat rheumatoid arthritis and other inflammatory diseases , are currently the biggest grossing drugs in the world ( Monaco et al . , 2015 ) . Despite the role of the iRhom/ADAM17 shedding complex in controlling signalling , much is yet to be understood about the molecular mechanisms that control this inflammatory trigger . To identify the wider machinery by which iRhoms regulate ADAM17 , we report here a proteomic screen to identify their binding partners . We have identified the poorly characterised FERM domain-containing protein 8 ( FRMD8 ) as having a strong and specific interaction with the cytoplasmic N-terminus of iRhoms . The functional significance of this interaction is demonstrated by loss of FRMD8 causing a similar phenotype to iRhom deficiency in cells: loss of mature ADAM17 and severely reduced shedding of ADAM17 substrates from the cell surface . We show that loss of FRMD8 leads to lysosomal degradation of mature ADAM17 and iRhom2 , indicating that its function is to stabilise the iRhom/ADAM17 sheddase complex once it reaches the plasma membrane . Overall , our results imply that FRMD8 is an essential component of the inflammatory signalling machinery . To test this proposal in vivo we deleted the FRMD8 gene in human induced pluripotent stem cells ( iPSCs ) and differentiated them into macrophages . Consistent with our biochemical data , these mutant macrophages were defective in their ability to release TNFα in response to lipopolysaccharide ( LPS ) stimulation , demonstrating the pathophysiological importance of FRMD8 in the normal inflammatory response by human macrophages . The in vivo significance of FRMD8 in regulating the stability of the iRhom/ADAM17 shedding complex was further reinforced by our observation that mature ADAM17 and iRhom2 protein levels are strongly reduced in tissues of FRMD8-deficent mice .
To investigate the molecular mechanisms that underlie iRhom2 functions , we performed a mass spectrometry-based screen to identify new proteins that interact with human iRhom2 . iRhom2-3xHA was stably expressed in human embryonic kidney ( HEK ) 293T cells and immunoprecipitated . The bead eluates containing immunoprecipitated iRhom2 and its interacting proteins were analysed by label-free mass spectrometry . As a negative control , we did the same analysis in parallel with 3xHA-tagged UNC93B1 , an unrelated polytopic protein that , like iRhom2 , is predominantly located in the ER ( Koehn et al . , 2007 ) ( Figure 1—figure supplement 1A ) . Quantitative protein abundance data from three biological replicates of iRhom2 and UNC93B1 co-immunoprecipitations were statistically analysed using the Perseus software platform ( Tyanova et al . , 2016 ) . Validating the overall approach , we detected ADAM17 , the known iRhom2 interacting protein ( Adrain et al . , 2012; McIlwain et al . , 2012; Christova et al . , 2013 ) as a statistically significant hit ( Figure 1A , Table 1 ) . Among the hits were several 14-3-3 proteins ( eta , epsilon , gamma , sigma , theta , zeta/delta ) and MAPK1/3 ( Table 1 ) , which we have previously reported to participate in the regulation of inflammatory signalling by phosphorylation of iRhom2 ( Grieve et al . , 2017 ) . The top hit by a long way , however , was FRMD8 ( Figure 1A , Table 1 ) , a poorly studied protein that has not previously been implicated in iRhom function , ADAM17 regulation , and growth factor or cytokine signalling . We confirmed the interaction between iRhom2 and FRMD8 by immunoprecipitation . C-terminally V5-tagged FRMD8 co-immunoprecipitated with either iRhom1-3xHA or iRhom2-3xHA ( Figure 1B ) . Conversely , we pulled down both iRhom1-3xHA and iRhom2-3xHA with an antibody against the V5 tag . Finally , we were also able to co-immunoprecipitate endogenous FRMD8 with iRhom2-3xHA ( Figure 1—figure supplement 1B ) . Together these results identify FRMD8 as a bona fide binding partner of iRhom1 and iRhom2 in human cells . As its name indicates , FRMD8 is a FERM ( 4 . 1/ezrin/radixin/moesin ) domain-containing protein . It is predicted to be a soluble cytoplasmic protein , and the only report about its function describes it as binding to the Wnt accessory receptor low-density lipoprotein receptor-related protein 6 ( LRP6 ) , and negatively regulating Wnt signalling ( Kategaya et al . , 2009 ) . To investigate the functional significance of FRMD8 binding to iRhoms , we examined the effects of loss of FRMD8 on iRhom function in HEK293T cells , using both siRNA and CRISPR/Cas9-mediated gene deletion ( Figure 2A , B ) . In both cases , loss of FRMD8 drastically reduced the protein levels of mature ADAM17 ( Figure 2A , B ) . This effect was specific to ADAM17 , as the maturation of its closest homologue , ADAM10 , was unaffected by loss of FRMD8 ( Figure 2B ) . Moreover , mature ADAM17 levels were rescued by expression of FRMD8-V5 in FRMD8 knockout HEK293T cells ( Figure 2C ) , confirming that the phenotype was caused by FRMD8 loss . Finally , in addition to this reduction of mature ADAM17 caused by FRMD8 loss , we found a striking loss of ADAM17 , but not ADAM10 , on the cell surface ( Figure 2D ) . These phenotypes partially phenocopy the loss of iRhoms ( Christova et al . , 2013; Grieve et al . , 2017 ) , consistent with FRMD8 being needed for iRhoms to act as positive regulators of ADAM17 . We also examined the consequences of loss of FRMD8 on ADAM17-dependent signalling . The shedding of alkaline phosphatase ( AP ) -tagged EGF receptor ligands AREG and HB-EGF , after stimulation with phorbol 12-myristate 13-acetate ( PMA ) , were both substantially reduced in FRMD8 knockout cells ( Figure 2E ) . To exclude the possibility that the defect in FRMD8 knockout cells is an inability to respond to PMA , we measured both PMA-stimulated and unstimulated , constitutive shedding of AP-tagged TGFα , another major EGFR ligand . Again , FRMD8 knockout cells released significantly less AP-TGFα compared to wild-type cells , both after stimulation but also after 20 hr of constitutive shedding ( Figure 2F ) , implying that mutant cells had fundamental defects in their ability to shed ADAM17 ligands , regardless of PMA stimulation . To demonstrate that the release of ligands was indeed caused by metalloprotease shedding and not simply an indication of leakage caused by cell death , we showed that it was sensitive to the ADAM10/17 inhibitor GW280264X ( GW ) ( Figure 2E , F ) . Overall , as with ADAM17 maturation , the shedding defects in FRMD8-deficient cells resemble those caused by the loss of iRhoms . As described above , iRhoms regulate ADAM17 function at multiple stages: from ER-to-Golgi trafficking , to the activation of the sheddase at the cell surface . To address where FRMD8 fits in this long-term relationship between iRhoms and ADAM17 , we started by analysing the FRMD8 binding site within iRhom2 . As a cytoplasmic protein , FRMD8 was likely to bind to the only substantial cytoplasmic region of iRhom2 , its N-terminus . We therefore made a set of iRhom2-3xHA N-terminal deletion constructs ( Figure 3A ) to locate the binding site . Deletion of the first 200 amino acids in the N-terminus of iRhom2 ( iRhom2∆200 ) did not disrupt FRMD8 binding , but no interaction was detected in mutants greater than ∆300 ( Figure 3B ) , implying that the region between 200 and 300 amino acids was necessary for FRMD8 binding . An internal deletion of amino acids 201–300 within iRhom2 ( iRhom2∆201-300 ) led to the loss of exogenous FRMD8 binding ( Figure 3C ) , confirming that the FRMD8 binding site lies within this region . In line with this , an iRhom1/2 DKO cell line reconstituted with iRhom2∆201-300 showed a similar deficiency to FRMD8 KO cells in ADAM17-mediated shedding of AREG ( compare Figures 3D and 2E ) . This reduction in shedding correlates with a reduction in the level of mature ADAM17 ( Figure 3E ) . Overall , this makes the iRhom2∆201-300 mutant a useful tool to study the loss of FRMD8 binding to iRhom2 and highlights that FRMD8 binding affects levels of mature ADAM17 , presumably either through controlling ADAM17 maturation or stability . Interestingly , the FRMD8 binding site is also absent in a mouse iRhom2 mutant called curly-bare ( cub ) , which lacks residues 1–268 ( Hosur et al . , 2014; Siggs et al . , 2014 ) . Sequence alignment shows that the deletion of 268 amino acids in mouse iRhom2 corresponds to the loss of residues 1–298 in the human protein ( Figure 3—figure supplement 1A ) . Consistent with this mapping data , we found that whereas full-length mouse iRhom2 bound human FRMD8 , the cub mutant cannot ( Figure 3—figure supplement 1B ) . This failure of FRMD8 binding presumably contributes to the complex defects that underlie the cub phenotype ( Johnson et al . , 2003; Hosur et al . , 2014; Siggs et al . , 2014 ) . Combined , these data show that FRMD8 is recruited to a discrete 201–300 amino acid region of the iRhom2 N-terminus , and that this binding is required for sufficient levels of mature ADAM17 , as well as ADAM17-dependent shedding . We next investigated the nature of a putative tripartite complex between iRhom2 , FRMD8 and ADAM17 . Previous work has shown a strong interaction between iRhom2 and ADAM17 ( Adrain et al . , 2012; McIlwain et al . , 2012; Grieve et al . , 2017 ) , yet how FRMD8 intersects with this complex is not known . When performing an immunoprecipitation of FRMD8 , we found both immature and mature ADAM17 as well as iRhom2 ( Figure 4A ) . This indicates that FRMD8 does indeed form a tripartite complex with ADAM17 and iRhom2 . To test this further , we performed a series of pairwise co-immunoprecipitations in the absence of FRMD8 , ADAM17 or iRhoms . First we tested the requirement for ADAM17 in the iRhom2/FRMD8 interaction . We found that exogenous FRMD8 and iRhom2 co-immunoprecipitated with each other in ADAM17 knockout cells ( Figure 4B ) . In turn , iRhom2 and ADAM17 were still able to interact with each other in FRMD8 knockout cells ( Figure 4C ) , showing that FRMD8 is not essential for the iRhom/ADAM17 sheddase complex to form . In contrast , FRMD8 did not pull down ADAM17 in cells mutant for both iRhoms ( Figure 4D ) . This demonstrated that there is no direct link between FRMD8 and ADAM17; instead both bind simultaneously to iRhom2 . Supporting this , FRMD8 co-immunoprecipitated with pro- or mature ADAM17 in iRhom1/2 DKO cells reconstituted with iRhom2WT , but not with iRhom2∆201-300 ( Figure 4D ) , the mutant that does not bind to FRMD8 ( Figure 3C ) . FRMD8 binds to both iRhom2/proADAM17 and iRhom2/mature ADAM17 complexes but associates preferentially with iRhom2/mature ADAM17 complexes ( Figure 4A , D ) , which have been shown to exist at the cell surface ( Grieve et al . , 2017; Cavadas et al . , 2017 ) . This is consistent with the observation of specific effects of FRMD8 loss on mature ADAM17 at the cell surface , but not immature ADAM17 ( Figure 2A–C ) . To further investigate a potential role for FRMD8 at the cell surface , we first assessed its effects on iRhom2 localisation by immunofluorescence . Overexpression of FRMD8-V5 in iRhom1/2 DKO cells reconstituted with wild-type iRhom2 led to a striking increase in plasma membrane iRhom2 ( Figure 5A ) , which in wild-type cells is almost exclusively observed within the endoplasmic reticulum ( Figure 1—figure supplement 1A ) . As a control , the iRhom2∆300 mutant , which cannot bind to FRMD8 ( Figure 3B ) , did not undergo the same ER-to-plasma membrane relocalisation upon FRMD8 overexpression . Indicating a reciprocal relationship between the two proteins , we also observed that the iRhom2 N-terminus was required for FRMD8 localisation at the cell surface ( Figure 5A , B ) . To test whether FRMD8 is sufficient to target iRhom2 to the cell surface , we fused FRMD8 to the N-terminus of the ER-localised iRhom2∆300 mutant ( FRMD8-iRhom2∆300 ) . Strikingly , we saw that the normal ER localisation of iRhom2∆300 ( Figure 5D ) shifted to the plasma membrane upon fusion to FRMD8 ( Figure 5E ) . Furthermore , we found that the localisation of ADAM17-V5 followed that of iRhom in both conditions: in iRhom2∆300 cells ADAM17 localised to the ER , and in FRMD8-iRhom2∆300 cells it was readily observed at the cell surface . We also noted that FRMD8-iRhom2∆300 showed strikingly higher total levels of iRhom2 ( Figure 5E ) , which hinted that FRMD8 may play a role in the protein turnover of iRhoms at the cell surface . In line with these observations , we found that FRMD8-iRhom2∆300 was much more stable compared to iRhom2∆300 or iRhom2WT as seen in cells incubated with cycloheximide ( CHX ) to block the synthesis of new proteins ( Figure 5F ) . Taken together , these data suggest that FRMD8 binding to iRhom2 stabilises the iRhom2 pool in the late secretory pathway and increases the cell surface localisation of the iRhom2/ADAM17 sheddase complex . Previous studies have shown that the cytoplasmic N-terminal region of iRhom2 is required to prevent lysosomal degradation of ADAM17 ( Grieve et al . , 2017 ) . Therefore , we questioned whether the absence of FRMD8 recruitment to the iRhom2 N-terminus led to delivery of iRhom and ADAM17 to lysosomes . By immunofluorescence microscopy , iRhom2WT localisation is indistinguishable from iRhom2∆300 ( Figure 6A , B ) within the endoplasmic reticulum ( Figure 1—figure supplement 1A ) . However , upon treatment with the lysosomal degradation inhibitor , bafilomycin A1 , both proteins accumulated in LAMP1-positive lysosomal puncta ( Figure 6C , D ) . This suggests that there is a constant turnover of iRhom2 through the endo-lysosomal pathway , with iRhoms presumably cycling via the plasma membrane , before being degraded . Interestingly , unlike the partial colocalisation between LAMP1 and iRhom2WT ( Figure 6C ) , iRhom2∆300 overlapped completely with LAMP1 ( indicated by the arrows in Figure 6D ) . This confirmed that in the absence of FRMD8 recruitment , iRhom2 is constitutively sent to lysosomes . Importantly , this lysosomal pool of iRhom2∆300 also colocalised with ADAM17-V5 after bafilomycin treatment ( highlighted with arrows in Figure 6F ) . All these data together indicate that the iRhom2/ADAM17 complex follow the same fate in the absence of FRMD8 recruitment ( Figure 6E , F ) . Using a complementary approach , we tested the stability of ADAM17 in FRMD8 knockout cells . After 16 hr of treatment with the lysosomal degradation inhibitors bafilomycin and ammonium chloride , the mature form of ADAM17 was partially restored ( Figure 6G; Figure 7—figure supplement 1B ) . Combined , these results explain the reduced level of mature ADAM17 in FRMD8 knockout cells: it implies that the defect caused by loss of FRMD8 is not a failure of ADAM17 maturation , but instead a failure to stabilise the mature form . In line with this interpretation , the proteasomal inhibitor MG132 had no effect on the stability of mature ADAM17 ( Figure 7A ) . We conclude that FRMD8 binding to iRhom2 acts to promote ADAM17 function by ensuring its stability after its maturation in the trans-Golgi network . If FRMD8 acts as a stabilising factor for the plasma membrane-localised iRhom2/ADAM17 sheddase complex , a difference in the cell surface level of iRhom2 is expected in the absence of FRMD8 . Most tagged iRhom2 is ER-localised ( Figure 1—figure supplement 1A , Figure 6A ) and the cell surface fraction is relatively small ( Maney et al . , 2015; Grieve et al . , 2017 ) . Therefore , we used cell surface immunostaining of iRhom2 followed by flow cytometry to measure specifically the pool of iRhom2 at the cell surface . In the absence of FRMD8 we detected a significant loss of cell surface iRhom2 ( Figure 7A ) . In line with our observation that cell surface iRhom2 represents only a small fraction of the total pool , a reduction of total iRhom2 levels was not detectable ( Figure 7B ) . This further supports our observations that FRMD8 binding to iRhoms is required to stabilise the cell surface pool of iRhoms . Consistent with our conclusion that FRMD8 primarily functions late in the iRhom2/ADAM17 relationship , we detected no defects in the ER-based iRhom2/proADAM17 interaction in FRMD8 knockout cells ( Figure 4C ) , nor in the trafficking of iRhom2 from the ER to the Golgi ( Figure 7—figure supplement 1C ) . Our results show that by binding to iRhom2 , FRMD8 stabilises both iRhom2 and mature ADAM17 , protecting them from degradation . A more direct demonstration of this stabilising function is provided by overexpressing FRMD8 , which leads to increased levels of tagged iRhom2 ( Figure 7C ) , as well as iRhom1 ( Figure 7—figure supplement 1D ) . Note that the 50 kDa N-terminally truncated fragment of iRhoms detected in western blots ( Nakagawa et al . , 2005; Adrain et al . , 2012; Maney et al . , 2015 ) is not stabilised by FRMD8 expression ( Figure 7C , Figure 7—figure supplement 1D ) . This iRhom fragment lacks the cytoplasmic tail , and therefore the binding site for FRMD8 , so its insensitivity to FRMD8 is consistent with our model . Intriguingly , the stabilisation of iRhom2 and FRMD8 is mutual: overexpression of iRhom2 consistently led to the stabilisation of endogenous FRMD8 protein ( Figure 7D ) without affecting FRMD8 mRNA levels ( Figure 7E ) . This indicates that the iRhom2-FRMD8 interaction leads to mutual stabilisation of both proteins as well as mutual effects on plasma membrane localisation ( Figure 5A ) . To ensure that our conclusion that FRMD8 stabilises iRhoms was not distorted by our use of overexpressed proteins , and in the absence of a usable antibody against human iRhom2 , we used CRISPR/Cas9 to insert a triple HA tag into the RHBDF2 locus to express endogenously C-terminally tagged iRhom2 . siRNA-mediated knockdown of iRhom2 confirmed that this editing was successful ( Figure 8A ) . The cells showed no defect in ADAM17 maturation ( Figure 8A , Figure 7—figure supplement 1E ) , indicating that the tagged protein was functional . In these cells FRMD8 overexpression led to an increase in endogenous iRhom2 levels ( Figure 8A ) ; conversely , siRNA knockdown of FRMD8 caused a reduction of iRhom2 protein ( Figure 8B ) , but no change of iRhom2 mRNA levels ( Figure 8C ) . Again , the 50 kDa iRhom2 fragment was not affected by FRMD8 levels ( Figure 8A , B ) . Parenthetically , this is the first reported evidence that this iRhom fragment exists endogenously , although its functional significance remains unclear . To summarise our results to this point , we have discovered that by binding to the iRhom2 cytoplasmic N-terminus , FRMD8 stabilises the cell surface iRhom2/ADAM17 sheddase complex . In the absence of FRMD8 recruitment to iRhom2 , this enzyme complex is sent to lysosomes and degraded . We tested the pathophysiological significance of our conclusions by analysing the consequence of loss of FRMD8 in human macrophages , which release TNFα in response to tissue damage and inflammatory stimuli . To generate mutant human macrophages , we used CRISPR/Cas9 to knock out FRMD8 in an iPSC line that had previously been generated from dermal fibroblasts of a healthy female donor ( Fernandes et al . , 2016 ) . The FRMD8 knockout and control iPSCs were analysed for deletions in the FRMD8 gene by PCR ( Figure 9—figure supplement 1A ) , and a normal karyotype was confirmed by single nucleotide polymorphism ( SNP ) analysis ( Figure 9—figure supplement 1B ) before differentiation into macrophages ( Figure 9A ) . These mutant macrophages expressed no detectable FRMD8 and , as in the HEK293T cells , showed severely reduced levels of mature ADAM17 ( Figure 9B ) . When challenged with the inflammatory trigger LPS , TNFα shedding from the cells , as measured by ELISA , was reduced ( Figure 9C ) . Confirming the expected specificity , the ADAM10 inhibitor GI254023X ( GI ) had no effect on TNFα release from these cells , whereas GW , an inhibitor of both ADAM10 and ADAM17 , further reduced TNFα release ( Figure 9—figure supplement 1C ) . Although shedding was inhibited , TNFα expression by LPS was normal in these cells ( Figure 9—figure supplement 1D ) . These results demonstrate that our conclusions about the requirement for FRMD8 in ADAM17 function in cell culture models does indeed apply to human macrophages . To investigate further the physiological significance of our discovery of the role of FRMD8 in stabilising iRhom/ADAM17 sheddase complexes , we analysed the levels of ADAM17 and iRhom2 in tissues from FRMD8-deficient mice . These mice were generated from embryonic stem ( ES ) cells from the KOMP Repository , University of California Davis , in which all coding exons ( 2-11 ) of the Frmd8 gene were deleted ( Figure 9—figure supplement 2A ) . Frmd8-/- mice are viable ( Figure 9—figure supplement 2B ) and fertile . The knockout was confirmed by western blot ( Figure 9D ) . Western blot analysis of tissues of Frmd8-/- mice showed that mature ADAM17 levels were reduced in all tissues examined compared to tissues from wild-type littermates ( Figure 9D ) . This confirms that FRMD8 controls the level of mature ADAM17 in vivo . Of note , there was a major reduction of mature ADAM17 levels in the brain , a tissue in which iRhom2 in almost completely absent but iRhom1 levels are high ( Christova et al . , 2013; Li et al . , 2015 ) . This supports our hypothesis that FRMD8 regulates mature ADAM17 levels through iRhom1 as well as iRhom2 . We also tested in vivo our conclusion that FRMD8 loss destabilises endogenous iRhoms ( Figure 8B ) . Using an antibody that we had previously generated against mouse iRhom2 ( Adrain et al . , 2012 ) , we analysed iRhom2 levels in Frmd8+/+ and Frmd8-/- mouse tissues . In lung and skin , both tissues with high iRhom2 expression ( Christova et al . , 2013 ) , we detected a strong decrease of iRhom2 protein levels in Frmd8-/- compared to wild-type ( Figure 9E , Figure 9—figure supplement 2C ) . Tissue from Rhbdf2-/- mice served as a control for the iRhom2 antibody specificity ( Figure 9E , Figure 9—figure supplement 2C ) . The reduction of endogenous iRhom2 and mature ADAM17 levels in mouse lung was about 75% ( Figure 9E ) , which is comparable to the reduction of mature ADAM17 levels in iPSC-derived human macrophages ( Figure 9B ) . In summary , our experiments in mice confirm the physiological importance of our prior conclusions: FRMD8 is required in vivo to regulate the stability of the iRhom/ADAM17 sheddase complex and is therefore a previously unrecognised essential component in regulating cytokine and growth factor signalling .
ADAM17 is the shedding enzyme that is responsible for not only the activation of inflammatory TNFα signalling , but also the release from the cell surface of multiple EGF family growth factors and other proteins . Its regulation has therefore received much attention , both from the perspective of fundamental cell biology and because of the proven therapeutic significance of blocking TNFα ( Monaco et al . , 2015 ) . Here we report that FRMD8 is a new component of the regulatory machinery that controls the release of ADAM17 substrates , including TNFα . We identified FRMD8 as a prominent binding partner of iRhoms , which are rhomboid-like proteins that act as regulatory cofactors of ADAM17 . Our subsequent experiments demonstrate that although FRMD8 binds to iRhoms throughout their life cycle , its function appears to be confined to the later stages of their role in regulating ADAM17 . FRMD8 stabilises the iRhom2/ADAM17 complex at the cell surface , ensuring it is available to shed TNFα and growth factors . We took advantage of iPSC technology to generate human FRMD8 knockout macrophages , allowing us to confirm that the mechanistic conclusions derived mostly from HEK293T cell models were indeed relevant to the human cells that provide the primary inflammatory response . Finally , tissues from FRMD8 knockout mice demonstrate the physiological importance of FRMD8 in a whole organism , and confirm that it stabilises the iRhom/mature ADAM17 complex in vivo . Bringing together all our results , we propose the following model of FRMD8 function in ADAM17-dependent signalling: FRMD8 binds to the cytoplasmic domain of iRhoms throughout the secretory pathway , forming a tripartite complex when iRhoms are also bound to ADAM17 . Despite this long-term relationship , we have found no evidence for a functional role for FRMD8 in ER-to-Golgi trafficking or ADAM17 maturation . Instead , FRMD8 acts later , to prevent the endolysosomal degradation of the iRhom/ADAM17 complex ( Figure 10 ) . The exact molecular detail of FRMD8 action on the iRhom2/ADAM17 sheddase complex is unclear . It is possible that FRMD8 increases the delivery of the iRhom2/ADAM17 sheddase complex from the Golgi apparatus to the cell surface , stabilises the complex by preventing its internalisation , or promotes the endosomal retrieval to the cell surface . In all cases , it is likely that the recruitment of additional proteins is required . Therefore , understanding the molecular interactions of FRMD8 , as well as the FRMD8/iRhom2/mature ADAM17 complex at the cell surface , will shed light into the molecular mechanism . As we have previously reported , it is the iRhom2/ADAM17 complex that is responsible for shedding ADAM17 substrates including TNFα . Without FRMD8 , iRhoms and mature ADAM17 are destabilised and the cell cannot shed TNFα in response to an inflammatory challenge . Combined with our previous studies ( Grieve et al . , 2017 ) , this work has changed our perspective on ADAM17 , the central enzyme in cytokine and growth factor shedding . Our evidence implies that it would be more appropriate to consider it as the active subunit of a regulatory complex at the cell surface , where iRhoms provide regulatory functions ( Maney et al . , 2015; Cavadas et al . , 2017; Grieve et al . , 2017 ) , and FRMD8 maintains the stability of the iRhom/ADAM17 complex post-ADAM17 maturation . It is essential that a pool of the sheddase is available on the cell surface to execute , for example , rapid cytokine release in response to inflammatory signals induced by bacterial infection . In the only other paper about FRMD8 function , it was reported that FRMD8 ( named Bili , after the Drosophila mutation ) negatively regulates Wnt signalling by binding to the LRP6 co-receptor , thereby preventing the recruitment of the signal transduction protein axin ( Kategaya et al . , 2009 ) . Although the signalling event being regulated is different , there is the obvious parallel that in both cases FRMD8 binds to the cytoplasmic tail of a transmembrane protein . In the case of Wnt signalling , this prevents the recruitment of axin; in the case of iRhom function , we do not yet know what the next step in the molecular chain of events is , but the cellular consequence is to prevent recruitment of iRhoms into the endolysosomal degradation system . Our results extend an important theme to emerge from a number of studies , namely the significance of the iRhom cytoplasmic N-terminal region in regulating iRhom/ADAM17 function . Several reports indicate that N-terminal mutations in iRhoms cause complex phenotypes that combine aspects of gain and loss of iRhom function , which is consistent with a regulatory function for this region . First , the cub mutation , an N-terminal deletion in mouse iRhom2 , does not abolish protein function but instead modulates it in complex ways that are still poorly understood ( Hosur et al . , 2014; Siggs et al . , 2014 ) . cub was described as a gain-of-function mutation that leads to constitutively elevated release of amphiregulin , but is also reported to be defective in releasing TNFα ( Hosur et al . , 2014 ) . Second , specific point mutations in the N-terminus of human iRhom2 are the cause of a rare genetic disorder called tylosis with oesophageal cancer ( TOC ) ( Blaydon et al . , 2012; Saarinen et al . , 2012 ) . TOC mutations , as well as truncation of parts of the N-terminus have been reported to enhance the activity of ADAM17 ( Maney et al . , 2015 ) , leading to the conclusion that parts of the N-terminus have inhibitory functions on ADAM17 function . Third , phosphorylation of specific sites in the iRhom2 N-terminus result in 14-3-3 binding and consequent activation of substrate shedding by associated ADAM17 ( Grieve et al . , 2017; Cavadas et al . , 2017 ) , demonstrating that the N-terminus of iRhom2 also positively regulates ADAM17 . The FRMD8 binding region does not overlap with these sites required for phosphorylation-dependent 14-3-3 binding , however it is formally possible that there is some functional overlap between them . We could not detect major changes in the interaction of FRMD8 with iRhom2 upon PMA stimulation ( Figure 10—figure supplement 1 ) , which leads to the phosphorylation of iRhoms ( Grieve et al . , 2017; Cavadas et al . , 2017 ) . Moreover , an iRhom2 mutant , in which 15 conserved phosphorylation sites have been mutated to alanine ( iRhom2pDEAD; Figure 3—figure supplement 1A ) ( Grieve et al . , 2017 ) , did not abolish the interaction with FRMD8 ( Figure 10—figure supplement 1 ) . This conclusively demonstrates that the binding of FRMD8 to iRhom2 does not require phosphorylation of iRhom2 . However , it is still formally possible that phosphorylation of iRhom2 affects FRMD8 binding specifically at the cell surface . This change cannot be detected by analysing the entire iRhom2 pool , which is primarily localised in the early secretory pathway . Therefore , we cannot exclude that the phosphorylation state of the relatively small cell surface pool of iRhom2 regulates the interaction with FRMD8 . Consistent with our current results , we reported previously that iRhom2 lacking the entire N-terminus is not sufficient to support ADAM17-mediated shedding in iRhom1/2-deficient cells , although it can promote ER-to-Golgi trafficking of ADAM17 ( Grieve et al . , 2017 ) . Complementary to the conclusion that iRhom N-termini are regulatory , the core TMD binding function of iRhoms depends on their membrane-embedded region ( Grieve et al . , 2017; Cavadas et al . , 2017 ) . A picture therefore begins to emerge of iRhoms having a modular structure , with a core , highly conserved TMD recognition domain in the membrane ( and perhaps the lumen ) , regulated by a more variable N-terminal domain that can integrate cytoplasmic signals . In light of the growing value of therapeutics that block TNFα signalling , and the wider potential of modulating a wide range of ADAM17 substrates , it is tempting to speculate that the cytoplasmic N-termini of iRhoms could provide potential new drug target opportunities . For example , the limited expression of iRhom2 makes it a theoretically attractive anti-inflammatory target ( Issuree et al . , 2013; Lichtenthaler , 2013 ) . iRhom2 knockout mice are broadly healthy , beyond defects in TNFα and type I interferon signalling that are only apparent upon challenge by bacterial and viral infections ( McIlwain et al . , 2012; Luo et al . , 2016 ) . Our work now implies that the interface between FRMD8 and iRhoms might be a useful target . This is supported , at least in principle , by our observation that even in cells with complete loss of FRMD8 , there is still a low level of mature ADAM17 at the cell surface , and consequently residual TNFα shedding . Even very efficient pharmacological blocking of the FRMD8/iRhom interaction would not , therefore , fully abolish inflammatory responses , potentially reducing side effects . Consistent with this idea , mice with a hypomorphic mutation in ADAM17 ( termed ADAM17ex/ex ) show that even only 5% of normal ADAM17 expression is sufficient to rescue many aspects of the loss of function phenotype ( Chalaris et al . , 2010 ) . Moreover , a recent study has shown that reducing ADAM17 levels has great pharmaceutical potential: reduced levels of ADAM17 in the Adam17ex/ex mouse limits colorectal cancer formation and any residual tumours are low-grade dysplasias ( Schmidt et al . , 2018 ) . In conclusion , our work demonstrates the cellular and physiological significance of FRMD8 binding to iRhoms , and how it stabilises the iRhom/ADAM17 sheddase complex at the cell surface . It also reinforces the picture that has begun to emerge of ADAM17 not acting alone but instead being supported by at least two other regulatory proteins that act as subunits of what is effectively an enzyme complex . This concept would help to explain how the activity of such a powerful and versatile – and therefore potentially dangerous – shedding enzyme is controlled with necessary precision . The next steps in fully revealing the role of FRMD8 will be to analyse the phenotypic consequences of its loss in mice , which should allow us to understand how the roles of FRMD8 in ADAM17 activation , Wnt signalling , and any other potential functions , are integrated . Notwithstanding these physiological questions , the work described here already provides a basis for beginning to investigate the potential of targeting the FRMD8/iRhom interface for modulating the release of ADAM17 substrates .
Human UNC93B1 , human iRhom2WT , iRhom2Δ100 , iRhom2Δ200 , iRhom2Δ300 , iRhom2Δ382 iRhom2Δ201-300 , and FRMD8-iRhom2Δ300 were amplified from human UNC93B1 ( BC025669 . 1 ) , human iRhom2 cDNA ( NM_024599 . 2; Origene ( SC122961 ) ) and human FRMD8 cDNA ( NM_031904; Addgene ( SC107202 ) ) by PCR and cloned with an C-terminal 3xHA tag into the lentiviral vector pLEX . puro using Gibson assembly ( New England Biolabs ) following the manufacturer’s instructions . C-terminal V5-tagged FRMD8 ( FRMD8-V5 ) was amplified from human FRMD8 cDNA ( Addgene ( SC107202 ) ) by PCR and cloned into pcDNA3 . 1 ( + ) using Gibson assembly . All constructs were verified by Sanger sequencing ( Source Bioscience , Oxford , UK ) . pM6P . blast plasmids expressing mouse iRhom2WT , iRhom2Δ268 ( iRhom2 cub ) , and iRhom2pDEAD were described previously ( Grieve et al . , 2017 ) . Human embryonic kidney ( HEK ) 293T cells were cultured in DMEM ( Sigma-Aldrich ) supplemented with 10% fetal calf serum ( FCS ) and 1x penicillin-streptomycin ( PS ) ( all Gibco ) at 37°C with 5% CO2 . Cells were transiently transfected with DNA using FuGENE HD ( Promega ) . Per 10 cm2 growth area 4 μl FuGENE HD was added to 1 μg DNA diluted in OptiMEM ( Gibco ) . The transfection mix was incubated for 20 min at room temperature and added to cells . Protein expression was analysed 48–72 hr after transfection . For knockdown experiments , siRNA was transfected using Lipofectamin RNAiMax ( Invitrogen ) following the manufacturer’s instructions . Per 6 well 50 pmol of FRMD8 SMARTpool siRNA ( Dharmacon; siGENOME Human FRMD8 ( 83786 ) siRNA; M-018955-01-0010 ) , non-targeting siRNA control pools ( Dharmacon; siGENOME D-001206-13-50 ) , RHBDF2 siRNA ( Thermo Fisher Scientific; HSS128594 and HSS128595 ) were used . Protein expression was analysed 72 hr after transfection . HEK293T wild-type cell lines stably expressing human UNC93B1-3xHA or human iRhom2-3xHA , and HEK293T iRhom1/2 DKO cell lines expressing iRhom2WT , iRhom2Δ300 , iRhom2Δ201-300 , or FRMD8-iRhom2Δ300 were generated by lentiviral transduction using the pLEX . puro vector as described previously ( Adrain et al . , 2012 ) . Cells were selected by adding 2 . 5 μg/ml puromycin ( Gibco ) . For CRISPR/Cas9-mediated knockout of FRMD8 the plasmid pSpCas9 ( BB ) −2A-Puro ( pX459; Addgene plasmid #48139 ) co-expressing the wild-type Streptococcus pyogenes Cas9 and the guide RNA ( gRNA ) was used . For gRNA design target sequences with a low chance of off targets were selected using online tools ( http://crispr . mit . edu; http://www . sanger . ac . uk/htgt/wge ) . A gRNA targeting exon 7 ( ACCCATAAAACGGCAGCTCG ) , which is present in all FRMD8 isoforms , was cloned into pX459 . 1 µg plasmid was transfected into a 6-well of HEK293T cells . Cells were selected with puromycin 48 hr after transfection to eliminate non-transfected cells . Single colonies were selected to establish clonal cell lines . Loss of FRMD8 expression was analysed by western blot and quantitative PCR . HEK293T iRhom1/2 double-knockout cells were generated using the plasmid pSpCas9 ( BB ) −2A-Puro V2 . 0 ( pX462 V2 . 0 ) co-expressing the S . pyogenes Cas9 nickase mutant D10A and a guide gRNA . gRNAs targeting exon 3 ( GGAACCATGAGTGAGGCCCC , GGGTGGCTTCTTGCGCTGCC ) and exon 10 ( AGCCGTGTGCATCTATGGCC , CCGTCTCATGCTGCGAGAAC ) of RHBDF1 , and exon 2 ( GCAGAGCCGGAAGCCACCCC , GGGTCTCTTTCTCGGGTGGC ) and exon 9 ( AAACTCGTCCATGTCATCATCACC , ACGGGTGCGATGCCATACGC ) of RHBDF2 were individually cloned into pX462 V2 . 0 . 250 ng of each plasmid were transfected together into a 6-well of HEK293T cells ( eight plasmids in total per well ) . Cells were selected with puromycin 48 hr after transfection and single colonies were selected to establish clonal cell lines . Loss of iRhom1 and iRhom2 was analysed by PCR . To generate a knock-in of a triple HA tag at the C-terminus of endogenous iRhom2 , a homology construct consisting of the triple HA tag ( 3xHA ) flanked at both sides with homology arms of approximately 800 bp was cloned into pcDNA3 . 1 ( + ) . The RHBDF2 locus was targeted in exon 19 in close proximity to the stop codon using a gRNA ( AGCGGTCAGTGCAGCACCT or CAGCGGTCAGTGCAGCACC ) cloned into vector epX459 ( 1 . 1 ) ( generated by subcloning enhanced Cas9 ( eSpCas9 ) v1 . 1 into plasmid pX459; a kind gift from Dr Joey Riepsaame , University of Oxford ) . HEK293T cells were treated with 200 ng/ml nocodazole ( Sigma-Aldrich ) for 17 hr and then transfected with epX459 ( 1 . 1 ) and the pcDNA3 . 1 ( + ) homology plasmid ( 0 . 5 µg each per 6-well ) . After puromycin selection and single cell cloning , cell clones were tested for the insertion of the 3xHA tag by PCR . HEK293 ADAM17 knockout cells were kindly provided by Dr Stefan Düsterhöft and have been published previously ( Riethmueller et al . , 2016 ) . HEK293T cells expressing human UNC93B1-3xHA ( control ) and human iRhom2-3xHA were used for anti-HA co-immunoprecipitation and analysed by mass spectrometry as described previously ( Grieve et al . , 2017 ) . Peptides were injected into a nano-flow reversed-phase liquid chromatography coupled to Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer ( Thermo Scientific ) . The raw data files generated were processed using the MaxQuant ( version 1 . 5 . 0 . 35 ) software , integrated with the Andromeda search engine as described previously ( Cox and Mann , 2008; Cox et al . , 2011 ) . Differential protein abundance analysis was performed with Perseus ( version 1 . 5 . 5 . 3 ) . A two-sample t-test was used to assess the statistical significance of protein abundance fold-changes . P-values were adjusted for multiple hypothesis testing with the Benjamini-Hochberg correction ( Hochberg and Benjamini , 1990 ) . Cells were washed with ice-cold PBS and then lysed on ice in Trition X-100 lysis buffer ( 1% Triton X-100 , 150 mM NaCl , 50 mM Tris-HCl pH 7 . 5 ) supplemented with EDTA-free protease inhibitor mix ( Roche ) and 10 mM 1 , 10-Phenanthroline ( Sigma-Aldrich ) . Cell debris were pelleted by centrifugation at 20 , 000 g at 4°C for 10 min . Proteins were immunoprecipitated by incubation with anti-HA magnetic beads ( Thermo Scientific ) or anti-V5 magnetic beads ( MBL International ) for 1 hr at 4°C . Beads were washed with Trition X-100 wash buffer ( 1% Triton X-100 , 300 mM NaCl , 50 mM Tris-HCl pH 7 . 5 ) . Proteins were eluted in 2x LDS buffer ( life technologies ) supplemented with 50 mM DTT for 10 min at 65°C . N-glycosylated proteins were enriched by incubating cells lysates with concanavalin A sepharose ( Sigma-Aldrich ) at 4°C for at least 3 hr with over-head rotation . Beads were pelleted ( 2 , 500 g , 5 min , 4°C ) and washed with Triton X-100 wash buffer . Proteins were eluted in 2x LDS buffer supplemented with 50 mM DTT and 50% sucrose for 10 min at 65°C . To access protein stability , HEK293T cells were treated with 100 µg/ml cycloheximide ( CHX; Sigma-Aldrich ) for 0–8 hr to block protein synthesis . After incubation , cells were washed with ice-cold PBS and then lysed on ice in Trition X-100 lysis buffer supplemented with EDTA-free protease inhibitor mix and 10 mM 1 , 10-Phenanthroline . Lysates were centrifuged at 20 , 000 g at 4°C for 10 min and analysed by SDS-PAGE . Cell lysates were mixed with 4x LDS buffer ( life technologies ) supplemented with 50 mM DTT and denatured for 10 min at 65°C prior to loading on 4–12% Bis-Tris gradient gels run in MOPS running buffer ( both Invitrogen ) . Proteins were transferred to a polyvinylidene difluoride ( PVDF ) membrane ( Millipore ) in transfer buffer ( Invitrogen ) . The membrane was blocked in 5% milk-TBST ( 150 mM NaCl , 10 mM Tris-HCl pH 7 . 5 , 0 . 05% Tween 20 , 5% dry milk powder ) and then incubated with the primary antibody: mouse monoclonal anti-β-actin-HRP ( Sigma-Aldrich , A3854 , 1:5000 ) , rabbit polyclonal anti-ADAM17 ( abcam; ab39162; 1:2000 ) , rabbit polyclonal anti-FRMD8 ( abcam; ab169933; 1:500 ) , rat monoclonal anti-HA-HRP ( Roche , 11867423001 , 1:2000 ) , goat polyclonal anti-V5 ( Santa Cruz , sc-83849 , 1:2000 ) , mouse monoclonal anti-transferrin receptor 1 , ( Thermo Fisher Scientific , 13–6800 , 1:2000 ) , and rabbit polyclonal anti-iRhom2 ( [Adrain et al . , 2012]; 1:500 ) . After three washing steps with TBST ( 150 mM NaCl , 10 mM Tris-HCl pH 7 . 5 , 0 . 05% Tween 20 ) , membranes were incubated with the secondary antibody for 1 hr at room temperature using either goat polyclonal anti rabbit-HRP ( Sigma-Aldrich , A9169 , 1:20000 ) , mouse monoclonal anti-goat-HRP ( Santa Cruz , sc-2354 , 1:5000 ) or goat polyclonal anti-mouse-HRP ( Santa Cruz , sc-2055 , 1:5000 ) . Cells were harvested in PBS and pelleted at 3000 g , 5 min , 4°C . RNA was isolated using the RNeasy kit ( Qiagen ) and reverse transcribed using the SuperScript VILO cDNA synthesis kit ( Invitrogen ) . Resulting cDNA was used for quantitative PCR ( qPCR ) using the TaqMan Gene Expression Master Mix ( Applied Biosystems ) and the following TaqMan probes ( all Thermo Fisher Scientific ) : human ACTB ( Hs99999903_m1 ) , human FRMD8 ( Hs00607699_mH ) , human RHBDF2 ( Hs00226277_m1 ) , and human TNFα ( Hs00174128_m1 ) . qPCR was performed on a StepOnePlus system ( Applied Biosystems ) . Gene expression was normalized to ACTB expression and expressed as relative quantities compared to the corresponding wild-type cell line . Error bars indicate the standard derivation of technical replicates . eight × 104 HEK293T cells were seeded in triplicates per condition into poly- ( L ) -lysine ( PLL; Sigma-Aldrich ) -coated 24-well plates and transfected the next day with 30 ng plasmid DNA encoding Alkaline Phosphatase ( AP ) -conjugated AREG , HB-EGF or TGFα ( received from Prof Carl Blobel ) . 48 hr after transfection , cells were washed with OptiMEM and then incubated with 200 µl phenolred-free OptiMEM ( Gibco ) containing either 200 nM PMA , the corresponding volume of the solvent ( DMSO ) , or 200 nM PMA and 1 µM GW ( synthesized by Iris Biotech ( Marktredwitz , Germany ) and kindly provided by Dr Stefan Düsterhöft ) for 30 min at 37°C . Cell supernatants were collected , the cells were washed in PBS and lysed in 200 µl Trition X-100 lysis buffer . The activity of AP in cell lysates and supernatants was determined by incubating 100 µl AP substrate p-nitrophenyl phosphate ( PNPP ) ( Thermo Scientific ) with 100 µl cell lysate or cell supernatant at room temperature followed by the measurement of the absorption at 405 nm . The percentage of AP-conjugated material released from each well was calculated by dividing the signal from the supernatant by the sum of the signal from lysate and supernatant . The data was expressed as mean of at least three independent experiments , each of which contained three biological replicates per condition . Cells were lysed in Triton X-100 lysis buffer as described above . Lysates were first denatured with Glycoprotein Denaturing Buffer ( New England Biolabs ) at 65°C for 15 min and then treated with endoglycosidase H ( Endo H ) or peptide:N-glycosidase F ( PNGase F ) following the manufacturer’s instructions ( New England Biolabs ) . For ADAM10 and ADAM17 cell surface staining , HEK293T cells were stimulated with 200 nM PMA for 5 min before harvest in PBS . 0 . 5 × 106 HEK293T cells were washed with ice-cold FACS buffer ( 0 . 25% BSA , 0 . 1% sodium azide in PBS ) and stained with rabbit polyclonal anti-HA antibody ( Santa Cruz , sc-805; 0 . 5 µg diluted in FACS buffer ) , mouse monoclonal anti-ADAM10 ( Biolegend , 352702; 4 µg diluted in FACS buffer ) or mouse monoclonal anti-ADAM17 ( A300E antibody ( Yamamoto et al . , 2012 ) , kindly provided by Dr Stefan Düsterhöft; 8 µg diluted in FACS buffer ) on ice for 45 min . After two washes with FACS buffer , the cells were incubated with Alexa Fluor 488-coupled secondary antibody ( Invitrogen , A21202 or A21206 ) ; 1:1000 dilution in FACS buffer ) on ice for 30 min . Cells were washed twice with ice-cold FACS buffer and then analysed with a BD FACSCalibur ( BD Biosciences ) and FlowJo software . Cells stained only with the secondary antibody or anti-HA negative cells served as control . HEK293T iRhom1/2 DKO cells ( 1 . 5 × 105 ) transduced with indicated iRhom2 constructs were plated on 13 mm PLL-coated glass coverslips in 12-well dishes . In FRMD8-V5 or TACE-V5 overexpression experiments , cells were transfected with 200 ng vector and grown for 72 hr prior to fixation . As indicated , cells were treated with 100 nM bafilomycin for 16 hr before fixation , to inhibit lysosomal degradation . Cells were washed three times in PBS at room temperature and fixed with 4% paraformaldehyde in PBS at room temperature for 20 mins . Fixative was quenched with 50 mM NH4Cl for 5 min . Cells were permeabilised in 0 . 2% Triton X-100 in PBS for 30 min and epitopes blocked with 1% fish-skin gelatin ( Sigma-Aldrich ) in PBS for 1 hr . Coverslips were then incubated at room temperature for 2 hr with rabbit anti-HA tag ( Cell Signalling , 3724 ) and goat anti-V5 probe ( Santa Cruz , sc-83849 ) in 1% fish-skin gelatin/PBS . After three PBS washes , coverslips were incubated with Alexa Fluor-coupled secondary antibodies raised in donkey ( Invitrogen ) for 45 min at room temperature . Cells were subsequently washed three times with PBS and once with H2O , prior to mounting on glass slides with mounting medium containing DAPI ( ProLong Gold; ThermoFisher Scientific ) . Images were acquired with a laser scanning confocal microscope ( Fluoview FV1000; Olympus ) with a 60 × 1 . 4 NA oil objective and processed using Fiji ( ImageJ ) . To generate iPSC-derived FRMD8 knockout macrophages , the human iPSC line AH017-13 was used . The AH017-13 line was derived from dermal fibroblasts of healthy donor in the James Martin Stem Cell Facility , University of Oxford as published previously ( Fernandes et al . , 2016 ) . Donors had given signed informed consent for the derivation of human iPSC lines from skin biopsies and SNP analysis ( Ethics Committee: National Health Service , Health Research Authority , NRES Committee South Central , Berkshire , UK ( REC 10/H0505/71 ) ) . AH017-13 iPSCs were cultured feeder cell-free in mTeSR1 ( STEMCELL Technologies ) on hESC-qualified geltrex ( Gibco ) . iPSCs were fed daily and routinely passaged with 0 . 5 mM EDTA , or when required using TrypLE ( Gibco ) and plated in media containing 10 μmol/l Rho-kinase inhibitor Y-27632 ( Abcam ) . AH017-13 iPSCs were transfected by electroporation using the Neon Transfection System ( Invitrogen ) . 3 × 106 AH017-13 iPSCs were electroporated ( 1400 mV , 20 ms , one pulse ) in a 100 μl tip with 15 μg pX459-FRMD8-exon7 plasmid DNA ( endotoxin-free quality ) , then plated at a density of 4 × 105 cells/cm2 and selected 48 hr after transfection with 0 . 25 µg/ml puromycin . After 48 hr of selection , surviving cells were plated on a feeder-layer of 4 × 106 irradiated mouse embryonic fibroblasts ( MEFs ) in 0 . 1% gelatin-coated 10 cm culture dishes and cultured in hES medium ( KnockOut DMEM , 20% KnockOut serum replacement , 2 mM L-Glutamine , 100 µM nonessential amino acids , 50 µM 2-Mercaptoethanol ( all Gibco ) and 10 ng/mL basic fibroblastic growth factor ( bFGF , R and D ) ) . Colonies were manually selected and grown on geltrex in mTeSR1 . Clones were analysed by western blot using the anti-FRMD8 antibody , and PCR followed by Sanger sequencing . For PCR DNA was isolated from iPSCs by incubation in DNA isolation buffer ( 10 mM Tris-HCl ( pH 8 ) , 1 mM EDTA , 25 mM NaCl , 200 µg/ml proteinase K added freshly ) at 65°C for 30 min . Proteinase K was inactivated at 95°C for 2 min . PCR using Q5 polymerase was performed according to the manufacturer’s instructions ( New England Biolabs ) using primers FRMD8_fw ( tgcagATCCATGACGAGGA ) and FRMD8_rev ( gtgctcgtgacaagacac ) . The PCR product was purified and sequenced using the primer FRMD8_exon7_fw ( GCCAGAGTCTCTTTGCTG ) for Sanger sequencing ( Source Bioscience , Oxford ) . AH017-13 wild-type and FRMD8 knockout clones were analysed by Illumina HumanOmniExpress24 single nucleotide polymorphism ( SNP ) array at the Wellcome Trust Centre for Human Genetics at the University of Oxford and assessed using KaryoStudio software to confirm normal karyotypes before differentiation into macrophages . For this study iPSCs were differentiated into embryoid bodies ( EBs ) by mechanical lifting of iPSC colonies and differentiated into macrophages as described in ( van Wilgenburg et al . , 2013 ) . Briefly , iPSCs were grown on a feeder layer of MEFs in hES medium . A dense 10 cm2 well of iPSCs was scored into 10 × 10 sections using a plastic pipette tip . The resulting 100 patches were lifted with a cell scraper and cell clumps were transferred into a 6-well ultra-low adherence plate ( Corning ) containing EB formation medium ( hES medium supplemented with 50 ng/ml BMP4 ( Invitrogen ) , 50 ng/ml VEGF ( Peprotech ) and 20 ng/ml SCF ( Miltenyi ) ) to form EBs . A 50% medium change was performed every second day . On day 5 EBs were harvested . Approximately 60–80 EBs were transferred into a T75 flask containing factory medium ( X-VIVO 15 ( Lonza ) supplemented with 2 mM L-Glutamine , 50 µM 2-Mercaptoethanol , 100 ng/ml M-CSF and 25 ng/mL IL-3 , 100 U/ml penicillin and 100 µg/ml streptomycin ( all Gibco ) ) . The EBs were fed weekly with fresh factory medium . After approximately two weeks EBs started to produce non-adherent macrophage precursors , which were harvested from the supernatant of EB cultures through a 70 µM cell strainer . Cells were differentiated into mature adherent macrophages for 7 days in macrophage medium ( X-VIVO 15 supplemented with 2 mM L-Glutamine , 100 ng/ml M-CSF , 100 U/ml penicillin and 100 µg/ml streptomycin ) . iPSC-derived macrophages were harvested from EB cultures , counted and seeded at 25 , 000 cells per well into 96-well tissue culture plates in triplicates per condition . Macrophages were cultured in macrophage differentiation medium for 7 days , and then activated with 50 ng/ml LPS ( Sigma-Aldrich ) in fresh macrophage differentiation medium for 4 hr . For inhibitor treatments cells were incubated with 50 ng/ml LPS and 3 µM GW or GI ( synthesized by Iris Biotech ( Marktredwitz , Germany ) and kindly provided by Dr Stefan Düsterhöft ) for 4 hr . Cell culture supernatants were collected and cleared from cells by centrifugation . TNFα in supernatants was measured by ELISA ( Human TNF alpha ELISA Ready-SET-Go , eBioscience ( 88-7346-86 ) ) according to the manufacturer’s instructions . Macrophages were lysed in Trition X-100 lysis buffer and protein concentration was determined using a BCA assay ( Thermo Scientific ) . The amount of TNFα in the supernatant was normalised to the protein concentration of the corresponding cell lysate to adjust for differences in TNFα release due to cell numbers . Commercially available Frmd8-/- mouse ES cells from KOMP Repository at UC Davis were used to generate Frmd8-/- mice . The mouse ES cells ( C57BL/6NTac strain ) were injected into blastocysts of Balb/c mice . Chimeras were bred to C57BL/6 to generate Frmd8+/- mice that were used for breeding of the colony and the generation of Frmd8-/- mice . For mice described in Figure 9—figure supplement 2B , we excised the LoxP-flanked neomycin resistance gene by breeding Frmd8-/- mice with homozygous Sox2-Cre deleter strain mice . The mouse work was performed under project licenses 80/2584 and 30/2306 . Mouse tissues were collected from sacrificed animals and stored on dry ice or at −80°C . Tissues were lysed in Triton X-100 RIPA buffer ( 1% Triton X-100 , 150 mM NaCl , 50 mM Tris-HCl ( pH 7 . 5 ) , 0 . 1% SDS , 0 . 5% sodium deoxycholate ) supplemented with EDTA-free protease inhibitor mix and 10 mM 1 , 10-Phenanthroline using a tissue homogeniser ( Omni International ) . Lysates were cleared from cell debris by centrifugation ( 20 , 000 g , 4°C , 10 min ) . Protein concentrations of tissue lysates were determined using a BCA assay . Values are expressed as means of at least three independent experiments with error bars representing the standard deviation . Unpaired , two‐tailed t‐tests were used for statistical analysis . Shedding assays and ELISA data was analysed using a Mann-Whitney test . Flow cytometry blots shown represent one from at least three experiments with similar outcome . Human iPSC lines were derived from dermal fibroblasts of donors that had given signed informed consent for the derivation of human iPSC lines from skin biopsies and SNP analysis ( Ethics Committee: National Health Service , Health Research Authority , NRES Committee South Central , Berkshire , UK ( REC 10/H0505/71 ) ) . All procedures on mice were conducted in accordance with the UK Scientific Procedures Act ( 1986 ) under a project license ( PPL ) authorized by the UK Home Office Animal Procedures Committee , project licenses 80/2584 and 30/2306 , and approved by the Sir William Dunn School of Pathology Local Ethical Review Committee . We used HEK293T cells ( RRID: CVCL_0063 ) for analysis of protein-protein interactions , subcellular localisation and loss-of-function experiments . These cells were used for experiments that provided a strong platform of in vitro evidence of a relationship between FRMD8 and iRhoms , prior to the generation of iPSC-derived macrophages and FRMD8 knock-out mice . The HEK293T cell line has been tested negative for mycoplasma contamination . | Cells in the human body communicate with one another for many different reasons , including to help organs develop correctly and to produce a healthy reponse to injury and infection . Signalling proteins , such as growth factors and cytokines , form the main language of this communication . Initially , many growth factors and cytokines remain attached to the surface of the cell that made them . When cells need to send a message to another one , an enzyme called ADAM17 acts like a pair of scissors to release the proteins from the cell surface , allowing them to travel towards other cells . This process must be carefully controlled because releasing too many growth factors or cytokines ( or releasing them at inappropriate times ) can lead to cancer and inflammatory diseases such as rheumatoid arthritis . Another group of proteins called iRhoms bind to ADAM17 to regulate the enzyme’s activity . But what controls the activity of the iRhom proteins themselves ? To find out , Künzel et al . used a technique called a proteomic screen that can identify which proteins bind to each other . This revealed that a protein called FRMD8 binds to iRhoms . Further experiments in human cells and mice revealed that FRMD8 maintains adequate levels of both ADAM17 and iRhoms at the surface of the cell . Cells that lack FRMD8 break down ADAM17 and iRhom proteins and release fewer growth factors and cytokines . Further work could help us to learn whether stopping FRMD8 from interacting with iRhoms could reduce cell communication . This , in turn , might reduce inflammation or cell growth . If so , then developing drugs that prevent FRMD8 from binding to iRhoms could lead to new treatments for inflammatory diseases and cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2018 | FRMD8 promotes inflammatory and growth factor signalling by stabilising the iRhom/ADAM17 sheddase complex |
Bacterial type IV secretion systems ( T4SSs ) are molecular machines that can mediate interbacterial DNA transfer through conjugation and delivery of effector molecules into host cells . The Helicobacter pylori Cag T4SS translocates CagA , a bacterial oncoprotein , into gastric cells , contributing to gastric cancer pathogenesis . We report the structure of a membrane-spanning Cag T4SS assembly , which we describe as three sub-assemblies: a 14-fold symmetric outer membrane core complex ( OMCC ) , 17-fold symmetric periplasmic ring complex ( PRC ) , and central stalk . Features that differ markedly from those of prototypical T4SSs include an expanded OMCC and unexpected symmetry mismatch between the OMCC and PRC . This structure is one of the largest bacterial secretion system assemblies ever reported and illustrates the remarkable structural diversity that exists among bacterial T4SSs .
Bacterial pathogens are a threat to global health and have evolved elaborate strategies to infect their hosts . Many effects of bacteria on host cells require the actions of bacterial secretion systems . Bacterial type IV secretion systems ( T4SS ) are a diverse class of molecular machines that mediate interbacterial DNA transfer through conjugation as well as delivery of effector proteins into host cells . T4SSs are found in a wide range of bacterial species , including many species that cause human disease , such as Helicobacter pylori , Legionella pneumophila , Bordetella pertussis , Brucella , and Bartonella ( Christie et al . , 2014; Grohmann et al . , 2018 ) . T4SSs in Gram-negative bacteria contain a minimum of 12 components ( designated VirB1-VirB11 and VirD4 in prototype systems ) , organized into an outer membrane core complex ( OMCC ) , an inner membrane complex ( IMC ) , and in some species an extracellular pilus ( Christie et al . , 2014; Grohmann et al . , 2018; Waksman , 2019 ) . High-resolution structures have been determined for OMCCs from two minimized T4SSs ( Xanthomonas citri T4SS and a portion of the OMCC from the pKM101 conjugation system ) ( Sgro et al . , 2018; Chandran et al . , 2009 ) . T4SSs in several bacterial species , including Helicobacter pylori and Legionella pneumophila , contain additional components , which are not present in the minimized systems ( Grohmann et al . , 2018; Frick-Cheng et al . , 2016; Schroeder , 2017; Ghosal et al . , 2017; Chang et al . , 2018; Chetrit et al . , 2018; Ghosal et al . , 2019; Hu et al . , 2019 ) . The H . pylori Cag T4SS is of particular interest because of its role in translocating CagA ( a bacterial oncoprotein ) into host cells , an important step in gastric cancer pathogenesis ( Fischer , 2011; Backert et al . , 2017 ) . In addition to containing unique components , the H . pylori Cag T4SS has an OMCC much larger in size and more intricate than those in minimized systems , and contains a periplasmic sub-complex not seen in minimized T4SSs ( Grohmann et al . , 2018; Chang et al . , 2018; Hu et al . , 2019 ) . Here we report the use of single particle cryo-EM to determine the structure of a transmembrane Cag T4SS complex extracted and purified from H . pylori . The structure can be divided into three major regions: the OMCC , a periplasmic ring complex ( PRC ) , and a central stalk . The OMCC has a structural organization markedly different from that of T4SS OMCCs in other species , and there is an unexpected symmetry mismatch between the 14-fold-symmetric OMCC and a contiguous 17-fold-symmetric PRC . We propose that the observed structural differences between the Cag T4SS and previously described minimized T4SSs have important functional implications for how T4SSs secrete various kinds of effectors .
Cag T4SS complexes were purified from H . pylori as described ( Frick-Cheng et al . , 2016 ) and visualized by single particle cryo-EM ( Figure 1 and Figure 1—figure supplement 1A , B ) . The resulting structure is a large mushroom-shaped complex , ~410 Å wide by ~460 Å long , with features that closely match those of the Cag T4SS detected in intact H . pylori using cryo-electron tomography ( Chang et al . , 2018; Hu et al . , 2019 ) ( Figure 1A , B ) . The global resolution of the map is 5 . 4 Å when no symmetry is imposed , with the highest resolution regions found near the center of the complex ( Figure 1A and Figure 1—figure supplement 1C–E ) . Although particles adopt a preferred orientation in vitrified ice , both en face and side views are present , which allowed for 3D reconstruction ( Figure 1—figure supplement 1B , F ) . The map can be divided into three major regions: an intricately organized domed cap that is associated with the outer membrane ( OMCC ) , a hollow ring-like mid-region localized within the periplasm ( PRC ) , and a tapered density that extends from the PRC to the inner membrane ( Stalk ) ( Figure 1A and Video 1 ) . Blurry density around the PRC , also visible in cryo-ET images ( Chang et al . , 2018; Hu et al . , 2019 ) , may represent a dynamic or less-structured portion of the T4SS ( Figure 1B ) . Using symmetry , focused refinement , and symmetry expansion , we determined a 3 . 7 Å resolution map of the OMCC and a 3 . 5 Å map of the PRC ( Figure 1C , Figure 1—figure supplement 2 , Videos 2 and 3 ) . The resolution of the OMCC and PRC maps made it possible to trace the secondary structure in these regions ( Figure 1D and Videos 2 and 3 ) . Since only a small number of T4SS particles seen in vitrified ice contained the Stalk density , it was not possible to determine the symmetry or obtain high resolution of this part of the complex . An axial section through the map in Figure 1D shows a large cavity running through the Cag T4SS , starting from where the OMCC spans the outer membrane ( OM ) and extending to the bottom of the PRC ( Figure 1E , F ) . The tapered end of the Stalk begins in the PRC channel and continues through the inner membrane ( IM ) . Models developed from cryo-ET analyses of both the H . pylori and L . pneumophila T4SSs propose a channel in this region of the complex ( Ghosal et al . , 2017; Chang et al . , 2018; Chetrit et al . , 2018; Ghosal et al . , 2019; Hu et al . , 2019 ) . A central section through the longitudinal plane of 3D density suggests there may be a channel that runs through the Stalk ( Figure 1B , F ) . However , due to the low resolution of the Stalk , this channel cannot be clearly visualized in the 3D map . The OMCC , with 14-fold symmetry , makes up the ‘mushroom cap’ of the T4SS and is organized into a central and outer ring connected by ‘spokes’ ( Figure 2 ) . The outer ring of the OMCC surrounds a central chamber that is ~270 Å wide and tapers to a ~ 35 Å opening at the top ( Figure 2B ) . As seen in OMCCs from minimized T4SSs ( Sgro et al . , 2018; Chandran et al . , 2009 ) , the H . pylori Cag T4SS OMCC contains both an outer layer ( O-layer ) and a thinner inner layer ( I-layer ) ( Figure 3A ) . ‘Spokes’ and outer rings are visible in both the O-layer and I-layer of the Cag T4SS ( Figure 3B and Figure 3—figure supplement 1 ) , but are not present in OMCCs from minimized T4SSs . The resolution of the cap is high enough to begin mapping the molecular organization of individual OMCC components ( Figure 4A , B ) . Previous mass spectrometry and Western blot analyses indicated that the isolated complexes contain CagY , CagX , CagM , CagT , and Cag3 ( Frick-Cheng et al . , 2016 ) . In the map , we can trace and identify portions of CagY ( residues 1677–1816 and 1850–1907 ) , CagX ( residues 349–510 ) , and CagT ( residues 26–269 ) ( Figure 4C and Figure 4—figure supplement 1A–C , left panels ) . CagY forms the crest of the cap-like structure of the OMCC and is comprised of β-sheets intertwining to position two helices per protomer atop the cap ( Figure 4A–C and Figure 4—figure supplement 1A , left panel ) . We predict that these helices breach the outer membrane , resulting in the formation of a channel ( Figure 4B ) . The C-terminal portion of CagX is comprised of two β-sheets preceded by a long helix ( Figure 4C and Figure 4—figure supplement 1B , left panel ) . This helix extends from the top of the O-layer through the central chamber into the I-layer of the OMCC ( Figure 4B ) . Adjacent to CagX , we traced a continuous chain 243 residues long corresponding to CagT ( Figure 4C ) . It includes a globular subdomain consisting of two β-sheets followed by a long C-terminal extension of three helices that contribute to the spoke and extend toward the edge of the map ( Figure 4B and Figure 4—figure supplement 1C , left panel ) . A predicted lipidation site within the N-terminal tail of CagT ( Fischer , 2011; Akopyants et al . , 1998 ) was not observed in the cryo-EM map , but the N-terminus of CagT is positioned for this interaction with the outer membrane . Extending to the edge of the O-layer , we have constructed poly-alanine models of protein ( s ) that we could not identify ( Figure 4A , B and Figure 4—figure supplement 1D ) . These components fill the remaining density within the spokes and consist predominantly of a repetitive fold composed of repeating units of β-sheets flanked by helices ( Figure 4A , B and Figure 4—figure supplement 1D ) . We predict that Cag3 is a component of the OMCC periphery based on a previous study showing that Cag T4SS complexes isolated from a Δcag3 mutant strain lacked peripheral components of the OMCC ( Frick-Cheng et al . , 2016; Hu et al . , 2019 ) , but we were unable to obtain a register . Within the I-layer we have observed at least two distinct bundles of helices of ~200 residues and ~300 residues ( Figure 4A , B and Figure 4—figure supplement 1E ) . Based on structural studies of the X . citri T4SS ( Grohmann et al . , 2018 ) , the I-layer of the Cag T4SS is predicted to contain portions of CagY and CagX , but none of the poly-alanine models allowed us to unambiguously attribute any portion of CagY , CagX , or other T4SS components to this region . The PRC , with 17-fold symmetry ( Figure 4D , E and Figure 4—figure supplement 2 ) , is a short hollow tube , 90 Å tall and 185 Å wide with 96 Å internal diameter ( Figure 4D , E ) , connecting the OMCC and the Stalk regions of the Cag T4SS ( Figure 1A ) . Although PRCs have not been detected in structural studies of E . coli or X . citri T4SSs ( Waksman , 2019 ) , this region was identified in cryo-ET studies of both H . pylori and L . pneumophila T4SSs ( Ghosal et al . , 2017; Chang et al . , 2018; Chetrit et al . , 2018; Ghosal et al . , 2019; Hu et al . , 2019 ) . The protein ( s ) in this region of the cryo-EM maps were modeled as three distinct polyalanine chains ( Figure 4—figure supplement 1F–H ) . This portion of the H . pylori Cag T4SS structure is predicted to contain CagM ( Waksman , 2019; Chang et al . , 2018 ) , but the three polyalanine chains could not be unambiguously attributed to CagM or other T4SS components . A symmetry mismatch occurs between the OMCC and PRC , going from 14-fold symmetry in the OMCC to 17-fold symmetry in the PRC . While the OMCC , PRC , and Stalk make physical contact in the lower resolution structure with no applied symmetry ( Figure 5 ) , in the refined structures these connections are lost because of the symmetry mismatch . Analysis of the 5 . 4 Å structure suggests that the connection between OMCC and PRC may occur by contacts between the 14 α-helices extending from the N-terminus of CagX and the 17 unidentified α-helices extending up from the PRC ( Figure 1A , E , and Figure 5 ) . The tips of these α-helices are presumed to be flexible , since the corresponding densities are not well defined in the higher resolution maps ( Figure 4D ) .
Overall , the H . pylori OMCC is much larger ( 410 Å diameter ) than other structurally characterized OMCCs ( 225 Å in X . citri and 170 Å in pKM101 conjugation system ) ( Sgro et al . , 2018; Chandran et al . , 2009 ) ( Figure 6 ) , and the components are more intertwined . While the inner chamber of the H . pylori OMCC is larger than what is seen in the E . coli and X . citri structures , the dimension of the H . pylori outer membrane pore , at 35 Å , is smaller than the X . citri outer membrane pore ( 45 Å ) but larger than the E . coli outer membrane pore ( 25 Å ) ( Figure 6B ) . The dimensions of the H . pylori OMCC resemble those of the L . pneumophila Dot/Icm T4SS ( Ghosal et al . , 2017; Chetrit et al . , 2018; Ghosal et al . , 2019 ) , but there is very little sequence relatedness when comparing components of the H . pylori and L . pneumophila T4SSs , and the L . pneumophila T4SS OMCC has 13-fold symmetry instead of 14-fold symmetry ( Chetrit et al . , 2018; Ghosal et al . , 2019 ) . The C-terminal portion of CagY is structurally similar to VirB10 and TraF homologs ( Figure 6 and Figure 4—figure supplement 1A ) , and the C-terminal portion of CagX is structurally similar to VirB9 and TraO homologs ( Figure 6 and Figure 4—figure supplement 1B ) . Both CagY and CagX are much larger in size than characterized VirB/Tra homologs ( Fischer , 2011; Backert et al . , 2017 ) , and correspondingly , CagX has a long α-helix that is absent in the structures of TraO or VirB9 ( 4 , 5 ) ( Figure 6 and Figure 4—figure supplement 1B ) . Consistent with the limited sequence relatedness of CagT to VirB7 homologs ( Fischer , 2011; Backert et al . , 2017 ) , there is relatively little structural relatedness when comparing CagT with VirB7 except for an N-terminal portion that contains little secondary structure ( Figure 6 and Figure 4—figure supplement 1C ) . In summary , these results provide the first high resolution structure of a transmembrane complex from a non-canonical T4SS and provide important new insights into a bacterial T4SS that contributes to the pathogenesis of gastric cancer . The Cag T4SS structure differs markedly from structures of previously described T4SSs , including the presence of an expanded OMCC with components that are more structurally intertwined and an unexpected symmetry mismatch between the OMCC and PRC . We predict that these differences will have important functional implications for the mechanism of CagA translocation .
The purification of the Cag T4SS complex was done using a previously described approach ( Frick-Cheng et al . , 2016 ) with reduced deoxycholate concentration ( 0 . 025% ) . For cryo-EM , 5 ul of the Cag T4SS sample ( as purified ) was applied to a glow discharged ultrathin continuous carbon film on Lacey 400 mesh copper grids ( TED PELLA ) . The sample was applied to a grid , incubated for 60 s and vitrified by plunge-freezing in a slurry of liquid ethane using a FEI Vitrobot at 4°C and 100% humidity . All the images were collected on the Titan Krios electron microscope ( Thermo Fisher ) equipped with a K2 Summit Direct Electron Detector ( Gatan ) operated at 300 kV and having a nominal pixel size of 1 . 64 Å per pixel . Micrographs were acquired using Leginon software ( Suloway et al . , 2005 ) . The total exposure time was 8 s and frames were recorded every 0 . 2 s , resulting in a total accumulated dose of ~60 e− Å−2 using a defocus range of −0 . 5 to −3 . 5 μm . All the Video frames were first dose-weighted and aligned using Motioncor2 ( Zheng et al . , 2017 ) . The contrast transfer function ( CTF ) values were determined by Gctf ( Zhang , 2016 ) . Image processing was carried out using cryoSPARC , RELION 2 . 1 and 3 . 0 ( Punjani et al . , 2017; Scheres , 2012; Zivanov et al . , 2018 ) . Using RELION , approximately 25 , 000 particles were manually picked from 4600 micrographs and extracted using a box size of 510 pixels ( 836 . 4 Å ) ( Supplementary file 1 ) . The extracted particles were exported to cryoSPARC and used to generate representative two-dimensional ( 2D ) class averages in both cryoSPARC and RELION , and approximately 24 , 000 ( cryoSPARC ) and 23 , 000 ( RELION ) particles were kept in good class averages . These particles were then subjected to 3D classification with a reference-free initial 3D model . The best 3D class ( ~17 , 000 in cryoSPARC and ~20 , 000 particles in RELION ) was used as reference for 3D auto-refinement with and without C14 symmetry ( low-pass filtered to 60 Å ) . Finally , a solvent mask and B-factor were applied to improve the overall features and resolution of the 3D maps with and without C14 symmetry , resulting in reconstruction of 3D maps with a global resolution of 4 . 1 Å and 5 . 4 Å , respectively ( Supplementary file 1 and Figure 1—figure supplement 1 ) . Estimation of per-particle defocus values ( CTF-refinement ) was applied to the selected particles using RELION . With the CTF-refined particle stack , C14 symmetry-imposed refinement with a soft mask around the OMCC region of the Cag T4SS core complex was done , resulting in a 3 . 8 Å resolution 3D map that contained improved features ( Figure 1—figure supplement 1 and Figure 1—figure supplement 2C ) . Based on the inspection of the asymmetrically refined 3D map showing 14-fold symmetry , relion_particle_symmetry_expand was used to enlarge the particle stack ( Zivanov et al . , 2018 ) . After signal subtraction and alignment-free 3D classification of the expanded particles using a soft mask around an asymmetric unit ( ~290 , 000 particles ) , one highly populated 3D class was produced which contains ~250 , 000 expanded particles . The 3D class was then subjected to the masked 3D refinement with local angular searches , resulting in reconstruction of the 3D map at 3 . 7 Å resolution ( Figure 1—figure supplement 2D ) . All resolutions were calculated using the gold-standard 0 . 143 FSC . To resolve the lower region of the core complex ( PRC ) , signal subtraction for each individual particle containing the PRC signal was used with a soft mask ( Supplementary file 1 and Figure 4—figure supplement 2A ) . The subtracted particles were then subjected to the alignment-free focused 3D classification ( five classes ) . Visual inspection of the 3D classes showed that the PRC has 17-fold symmetry . The best 3D class of the PRC ( ~20 , 000 particles ) was then subjected to a masked 3D refinement with local angular searches . This process was done using either C1 , C12 , C14 or C17 symmetry . Only the reconstruction using C17 symmetry improved in resolution , resulting in 3D reconstruction of the PRC at 3 . 5 Å ( Figure 4—figure supplement 2C–E ) . Resolutions were calculated using the gold-standard 0 . 143 FSC . Image processing steps are summarized in Supplementary file 1 . There were no high-resolution structures for any of the known components of the OMCC except for a crystal structure of a small region in CagX ( Zhang et al . , 2017 ) . Density maps of the OMCC and PRC were of sufficient quality for building de novo models of regions of CagY , CagX , and CagT in COOT ( Figure 4 and Figure 4—figure supplement 1 ) , facilitated by the small crystal structure of CagX and homology models of OMCC components from E . coli and X . citri ( Sgro et al . , 2018; Chandran et al . , 2009 ) . A homology model for a C-terminal region of CagY was constructed from VirB10 ( PDB-3JQO ) using the Swiss Model server . The CagX crystal structure ( PDB-5H3V ) and the CagY model were placed within the asymmetric unit of the electron density map using UCSF Chimera ( Pettersen et al . , 2004 ) . These models were manually adjusted and extended in COOT ( Emsley and Cowtan , 2004 ) . Density corresponding to CagT was modeled de novo in COOT , and the three proteins ( CagT , CagX , and CagY ) were further refined in real space using Phenix while applying secondary structure restraints ( Adams et al . , 2002 ) . Once the asymmetric unit was constructed , 14-fold symmetry was applied in Phenix , and the entire model of the OMCC was subjected to one more round of refinement . Secondary structure elements were built de novo into the remaining density of the OMCC and the PRC asymmetric units in COOT . These models were refined in Phenix as described above , and the entire model was generated by applying 14-fold and 17-fold symmetry to the OMCC and PRC , respectively . The resolution of each individual model was estimated by Fourier Shell correlation against the map used to construct it using the Phenix Cryo-EM Validation tool . Molprobity scores , Clashscores and Ramachandran plots were used to validate the models that were constructed ( Chen et al . , 2010; Afonine et al . , 2018 ) ( Supplementary file 2 ) . Supplementary file 3 shows the FSCs of the half maps against the refined model agree with each other , suggesting that the models are not over-refined . Programs used for structure determination and refinement were accessed through SBGrid ( Morin et al . , 2013 ) . Structures were rendered using Chimera and ChimeraX ( Pettersen et al . , 2004; Goddard et al . , 2018 ) . The cryo-EM volumes have been deposited in the Electron Microscopy Data Bank under accession codes EMD-20023 ( T4SS C1 reconstruction ) , EMD-20020 ( Focused OMCC Reconstruction ) , EMD-20022 ( OMCC Asymmetric Reconstruction ) , EMD-20021 ( Focused PRC Reconstruction ) . Map coordinates have been deposited in the Protein Data Bank under accession numbers 6OEE ( CagT ) , 6OEG ( CagX ) , and 6ODI ( CagY ) , 6OEF ( O-layer ) , 6OEH ( I-Layer ) , and 6ODJ ( PRC ) . | Helicobacter pylori is a species of bacterium that can colonize the human stomach , causing changes that greatly increase the risk of ulcers and stomach cancer . Some strains of H . pylori produce a protein called CagA , which alters how stomach cells grow and divide . The bacterium injects CagA directly into stomach cells using a syringe-like structure called a type IV secretion system . Type IV secretion systems are found in many species of bacteria and are involved in a variety of processes , including the exchange of genes between neighboring bacteria . The systems typically have at least 12 components . Previous studies have revealed how the components of some of these systems fit together to form working machines . However , the type IV secretion system that delivers CagA ( called the Cag T4SS ) contains additional components and it remains unclear how these components are organized in the structure . A technique called cryo-electron microscopy uses electrons to visualize proteins that have been rapidly frozen so they can be captured and imaged in their natural shape and form . Chung , Sheedlo et al . extracted the Cag T4SS apparatus directly from H . pylori and used cryo-electron microscopy to determine its shape to a high level of detail . These images were then used to build a detailed model of the Cag T4SS that included many of its components . The model shows that the Cag T4SS is larger and more complex than other type IV secretion systems that have been studied previously . Therefore , Chung , Sheedlo et al . propose that the Cag T4SS is specially adapted to work in the stomach . These findings open the door for future research to define how individual components of the Cag T4SS help to inject CagA into stomach cells . In addition , future research will allow researchers to understand how the type IV secretion systems found in different bacterial species carry out a wide range of roles . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | Structure of the Helicobacter pylori Cag type IV secretion system |
Differentiation programs such as meiosis depend on extensive gene regulation to mediate cellular morphogenesis . Meiosis requires transient removal of the outer kinetochore , the complex that connects microtubules to chromosomes . How the meiotic gene expression program temporally restricts kinetochore function is unknown . We discovered that in budding yeast , kinetochore inactivation occurs by reducing the abundance of a limiting subunit , Ndc80 . Furthermore , we uncovered an integrated mechanism that acts at the transcriptional and translational level to repress NDC80 expression . Central to this mechanism is the developmentally controlled transcription of an alternate NDC80 mRNA isoform , which itself cannot produce protein due to regulatory upstream ORFs in its extended 5’ leader . Instead , transcription of this isoform represses the canonical NDC80 mRNA expression in cis , thereby inhibiting Ndc80 protein synthesis . This model of gene regulation raises the intriguing notion that transcription of an mRNA , despite carrying a canonical coding sequence , can directly cause gene repression .
Cellular differentiation programs depend on temporally controlled waves of gene activation and inactivation . These waves in turn drive the morphogenetic events that ultimately transform one cell type into another . Differentiation models ranging from Bacillus subtilis sporulation to mouse embryogenesis have elucidated how transcription factor handoffs temporally activate the expression of gene clusters ( Errington , 2003; Zernicka-Goetz et al . , 2009 ) . In comparison , much less is understood about how gene repression is coordinated with the transcription factor-driven waves of gene expression and how this inactivation is mechanistically achieved . One critical morphogenetic event that relies on inactivation is the loss of kinetochore function during meiotic prophase . The kinetochore is a protein complex that binds to centromeric DNA and serves as the attachment site for spindle microtubules to mediate chromosome segregation ( Musacchio and Desai , 2017 ) ( Figure 1A ) . In multiple systems , it has been shown that kinetochores do not bind to microtubules in meiotic prophase ( Asakawa et al . , 2005; Kim et al . , 2013; Meyer et al . , 2015; Miller et al . , 2012; Sun et al . , 2011 ) . Furthermore , this temporal inactivation is achieved through removal of the outer kinetochore , the site where microtubule attachments occur ( Asakawa et al . , 2005; Kim et al . , 2013; Meyer et al . , 2015; Miller et al . , 2012; Sun et al . , 2011 ) ( Figure 1B ) . In the presence of a spindle , cells that fail to disassemble the outer kinetochore undergo catastrophic missegregation of meiotic chromosomes , underlying the essential nature of kinetochore downregulation during meiotic prophase ( Miller et al . , 2012 ) . Importantly , the kinetochore is reactivated when the outer kinetochore reassembles upon transition from prophase to the meiotic divisions . How the initial removal and subsequent reassembly of the outer kinetochore is coordinated with the meiotic gene expression program is unknown . Budding yeast provides a powerful model to address how the dynamic regulation of kinetochore function is integrated into the meiotic gene expression program . Entry into meiosis marks a clear cell-fate transition defined by the induction of Ime1 , a master transcription factor . Ime1 activates the expression of genes involved in DNA replication and meiotic recombination ( Kassir et al . , 1988; van Werven and Amon , 2011 ) . Successful completion of recombination , in turn , induces a second transcription factor Ndt80 , which activates the expression of genes involved in meiotic divisions and gamete development ( Chu and Herskowitz , 1998; Xu et al . , 1995 ) . Thus , the landmark morphogenetic events in budding yeast meiosis are coordinated by the relay between these two transcription factors . Furthermore , a high-resolution map of the gene expression waves that drive meiosis has been generated for budding yeast ( Brar et al . , 2012 ) . Importantly , analysis of this dataset revealed that , of the 38 genes that encode kinetochore subunits , NDC80 displays the most regulated expression pattern between meiotic prophase and the subsequent division phases ( Miller et al . , 2012 ) . Ndc80 is the namesake member of an evolutionarily conserved complex that forms the microtubule-binding interface of the outer kinetochore ( Tooley and Stukenberg , 2011 ) ( Figure 1A ) . Numerous lines of evidence indicate that the tight regulation of NDC80 is essential for the timely function of kinetochores during meiosis . First , the decline of Ndc80 protein in meiotic prophase correlates with the dissociation of the outer kinetochore from the chromosomes ( Kim et al . , 2013; Meyer et al . , 2015; Miller et al . , 2012 ) . Second , even though the other outer kinetochore subunits are expressed in meiotic prophase , they do not localize to the kinetochores ( Meyer et al . , 2015 ) . Third , the subsequent increase in Ndc80 protein coincides with outer kinetochore reassembly ( Meyer et al . , 2015; Miller et al . , 2012 ) . Finally , in the presence of a spindle , prophase misexpression of NDC80 disrupts proper meiotic chromosome segregation ( Miller et al . , 2012 ) . Together , these results indicate that NDC80 regulation is necessary for the proper timing of kinetochore function in meiosis and highlight the importance of controlling Ndc80 protein levels during meiotic differentiation . Here we uncovered how the timely function of kinetochores is achieved through the regulation of Ndc80 protein synthesis during budding yeast meiosis . This mechanism is based on the use of two NDC80 mRNA isoforms , which have opposite functions and display distinct patterns of expression . In addition to the canonical protein-translating NDC80 mRNA , we found that meiotic cells also expressed a 5’-extended NDC80 isoform . Despite carrying the entire NDC80 open reading frame ( ORF ) , this alternate isoform cannot produce Ndc80 protein due to the presence of regulatory upstream ORFs ( uORFs ) in its extended 5’ leader . Rather , its transcription plays a repressive role to inhibit transcription of the canonical NDC80 mRNA in cis and thereby inhibit Ndc80 protein synthesis . Furthermore , we found that the expression of the 5’-extended isoform was activated by the meiotic initiator transcription factor Ime1 . Upon exit from prophase , the mid-meiotic transcription factor Ndt80 activated the expression of the canonical NDC80 mRNA isoform . Taken together , this study uncovers how NDC80 gene repression is achieved and how inactivation and subsequent reactivation of the kinetochore is coordinated with the transcription factor-driven waves of meiotic gene expression .
The Ndc80 complex consists of four subunits , namely Ndc80 , Nuf2 , Spc24 , and Spc25 ( Figure 1A ) . All the subunits other than Ndc80 persist in meiotic prophase ( Meyer et al . , 2015 ) . Consistent with this report , we found that even in an extended meiotic prophase arrest , Ndc80 was the only subunit of its complex whose abundance decreased at this meiotic stage ( Figure 1C and Figure 1—figure supplement 1 ) . Nuf2 , Spc24 , and Spc25 were all expressed , though it has been reported that these proteins fail to localize to the kinetochores during meiotic prophase ( Meyer et al . , 2015 ) . These observations raised the possibility that Ndc80 could be the limiting kinetochore subunit in meiosis . If correct , then the elevation of Ndc80 protein levels , but not the other subunits , should reactivate kinetochore function in meiotic prophase . To test this prediction , we overexpressed each of the Ndc80 complex subunits ( Figure 1—figure supplement 2 ) , in conjunction with the B-type cyclin Clb3 , under an inducible CUP1 promoter ( pCUP ) . CLB3 misexpression causes bipolar spindle assembly in meiotic prophase ( Miller et al . , 2012 ) . In pCUP-CLB3 cells , if kinetochores are functional in meiotic prophase , they attach to the spindle microtubules prematurely . These premature attachments , in turn , cause sister chromatid segregation in meiosis I , essentially disrupting proper meiotic chromosome segregation ( Miller et al . , 2012 ) . When NDC80 was overexpressed in pCUP-CLB3 cells during meiotic prophase , over 30% of the cells displayed an abnormal segregation pattern in meiosis I . In contrast , misexpression of CLB3 alone resulted in only a 7% segregation defect . Importantly , this defect was not further enhanced by the overexpression of NUF2 , SPC24 or SPC25 ( Figure 1D ) . Based on this observation , we conclude that kinetochore function is repressed in meiotic prophase due to limiting levels of Ndc80 . Following prophase , Ndc80 becomes highly abundant during the meiotic divisions ( Miller et al . , 2012 ) ( Figures 1C , 10 h time point ) , consistent with its role in facilitating chromosome segregation ( Wigge and Kilmartin , 2001 ) . Together , these results demonstrate that Ndc80 is the sole subunit of its complex that is tightly regulated during meiotic differentiation and strongly support the notion that NDC80 downregulation and re-synthesis govern kinetochore functionality in meiosis . To dissect the molecular mechanism for the strict temporal regulation of the NDC80 gene in meiosis , we first took advantage of the high-resolution RNA-seq and ribosome profiling dataset generated for budding yeast meiosis ( Brar et al . , 2012 ) . Analysis of this dataset revealed the presence of meiosis-specific RNA-seq reads that extend to ~500 base pairs ( bp ) upstream of the NDC80 ORF ( Figure 2A ) . These reads appeared after meiotic entry and persisted until the end of meiosis , but were absent during vegetative growth ( Figure 2—figure supplement 1 , vegetative ) or starvation ( Figure 2—figure supplement 1 , MATa/MATa ) . To monitor the different RNA molecules generated from the NDC80 locus , we performed northern blotting . In the absence of meiotic progression , when cells were subject to nutrient poor conditions , we detected only a single NDC80 transcript throughout the starvation regime ( no CuSO4 , Figure 2—figure supplement 2A ) . However , in cells undergoing synchronous meiosis , two distinct NDC80 transcript isoforms became evident: a longer , meiosis-specific isoform , and a shorter isoform that was also present under non-meiotic conditions ( Figure 2B and Figure 2—figure supplement 2 ) . The longer isoform appeared after meiotic entry , persisted throughout meiotic prophase and gradually disappeared during the meiotic divisions . The shorter isoform was present in vegetative cells prior to meiotic entry , but was weakly expressed during S phase and meiotic prophase . Its abundance dramatically increased during the meiotic divisions ( Figure 2B and Figure 2—figure supplements 2B and 3 ) . Interestingly , the Ndc80 protein levels were noticeably higher during the meiotic stages when the shorter transcript was the predominant isoform , but lower when the longer transcript was predominant ( Figure 2B ) . In addition to northern blotting , we used single molecule RNA fluorescence in situ hybridization ( smFISH ) to assess the cell-to-cell variability in transcript expression and subcellular localization of these two NDC80 transcript isoforms . With two sets of probes that bind to the same region of NDC80 ORF ( odd/even probes ) , we verified that our smFISH could uniquely pair the FISH spots from these two probe sets with an accuracy of 88% ( Figure 2—figure supplement 4 ) , a value similar to what was reported previously ( Raj et al . , 2008 ) . Furthermore , we confirmed that the number of cells analyzed per sample per experimental repeat ( >95 cells ) exceeded the minimal number of cells required to achieve a stable sampling average ( Figure 2—figure supplement 5 ) , and thus our sample size is large enough to reflect the population mean . To differentiate between the two NDC80 isoforms , we used another two sets of probes: one set ( Q 670 ) , conjugated to Quasar 670 , is complementary to the sequences common between the short and the long isoforms . The other set ( CF 590 ) , conjugated to CAL Fluor Red 590 , is unique to the long isoform . The long isoforms were identified as the spots where the signal from both probe sets colocalized , whereas the short isoforms were identified as the spots with signal only from Q 670 ( Figure 2C and D ) . The smFISH analysis revealed that the expression of the two NDC80 isoforms was temporally regulated . Vegetative cells expressed only the short NDC80 isoform; fewer than 2% of these cells expressed the long isoform ( Figure 2C and D ) . In meiotic prophase , a stage defined by the presence of the synaptonemal complex component Zip1 , 100% of cells expressed the long isoform , and over 50% of them had more than 20 transcripts per cell . During the same stage , the level of the short isoform significantly decreased in comparison to its levels in vegetative growth ( p=0 . 0260 , two-tailed Wilcoxon Rank Sum test , Figure 2D ) and pre-meiotic starvation ( p=0 . 0090 , Figure 2—figure supplement 6 ) . As cells entered meiosis I , the level of the short isoform dramatically increased while that of the long isoform declined , in comparison to the levels of these isoforms during meiotic prophase ( p<0 . 0001 for both NDC80short and NDC80long mRNAs , Figure 2D ) . Thus , the two NDC80 isoforms have expression signatures specific to different cellular states . In addition , the two NDC80 isoforms localized to both the nucleus and cytoplasm ( Figure 2C ) . We saw no evidence that the NDC80long isoform was solely retained in the nucleus; all of the Zip1-positive cells had at least one NDC80long mRNA localized outside of the DAPI-stained region . This localization pattern was consistent with the possibility that both transcripts were translated , as shown by ribosome profiling ( Figure 2A , bottom panel ) ( Brar et al . , 2012 ) . Altogether , the combined analyses of northern and western blotting , as well as smFISH , reveal two interesting trends: ( 1 ) In meiosis , the expression of the long and short NDC80 isoforms are anti-correlated . ( 2 ) Ndc80 protein levels positively correlate with the presence of the short isoform and negatively correlate with the long isoform ( Figure 2B ) . The negative correlation between the longer NDC80 isoform and Ndc80 protein levels suggested that this longer isoform was unable to support the synthesis of Ndc80 protein . In addition to the NDC80 ORF , the longer isoform contains nine uORFs , each with an AUG start codon . The first six of these uORFs , those closest to the 5’ end of the mRNA , have ribosome profiling signatures consistent with them being translated in meiosis ( Figure 3—figure supplement 1 ) . Upstream start codons in transcript leaders can capture scanning ribosomes to alternate reading frames , thereby restricting ribosome access to the main ORF ( Arribere and Gilbert , 2013; Calvo et al . , 2009; Johnstone et al . , 2016 ) . We mutated the start codon of the first six uORFs ( ∆6AUG ) to test whether translation of the uORFs within the longer NDC80 isoform represses translation of Ndc80 protein from this mRNA . In the ∆6AUG strain , the negative correlation between the long isoform and Ndc80 protein level persisted ( Figure 3 ) , potentially because translation of the remaining three uORFs could still repress translation of the ORF . Indeed , when all nine AUGs were mutated , Ndc80 protein became highly abundant during meiotic prophase , even though the long isoform remained the predominant NDC80 transcript in these cells ( Figure 3 ) . These results demonstrate that although the longer isoform of NDC80 contains the entire ORF , the presence of the uORFs in its 5’ leader prevents Ndc80 translation from this mRNA . Next , we tested whether the repressive role of the uORFs resulted from the act of translation or the peptides encoded by these uORFs . We modified the long isoform , such that it still contained all the upstream AUG start codons , but each start codon was followed by a single amino acid and then immediately by a stop codon ( mini uORF ) . Thus , this construct retained the translation ability of the uORFs but rendered them incapable of producing a peptide chain . We found that Ndc80 levels were still reduced during meiotic prophase in the mini uORF strain ( Figure 3 ) . Therefore , translation of the uORFs represses translation of the NDC80 ORF from the long NDC80 isoform , rendering this isoform unable to synthesize Ndc80 protein . Our analyses so far demonstrate that the two NDC80 mRNA isoforms differ with regards to their size and ORF coding capacity . The shorter isoform is capable of translating NDC80 ORF . In contrast , although the longer isoform contains the entire ORF , it does not support Ndc80 synthesis . The coding information is not decoded from this isoform because uORF translation prevents ribosomes from accessing the actual ORF . To signify the unique features of each NDC80 transcript isoform , we named the short mRNA NDC80ORF , and the longer mRNA NDC80luti for long un-decoded transcript isoform . Given that NDC80luti does not appear to produce Ndc80 protein , we set out to understand why meiotic cells express this mRNA isoform . Based on the observation that the expression levels of these two isoforms are anti-correlated , we posited that the transcription of NDC80luti represses NDC80ORF . To test this hypothesis , we first eliminated NDC80luti production by deleting its promoter along with different portions of the NDC80luti transcript ( ∆NDC80luti , Figure 4—figure supplement 1 ) . As shown by northern blotting , NDC80ORF was detected during meiotic prophase in two different ∆NDC80luti mutant strains ( Figure 4A and Figure 4—figure supplement 2 ) . Analysis of smFISH also confirmed that the level of NDC80ORF in ∆NDC80luti cells significantly increased during meiotic prophase ( Figure 4B and C , p=0 . 0004 ) , with a median exceeding that of pre-meiotic cells ( Figure 2—figure supplement 6 ) . Accordingly , Ndc80 protein levels increased throughout meiotic prophase ( Figure 4A ) . Additionally , we inserted a termination sequence ~220 bp downstream of the NDC80luti transcription start site ( NDC80luti-Ter ) . We observed that , upon early termination of NDC80luti , NDC80ORF mRNA and Ndc80 protein persisted in meiotic prophase ( Figure 4—figure supplement 3 ) . This observation suggests that continuous transcription through the NDC80ORF promoter is necessary for NDC80ORF repression . It also indicates that the repression of NDC80ORF is not due to competition between the NDC80ORF promoter and the NDC80luti promoter for RNA polymerase and the general transcription machinery . Altogether , we conclude that expression of the NDC80luti mRNA is required to repress the NDC80ORF transcript and reduce Ndc80 protein levels during meiotic prophase . By what mechanism does NDC80luti reduce the steady-state level of NDC80ORF ? We posited that NDC80luti acts in cis based on other instances of overlapping transcription in budding yeast ( Bird et al . , 2006; Martens et al . , 2004; van Werven and Amon , 2011 ) . To test this , we engineered strains to have one wild type NDC80luti allele and another allele in which the promoter of NDC80luti has been deleted ( ∆NDC80luti ) . In order to monitor Ndc80 protein levels , we inserted a 3V5 epitope as a C-terminal fusion to NDC80 in either the wild type or the ∆NDC80luti allele . If NDC80luti functions in trans , then Ndc80-3V5 should be downregulated to the same extent in both strains . Instead , we found that Ndc80-3V5 was downregulated only when NDC80luti was generated on the same chromosome , directly upstream of NDC80-3V5 ( Figure 4D , middle panel ) . This result demonstrates that NDC80luti-mediated repression occurs in cis , since NDC80luti cannot reduce Ndc80 protein expression from a copy of NDC80 on another chromosome ( Figure 4D , right panel ) . In the accompanying manuscript , Chia et al . revealed that this cis-acting mechanism is a result of alterations to the chromatin landscape across the NDC80ORF promoter caused by NDC80luti transcription ( Chia et al . , 2017 ) . Since NDC80luti is necessary to repress NDC80ORF during meiosis , we next investigated whether the NDC80luti leader is sufficient to regulate other genes in meiosis . We replaced the promoter and 5’ leader of NUF2 , the gene encoding the binding partner of Ndc80 , with the promoter and 5’ leader region of NDC80luti ( NDC80luti-NUF2 ) . In wild type cells , a single NUF2 mRNA species was expressed in meiotic prophase , a stage when NUF2 mRNA levels and Nuf2 protein levels were stable ( Figure 5A and B ) . In contrast , NDC80luti-NUF2 cells expressed a longer mRNA ( NUF2luti ) in meiotic prophase ( Figure 5A ) , and the abundance of NUF2ORF transcripts was reduced by ~60% compared to that in the pre-meiotic stage ( Figure 5—figure supplement 1 ) , a reduction level similar to that of the Nuf2 protein ( Figure 5B ) . This result demonstrates that the promoter and 5’ leader sequence of NDC80luti is sufficient to downregulate another protein in meiotic prophase . As NDC80luti expression is naturally restricted to meiosis , we tested whether the expression of NDC80luti was sufficient to downregulate NDC80ORF outside of meiosis . We artificially expressed NDC80luti during mitosis , a time when NDC80luti is naturally absent . We engineered strains in which the sole copy of the NDC80 gene had a modified upstream region , such that the endogenous promoter of NDC80luti was replaced by the inducible GAL1-10 promoter ( pGAL-NDC80luti ) . This alteration had minimal effect on cell growth ( Figure 8C , uninduced ) , suggesting that NDC80ORF transcript and Ndc80 protein expression is largely unaffected in the absence of induction . In wild type cells synchronously progressing through the mitotic cell cycle , a single mRNA isoform , NDC80ORF , was present at all stages ( Figure 5C , left panel ) . In contrast , the NDC80ORF transcript became undetectable in pGAL-NDC80luti cells one hour after NDC80luti induction ( Figure 5C , right panel ) . Four hours after induction , Ndc80 protein levels were reduced to 20% of the initial level , while in wild type cells it was increased to 116% ( Figure 5D ) . Based on these data , we conclude that NDC80luti expression is sufficient to repress NDC80ORF outside of meiosis . The reduction in NDC80ORF expression , in turn , causes reduced synthesis of Ndc80 protein , thus essentially turning off the NDC80 gene . Since the timely expression of NDC80luti and NDC80ORF is crucial to establish the temporal pattern of Ndc80 protein levels in meiosis , we next investigated which transcription factors directly control NDC80luti and NDC80ORF expression . In S . cerevisiae , meiotic gene expression is orchestrated by two master transcription factors: Ime1 and Ndt80 ( Chu and Herskowitz , 1998; Kassir et al . , 1988; Xu et al . , 1995 ) . Diploid MATa/MATα cells initiate meiosis by expressing IME1 in response to nutrient deprivation ( van Werven and Amon , 2011 ) . Interestingly , IME1 expression correlated with the time of NDC80luti expression , suggesting that Ime1 might regulate NDC80luti transcription . Indeed , deletion of IME1 abolished NDC80luti production and resulted in persistent levels of NDC80ORF transcript and Ndc80 protein ( Figure 6A and Figure 6—figure supplement 1 ) . Ime1 does not directly bind to DNA , but functions as a co-activator for Ume6 ( Washburn and Esposito , 2001 ) . In the absence of Ime1 , Ume6 represses early meiotic genes in mitosis by binding to a consensus site called the upstream repressive sequence ( URS1 ) in the promoters of these genes . Upon meiotic entry and subsequent interaction with Ime1 , the Ume6-Ime1 complex activates the transcription of these early meiotic genes ( Bowdish et al . , 1995; Park et al . , 1992 ) . Given the close relationship between Ime1 and Ume6 , we inspected the 5’ intergenic region of NDC80 and identified a consensus site for Ume6 583 bp upstream of the Ndc80 translation start site ( Figure 6B and Figure 4—figure supplement 1 ) , within the NDC80luti promoter . ChIP analysis revealed that Ume6 binding was enriched over the predicted URS1 site in mitosis and early meiosis ( Figure 6C and Figure 6—figure supplement 2 ) , whereas Ume6 binding was undetectable within the NDC80ORF promoter ( Figure 6C and Figure 6—figure supplement 2 ) . Deletion of the URS1 site ( ndc80-urs1∆ ) completely abolished Ume6 binding to the NDC80luti promoter ( Figure 6C ) , but did not affect another Ime1-Ume6 target gene IME2 ( Figure 6—figure supplement 3 ) . Consistent with the role of Ume6 as a transcriptional repressor in mitosis , deletion of the URS1 site resulted in leaky expression of NDC80luti during vegetative growth ( Figure 6D and E , p<0 . 0001 ) and reduced expression of NDC80ORF ( Figure 6E , p=0 . 0057 ) . Abolishing Ume6 binding eliminated strong induction of NDC80luti in meiosis ( Figure 6G and H , p<0 . 0001 ) , causing moderately increased levels of NDC80ORF transcript by northern blot and Ndc80 protein in meiotic prophase ( Figure 6I ) . We did not detect significant increase in NDC80ORF in the urs1∆ cells by smFISH ( Figure 6H ) , likely due to technical reasons ( See Materials and Methods ) . We conclude that similar to early meiotic genes , Ime1 and Ume6 directly regulate the transcription of NDC80luti . The second key meiotic transcription factor , Ndt80 , is required for meiotic chromosome segregation and spore formation ( Chu and Herskowitz , 1998; Xu et al . , 1995 ) . Expression of NDT80 occurs shortly before the reappearance of NDC80ORF transcript . Within the budding yeast lineage , an Ndt80 consensus site , called the mid-sporulation element ( MSE ) , was identified at 184 bp upstream of the Ndc80 translation start site ( Figure 6B and Figure 4—figure supplement 1 ) , within the NDC80ORF promoter . One hour after Ndt80 expression was induced in the pGAL-NDT80 GAL4-ER system , Ndt80 binding was enriched over the predicted MSE by ChIP analysis; moreover , mutations in the MSE ( ndc80-mse ) led to a complete loss of Ndt80 enrichment ( Figure 6F ) , but did not affect another Ndt80 target gene MAM1 ( Figure 6—figure supplement 4 ) . Furthermore , the defect in Ndt80 binding to the NDC80ORF promoter reduced both NDC80ORF transcript and Ndc80 protein levels during the meiotic divisions ( Figure 6J ) . These results demonstrate that Ndt80 directly induces NDC80ORF expression after meiotic prophase , and this timely induction of NDC80ORF elevates the levels of Ndc80 protein prior to the meiotic divisions . Since Ndc80 appears to be the limiting subunit of the kinetochore , we posited that the regulated expression of NDC80luti and NDC80ORF serves to inactivate and reactivate kinetochores , respectively , through modulating Ndc80 protein levels . In budding yeast , kinetochores are inactive in meiotic prophase ( Miller et al . , 2012 , and Figure 1D ) , but they can be activated upon Ndc80 overexpression ( Miller et al . , 2012 , and Figure 1D ) . We asked whether functional kinetochores could also be generated in meiotic prophase if cells failed to express NDC80luti ( ∆NDC80luti ) or expressed a version of NDC80luti that could translate Ndc80 protein ( ∆9AUG ) . Both conditions caused an increase in Ndc80 levels in meiotic prophase ( Figures 3 and 4A ) . Using the same assay described in Figure 1D , we observed that over 50% of the ∆NDC80luti or ∆9AUG cells displayed abnormal chromosome segregation in meiosis I ( Figure 7A ) , suggesting premature kinetochore activity in meiotic prophase . The extent of this phenotype was indistinguishable from that when Ndc80 was overexpressed in meiotic prophase ( pCUP-NDC80 ) ( Figure 7A ) . Therefore , repression of NDC80ORF by NDC80luti transcription is crucial to inhibit untimely kinetochore function during meiotic prophase . Functional kinetochores must be present after meiotic prophase to faithfully execute chromosome segregation during the two meiotic divisions . Since Ndc80 protein levels become nearly undetectable during prophase ( Figure 1C ) , Ndc80 must be resynthesized to restore the ability of kinetochores to interact with microtubules upon exit from prophase . This resynthesis relies on the transcription factor Ndt80 to induce transcription of NDC80ORF ( Figure 6F and J ) . To test the significance of Ndt80-dependent induction of NDC80ORF in meiosis , we monitored the segregation pattern of chromosome V in cells with a mutated Ndt80 binding site in the NDC80ORF promoter ( ndc80-mse ) . Only 1% of wild type cells missegregated chromosome V , whereas 98% of the ndc80-mse cells failed to properly segregate this chromosome ( Figure 7B ) , suggesting that kinetochores are not functional in ndc80-mse cells . In support of this conclusion , in ndc80-mse cells , elongated bipolar spindles ( over 2 μm ) appeared earlier and persisted longer than in wild type cells ( Figure 7C ) , a phenomenon consistent with defective microtubule-kinetochore attachments ( Wigge et al . , 1998; Wigge and Kilmartin , 2001 ) . Additionally , the abundance of short meiosis II spindles ( less than 2 µm ) was reduced in the ndc80-mse cells ( Figure 7D ) , and at the end of meiosis , more than four nuclei were often observed ( representative images shown in Figure 7B ) . The ndc80-mse mutation also severely affected the sporulation efficiency ( Figure 7—figure supplement 1 ) . All of these results demonstrate that Ndt80-dependent induction of NDC80ORF is essential for re-establishing kinetochore function to mediate meiotic chromosome segregation . Unlike NDC80ORF transcript , NDC80luti is absent in vegetative growth due to repression by Ume6 ( Figure 6D and E ) . We hypothesized that NDC80luti is repressed during the mitotic cell cycle because its expression could inactivate kinetochore function ( Figure 5C and D ) . Indeed , when the Ume6 repressor-binding site within the NDC80luti promoter was deleted ( urs1∆ ) , these cells grew similar to wild type cells at 30°C , but they had a severe growth defect at 37°C due to reduced Ndc80 levels ( Figure 8A and B ) . Thus , the repression of NDC80luti by Ume6 is critical for the fitness of mitotically dividing cells . When NDC80luti was strongly induced in vegetative growth using the inducible GAL1-10 promoter , these cells had a severe growth defect ( Figure 8C ) . This defect was rescued by a second copy of NDC80 at an ectopic locus , consistent with the notion that NDC80luti-mediated repression of NDC80ORF occurs in cis ( Figure 8C and Figure 4D ) . Cell death was also rescued by silencing the pGAL-induced NDC80luti expression using CRISPRi ( Qi et al . , 2013 ) ( Figure 8D ) , presumably due to the activation of the NDC80ORF promoter in the absence of NDC80luti transcription . Induction of the uORF-free NDC80luti ( ∆9AUG ) caused no appreciable growth defect ( Figure 8C ) , consistent with the observation that the ∆9AUG cells could express Ndc80 protein ( Figure 3 ) . The inducible nature of the GAL1-10 promoter allowed us to directly test whether the growth defect associated with the mitotic NDC80luti expression arose from defects in kinetochore function . We performed fluorescence microscopy to track spindle length ( Spc42-mCherry ) and chromosome segregation ( CENV-GFP dots ) . Cells expressing NDC80luti displayed a range of kinetochore-microtubule attachment defects ( Figure 8E , bottom panel ) . In cells with separated spindle pole bodies , ~30% of the cells expressing NDC80luti had metaphase spindles ( ≤2 μm ) improperly localized to either the bud or the bud neck , whereas only 3% of the wild type cells displayed this phenotype ( Figure 8F ) . Furthermore , in cells expressing NDC80luti , an abnormal distribution of spindle length was observed , characteristic of a metaphase arrest ( Figure 8—figure supplement 1 ) . Spindle elongation was also observed prior to chromosome capture , suggesting improper kinetochore function ( Figure 8G ) . Collectively , these analyses revealed that the strict temporal regulation of NDC80luti and NDC80ORF transcription in both mitosis and meiosis is essential to ensure the proper timing of kinetochore function and high fidelity chromosome segregation .
In meiosis , kinetochore function is transiently inactivated to facilitate accurate chromosome segregation ( Miller et al . , 2013 ) . This transient inactivation is achieved by the removal of the outer kinetochore from chromosomes and has been described in organisms ranging from yeast to mice ( Asakawa et al . , 2005; Kim et al . , 2013; Meyer et al . , 2015; Miller et al . , 2012; Sun et al . , 2011 ) . In budding yeast , we found that outer kinetochore removal is mediated by limiting the abundance of a single subunit , Ndc80 . Ndc80 is the only member of its complex whose protein abundance is essentially absent in meiotic prophase ( Meyer et al . , 2015 and Figure 1C ) . Furthermore , prophase overexpression of NDC80 , but none of the other Ndc80 complex subunits , promotes premature spindle attachments and causes meiotic chromosome segregation errors ( Figure 1D ) . Thus , in the case of the meiotic kinetochore , the cell regulates the activity of a multi-protein complex by limiting the availability of a single subunit . The control of protein complex activity through the limitation of a key subunit is a more general principle . A genome-wide study that analyzed the composition of protein complexes during the cell cycle revealed that in budding yeast , most protein complexes have both constitutively and periodically expressed subunits ( de Lichtenberg et al . , 2005 ) . It is proposed that due to the periodically expressed subunits , these protein complexes assemble "just-in-time" to restrict their function to specific cell cycle stages ( de Lichtenberg et al . , 2005 ) . The luti-mRNA-dependent regulatory circuit described here may more broadly address how regulated subunits are provided "just-in-time" and , importantly , at no other time . A key aspect of the work presented here is the surprising finding that an mRNA can serve a purely regulatory function . Indeed , NDC80luti is a bona fide mRNA . It is poly-adenylated , is engaged by the ribosome and , most importantly , when the uORF start codons are ablated , Ndc80 protein is translated from this extended mRNA isoform ( Brar et al . , 2012 and Figure 3 ) . Moreover , NDC80luti is likely a RNA Polymerase II transcript because its promoter is occupied by the pre-initiation complex member Sua7 ( TFIIB ) and because Pol II-associated chromatin marks are detected downstream of the NDC80luti promoter when this transcript is made ( Chia et al . , accompanying manuscript ) . NDC80luti cannot be decoded by the ribosome due to the presence of AUG-uORFs contained in its extended 5’leader . By competitively engaging the ribosome , these uORFs prevent translation of Ndc80 protein . The polypeptides that the uORFs encode are unlikely to play a role in the repression of kinetochore function as the uORFs can be minimized to 2-codon units while maintaining NDC80luti-based repression ( Figure 3 ) . Interestingly , upstream AUG codons are also present in the putative NDC80luti mRNAs predicted from the other fungal species . Three regions were enriched for the presence of such AUGs ( Figure 9—figure supplements 1 and 2 ) , but the sequences and the length of these putative uORFs did not seem to be conserved ( Supplementary file 1G ) . This observation is consistent with the idea that the act of uORF translation , rather than the identity of the uORF peptides , serves as a conserved feature in evolution . The repressive nature of the uORFs contained in NDC80luti mirrors those found in the uORF-containing prototype transcript , GCN4 ( Mueller and Hinnebusch , 1986 ) . However , in the case of GCN4 , changes in nutrient availability can relieve the uORF-mediated translational repression , whereas for NDC80luti , the uORF-mediated repression appears to be constitutive . In both cases , GCN4 and NDC80 can exist in on and off states . For GCN4 , this switch is manifested in the two translational states of the same mRNA molecule . For NDC80 , the switch is manifested instead by two distinct transcripts , one , which results in protein synthesis and one , which represses protein synthesis . It is important to note that for other potential luti-mRNAs , the precise mechanism of translational repression may not be conserved and could instead involve other means such as RNA hairpins or binding sites for translational repressors . Why do meiotic cells express an mRNA that does not encode any functional polypeptides ? We propose that the biological purpose of NDC80luti is to shut down Ndc80 protein synthesis by repressing NDC80ORF in cis , thereby inactivating kinetochore function during meiotic prophase . Multiple lines of evidence support this model . First , disruption of NDC80luti expression in meiosis results in elevated levels of NDC80ORF and Ndc80 protein in meiotic prophase , leading to premature kinetochore activation ( this study and Chia et al . , accompanying manuscript ) . Second , induction of NDC80luti transcription in cis is sufficient to repress NDC80ORF and inactivate kinetochore function in mitotic cells ( this study ) . Third , transcription of NDC80luti introduces repressive chromatin marks at the NDC80ORF promoter that are necessary for the downregulation of NDC80ORF and Ndc80 protein ( Chia et al . , accompanying manuscript ) . Altogether , these findings strongly suggest that the primary function of the NDC80luti mRNA is to turn off the NDC80 gene . It is important to note that our study only addresses the mechanism of how Ndc80 protein synthesis is repressed in meiotic prophase . Indeed , efficient and timely reduction of Ndc80 protein levels may require regulated proteolytic mechanisms not yet elucidated . Further studies are necessary to determine if proteolysis plays a role in the rapid removal of the outer kinetochore in meiotic prophase and if so , by what means this proteolysis is achieved . Why do budding yeast cells use this seemingly complex mechanism , which relies on the transcription of an undecoded mRNA isoform , to repress a kinetochore gene during meiosis ? We would argue from an evolutionary point of view that this solution could be both economical and highly flexible . First , the meiotic cell is co-opting two existing transcription factors , Ime1 and Ndt80 , for roles in activating and repressing gene expression , obviating the need to evolve novel trans-acting factors . This mechanism also ensures temporal coordination of gene activation and inactivation using the same transcription factor . In the case of NDC80 , the luti-mRNA rides the Ime1 wave of gene expression to shutoff kinetochore function while the protein-coding mRNA rides the subsequent Ndt80 wave to reactivate the kinetochore for the division phases . While transcription factors have previously been implicated in the repression of downstream promoters ( Bird et al . , 2006; Martens et al . , 2004; Shearwin et al . , 2005; van Werven et al . , 2012 ) , our study is the first clear demonstration that it is the choice of promoter and the identity of the resulting mRNA isoform that governs whether a gene is turned on or turned off by a given transcription factor . This mode of gene repression relies on two sets of cis-regulatory sequences , which are evolutionarily flexible ( Carroll , 2008; Stern and Orgogozo , 2008; Wittkopp and Kalay , 2011 ) . The first cis-acting sequence is the distal transcription factor-binding site , which induces transcription of NDC80luti , and , in concert with co-transcriptional chromatin modifications , silences the downstream canonical promoter activity . The second cis-acting sequence is the AUG-uORFs within the extended 5’ leader of the luti-mRNA , which prevents downstream ORF translation . Inherent to a mechanism that is so heavily reliant on cis-regulatory elements is the notion that minor changes in the DNA sequence can impact gene expression at a multitude of levels , thus tuning gene output . This tuning can be manifested at the level of nucleosome spacing , strength of transcription factor binding and translational regulation . Therefore , the cell has a vast evolutionary space , which can be explored through small changes in DNA sequence . The defining sequence features of the NDC80 luti-mRNA are a 5’-extended mRNA leader coupled with repressive uORFs contained in this extended leader . Analysis of the mRNA-seq and ribosome profiling datasets of meiotic yeast revealed hundreds of transcripts with potential luti-like signatures ( Brar et al . , 2012 ) . In support of this idea , two other genes , ORC1 and BOI1 , have been shown to express meiosis-specific transcript isoforms with uORF-containing leader extensions ( Xie et al . , 2016 and Liu et al . , 2015 ) . Rather than dissecting each candidate luti-mRNA on a case by case basis , future studies that integrate additional genome-wide datasets to measure stage-specific transcription factor binding sites , transcription-coupled chromatin modification states , mRNA translation status with isoform specificity and protein abundance would result in a high-confidence map of luti-mRNAs and aid in the dissection of their cellular functions . Beyond budding yeast meiosis , can the regulatory circuit described in our study be present in other developmental programs and in other organisms ? We would argue so , because various organisms also possess the three principles of this module , namely , alternative promoter usage , transcription-coupled repression , and uORF-mediated translational repression . Alternative promoter usage is widespread in development and among different cell types . For example , in the fruit fly , more than 40% of developmentally expressed genes have at least two promoters with distinct regulatory programs ( Batut et al . , 2013 ) . Half of human genes have more than one promoter , resulting in the expression of mRNA isoforms with 5’ heterogeneity ( Kimura et al . , 2006 ) . Furthermore , transcription-based interference mechanisms , as well as transcription-coupled histone modifications , have been described in a variety of organisms ( Corbin and Maniatis , 1989; Eissenberg and Shilatifard , 2010; Shearwin et al . , 2005; Wagner and Carpenter , 2012 ) . Finally , recent studies have shown that uORF translation is much more widespread than traditionally believed and acts in a regulatory manner ( Calvo et al . , 2009; Chew et al . , 2016; Johnstone et al . , 2016 ) . Therefore , we envision that the regulatory circuit described here can be used as a roadmap in future studies to uncover transcription-coupled gene repression during cell fate transitions across multiple species . A key implication of this model of gene regulation is a blurring of the line between "coding" and "non-coding" RNAs . Seminal work has uncovered multiple classes of non-coding RNAs that play regulatory functions in the cell , such as long non-coding RNAs , microRNAs , small interfering RNAs , and piwiRNAs ( Ambros , 2001; Batista and Chang , 2013; Cech and Steitz , 2014; Guttman et al . , 2009 ) . Our study demonstrates that mRNAs , which are deemed protein coding units , can themselves be direct regulators of gene expression by at least two simultaneous means: they can induce transcription-coupled silencing of a downstream promoter , and features in their 5’ leaders , such as the presence of uORFs or secondary structures , could directly impact translation efficiency in a positive or negative manner ( Arribere and Gilbert , 2013; Brar et al . , 2012; Rojas-Duran and Gilbert , 2012 ) . Notably , multiple studies have reported poor correlation between mRNA and protein abundance ( Maier et al . , 2009 ) . For those mRNAs that anti-correlate with their protein levels , this apparent contradiction might be due to a luti-mRNA being misattributed as a canonical protein-coding transcript . Our study could dramatically transform the way we understand the function of alternate mRNA isoforms and aid in the proper biological interpretation of genome-wide transcription studies .
All the strains used in this study are described in Supplementary file 1A and are derivatives of SK1 . The pGAL-NDT80 GAL4-ER and the pCUP-IME1 pCUP-IME4 synchronization systems have been described previously ( Benjamin et al . , 2003; Berchowitz et al . , 2013 ) . The centromeric TetR/TetO GFP dot assay is described in ( Michaelis et al . , 1997 ) . The ndc80-1 temperature-sensitive mutant was first described in ( Wigge et al . , 1998 ) , the Zip1::GFP ( 700 ) described in ( Scherthan et al . , 2007 ) , and pCUP-NDC80 pCUP-CLB3 described in ( Miller et al . , 2012 ) . NDC80-3V5 , NUF2-3V5 , SPC24-3V5 , SPC25-3V5 , pCUP-NUF2 , pCUP-SPC24 , pCUP-SPC25 , pGAL-NDC80luti , pGAL-∆9AUG , ndc80∆ , nuf2∆ , ( ∆−600 to −300 ) -NDC80 , and ( ∆−600 to −400 ) -NDC80 were generated at the endogenous gene loci using PCR-based methods ( Longtine et al . , 1998 ) . The V5 plasmid is kind gift from Vincent Guacci . Primer sequences used for strain construction can be found in Supplementary file 1B . Single integration plasmids carrying either NDC80 or NUF2 were constructed by Gibson Assembly ( Gibson et al . , 2009 ) , and were digested with PmeI to integrate at the LEU2 locus . For NDC80 , the LEU2 integration plasmid included the SK1 genomic sequence spanning from 1000 bp upstream to 357 bp downstream of the NDC80 coding region; and for NUF2 , spanning from 1000 bp upstream to 473 bp downstream of the NUF2 coding region . Both constructs included a C-terminal fusion of the 3V5 epitope to NDC80 and NUF2 , and both completely rescued the full deletion of NDC80 or NUF2 , respectively . Deletions ( ndc80-urs1∆ and ( ∆−600 to −479 ) -NDC80 ) and point mutations ( ndc80-mse ) were generated from the NDC80 LEU2 single integration plasmid using the site-directed mutagenesis kit ( Q5 Site-Directed Mutagenesis Kit , NEB , Ipswitch , MA ) . The entire URS1 site and the "A" right upstream of the site were deleted in the ndc80-urs1Δ strain . The ndc80-mse construct has two C to A mutations , marked using black diamonds in Figure 6B . The ∆6AUG , ∆9AUG , mini uORF , NDC80luti-Ter , and NDC80luti-NUF2 constructs were generated by Gibson assembly ( Gibson et al . , 2009 ) using the NDC80 and NUF2 LEU2 integration plasmids , as well as gBlocks gene fragments ( IDT , Redwood City , CA ) for the ∆9AUG and mini uORF constructs . SNR52 promoter-controlled guide RNAs targeting NDC80luti ( A-D ) were cloned into a 2-micron plasmid carrying a LEU2 selectable marker ( pRS425 backbone ) . See Supplementary file 1C for the full list of the integration and 2-micron plasmids . Synchronously sporulating cell cultures were prepared as in ( Berchowitz et al . , 2013 ) . In short , the endogenous promoters of IME1 and IME4 were replaced with the inducible CUP1 promoter . Diploid cells were grown in YPD ( 1% yeast extract , 2% peptone , 2% glucose , and supplemented with 22 . 4 mg/L uracil and 80 mg/L tryptophan ) for 20–24 hr at room temperature . For optimal aeration , the total volume of the flask exceeded the volume of the medium by 10 fold . Subsequently , cells were transferred to BYTA ( 1% yeast extract , 2% bacto tryptone , 1% potassium acetate , 50 mM potassium phthalate ) and grown for another 16–18 hr at 30°C . The cells were then pelleted , washed with sterile milliQ water , and resuspended at 1 . 85 OD600 in sporulation ( SPO ) media ( 0 . 5% ( w/v ) potassium acetate [pH 7] , 0 . 02% ( w/v ) raffinose ) at 30°C . To initiate synchronous sporulation , expression of IME1 and IME4 was induced 2 hr after cells were transferred to SPO by adding copper ( II ) sulphate to a final concentration of 50 µM . The pGAL-NDT80 GAL4-ER system was used to generate populations of cells synchronously undergoing the meiotic divisions ( Carlile and Amon , 2008 ) . Cells were prepared for meiosis as in the pCUP-IME1 pCUP-IME4 protocol , and resuspended at 1 . 85 OD600 in SPO . The flasks were placed at 30°C for 5–8 hr to block cells in meiotic prophase ( See figure legend for the specific arrest duration for each experiment ) . To release cells from pachytene , NDT80 expression was induced with 1 μM β-estradiol . Subsequently , cells progressed through meiosis synchronously . MATa cells were first grown to an OD600 of 1–2 at 30°C in YPD , diluted back to OD600 0 . 005 in YEP-RG ( 2% raffinose and 2% galactose in YEP supplemented with 22 . 4 mg/L uracil and 80 mg/L tryptophan ) , and then grown at room temperature for 15–17 hr . Exponentially growing cells were diluted again to an OD600 of 0 . 19 in YEP-RG , and arrested in G1 with 4 . 15 μg/mL alpha-factor , and 1 . 5 hr later , an additional 2 . 05 μg/mL of alpha-factor was added to the cells . After 2 hr in alpha-factor , 1 μM β-estradiol was added to cultures to induce pGAL expression . One hour after the β-estradiol addition , cells were filtered , rinsed with YEP ( 10 times volume of the culture volume ) to remove the alpha-factor , and placed into a receiving flask containing YEP-RG with 1 μM β-estradiol . Time points were taken before β-estradiol induction , before release , and every 15 min after release , for 3 hr . Clustal analysis ( Goujon et al . , 2010; Sievers et al . , 2011 ) was performed using the genomic sequences of S . bayanus , S . kudriavzevii , S . mikatae , S . cerevisiae and S . paradoxus from Saccharomyces sensu stricto genus ( Scannell et al . , 2011 ) , and imported into the Webpage of the Clustal Omega Multiple Sequence Alignment tool < http://www . ebi . ac . uk/Tools/msa/clustalo/> . The Ume6-3V5 chromatin immunoprecipitation experiments were performed as described previously with the following modifications ( van Werven et al . , 2012 ) . Cells were fixed with formaldehyde ( 1% v/v ) for 15 min . Frozen cell pellets were disrupted 4 times ( 5 min each ) using a Beadbeater ( Mini-Beadbeater-96 , Biospec Products , Bartlesville , OK ) . Chromatin was sheared 5 × 30 s ON/30 s OFF with a Bioruptor Pico ( Diagenode , Denville , NJ ) to a fragment size of ~200 bp . Chromatin extracts were incubated with 20 μL of anti-V5 agarose beads ( A7345 , Sigma , St . Louis , MO ) at 4°C . The Ndt80-3V5 chromatin immunoprecipitation experiments were performed as described previously with the same modifications as used for Ume6-3V5 except for the sonication conditions ( Strahl-Bolsinger et al . , 1997 ) . Chromatin was sheared 5 × 10 s ON/30 s OFF with a Bioruptor Pico ( Diagenode ) to a fragment size of ~500 bp . Reverse crosslinked input DNA and immunoprecipitated DNA fragments were amplified with Absolute SYBR green ( AB4163/A , Thermo Fisher , Waltham , MA ) and quantified with a 7500 Fast Real-Time PCR machine ( Thermo Fisher ) using the primer pairs directed against the upstream region and the coding region of NDC80 , the MAM1 promoter , and the IME2 promoter . We also measured the signals from the NUF2 promoter and HMR , regions that do not display significant binding for either of the transcription factors . The oligonucleotide sequences used are listed in Supplementary file 1D . Cells were fixed with 3 . 7% formaldehyde at room temperature for 15 min , washed once with potassium phosphate/sorbitol buffer ( 100 mM potassium phosphate [pH 7 . 5] , 1 . 2 M sorbitol ) , and then permeabilized with 1% Triton X-100 with 0 . 05 μg/mL DAPI in potassium phosphate/sorbitol buffer . Cells were imaged using a DeltaVision microscope with a 100x/1 . 40 oil-immersion objective ( DeltaVision , GE Healthcare , Sunnyvale , CA ) and filters: DAPI ( EX390/18 , EM435/48 ) , GFP/FITC ( EX475/28 , EM525/48 ) , and mCherry ( EX575/25 , EM625/45 ) . Images were acquired using the softWoRx software ( softWoRx , GE Healthcare ) . For Figure 8E–G and Figure 8—figure supplement 1 , diploid cells were first grown to an OD600 of 1–2 at 30°C in YPD . They were then diluted to an OD600 of 0 . 002 in YEP-RG and grown at 30°C for 16 hr . Exponentially growing cells were diluted back to an OD600 of 0 . 2 in YEP-RG and induced to express NDC80luti with 1 μM β-estradiol . Samples were taken before induction and 6 hr after induction . Images were acquired as described in the fluorescence microscopy method section , and analysed using the FIJI image processing software ( RRID:SCR_002285 , Schindelin et al . , 2012 ) . First , maximum-intensity projection was performed . Second , projected spindle length ( defined as the distance between Spc42-mCherry foci ) was measured using the "measure" plugin . The distribution of the projected spindle length was graphed as violin plots using ( BoxPLotR RRID:SCR_015629 , Spitzer et al . , 2014 ) . Third , in cells with separated spindle poles , the status of the Spc42-mCherry association with CENV-GFP dots was categorized as 1 ) each Spc42-mCherry focus is associated with a CENV-GFP dot , 2 ) only one Spc42-mCherry focus is associated with CENV-GFP dots ( either one or both of the GFP dots ) , or 3 ) neither Spc42-mCherry focus is associated with a CENV-GFP dot . After categorizing the localization of the CENV-GFP dots , the projected spindle length was measured for spindles in category 2 and 3 , and the spindle length distributions were graphed as violin plots using ( BoxPLotR RRID:SCR_015629 , Spitzer et al . , 2014 ) . Finally , in cells with separated spindle poles , the location of the spindle was recorded as 1 ) in the mother , 2 ) across the bud neck , or 3 ) in the bud . The percentage of spindles that were both less than 2 . 0 μm and abnormally localized ( across the bud neck or in the bud ) was calculated . For each analysis , 100 cells were counted . Tubulin indirect immunofluorescence was performed as described ( Kilmartin and Adams , 1984 ) using a rat anti-tubulin antibody ( MCA78G , Bio-rad Antibodies , Kidlington , UK ) at a dilution of 1:200 and a pre-absorbed anti-rat FITC antibody ( 712-095-153 , Jackson ImmunoResearch Laboratories , Inc . West Grove , PA ) at a dilution of 1:200 . The meiotic stage of a cell was determined based on its spindle and DAPI morphologies . Metaphase I spindles were defined as a short bipolar spindle spanning a single DAPI mass; an anaphase I spindle was defined as a single elongated spindle spanning two DAPI masses; a pair of metaphase II spindles were defined as two short bipolar spindles each spanning a distinct DAPI mass within a single cell; and finally , a pair of anaphase II spindles was defined as two elongated spindles with 4 DAPI masses within a single cell . To image spindle samples for characterization of spindle length , z stacks ( 8–10 slices ) were acquired with a step size of 0 . 5 μm using the DeltaVision microscope ( GE Healthcare ) described in the fluorescence microscopy section . To measure the projected spindle length , maximum-intensity projection of these images was generated by FIJI ( RRID:SCR_002285 , Schindelin et al . , 2012 ) . Next , the projected spindle length ( defined as the spindle pole-to-pole distance ) was measured using the "measure" plugin ( Schindelin et al . , 2012 ) , and cells were staged to be in either meiosis I or meiosis II depending on the number of bipolar spindles . For cells undergoing meiosis II , both spindles were quantified , but only the longer of the two was reported . For each time point , the percentage of cells in each category was quantified: 1 ) meiosis I spindles that were less than 2 μm , 2 ) meiosis I spindles that were over 2 μm , 3 ) meiosis II spindles that were less than 2 μm , and 4 ) meiosis II spindles that were over 2 μm . Over 100 cells per time point were quantified . A previously described northern blot protocol was modified as below ( Koster et al . , 2014 ) . RNA was extracted with acid phenol:chloroform:isoamyl alcohol ( 125:24:1; pH 4 . 7 ) and then isopropanol precipitated . RNA samples ( 8–10 µg ) were denatured in a glyoxal/DMSO mix ( 1 M deionized glyoxal , 50% v/v DMSO , 10 mM sodium phosphate buffer pH 6 . 5–6 . 8 ) at 70°C for 10 min and then separated on a 1 . 1% agarose gel for 3 hr at 80 V . RNAs were transferred onto nylon membranes overnight by capillary transfer . The membranes were blocked for at least 3 hr at 42°C in ULTRAhyb Ultrasensitive Hybridization Buffer ( Thermo Fisher ) before hybridization . Radioactive probes were synthesized using a Prime-It II Random Primer Labelling Kit ( Agilent , Santa Clara , CA ) . The oligonucleotide sequences of the primers used to amplify the NDC80 , NUF2 , SCR1 , and CIT1 DNA template are displayed in Supplementary file 1D . Quantification was performed with FIJI ( RRID:SCR_002285 , Schindelin et al . , 2012 ) . For all the images , the LUT ( lookup table ) was inverted . Then , a rectangular box was drawn around a band of interest . The mean signal intensity ( gray-scale ) within the box area was calculated using the "measure" plugin . For background subtraction , the same box was moved directly above and below the band , the signal intensity of these two regions was measured , and the average background intensity ( top and bottom ) was calculated . After subtracting the average background intensity of a given lane from the signal intensity of the band in that lane , this corrected value for each time point was then normalized to the initial time point . The same-sized box was used for all the time points in one experiment . Single-molecule RNA FISH was performed as described ( Raj et al . , 2008 ) with modifications . All the probes ( Supplementary file 1E for probe sequences ) were designed , synthesized , and labelled by Stellaris ( Biosearch Technologies , Novato , CA ) . The unique region of NDC80luti was targeted by twenty 20-mer oligonucleotide probes coupled to CAL Fluor Red 590 . Thirty 20-mer probes , coupled to Quasar 670 dye , were targeted to the coding region of NDC80 . To measure our detection quality , 54 alternating probes ( odd and even probes , 27 probes in each set ) were designed to target the common region of NDC80luti and NDC80ORF , and coupled with Quasar 670 dye and CAL Fluor Red 590 dye , respectively . For meiosis experiments , cells were sporulated as described above . To fix cells , 160 μL of 37% formaldehyde was added into 1840 μL of meiotic cultures and incubated at room temperature for 20 min with gentle agitation . The fixed samples were moved to 4°C to continue fixing overnight . For vegetative samples , cells were grown in YPD to an OD600 of 0 . 4–0 . 6 , fixed in formaldehyde at room temperature for 20 min , and then prepared for digestion as below . Cells were washed three times in 1 . 5 mL cold Buffer B ( 0 . 1 M potassium phosphate [pH 7 . 5] , 1 . 2 M sorbitol ) and resuspended in 425 μl digestion buffer ( 425 μL Buffer B mixed with 40 μL 200 mM Vanadyl ribonucleoside complex ( VRC ) ( NEB ) with 50 μg of zymolyse ( zymolase 100T , MP Biomedicals , Santa Ana , CA ) . Cells were digested at 30°C until approximately 70% of cells were digested . This took about 15–20 min for early meiotic and vegetative samples and 30–35 min for pachytene and post meiotic prophase samples . Digested cells were gently washed with 1 mL of cold Buffer B and resuspended in 1 mL of 70% EtOH for 3 . 5–5 hr to allow permeabilization . To prepare for hybridization , cells were first incubated in 1 mL of 10% formamide wash buffer ( 10% formamide , 2X SSC ) at room temperature for at least 15 min . For hybridization , each probe set ( to a final concentration of 500 nM ) and 20 mM VRC were added to hybridization buffer ( 1% Dextran sulfate ( EMD Millipore , Billerica , MA ) , 1 mg/mL E . coli tRNA ( Sigma ) , 2 mM VRC , 0 . 2 mg/mL BSA , 1X SSC , 10% formamide ( Thermo Fisher ) in nuclease-free water ) . Hybridization was performed overnight at 30°C with gentle agitation . Samples were then incubated in the dark for 30 min at 30°C in 1 mL of 10% formamide wash buffer , the buffer was then washed away , cells were stained with DAPI , and resuspended in 50 μL of glucose-oxygen-scavenging buffer ( GLOX buffer ( 10 mM Tris [pH 8 . 0] , 2x SSC , 0 . 4% glucose ) ) solution without enzymes . Prior to imaging , 15 μL of GLOX solution with enzyme ( 1% v/v catalase , 1% v/v glucose oxidase ( Sigma ) , 2 mM Trolox ( Sigma ) ) was added to the sample . Images were acquired with the DeltaVision microscope ( GE Healthcare ) as described in the fluorescence microscopy section with two additional filters: TRITC ( EX542/27 , EM597/45 ) for CAL Fluor Red 590 and CY5 ( EX632/22 , EM679/34 ) for Quasar 670 . Series of z-stacks ( 15–25 slices ) were acquired with a step size of 0 . 2 μm . To quantify FISH spots , maximum-intensity projection of the z-stacks was first generated in ( RRID:SCR_002285 , Schindelin et al . , 2012 ) , different channels were split , and these processed images were analysed with custom software written in Matlab ( McSwiggen , 2017 ) ( Mathworks , Sunnyvale , CA ) . Cell boundaries were hand-drawn . The spot detection code first filtered the raw images using an eight pixel Gaussian kernel to remove background signal . Diffraction-limited spots corresponding to single mRNA were detected using an adaptation of the MTT spot-detection algorithm ( Sergé et al . , 2008 ) , using the following detection parameters: NA: 1 . 4; detection box: 5 pixels; error rate: 0 . 1; deflation loops: 0 . With these detection settings , many low-intensity fluctuations in background fluorescence were detected as spots . To identify bona fide mRNA molecules , we plotted the signal ( defined as the integrated value of the pixel intensities ) against the signal-to-noise ratio ( SNR; defined as the signal divided by the variance of the pixel values around the detected spot ) , identified a population of detections that were well separated from the background detections , and chose these signal and SNR values as thresholds . To confirm these threshold choices , we plotted the number of spots detected as a function of the threshold chosen , and found that these thresholds fell within a ‘plateau’ , as others have described ( Senecal et al . , 2014; Raj et al . , 2006 ) , where an increase in the choice of threshold has little effect on the total number of mRNA detected . Inspection of detected mRNAs , post-threshold , was in good agreement with spots that were manually counted . Once chosen , the same "signal" and "SNR" thresholds were applied to all the images within a replicate . In general , we found that thresholds between replicates varied only slightly ( For CF 590 probes , signal = 1100–1500 and SNR = 2 . 5–3; for Q670 probes , signal = 1000–2000 and SNR = 2–3 ) . After detection , spots between the CF 590 and Q 670 probe sets need to be paired to identify NDC80luti and NDC80ORF transcripts . Pairing was done using the knnsearch Matlab function to separately identify the closest CF 590 spot for each detected Q 670 spot , and vice versa . Two spots are only considered paired if they are mutual nearest neighbors . Using this as a criterion for pairing , greater than 95% of spot pairs occurred within 2 pixels of each other , which is well within the expected value given any chromatic and detection artifacts between the two color channels . By comparison , fewer than 10% of unpaired spots had nearest neighbor distance of less than four pixels , showing that the probability of misidentifying a spot pair is low . The number of cells with a given number of NDC80luti or NDC80ORF transcripts per cell was graphed as relative frequency histograms . The largest bin of each histogram was normalized to the same length across all the histograms . Per-cell statistics of paired spots ( NDC80luti mRNA ) , Q 670-only spots ( NDC80ORF ) , and CF 590-only spots ( false positives , early terminated transcripts , and degradation products ) were collected and pooled between biological replicates . First , to determine whether sufficient data had been collected for a given data set , bootstrap analysis of the data was performed . For 500 iterations , statistics from a single cell was randomly sampled from the data , and the mean and variance calculated . This process was repeated for two cells randomly selected from the data , without replacement; then for three cells randomly selected , etc . until one half of the total data set size was reached . A plot of the mean and standard deviation of paired and unpaired spots shows that the mean is stable and that the change in the variance plateaus at a number far below the number of cells assayed , suggesting that our sample size is sufficiently large ( Figure 2—figure supplement 5 ) . For each sample , over 95 cells were counted and three independent experiments were performed . Thus , for each data set , we could ensure that enough cells were measured to accurately account for the biological variation intrinsic to the data set . To compare across different strains and conditions , the two-tailed non-parametric Wilcoxon Rank Sum test was applied to the pooled data obtained from three independent experimental repeats . The p-value was determined using the ranksum function in Matlab ( Mathworks ) . Explanation about Figure 6H: Based on our smFISH statistical analysis , the NDC80ORF transcript level in the urs1∆ cells did not differ significantly from that of wild type cells , even though there was a clear difference in the northern blot analysis ( Figure 6I ) . We consider the possibility that our smFISH quantification method has a technical limitation when the NDC80luti isoform is highly expressed . Since we identified NDC80ORF based on the presence of the Q 670 signal ( both transcripts ) and the absence of CF 590 signal ( NDC80luti unique probes ) , a missed localization in the CF 590 channel would cause us to over-estimate the number of NDC80ORF . In our control experiments using alternating probes ( Figure 2—figure supplement 4 ) , we measured that ~ 6% of the Q 670 spots lack colocalizing signal from the CF 590 channel . In conditions where the NDC80luti isoform is expressed to the high level observed in wild type meiotic prophase , we expect to miss ~1 CF 590 spot per cell , which would then be interpreted as an extra NDC80ORF molecule . Since the total number of NDC80luti transcripts between wild type and the urs1∆ mutant was quite different during meiotic prophase ( the median of NDC80luti transcripts is 15 in wild type , that of urs1∆ is merely 5 , Figure 6H ) , the number of mRNA being mis-classified as NDC80ORF mRNA would also be higher in wild type cells , due to a missed signal from the CF 590 channel . Therefore , it is possible that we over-estimated the number of NDC80ORF mRNA in the wild type strain . Given these limitations , we propose that the difference in transcript levels between the wild type and urs1∆ mutant is too subtle to be detected by our smFISH analysis . Cells were grown on YPG ( 2% glycerol + YEP ) plates overnight , resuspended in milliQ H2O , and then diluted to an OD600 of 0 . 2 . 5-fold serial dilutions were performed , and cells were spotted onto YEP-RG plates with or without supplement of 1 μM β-estradiol . The cells were incubated at 30°C for 1–2 days . For experiments in which dCas9 was used to repress NDC80luti , cells were first grown on SC-G -leu ( 0 . 67% yeast nitrogen base , 2% glycerol , supplemented with adenine , lysine , tyrosine , phenylalanine , threonine , uracil , tryptophan , and histidine ) . Serial dilutions were performed as above and cells were spotted onto SC-RG -leu plates ( 0 . 67% yeast nitrogen base , 2% raffinose , 2% galactose , supplemented with adenine , lysine , tyrosine , phenylalanine , threonine , uracil , tryptophan , and histidine ) with or without 1 μM β-estradiol . Protein extracts were prepared using a trichloroacetic acid ( TCA ) extraction protocol . Briefly , ∼4 OD600 units of cells were treated with 5% trichloroacetic acid for at least 15 min at 4°C . Following an acetone wash , the cell pellet was subsequently dried . The cell pellet was lysed with glass beads in lysis buffer ( 50 mM Tris–HCl [pH 7 . 5] , 1 mM EDTA , 2 . 75 mM DTT , protease inhibitor cocktail ( cOmplete EDTA-free , Roche , Basel , Switzerland ) using a Mini-Beadbeater-96 ( Biospec Products ) . Next , 3x SDS sample buffer ( 187 . 5 mM Tris [pH 6 . 8] , 6% ß-mercaptoethanol , 30% glycerol , 9% SDS , 0 . 05% bromophenol blue ) was added and the cell lysate was boiled for 5 min . Proteins were separated by PAGE using 4–12% Bis-Tris Bolt gels ( Thermo Fisher ) and transferred onto nitrocellulose membranes ( 0 . 45 μm , Bio-rad , Hercules , CA ) using a semi-dry transfer apparatus ( Trans-Blot Turbo Transfer System , Bio-rad ) . The membranes were blocked for at least 30 min with Odyssey Blocking Buffer ( PBS ) ( LI-COR Biosciences , Lincoln , NE ) before incubation overnight at 4°C with a mouse anti-V5 antibody ( RRID:AB_2556564 , R960-25 , Thermo Fisher ) at a 1:2000 dilution . We monitored Hxk1 levels using a rabbit anti-hexokinase antibody ( RRID:AB_2629457 , H2035 , US Biological , Salem , MA ) at a 1:10 , 000 dilution , Pgk1 levels with a 1:10 , 000 diluted mouse anti-Pgk1 antibody ( RRID:AB_2532235 , SC7167 , Molecular Probes , Carlsbad , CA ) , and Kar2 levels with a 1:200 , 000 rabbit anti-Kar2 antibody ( provided by Mark Rose ) . Membranes were washed in PBST ( phosphate buffered saline with 0 . 01% tween-20 ) and incubated with an anti-mouse secondary antibody conjugated to IRDye 800CW at a 1:15 , 000 dilution ( RRID:AB_621847 , 926–32212 , LI-COR Biosciences ) and an anti-rabbit antibody conjugated to IRDye 680RD at a 1:15 , 000 dilution ( RRID:AB_10956166 , 926–68071 , LI-COR Biosciences ) to detect the V5 epitope and Hxk1 , respectively . Immunoblot images were generated and quantified using the Odyssey system ( LI-COR Biosciences ) . All code used for the analysis of smFISH images has been made available by the authors in the following code repository: https://gitlab . com/tjian-darzacq-lab/Chen_Tresenrider_et_al_2017 ( copy archived at https://github . com/elifesciences-publications/Chen_Tresenrider_et_al_2017 ) . | DNA stores the genetic information needed to make proteins and other molecules inside cells . To make a protein , cells use a particular section of DNA as a template to make molecules of messenger RNA ( or mRNA for short ) , which are then translated into the corresponding protein . Inside yeast , humans and other eukaryotic organisms , DNA is organized into structures called chromosomes . When these cells divide to make sex cells , such as egg cells and sperm , they undergo a process known as meiosis to make four daughter cells with only half as many chromosomes as the parent cell . During meiosis the parent cell’s chromosomes need to be separated twice in quick succession . Large assemblies of proteins known as kinetochores are essential for this process . At the beginning of meiosis the kinetochores are inactive , which prevents the chromosomes from being separated too soon . Later on , the kinetochores are activated to allow the chromosomes to be separated . In budding yeast cells , control of kinetochore activity is achieved by regulating the levels of a single protein within the kinetochore known as Ndc80 . It was not clear , however , how this regulation occurs . Chen , Tresenrider et al . show that two yeast mRNAs with opposing activities are the key to regulating when Ndc80 is produced during meiosis . These two mRNAs carry the same protein-coding message but one mRNA is longer and contains an extension that prevents it from being translated into protein . The sole role of the longer mRNA is to prevent the shorter mRNA from being produced . The longer mRNA is only made in early meiosis , which prevents new Ndc80 proteins from being made during this time and results in the kinetochores being inactivated . The chromosomes only separate when production of the shorter mRNA returns later in meiosis and higher levels of Ndc80 protein allow the kinetochores to reform . These findings demonstrate that even if an mRNA molecule does carry protein-coding information , it may not actually act as a messenger to make that protein , but instead , it can serve a regulatory role by blocking that protein’s production . The key components involved in regulating Ndc80 during meiosis are found in many organisms ranging from yeast to humans , suggesting that similar mechanisms could be used to control proteins involved in other processes in cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2017 | Kinetochore inactivation by expression of a repressive mRNA |
To date , microglia subsets in the healthy CNS have not been identified . Utilizing autofluorescence ( AF ) as a discriminating parameter , we identified two novel microglia subsets in both mice and non-human primates , termed autofluorescence-positive ( AF+ ) and negative ( AF− ) . While their proportion remained constant throughout most adult life , the AF signal linearly and specifically increased in AF+ microglia with age and correlated with a commensurate increase in size and complexity of lysosomal storage bodies , as detected by transmission electron microscopy and LAMP1 levels . Post-depletion repopulation kinetics revealed AF− cells as likely precursors of AF+ microglia . At the molecular level , the proteome of AF+ microglia showed overrepresentation of endolysosomal , autophagic , catabolic , and mTOR-related proteins . Mimicking the effect of advanced aging , genetic disruption of lysosomal function accelerated the accumulation of storage bodies in AF+ cells and led to impaired microglia physiology and cell death , suggestive of a mechanistic convergence between aging and lysosomal storage disorders .
Microglia are a unique population of tissue resident macrophages residing in the central nervous system ( CNS ) accounting for 10% to 15% of all cells within the CNS . While displaying some canonical macrophage activities such as the phagocytosis of debris and apoptotic bodies ( Chan et al . , 2001; Janda et al . , 2018 ) , microglia are also endowed with functions specific to the CNS microenvironment ( Clayton et al . , 2017; Li and Barres , 2018; Ransohoff , 2016; Ransohoff and Khoury , 2016 ) , such as synaptic remodeling ( Paolicelli et al . , 2011; Stephan et al . , 2012; Stevens et al . , 2007; Weinhard et al . , 2018 ) , neuronal support ( Parkhurst et al . , 2013; Ueno et al . , 2013 ) , and oligodendrogenesis ( Hagemeyer et al . , 2017; Wlodarczyk et al . , 2017 ) . However , despite this diversity of functions , no durable subsets have been identified in the healthy adult brain at steady-state . The disease-associated microglia ( DAM ) and the closely-related microglia subset expressing a neurodegenerative phenotype ( MGnD ) were reported to arise in response to the accumulation of β-amyloid plaques in Alzheimer’s Disease ( AD ) transgenic mouse models ( Kamphuis et al . , 2016; Keren-Shaul et al . , 2017; Krasemann et al . , 2017; Mrdjen et al . , 2018 ) , as well as in other models of neurodegeneration and aging ( Chiu et al . , 2013; Holtman et al . , 2015; Spiller et al . , 2018; Wlodarczyk et al . , 2014 ) . Mirroring the gene signature and function of DAM and MGnD subsets in disease conditions , the proliferative-region-associated microglia ( PAM ) subset was associated with the phagocytosis of newly formed and dying oligodendrocytes during normal post-natal development ( Felsky et al . , 2019; Hagemeyer et al . , 2017; Li et al . , 2019 ) . Dark microglia , a subset identified by their condensed , electron-dense cytoplasm visible by transmission electron microscopy , are frequently observed during pathologic states and believed to actively associate with neuronal synapses ( Bisht et al . , 2016 ) . Beyond conditions of neuronal or oligodendrocyte cell death , microglia heterogeneity was also observed within the adult steady-state CNS . In particular , regional variations in cellular density ( Lawson et al . , 1990 ) , but also in microglia gene expression profiles were reported , pointing to region-specific microglia functions and possibly subsets in healthy brain ( Ayata et al . , 2018; Grabert et al . , 2016 ) . While these emerging transcriptomic findings suggest that steady-state microglia subsets are likely present , they remain to be identified and characterized . By using cellular autofluorescence ( AF ) as a novel photophysical parameter to explore microglia heterogeneity in unperturbed conditions , we report that steady-state microglia exist in two discrete states at a regulated ratio throughout the entire lifespan of rodent and non-human primate species . Leveraging this physical property , we devised a novel probe-free method to isolate and characterize these subsets and established that AF+ and AF− microglia differed in their ultrastructural features , homeostatic dynamics , proteomic content and physiological properties .
Based on the observation that certain peripheral myeloid populations are highly autofluorescent , we initially included an empty fluorescence channel in our microglia flow cytometry analyses . Doing so revealed an unexpected bimodal distribution of autofluorescence ( AF ) intensity in CD45dimCD11B+ microglia isolated from 6-month-old naïve C57BL/6 mice ( Figure 1A and Figure 1—figure supplement 1A ) . This bimodal distribution identified two subsets of microglia: an AF-positive ( AF+ ) subset showing a strong AF signal , and an AF-negative ( AF− ) subset which displayed no or minimal levels of AF . The two subsets appeared at a highly reproducible ratio of 1:2 . 5 ( AF−:AF+ ) , with an average frequency of AF+ microglia of 71 ± 1 . 2% ( n = 13; CV = 1 . 6% ) ( Figure 1—figure supplement 1B ) . To further characterize the spectral properties of the AF signal , we included additional empty fluorescence channels . The AF signal showed maximal intensity in the 660–735 nm emission range upon excitation with a 488 nm ( Blue ) laser but was also detected across multiple combinations of laser excitation wavelengths and emission filter ranges ( Figure 1B ) . Extraction of single cell-level data from flow cytometry analyses revealed that the majority of microglia that were either positive or negative for AF in the Blue-710 nm AF channel were also positive or negative across most other AF channels tested , respectively ( Figure 1C ) . In a few channels such as Red-780 nm for instance , 10% to 20% of the cells that were positive in Blue-710 nm appeared low or negative , which likely resulted from a differential sensitivity of the channels used to detect AF as shown in Figure 1B by the differential brightness of the AF+ population in those two channels . Both AF microglia subsets were detected across different regions of the brain and while there were minor differences in the frequency of AF+ microglia between regions , AF+ cells displayed similar levels of AF signal in cerebellum , cortex and hippocampus ( Figure 1—figure supplement 1C , D ) . Both AF+ and AF− subsets of microglia were positive for microglia homeostatic markers , including CX3CR1 , P2RY12 and TMEM119 ( Figure 1D and Figure 1—figure supplement 1E ) . Finally , there were no gender differences observed in the frequency of AF+ microglia nor the intensity of AF ( Figure 1—figure supplement 1F , G ) . This observation was conserved across species as microglia isolated from the brains of 3- to 4-year-old Cynomolgus monkeys showed a very consistent bimodal pattern of AF in the Blue-710 nm channel with 80 ± 3 . 6% of microglia that were AF+ on average ( Figure 1E , F ) . Altogether , these analyses identified cellular AF as a novel photophysical property for discriminating two discrete subsets of microglia present at steady-state and conserved between rodents and non-human primates . To define the subcellular origin of the AF signal in microglia , we utilized imaging cytometry . Microscopy images of AF+ cells identified by the fluorescence intensity detected in two empty channels ( Figure 1G ) revealed that the AF signal was not diffusely distributed throughout the cellular volume but was punctate and localized within intracellular organelles . The subcellular AF compartments were observed in all tested AF channels and systematically colocalized ( Figure 1—figure supplement 1H and I ) . AF− cells did not display detectable AF subcellular compartments in any of the tested channels , altogether establishing that these two subsets of microglia , identified solely by their AF profiles , differed by the presence of highly autofluorescent intracellular organelles that were restricted to the AF+ microglia subset . In contrast to the punctate AF signal detected in 3-month-old animals , imaging flow cytometry applied to naïve mice from a range of ages revealed that AF subcellular structures became largely confluent in 10-month-old animals and occupied a larger fraction of the cytosol ( from 11 μm2 to 17 μm2 on average in 3- and 10- month-old animals , respectively ) ( Figure 2A , B ) . Consistent with these results , flow cytometry analyses established a nearly linear increase of AF signal intensity in AF+ microglia with aging ( Figure 2C , D ) , resulting in a cumulative 3-fold increase of AF signal between 3- and 12-month-old mice across multiple fluorescence channels ( Figure 2D and Figure 2—figure supplement 1A ) . While a clear AF+ population was not detectable in mice aged 15 days post-natal or younger , a bimodal distribution appeared as early as 30 days post-natal and AF increased linearly with age from that point on ( Figure 2—figure supplement 1B , C ) . In contrast , age-dependent increases in cellular AF were not detected in peripheral CD11B+ cells residing in the spleen ( Figure 2—figure supplement 1D ) . Despite the large increase in the AF signal observed in AF+ cells , their frequency remained largely unchanged during aging , decreasing only slightly from 75% to 72% from 3 to 12 months of age ( Figure 2E ) . Finally , neither the proportion of AF− cells ( Figure 2E ) nor the area of intracellular AF signal detected in AF− microglia ( Figure 2A , B ) were altered by aging , revealing a selective impact of aging on the AF+ microglia subset . We next isolated AF+ and AF− microglia from 3- and 18-month-old mice using fluorescence-activated cell sorting ( FACS ) and performed transmission electron microscopy ( TEM ) analyses . At 3 months of age , the intracellular organization of AF microglia subsets differed in that AF+ cells almost systematically contained large storage bodies filled with osmophilic electron-dense deposits ( Figure 3A ) . These electron-dense storage bodies were observed in 80% of AF+ microglia whereas only 46% of AF− microglia contained electron-dense organelles ( Figure 3A , C ) . When observed in AF− cells , storage bodies were devoid of complex storage material and were more regularly shaped than those observed in AF+ cells ( Figure 3A ) . In aged mice , the storage bodies within AF+ microglia changed dramatically in both size and complexity ( Figure 3A ) . Between 3 and 18 months of age , the percentage of visible cytoplasm occupied by storage bodies increased from 9% to 23% on average and approximately 30% of AF+ cells from aged mice contained AF material that occupied more than a third of the cytoplasm ( Figure 3D ) . The ultrastructural composition of the storage bodies in AF+ cells varied with frequent curvilinear ( black arrowheads ) and fingerprint-like profiles ( red arrowheads ) ( Figure 3B ) . White , linear , rod-like material with fine-tipped ends ( white arrowheads ) were observed within membrane-bound lipoid bodies and closely resembled cholesterol crystal deposits ( Figure 3B ) . Most prominent , however , was the proportion and volume of lipid droplets ( asterisks ) ( Figure 3B ) . In contrast , both the frequency of cells containing storage bodies and the proportion of the cytosol occupied by storage bodies remained unchanged in the AF− microglia subset ( Figure 3A , C , D ) . In addition to these ultrastructural differences , AF+ cells expressed higher levels of LAMP1 and CD68 compared to AF− cells ( Figure 3E , F ) , indicating an enlargement of endolysosomal storage compartments in AF+ cells . Furthermore , a gradual age-dependent increase in LAMP1 and CD68 protein levels was observed in the AF+ subset ( Figure 3E , F ) whereas AF− microglia did not show changes . Altogether these observations indicated that AF+ microglia differed from AF− cells by their unique accumulation of endolysosomal storage compartments with aging . To explore the cellular dynamics and ontogenic relationships between these two novel subsets of microglia , we depleted microglia in 14-month-old mice with the CSF1R-small molecule antagonist BLZ945 ( Krauser et al . , 2015; Pyonteck et al . , 2013 ) . Twenty-four hours following the treatment period , depletion was nearly complete , as assessed by the remaining frequency and absolute numbers of microglia ( Figure 4A , B and data not shown ) . Consistent with previous reports ( Huang et al . , 2018; O’Neil et al . , 2018 ) , microglia rapidly repopulated the CNS at 7 and 14 days post-treatment , recovering on average to 25% ( at 7 days ) and 87% ( at 14 days ) of steady-state microglia numbers ( Figure 4A and data not shown ) . However , repopulation by AF+ and AF− subsets showed distinct kinetics . While AF− microglia cell numbers reached 67% of steady-state levels by day 7 , AF+ cells only repopulated to 4% of steady-state levels by that time ( Figure 4B ) . At 14 days repopulating AF+ cell numbers only reached 14% of steady-state AF+ values in vehicle-treated mice while the repopulating AF− subset surpassed steady-state levels by approximately 2-fold ( Figure 4B ) before normalizing to steady-state levels at day 70 . Altogether these results established that the AF− subset was the first subset to repopulate the depleted brain while the repopulation by AF+ cells was delayed , suggesting a possible conversion from AF− to AF+ state during repopulation . Supporting this hypothesis , the AF intensity of the repopulated microglia slowly increased from barely detectable levels at day seven to slightly increased levels at day 14 , ultimately returning to a bimodal distribution by day 70 ( Figure 4C , D ) . Even at the latter timepoint , only a small fraction of this newly-formed AF+ subset ( 19% on average ) displayed AF intensity levels comparable to those seen in vehicle-treated mice ( Figure 4C , D and Figure 4—figure supplement 1A , B ) while most repopulating AF+ cells displayed 35% weaker AF signal intensity on average . This gradual accumulation of AF material over time in repopulating microglia supported the possibility that AF+ microglia were derived from the conversion of AF− cells during replenishment . Further validating this conclusion , 56% of microglia displayed positivity for the proliferation-associated marker KI-67 at day seven before returning to steady-state values by day 14 ( Figure 4E , F ) . Because AF+ microglia were virtually absent during this early repopulation phase ( Figure 4B ) , these proliferation kinetics implied that the AF− subset of microglia was the predominant subset responsible for the repopulation of the microglia compartment following depletion and that AF+ microglia were derived from the conversion over time of a subset of AF− microglia . To determine whether microglia AF subsets were distinct at the molecular level , FACS-isolated AF+ and AF− microglia were analyzed by nano liquid chromatography mass spectrometry ( nLC-MS ) . A total of 4231 proteins were detected by label-free LC-MS/MS and after filtering for proteins quantified in at least 50% of samples in either subset , 3492 remained for analysis . 351 proteins showed significant differences in abundance between AF+ and AF− microglia ( Benjamini-Hochberg adj . p value < 0 . 01 , fold-change > |1 . 3| ) , with 254 and 97 upregulated and downregulated differentially expressed proteins ( DEPs ) , respectively ( Figure 5A and Figure 5—source data 1 ) . When ranked by significance , 32 of the top 50 DEPs upregulated in AF+ microglia were associated with endolysosomal biology , among which was LAMP1 , validating prior flow cytometry results ( Figure 3E , F ) . Furthermore , a large number of lysosomal enzymes were upregulated in AF+ microglia including cathepsins ( CTSA , CTSB , CTSD , CTSF , CTSL , CTSZ ) as well as several enzymes involved in lysosomal degradation such as amidases ( ASAH1 , GBA , NAAA ) , thioesterases ( PPT1 , PSAP , NAGA ) , proteases ( TPP1 , LGMN ) and glycosyl hydrolases ( HEXA , HEXB , GLB1 ) . Many other DEPs involved in the biology , trafficking , or fusion of endosomes with lysosomes and phagosomes ( ATP6V0D1 , SCARB2 , GRN , ARL8B , TOM1 , STX7 , TMEM55B ) were also upregulated in AF+ microglia . Underscoring the importance of these proteins towards maintaining proper CNS homeostasis , genetic perturbations in 15 of the top 50 DEPs were associated with severe CNS-related storage disorders , such as Neuronal Ceroid Lipofucinosis and Niemann-Pick . Lastly , DEPs that are typically associated with neurons ( CBLN1 , CBLN4 , SNAP25 ) , oligodendrocytes ( MOBP , SYNE2 ) and astrocytes ( GJA1 ) were detected at significantly higher levels in AF+ microglia , suggesting functional differences in either the phagocytic capacity of this subset or its ability to fully degrade ingested material . To systematically explore additional pathways distinguishing AF subsets at the molecular level , Gene Ontology ( GO ) term enrichment was performed with Panther ( Figure 5B and Figure 5—source data 2 ) . DEPs upregulated in AF+ microglia were enriched in pathways related to intracellular vesicle-mediated transport , lysosomal organization , protein transport , lipid catabolic processes and regulation of TOR signaling ( Figure 5B , C ) . DEPs downregulated in AF+ microglia showed enrichment in RNA-related biological processes ( transcription , RNA metabolism , splicing ) and chromatin silencing . In agreement with GO term enrichment analysis , the top canonical pathways identified by Ingenuity Pathway Analysis ( IPA ) as enriched in AF+ microglia included phagosome maturation , autophagy , numerous catabolic pathways ( amino acid and glutaryl-coA catabolism , ketogenesis and fatty acid β-oxidation ) , mitochondrial dysfunction , and pathways that were indicative of mTOR deregulation ( mTOR , AMPK ) ( Figure 5D and Figure 5—source data 3 ) . Finally , upstream regulators predicted to explain the distinct proteome displayed by AF+ cells included transcription factors involved in the regulation of cell cycle , senescence and apoptosis ( TP53 , MYC , CDKN2A ) , ER stress and unfolded protein response ( XBP1 ) , autophagy and lysosomal biogenesis ( TFEB ) and inflammatory responses ( NFKBIA , PPARA ) ( Figure 5E and Figure 5—source data 3 ) . Altogether , these results established that AF+ microglia expressed a unique proteome characterized by an increased representation of endolysosomal , autophagic , catabolic , and mTOR-related proteins . Given the abundance of intracellular storage bodies and the increased detection of proteins involved in endocytosis and phagosome maturation observed in AF+ microglia , we investigated whether differential phagocytosis of myelin debris contributed to the subset-specific accumulation of AF storage material . Shiverer mice ( Mbpshi/shi ) harbor a spontaneous autosomal-recessive mutation in myelin basic protein ( MBP ) which results in defective myelin compaction and physical instability of the myelin sheath ( Privat et al . , 1979; Weil et al . , 2016 ) . Despite this myelin deficiency , there were only small differences in the frequency of AF+ cells , AF signal intensity , and LAMP1 levels in microglia from 70 day old Mbpshi/shi mice as compared to control animals ( Figure 6A , B ) . Importantly , however , microglia from Mbpshi/shi mice upregulated markers known to be induced by the phagocytosis of apoptotic neuronal bodies such as CLEC7A/DECTIN1 ( Krasemann et al . , 2017; Figure 6—figure supplement 1A , B ) . Altogether , these observations validated that myelin instability was associated with increased microglia phagocytic activities in this model while also indicating that phagocytosis of unstable myelin sheaths or cellular debris was unlikely to be the primary driver of AF accumulation with aging . To explore whether other phagocytic activities were contributors to AF accumulation , we examined Fc-receptor mediated and Triggering Receptor Expressed on Myeloid cells 2 ( TREM2 ) -mediated phagocytosis using Fcer1g−/− mice , which lack the common Fc-gamma chain required for functional expression of activating Fc-receptors ( Takai et al . , 1994 ) , and Trem2−/− mice , which were reported to have microglia defective in the engulfment of apoptotic neurons , myelin debris , and synapses ( Cantoni et al . , 2015; Hsieh et al . , 2009; Wang et al . , 2015 ) . Microglia isolated from 9-month-old Fcer1g−/− and 6-month-old Trem2−/− mice displayed no differences in the frequency of AF+ microglia or AF intensity ( Figure 6C , E ) , nor did they differ in LAMP1 levels when compared to controls ( Figure 6D , F ) . Together , these results indicated that the generation of microglia AF was not dependent on either Fc-receptor- or TREM2-mediated phagocytosis . Given that the proteomic analysis identified upregulation of autophagic proteins in AF+ microglia , we reasoned that the recycling of cellular components and their subsequent lysosomal degradation might contribute to the accumulation of AF material in microglia . To disrupt the formation of autophagic vacuoles in microglia , Atg5flox/flox mice were crossed to the Cx3cr1CRE-ERT2 mouse line and treated with tamoxifen at 6 weeks of age to induce Atg5 conditional deletion in CX3CR1+ cells ( Figure 6—figure supplement 2A ) , hereafter referred to as Atg5−/− . In 12-month-old Atg5−/− mice , there was a significant decrease both in the frequency and the relative AF intensity of the AF+ subset ( Figure 6G ) . Interestingly , in Atg5−/− mice the distribution of AF intensity in AF+ cells broadened significantly compared to controls . To better characterize this change in distribution , we divided the AF+ subset into AFdim and AFhi populations ( Figure 6G ) . In Atg5−/− mice , the proportion of AFhi microglia was decreased compared to Atg5+/+ ( 30% versus 36% , p=0 . 012 , Figure 6H ) . While no difference was observed in LAMP1 ( Figure 6I ) , CD68 was significantly lower in AF+ microglia from Atg5−/− mice ( Figure 6—figure supplement 2B ) . We next examined the contribution of lysosomal pathways in the regulation of AF accumulation . To this end , we isolated microglia from Cln3Δex7/8 homozygous knock-in mice ( hereafter referred as Cln3KI/KI ) , which express the most common loss-of-function allelic variant of the lysosomal gene Cln3 found in patients with juvenile-forms of neuronal ceroid lipofuscinosis ( NCL ) ( Cotman et al . , 2002 ) . While the proportion of AF+ microglia was modestly decreased in 5-month-old Cln3KI/KI mice , AF+ microglia isolated from Cln3KI/KI animals showed a striking increase in AF signal intensity compared to microglia from control mice ( Figure 6J ) . Furthermore , the AF signal intensity was not equivalently increased across the entire light emission spectrum , with the highest difference ( 2 . 4-fold on average ) observed in the short emission wavelengths ( 525 nm ) whereas differences were more limited at 585 nm and no longer detected at 710 nm ( Figure 6J ) , likely reflecting differences in both the magnitude and composition of AF accumulation in Cln3KI/KI animals . Consistent with a lysosomal origin of AF signal in wild-type mice ( Figure 3E , F ) , LAMP1 protein was increased by more than 2-fold in AF+ microglia from Cln3KI/KI mice ( Figure 6K ) . In contrast , LAMP1 levels in AF− microglia remained unchanged across all genotypes ( Figure 6K ) , establishing that lysosomal dysfunction selectively affected the subset of microglia accumulating AF material . In summary , autophagy and lysosomal dysfunction respectively attenuated and increased the accumulation of AF material in AF+ microglia , thereby establishing autophagy and lysosomal biology as important mechanisms contributing to the formation of the AF+ microglia subset . Given that the relative proportion of AF+ microglia remained steady during the first year of life ( Figure 2E ) , we next analyzed AF microglia subsets during advanced aging , defined in mice as 18 to 24 months of age . Similar to what we observed between 3 and 12 months of age ( Figure 2C ) , the maximal intensity of AF in AF+ cells continued to increase between 12 and 24 months of age ( Figure 7A ) . In contrast to the largely bimodal distribution of AF seen during the first 12 months of age , the distribution observed at 24 months of age was more continuous , with a noticeable downward shift in AF intensity detected on a per cell basis ( Figure 7A ) . To quantify this change in distribution , we divided the AF+ subset into AFdim and AFhi populations as done previously ( Figure 6G ) . While at 12 months of age 50% of microglia were AFhi , this population decreased to 32% on average by 24 months of age , with a concomitant increase in AFdim and AF− cells ( Figure 7B ) . This apparent relative loss of the AFhi microglia subset was accompanied by a 31% decrease in the overall number of microglia recovered from 24-month-old mice as compared to 12-month-old animals . This decreased recovery did not affect all AF microglia subsets equally but was uniquely attributed to a 58% decrease in the absolute number of AFhi cells ( Figure 7C ) . To explore the possible factors involved in the preferential loss of AFhi microglia with age , we examined rates of proliferation and cell death in adult and aged animals . The frequency of microglia positive for AnnexinV increased in aged mice ( 4 . 0 ± 0 . 9% , n = 8 ) as compared to adult mice ( 1 . 9 ± 0 . 2% , n = 7 ) and was mostly attributed to AFhi and AFdim cells , which showed a 2-fold increase in the cell death rate at 24 months of age as compared to 12 months of age ( Figure 7D , E ) . By contrast , the cell death rate seen in AF− cells did not significantly increase between 12 and 24 months of age . When quantifying specifically early-apoptotic ( AnnexinV+ DAPI− ) cells , AFhi and AFdim microglia exhibited similar relative increases in early-apoptotic rates between 12 and 24 months of age ( Figure 7—figure supplement 1A ) , altogether suggesting that increased rates in apoptotic and necrotic death were both contributing to the overall 2-fold higher cell death rate seen in AFhi and AFdim microglia with aging . Unlike cell death rates , homeostatic proliferation rates at 12 months of age were generally higher in AF− microglia than in AFdim ( by 2-fold on average ) and AFhi cells ( by 7-fold on average ) and did not change with aging , with the exception of AFhi microglia which showed a 2 . 6-fold increase in proliferative rate between 12- and 24 months of age ( Figure 7F , G ) . Because microglia from Cln3KI/KI mice accumulated higher levels of AF in the AF+ population ( Figure 6J ) , we reasoned that accelerated accumulation of AF material may exacerbate the differences in AF subset survival . Accordingly , as early as at 9 months of age and more prominently at 18 months of age , we observed a marked decrease in the frequency of the AFhi subset , using channels that either showed increased AF signal at 5 months of age ( e . g . , Blue-525 nm ) or channels that showed no increase in AF signal at this age ( e . g . , Blue-710 nm ) ( Figure 7H and Figure 7—figure supplement 1B , C ) . This decreased proportion also correlated with AF dose-dependent reductions in microglia cell numbers ( by 30% and 85% , respectively , for AFdim and AFhi microglia , Figure 7I ) and increases in overall cell death and early-apoptotic rates in AFhi and AFdim microglia , but not AF− cells from Cln3KI/KI mice as compared to control mice ( by 3 . 0- and 2 . 0-fold , respectively , Figure 7J and Figure 7—figure supplement 1D ) . Altogether , these observations supported a model where the age-associated progressive accumulation of AF material directly caused increased cell death rates , thereby leading to the selective decline of the AF+ subset observed during natural aging and the premature collapse of this subset upon accelerated accumulation of AF caused by genetic manipulation of the Cln3 pathway . Based on proteomic analyses , we hypothesized that the higher apoptotic rates observed in AF+ cells could result from mitochondrial dysfunction , the highly catabolic metabolism of AF+ cells or their reliance on fatty acid β-oxidation as a source of energy , all metabolic processes which are known to generate high levels of ROS . As measured by a ROS-sensitive cellular dye , AF+ microglia from 3- and 24-month-old mice displayed baseline ROS levels that were on average 2-fold and 2 . 4-fold higher than those observed in AF− microglia , respectively ( Figure 7K ) . After treatment with tert-Butyl hydroperoxide ( tBHP ) , a potent inducer of cellular ROS , AF+ microglia generated ROS levels that were 1 . 5-fold and 2-fold higher than those observed in AF− microglia at 3 and 24 months of age , respectively ( Figure 7L ) . tBHP-induced ROS generation increased with aging and this increase was selectively observed in AF+ microglia that generated 1 . 6-fold more ROS at 24 months than at 3 months of age ( Figure 7L ) . To determine whether differences in mitochondrial content were contributing to the increased levels of ROS observed in AF+ cells , we assayed microglia cellular content using the Mitotracker probe . AF+ microglia exhibited , 1 . 5-fold and 1 . 6-fold higher levels of Mitotracker fluorescence on average than AF− cells at 3 and 24 months of age , respectively ( Figure 7M ) . There was no age-dependent increase in Mitotracker signal detected for AF+ microglia . Together , these data established that AF accumulation in AF+ microglia was associated with increased generation of ROS , which may contribute to the changes selectively observed in AF+ physiology with aging and lysosomal dysfunction .
Generally , a cellular subset should exhibit shared properties with other cells within the cell type while exhibiting unique features and selective physiological functions that are independent of their microenvironment or external stimuli . It has therefore been suggested that true microglial subsets should be defined in unchallenged conditions by their intrinsic properties which translate into unique physiological functions during development or aging ( Stratoulias et al . , 2019 ) . This report describes novel microglia subsets that precisely satisfy each criteria of this definition: ( 1 ) Cellular autofluorescence constitutes a cell-intrinsic photophysical parameter that unexpectedly distributed in a bimodal fashion in microglia; ( 2 ) The distinct subsets of microglia identified by this parameter were found in healthy , non-challenged conditions in both mice and non-human primates , irrespective of brain region or local microenvironment ( e . g . , various levels of myelin content such as in Shiverer mice ) and independently of any specific stimulus or age ( e . g . , these subsets were present in young healthy brain as well as in aged animals ) ; ( 3 ) AF+ cells selectively accumulated lysosomal storage bodies , therefore identifying the subset of microglial cells endowed with clearance functions in the brain and suggesting that this intrinsic AF property reflected unique physiological functions; ( 4 ) These subsets were detected longitudinally at a fixed ratio throughout most of adult life , implying the existence of homeostatic mechanisms maintaining a strict abundance ratio between them . Microglia that were negative for autofluorescence were detected even in advanced aging conditions , excluding the possibility that these subsets could be simple cellular phenotypes . While DAM , MgnD , LDAM and PAM microglia states ( Hagemeyer et al . , 2017; Kamphuis et al . , 2016; Keren-Shaul et al . , 2017; Krasemann et al . , 2017; Li et al . , 2019; Marschallinger et al . , 2020 ) were previously described in disease settings , aging and during development , functionally-defined populations of microglia have not been identified or characterized in steady-state settings ‒ to the best of our knowledge ‒ , despite the heterogeneity of microglia observed by mass cytometry and single cell transcriptomic methods ( Hammond et al . , 2019; Li et al . , 2019; Masuda et al . , 2019; Matcovitch-Natan et al . , 2016; Mrdjen et al . , 2018 ) . The novel subsets described here , which we propose to name AF+ and AF− , are predicted to have distinct functions and potentially represent the first defined subsets of microglia in the unchallenged healthy brain . Lines of evidence point to a model where AF+ microglia derive from the gradual and synchronous conversion of a subset of AF− cells: ( 1 ) AF was found to linearly increase from 3 months to 12 months of age , indicative of a progressive and cumulative biological process during aging; ( 2 ) Microglia proliferation following depletion was largely restricted to the AF− subset which fully replenished within 14 days; ( 3 ) The post-depletion re-emergence of the AF+ subset was delayed until after proliferation had subsided . One limitation of these repopulation experiments is that we were unable to exclude the possibility that a small number of surviving AF+ cells were responsible for repopulating the AF+ niche over a longer period of time . Only future lineage tracing experiments or the adoptive-transfer of congenically-labeled AF+ and AF− microglia into neonatal hosts as recently described ( Hasselmann et al . , 2019 ) will definitely establish the ontogenic relationship between these microglia subsets . Our repopulation results are in agreement with the reported decrease in both lipofuscin content and CD68+ lysosomal volume described by O’Neil and colleagues at 21 days following PLX5622-induced microglia depletion and repopulation ( O’Neil et al . , 2018 ) . While the authors concluded that enforced microglia turnover reversed age-related lipofuscin accumulation , our data indicate that a discrete AF+ microglia population re-emerges post depletion and that repopulating AF+ microglia progressively accumulate AF with time , further establishing that AF accumulation is a progressive physiological process directly linked to aging . However , several important questions remain to be addressed in future studies , among which are: ( 1 ) the formal demonstration of the irreversibility of AF accumulation with aging; ( 2 ) the identification of the molecular cues responsible for the conversion from AF− to AF+; and ( 3 ) the elucidation of the mechanisms maintaining AF subset ratios . Historically , the gradual increase of AF in the brain with aging has been attributed to the accumulation of lipofuscin within postmitotic neurons and glial cells ( Nakanishi and Wu , 2009; Seehafer and Pearce , 2006; Xu et al . , 2008 ) . Lipofuscin is characterized by a distinctive cytoplasmic accumulation of non-degradable fluorescent storage material comprised of highly crosslinked , polymeric and oxidized macromolecules as well as metal cations ( Brunk and Terman , 2002 ) . While lipofuscin typically displays spectral excitation and emission maxima in the range of 320–480 nm and 460–630 nm , respectively ( Jung et al . , 2010; Moreno-García et al . , 2018; Warburton et al . , 2007 ) , the exact composition is known to vary between brain regions and cell types , likely impacting its precise spectral properties ( Gilissen and Staneva-Dobrovski , 2013 ) . In contrast , several exciting laser wavelengths ( 405 , 488 , 561 , 640 nm ) led to the emission of AF in microglia while the AF emission spectrum was also characterized by an equally wide range of wavelengths ( 450–780 nm ) observed on a per cell basis . The latter property suggests that the AF signal detected within each AF+ microglia originates from complex subcellular content that may only partially overlap with traditional lipofuscin detected in other cell types . For example , while ceroid and lipofuscin exhibit some overlapping characteristics , these storage materials have unique compositions that in turn display subtle spectral differences ( Seehafer and Pearce , 2006 ) . The possibility that microglia AF observed by flow cytometry is detecting a wide range of storage material is further supported by the genotype-dependent spectral differences observed in Cln3KI/KI microglia , which displayed increased fluorescence intensity in select channel wavelengths , likely reflecting the specific accumulation of ceroid and not all possible subtypes of autofluorescence-emitting storage materials that may accumulate in AF+ microglia with normal aging . One limitation of this study is that the identification of AF subsets currently requires analysis of a single cell suspension by flow cytometry . Efforts leveraging proteomics-generated DEPs are currently ongoing to understand the spatial distribution in situ of AF+ and AF− microglia in the context of different CNS regions and in relation to other cell types or subcellular structures such as synapses or apoptotic bodies . Understanding how aging and lysosomal dysfunction modify the spatial distribution of these two microglia subsets within the CNS should ultimately improve our understanding of the differential origin and composition of storage bodies accumulating in AF+ microglia . When imaged by TEM , several types of storage compartments were detected in AF+ microglia , among which were: lipid-based storage material which was reminiscent of the lipid droplets recently described in lipid droplet-accumulating microglia ( LDAM ) ( Marschallinger et al . , 2020; crystalline deposits which resembled cholesterol crystals previously described in foamy macrophages , lipid laden microglia and endothelial cells ( Baumer et al . , 2017; Cantuti-Castelvetri et al . , 2018; Klinkner et al . , 1995; Tangirala et al . , 1994 ) ; and finally , curvilinear and fingerprint-like pattern depositions , which are hallmark features of storage disorders such as NCL and progranulin haploinsufficiency ( Anderson et al . , 2006; Cotman et al . , 2002; Ward et al . , 2017 ) and suggested a mechanistic convergence between lipofuscinosis/ceroid pathologies and microglia dysfunction associated with aging ( Spittau , 2017 ) . In addition to differences in subcellular content and population dynamics , AF+ microglia showed distinct differences at the proteome level , characterized by increased representation of endolysosomal and autophagic pathways as well as proteins involved in oxidative phosphorylation , catabolic metabolism and fatty acid β-oxidation . These proteomic differences are reminiscent of previously reported effects of aging on the microglia proteome , which consisted of a bioenergetic shift towards fatty-acid utilization , perturbations in the regulation of gene expression and an upregulation of proteins suggestive of mitochondrial dysfunction ( Flowers et al . , 2017 ) . However , while those changes were observed in microglia from 24-month-old mice , we detected these differences in AF+ microglia from mice as young as 12-month-old , suggesting that this aging signature may be primarily driven by the AF+ subset . GO term enrichment and IPA analyses also identified dysregulation of mTOR-related proteins and pathways in AF+ microglia . mTORC1 , which integrates upstream pathways sensing nutrient availability , cellular energy and growth factors , regulates cell growth through its control of metabolism , autophagy , protein translation and organelle biogenesis ( Liu and Sabatini , 2020 ) . The activity of mTORC1 is tightly controlled by its recruitment to the lysosomal surface by the Rag-Ragulator complex , where is it brought into close proximity to the activating GTPase Rheb . The elevated levels of Ragulator components ( LAMTOR1 , 2 , 4 , 5 ) and Rag members RRAGA/B and RRAGC in AF+ cells are likely reflective of the increased lysosomal content present in these cells . The oxidative and catabolic metabolic profile of AF+ microglia along with their overrepresentation of autophagic proteins points to an impaired activation of mTOR in AF+ microglia and suggests that pharmacological modulation of mTOR may restore AF+ microglia physiology , a hypothesis that remains to be tested . Intriguingly , recent work exploring the effects of elevated mTORC1 signaling in Tsc1-deficient microglia described increased lysosomal biogenesis and phagocytosis that was coincident with decreased synaptic density and a more reactive microglial morphology ( Zhao et al . , 2018 ) . In prior studies which analyzed lipofuscin in microglia during aging ( Mendes-Jorge et al . , 2009; Moreno-García et al . , 2018; Nakanishi and Wu , 2009 ) , the assumption was that the accumulation occurred equivalently across all microglial cells . Instead , we have established that the progressive increase of AF material and associated accumulation of LAMP1+ storage bodies are occurring exclusively within AF+ microglia , highly suggestive of functional differences between subsets . In contrast , AF− microglia remained devoid of AF signal during natural aging and even when lysosomal function was impaired by the Cln3Δex7-8 mutation . The persistence of the AF− subset in those perturbed conditions suggests specific mechanisms or regional cues governing the distinct functions unique to each AF subset that will be the object of future investigation . Interestingly , aging-related myelin degradation , shedding and subsequent phagocytosis by microglia were proposed to be the primary contributors to the accumulation of lipofuscin ( Safaiyan et al . , 2016 ) . However , our observations in the MBP-mutant Shiverer mice ( Molineaux et al . , 1986 ) , which exhibit CNS hypomyelination and myelin sheath instability ( Bird et al . , 1978; Kirschner and Ganser , 1980; Privat et al . , 1979; Safaiyan et al . , 2016; Weil et al . , 2016 ) , do not fully support this model as the bimodal AF distribution was largely unaltered in Mbpshi/shi microglia despite their upregulation of CLEC7A/DECTIN1 , which likely reflected their active phagocytosis of apoptotic bodies and cellular debris caused by myelin instability . Therefore , myelin phagocytosis and degradation are unlikely to be the primary contributor to the microglia AF signal . This conclusion is compatible with observations made in the subretinal layers of the eye where the increased lipofuscin signal observed in microglia with aging or age-related macular degeneration was not linked to myelin phagocytosis , but to the phagocytosis of rod outer segments and apoptotic retinal pigment epithelial cells ( Santos et al . , 2010; Xu et al . , 2008 ) . While the lack of alterations in AF with the loss of Fc-receptor- or TREM2-mediated phagocytosis indicates that those particular pathways are not key contributors , AF accumulation may still be impacted by alternative phagocytic pathways active in microglia that were not tested in this study , such as TAM ( Tyro3 , Axl , Mer ) ( Fourgeaud et al . , 2016; Tufail et al . , 2017 ) , SIRPα-CD47 ( Hutter et al . , 2019 ) and C3-CD11b ( Fu et al . , 2012; Schafer et al . , 2012 ) . In contrast , we established that lysosomal and autophagosomal degradation pathways are primary mechanisms of accumulation of AF material , as evidenced by the accelerated and decreased AF accumulation respectively observed in AF+ microglia from Cln3Δex7-8 mice and Atg5-deficient microglia . Of all cell types in the brain , microglia express the highest transcript levels of Cln3 , which is involved in the regulation of lysosomal pH , cathepsin activity , and endocytic trafficking ( Cárcel-Trullols et al . , 2017; Cotman and Staropoli , 2012; Golabek et al . , 2000; Schmidtke et al . , 2019 ) . Although the disruption of Cln3 expression was present in both AF+ and AF− cells in Cln3Δex7-8 mice , the latter subset was unaffected by the perturbation of the lysosomal pathway , suggesting that CLN3-dependent lysosomal degradation is dispensable in AF− microglia , which further highlights the molecular differences between AF+ and AF− microglia . While lipofuscin accumulation has been linked to age-related neuronal and microglial functional decline ( von Bernhardi et al . , 2015; Brunk and Terman , 2002; Höhn and Grune , 2013; Jung et al . , 2007; Kurz et al . , 2008; Safaiyan et al . , 2016; Sierra et al . , 2007 ) , the ability to precisely characterize its functional impact was impeded by the inability to isolate microglia which accumulated AF storage material from those which did not . Here , we overcame this limitation by developing a method permitting the isolation of these microglia subsets and established that the age-dependent accumulation of AF storage material was associated with increased mitochondrial mass and ROS production , decreased homeostatic proliferation and increased cell death rates in AF+ cells . Increased ROS production by aging AF+ microglia could be a direct consequence of the accumulation of AF material as lipofuscin can incorporate transition metals such as copper and iron , which form a redox-active surface catalyzing the Fenton reaction and promote mitochondria-independent cytotoxic effects ( Höhn et al . , 2010 ) . Alternatively , the increased generation of ROS in AF+ could result from mitochondrial dysfunction , a pathway that was enriched in the AF+ proteome or the highly catabolic metabolism of AF+ cells and their reliance on fatty acid β-oxidation which is another major a source of ROS . In addition to ROS , the accumulation of undegraded macromolecules within lysosomes , which is known to inhibit key catabolic enzymes and permeases ( Lamanna et al . , 2011; Prinetti et al . , 2011; Walkley and Vanier , 2009 ) , could further propel the lysosomal system into complete dysfunction . The subsequent overaccumulation of storage material may also physically damage the lysosomal membranes , causing dispersal of storage contents into the cytosol or the extracellular space and subsequent inflammasome activation and neuronal cell death , as it was shown recently in a genetic model for acid sphingomyelinase deficiency and in a demyelination model ( Cantuti-Castelvetri et al . , 2018; Gabandé Rodríguez et al . , 2018 ) . Foam cell macrophages , which are generated by the overaccumulation of oxidized low-density lipid protein and cholesterol in atherosclerotic plaques , have been shown to contribute to vascular pathology via these precise mechanisms ( Childs et al . , 2016; Duewell et al . , 2010; Gibson et al . , 2018; Hakala et al . , 2003 ) . Interestingly , foam cell macrophages exhibit many of the ultrastructural features observed in aged AF+ microglia and thus may share common pathways precipitating cellular dysfunction and tissue pathology . Lastly , another possible mechanism by which AF material could trigger microglia dysfunction and cytotoxic effects is via its binding and inhibition of the proteasome and the associated perturbation of cellular proteolytic functions , as it was previously reported for lipofuscin ( Höhn et al . , 2011; Sitte et al . , 2000; Szweda et al . , 2003 ) . Despite the striking stability of the AF subset ratio throughout most adult mice , cumulative and age-dependent AF accumulation ultimately resulted in the collapse of the ratio of AF+ to AF− microglia in advanced aging , likely due to one of the aforementioned mechanisms . The accelerated accumulation of AF material in AF+ microglia from Cln3Δex7-8 mice phenocopied the effect of advanced aging on microglia by selectively increasing apoptotic rates of AF+ cells and resulting in an early and precipitous decline of the AF+ subset . Altogether , these results suggest a convergence of mechanisms operating during aging and in lysosomal storage disorders whereby the accumulation of storage bodies in AF+ cells drives microglia dysfunction , possibly contributing to neurodegeneration known to be associated with both conditions . Together , these results provide novel insight into microglia physiology by identifying two previously unknown subsets which exist at steady-state in the murine and non-human primate CNS . The ability to identify and isolate distinct microglia subsets via a cell-intrinsic , label-free method presents future opportunities to further characterize any additional functions that may shed light on their subset-specific roles in maintaining the delicate balance between CNS homeostasis and neurodegeneration . The significance of these subsets is clearly highlighted by a model of juvenile neuronal ceroid lipofuscinosis in which lysosomal dysfunction was born solely by AF+ microglia , whereas AF− microglia appeared unaffected . Lastly , the finding that the progressive depletion of AF+ microglia occurred as function of natural brain aging raises the possibility that this subset may be functionally protective and warrants further investigation into the biological factors regulating their survival .
This study was performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals . Research animals at Biogen were housed under specific pathogen free conditions in an AAALAC accredited facility with a 12 hr - 12 hr light - dark cycle and environmental conditions controlled at 21°C and 40–60% humidity . Animals handled according to an approved institutional animal care and use committee ( IACUC ) protocol ( #756 ) . This study was reviewed and approved by the Massachusetts General Hospital ( MGH ) Subcommittee of Research Animal Care ( SRAC ) , which serves as the Institutional Animal Care and Use Committee ( IACUC ) for MGH ( Protocol #2008N000013 ) . C57BL/6J , B6 . 129P2-Apoetm1Unc/J , C3Fe . SWV-Mbpshi/J ( Chernoff , 1981 ) , B6J . B6N ( Cg ) -Cx3cr1tm1 . 1 ( cre ) Jung/J ( Cx3cr1CreERT2 ) ( Yona et al . , 2013 ) were purchased from Jackson Labs . B6 . 129S-Atg5<tm1Myok> ( Atg5flox ) ( Hara et al . , 2006 ) were acquired from Dr . Noboru Mizushima ( University of Tokyo ) . Cln3Δex7/ex8 knock-in mice ( Cotman et al . , 2002 ) were generously provided by Dr . Susan Cotman ( Harvard Medical School ) and carry a ~ 1 kb genomic deletion in the lysosomal gene Cln3 that is orthologous to the commonly observed mutation in Juvenile Neuronal Ceroid Lipofuscinosis patients . This exonic deletion results in the production of an alternatively spliced mRNA species encoding a detectable non-truncated CLN3 mutant protein . B6 . 129P2-Fcer1gtm1Rav N12 ( Fcer1g−/− ) ( Takai et al . , 1994 ) were purchased from Taconic . Trem2tm1 ( KOMP ) Vlcg ( Trem2−/− ) were acquired from the KOMP repository . Atg5−/− mice were generated by crossing the Atg5flox and Cx3cr1CreERT2 lines . Age-matched cohorts were used in all experiments . All mice used in studies were female unless otherwise noted . Mice were euthanized by CO2 inhalation . One gram of tamoxifen ( Cat# T5648 , Millipore-Sigma ) was dissolved in 5 mL ethanol and then mixed with 45 mL of corn oil USP ( Cat# CO136 , Spectrum Chemical ) warmed at 37°C for a final concentration of 20 mg/mL . 6 week old mice were given two 200 µL subcutaneous injections of tamoxifen-corn oil , spaced 48 hr apart to induce genetic recombination . All solutions were used ice-cold unless otherwise noted . Immediately following CO2 euthanasia , mice were transcardially perfused with 20 mL of phosphate buffered saline ( PBS ) containing 3 mM EDTA . Brains were dissected and transferred into a 15 mL conical tube filled with Hanks’ Balanced Salt Solution ( HBSS ) on ice . For tissue homogenization , brains were transferred to a Petri dish and minced over ice with a scalpel . Minced tissue was transferred to a 7 mL glass dounce homogenizer ( Cat# 57542 , Wheaton ) containing 5 mL of HBSS + 25 mM HEPES and homogenized with a pestle for approximately 10 strokes each . The single cell suspension was transferred to a 15 mL conical tube and centrifuged at 300xg for 5 min at 4°C . The supernatant was aspirated , and the cell pellet was gently resuspended in 1 mL of fetal bovine serum ( FBS ) . A 33% isotonic Percoll solution was prepared by combining 9 mL of Percoll ( Cat# 17-0891-01 , GE Healthcare ) with 1 mL of 10X HBSS and 20 mL of 1X HBSS + 25 mM HEPES . 9 mL of the 33% isotonic Percoll solution was then added and the cell suspension mixed by inversion . 1 mL of a 10% FBS/HBSS solution was overlaid on the Percoll suspension and the cells were centrifuged at 800 g for 15 min at 4°C with no brake . The resulting cell pellet was washed once in 10 mL HBSS + 25 mM HEPES , centrifuged at 300xg for 5 min at 4°C and resuspended in a final volume of 1 mL HBSS + 25 mM HEPES . Antibody staining was performed in a 96 well v-bottom plate ( Cat# 3897 , Corning ) with the following antibodies after Fc-receptor blocking for 5 min ( anti-CD16/32 , clone 93 , Biolegend ) : anti-CD45 BV785 ( clone 30-F11 , BioLegend ) , anti-CD11B BV510 ( clone M1/70 , BioLegend ) , anti-CX3CR1 BV421 ( clone SA011F11 , BioLegend ) , and anti-P2RY12 PE ( clone S16007D , BioLegend ) . For unconjugated anti-TMEM119 ( clone 101–6 , Abcam ) an anti-rabbit AlexaFluor488 secondary antibody ( clone Poly4064 , BioLegend ) was used . DAPI ( Cat# 62248 , ThermoFisher Scientific ) was used to assess viability . For apoptotic cell analysis , AnnexinV conjugated to AlexaFluor647 ( Cat# A23204 ThermoFisher Scientific ) was used according to the supplier protocol after surface staining . For intracellular antigens , cell suspensions were fixed and permeabilized using the Transcription Factor Staining kit ( Cat# 00-5523-00 , ThermoFisher Scientific ) after surface staining with ZombieUV Fixable Viability dye ( Cat# 423107 BioLegend ) . Fixation was followed by staining with CD68 FITC ( clone FA11 , BioLegend ) LAMP1 AlexaFluor647 ( clone 1D4B , BioLegend ) and KI-67 BV421 ( clone 11F6 , BioLegend ) . Absolute cell counts were determined using 123count eBeads ( Cat# 01-1234-42 , ThermoFisher Scientific ) Flow cytometry data were acquired on a five laser-equipped LSRFortessa X-20 ( BD Biosciences ) and analyzed in FlowJoV10 ( Treestar ) . Non-autofluorescent peripheral immune cells were used as negative controls to determine the initial bifurcating gate to identify AF− and AF+ microglia subsets . To subdivide the AF+ subset , a gate encompassing AFhi microglia was drawn to capture the AF+ gaussian peak . The AFdim gate was drawn between the edge of the AF− population and the base of the AF+ gaussian peak ( lower boundary of AFhi gate ) . Imaging flow cytometry data were acquired with an AMNIS Imagestream MKII ( Millipore-Sigma ) and analyzed with AMNIS IDEAS software . Immediately following euthanasia , Cynomolgus macaques were transcardially perfused with an ice-cold Lactated Ringers Solution ( LRS ) . Brains were removed and placed into ice-cold LRS and kept chilled until processing . Approximately 500 mg of tissue from the frontal cortex was excised , minced and gently homogenized on ice in a 7 mL glass dounce homogenizer ( Cat# 57542 , Wheaton ) containing 5 mL of HBSS + 25 mM HEPES . Microglia were then isolated using the same procedure as described above for murine microglia . Antibody staining was performed in a 96 well v-bottom plate ( Cat# 3897 , Corning ) with the following antibodies after Fc-receptor blocking ( Human TruStain FcX , Cat#422301 , Biolegend ) : anti-CD45 BV786 ( clone D058-1283 , BD Biosciences ) , anti-CD11B BV510 ( clone ICRF44 , BD Biosciences ) . Single cell suspensions were prepared as previously described for flow cytometry microglia isolations through the initial centrifugation . The resulting cell pellet was resuspended in 5 mL of a room-temperature ( RT ) 70% isotonic Percoll solution , which was carefully overlaid with 5 mL of a 37% RT isotonic Percoll solution . The Percoll cell suspension was then centrifuged at 800 g for 25 min at 22°C with acceleration set to 50% and brake set to 10% . Enriched microglia were then carefully collected from the 37%/70% Percoll interphase to a new 15 mL conical tube previously blocked with 2% bovine serum albumin ( BSA ) in PBS at RT for 3 hr . Microglia were washed once with 10 mL of ice-cold HBSS + 25 mM HEPES and centrifuged at 300 g for 5 min at 4°C . The resulting microglia cell pellet was resuspended in 100 μL of HBSS + 2% BSA . Cell suspensions were Fc-blocked and then stained with anti-CD45 AF488 ( clone 30-F11 , BioLegend ) , anti-CD11B APC ( clone M1/70 , BioLegend ) , anti-CX3CR1 BV785 ( clone SA011F11 , BioLegend ) . DAPI was used for viability and a dump gate in the PE channel was created with the following antibodies: anti-Ly6G PE ( clone 1A8 , BD Biosciences ) , anti-Ly6C PE ( clone HK1 . 4 , BioLegend ) , anti-CD3 PE ( clone 17A2 , BD Biosciences ) , anti-CD19 PE ( clone 1D3 , BD Biosciences ) , anti-NK1 . 1 PE ( clone PK136 , BD Biosciences ) . Cell sorting was performed on a BD FACSAria Fusion instrument equipped with an 85 μm nozzle . 75 , 000 microglia were FACS-isolated into tubes containing the dried residual of 150 μL 8M urea , 5 mM EDTA , 0 . 1M Tris/HCl pH 8 . 5 , resulting in final volumes of 300 μL . The suspensions were mixed and stored frozen until used . Lysates were thawed and treated with a E220 focused beam ultrasonicator ( Covaris ) 20 times at 2 s each with 150W output . Lysates were clarified by centrifugation , reduced with 10 mM final concentration of DTT for 20 min and subsequently alkylated with 30 mM 2-iodoacetamide for 30 min at room temperature . After a two-fold dilution in water , samples were digested overnight with 0 . 05 μg LysC ( Wako ) and 0 . 2 μg modified trypsin ( Promega ) . Peptides were acidified to a final concentration 0 . 5% ( v/v ) trifluoroacetic acid and desalted using C18 StageTips ( Rappsilber et al . , 2007 ) . 50% of resulting peptides were separated on a 50 cm , 75 μm inner diameter EasySpray column packed with 2 μm PepMap C18 RSLC material ( ThermoFisher ) over 120 min using an EASY-nLC 1200 system ( ThermoFisher ) . Peptides were analyzed online with an Orbitrap Fusion-Lumos mass spectrometer ( ThermoFisher ) in data-dependent acquisition mode . Full scans were acquired at a resolution of 240 , 000 in the Orbitrap analyzer and the most abundant precursors were selected in a 1 s scan cycle for higher-energy dissociation ( HCD ) with a 0 . 7 Th isolation window , a target of 10 , 000 ions and a maximum injection time of 25 ms , followed by detection fragment ion in the ion trap . Raw data were processed in MaxQuant Version 1 . 6 . 0 . 16 ( Cox and Mann , 2008 ) and searched with Andromeda ( Cox et al . , 2011 ) against a comprehensive SwissProt release for mouse with a false discovery rate of 1% at the peptide and protein level . Identifications of MS1 features were transferred between samples using the Match Between Runs option with a 0 . 2 min matching tolerance . Label-free protein quantification was performed using the MaxLFQ algorithm ( Cox et al . , 2014 ) . Data analysis was performed in Perseus ( Tyanova et al . , 2016 ) . After log transformation and applying a filter for proteins quantified in at least 50% of samples in either group , missing values were imputed with a normal distribution with a mean shifted down by 1 . 8-fold and a width of 0 . 3-fold in relation to the standard distribution of measured values of the respective sample . p values were calculated using Student’s t-test and the false discovery rate was controlled with the Benjamini-Hochberg method . Proteins were determined to be differentially expressed if they met an adjusted p value threshold of <0 . 01 and an absolute fold change difference threshold of >1 . 3 . Human orthologs of identified mouse proteins were assigned using Panther and GO term enrichment of differentially expressed proteins was calculated with the Panther Overpresentation Test ( Mi et al . , 2019 ) . Canonical pathway analysis and identification of upstream regulators was performed with Ingenuity Pathway Analysis ( Qiagen ) . Microglia from pooled animals ( n = 10 per age ) were FACS-isolated directly into tubes containing 4% glutaraldehyde/0 . 1M sodium-cacodylate solution , pH 7 . 4 ( Electron Microscopy Sciences , Hatfield , PA ) . Fixed cells were pelleted and post-fixed in 1 . 0% osmium tetroxide in cacodylate buffer for 1 hr at RT , rinsed in cacodylate buffer , and stabilized using 2% agarose in PBS . Agarose-embedded pellets were dehydrated through a graded series of ethanols to 100% , then dehydrated briefly in 100% propylene oxide . Specimens were infiltrated in a mix of 2:1 propylene oxide/Eponate resin ( Ted Pella , Redding , CA ) 2 hr at room temperature on a gentle rotator , then switched into a 1:1 mix of propylene oxide/Eponate resin and allowed to infiltrate overnight on a gentle rotator . 24 hr later , specimens were placed into fresh 100% Eponate resin and allowed to infiltrate for several hours , then embedded in flat molds with fresh 100% Eponate . Polymerization occurred within 24–48 hr at 60°C . Thin ( 70 nm ) sections were cut using a Leica EM UC7 ultramicrotome , collected onto formvar-coated grids , contrast-stained with uranyl acetate and Reynold's lead citrate and examined in a JEOL JEM 1011 transmission electron microscope at 80 kV . Images were collected using an AMT digital imaging system with proprietary image capture software ( Advanced Microscopy Techniques , Danvers , MA ) . TEM image analysis and area determinations for subcellular features were calculated in FIJI ( Schindelin et al . , 2012 ) . Cytosolic area values were calculated by determining the full cellular area and subtracting the nuclear region . Mitotracker DeepRedFM ( Cat# M22426 , ThermoFisher Scientific ) reagent was used according to the supplier protocol . Briefly , a single cell suspension of isolated microglia was surface stained with CD45 BV786 and CD11B BV510 . Cells were pelleted and then resuspended gently in 500 μL prewarmed 37°C staining solution containing 50 nM of Mitotracker reagent . Cells were incubated for 30 min at 37°C in a humidified 5% CO2 incubator then pelleted and resuspended in 200 μL of HBSS+2% BSA for analysis by flow cytometry . CellROX Deep Red ( Cat# C10491 , ThermoFisher Scientific ) was used according to supplier protocol . Briefly , a single cell suspension of isolated microglia was surface stained with CD45 BV786 and CD11B BV510 . Cells were pelleted and then resuspended gently in prewarmed RPMI1640 ( Cat# 11875085 , ThermoFisher Scientific ) containing 10% FBS . Cells were treated with a 100 μM solution containing tert-butyl hydroperoxide for 30 min at 37°C in a humidified 5% CO2 incubator to induce ROS . CellROX reagent was then added to a final concentration of 500 nM and the cells were incubated for 45 min at 37°C . Cells were pelleted , resuspended in 200 μL of prewarmed RPMI1640/10%FBS containing DAPI ( 0 . 1 μg/mL final concentration ) and then promptly analyzed on a flow cytometer . To deplete microglia , 14-month-old mice were orally treated q . d . for 7 days with BLZ945 at a dose of 200 mg/kg body-weight ( Krauser et al . , 2015 ) in a 20% captisol solution . Vehicle-treated mice received only 20% captisol solution . Tissues were collected at the indicated time after last dose . 50 , 000–100 , 000 microglia were FACS-isolated directly into chilled RNA/DNA free microfuge tubes containing 500 μL RLT-plus lysis buffer supplemented with 1% 2-mercaptoethanol . RNA was isolated with the RNEasy Plus Micro Kit ( Cat# 74034 , Qiagen ) according to the kit protocol . RNA quality was assessed using the Agilent RNA6000 Pico kit ( Cat# 5067–1513 , Agilent ) on an Agilent Bioanalyzer system . RNA quantity was determined with the Quant-iT RiboGreen Fluorescence assay ( Cat# R11490 , ThermoFisher Scientific ) . Isolated RNA was then converted to cDNA using the SuperScript IV VILO kit ( Cat# 11754050 , ThermoFisher Scientific ) per manufacturer protocol . Microglia cDNA was combined with a QuantiTect Multiplex PCR ( Cat# 204541 Qiagen ) master mix containing Taqman primers spanning Atg5 exons 3–4 ( AssayID Mm01187301_m1 , Cat# 4351372 , ThermoFisher Scientific ) and Gapdh endogenous control primers ( Cat# 4352339E , ThermoFisher Scientific ) . qRT-PCR reaction was run for 40 cycles on a QuantStudio12K Flex instrument ( ThermoFisher Scientific ) and relative quantity ( RQ ) of Atg5 transcript was calculated using the 2ΔΔCt method . Statistical analysis was performed using Prism software ( GraphPad , San Diego , CA ) . To determine significance values when comparing AF subsets , a paired , two-tailed t-test or 2-way repeated measures ANOVA with Tukey’s or Sidak’s post-hoc multiple comparison’s test were utilized . For comparisons between genotypes , significance testing was performed using two-tailed independent t-test or 1-way ANOVA followed by Tukey’s post-hoc for multiple comparisons . For BLZ945 experiments , a 1-way ANOVA followed by Dunnett’s post-hoc test for multiple comparisons was used . A value of p<0 . 05 was considered statistically significant . All data were presented as the mean with standard deviation . | Microglia are a unique type of immune cell found in the brain and spinal cord . Their job is to support neurons , defend against invading microbes , clear debris and remove dying neurons by engulfing them . Despite these diverse roles , scientists have long believed that there is only a single type of microglial cell , which adapts to perform whatever task is required . But more recent evidence suggests that this is not the whole story . Burns et al . now show that we can distinguish two subtypes of microglia based on a property called autofluorescence . This is the tendency of cells and tissues to emit light of one color after they have absorbed light of another . Burns et al . show that about 70% of microglia in healthy mouse and monkey brains display autofluorescence . However , about 30% of microglia show no autofluorescence at all . This suggests that there are two subtypes of microglia: autofluorescence-positive and autofluorescence-negative . But does this difference have any implications for how the microglia behave ? Autofluorescence occurs because specific substances inside the cells absorb light . In the case of microglia , electron microscopy revealed that autofluorescence was caused by structures within the cell called lysosomal storage bodies accumulating certain materials . The stored material included fat molecules , cholesterol crystals and other substances that are typical of disorders that affect these compartments . Burns et al . show that autofluorescent microglia contain larger amounts of proteins involved in storing and digesting waste materials than their non-autofluorescent counterparts . Moreover , as the brain ages , lysosomal storage material builds up inside autofluorescent microglia , which increase their autofluorescence as a result . Unfortunately , this accumulation of cellular debris also makes it harder for the microglia to perform their tasks . Increasing evidence suggests that the accumulation of waste materials inside the brain contributes to diseases of aging . Future work should examine how autofluorescent microglia behave in animal models of neurodegenerative diseases . If these cells do help protect the brain from the effects of aging , targeting them could be a new strategy for treating aging-related diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"tools",
"and",
"resources",
"immunology",
"and",
"inflammation"
] | 2020 | Differential accumulation of storage bodies with aging defines discrete subsets of microglia in the healthy brain |
The speciose mammalian order Eulipotyphla ( moles , shrews , hedgehogs , solenodons ) combines an unusual diversity of semi-aquatic , semi-fossorial , and fossorial forms that arose from terrestrial forbearers . However , our understanding of the ecomorphological pathways leading to these lifestyles has been confounded by a fragmentary fossil record , unresolved phylogenetic relationships , and potential morphological convergence , calling for novel approaches . The net surface charge of the oxygen-storing muscle protein myoglobin ( ZMb ) , which can be readily determined from its primary structure , provides an objective target to address this question due to mechanistic linkages with myoglobin concentration . Here , we generate a comprehensive 71 species molecular phylogeny that resolves previously intractable intra-family relationships and then ancestrally reconstruct ZMb evolution to identify ancient lifestyle transitions based on protein sequence alone . Our phylogenetically informed analyses confidently resolve fossorial habits having evolved twice in talpid moles and reveal five independent secondary aquatic transitions in the order housing the world’s smallest endothermic divers .
A fundamental challenge of evolutionary biology is to understand the phylogenomic foundations of biochemical and physiological specialisations that have enabled organisms to proliferate into new adaptive zones . While species of most mammalian orders generally occupy a single niche , members of a few mammalian lineages have radiated into a diverse range of environmental habitats . The order Eulipotyphla is one such assemblage , in that terrestrial forms repeatedly evolved into high-altitude , scansorial ( climbing ) , fossorial ( subterranean ) , semi-fossorial , and semi-aquatic niches ( Burgin and He , 2018; He et al . , 2017 ) . The eulipotyphlan clade consists of 527 recognized species of small insectivorous mammals distributed among four extant families ( Burgin et al . , 2018 ) : Erinaceidae ( hedgehogs , gymnures , and moonrats ) , Soricidae ( shrews ) , Talpidae ( moles , shrew moles , shrew-like moles , and desmans ) , and Solenodontidae ( solenodons ) . A fifth family ( Nesophontidae , Caribbean ‘island-shrews’ ) only became extinct in the past 500 years following the arrival of Europeans in the West Indies ( MacPhee et al . , 1999 ) . Fossil evidence supports terrestrial habits as ancestral for Eulipotyphla , with shrew-like moles , erinaceids , and the majority of shrews retaining morphological characteristics ( e . g . slender legs and a long , pointed mouse-like snout; Figure 1 ) effective for hunting and avoiding predation on land ( Churchfield , 1990; Nowak , 1999 ) . By contrast , fossorial moles have evolved powerful forelimbs that are held out to the sides of their tube-shaped bodies and have large , spade-like hands facing posteriorly and oriented vertically allowing them to be effective burrowers ( Gorman and Stone , 1990 ) . Semi-fossorial shrews , solenodons , and shrew moles exhibit specialisations in-between that of terrestrial and fossorial lineages , often retaining pointed mouse-like snouts yet possessing forelimbs that are oriented downwards and backwards to facilitate digging ( Gorman and Stone , 1990; Nowak , 1999 ) . Finally , semi-aquatic moles and shrews possess a stream-lined body and limb adaptations for underwater locomotion ( Nowak , 1999; Burgin and He , 2018 ) , with the larger-bodied desmans having an outward appearance more similar to that of semi-aquatic rodents ( e . g . muskrat ) than to that of any other eulipotyphlan mammal ( Figure 1 ) . The secondary aquatic transitions of water shrews—which represent the world’s smallest endothermic divers—are especially remarkable , given they have the lowest on-board oxygen stores , the highest mass-specific oxygen requirements , and most unfavorable surface area to volume ratios for body heat retention among all mammalian breath-hold divers ( Calder , 1969; Gusztak , 2008 ) . While the lifestyle of extant eulipotyphlan mammals is easily predicted based on external morphological features alone ( e . g . Woodman and Gaffney , 2014; Sansalone et al . , 2016; Sansalone et al . , 2019 ) , definitive assignments of transitional evolutionary stages have remained elusive due to an incomplete , fragmentary fossil record ( Hickman , 1984; Sanchez-Villagra et al . , 2006; Piras et al . , 2012; Hooker , 2016 ) . Additionally , because morphological specialisations coincide with lifestyle across the different clades of eulipotyphlan mammals , it has been difficult to discern whether fossorial and semi-aquatic specialisations found in Eurasian and North American moles arose in a common ancestor or are due to convergent evolution ( Schwermann and Thompson , 2015; He et al . , 2017 ) . As a result , long-standing questions persist regarding the evolutionary pathways leading to the diverse ecomorphological variation seen in eulipotyphlan mammals . For example , the shoulder and humeroclavicular joint morphology of fossorial talpid species aligns closely with semi-aquatic locomotion , suggesting that the ancestor of Talpinae moles was semi-aquatic ( Campbell , 1939; Whidden , 1999 ) . Conversely , Reed , 1951 and Grand et al . , 1998 proposed that semi-fossorial and fossorial forms evolved from a terrestrial ancestor and that the semi-aquatic lifestyle of desmans was secondary to a semi-fossorial phase . Skeletal and fossil evidence also suggested that a period of strictly fossorial life preceded the invasion of semi-aquatic habits by the ancestors of star-nosed moles ( Condylura cristata ) ( Hickman , 1984; Grand et al . , 1998; Sanchez-Villagra et al . , 2006; Piras et al . , 2012; Hooker , 2016 ) , although Sansalone et al . , 2016 proposed that the semi-aquatic lifestyle of this genus instead represents an autapomorphy from a semi-fossorial ( shrew mole-like ) ancestor . Adaptive morphology and behavioural/physiological specialisations for diving have also been well documented for members of the family Soricidae ( Hutterer , 1985; Catania et al . , 2008; Burgin et al . , 2018 ) , although uncertainties regarding the evolution of semi-aquatic habits in this clade are also pervasive . For example , semi-aquatic members of four genera , Sorex ( North American water shrews ) , Neomys ( European water shrews ) , Chimarrogale ( Asian water shrews ) , and Nectogale ( Elegant water shrews ) show a graded series of progressively greater morphological adaptations for underwater foraging ( Hutterer , 1985 ) . However , Churchfield , 1990 suggested these genera represent four convergent aquatic invasions , with a subsequent phylogenetic analysis instead providing support for semi-aquatic habits having evolved three times ( He et al . , 2010 ) . By contrast , it has been hypothesised that a shared Late Miocene nectogaline ancestor of Neomys , Chimarrogale , and Nectogale may have been adapted to humid environments with permanent open water ( Rofes and Cuenca-Bescos , 2006 ) , thus allowing the possibility that semi-aquatic habits evolved only twice in shrews . Thus , the evolutionary pathways underlying the origins of the diverse phenotypes of eulipotyphlan mammals remain unresolved , calling for a novel approach . The monomeric oxygen-binding protein myoglobin , which plays an essential role in O2 storage and facilitates O2 diffusion within the heart and skeletal muscle of most vertebrates , provides an intriguing , objective molecular target to address this question . Indeed , the O2 storage function of myoglobin has long been known to contribute integrally to the mammalian dive response , with maximum active submergence times of birds and mammals being strongly correlated to muscle myoglobin concentrations ( Noren and Williams , 2000; Lestyk et al . , 2009; Ponganis et al . , 2011 ) . Links between the molecular evolution of this molecule and aquatic life had already been suspected for a half century or more ( Scholander , 1962; Romero-Herrera et al . , 1973 ) , leading to the speculation that ‘there might be functional reasons , perhaps associated with diving’ underlying the presence of parallel amino acid replacements in the myoglobins of seals and cetaceans ( Romero-Herrera et al . , 1978 ) . Mirceta et al . , 2013 extended this work to reveal that maximal muscle myoglobin concentration was mechanistically linked to myoglobin net surface charge ( ZMb ) in mammals via adaptive changes in primary structure , with convergent increases in ZMb found in members of all eight lineages with an extended aquatic/semi-aquatic evolutionary history . This trait presumably represents an adaptive response to combat the general propensity for proteins to precipitate at high concentration , thereby allowing for advantageous elevations of this muscle O2 store without deleterious self-aggregation ( Mirceta et al . , 2013 ) . An increase in ZMb may also reduce the aggregation of newly synthesised , unfolded myoglobin chains ( apomyoglobin ) thereby permitting efficient heme uptake and apomyoglobin folding to out compete precipitation at the high rates of translation needed to increase the cytoplasmic concentration of myoglobin while also limiting hydrophobic interactions of partially folded proteins ( Samuel et al . , 2015; Isogai et al . , 2018 ) . It should be stressed that increases in myoglobin concentration afforded by an elevated ZMb presumably represent an evolutionary trade off because there is less room for contractile units ( sarcomeres ) and organelles ( e . g . mitochondria ) in muscle tissue as more energy and space is allocated to myoglobin . Therefore , an increased ZMb only offers an advantage to species experiencing hypoxic episodes as for example occurs during breath-hold dives when access to environmental air is impeded . Importantly , the finding that ZMb and maximal myoglobin concentrations are only slightly elevated within mammals that live at high elevations or have fossorial lifestyles ( McIntyre et al . , 2002; Mirceta et al . , 2013 ) suggests that myoglobin primary structure may be useful to discern past semi-aquatic versus high-altitude/fossorial evolutionary histories within eulipotyphlan mammals . Support for the above contention is provided by the molecular modelling and phylogenetic reconstruction of ZMb in six eulipotyphlan species ( Mirceta et al . , 2013 ) , which revealed convergent elevations in ZMb in both semi-aquatic taxa included in the dataset—the American water shrew ( Sorex palustris ) and the star-nosed mole . The sparse taxon sampling in the cited study ( two moles , three shrews , and a hedgehog ) , however , did not account for the broad phylogenetic and ecomorphological diversity within these families . Additionally , current phylogenetic hypotheses for Eulipotyphla lack definitive resolution below the family level ( He et al . , 2010; He et al . , 2017 ) thereby precluding reliable ancestral reconstructions . To overcome these shortcomings , we used a capture hybridisation approach to target coding sequences of myoglobin together with 25 tree-of-life genes from 61 eulipotyphlan DNA libraries ( 44 moles , 11 shrews , five hedgehogs , and one solenodon ) that included representatives from all seven recognized semi-aquatic genera within this order . We then tested whether ZMb is elevated within members of all living genera of semi-aquatic Eulipotyphla , and if so , whether there is a significant correlation between this lifestyle and ZMb . Having shown that these important conditions are met , we then traced ZMb across eulipotyphlan evolutionary history , thereby allowing us to determine when and how many times semi-aquatic specialisations for increased dive durations evolved in both shrews and moles , and evaluate alternative evolutionary scenarios of talpid lifestyle evolution .
To obtain mammalian tree-of-life genes ( Meredith et al . , 2011 ) for phylogenetic estimation , we conducted in-solution probe-hybridisation for segments of 25 single-copy genes from 61 eulipotyphlan DNA libraries ( Supplementary file 1a ) . Twenty-three of these loci were efficiently captured , with usable sequence obtained from 51 to 61 libraries per locus , equivalent to a ~95% ( 1330/1403 ) success rate ( Supplementary file 1b ) . By contrast , probes for two gene segments only successfully hybridised to 21 and 15 of the 61 libraries , respectively , and were subsequently not included in the phylogenetic analyses . Specifically , we failed to capture the prepronociceptin ( PNOC ) gene for any shrew and many of the talpid species ( Supplementary file 1b ) . Probes for the interphotoreceptor retinoid binding protein ( IRBP ) gene also failed to retrieve any sequence from the solenodon and 37 of the 38 non-uropsiline talpid specimens . Notably , the putative 234 bp IRBP fragment recovered from True’s shrew mole ( Dymecodon pilirostris ) contained gaps and premature stop codons , suggesting this gene was inactivated in the common ancestor of the subfamily Talpinae . Similarly , the Chinese mole shrew ( Anourosorex squamipes ) IRBP sequence contained premature stop codons and was presumed to be non-functional . After incorporating orthologous tree-of-life sequence data from ten additional eulipotyphlan specimens and five outgroup species downloaded from GenBank ( see Materials and methods for details ) , our 76 specimen dataset ( 71 eulipotyphlans ) resulted in a final alignment of 39 , 414 bp ( Supplementary file 1b; Figure 2—figure supplement 1 ) . To estimate evolutionary relationships , we used maximum likelihood ( RAxML ) and Bayesian ( BEAST ) approaches on the concatenated alignment , and coalescent-based species tree methods ( ASTRAL-III , *BEAST ) on the 23 individual gene trees . These analyses resulted in highly congruent topologies ( Figure 2 and Figure 2—figure supplements 2–4 ) , with family level relationships corresponding to those obtained in recent eulipotyphlan ( Brace et al . , 2016; Springer et al . , 2018 ) and mammal wide studies ( Meredith et al . , 2011; Esselstyn et al . , 2017 ) . Higher level relationships within Soricidae and Erinaceidae also corresponded closely with previous studies on these families ( He et al . , 2010; He et al . , 2012 ) . Principal among these is the non-sister relationship of the Nectogalini water shrew clades Neomys and Chimarrogale + Nectogale , as the latter were strongly supported as sister to terrestrial Episoriculus by RAxML ( bootstrap support [ML-BS]=97; Figure 2—figure supplement 2 ) , BEAST ( posterior probability [PP]=1 . 0; Figure 2 ) , ASTRAL-III ( coalescent bootstrap support based on gene-wise resampling [C-BS]=76; Figure 2—figure supplement 3 ) , and *BEAST ( posterior probability [C-PP]=0 . 99; Figure 2—figure supplement 4 ) . By contrast , previously intractable relationships within Talpidae are now resolved , with the interrelationships among fossorial and semi-aquatic clades consistently recovered ( Figure 2 and Figure 2—figure supplements 2–4 ) . Specifically , desmans are placed sister to a clade containing shrew moles , star-nosed moles , and Eurasian fossorial moles ( BS = 100 , PP = 1 . 0 , C-BS = 87 , C-PP = 0 . 97 ) , with Condylurini and Talpini recovered as sister lineages ( BS = 72 , PP = 1 . 0 , C-BS = 52 , C-PP = 0 . 93 ) . North American fossorial Scalopini moles are also supported as sister to all other Talpinae moles with high support scores ( BS = 97 , PP = 1 . 0 , C-BS = 87 , C-PP = 0 . 97 ) . Accordingly , a sister group relationship of fully fossorial Scalopini and Talpini moles was statistically rejected by the Shimodaira–Hasegawa ( SH ) test ( p<0 . 01 , Supplementary file 1c ) . The only minor incongruences among phylogenies was the position of the tribe Soricini and of the Taiwanese brown-toothed shrew ( Episoriculus fumidus ) within Nectogalini in the ASTRAL tree ( Figure 2 and Figure 2—figure supplements 2–4 ) . Complete myoglobin coding sequences ( 465 base pairs including initiation and stop codons ) were obtained for 55 eulipotyphlan species ( 38 moles , 14 shrews , 2 erinacids , and 1 solenodon ) using capture-hybridisation , transcriptome sequencing , genome mining , and PCR approaches . Consistent with previous surveys that indicate that myoglobin occurs as a single-copy , orthologous gene in the genomes of mammals and other jawed vertebrates ( Schwarze et al . , 2014 ) , with rare , lineage-specific gene duplications being restricted to certain aquatic lineages such as some Cyprininae ( carp and goldfish; Helbo et al . , 2012 ) and Dipnoi ( lungfishes; Lüdemann et al . , 2020 ) , we found no evidence for gene paralogues in any of the species examined , with conceptual translations additionally revealing the expected 153 amino-acid peptides in most cases . However , the translated myoglobin proteins of the Chinese mole shrew , both members of the genus Scaptonyx , and the two desman genera are only 152 amino acids in length . In every case , the shorter myoglobin sequences are the result of 3 bp deletions in exon three that corresponded to residue position 121 ( Supplementary file 1d ) and shorten the loop between helices G and H from 6 to 5 residues . These deletions were confirmed for each lineage using at least two of the three sequencing approaches noted above . To our knowledge , a similar deletion was previously only known for three bird of prey species ( Enoki et al . , 2008 ) , but we have detected it also in the predicted myoglobin sequences of the draft genomes of several burrowing species of the order Rodentia ( data not shown ) , notably including the fully fossorial Transcaucasian and Northern mole voles ( Ellobius lutescens and E . talpinus , respectively; Mulugeta et al . , 2016 ) . Homology modelling of myoglobin structure was conducted for one extant species from each of the five diving lineages together with that predicted for the last common eulipotyphlan ancestor based on ancestral sequence reconstruction ( see next section ) . This analysis confirmed that all charge-changing substitutions were located at the solvent-exposed surface of the protein ( Supplementary file 1d ) . These comparisons further suggested that deletion of position 121 in the loop between helices G and H in the myoglobin of the Russian desman has minimal effect on the tertiary structure of the protein ( Figure 3—figure supplement 1 ) . In addition to their effect on ZMb , some of the charge-increasing substitutions were associated with a complex re-arrangement of the network of salt bridges in the tertiary structure of the protein . Thus , for example , during the evolution of the Russian desman , arguably the most aquatic of extant Eulipotyphla , the two neighbouring substitutions Gln26→Arg26 ( charge-increasing ) and Glu27→Asp27 ( charge-neutral ) in the B-helix allow for the formation of a new intra-helical salt bridge between Arg26 and Asp27 , while at the same time breaking the inter-helical salt bridge between Glu27 and Lys118 in the B- and G-helices , respectively ( Figure 3 ) . However , charge-decreasing Asn35→Asp35 followed by charge-increasing Gln113→Lys113 subsequently again tethered the B- and G-helices to each other by the salt bridge Asp35- Lys113 . Further , the removal of a negative charge by the substitution Asp44→Ala44 destroyed a salt bridge between Asp44 and Lys47 in the CD-corner of the protein , thereby potentially affecting the flexibility of the loop between the C- and D-helices ( Figure 3 ) . However , a detailed assessment of any changes in the folding stability of the proteins that are associated with the identified charge-changing substitutions , let alone with any charge neutral replacements ( see , e . g . Isogai et al . , 2018 ) , is difficult and beyond the scope of this study , not least because of the potentially opposing effects of salt bridge formation and the associated desolvation of charges for the folding stability of proteins ( for discussion see Bosshard et al . , 2004 ) . Without exception , the modelled ZMb values of extant semi-aquatic taxa ( 2 . 07 to 3 . 07 ) were substantially higher than those of terrestrial Eulipotyphla ( −0 . 46 to 0 . 63 ) , with fossorial species generally exhibiting intermediate ZMb values ( typically 1 . 07; Figure 4A and Figure 4—figure supplement 1 ) . To assess the reliability of our ZMb determinations , we measured the electrophoretic mobility of the myoglobin band of muscle extracts from two semi-aquatic , two strictly fossorial , and one terrestrial eulipotyphlan species ( Figure 4B ) . The close correspondence between the two variables validates ZMb as a molecular marker for inferring present and past semi-aquatic habits in Eulipotyphla . Our myoglobin nucleotide and amino acid gene trees retrieved few well-supported phylogenetic relationships ( especially at deeper nodes ) , and no compelling evidence of an evolutionary history that deviated from the species trees ( Figure 4—figure supplement 2A , B ) , which would potentially result in erroneous ancestral reconstructions ( Hahn and Nakhleh , 2016 ) . We thus conducted a maximum likelihood ancestral amino acid sequence reconstruction using the species tree in Figure 1 as the phylogenetic backbone and the best fitting model of the Dayhoff amino acid substitution matrix with a gamma distribution of rate variation among sites . This analysis yielded highly supported ancestral amino acid identities across all 153 residues and 58 internal nodes of the species tree ( Figure 5—figure supplement 1A ) , which was presumably due to our dense taxon sampling and the relatively highly conserved primary structure of mammalian myoglobins . Of the 8874 ( =58 × 153 ) reconstructed ancestral sites on the species tree , 8799 ( 99 . 15% ) had maximal probabilities of reconstructed amino acid identities of p>0 . 95 under the given phylogeny and amino acid substitution model . In 50 cases ( 0 . 55% ) , alternatively reconstructed amino acids with p>0 . 05 were of the same charge and thus did not affect the calculated ZMb values . Only in 25 cases ( 0 . 28% ) , one or more alternatively reconstructed amino acids with p>0 . 05 carried a different charge from the most probable amino acid at that site . However , even in those cases this usually had only a minimal effect on ZMb ( ±0 . 11; three sites ) , or the summed probabilities of alternative charge states were comparatively small ( p≤0 . 21; 23 sites ) such that we regard both the results of the ancestral sequence reconstruction and of the resulting values of ZMb as robust . This is supported by generally congruent results obtained by codon-based ancestral sequence reconstructions ( Figure 5—figure supplement 1B ) , which primarily deviated from the above empirical amino acid matrix-based method at some deeper nodes , where a higher number of negatively charged amino acid residues ( and thus anomalously low values for ZMb; e . g . −1 . 93 ) were reconstructed relative to those obtained for extant mammalian species ( Mirceta et al . , 2013; this study ) . Consequently , the following discussion largely focuses on the ZMb reconstruction using the empirical Dayhoff substitution matrix ( Figure 5—figure supplement 1A ) , which moreover—in contrast to the codon-based analysis ( Figure 5—figure supplement 1B ) —includes the effects of purifying selection acting on replacements with dissimilar amino acid properties . Ancestral ZMb estimates arising from the best fitting amino-acid based reconstruction model indicated that the most recent common ancestor of Eulipotyphla displayed a ZMb of 0 . 18 . Within Soricidae , distinct increases in ZMb were found on the branches leading to the three semi-aquatic clades ( Figure 5 ) . Specifically , the ZMb of the American water shrew ( Sorex palustris ) branch increased from −0 . 46 to 2 . 48 and was characterised by three integral ( +1 ) charge increasing residue replacements ( Asp122→Asn122 , Ser132→Lys132 , and Glu136→Gln136; Figure 3—figure supplement 2 ) . The other three genera of water shrews reside in the tribe Nectogalini , the stem branch of which evolved a single charge increasing substitution ( Glu27→Ala27 ) and thus had a ZMb of 1 . 07 ( Figure 5 ) . The European water shrew branch ( Neomys fodiens ) subsequently acquired two charge increasing substitutions ( Asn12→Lys12 , Asp53→Ala53; ZMb = 3 . 07; Figure 3—figure supplement 2B ) , while the common ancestor of Nectogale + Chimarrogale evolved a separate charge increasing replacement ( Asp44→Ser44; ZMb = 2 . 07; Figure 3—figure supplement 2C ) . By contrast , members of the terrestrial genus Episoriculus , which are nested between the semi-aquatic nectogaline lineages , exhibited secondary reductions in ZMb towards neutrality via different residue substitutions in each case ( Figure 5—figure supplement 1A ) . Importantly , similar results were obtained when the above amino-acid-based reconstructions were re-run on an alternative topology that supported the monophyly of Episoriculus ( Figure 5—figure supplement 1C ) . Within the Talpidae , ZMb exhibited an increase from an ancestral value of 0 . 07 to 1 . 07 ( Ser132→Lys132 ) in the stem Talpinae branch ( Figure 5 ) , with the probabilities of the identities of reconstructed amino acids at this node under the given tree and substitution model being 1 . 00 for all 153 sites ( Figure 5—figure supplement 1A ) . With few exceptions , these values remained highly conserved in the clades containing either semi-fossorial or fossorial taxa ( Figure 5 and Figure 5—figure supplement 1A , B ) . By contrast , ZMb increased to >2 in members of both semi-aquatic lineages—star-nosed moles and the desmans ( Figure 3 ) . For the star-nosed mole branch this entailed two charge increasing substitutions ( Asp53→Gly53 , Gln128→Lys128 ) and one charge decreasing substitution ( Gly129→Glu129; Figure 3—figure supplement 2D and Figure 5—figure supplement 1A ) . The ancestral branch of the desmans also acquired two charge increasing replacements ( Gln26→Arg26 , Asp44→Ala44 ) together with one charge decreasing ( Asn35→Asp35 ) replacement , with ZMb further being elevated in the Russian desman branch via the acquisition of an additional charge increasing substitution ( Gln113→Lys113; ZMb = 3 . 07; Figures 3 and 5 ) . Notably , while the same residue positions are occasionally recruited in the charge altering replacements of semi-aquatic taxa , the derived residues are different in all cases ( Supplementary file 1e ) . To evaluate the above reconstructions of semi-aquatic lifestyles based on ZMb , we coded each species as semi-aquatic or non-aquatic and estimated ancestral lifestyles using both maximum parsimony and threshold models . Although both of these latter analyses suggested that semi-aquatic lifestyles evolved five times independently , this result was not strongly supported by the threshold model ( Figure 5—figure supplement 2A ) . For example , the posterior probabilities of the most recent common ancestor of Desmana + Galemys and Chimarrogale + Nectogale being semi-aquatic was only 0 . 85 and 0 . 75 , respectively , while the posterior probability for the most recent common ancestor of Nectogalini being semi-aquatic was 0 . 40 . To account for the alternative placement of E . fumidus , we repeated this analysis using the results of the *BEAST species tree ( Figure 2—figure supplement 4 ) . The maximum parsimony reconstruction yielded two equi-parsimonious ancestral reconstructions of semi-aquatic lifestyle in nectogaline shrews , encompassing a single origin at the base of nectogaline shrews with a secondary loss at the base of the Episoriculus clade ( Figure 5—figure supplement 2B ) . By contrast , the threshold model only weakly supported two independent origins of a semi-aquatic lifestyle in Neomys and Nectogale + Chimarrogale . As a final test regarding the reliability of using ZMb to predict ancient semi-aquatic lifestyles , we first assigned species as semi-aquatic or non-aquatic ( see Figure 2 and Supplementary file 1a for lifestyle assignments ) , and used a threshBayes analysis to estimate covariances between ZMb and a semi-aquatic lifestyle . This analysis revealed a strong correlation between ZMb and aquatic adaptation ( correlation coefficient r = 0 . 78 , 95% highest posterior density [HPD]=0 . 48–0 . 93; Figure 5—figure supplement 3A ) . Conversely , threshBayes analyses did not support a correlation between ZMb and adaptations for digging , or between ZMb and a fully fossorial habit , when terrestrial and semi-aquatic eulipotyphlan species were included ( Figure 5—figure supplement 3B , C ) . When applying threshBayes to subsets of habits that included only terrestrial and fully fossorial species , or terrestrial and burrowing species , weak correlations between ZMb and fossoriality ( r = 0 . 55 ) and digging habits ( r = 0 . 32 ) were revealed , but not significantly supported ( i . e . 95%HPD overlaps with 0; Figure 5—figure supplement 3D , E ) . Importantly , ZMb comparisons between semi-aquatic and terrestrial , and between semi-aquatic and fossorial habits ( Figure 5—figure supplement 3F , G ) were both significant .
The phylogenetic estimates constructed from our comprehensive tree-of-life gene set provide a robust framework to interpret ecomorphological evolution within Eulipotyphla . The close correspondence of our concatenation and coalescent phylogenetic topologies not only support key findings of previous studies—that is , the monophyly of shrew moles , the non-monophyly of nectogaline water shrews ( Whidden , 2000; He et al . , 2010; He et al . , 2017 ) —but finally puts to rest the long hypothesised monophyletic origin of the fully fossorial tribes Talpini and Scalopini , which have been routinely grouped together based on morphological data ( see , e . g . Whidden , 2000; Piras et al . , 2012; Schwermann and Thompson , 2015; Hooker , 2016; Sansalone et al . , 2019 ) . Results of the present study thus provide compelling evidence that the extreme anatomical specialisations for subterranean life evolved convergently in these two tribes . The molecular phylogenetic position of the amphibious desmans and semi-aquatic/fossorial star-nosed mole are also finally resolved ( Figure 2 ) , with the latter placed sister to Talpini , a relationship not supported by morphological-based hypotheses ( see , e . g . Whidden , 2000; Motokawa , 1999; Sanchez-Villagra et al . , 2006 ) . Previous studies have failed to reach consensus on the lifestyle evolution of Eulipotyphla . Here , we show that ancestral sequence reconstruction of myoglobin primary structure and ZMb modelling , with their well established mechanistic and biophysical underpinnings , outperform discrete , two character-state lifestyle reconstructions based on maximum parsimony and threshold approaches . These attributes , together with demonstrated links between ZMb and maximal Mb concentration , and hence muscle oxygen storage capacity ( Mirceta et al . , 2013; Berenbrink , 2021 ) , provide strong support that this quantitative metric is well suited to resolve long-standing questions regarding lifestyle evolution within Eulipotyphla . For example , despite being less specialised for aquatic life than marine mammals , the strong positive correlation between ZMb and semi-aquatic specialisation indicates that ZMb is a powerful marker to identify secondary aquatic transitions in even the world’s smallest mammalian divers . Although a strong correlation was not recovered between ZMb and strictly fossorial habitation , ZMb is highly conserved in fossorial Scalopini and Talpini , with 23 of 25 species exhibiting a value of 1 . 07 . This conservation is presumably driven by selective pressures to maintain moderately elevated tissue myoglobin levels to help foster burst burrowing activities in their hypoxic underground environment . By contrast , the ZMb of terrestrial species in our dataset were consistently close to neutrality , although many lineages exhibited clear signals of ZMb fluctuation over time ( Figure 5—figure supplement 1 ) . For example , the shrew gymnure ( Neotetracus sinensis ) branch evolved nine charge altering residue substitutions since its split from European hedgehogs ( Erinaceus europaeus ) , including the charge inversion Asp126→Lys126 that increases ZMb by +2 . Similarly , the Taiwanese brown-toothed shrew branch ( Episoriculus fumidus ) has fourteen charge altering residue substitutions , including one negative-to-positive ( Glu109→Lys109 ) and two positive-to-negative charge inversions ( Lys102→Glu102 and Lys132→Glu132 ) . These observations are consistent with a stochastic evolutionary process operating under purifying selection . Charge fluctuation is also apparent on the Hispaniolan solenodon ( Solenodon paradoxus ) branch , with the moderately elevated ZMb value ( 1 . 85 ) in line with their terrestrial/burrowing lifestyle ( Nowak , 1999 ) . Myoglobin sequence data from additional extant/extinct members of this family ( e . g . Cuban solenodons ) together with that from the recently extinct terrestrial/fossorial Antillean family Nesophontidae ( MacPhee et al . , 1999 ) —which is placed sister to solenodons ( Brace et al . , 2016 ) —may help resolve the life history evolution of this poorly understood insectivore clade . Similar , albeit less pronounced charge fluctuations are evident in several fossorial mole branches ( as well as the stem Condylura and desman branches ) , suggesting that the single fossorial ZMb outlier ( hairy-tailed mole , Parascalops breweri; ZMb = 2 . 07 ) may also represent stochastic variation that has not yet been selected against and that presumably will return to an ecologically normalised value over evolutionary time . Our results indicate that charge increases in ZMb to >2 are essential for members of eulipotyphlan mammals to successfully exploit a semi-aquatic lifestyle . Because ZMb of the most recent common ancestor of Talpinae was confidently estimated to be below this value ( Figure 5 ) , our results do not support the ‘aquatic mole’ hypothesis which posits that a semi-aquatic stage predated the invasion of fossorial habits in talpid moles ( Campbell , 1939; Whidden , 1999 ) . This conclusion is supported by the results of our lifestyle reconstructions , which uniformly revealed that stem Talpinae had a low probability of semi-aquatic habits . These findings are instead consistent with the interpretation that fossorial forms evolved directly from terrestrial/semi-fossorial ancestors , without passing through a semi-aquatic phase ( Reed , 1951; Hickman , 1984; Grand et al . , 1998 ) . The interpretation of a semi-fossorial ancestry for early non-uropsiline ( Talpinae ) moles is further supported by our finding of a presumed inactivation/deletion of IRBP on this branch ( Supplementary file 1b ) . This eye-specific locus has been shown to be inactivated/lost in numerous mammalian lineages that inhabit subterranean and other dim-light niches ( Emerling and Springer , 2014 ) , and also appears to be non-functional in solendons and the Chinese mole shrew ( this study ) . The ZMb charge increase ( to 2 . 07 ) in the stem desman branch , paired with a lack of pronounced forelimb specialisations for digging ( He et al . , 2017 ) and a basal placement among non-scalopine Talpinae ( Figure 2 ) , also chimes with Reed , 1951 suggestion that the semi-aquatic lifestyle of this clade was secondary to a semi-fossorial phase . By contrast , semi-aquatic/fossorial star-nosed mole exhibits prominent morphological adaptations for burrowing and is placed sister to fossorial Talpini , consistent with Grand et al . , 1998 hypothesis that this lineage passed through a specialised fossorial stage prior to invasion of the semi-aquatic niche . However , a semi-fossorial—as opposed to a fully fossorial—ancestry for Condylura ( Sansalone et al . , 2016 ) cannot be excluded based on the pattern of ZMb evolution on this branch . The charge elevation to 1 . 07 in stem Nectogalini shrews is enticing , as it temporally corresponds to the early Miocene fossil Asoriculus , which was theorised to have inhabited wet environments though was unlikely to have been an efficient semi-aquatic predator ( Rofes and Cuenca-Bescos , 2006 ) . The single charge increasing substitution that evolved at this stage ( Asp27→Ala27 ) is retained in the three semi-aquatic nectogaline genera and presumably facilitated the adaptation of early Neomys and Chimarrogale + Nectogale for aquatic food resources starting some 15 and 10 million years ago , respectively . The presence of separate charge increasing replacements within the evolutionary branches of each of the three Episoriculus species also opens the possibility of additional semi-aquatic ‘experiments’ in this genus . Regardless , independent reductions in ZMb to neutrality or below for each of the latter extant species , which today inhabit damp areas in vegetated environments ( Nowak , 1999 ) , is consistent with convergent reversions of Episoriculus lineages to a predominantly terrestrial foraging habit . Our results provide strong evidence that the exploitation of semi-aquatic habits by extant shrews and talpids occurred at least five times , and was accompanied by convergently evolved charge-increasing substitutions at different surface sites on the myoglobin protein in each case ( Figure 3 , Figure 3—figure supplement 1 , Figure 5 , Figure 5—figure supplement 1 ) . This finding provides additional support for our contention that adaptive increases in ZMb underlie the invasion of ( semi- ) aquatic niches by mammals , presumably by allowing for higher skeletal muscle myoglobin concentration ( Mirceta et al . , 2013; Berenbrink , 2021 ) . This elevated ZMb presumably underlies the elevated O2 reserves in muscle of star-nosed moles and American water shrews compared to non-aquatic relatives ( McIntyre et al . , 2002; Gusztak , 2008 ) , and likely contributes to the extended dive times and remarkable underwater foraging efficiency of these species ( Catania et al . , 2008 ) . The increase in ZMb must be particularly important for semi-aquatic soricine shrews due to allometric considerations that have resulted in extremely high muscle mitochondrial contents ( which may comprise up to 45% of the cell volume; Weibel , 1985 ) and mass-specific tissue O2 requirements that may be >100 fold higher than those of large-bodied marine mammals ( Butler , 1998; Gusztak et al . , 2005 ) . It is notable that the highest ZMb values were found for Russian desmans ( 3 . 07 ) and European water shrews ( 3 . 07 ) , consistent with the exceptional diving abilities of these species ( Vogel et al . , 2007; Ivlev et al . , 2010 ) , and it is predicted that these two species will also possess the highest muscle myoglobin concentrations among Eulipotyphla . Adaptive evolution of similar phenotypic and physiological features occurring in distantly related lineages are not uncommon in mammals ( Madsen et al . , 2001 ) . For example , adaptive radiations in Afrotheria and Laurasiatheria resulted in striking morphological convergence of species occupying semi-aquatic ( otter shrews vs . desmans ) and subterranean ( golden moles vs . true moles ) habitats ( Madsen et al . , 2001; Springer et al . , 2004 ) . A unifying pattern underlying these and most other large-scale mammalian radiations over the past 200 million years is that they all involved ecological and locomotory diversification from ancestral lineages of small insectivores ( Grossnickle et al . , 2019 ) . The extensive radiation of small terrestrial Eulipotyphla into different adaptive zones , including four independent origins of venom systems in shrews and solenodons ( Casewell et al . , 2019 ) and multiple independent invasions of shrews and moles to semi-aquatic , semi-fossorial , and subterranean environments that occurred on shallow timescales of only a few million years , further demonstrates the high intrinsic evolutionary potential of this Bauplan . Morphological , physiological , and even behavioral convergence have previously been identified within semi-aquatic eulipotyphlan species . For example , D . moschatus , C . cristata , and S . palustris can all detect prey scent while under water via the rapid exhalation and inhalation of air bubbles ( Catania , 2006; Catania et al . , 2008; Ivlev et al . , 2013 ) , with the latter two species also being characterised by an elevated proton-buffering capacity in muscle ( McIntyre et al . , 2002; Gusztak , 2008 ) . The results presented here add to this list of convergences , and indicate that semi-aquatic eulipotyphlans have evolved similar ZMb ( and presumably elevated myoglobin concentration ) phenotypes via the same selection pressure acting on different sites of the protein and by dissimilar combinations of amino acid substitutions ( i . e . differential gains and losses of cationic and anionic residues , respectively ) . In other words , molecular adaptation of myoglobin towards life in a semi-aquatic environment is predictable at the protein level but underpinned by unpredictable genotypic evolution . As such , the phylogenomic analysis of myoglobin loci from tissue samples is not only able to provide insights into the lifestyles of rare and recently extinct mammalian species ( e . g . museum specimens and subfossil material from the obscure Caribbean nesophontids ) , but also offers a useful tool to infer past semi-aquatic transitions based on myoglobin primary structure alone .
Our taxon sampling of eulipotyphlan mammals included 44 talpids , 11 shrews , 5 erinaceids , and 1 solenodon ( 61 specimens encompassing 60 species ) . Note that this sampling incorporates talpid specimens from five putative ‘cryptic lineages’ ( denoted by ‘sp . ’ , ‘sp . 1’ , or ‘sp . 2’ in the Figures and Supplementary file 1 ) ; for the purpose of this study , each of these genetically distinct lineages are considered independent species . The tissue samples were from various resources ( Supplementary file 1a ) , with most tissue samples provided by co-authors from China , Japan , Canada , and the USA . Voucher specimens collected by co-authors were deposited in the Kunming Institute of Zoology ( KIZ , China ) , the National Museum of Nature and Science ( NMNS , Japan ) or kept in personal collections ( A . S . and S . I . K . ) . Additional tissue samples were obtained with permission from the National Museum of Natural History ( USNM , USA ) , the Burke Museum of Natural History and Culture ( NWBM , USA ) , the Field Museum of Natural History ( FMNH , USA ) , and the New Mexico Museum of Natural History ( NMMNH , USA ) . For each specimen , we used a capture hybridisation approach ( Mason et al . , 2011; Horn , 2012 ) to enrich myoglobin exons and segments of 25 mammalian tree-of-life genes ( Meredith et al . , 2011 ) for phylogenetic analyses . We first downloaded tree-of-life sequences from three eulipotyphlan whole genome sequences available in GenBank ( Erinaceus europaeus , Sorex araneus , Condylura cristata ) , together with 60 bp of 5’- and 3’- flanking sequence for each target . We then aligned each gene segment using MAFFT ( Katoh and Standley , 2013 ) . The resulting alignments were used to design 120 mer RNA probes ( baits ) that overlapped by 90 bp ( 4x tiling ) , and collapsed any replicates with up to six mismatches ( 95% similarity ) for each segment . For example , if the 120 bp gene fragments from all species were 95% similar with each other , only one probe was designed for this region , otherwise two or more probes were designed to cover the heterogeneity . The myBaits probes were synthesised by Arbor Biosciences ( Ann Arbor , MI , USA ) . As a first step in DNA library construction we extracted total DNA from each specimen using a Qiagen DNeasy Blood and Tissue Kit ( Qiagen , Canada ) . The quality and quantity of each DNA sample was measured using a Nanodrop 2000 . We then sheared the total DNA into smaller fragments using NEBNext dsDNA Fragmentase ( New England Biolabs , Canada ) , and used this as template to construct DNA libraries using a NEBNext Fast DNA Library Prep Set for Ion Torrent kit ( New England Biolabs , Canada ) . Each sample library contained a unique barcode adapter ( NEXTflex DNA Barcodes for Ion Torrent , BIOO Scientific , USA ) . We selected libraries within the size range of 450–500 bp using a 2% E-gel on an E-Gel Electrophoresis System ( Invitrogen , Canada ) , and re-amplified the size-selected libraries using a NEBNext High-Fidelity 2X PCR Master Mix ( New England Biolabs , Canada ) . Finally , we purified the libraries using Serapure magnetic beads , and measured DNA concentrations using a Qubit 2 Fluorometer ( Thermo Fisher Scientific , Canada ) . We pooled up to four DNA libraries of similar quality and concentrations before hybridisation to avoid biased target captures ( e . g . baits being used up by one sample ) . Approximately 500 ng ( 100–1000 ng ) pooled DNA library was used for each hybridisation . We conducted in-solution hybridisation using a myBaits custom target capture kit ( Arbor Biosciences , Ann Arbor , MI , USA ) following the myBaits user manual v3 . 0 . The enriched libraries were re-amplified and purified as above . We thereafter measured the DNA concentration using a Qubit flourometer and pooled the enriched libraries for sequencing . The libraries were sequenced using either v318 chips on an Ion Torrent Personal Genome Machine ( PGM ) or an Ion PI Chip v3 via an Ion Proton Machine . Ion Torrent sequencing technology is characterised by higher error rates than Illumina ( Jünemann et al . , 2013 ) , and Ion Torrent platforms produce single-end ( rather than pair-end ) reads . We therefore conducted comprehensive data cleaning and reconciliation procedures , and selected software which could handle single-end sequencing data . The raw data were automatically demultiplexed , trimmed , and converted to FASTQ format on the Torrent Suite v4 . 0 . 2 ( Thermo Fisher Scientific , Canada ) after sequencing . Briefly , we trimmed contaminant ( adapters and barcodes ) sequences with AlienTrimmer ( Criscuolo and Brisse , 2013 ) using conservative parameters ( -k 15 m 5 l 15 -q 0 p 0 ) . To remove poor quality data , we used the DynamicTrim function of the software SolexaQA ++v3 . 1 ( Cox et al . , 2010 ) to trim sequences dynamically and crop the longest contiguous segment for each read . We set the probability value to 0 . 01 ( i . e . one base call error every 100 nucleotides ) in this analysis . We removed duplicated and near-duplicated reads for each sample as implemented in ParDRe using all default parameters ( González-Domínguez and Schmidt , 2016 ) . Finally , we conducted data correction using Karect , a multiple sequence alignment-based approach ( Allam et al . , 2015 ) , because this software handles substitution , insertion , and deletion errors . The output files of Karect were used for sequence assembly . We de novo assembled the raw sequences for each sample using Abyss v2 . 0 ( Simpson et al . , 2009 ) , MIRA v4 . 0 ( Chevreux et al . , 2004 ) , and SPAdes v3 . 10 ( Bankevich et al . , 2012 ) , all of which were designed for short read sequencing data . Abyss is able to use a paired de Bruijn graph instead of a standard de Bruijn graph by specifying a k-mer size ( K ) and a k-mer pair span ( k ) . We set the K and k to 17 and 33 , respectively , and set the maximum number of branches of a bubble to five in our analyses . MIRA is based on a Smith-Waterman algorithm . We ran MIRA using specific parameters including bases_per_hash = 31 and minimum_read_length = 35 . The SPAdes assembler is also based on a de Bruijn graph , and we set only one k-mer value of 33 for analyses . It is known that merging different draft assemblies ( i . e . reconciliation ) could improve the assembly quality ( Zimin et al . , 2008 ) . We therefore conducted reconciliation using Geneious R11 ( https://www . geneious . com ) . We concatenated the assembled draft contigs generated in three assemblers into a list . We removed contigs shorter than 120 bps , and used the BBMap dedupe function to remove duplicate contigs . We conducted assemblies using the Geneious assembler to group draft contigs with a minimum overlap identity of 96% to a new contig . Finally , all the new contigs and the leftover draft contigs were grouped into a contig list for subsequent analyses . We used four strategies to obtain myoglobin coding sequences ( Supplementary file 1a ) . We first extracted available eulipotyphlan myoglobin mRNA and gene sequences from GenBank . The three coding exons were individually used both as templates for capture hybridisation probe design ( see above ) and to map the hybridisation contigs/generate consensus sequences for each exon . The 5’- and 3’- ends of introns were confirmed based on the GT-AG splice site rule . Of the 61 samples that we used for hybridisation capture , complete coding sequence was obtained for 27 samples , partial myoglobin sequences were obtained for 30 samples , and no sequence obtained for four samples ( Supplementary file 1a ) . To cross-validate the results of our hybridisation experiments , fill sequencing gaps , and extend our taxon sampling , we also PCR amplified and Sanger sequenced whole myoglobin exons from existing DNA samples and/or employed transcriptome sequencing on additional eulipotyphlan specimens . For the latter , we collected heart and lung samples from five shrews , one shrew-like mole , and one gymnure ( Supplementary file 1a ) . Tissues were preserved in RNAlater ( Qiagen , China ) , and stored at −80C . Total RNA was extracted using a RNeasy Mini kit ( Qiagen , China ) , and mRNA subsequently enriched using immobilised oligo ( dT ) . mRNA was sheared and reverse transcribed to DNA . The cDNA libraries were purified and re-amplified using PCR for de novo sequencing using a HiSeq X Ten Sequencing System . Approximately 6 Gb data were obtained for each sample . Experiments and sequencing were conducted by BioMarker Co . ( Beijing , China ) . We used FastQC v0 . 11 . 5 ( Andrews , 2010 ) to access sequence quality , and trimmed adapter sequences using Trimmomatic v0 . 39 ( Bolger et al . , 2014 ) . We conducted de novo assembly using Trinity v2 . 4 with default parameters ( Grabherr et al . , 2011 ) . Finally , primers for PCR were designed for conserved exon flanking regions from available eulipotyphlan genomes , hybridisation capture , and mRNA sequences , and were used for both PCR and Sanger sequencing ( Supplementary file 1j ) . These procedures resulted in complete coding sequences being obtained for 55 eulipotyphlan species ( Supplementary file 1d ) . We additionally extracted 25 mammalian tree-of-life gene segments from seven publicly available eulipotyphlan genomes on GenBank using PHYLUCE ( Faircloth , 2016 ) and Geneious R11: Indochinese shrew ( Crocidura indochinensis ) , gracile shrew-like mole ( Uropsilus nivatus ) , Eastern mole ( Scalopus aquaticus ) , Hispaniolan solenodon ( Solenodon paradoxus ) , European hedgehog ( Erinaceus europaeus ) , common shrew ( Sorex araneus ) , and the star-nosed mole ( Condylura cristata ) . Corresponding sequences from five outgroup taxa were also mined: guinea pig ( Cavia porcellus ) , horse ( Equus caballus ) , cat ( Felis catus ) , pig ( Sus scrofa ) , and bat ( Pteropus alecto ) . PHYLUCE was originally developed for ultra-conserved elements ( UCE ) . We followed the ‘harvesting UCE loci from genomes’ protocol , but used the tree-of-life reference genes as probes instead of the original UCE probe sets . We extracted genomic regions which were at least 75% similar to the tree-of-life reference sequences . We also mapped the genomes to the tree-of-life references using Geneious with a minimum overlap identity of 75% . These two packages were generally equally efficient at capturing target genes , although in a few cases only one successfully captured the target genes from the genome . We used the above eulipotyphlan myoglobin and tree-of-life gene sequences to generate consensus sequences for the 61 specimens employed for the hybridisation capture experiments ( Supplementary file 1b ) . Briefly , the GenBank sequences were used as reference scaffolds to individually map the Ion Torrent generated reads of each sample using the Geneious ‘Map to Reference’ function , and allowing for a mismatch of 35% per contig . This package conducts iterative mapping and outperforms many other algorithms by higher mapping rates and better consensus accuracy ( Kearse et al . , 2012 ) . Approximately 1–4 contigs from each sample were mapped to each gene reference . For the TTN gene segment , whose reference sequence was 4452 bp in length , as many as 10 contigs from each sample could be mapped to the reference . In addition , sequences of 19 nuclear gene segments obtained from 21 eulipotyphlan samples collected as part of previous studies ( He et al . , 2014; He et al . , 2017 ) , were also used for assemblies as above ( and included in the final assemblies ) . Three shrew species ( Episoriculus umbrinus , Episoriculus fumidus , and Sorex bedfordiae ) for which we obtained myoglobin coding sequences via transcriptome sequencing ( see below ) were not included in our hybridisation capture experiments . We thus downloaded the available tree-of-life genes from these species ( APOB , BRCA1 , and RAG2 ) on GenBank and included them in our analysis . The resulting 25 tree-of-life gene segments were aligned separately using MAFFT . We then removed sequences shorter than 247 bp , and estimated gene trees using FastTree v2 . 1 . 5 to check for potential paralogues for each gene . We also checked each alignment by eye , and removed ambiguous regions from the alignment . As segments of two genes ( IRBP , PNOC ) failed to hybridise to the majority of the DNA libraries , they were removed from subsequent analysis . The final 23 gene segment alignment ( 39 , 414 bp ) included sequences from 76 samples ( Figure 2—figure supplement 1 ) . Because ASTRAL assumes no intra-locus recombination , we also tested recombination events using both RDP and GENECONV methods to examine each gene as implemented in RDP v5 . 5 ( Martin et al . , 2015 ) . When the program detected a signal of recombination , we checked UPGMA trees estimated using the two non-overlapping fragments and also examined the original alignment to see whether the signal was likely due to recombination or other evolutionary processes . We did not observe strong evidence of cross-species recombination in gene alignments ( data not shown ) . We estimated evolutionary relationships using several different approaches . We first constructed a summary-coalescent tree whereby we simultaneously conducted rapid bootstrap analyses and searched for the best scoring maximum likelihood tree using RAXML v8 . 2 ( Stamatakis , 2014 ) for each gene alignment , allowing the program to determine the number of bootstraps ( -#autoMRE ) . We used Cavia porcellus as the root of the tree employing GTR+ γ during both ML searches and bootstrapping phases , and disabled the BFGS searching algorithm for optimizing branch lengths and GTR parameters ( --no-bfgs ) . We followed Simmons and Kessenich , 2020 recommendation to remove dubiously supported clades and increase accuracy of tree estimation . We used RAxML to estimate SH-like aLRT support values for the best-scoring gene trees ( -f j; Supplementary file 1g ) . Then we collapsed branches whose SH-like aLRT support values equal zero using Newick utilities and TreeGraph 2 ( Junier and Zdobnov , 2010; Stöver and Müller , 2010 ) . Then we used the collapsed trees to estimate the coalescent species tree using ASTRAL III v5 . 15 . 0 ( Zhang et al . , 2018 ) . Because gene-wise bootstrap could provide more conservative support than site-wise bootstrap analyses ( Simmons et al . , 2019 ) , we conducted gene-wise bootstrapping ( --gene-only ) instead of the typical site-wise bootstrapping . We also allowed the program to explore a larger search space by adding extra bipartitions to the search space ( --extraLevel 2 ) . In addition to the summary coalescent analyses , we also constructed a concatenation-based tree using the same dataset . We partitioned the alignment by gene , searched for the best-scoring tree and conducted rapid bootstrapping under the GTR+ γ model using RAxML as described above . We used seven calibrations ( Supplementary file 1h ) and BEAST v2 . 5 ( Bouckaert et al . , 2014 ) to estimate divergence times for all 76 samples as well as for the 60 species ( 55 eulipotyphlans and five outgroup species ) for which complete myoglobin coding sequences were obtained . For this analysis , we first partitioned the alignment by gene and used the bModelTest package of BEAST2 to estimate the most appropriate substitution model for each gene ( Bouckaert and Drummond , 2017 ) . We used a relaxed clock model with lognormal distribution for estimating the branch lengths , a birth-death model for the prior of the tree , and ran the analysis for 100 million generations . We used Cavia porcellus as the outgroup to Laurasiatheria , and also fixed the relationships of the other four outgroup species , because a biased sampling toward the ingroup ( i . e . Eulipotyphla ) may lead to an inaccurate estimation of outgroup relationships ( Springer et al . , 2018 ) . Secondly , we used the BModelAnalyzer package of BEAST2 to determine the best models for each gene based on the results of the bModelTest ( Supplementary file 1f; Barido-Sottani et al . , 2018 ) . We fixed the models of evolution based on the results of BModelAnalyzer and re-ran BEAST using the same parameters described above . Finally , we employed a multispecies coalescent model ( *BEAST; Heled and Drummond , 2010 ) as implemented in BEAST v2 . 5 . We grouped the samples by species . We used linear and constant root as the prior for the population model . Substitution model , tree model , and calibrations were set as above . All gene trees and species trees are given in Supplementary file 1i . To examine whether alternative evolutionary hypotheses of life histories ( e . g . a single origination of fully fossorial lifestyle within Talpidae ) could be statistically rejected , we performed Shimodaira–Hasegawa ( SH ) tests . For these analyses , we constrained the monophyletic relationships of: ( i ) fully fossorial Talpini and Scalopini moles , ( ii ) semi-aquatic desmans and the star-nosed mole , and ( iii ) semi-aquatic nectogaline shrew genera ( Chimarrogale , Nectogale , and Neomys ) , one at a time and estimated the maximum likelihood concatenation trees using RAxML as described above ( Supplementary file 1c ) . Then we computed the log likelihood between the best scoring maximum likelihood tree and the constrained alternative phylogenies as implemented in RAxML ( -f H ) . We estimated ancestral myoglobin sequences for each node of a 60 species phylogeny that utilised both DNA sequences and amino acids . For comparison , we also estimated myoglobin gene trees utilizing both nucleotide and amino acid sequences as implemented in RAxML . We used Dayhoff+ γ model ( Dayhoff et al . , 1978 ) for the amino acid gene tree estimation using the same settings described above . Prior to analysis , the start ( methionine ) and stop codons were removed from the alignment . As in our previous study ( Mirceta et al . , 2013 ) , we performed maximum likelihood ancestral amino acid sequence reconstruction as implemented in MEGA ( Kumar et al . , 2018 ) using the Dayhoff+ γ model that was obtained as the best-fitting substitution model using the model test function in MEGA-X . Prior to conducting the codon-based analysis , we removed codons corresponding to residue position 121 ( which was absent for 5 of the 60 species; Supplementary file 1d ) . We then used the PAML package CodeML ( Yang , 2007 ) as implemented in EasyCodeML ( Gao et al . , 2019 ) , and compared codon substitution models ( site models ) including M0 , M1a , M2a , M3 , M7 , M8 , and M8a using likelihood-ratio tests . We relied on the model with the highest likelihood ( M8a ) . Because PAML does not take account of insertion/deletion events ( indels ) , and instead treats gaps as missing data , the ancestral states of the gapped codon position 121 was reconstructed separately using a likelihood-based mixture model as implemented in FastML ( Ashkenazy et al . , 2012 ) . To assess the three-dimensional location and any secondary or tertiary structural implications of amino acid replacements or insertions/deletions , we used the fully automated homology modelling facilities of the SWISS-MODEL server ( Waterhouse et al . , 2018 ) to build protein structural models from the reconstructed ancestral primary structure of myoglobin in the last common eulipotyphlan ancestor and from the primary structure of the sequenced myoglobins of one species of each of the five semiaquatic lineages . Implications of the gapped position 121 on the tertiary structure of myoglobin in the Russian desman compared to the last eulipotyphlan ancestor were visualised in PyMol ( The PyMOL Molecular Graphics System , Version 2 . 1 . 1 , Schrödinger , LLC ) . We calculated ZMb as the sum of the charge of all ionizable groups in myoglobin at pH = 6 . 5 by modelling Mb primary structures onto the tertiary structure and using published , conserved , site-specific ionisation constants ( McLellan , 1984; Mirceta et al . , 2013 ) . The reliability of modelled ZMb values was assessed by determining the electrophoretic mobility of native myoglobin bands at the same pH in muscle extracts of representative eulipotyphlan species and the grey seal , Halichoerus grypus , as an example of a marine mammal . Approximately 0 . 2 g of skeletal or cardiac muscle tissue from selected species , freed from any obvious fat or connective tissue remnants and rinsed with homogenisation buffer to move any remaining blood , was homogenised in 5 volumes of ice-cold 0 . 2 M MES buffer [2- ( N-morpholino ) ethanesulfonic acid] adjusted to pH 6 . 5 , using an Ultra-turrax T25 homogeniser for 10 s at first 9500 rpm and then three times at 13 , 500 rpm , leaving samples to cool down between steps for 1 min on ice to avoid heat denaturation of proteins . The homogenised muscle extracts were then centrifuged at 10 , 500 g ( 20 min at 4°C ) and the supernatants stored at −80°C until further use . Electrophoretic mobility of thawed muscle extracts was assessed in 9% polyacrylamide gels containing 0 . 3 M MES buffer pH 6 . 5 , using a Bio-Rad Mini-PROTEAN II gel system with 0 . 2 M MES pH 6 . 5 as the running buffer at 100 V and room temperature for a minimum of 3 hr . Native myoglobin bands were identified by their distinct red-brown color before general protein staining with EZBlue ( G104 , Sigma-Aldrich ) . Electrophoretic mobility was assessed on digital gel images and expressed as distance travelled relative to the grey seal myoglobin , which was used as a standard of a marine mammal myoglobin with high net surface charge ( Mirceta et al . , 2013 ) . The correlation between measured relative electrophoretic mobility and modelled ZMb values was assessed using Phylogenetic Generalised Least Squares ( PGLS ) analysis using the CAPER package ( Orme et al . , 2013 ) as implemented in R v3 . 6 and the tree from the BEAST analysis in Figure 2 . Because of low sample size , the parameter lambda was not estimated from the data but fixed at a value of 1 . 0 . We analyzed the relationship between lifestyles and ZMb based on a threshold model ( Felsenstein , 2012 ) using the phytools function threshBayes ( Revell , 2012 ) as implemented in R v3 . 6 . The threshold model hypothesises that each lifestyle is determined by an underlying , unobserved continuous trait ( i . e . liability ) . We first categorised the 55 eulipotyphlan species for which ZMb was calculated as either semi-aquatic ( including the semi-aquatic/fossorial star-nosed mole in this category ) or non-aquatic based on the habits described in Burgin and He , 2018 . We ran Markov chain Monte Carlo ( MCMC ) for 107 generations , sampling every 500 generations , and discarded the first 20% generations as burn-in . We plotted the posterior sample for the correlation to examine whether analyses reached a stationary state . We also estimated the correlation between ZMb and full fossoriality ( i . e . fossorial species versus non-fossorial species ) , as well as that between ZMb and ‘digging’ ( a category that included both fossorial and semi-fossorial species ) habits using the same approach . Finally , we also created subsets of our dataset to enable comparisons between only two ecomorphotypes , with the following four threshBayes analyses conducted: terrestrial ZMb versus semi-aquatic ZMb , terrestrial ZMb versus fully fossorial ZMb , terrestrial ZMb versus semi-fossorial/fossorial ZMb , fully fossorial ZMb versus semi-aquatic ZMb . We estimated the ancestral lifestyle using a maximum parsimony and a threshold model based on the 76 species time-calibrated concatenated gene tree ( Figure 2 ) and the *BEAST coalescent species tree ( Figure 2—figure supplement 4 ) . We categorised the species into non-aquatic or semi-aquatic as above . We used the R package castor to reconstruct ancestral lifestyles using maximum parsimony ( Louca and Doebeli , 2018 ) , treating the transition cost between non-aquatic and semi-aquatic equally . We then used phytools to estimate ancestral states using a threshold model ( Revell , 2014 ) . We ran MCMC for 1 million generations , sampling every 1000 generations , and discarded the first 20% generations as burn-in . We performed the threshold analyses using either a Brownian motion or lambda model for estimating the liability , and compared the results based on deviance information criterion ( DIC ) . We selected the result of the lambda model because it outperformed the Brownian motion model ( ΔDIC = 99 , data not shown ) . | The shrews , moles and hedgehogs that surround us all belong to the same large group of insect-eating mammals . While most members in this ‘Eulipotyphla order’ trot on land , some , like moles , have evolved to hunt their prey underground . A few species , such as the water shrews , have even ventured to adopt a semi-aquatic lifestyle , diving into ponds and streams to retrieve insects . These underwater foragers share unique challenges , burning a lot of energy and losing heat at a high rate while not being able to store much oxygen . It is still unclear how these semi-aquatic habits have come to be: the fossil record is fragmented and several species tend to display the same adaptations even though they have evolved separately . This makes it difficult to identify when and how many times the Eulipotyphla species started to inhabit water . The protein myoglobin , which gives muscles their red color , could help in this effort . This molecule helps muscles to capture oxygen from blood , a necessary step for cells to obtain energy . Penguins , seals and whales , which dive to get their food , often have much higher concentration of myoglobin so they can spend extended amount of time without having to surface for air . In addition , previous work has shown that eight groups of mammalian divers carry genetic changes that help newly synthetized myoglobin proteins to not stick to each other . This means that these animals can store more of the molecule in their muscles , increasing their oxygen intake and delivery . He et al . therefore speculated that all semi-aquatic Eulipotyphla species would carry genetic changes that made their myoglobin less likely to clump together; underground species , which also benefit from absorbing more oxygen , would display intermediate alterations . In addition , reconstructing the myoglobin sequences from the ancestors of living species would help to spot when the transition to aquatic life took place . A variety of approaches were harnessed to obtain myoglobin and other sequences from 55 eulipotyphlan mammals , which then were used to construct a strongly supported family tree for this group . The myoglobin results revealed that from terrestrial to subterranean to semi-aquatic species , genetic changes took place that would diminish the ability for the proteins to stick to each other . This pattern also showed that semi-aquatic lifestyles have independently evolved five separate times – twice in moles , three times in shrews . By retracing the evolutionary history of specific myoglobin properties , He et al . shed light on how one of the largest orders of mammals has come to be fantastically diverse . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2021 | Myoglobin primary structure reveals multiple convergent transitions to semi-aquatic life in the world's smallest mammalian divers |
To capture the functional diversity of microbiota , one must identify metabolic functions and species of interest within hundreds or thousands of microorganisms . We present Metage2Metabo ( M2M ) a resource that meets the need for de novo functional screening of genome-scale metabolic networks ( GSMNs ) at the scale of a metagenome , and the identification of critical species with respect to metabolic cooperation . M2M comprises a flexible pipeline for the characterisation of individual metabolisms and collective metabolic complementarity . In addition , M2M identifies key species , that are meaningful members of the community for functions of interest . We demonstrate that M2M is applicable to collections of genomes as well as metagenome-assembled genomes , permits an efficient GSMN reconstruction with Pathway Tools , and assesses the cooperation potential between species . M2M identifies key organisms by reducing the complexity of a large-scale microbiota into minimal communities with equivalent properties , suitable for further analyses .
Understanding the interactions between organisms within microbiomes is crucial for ecological ( Tara Oceans coordinators et al . , 2015 ) and health ( Integrative HMP ( iHMP ) Research Network Consortium , 2014 ) applications . Improvements in metagenomics , and in particular the development of methods to assemble individual genomes from metagenomes , have given rise to unprecedented amounts of data which can be used to elucidate the functioning of microbiomes . Hundreds or thousands of genomes can now be reconstructed from various environments ( Pasolli et al . , 2019; Forster et al . , 2019; Zou et al . , 2019; Stewart et al . , 2018; Almeida et al . , 2020 ) , either with the help of reference genomes or through metagenome-assembled genomes ( MAGs ) , paving the way for numerous downstream analyses . Some major interactions between species occur at the metabolic level . This is the case for negative interactions such as exploitative competition ( e . g . for nutrient resources ) , or for positive interactions such as cross-feeding or syntrophy ( Coyte and Rakoff-Nahoum , 2019 ) that we will refer to with the generic term of cooperation . In order to unravel such interactions between species , it is necessary to go beyond functional annotation of individual genomes and connect metagenomic data to metabolic modelling . The main challenges impeding mathematical and computational analysis and simulation of metabolism in microbiomes are the scale of metagenomic datasets and the incompleteness of their data . Genome-scale metabolic networks ( GSMNs ) integrate all the expected metabolic reactions of an organism . Thiele and Palsson , 2010 defined a precise protocol for their reconstruction , associating the use of automatic methods and thorough curation based on expertise , literature , and mathematical analyses . There now exists a variety of GSMN reconstruction implementations: all-in-one platforms such as Pathway Tools ( Karp et al . , 2016 ) , CarveMe ( Machado et al . , 2018 ) or KBase that provides narratives from metagenomic datasets analysis up to GSMN reconstruction with ModelSEED ( Henry et al . , 2010; Seaver et al . , 2020 ) . In addition , a variety of toolboxes ( Aite et al . , 2018; Wang et al . , 2018; Schellenberger et al . , 2011 ) , or individual tools perform targeted refinements and analyses on GSMNs ( Prigent et al . , 2017; Thiele et al . , 2014; Vitkin and Shlomi , 2012 ) . Reconstructed GSMNs are a resource to analyse the metabolic complementarity between species , which can be seen as a representation of the putative cooperation within communities ( Opatovsky et al . , 2018 ) . SMETANA ( Zelezniak et al . , 2015 ) estimates the cooperation potential and simulates flux exchanges within communities . MiSCoTo ( Frioux et al . , 2018 ) computes the metabolic potential of interacting species and performs community reduction . NetCooperate ( Levy et al . , 2015 ) predicts the metabolic complementarity between species . In addition , a variety of toolboxes have been proposed to study communities of organisms using GSMNs ( Kumar et al . , 2019; Sen and Orešič , 2019 ) , most of them relying on constraint-based modelling ( Chan et al . , 2017; Zomorrodi and Maranas , 2012; Khandelwal et al . , 2013 ) . However , these tools can only be applied to communities with few members , as the computational cost scales exponentially with the number of included members ( Kumar et al . , 2019 ) . Only recently has the computational bottleneck started to be addressed ( Diener et al . , 2020 ) . In addition , current methods require GSMNs of high quality in order to produce accurate mathematical predictions and quantitative simulations . Reaching this level of quality entails manual modifications to the models using human expertise , which is not feasible at a large scale in metagenomics . Automatic reconstruction of GSMNs scales to metagenomic datasets , but it comes with the cost of possible missing reactions and inaccurate stoichiometry that impede the use of constraint-based modelling ( Bernstein et al . , 2019 ) . Therefore , development of tools tailored to the analysis of large communities is needed . Here , we describe Metage2Metabo ( M2M ) , a software system for the characterisation of metabolic complementarity starting from annotated individual genomes . M2M capitalises on the parallel reconstruction of GSMNs and a relevant metabolic modelling formalism to scale to large microbiotas . It comprises a pipeline for the individual and collective analysis of GSMNs and the identification of communities and key species ensuring the producibility of metabolic compounds of interest . M2M automates the systematic reconstruction of GSMNs using Pathway Tools or relies on GSMNs provided by the user . The software system uses the algorithm of network expansion ( Ebenhöh et al . , 2004 ) to capture the set of producible metabolites in a GSMN . This choice answers the needs for stoichiometry inaccuracy handling , and the robustness of the algorithm was demonstrated by the stability of the set of reachable metabolites despite missing reactions ( Handorf et al . , 2005; Kruse and Ebenhöh , 2008 ) . Consequently , M2M scales metabolic modelling to metagenomics and large collections of ( metagenome-assembled ) genomes . We applied M2M on a collection of 1520 draft bacterial reference genomes from the gut microbiota ( Zou et al . , 2019 ) in order to illustrate the range of analyses the tool can produce . This demonstrates that M2M efficiently reconstructs metabolic networks for all genomes , identifies potential metabolites produced by cooperating bacteria , and suggests minimal communities and key species associated to their production . We then compared metabolic network reconstruction applied to the gut reference genomes to the results obtained with a collection of 913 cow rumen MAGs ( Stewart et al . , 2018 ) . In addition , we tested the robustness of metabolic prediction with respect to genome incompleteness by degrading the rumen MAGs . The comparison of outputs from the pipeline indicates stability of the results with moderately degraded genomes , and the overall suitability of M2M to MAGs . Finally , we demonstrated the applicability of M2M in practice to metagenomic data of individuals . To that purpose , we reconstructed communities for 170 samples of healthy and diabetic individuals ( Forslund et al . , 2015; Diener et al . , 2020 ) . We show how M2M can help connect sequence analyses to metabolic screening in metagenomic datasets .
M2M is a flexible software solution that performs automatic GSMN reconstruction and systematic screening of metabolic capabilities for up to thousands of species for which an annotated genome is available . The tool computes both the individual and collective metabolic capabilities to estimate the complementarity between the metabolisms of the species . Then based on a determined metabolic objective which can be ensuring the producibility of metabolites that need cooperation , that we call cooperation potential , M2M performs a community reduction step that aims at identifying a minimal community fulfilling the metabolic objective , as well as the set of associated key species . M2M’s main pipeline ( Figure 1a ) consists in five main steps that can be performed sequentially or independently: ( i ) reconstruction of metabolic networks for all annotated genomes , ( ii ) computation of individual and ( iii ) collective metabolic capabilities , ( iv ) calculation of the cooperation potential , and ( v ) identification of minimal communities and key species for a targeted set of compounds . Sets of producible metabolites for individual or communities of species are computed using the network expansion algorithm ( Ebenhöh et al . , 2004 ) that is implemented in Answer Set Programming in dependencies of M2M . Network expansion enables the calculation of the scope of one or several metabolic networks in given nutritional conditions , described as seed compounds . The scope therefore represents the metabolic potential or reachable metabolites in these conditions ( see Materials and methods ) . M2M calculates individual scopes for all metabolic networks , and the community scope comprising all reachable metabolites for the interacting species . Network expansion is also used in the community reduction optimisation implemented in MiSCoTo ( Frioux et al . , 2018 ) , the dependency of M2M , as reduced communities are expected to produce the metabolites of interest . The inputs to the whole workflow are a set of annotated genomes , a list of nutrients representing a growth medium , and optionally a list of targeted compounds to be produced by selected communities that will bypass the default objective of ensuring the producibility of the cooperation potential . Users can use the annotation pipeline of their choice prior running M2M . The whole pipeline is called with the command m2m workflow but each step can also be run individually as described in Table 1 . A main characteristic of M2M is to provide at the end of the pipeline a set of key species associated to a metabolic function together with one minimal community predicted to satisfy this function . We define as key species organisms whose GSMNs are selected in at least one of the minimal communities predicted to fulfill the metabolic objective . Among key species , we distinguish those that occur in every minimal community , suggesting that they possess key functions associated to the objective , from those that occur only in some communities . We call the former essential symbionts , and the latter alternative symbionts . These terms were inspired by the terminology used in flux variability analysis ( Orth et al . , 2010 ) for the description of reactions in all optimal flux distributions . If interested , one can compute the enumeration of all minimal communities with m2m_analysis , which will provide the total number of minimal communities as well as the composition of each . Figure 1b illustrates these concepts with an initial community formed of eight species . There are four minimal communities satisfying the metabolic objective . Each includes three species , and in particular , the yellow one is systematically a member . Therefore , the yellow species is an essential symbiont whereas the four other species involved in minimal communities constitute the set of alternative symbiont . As key species represent the diversity associated to all minimal communities , it is likely that their number is greater than the size of a minimal community , as this is the case in Figure 1b . In order to illustrate its applicability to real data , M2M was applied to a collection of 1520 bacterial high-quality draft reference genomes from the gut microbiota presented in Zou et al . , 2019 . The genomes were derived from cultured bacteria , isolated from faecal samples covering typical gut phyla ( Costea et al . , 2018 ) : 796 Firmicutes , 447 Bacteroidetes , 235 Actinobacteria , 36 Proteobacteria , and 6 Fusobacteria . The dereplicated genomes represent 338 species . The genomes were already annotated and could therefore directly enter M2M pipeline . The full workflow ( from GSMN reconstruction to key species computation ) took 155 min on a cluster with 72 CPUs and 144 Gb of memory . We illustrate in the next paragraphs the scalability of M2M and the range of analyses it proposes by applying the pipeline to this collection of genomes .
M2M allows metabolic modelling of large-scale communities , based on reference genomes or de novo constructed MAGs , inferring metabolic complementarity found within communities . M2M is a flexible framework that automates GSMN reconstruction , individually and collectively analyse GSMNs , and performs community selection for targeted functions . The large combinatorics of minimal communities due to functional redundancy in microbiotas is addressed by providing key species associated to metabolic end-products . This could allow targeting specific members of the community through pro- or prebiotics , to model the metabolites the human host will be exposed to . We validated the flexibility of the software and the range of analyses it can offer with several datasets , corresponding to multiple use-cases in the microbiome field . This allowed us to characterise metabolic complementarity in a large collection of draft reference genomes . We further assessed the robustness of M2M to data incompleteness by performing analyses on collections of MAGs . Finally , we applied M2M to a common use-case in metagenomics: the study of communities associated to individuals , in a disease context . Our method is robust against the uncertainty inherent to metagenomics data . It scales to typical microbial communities found in the gut and predicts key species for functions of interest at the metabolic level . Future developments will broaden the range of interactions to be modelled and facilitate the incorporation of abundance data . This software is an answer to the need for scalable predictive methods in the context of metagenomics where the number of available genomes continues to rise .
M2M can process existing metabolic networks in SBML format or proposes the automatic reconstruction of non-curated metabolic networks ( m2m recon ) . As a multi-processing solution , it facilitates the treatment of hundreds or thousands of genomes that can be retrieved from metagenomic experiments . The underlying GSMN reconstruction software is Pathway Tools ( Karp et al . , 2016 ) , a graphical user interface ( GUI ) based software suite for the generation of individual GSMNs , called Pathway/Genome Databases ( PGDBs ) . Typically , a PGDB is obtained from an annotated genome using PathoLogic , the software prediction component of Pathway Tools , and curated afterwards . We developed Mpwt ( Multiprocessing Pathway Tools ) , a command-line Python wrapper ( also available as a standalone tool ) for Pathway Tools . Mpwt and M2M ( i ) format the genomic inputs , ( ii ) automate the reconstruction step by initialising a PathoLogic environment for each genome , and ( iii ) extract and convert the resulting GSMNs in PGDB and SBML ( Hucka et al . , 2003; Hucka et al . , 2018 ) formats using the PADMet library ( Aite et al . , 2018 ) . Mpwt handles three types of genomic inputs ( Genbank , Generic Feature Format ( GFF ) or PathoLogic format ) that must contain GO-terms and EC-numbers annotations necessary for Pathway Tools . These annotations are for example found in the Genbank files generated by Prokka ( Seemann , 2014 ) . In addition , we specifically developed Emapper2gbk , a Python package dedicated to the connection between the Eggnog-mapper annotation tool ( Huerta-Cepas et al . , 2017 ) and Mpwt in order to generate these inputs . This part of the workflow encompasses three steps: computation of the ( i ) individual ( m2m iscope ) and ( ii ) collective ( m2m cscope ) metabolic potentials , and ( iii ) the characterisation of the cooperation potential of the GSMN collection ( m2m addedvalue ) . The former two rely on the network expansion algorithm ( Ebenhöh et al . , 2004 ) , the latter being a set difference between the results of the first two steps . The network expansion algorithm computes the scope of a metabolic network from a description of the growth medium called seeds . The scope consists in the set of metabolic compounds which are reachable , or producible , according to a boolean abstraction of the network dynamics assuming that cycles cannot be self-activated . More precisely , the algorithm recursively considers products of reactions to be producible if all reactants of the reactions are producible , provided an initiation with a set of seed nutrients . The underlying implementation of the network expansion algorithm used in M2M relies on Answer Set Programming ( ASP ) ( Schaub and Thiele , 2009 ) . We define a metabolic network as a bipartite graph G= ( R∪M , E ) , where R and M stand for reaction and metabolite nodes . When ( m , r ) ∈E ( respectively ( r , m ) ∈E ) , with m∈M and r∈R , the metabolite is called a reactant ( respectively product ) of the reaction r . The scope of a set of seed compounds S according to a metabolic network G , denoted by scope ( G , S ) , is iteratively computed until it reaches a fixed point ( Handorf et al . , 2005 ) . It is formally defined byScope ( G , S ) =⋃iMi , where M0=S and Mi+1=Mi∪products ( {r∈R∣reactants ( r ) ⊆Mi} ) . A minimal community 𝒞 enabling the producibility of a set of targets T from the seeds S is a sub-family of the community G1 , … , Gn which is solution of the following optimisation problem:minimize{Gi1 . . . GiL} ⊂{G1 . . . GN}size ( {Gi1 . . . GiL} ) subject toT⊂collectiveScope ( Gi1 . . . GiL , S ) . Solutions to this optimisation problem are communities 𝒞= ( Gi1… , GiL ) of minimal size . We define minimalCommunities ( G1 . . . Gn , S , T ) to be the set of all such minimal communities . A first output of the m2m mincom command is the ( minimal ) size L of communities solution of the optimisation problem . The composition of one optimal community is also provided . The targets are by default the components of the cooperation potential , T=cooperationPotential ( G1 , . . . , Gn , S ) , but can also be a group of target metabolites defined by the user . Many minimal communities are expected to be equivalent for a given metabolic objective but their enumeration can be computationally costly . We define key species which are organisms occurring in at least one community among all the optimal ones . Key species can be further distinguished into essential symbionts and alternative symbionts . The former occur in every minimal community whereas the latter occur only in some minimal communities . More precisely , the key species keySpecies ( G1 . . . Gn , S , T ) , the essential symbionts essentialSymbionts ( G1 . . . Gn , S , T ) , and the alternative symbionts alternativeSymbionts ( G1 . . . Gn , S , T ) associated to a set of metabolic networks , seeds S and a set of target metabolites T are defined bykeySpecies ( G1 . . . Gn , S , T ) ={G∣∃𝒞∈minimalCommunities ( G1 . . . Gn , S , T ) , G∈𝒞} . essentialSymbionts ( G1 . . . Gn , S , T ) ={G∣∀𝒞∈minimalCommunities ( G1 . . . Gn , S , T ) , G∈𝒞} . alternativeSymbionts ( G1 . . Gn , S , T ) =keySpecies ( G1 . . . Gn , S , T ) ∖essentialSymbionts ( G1 . . . Gn , S , T ) . As a strategy layer over MiSCoTo , M2M relies on the Clasp solver ( Gebser et al . , 2012 ) for efficient resolution of the underlying grounded ASP instances . Although this type of decision problem is NP-hard ( Julien-Laferrière et al . , 2016 ) , as with many real-world optimisation problems worst-case asymptomatic complexity is less informative for applications than practical performance using heuristic methods . The Clasp solver implements a robust collection of heuristics ( Gebser et al . , 2007; Andres et al . , 2012 ) for core-guided weighted MaxSAT ( Manquinho et al . , 2009; Morgado et al . , 2012 ) that provide rapid set-based solutions to combinatorial optimisation problems , much in the same way that heuristic solvers like CPlex provide rapid numerical solutions to mixed integer programming optimisation problems . The kinds of ASP instances constructed by MiSCoTo for M2M are solved in a matter of minutes for the identification of key species and essential/alternative symbionts . Indeed the space of solutions is efficiently sampled using adequate projection modes in ASP , which enables the computation of these groups of species without the need for a full enumeration . | All the microbes that live in a specific environment , for example an organ , are collectively called the microbiota . In humans , the microbiota of the gut has been extensively studied by sequencing the DNA of the different microbes to identify them and determine the roles they play in health and disease . The DNA sequences of all the members of the microbiota is called the metagenome . The chemical reactions that the gut microbiota perform to produce energy and make the biomolecules they need to survive are collectively referred to as the metabolism of these microbes . Studying the metabolism of the gut microbiota can help researchers understand the roles of the different microbes . However , the large variety of species in the gut microbiota and gaps in the information about them render these studies difficult , despite technology improving quickly . To tackle this issue , Belcour , Frioux et al developed a new piece of software called Metage2Metabo ( M2M ) that simulates the metabolism of the gut microbiota and describes the metabolic relationships between the different microbes . Metage2Metabo analyses the roles of the metabolic genes of a large number of microbe species to establish how they complement each other metabolically . Then , it can calculate the minimum number of species needed to perform a metabolic role of interest within that microbiota , and which key species are associated with that role . To test the new software , Belcour , Frioux et al . used Metage2Metabo to analyse genomes from the human gut microbiota and from the cow rumen ( one of the cow’s stomachs ) . They showed that even if the metagenome was incomplete , the software was able to make stable predictions of key species involved in metabolic complementarity . Additionally , they also illustrated how the method can be used to study the gut microbiota of individuals . This work presents a new method for determining the metabolic relationships between species within a microbiota . The software is highly flexible and could be adapted to identify key members within different communities . In the context of the gut microbiota , the predictions of Metage2Metabo could shed lights on the interactions between the host and the microbes and contribute to a better understanding of microbe environments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"tools",
"and",
"resources"
] | 2020 | Metage2Metabo, microbiota-scale metabolic complementarity for the identification of key species |
New discoveries and dating of fossil remains from the Rising Star cave system , Cradle of Humankind , South Africa , have strong implications for our understanding of Pleistocene human evolution in Africa . Direct dating of Homo naledi fossils from the Dinaledi Chamber ( Berger et al . , 2015 ) shows that they were deposited between about 236 ka and 335 ka ( Dirks et al . , 2017 ) , placing H . naledi in the later Middle Pleistocene . Hawks and colleagues ( Hawks et al . , 2017 ) report the discovery of a second chamber within the Rising Star system ( Dirks et al . , 2015 ) that contains H . naledi remains . Previously , only large-brained modern humans or their close relatives had been demonstrated to exist at this late time in Africa , but the fossil evidence for any hominins in subequatorial Africa was very sparse . It is now evident that a diversity of hominin lineages existed in this region , with some divergent lineages contributing DNA to living humans and at least H . naledi representing a survivor from the earliest stages of diversification within Homo . The existence of a diverse array of hominins in subequatorial comports with our present knowledge of diversity across other savanna-adapted species , as well as with palaeoclimate and paleoenvironmental data . H . naledi casts the fossil and archaeological records into a new light , as we cannot exclude that this lineage was responsible for the production of Acheulean or Middle Stone Age tool industries .
Geological evidence from the Dinaledi Chamber , including direct dating of the fossil hominin remains , places H . naledi in the later Middle Pleistocene . All considerations of the phylogenetic placement of H . naledi to date agree that its branch on the hominin phylogeny must have originated earlier than 900 ka ( Dembo et al . , 2016 ) , and earlier branch points are credible ( Berger et al . , 2015; Dembo et al . , 2016; Thackeray , 2015 ) . If H . naledi stems from a basal node within Homo , a scenario that is not rejected by phylogenetic analyses ( Dembo et al . , 2016; Hawks and Berger , 2016 ) , its branch may have originated much earlier , in the Pliocene . How did this H . naledi lineage fit into the hominin diversity of the Early and Middle Pleistocene , and is it possible that palaeoanthropologists may already have discovered—but not recognized—other fossils that represent this branch ? The hominin fossil record of the African Middle Pleistocene is extremely sparse ( Klein , 1973; Berger and Parkington , 1995; Grün et al . , 1996; Stynder et al . , 2001; Marean and Assefa , 2005; Klein et al . , 2007a; Millard , 2008; Stringer , 2011; Wood , 2011; Smith et al . , 2015 ) . Fossils that putatively derive from this period between 780 , 000 and 130 , 000 years ago are limited and typically fragmentary ( Table 1; Figure 1 ) . Only a small number are thought to come from the period before 200 , 000 years ago . To this number , it is possible to add perhaps a half dozen partial mandibles and a somewhat greater number of postcranial fragments or dental remains . Many of these were found prior to 1960 and lack adequate provenience . Some have been ‘dated’ mainly by their morphological appearance , or by examination of vertebrate faunal remains that were not excavated together in association with the hominin specimen . Others have been subject to direct dating , but this has often been reported in ways that do not reflect the full statistical uncertainty ( Millard , 2008 ) . Among these finds , crania have been considered the most diagnostically important as indicators of the presence of archaic humans with large brain sizes in the Middle and Late Middle Pleistocene . 10 . 7554/eLife . 24234 . 003Figure 1 . African fossil sites from the Middle and earliest Late Pleistocene . Sites discussed in the text are highlighted in pink here . Geological age estimations for each fossil hominin assemblage are given in Table 1 , along with references . DOI: http://dx . doi . org/10 . 7554/eLife . 24234 . 00310 . 7554/eLife . 24234 . 004Table 1 . Significant hominin fossil remains from the Middle and Early Late Pleistocene of Africa . Included are those sites that have geological age estimates between 780 , 000 and 120 , 000 years ago , and some sites for which claims of Middle Pleistocene age have been made but without chronometric support . Sites denoted here with ‘no date’ are those for which no chronometric determinations based on samples of hominin material or securely associated faunal remains have been reported in the literature . Some chronometric determinations that were based only on morphology or associated fauna have given rise to broad age estimations; we omit the details of such determinations here . Some additional sites with fragmentary remains , especially isolated dental remains , are not listed . The first four entries ( KNM-OL 45500 , OH 12 , Daka and Buia ) are older than the beginning of the Middle Pleistocene but are included because they are discussed in text . DOI: http://dx . doi . org/10 . 7554/eLife . 24234 . 004SiteSpecimensLocationGeological age ( ka ) Source ( s ) Olorgesailie ( KNM-OL 45500 ) FrontalKenya900–970Potts et al . ( 2004 ) Olduvai Gorge ( OH 12 ) *Partial calvariaTanzania780–1 , 200Tamrat et al . ( 1995 ) ; Mcbrearty and Brooks ( 2000 ) DakaCalvaria , femurEthiopia~1 , 000Asfaw et al . ( 2002 ) BuiaCalvaria , postcranial fragmentsEritrea~1 , 000Abbate et al . ( 1998 ) Tighénif ( Ternifine ) Three mandibles , skull fragmentMorocco~700Geraads et al . ( 1986 ) Elandsfontein ( Saldanha ) Partial calvaria and mandible fragSouth Africa600–1 , 000Klein et al . ( 2007a ) BodoPartial calvaria , left parietal ( found roughly 400 m from Bodo 1 ) , distal humerusEthiopia550–640Conroy et al . ( 1978 ) ; Clark et al . ( 1994 ) Baringo ( Kaphturin Formation ) Mandible , ulnaKenya510–512Leakey et al . ( 1970 ) ; Deino and McBrearty ( 2002 ) SaléPartial calvaria and upper dentitionMorocco~300Jaeger , 1975 ) ; Geraads ( 2012 ) Ndutu*Partial calvariaTanzania370–990Tamrat et al . ( 1995 ) ; Mcbrearty and Brooks ( 2000 ) Berg AukasPartial femurNamibiaNo dateGrine et al . ( 1995 ) KabweCalvaria , material from at least three individualsZambiaNo dateKlein ( 1973 ) FlorisbadPartial craniumSouth Africa224–294Grün et al . ( 1996 ) Cave of HearthsPartial mandibleSouth AfricaNo dateCooke , 1962 ) HoedjiespuntTeeth , tibiaSouth AfricaNo dateBerger and Parkington ( 1995 ) ; Stynder et al . ( 2001 ) Eliye SpringsCalvariaKenyaNo dateBräuer and Leakey ( 1986 ) Dinaledi Chamber ( Rising Star ) Remains of at least 15 individualsSouth Africa236–335Berger et al . , 2015; Dirks et al . ( 2017 ) Lesedi Chamber ( Rising Star ) Partial skeleton , remains of at least three individualsSouth AfricaNo dateHawks et al . , 2017 ) Omo KibishTwo partial crania , partial skeletonEthiopia155–200McDougall et al . ( 2005 ) ; Aubert et al . ( 2012 ) HertoThree partial craniaEthiopia154–160White et al . ( 2003 ) ; Clark et al . ( 2003 ) Ileret ( KNM-ER 3884 ) Partial calvariaKenya162–∞Bräuer et al . ( 1997 ) Jebel IrhoudThree calvaria , mandible , fragments of seven individualsMorocco144–176Hublin ( 2001 ) ; Smith et al . ( 2007 ) Laetoli ( Ngaloba Beds ) CraniumTanzania130Day et al . ( 1980 ) ; Hay ( 1987 ) SingaCalvariaSudan131–135McDermott et al . ( 1996 ) Lake EyasiCalvariaTanzania88–132Mehlman ( 1984 , 1987 ) ; Domínguez-Rodrigo et al . , 2008*Many authors have studied the stratigraphy of Olduvai Gorge and nearby sites , resulting in varied dates being reported for these fossils . We report here the widest range as reviewed by Mcbrearty and Brooks ( 2000 ) , based on the paleomagnetic sequence . Only three of the fossil crania occur within 3 , 000 km of the location of the Rising Star cave system . The cranium and mandible fragment from Elandsfontein , or Saldanha , South Africa derives from an open-air deposit with multiple depositions of faunal remains , which , by comparison with East African vertebrate assemblages , come from about 600 , 000 to 1 million years ago ( Klein et al . , 2007a ) . However , the skull and mandible fragment are essentially surface finds . The Florisbad partial cranium from near Bloemfontein , South Africa , comes from mineral springs which accumulated fossils from around 40 ka to 400 ka; a direct electron spin resonance ( ESR ) assessment on a human M3 yielded an age estimate of 259 ± 35 ka ( Grün et al . , 1996 ) . This single tooth is assumed to be associated with the fossil hominin cranium . It is roughly 1 mm smaller in diameter than a sample of four H . naledi upper third molars , and lies within the size range of both H . erectus and African modern human populations ( Smith et al . , 2015 ) . Although third molars vary substantially in humans and fossil Homo , this tooth does not resemble the morphology of known H . naledi maxillary third molars . The Kabwe , or Broken Hill , cranium from Zambia has no direct geological age assessment and its context is very poorly known . The vertebrate fauna in the cave sediments where the skull presumably originated is Middle Pleistocene in age ( Klein , 1973; Millard , 2008 ) , although the context of the skull in relation to these faunal remains is not clear . Observing that the morphology of the skull is primitive compared to those of crania from early Late Pleistocene contexts , some workers have argued that the Kabwe skull may be 300 ka or earlier ( Bräuer , 2008 ) , though more recent work suggests it may be of later middle Pleistocene age ( Stringer , 2011 ) . The Lake Ndutu cranium is the only other subequatorial cranial specimen thought to be between 780 , 000 and 200 , 000 years old , but it also lacks secure provenience or a direct geological age estimate . Teeth from Hoedjiespunt ( Berger and Parkington , 1995; Stynder et al . , 2001 ) and the mandible from Cave of Hearths , South Africa , also provide evidence of hominins , probably of Middle Pleistocene age , but without further comparisons we cannot rule out the possibility that these fossils may themselves represent H . naledi . All of the cranial fossils discussed above share substantially larger brain size than H . naledi but are morphologically diverse in comparison to each other . The uncertain provenience and inexact geological ages of these fossils limit our ability to test when and where the populations that they represent may have existed , and it is conceivable that some of the remains are not Middle Pleistocene at all . In addition to the crania , a few postcranial specimens document individuals with a larger body size than has yet been observed for H . naledi . The Kabwe hominin collection includes several postcranial elements , which are not associated with certainty with the cranium , but that clearly represent individuals with a large body size . Also within the 3 , 000 km radius are a large femur from Berg Aukas , Namibia ( Grine et al . , 1995 ) , presumed to be of Middle Pleistocene age , and a large tibia from Hoedjiespunt , South Africa ( Berger and Parkington , 1995; Stynder et al . , 2001; Churchill et al . , 2000 ) that is Middle Pleistocene in age . Earlier than these Middle Pleistocene fossil specimens , the sites of Olorgesailie and Olduvai Gorge both preserve evidence of hominin crania ( KNM-OL 45500 , OH 12 ) , which although fragmentary , clearly belonged to individuals that had relatively small brain sizes comparable to some of the earliest H . erectus remains ( Potts et al . , 2004; Antón , 2004 ) . These have been called ‘H . erectus-like’ , but they are different from contemporary fossil specimens attributed to H . erectus from further to the north , such as BOU-VP-2/66 calvaria from Daka , Ethiopia ( Asfaw et al . , 2002; Gilbert et al . , 2003 ) , and the UA 31 cranium from Buia , Eritrea ( Abbate et al . , 1998 ) , both approximately 1 million years old . KNM-OL 45500 and OH 12 also differ from the large and robust OH 9 cranium , which is likewise from Olduvai Gorge but earlier in time . The fossil sites from Lake Baringo southward to Olduvai Gorge lie at the hinge point separating subequatorial from northeast African populations of many living species of mammals . It is possible that fossils such as KNM-OL 45500 , OH 12 and OH 28 represent northern excursions of a more diverse subequatorial hominin community that included H . naledi and its relatives . The great tropical forests of Africa pose a biogeographic barrier to species that are adapted to savanna and savanna-woodland-mosaic habitats . During the Pleistocene , these tropical forests repeatedly expanded eastward along the equator . During these times , Equatorial East Africa broke into a mosaic of small savanna remnants , while a large and contiguous area of savanna and savanna-woodland mosaic stretched southward from the equatorial forest ( Lorenzen et al . , 2012; Faith et al . , 2016 ) . During most of the Pleistocene , the area suitable for hominins located to the south of the equatorial forests was vastly larger than that to the north or east . During the last 1 . 3 million years of this period , the Lake Malawi basin became increasingly moist with greater climate stability ( Johnson et al . , 2016 ) . The paleoclimate record for the Kalahari is not as deep , but during the last 200 , 000 years , this area underwent both arid periods with dune formation and wetter periods with vast paleolakes ( Robbins et al . , 2016 ) , a pattern that likely held during earlier climate cycles . Hence the subequatorial area that was potentially suitable for hominins varied extensively , but may have been 5–15 times larger than the equivalent habitats in eastern Africa north of the equator . This biogeographic history is reflected in the phylogeography of savanna-adapted species today . Across several orders of savanna-adapted mammals , including ungulates , primates , and carnivores , a north–south dichotomy can be observed in genetic patterning ( Bertola et al . , 2016 ) , and this is thought to be the result of past climate patterns . A comparison of 19 ungulate species shows that many have a ‘suture zone’ in the East African equatorial region , where relatively distinct mtDNA clades from eastern and southern parts of the species’ range meet ( Lorenzen et al . , 2012 ) . In several species of ungulates , primates , and carnivores , today’s East African populations were colonized from the large and more contiguous region to the south; in some other species , East African diversity has been maintained through vicariance of populations in mosaic refugia within East Africa . A southern African origin is also inferred for baboons ( Papio ) . Likemany savanna ungulates this genus manifests an mtDNA suture zone in equatorial East Africa , which has recently been expanded by introgressive hybridization of a more southern mtDNA clade into P . cynocephalus ( Zinner et al . , 2013 ) . Similar to many other mammals in Africa , living humans carry a genetic legacy of greater subequatorial African diversity . Today , the human populations that have the highest observed genetic diversity are the hunter-gatherers of southwestern Africa , followed by the Mbuti rainforest hunter-gatherers of north eastern Congo ( Mallick et al . , 2016 ) . Hadza and Sandawe people from Tanzania also draw ancestry from a diverse source population similar to today’s southwestern African hunter-gatherers ( Pickrell et al . , 2014 ) . Today’s populations are descendants of human populations that have been relatively large and stable across most of the past 200 , 000 years . These populations did not undergo the population bottleneck evidenced in non-African populations and to a lesser degree in West African and northeastern African peoples ( Mallick et al . , 2016 ) . Additionally , the genomes of Hadza , Sandawe , Biaka , Baka , and San people bear evidence of a small fraction of introgression from highly genetically divergent populations that no longer exist ( Hammer et al . , 2011; Lachance et al . , 2012; Beltrame et al . , 2016; Hsieh et al . , 2016 ) . The implication of these observations is that tropical and subequatorial Africa were home to multiple genetically divergent populations of hominins . Some of these populations diverged in the Early Pleistocene , and had genomes that were equally or more diverse than those of Neanderthals , Denisovans , or contemporary modern humans . Some of these populations survived and hybridized after the initial diversification of modern humans , perhaps as recently as 35 , 000 years ago ( Hammer et al . , 2011 ) or even into the early Holocene ( Hsieh et al . , 2016 ) . As other have noted ( Stringer , 2016 ) , the fossil hominin record of the Middle and Late Pleistocene shows no simple linear progression towards modern humans , and different morphological forms overlapped in time . A small-brained hominin has been recognized from this time period in Asia on the island of Flores ( Brown et al . , 2004 ) , and we now include a small-brained species of hominin from Africa in this recognized diversity . Modern H . sapiens is a phylogenetic relict . In biology , a relict is a species that remains from a clade that was more diverse in the past ( Grandcolas and Trewick , 2016 ) . We have known for a long time that other hominin populations once inhabited Eurasia and island Southeast Asia , including the Neanderthals , Denisovans , and H . floresiensis ( Bocquet-Appel and Demars , 2000; Brown et al . , 2004; Cooper and Stringer , 2013; Li et al . , 2017 ) . Genetic evidence shows that equally diverse populations of archaic humans once existed in subequatorial Africa ( Hammer et al . , 2011; Stringer , 2011; Lachance et al . , 2012 ) , and although no fossil evidence can yet be associated with such evidence of genetic introgression , the Middle Pleistocene record of this region does speak to the presence of morphological diversity . Within this context , H . naledi provides fossil evidence of one subequatorial lineage , and we do not yet know whether it contributed to the modern human gene pool . Another implication of modern humans as a relict is that the features of today’s humans give a biased and incomplete picture of the diversity of the Homo clade ( cf . Grandcolas et al . , 2014 ) . These biases have had enormous consequences for the historical development of paleoanthropology . One of the most persistent biases has been to conceive of postcranial and dental adaptations of Homo as mere adjuncts to the extraordinary increase in brain size evidenced in living humans . Poor fossil evidence once appeared to support the notion that human-like aspects of locomotor , manipulatory , and dietary strategy evolved in tandem with larger brains , and that H . erectus combined these for the first time ( Wood and Collard , 1999; Hawks et al . , 2000 ) . But newer evidence shows that some fossils attributed to H . erectus had a mosaic of human-like and primitive postcranial features ( Lordkipanidze et al . , 2007 ) , that some fossil samples of H . erectus had brain sizes equivalent to those of H . habilis and H . rudolfensis ( Lordkipanidze et al . , 2013 ) , and that H . habilis may be closer to Au . sediba than to other species of Homo ( Dembo et al . , 2015 , 2016 ) . H . naledi shows that many human-like anatomical aspects of the hand , foot , lower limb , dentition and cranium , including some aspects that are not present in H . erectus , occurred in a species with a brain size equal to that of australopiths ( Berger et al . , 2015 ) . The full geographic extent of H . naledi is unknown , though aspects of its anatomy might be used to argue that this species is unlikely to be endemic only to the region where its fossils are presently found . With its human-like pattern of lower limb and foot anatomy ( Harcourt-Smith et al . , 2015; Marchi et al . , 2017 ) , human-like stature ( Berger et al . , 2015 ) , and dental morphology consistent with reliance on a high-quality diet ( Berger et al . , 2015 ) , H . naledi appears to have used its environment in a similar way to H . erectus and H . sapiens . Indeed , even earlier small-bodied Australopithecus and Paranthropus have been inferred to be eurytopic , capable of using a wide range of habitats , and all well-sampled species are geographically widespread ( Wood and Strait , 2004; Behrensmeyer and Reed , 2013 ) . Aside from these considerations , the hominin fossil sample is insufficient to support further conclusions about the geographic range of H . naledi . The late survival of H . naledi from origins deep in the Pleistocene up to the Dinaledi and Lesedi Chamber deposits is surprisingly unhelpful in testing hypotheses about its evolutionary origin or its morphological pattern . How the traits of H . naledi evolved does not depend on the geological age of the Dinaledi Chamber fossils , but on the phylogenetic position of H . naledi and the morphological patterns of other hominin taxa . Phylogenetic scenarios for H . naledi place its origin either: ( 1 ) somewhere among the poorly resolved branches leading to H . habilis , H . rudolfensis , H . floresiensis and Au . sediba ( Berger et al . , 2015; Dembo et al . , 2016 ) ; Thackeray , 2015 ) : ( 2 ) as a sister to H . erectus and larger-brained Homo including H . sapiens ( Dembo et al . , 2016 ) ; or ( 3 ) as a sister to a clade including H . sapiens , H . antecessor , and other archaic humans ( Dembo et al . , 2016 ) ( Figure 2 ) . Maximum parsimony analysis of a large dataset of cranial and dental traits supports scenario 1 , placing H . naledi among the most basal nodes of the Homo phylogeny ( Dembo et al . , 2016 ) . Bayesian analysis of the same dataset supports scenario 3 , placing H . naledi closer to modern humans than any H . erectus sample ( Dembo et al . , 2016 ) . An informal consideration of postcranial traits suggests that Dembo et al . ( 2016 ) analysis , if it included postcrania , might more likely support scenario 2 . This is because H . naledi shares many derived features of the hand , foot , and lower limb with H . erectus and H . sapiens that are apparently absent from H . habilis , H . floresiensis , or Au . sediba , yet lacks several derived traits of the shoulder , trunk , and hip shared by H . erectus and H . sapiens ( Hawks et al . , 2017; Williams et al . , 2017; Marchi et al . , 2017; Feuerriegel et al . , 2017 ) . However , the fossil record for these areas of anatomy in early hominins other than H . naledi is admittedly limited . 10 . 7554/eLife . 24234 . 005Figure 2 . Phylogenetic scenarios for H . naledi . A simplified cladogram of Homo , with the possible placements of H . naledi indicated . The cladogram places A . africanus as an outgroup to the Homo + Au . sediba clade , as consistent with nearly all phylogenetic analyses of these species ( Berger et al . , 2010; Dembo et al . , 2015 , 2016 ) . To simplify the tree , we have omitted H . antecessor , H . heidelbergensis and Neanderthals , which all phylogenetic analyses place as sisters to H . sapiens relative to H . erectus . There is no present consensus about the branching order among H . habilis , H . rudolfensis , H . floresiensis and Au . sediba ( Dembo et al . , 2015 , 2016 ) , and so these are depicted as a polytomy . DOI: http://dx . doi . org/10 . 7554/eLife . 24234 . 005 No interpretation of this anatomy can eliminate the necessity of some reversals or parallelism . If H . naledi is a sister to H . sapiens ( scenario 3 ) , then all of the primitive traits it does not share with H . erectus , including its small brain size ( Berger et al . , 2015 ) , shoulder morphology ( Feuerriegel et al . , 2017 ) , ilium form ( VanSickle et al . , personal communication ) , long , anteroposteriorly flattened femur neck ( Marchi et al . , 2017 ) , thorax shape ( Williams et al . , 2017 ) , and markedly curved finger bones ( Kivell et al . , 2015 ) , might be interpreted as evolutionary reversals . If H . naledi is a sister taxon to a clade including H . habilis , H . rudolfensis and all other large-brained species of Homo , then the larger brain size of these other species of Homo could be homologous . But this scenario would require many parallel evolutionary developments in H . naledi and H . sapiens , including hand and wrist morphology ( Kivell et al . , 2015 ) , foot morphology ( Harcourt-Smith et al . , 2015 ) , lower limb morphology ( Marchi et al . , 2017 ) , and some cranial and dental morphologies ( Laird et al . , 2017; Schroeder et al . , 2017 ) . Some of these derived traits are also shared with H . erectus , others are not evidenced in any known H . erectus fossils . Whatever phylogenetic scenario we accept , H . naledi is not unique in demonstrating homoplasy ( Wood and Harrison , 2011 ) , but it does present a uniquely strong postcranial record documenting its mosaic anatomy . The long evolutionary branch leading to H . naledi as represented in the Rising Star cave system may have implications for its mosaicism , at least with respect to cranial and mandibular form . Much of the evolution of cranial form among species of Homo in the Pleistocene appears to be consistent with neutral evolution by genetic drift , with a few features showing evidence of adaptive evolution ( Ackermann and Cheverud , 2004; Weaver et al . , 2007; Schroeder et al . , 2014 ) . If the correlations among some aspects of H . naledi cranial anatomy were not constrained by selection , then a long evolutionary branch would create substantial opportunity for divergence over time by drift . Such non-adaptive evolution , combined with the adaptive evolution of some traits , might create a unique pattern in this species ( Laird et al . , 2017 ) , although it seems likely that postcranial features would be subject to greater adaptive constraints . An alternative hypothesis for the homoplastic appearance of H . naledi is hybridization among two or more hominin lineages . As ancient DNA evidence has grown , it has become clear that hybridization among genetically distant human lineages occurred many times ( Kuhlwilm et al . , 2016; Meyer et al . , 2012; Prüfer et al . , 2014 ) , as is the case in chimpanzees and bonobos ( de Manuel et al . , 2016 ) and in many other mammalian lineages ( Schaefer et al . , 2016 ) . The mosaic anatomy of H . naledi , which includes many shared derived characters of modern humans and H . erectus , might suggest the hypothesis that H . naledi resulted from the hybridization of a more human-like population and a late-surviving australopith . This hypothesis remains untestable with the current evidence , although it seems more parsimonious to suggest that H . naledi itself survived from an early period of diversification of Homo . Morphology does not rule out the possibility that H . naledi originated in the Early Pleistocene as a result of the hybridization of different populations , and persisted long after this hybrid speciation . The evidence of genetic mixture among more recent hominins makes this hypothesis seem reasonable , but again it is untestable unless genetic material is obtained from the fossils . Attempts to obtain aDNA from H . naledi remains have thus far proven unsuccessful . In addition , we have reported several apparent autapomorphies that are present across the skeleton of H . naledi . These include the morphology of the thumb , aspects of the morphology of the spine , and aspects of the morphology of the proximal femur ( Berger et al . , 2015; Kivell et al . , 2015; Marchi et al . , 2017 ) . Unfortunately , the Pliocene hominin record is poor , and without clearly understanding the ancestral lineage of H . naledi , and whether we have in fact already discovered its ancestors , we cannot know whether such features may have been present in the last common ancestor ( LCA ) of H . naledi and other hominin species , and are thus actually primitive in H . naledi’s lineage rather than uniquely derived . Therefore , the importance of these apparent autapomorphies in establishing the origins of H . naledi remain unresolved . Until now , palaeoanthropologists and archaeologists have generally assumed that morphologically primitive hominins such as H . naledi did not survive into the later parts of the Pleistocene in Africa . This assumption has guided the interpretation of fossil discoveries with poor geological or stratigraphic context , including the many surface finds that make up the majority of the record from ancient lacustrine and riverine deposits ( e . g . Taieb et al . , 1976; Yuretich , 1979; Kalb et al . , 1982; Tiercelin , 1986; Ward et al . , 1999; WoldeGabriel et al . , 2001; Clark et al . , 2003; Gathogo and Brown , 2006; McDougall and Brown , 2006; Campisano and Feibel , 2008; Campisano , 2012 ) . These and other studies have shown that in many African sedimentary contexts , Pliocene or Early Pleistocene sediments are overlain by deposits of Middle or Late Pleistocene age or even by Holocene-aged deposits . It is common knowledge that fragmentary fossils of Plio-Pleistocene age occur ex situ on the surface with Middle Stone Age ( MSA ) , Later Stone Age ( LSA ) , or historic artifacts; in the absence of in situ association , anthropologists often rely upon a fossil’s morphology as an indicator of its age . The discovery of H . naledi provided a natural experiment to test whether anthropologists can reliably establish the approximate age of hominin fossil fragments from their morphology . Before the publication of a geological age for H . naledi , many anthropologists examined its entire morphological pattern and concluded that the species must date to more than 1 . 5 million years ago . This includes one formal morphological study ( Thackeray , 2015 ) and many other published comments by experts . A second study concluded that the Dinaledi hominin sample could be 930 , 000 years old , though the confidence interval on this estimate ranged from the present to c . 2 . 5 Ma ( Dembo et al . , 2016 ) . These examples show that expert intuition about the ages of fossil samples is likely to be wrong when based on their morphology alone . We must therefore demand fuller information about the geological context both of surface finds and of finds that are reported as in situ . If fragments of H . naledi had been found in isolation—instead of in the cohesive assemblage of the Dinaledi Chamber—many parts of its anatomy individually may have been confused for hominin material of Pliocene age . As we have noted , parts of the H . naledi cranial vault , dentition , shoulder , manual phalanges , pelvis and proximal femur would be easily misattributed to Australopithecus . Other parts of the hand , dentition , foot , and lower limb exhibit morphology similar to that of modern humans or H . erectus . As we know neither the origination point nor the extinction time of H . naledi , it is conceivable that fragments from this species have already been misattributed to other hominin taxa . H . naledi has traits that were long considered to be adaptations for creating material culture . Its wrist , hand and fingertip morphology share several derived features with Neanderthals and modern humans that are absent in H . habilis , H . floresiensis , and Au . sediba ( Kivell et al . , 2015 ) . If these features evolved to support habitual tool manufacture in Neanderthals and modern humans , then it is reasonable to conclude that H . naledi was also fully competent in using tools . The use of tools and the consumption of higher-quality foodstuffs including meat and processed plant resources have been hypothesized as evolutionary pressures leading to dental reduction in hominins ( Zink and Lieberman , 2016 ) . The small dentition of H . naledi manifests this adaptive strategy to a greater extent than H . habilis , H . rudolfensis and most H . erectus samples ( Berger et al . , 2015; Hawks et al . , 2017 ) , though without the predicted encephalization . What tools did H . naledi make ? Its lineage may have existed across much or all of the time during which African hominin populations were manufacturing Acheulean and possibly even Oldowan assemblages ( e . g . Mcbrearty and Brooks , 2000 ) . The H . naledi lineage also existed during at least the first half of the MSA , which as an archaeological category seems to have commenced more than 400 ka in several instances in subequatorial and northeastern Africa ( Dusseldorp et al . , 2013; Wilkins and Chazan , 2012; McBrearty and Tryon , 2006; Mcbrearty and Brooks , 2000 ) . Many previous workers have grappled with the question of which hominin species were the makers of Early Stone Age industries ( e . g . Foley , 1987; Susman , 1991; Domalain et al . , 2016 ) . A key part of these considerations has been the role of brain size and behavioural ecology in sustaining traditions , which have supported the role of larger-brained H . habilis and H . erectus as toolmakers and have downplayed the possibility that small-brained Paranthropus may likewise have innovated ( e . g . Hopkinson et al . , 2013; Domalain et al . , 2016 ) . With some exceptions ( e . g . Stringer , 2011 ) , there has been a widespread assumption that MSA traditions were made by modern humans or their ancestors , whether denoted as ‘archaic H . sapiens’ or as a precursor such as ‘H . helmei’ ( Mcbrearty and Brooks , 2000; Lahr and Foley , 2001; Stringer , 2002; Henshilwood and Marean , 2003; Henshilwood and Marean , 2006; Dusseldorp et al . , 2013 ) . MSA variants are characterized by the manufacture of blades , by the presence of the Levallois flaking technique and of hafted implements , at some locations by the use of pigments , and by a lack of emphasis on large cutting tools such as the handaxes and cleavers of the Acheulean industry ( e . g . Mcbrearty and Brooks , 2000; Henshilwood and Marean , 2003; Marean and Assefa , 2005; Henshilwood and Marean , 2006 ) . Some of these technical innovations have even been considered as markers of modern human behaviour . However , it is now clear that the populations of subequatorial Africa had deep prehistoric divisions ( Stringer , 2016; Lachance et al . , 2012; Hsieh et al . , 2016 ) and that multiple genetically and morphologically divergent hominin populations probably created Acheulean and MSA archaeological traditions . This situation is paralleled outside of Africa , where most of the manufacturing techniques that characterize the MSA were also mastered by Neanderthals and possibly by Denisovans ( Roebroeks and Soressi , 2016; d’Errico and Banks , 2013 ) . These archaic populations diverged from African populations well before the appearance of such techniques either in Africa or in Eurasia ( Meyer et al . , 2016 ) , so these techniques must either have been invented independently multiple times or have been transferred by long-distance exchange of ideas across long-separated hominin populations . H . naledi existed contemporaneously with MSA archaeological industries across subequatorial and northeastern Africa ( Mcbrearty and Brooks , 2000; Henshilwood and Marean , 2003 , Marean , 2006; Marean and Assefa , 2005; Henshilwood and Marean , 2006; McBrearty and Tryon , 2006; Wilkins and Chazan , 2012; Dusseldorp et al . , 2013; Wurz , 2013 ) . Excavations in the Rising Star cave system have not yet uncovered artifacts in direct association with H . naledi . But considering the weak nature of the fossil hominin record , H . naledi may be the only hominin definitely known to be present during at least the early part of the MSA in the highvelt region of southern Africa ( Dusseldorp et al . , 2013 ) . Considering the context , it is possible that H . naledi sustained MSA traditions . Without extraordinary evidence , we cannot uncritically accept that such a broadly defined archaeological tradition was the exclusive product of a single population across Africa . Did H . naledi deliberately deposit bodies within the Rising Star cave system ? With respect to the deposition of the fossil material , it is appropriate to adopt a null hypothesis that the remains entered the Dinaledi and Lesedi Chambers without intentional hominin mediation , and to see whether the evidence can reject that hypothesis . We have previously examined depositional scenarios on the basis of evidence from the Dinaledi Chamber ( Dirks et al . , 2015 , 2016; Randolph-Quinney et al . , 2016 ) . The discovery of hominin material in the Lesedi Chamber adds a second instance of deposition of hominin skeletal material within the cave system . Some other cave systems in the Cradle of Humankind area likewise present evidence of multiple episodes of the deposition of hominin remains . Swartkrans has a complex series of infills that contain hominin and a broad array of macrofaunal remains , many of which bear evidence of carnivore or scavenger activity representing multiple accumulating agents ( Pickering et al . , 2004a ) . Further , evidence of cutmarks , percussion marks , and burned bone show that hominins were an accumulating agent of some Swartkrans faunal remains ( Pickering et al . , 2005 ) . Sterkfontein is another cave system that has a complex series of infills , in which much bone material bears traces of carnivore and scavenger activity . Within Sterkfontein , the Silberberg Grotto is a deep chamber that contains one hominin skeleton ( StW 573 ) together with faunal remains that appear to have fallen from above; it is a death trap ( Pickering et al . , 2004a ) . Also within the Sterkfontein system , the Jacovec Cavern breccia presents some evidence for water transport of material from the surface and water sorting of bone ( Kibii , 2007 ) . These examples provide several hypotheses for the deposition of hominin skeletal remains that do not involve intentional behaviour by the hominins themselves , and we have previously examined whether the Dinaledi Chamber evidence is compatible with any of them ( Dirks et al . , 2015 , 2016 ) . While geological and sedimentological studies of the Lesedi Chamber are still ongoing , we can consider how its taphonomic situation resembles the Dinaledi Chamber material . In the Dinaledi Chamber , the skeletal material showed invertebrate surface modification but a complete lack of markings from carnivores , scavengers , or hominins ( Dirks et al . , 2015 , 2016 ) . The Lesedi Chamber hominin material likewise presents no evidence of cutmarks , tooth marks , scoring , puncture marks , gnawing or bone cylinders , and only shows surface markings consistent with abrasion or pitting , many after the deposition of manganese and iron oxide coatings on the bones ( Hawks et al . , 2017 ) . These observations seem to exclude carnivores and scavengers as the primary accumulating agents for the assemblages . The Dinaledi Chamber is enormously challenging to reach today , and both sedimentological and geological evidence supports the hypothesis that the chamber itself and the entry chute from the neighboring Dragon’s Back Chamber had substantially the same configuration at the time at which the H . naledi skeletal remains entered ( Dirks et al . , 2015 , 2016 ) . Some have questioned whether one or more alternative entrances to the Dinaledi Chamber may once have existed , which might have made the physical situation much easier for H . naledi to enter the chamber from the outside ( Val , 2016; Thackeray , 2015 ) . But any such entrance would have needed to replicate most of the constraints of the present entrance , or else it would not produce the sedimentological distinctiveness of the Dinaledi Chamber or the lack of non-hominin macrofauna ( Dirks et al . , 2016; Randolph-Quinney et al . , 2016 ) . The situation in the Lesedi Chamber makes these constraints of the Dinaledi Chamber even more apparent . The Lesedi Chamber is similarly situated deep inside the cave system , far inside the dark zone , with no nearby surface entrance ( Hawks et al . , 2017 ) . However , no strong physical constraint prevents macrofauna , at least those smaller than humans , from entering . Faunal material in the chamber demonstrates that at least the remains of small carnivores and smaller fauna did enter the Lesedi Chamber , even though it is deep in the cave , well within the dark zone . Although we do not know the timing or manner in which these faunal elements entered the Lesedi Chamber , their presence reinforces the importance of physical constraints in impeding entry into the Dinaledi Chamber , where no such faunal remains have been found ( Dirks et al . , 2015 ) . Further sedimentological and geological assessment of the Lesedi Chamber , and direct dating of the faunal and hominin remains , may clarify the relation of faunal and hominin remains . Val ( 2016 ) proposed that the hominin skeletal material from the Dinaledi Chamber may have been transported from another location within the cave system , which we have not located , but which might itself have been consistent with carnivore accumulation or a death trap from the surface . In Sterkfontein , there may have been redeposition of sediments from higher chambers into the Silberberg Grotto ( Kramers and Dirks , 2017 ) , providing a possible example a process driven by gravity from above , although the StW 573 skeleton itself appears to be in near-primary context . No openings in the ceilings above the Dinaledi or Lesedi Chambers appear consistent with the gravity-driven transport of material from directly above . The Dinaledi Chamber skeletal material shows no evidence of high-energy fluvial transport , which would have been necessary to move such a quantity of bone any horizontal distance through the cave ( Dirks et al . , 2015 , 2016 ) . The same is true of the remains within the Lesedi Chamber ( Hawks et al . , 2017 ) . In both deposits , there is evidence of postdepositional reworking of sediments , but in both deposits , some articulated remains have been recovered , and neither the skeletal element representation nor the physical condition of the remains are consistent with wholesale secondary redeposition of the hominin assemblages from any third location ( Dirks et al . , 2016; Hawks et al . , 2017 ) . We consider it untenable to hypothesize that both the Dinaledi Chamber and the Lesedi Chamber were accidental death trap situations . We have previously written ( Dirks et al . , 2015 , 2016 ) that the accidental death trap hypothesis was one that the physical evidence from the Dinaledi Chamber might not reject . Still , the evidence that hominin individuals of all ages were deposited in the chamber over some period of time , as well as the sediment composition within the chamber itself , led us to view that hypothesis as less likely . Other cave systems in the region have been hypothesized as death-trap situations , including Malapa ( Dirks et al . , 2010; L'Abbé et al . , 2015; Val et al . , 2015 ) and the Silberberg Grotto of Sterkfontein ( Pickering et al . , 2004a ) . In these instances , a relatively direct vertical route existed either from the surface or from cave chambers above; furthermore , other non-cryptic macrofauna were present among the fossil remains , externally derived sediments were abundant , and ( in the case of Malapa ) plant remains were also present . No such evidence occurs in the Dinaledi Chamber . As we continue to study the geological history of the Rising Star system , it is possible that we will find that H . naledi or small carnivores accessed the Lesedi Chamber differently than excavators today , but we have found no evidence of a nearby vertical entry that would accommodate hominins . The presence of two such situations in the same cave system , with no remaining evidence of a death trap other than the hominin remains , would be unlikely . The evidence in its entirety appears incompatible with the hypothesis of accumulation in both the DInaledi and Lesedi Chambers without some hominin agency . However , with a later Middle Pleistocene date for the Dinaledi Chamber material , we must also consider the suggestion that modern humans or their immediate predecessors were accumulating agents for the H . naledi skeletal material ( e . g . , C . Marean , quoted in Gibbons ( 2015] . This hypothesis would require that the modern humans left no cutmarks or tooth marks on the H . naledi material , and that they treated H . naledi remains differently than those of any other species , including modern humans themselves . Also , although it is possible that H . naledi may have existed in contact with ancestors of modern humans , we have as yet no evidence of this . We will continue to explore the possible interactions of H . naledi and other hominin populations , but they do not appear to be a likely explanation for the deposition of skeletal remains in the Rising Star cave system . We propose that funerary caching by H . naledi is a reasonable explanation for the presence of remains in the Dinaledi and Lesedi Chambers . Mortuary behaviours , while culturally diverse , are universal among modern human cultural groups ( Pettitt , 2010 ) . Such behaviours are not seen in living non-human primates or in other social mammals , but many social mammals exhibit signs of grief , distress , or other emotional response when other individuals within their social group die ( King , 2013 ) . We have no information about whether H . naledi was a symbolic species , although with the possibility that it manufactured MSA toolkits , we do not rule out such abilities . But symbolic cognition is not likely to have been necessary to sustain a repeated cultural practice in response to the physical and social effects of the deaths of group members ( Pettitt , 2010 ) . Such behaviour may have many different motivations , from the removal of decaying bodies from habitation areas , to the prevention of scavenger activity , to social bonding , which are not mutually exclusive . We suggest only that such cultural behaviour may have been within the capabilities of a species that otherwise presents every appearance of technical and subsistence strategies that were common across the genus Homo . Fossil and genetic evidence shows that subequatorial Africa was home to diverse populations of hominins throughout the Pleistocene . The expansive savanna and open woodland habitats of this region have driven biodiversity in many mammals and birds that have similar habitat preferences to hominins ( e . g . Bannerman and Burns , 1953; Kingdon , 2015; Payne , 2013 ) . We suggest that for hominin populations too , subequatorial Africa appears to have been a source of biological diversity and innovation . No paleoanthropologist anticipated that a species like H . naledi existed in this region during the late Middle Pleistocene . However , considering a broad array of biogeographic , phylogenetic , and genetic evidence from humans and other mammals , the discovery of more members of a diverse community of hominin populations in this vast region should no longer be a surprise . This hypothesis should provoke greater examination of the paleoenvironments and regional paleoclimate across this region of Africa . Further , the presence of a diversity of hominin populations throughout most of the Pleistocene must lend caution to how we examine fragmentary specimens . Many of these populations and species are indistinguishable from each other in many parts of the skeleton , despite being very different in others . In particular , we must apply renewed caution to behavioural inferences . The transition to early MSA industries is likely to have involved a broad array of hominin populations and/or species that possibly interacted with each other . Is H . naledi the direct ancestor of humans or of any other hominins ? The later Middle Pleistocene age of the Dinaledi Chamber assemblage is substantially later than the date presently recognized for the first appearance of H . erectus some 1 . 8 million years ago . The Dinaledi occurrence of H . naledi is later in time than the hypothesised genetic divergence of Neanderthal and modern human populations more than 500 , 000 years ago ( Meyer et al . , 2016 ) . But a paleontological view recognizes that any particular set of fossils does not represent the entire time depth of a species or its relationships , and phylogenetic analyses of H . naledi show that the species and its branch must have existed much earlier than the Dinaledi fossils ( Dembo et al . , 2016; Hawks and Berger , 2016; Thackeray , 2015 ) . One analysis of craniodental evidence places H . naledi amid the branches leading to H . habilis , Au . sediba , and H . rudolfensis , suggesting that its anatomical pattern may have been present from the earliest origin of Homo . Another analysis has placed H . naledi as a sister taxon to archaic species of Homo and modern humans , closer to living humans than H . erectus ( Dembo et al . , 2016 ) . If this is true , an early H . naledi population may have been the ancestor of humans , placing H . erectus as a side branch . H . naledi is clearly a primitive species within the genus Homo , despite sharing many derived features with archaic and modern humans . The fossil record for other species attributed to early Homo is presently too incomplete to ascertain whether these species also show such mosaicism , or whether they express different manifestations of primitive and derived morphological patterns . A species like H . naledi might well have given rise repeatedly to other branches of Homo , each derived in a somewhat different way . A fresh look at the hominin fossil record , setting aside a history of linear assumptions about the evolution of H . erectus and H . sapiens , may set a new context for further fossil discoveries . Better analytical techniques , and increased knowledge provided by aDNA , may shed further light on these questions . | Species of ancient humans and the extinct relatives of our ancestors are typically described from a limited number of fossils . However , this was not the case with Homo naledi . More than 1 , 500 fossils representing at least 15 individuals of this species were unearthed from the Rising Star cave system in South Africa between 2013 and 2014 . Found deep underground in the Dinaledi Chamber , the H . naledi fossils are the largest collection of a single species of an ancient human-relative discovered in Africa . After the discovery was reported , a number of questions still remained . H . naledi had an unusual mix of ancient and modern traits . For example , it had a small brain like the most ancient of human-relatives , yet its wrists looked much like those of a modern human . This raised the question: where does H . naledi fit within the scheme of human evolution ? Now , Berger et al . —who include many of the researchers who were involved in the discovery of H . naledi—reconsider this question in the light of new findings reported in two related studies . First , Dirks et al . provide a long-anticipated estimate for the age of the fossils at between 236 , 000 and 335 , 000 years old . Second , Hawks et al . report the discovery of more H . naledi fossils from a separate chamber in the same cave system . These estimated dates fall in a period called the late Middle Pleistocene , and mean that H . naledi possibly lived at the same time , and in the same place , as modern humans . Berger et al . explain that the existence of a relatively primitive species like H . naledi living this recently in southern Africa is at odds with previous thinking about human evolution . Indeed , all other members of our family tree known from the same time had large brains and were generally much more evolved than our most ancient relatives . However , Berger et al . argue that we have only an incomplete picture of our evolutionary past , and suggest that old fossils might have been assigned to the wrong species or time period . Reassessing the old fossils might lead the scientific community to rethink what kinds of human-relative were around in southern Africa at different times , and what those ancient species were capable of . For example , archeologists had previously thought that modern humans made all the stone tools dating from around the late Middle Pleistocene found in southern Africa , but now we must consider whether some of them could have been made by H . naledi . | [
"Abstract",
"Introduction"
] | [
"evolutionary",
"biology",
"short",
"report"
] | 2017 | Homo naledi and Pleistocene hominin evolution in subequatorial Africa |
The HIV-1 Vpr accessory protein induces ubiquitin/proteasome-dependent degradation of many cellular proteins by recruiting them to a cullin4A-DDB1-DCAF1 complex . In so doing , Vpr enhances HIV-1 gene expression and induces ( G2/M ) cell cycle arrest . However , the identities of Vpr target proteins through which these biological effects are exerted are unknown . We show that a chromosome periphery protein , CCDC137/cPERP-B , is targeted for depletion by HIV-1 Vpr , in a cullin4A-DDB1-DCAF1 dependent manner . CCDC137 depletion caused G2/M cellcycle arrest , while Vpr-resistant CCDC137 mutants conferred resistance to Vpr-induced G2/M arrest . CCDC137 depletion also recapitulated the ability of Vpr to enhance HIV-1 gene expression , particularly in macrophages . Our findings indicate that Vpr promotes cell-cycle arrest and HIV-1 gene expression through depletion of CCDC137 .
Human and simian immunodeficiency viruses ( HIV-1 , HIV-2 and SIVs ) encode several accessory proteins; Vpr , Vpx , Vif , Nef , and Vpu . While accessory proteins are often dispensable for replication in immortalized cell lines , they are important in a physiological context and typically act by removing or displacing molecules that are deleterious to virus replication . Among the HIV-1 accessory proteins , the function of ~14 kDa Viral Protein R ( Vpr ) remains the most enigmatic . Replication deficits of inconsistent magnitude are evident in HIV-1 mutants lacking Vpr , particularly in primary macrophages ( Balliet et al . , 1994; Connor et al . , 1995; Fouchier et al . , 1998 ) , while deletion of Vpr from SIVmac modestly attenuates pathogenesis ( Hoch et al . , 1995; Gibbs et al . , 1995 ) . Vpr shares a common ancestor with an HIV-2/SIV accessory protein , Vpx . Both proteins bind to VprBP ( DCAF1 ) and in so doing recruit the cullin 4A-containing E3 ubiquitin ligase complex ( CRL4 ) ( Zhang et al . , 2001; Srivastava et al . , 2008; DeHart et al . , 2007; Hrecka et al . , 2007; Le Rouzic et al . , 2007; Schröfelbauer et al . , 2007; Tan et al . , 2007; Wen et al . , 2007 ) . Both Vpr and Vpx proteins are incorporated into virions through an interaction with the virion structural protein Gag ( Kondo et al . , 1995; Kewalramani et al . , 1996 ) . Recruitment of CRL4 by virion-associated Vpx and some SIV Vpr proteins can induce the degradation of the antiviral protein SAMHD1 shortly following viral entry ( Hrecka et al . , 2011; Laguette et al . , 2011; Spragg and Emerman , 2013 ) . However , HIV-1 , HIV-2 and many SIV Vpr proteins do not exhibit SAMHD1-depleting activity . Rather , HIV-1 Vpr mediated CRL4 recruitment has different biological effects , including G2/M cell-cycle arrest of infected cells ( Connor et al . , 1995; DeHart et al . , 2007; Hrecka et al . , 2007; Le Rouzic et al . , 2007; Schröfelbauer et al . , 2007; Tan et al . , 2007; Wen et al . , 2007; Jowett et al . , 1995; Rogel et al . , 1995; Belzile et al . , 2007 ) and activation of the ATR ( ataxia-telangiectasia and Rad3-related ) -mediated DNA damage response ( DDR ) ( Roshal et al . , 2003; Zimmerman et al . , 2004; Fregoso and Emerman , 2016 ) . HIV-1 gene expression is also enhanced by Vpr in some contexts , and similarly enhanced in cells arrested in G2/M ( Connor et al . , 1995; Goh et al . , 1998; Yao et al . , 1998; Gummuluru and Emerman , 1999 ) . Numerous cellular proteins including UNG2 , the SLX4 complex , helicase-like transcription factor ( HLTF ) , survival of motor neuron-1 ( SMN1 ) , cell division cycle associated 2 ( CDCA2 ) , and zinc finger protein 267 ( ZNF267 ) have been reported to be depleted by Vpr ( Laguette et al . , 2014; Hrecka et al . , 2016; Lahouassa et al . , 2016; Greenwood et al . , 2019 ) , However , while depletion of some of these proteins has been reported to have modest effects on cell cycle ( Greenwood et al . , 2019 ) , none of the previously identified Vpr target proteins have been demonstrated to be solely responsible for the G2/M-arrest and gene-expression enhancement effects of HIV-1 Vpr . Indeed , the identity of the host proteins whose Vpr-induced depletion induces cell cycle arrest , or otherwise facilitates viral propagation , remains controversial . Here , we identify a Vpr target protein , cPERP-B , also known as coiled-coil domain-containing-137 ( CCDC137 ) , whose depletion causes G2/M cell cycle arrest and enhances HIV-1 gene expression , thus recapitulating the salient biological effects of HIV-1 Vpr .
To identify HIV-1 Vpr target proteins , we used a proximity-dependent method ( Roux et al . , 2012 ) in which a biotin-ligase , BirA ( R118G ) fused an HIV-1NL4-3 Vpr bait was expressed in proteasome inhibitor treated cells . Biotinylated proteins were enriched using streptavidin magnetic beads , identified using mass spectroscopy and the biotinylated proteome in BirA ( R118G ) -Vpr , and BirA ( R118G ) expressing cells was compared . Multiple nuclear proteins were biotinylated in BirA ( R118G ) -Vpr , but not BirA ( R118G ) expressing , cells ( Figure 1—figure supplement 1A , Supplementary file 1 ) . Notably , Ki-67 , a proliferating cell marker , was the top ‘hit’ , with >90 Ki-67 peptides detected in replicate experiments . However , Vpr did not induce Ki-67 depletion ( Figure 1—figure supplement 1B , C ) and shRNA-induced Ki-67 depletion did not induce G2/M arrest ( Figure 1—figure supplement 1D , E ) , strongly suggesting that it is not a Vpr target protein . Based on the fact that Ki-67 was the top hit in our proximity biotinylation experiments but did not itself appear to be responsible for the Vpr-induced cell cycle arrest , we hypothesized that a Ki-67-proximal or interacting protein might represent the genuine target of Vpr . Ki-67 recruits a group of proteins termed ‘chromosome periphery proteins’ ( cPERPs ) , that localize within the nucleus , primarily the nucleolus , during interphase but are relocalized to chromosome peripheries during mitosis ( Booth et al . , 2014; Ohta et al . , 2010 ) . Therefore , we next conducted a focused screen of candidate target proteins that were either prominent hits in the BirA ( R118G ) -Vpr screen ( Supplementary file 1 ) , were nucleolar or nuclear proteins , members of the cPERP group , and/or were reported to bind Ki-67 . Of numerous candidates tested in transient co-transfection/western blot assays , Vpr only induced the depletion of cPERP-B , also termed CCDC137 ( Ohta et al . , 2010; Figure 1—figure supplement 2A–C ) . To assess the potency with which Vpr could induce depletion of CCDC137 , we cotransfected 293 T cells with various amounts of CCDC137 and HIV-1 Vpr expression plasmids . Both CCDC137 and Vpr proteins were tagged with the same epitope , a single copy of an HA tag , to assess their relative steady-state levels . This analysis revealed that co-expression of a barely detectable amount of Vpr resulted in the removal of much larger quantities of CCDC137 , underscoring the potency with which Vpr induced CCDC137 depletion ( Figure 1A , B , Figure 1—figure supplement 2D ) . Treatment of cells with MG132 abolished the ability of Vpr to induce CCDC137 depletion , suggesting that Vpr triggered proteasome-dependent CCDC137 degradation ( Figure 1C ) . Furthermore , when DCAF1 , an essential component of the CRL4 complex , was depleted by lentiviral vector-mediated RNA interference , the depletion of CCDC137 was abrogated ( Figure 1D ) . We set out to mapped the degrons through which CCDC137 is depleted by Vpr . First , CCDC137 was expressed as N-terminal ( residues 1–154 ) and C-terminal ( residues 155–289 ) fragments . Transfection experiments revealed that the isolated N-terminal but not the C-terminal portion was depleted by Vpr ( Figure 2—figure supplement 1A ) . Alanine scanning mutagenesis in the context of full length CCDC137 , whereby blocks of 5 residues were mutated throughout the N-terminal CCDC137 portion revealed that CCDC137 residues 61 to 75 were important for Vpr-induced depletion ( Figure 2A , Figure 2—figure supplement 1B ) . Additionally , alanine substitutions of an LxxLL motif ( positions 228–232 ) through which CCDC137 binds nuclear receptors ( Youn et al . , 2018 ) , also reduced Vpr-induced CCDC137 depletion ( L/A , mutation , Figure 2A ) . Combined substitution of CCDC137 residues 61–65 or 66–70 , coupled with the L/A ( 228-232 ) mutation caused CCDC137 to be substantially Vpr-resistant ( Figure 2A ) . As previously reported , CCDC137 formed foci in the nucleus that colocalized with fibrillarin , a nucleolar marker , during interphase ( Ohta et al . , 2010; Figure 2B ) and this property was unaffected by the aforementioned residue 61 to 65 , 66 to 70 , or the L/A ( 228-232 ) Vpr resistance-inducing substitutions ( Figure 2C ) . To determine whether Vpr and CCDC137 were physically associated , we co-expressed a glutathione S-transferase ( GST ) -Vpr fusion protein with CCDC137 in 293 T cells . Wild-type CCDC137 , but not a Vpr-resistant CCDC137 mutant ( L/A 66–70 ) , could be co-precipitated with GST-Vpr ( Figure 3A ) when coexpressed in 293 T cells . Conversely , a mutant GST-Vpr ( Q65R ) that did not cause CCDC137 depletion ( Figure 3—figure supplement 1 ) did not co-precipitate CCDC137 ( Figure 3A ) . Similarly , V5-tagged CCDC137 could be co-immunoprecipitated with HA-tagged Vpr , but other control proteins , including Ki-67-interacting protein ( MKI67IP ) and nucleolar-localized ribosomal RNA processing protein 1 ( RRP1 ) , did not coimmunoprecipitate with Vpr ( Figure 3B ) . When the biotinylation proximity assay ( Roux et al . , 2012 ) was coupled to Western blot analysis , CCDC137 was found to be biotinylated by Vpr-BirA ( R118G ) but not BirA ( R118G ) despite our inability to detect biotinylated CCDC137 in the initial mass spectroscopic screens ( Figure 3C ) . We infected 293T or U2OS cells with a minimal version of HIV-1 ( termed V1 ) in which Gag , Pol , Vif , Vpu and Env carry deletions or nonsense mutations . In V1 infected cells , Tat , Rev , HA-tagged Vpr and GFP are expressed , all driven by native HIV-1 LTR sequences , and infected cells can be identified by flow cytometry or microscopy ( Figure 4—figure supplement 1 ) . Infection with V1 lacking Vpr ( V1/δ-Vpr ) had no effect on endogenous CCDC137 levels , while infection with V1/HA-Vpr , encoding Vpr from one of several HIV-1 laboratory adapted or primary transmitted founder strains , caused profound CCDC137 depletion ( Figure 4A ) . HIV-2 Vpr induced partial CCDC137 depletion while SIVMAC Vpr did not affect CCDC137 levels ( Figure 4A ) , consistent with their abilities , or lack thereof in the case of SIVmac , to induce G2M arrest ( Figure 4—figure supplement 2 ) . Immunofluorescent staining of endogenous CCDC137 in 293 T cells , or ectopically expressed V5-tagged CCDC137 in U2OS cells , showed that CCDC137 was diminished to nearly undetectable levels in V1/HA-Vpr infected ( GFP+ ) cells compared to neighboring , uninfected ( GFP- ) cells or V1/δ-Vpr infected cells ( Figure 4B , Figure 4—figure supplement 3 ) . To ascertain whether CCDC137 depletion might be responsible for Vpr-induced cell cycle arrest , we next asked whether depletion of CCDC137 per se could induce G2/M cell cycle arrest . Lentiviral constructs expressing CAS9 and CCDC137-targeting CRISPR guide RNAs efficiently generated CCDC137 knockout alleles , but most transduced cells died and none of the numerous surviving cell clones that were analyzed contained frameshifting indels in both copies of CCDC137 , suggesting that CCDC137 is essential for proliferating cell viability . Two CCDC137-targeting , lentiviral shRNA vectors encoding a puromycin N-acetyl-transferase caused effective short term CCDC137 depletion after puromycin selection of 293 T cells , with CCDC137/shRNAII causing more profound depletion than CCDC137/shRNAI ( Figure 5A ) . Notably , CCDC137/shRNA transduced cells accumulated in G2/M ( Figure 5B ) as revealed by propidium iodide ( PI ) staining and the extent of CCDC137 depletion and G2/M accumulation were correlated ( Figure 5A , B ) . A CCDC137 cDNA construct lacking the 3’UTR targeted by CCDC137/shRNAII substantially rescued G2/M arrest ( Figure 5B ) . As an alternative way to assess the effect of CCDC137 knockdown on cell cycle , we utilized a Fluorescent Ubiquitination-based Cell Cycle Indicator ( FUCCI ) construct ( Sakaue-Sawano et al . , 2008 ) , which couples the cell-cycle regulated proteins with fluorescent proteins . G2/M accumulation was also evident upon CCDC137 depletion in several U2OS cell clones expressing mClover or mKusabira-Orange2 ( mKO2 ) fluorescent proteins fused to geminin ( 1–110 aa ) that are depleted during G1 but present during G2/M ( Figure 5C , Figure 5—figure supplement 1 ) . Accordingly , live cell imaging of U2OS/mClover-hGeminin ( 1–110 aa ) cells revealed fluctuating fluorescence that disappeared upon cell division , while CCDC137 depleted cells did not divide and retained mClover-hGeminin fluorescence , indicating G2/M growth arrest until apparent cell death ( Figure 5—video 1 ) . HIV-1 and HIV-2 Vpr are able to activate the DNA damage response ( DDR ) which activates response pathways through ATM/ATR and Chk1/2 kinases , leading to cell cycle arrest ( Roshal et al . , 2003; Zimmerman et al . , 2004; Fregoso and Emerman , 2016 ) . As part of the Vpr-induced DNA damage response , histone H2A variant H2AX , a marker for DNA damage , undergoes phosphorylation at Ser 139 ( γ-H2AX ) and forms nuclei foci . CCDC137 depletion using shRNA also caused accumulation of nuclear foci of γ-H2AX ( Figure 5D ) , mimicking the reported Vpr-induced DDR . Importantly , the CCDC137 depletion-induced DDR was rescued by expression of a shRNA-resistant CCDC137 cDNA ( Figure 5D ) . If Vpr induces cell-cycle arrest through depletion of CCDC137 , we reasoned that Vpr-induced cell-cycle arrest should be alleviated by overexpression of the Vpr-resistant CCDC137 mutants . U2OS cells containing doxyxcycline inducible CCDC137 expression constructs were treated with doxycycline to induce the expression of CCDC137 and then infected with V1/HA-Vpr or V1/δ-Vpr at an MOI of 2 for analysis of protein levels ( Figure 6A ) , or an MOI of 0 . 5 for analysis of DNA content ( Figure 6B ) . Wild-type CCDC137 was depleted following V1/HA-Vpr infection while mutant CCDC137 ( L/A 66–70 ) largely resisted Vpr-induced depletion ( Figure 6A ) . Crucially , overexpression of WT CCDC137 partly ameliorated the G2/M arrest effect of Vpr , while expression of the Vpr-resistant CCDC137 ( L/A 66–70 ) mutant conferred nearly complete resistance to Vpr-induced G2/M arrest ( Figure 6B , C ) , suggesting that depletion of CCDC137 is necessary for G2/M cell cycle arrest induction by Vpr . Prior work has shown that Vpr enhances HIV-1 HIV-1 gene expression in a variety of cell types ( Connor et al . , 1995; Goh et al . , 1998; Gummuluru and Emerman , 1999 ) . Both of these phenotypes have also been associated with G2/M arrest properties of Vpr . We therefore assessed the effects of Vpr expression or CCDC137 depletion on HIV-1 gene expression in various cell types . Live-cell imaging of U2OS/mClover-hGeminin cells infected with V1/δ-Vpr/mCherry resulted in accumulating mCherry fluorescence and fluctuating mClover-hGeminin fluorescence , indicative of a normal cell cycle . Conversely , V1/HA-Vpr/mCherry infection induced accumulation of mClover fluorescence in infected , mCherry+ cells ( Figure 7—video 1 ) . Notably , the level of mCherry fluorescence was clearly greater in V1/HA-Vpr/mCherry infected cells than in V1/δ-Vpr/mCherry infected cells ( Figure 7—video 1 ) . Moreover , live-cell imaging of U2OS/mClover-hGeminin cells transduced with a CCDC137-depleting shRNA vector ( CCDC137/shRNAII , Figure 5A ) and then infected with V1/δ-Vpr/mCherry , showed that CCDC137 depletion using shRNA recapitulated the effect of Vpr , causing both G2M arrest and increased HIV-1 gene ( mCherry ) expression ( Figure 7—video 2 ) . This finding was reproduced using FACS analysis of V1/δ-Vpr/GFP-infected U2OS cells . Specifically , prior transduction of U2OS cells with a CCDC137-depleting shRNA vector ( CCDC137/shRNAII , Figure 5A ) caused elevated GFP expression in infected cells ( Figure 7A , B ) . While Vpr expression caused CCDC137 depletion and increased HIV-1 gene expression in U2OS cells , these cells are not natural targets of HIV-1 . Therefore , we next examined the effects of Vpr on CCDC137 and HIV-1 gene expression in primary cells . Immunofluorescence assays in V1/HA-Vpr and V1/δ-Vpr infected primary macrophages showed that Vpr effectively depleted CCDC137 therein while the Q65R mutant Vpr did not ( Figure 8A , Figure 8—figure supplement 1 ) . Infection of primary CD4+ T-cells with V1/HA-Vpr and V1/δ-Vpr revealed that the presence of Vpr resulted in ~2 fold higher levels of GFP expression than did in its absence ( Figure 8B , C ) . The effect of Vpr on HIV-1 gene expression was more pronounced in primary macrophages , where V1/HA-Vpr but not V1/HA-Vpr ( Q65R ) resulted in higher levels of GFP expression than did infection with V1/δ-Vpr , even though marked donor-to-donor variation was evident ( Figure 8D , E , Figure 8—figure supplement 1 , Figure 8—videos 1 , 2 ) . Several different Vpr proteins from primary HIV-1 strains caused increased HIV-1 gene expression in macrophages ( Figure 8F ) and similar Vpr-induced enhancement of HIV-1 gene expression was evident in macrophages infected with a full-length reporter virus ( HIV-1NHG , Figure 8—figure supplement 2 ) , as previously reported ( Connor et al . , 1995 ) . Notably , the Vpr-induced increase in GFP levels in V1 infected macrophages was accompanied by elevated HIV-1 RNA levels , as assessed by in-situ hybridization ( Figure 8G , H ) indicating that Vpr enhances viral gene expression primarily by enhancing RNA synthesis or increasing RNA stability . To enable Vpr-independent CCDC137 depletion in infected primary cells , while simultaneously measuring HIV-1 gene expression , we constructed a derivative of V1/δ-Vpr carrying an shRNA expression cassette ( V1/sh , Figure 9—figure supplement 1 ) . Infection with V1/sh enabled effective Vpr-independent CCDC137 depletion in infected macrophages ( Figure 9A , Figure 9—figure supplement 2 ) . Importantly , Vpr-independent CCDC137 depletion in V1/sh infected primary CD4+ T-cells resulted in an enhancement of HIV-1 gene expression ( Figure 9B , Figure 9—figure supplement 3A ) . More strikingly , Vpr-independent CCDC137 depletion in V1/sh infected macrophages caused pronounced enhancement of HIV-1 reporter gene expression as assessed by live cell imaging or FACS assays ( Figure 9C , Figure 9—figure supplements 2 and 3B , C , Figure 9—videos 1 , 2 ) . This enhancing effect of CCDC137 depletion on HIV-1 gene expression was also evident when HIV-1 RNA levels were assessed by in-situ hybridization assays using probes targeting GFP ( Figure 9D , E ) . Similarly , qRT-PCR ( Figure 9F ) assays of GFP and Gag mRNA levels in multiple donors , indicated that CCDC137 depletion recapitulated the effect of Vpr on HIV-1 transcription or RNA stability . Overall , in three different cell types , shRNA-driven CCDC137 depletion had similar enhancing effects on HIV-1 gene expression as did Vpr expression .
A persistent paradox in the study of Vpr is that Vpr-defective viruses are often selected in long term replication experiments or in chronically infected cells ( Jowett et al . , 1995; Rogel et al . , 1995 ) . Presumably therefore , Vpr is deleterious to HIV-1 replication in these contexts , despite its positive effects on HIV-1 gene expression . This effect seems likely to be explained the interplay of infected cell life span and burst size ( Goh et al . , 1998 ) . In cell culture , the longevity of an infected cell plays a crucial role in determining burst size , and thus the effect of Vpr in curtailing the lifespan of an infected cell is expected to confer a competitive disadvantage . Conversely , the life span of infected T-cells in vivo is likely limited by factors other than Vpr and even a modest Vpr-induced enhancement in viral gene expression should confer advantage . Macrophages are not cycling cells , and lack ATR , Rad17 and Chk1 ( Zimmerman et al . , 2006 ) . Thus , G2/M arrest and DDR induction by Vpr are likely not relevant therein . We confirmed herein that the Vpr-induced enhancement of HIV-1 gene expression is particularly evident in macrophages ( Connor et al . , 1995 ) . We surmise that the key consequence of Vpr-induced CCDC137 depletion is enhancement of HIV-1 gene expression , with G2/M arrest and DDR in cycling cells perhaps representing an epiphenomenon . While this work was in progress , Greenwood et al reported the effects of HIV-1 infection on the cellular proteome and found that many proteins were depleted from HIV-1 and HIV-2 infected cells , in a Vpr-dependent manner ( Greenwood et al . , 2019 ) . Consistent with our findings , CCDC137 was among the proteins found to be depleted in HIV-1 and HIV-2 infected cells . While CCDC137 was not investigated further , Greenwood et al reported that depletion of other apparent Vpr targets , ( SMN1 , CDCA2 and ZNF267 ) caused a degree of G2/M arrest . While we cannot exclude the possibility that additional target proteins contribute to the G2/M arrest properties of Vpr , the data presented herein suggests that CCDC137 represents the dominant mediator of the G2/M arrest and HIV-1 gene expression effects of HIV-1 Vpr . Further work , including a detailed analysis of Vpr mutants with distinct target protein specificities should help delineate the relative contributions of various target proteins to Vpr-associated phenotypes . Of note , other viral genes , in particular Vif , are able to affect the cellular environment , including cell cycle perturbation . To exclude the influence of other viral genes , and to facilitate shRNA experiments in primary cells , we employed a minimal version of HIV-1 in most experiments to study the effects of Vpr on cell cycle and gene expression in the target cells . However , we note that the effects of Vpr on gene expression in macrophages were recapitulated with a near full length viral construct , and are consistent with data obtained using a different near full length viral reporter construct reported by Connor et al . , 1995 . That Vpr associates with a cPERP protein is consistent with prior findings that Vpr binds to chromatin and , with VprBP , forms chromatin-associated nuclear foci , a property that is associated with G2/M cell cycle arrest ( Lai et al . , 2005; Belzile et al . , 2010 ) . CCDC137 is a poorly characterized protein - little is known about its function . CCDC137 has been reported to sequester retinoic acid receptor ( RAR ) to the nucleolus and thus is dubbed RaRF ( Retinoic acid Resistance Factor ) ( Youn et al . , 2018 ) but whether this property is relevant to HIV-1 gene expression is unknown . Further work will be required to discern the mechanistic details of how CCDC137 affects G2/M transition and inhibits HIV-1 gene expression . Nevertheless , the findings reported herein reveal an important aspect of how HIV-1 manipulates host cells to facilitate its replication .
Monoclonal antibodies used herein included anti-FLAG ( Sigma ) , anti-HA ( BioLegend anti-HA . 11 ) , anti-HIV capsid p24 ( 183-H12-5C , NIH AIDS Research and Reference Reagent Program ) , anti-Ki67 ( abcam , ab16667 ) , anti-tubulin ( Sigma , T9026 ) , and anti-V5 ( Invitrogen , R960-25 ) . Polyclonal antibodies were purchased from ThermoFisher ( anti-V5 , PA1-993 ) , or Abcam ( anti-CC137 , ab185368 or ab183864; anti-gamma H2AX phosphor S139 , ab11174; anti-Fibrillarin , ab5821 ) . Secondary antibodies included goat anti-mouse or anti rabbit IgG conjugated to Alexa Fluor 488 or Alexa Fluor 568 ( Invitrogen ) for immunostaining , or IRDye 800CW and IRDye 680 , or IRDye 680RD Streptavidin ( LI-COR Biosciences ) for Western blot analysis . An HA-epitope was fused in-frame at the N-terminus of HIV-1NL4-3 Vpr and subcloned into pCR3 . 1 for transient cotransfection experiments . To express Vpr driven by an HIV-1 long terminal repeat ( LTR ) in the context of a viral genome , a proviral plasmid was generated harboring a minimal viral genome ( V1 ) ( Zennou and Bieniasz , 2006 ) , engineered from HIV-1NL4/3 ( R7/3 ) harboring large deletions or inactivating mutations in ( Gag , Pol , Vif , Vpu and Env ) and in which Nef is replaced with GFP . Sequences encoding an HA epitope were fused in frame at N-terminus of Vpr to generate V1/HA-Vpr while V1/δ-Vpr was constructed by deletion of the nucleotide sequences between the cPPT/FLAP and the SalI site within Vpr . In some constructs , the WT HIV-1NL4-3 Vpr was replaced with Vpr-encoding sequences from Q65R NL4-3 Vpr , transmitted HIV-1 founder strains , HIV-2 , SIVAGM Sab , or SIVmac . To construct the V1/HA-Vpr or V1/δ-Vpr expressing mCherry ( V1/mCherry ) , an open reading frame encoding mCherry was amplified , digested with NotI/XhoI , and inserted into V1 to replace GFP . To construct the V1/shCCDC137II carrying an shRNA targeting CCDC137 , the DNA sequence containing the U6 promoter and shRNA targeting CCDC137 ( I: CAGATGCTGCGGATGCTTCT; II: GGTGAAACATGATGACAACA ) was PCR amplified and , after digestion with KpnI , subcloned into V1 carrying inactivating mutations at the 5’ end of Vpr ( ATGGAACAA/GTGGAATAA ) , upstream 5’ of the RRE ( Fig . S6 ) . The control V1/shLuc vector contained the DNA sequence carrying U6 promoter-shRNA targeting luciferase ( CGCTGAGTACTTCGAAATGTC ) at the same position . V1 derivatives expressing BirA ( R118G ) and BirA ( R118G ) -Vpr were constructed by insertion of nucleotide sequences encoding BirA ( R118G ) or BirA ( R118G ) -Vpr fusion proteins into the GFP ( Nef ) position in V1 vector . Plasmids expressing the various human proteins ( CREB1 , CREB3L1 , hnRNP D , hnRNP F , POC1A , PPM1G , CCDC137 , PES1 , WDR18 , WDR74 , MAK16 , NOC2L , RRS1 , NIP7 , hnRNPA1 , hnRNP H , hnRNP K , hnRNP R , hnRNP U , hnRNP C , hnRNP A2B1 , and DHX9 ) with a V5 epitope fused to the C-terminus were from a pCSGW-based human ORF lentiviral library . Plasmid pCR3 . 1 was used to express Bop1 , NCL , MKI67IP , or WDR12 with an Flag epitope fused to their N-termini or to express hnRNP A2B1 or DHX9 with an HA epitope fused to their N-termini . Subsequently , CCDC137 was subcloned into pLNCX2 expression vector with an HA epitope fused to its C-terminus . CCDC137 alanine scanning mutants were generated by overlap-extension PCR amplification , using the wild-type CCDC137-HA expression plasmid as the template and inserted into the pLNCX2 expression vector . All aforementioned plasmids were constructed using PCR , Accession numbers for cDNA sequences and oligonucleotides listed in supplementary file 2 . GST-Vpr expression plasmids were based on pCAGGS and were constructed by PCR amplification of the Vpr coding region from HIV-1NL4-3 which was then inserted in-frame at the 3’ end of GST encoding sequences . To construct a tetracycline-inducible CCDC137 expression vector , wild-type and mutant CCDC137 were amplified , using the corresponding pCR3 . 1 expression plasmids as templates , and inserted into LKO-derived lentiviral expression vector pLKOΔ-puro ( Busnadiego et al . , 2014 ) which also included a puromycin resistance cassette . All cloned coding sequences were verified by DNA sequencing , oligonucleotide sequences used in construction are listed in supplementary file 2 . For knockdown experiments , the lentiviral vector pLKO . 1-TRC ( Moffat et al . , 2006 ) was used to deliver shRNAs . For CCDC137 shRNAs , the lentiviral vector contains two functional elements , shRNA targeting sequences ( I: CAGATGCTGCGGATGCTTCT; II: GGTGAAACATGATGACAACA ) and puromycin-resistance cassette . Human embryonic kidney HEK-293T ( ATCC CRL-3216 ) were maintained in DMEM supplemented with 10% fetal bovine serum ( Sigma F8067 ) and gentamycin ( Gibco ) . Human bone osteosarcoma epithelial cells ( U2OS , ATCC HTB-96 ) were grown in McCoy's 5a Medium Modified ( ATCC 30–2007 ) /10% FCS/gentamycin . MT4 cells ( RRID:CVCL_2632 ) were maintained in RPMI supplemented with 10% fetal bovine serum ( FCS ) and gentamycin . All cell lines used in this study were monitored by SYBR Green real-time PCR RT assay periodically to ensure the absence of retroviral contamination and stained with DAPI to ensure absence of mycoplasma . Cell line identification was documented by the suppliers . To construct cell-cycle reporter cell line , U2OS cells were transduced with a retroviral vector ( pLHCX ) encoding Clover ( a rapidly-maturing green/yellow fluorescent protein ) or mKusabira-Orange2 ( mKO2 , an orange fluorescent protein ) fused to the N-terminus of Geminin 1–110 aa . Single cell clones were isolated after hygromycin selection . Human lymphocytes were prepared from Leukopaks from NY Blood Center by spinning on top of lymphocyte separation medium ( Corning ) . Macrophages were then isolated by plastic adherence and differentiated using GM-CSF ( Thermo Fisher ) . CD4+ T cells were isolated using an EasySep kit ( StemCell ) and maintained in RPMI/10%FCS supplemented with IL2 . For transfection experiments in 293 T cells , cells were seeded at a concentration of 1 . 5 × 105 cells/well ( 24-well plate ) , 3 × 105 cells/well ( 12-well plate ) or 2 × 106 ( 10 cm dish ) and transfected on the following day using polyethylenimine ( PolySciences ) . To test whether Vpr could induce depletion of proteins , 293 T cells in 24-well plates were transfected with 200 ng of pCR3 . 1-based plasmids expressing Flag-tagged proteins or HA-tagged proteins , or pCSGW-based plasmids expressing V5-tagged proteins ( from human ORFs lentiviral library ) , along with increasing amounts ( 0 ng , 25 ng , or 50 ng ) of a pCR3 . 1/HA-Vpr expression plasmid . The total amount of DNA was held constant by supplementing the transfection with empty expression vector . Cells were harvested at 28 hr post transfection and subjected to Western blot analysis . To assess the potency with which HIV-1 Vpr induced CCDC137 depletion , 293 T cells in 24-well plates were transfected with varying amounts ( 0 ng , 100 ng , 200 ng , or 400 ng/well ) of a pCR3 . 1/CCDC137-HA expression plasmid and increasing amounts ( 0 ng , 25 ng , 50 ng , 100 ng , or 200 ng/well ) of a pCR3 . 1/HA-Vpr expression plasmid . The total amount of DNA was held constant by supplementing the transfection with empty expression vector . To generate V1-derived viral stocks , 293 T cells were transfected with 5 μg of pV1-derived proviral plasmids encoding no Vpr or the various HA-tagged Vpr proteins , 5 μg of an HIV-1 Gag-Pol expression plasmid ( pCRV1/GagPol ) and 1 μg of VSV-G expression plasmid into 293 T cells in 10 cm dishes . Virus-containing supernatant was collected and filtered ( 0 . 2 μm ) 2 days later . The lentiviral vectors that transduced CCDC137 cDNAs or shRNAs were similarly generated , except that pLKOΔ-puro or pLKO . 1-TRC-derived plasmids were used in place of pV1-derived plasmids . Ten million MT4 cells were transduced with V1-based constructs expressing BirA ( R118G ) or BirA ( R118G ) -Vpr at an MOI of 2 , treated with 50 µM biotin ( Sigma ) and , after treatment with 10 µM MG132 for 4 hr , cells were harvested 48 hr after transduction . Biotinylated proteins were purified using a previously described protocol ( Roux et al . , 2012 ) . In brief , cells were lysed in lysis buffer ( 50 mM Tris , pH 7 . 4 , 500 mM NaCl , 0 . 4% SDS , 5 mM EDTA , 1 mM DTT , and 1x complete protease inhibitor [Roche] ) and sonicated . After addition of Triton X-100 and further sonication , cell lysates were centrifuged and cleared supernatants were incubated with Dynabeads ( MyOne Steptavadin C1 [Invitrogen] ) for 4 hr . Beads were then washed with 2% SDS , followed by wash thoroughly with buffer 2 ( 0 . 1% deoxycholate , 1% Triton X-100 , 500 mM NaCl , 1 mM EDTA , and 50 mM Hepes , pH 7 . 5 ) , buffer 3 ( 250 mM LiCl , 0 . 5% NP-40 , 0 . 5% deoxycholate , 1 mM EDTA , and 10 mM Tris , pH 8 . 1 ) and buffer 4 ( 50 mM Tris , pH 7 . 4 , and 50 mM NaCl ) . Biotinylated proteins were eluted from the beads with NuPAGE LDS sample buffer supplemented with 200 µM biotin and separated by NuPAGE Bis-Tris Gels . Each gel lane was cut in five bands and subjected to LC-MS/MS analysis ( Proteomics Resource Center , Rockefeller University ) . HEK-293T cells were transiently transfected with plasmids expressing HA-Vpr and V5-tagged human protein factors , and treated with 10 µM MG132 for 4 hr before harvest and lysis with ice-cold lysis buffer ( 50 mM Tris , pH 7 . 4 , 150 mM NaCl , 0 . 5 mM EDTA , 1% digitonin [Sigma] , supplemented with 1X complete protease inhibitor [Roche] ) . After lysis on ice for 10 min , followed by centrifugation at 10 , 000 rpm for 10 min at 4°C , clarified lysates were mixed with 1 µg anti-HA monoclonal antibody and rotated with 30 µl pre-equilibrated Protein G Sepharose 4 Fast Flow resin ( GE healthcare ) for 3 hr at 4°C . The resin was then washed three times with wash buffer ( 50 mM Tris , pH 7 . 4 , 150 mM NaCl ) and the bound proteins were eluted with SDS-PAGE sample buffer and analyzed by Western blotting . Human 293 T cells in 6-well plates were co-transfected with 100 ng of GST or 1 µg of GST-Vpr expression plasmids and 500 ng of HA-tagged CCDC137 expression plasmids . The total amount of DNA was held constant by supplementing the transfection with empty expression vector . Two days later , cells were treated with MG 132 for 4 hr and then lysed in Lysis buffer ( 50 mM Tris , pH 7 . 4 , 150 mM NaCl , 5 mM EDTA , 5% glycerol , 1% Triton X-100 , and 1x complete protease inhibitor [Roche] ) . Cleared lysates were then incubated with glutathione sepharose ( GE healthcare ) for 4 hr at 4°C and , after wash with buffer ( 50 mM Tris , pH 7 . 4 , 150 mM NaCl , 5 mM EDTA , 0 . 1% Triton X-100 ) , bound proteins were eluted in SDS-PAGE sample buffer and subjected to Western blot analysis . Human 293 T cells in 6-well plates were transfected with a V5-tagged CCDC137 expression plasmid and BirA ( R118G ) or BirA ( R118G ) -Vpr expression plasmid . Cells were treated with 50 µM biotin at 24 hr after transfection , and 10 µM MG132 at 40 hr post transfection , and harvested at 44 hr post transfection . Cells were then lysed and cleared lysates were incubated with Dynabeads ( MyOne Steptavadin C1 ) above . After thorough wash , biotinylated proteins were eluted from the beads with SDS-PAGE sample buffer supplemented with 200 µM biotin and subjected to Western blot analysis . Cell lysates and immunoprecipitates were separated on NuPage Novex 4–12% Bis-Tris Mini Gels ( Invitrogen ) , and NuPAGE MES SDS running buffer ( Invitrogen , NP0002 ) was used when Vpr was detected . Proteins were blotted onto nitrocellulose membranes . Thereafter , the blots were probed with primary antibodies and followed by secondary antibodies conjugated to IRDye 800CW or IRDye 680 . Fluorescent signals were detected and quantitated using an Odyssey scanner ( LI-COR Biosciences ) . To determine effects of Vpr on cell cycle , V1 based viral stocks were used to inoculate 2 . 5 × 105 U2OS cells in 6-well plates at an MOI of 0 . 5 to 1 . At 48 hr post-infection , cells were trypsinized , fixed with paraformaldehyde ( PFA ) in phosphate-buffered saline ( PBS ) , washed with PBS , and fixed again in 70% ethanol . After an additional wash with PBS , the cells were resuspended in FxCyclePI/RNase Staining Solution ( Invitrogen ) and incubated at 30°C for 30 min . Flow cytometric analysis was performed using Attune NxT Acoustic Focusing Cytometer ( ThermoFisher Scientific ) . Alternatively , cells were transduced with pLKO . 1-TRC-derived vectors encoding shRNA targeting CCDC137 . After 48 hr , cells were selected in 1 μg ml−1 puromycin or 5 μg ml−1 blasticidin prior to propidium iodide staining , 40 hr later as described above . In some experiments , U2OS cells expressing mKO2-hGeminin ( 1–110 aa ) or mClover-hGeminin ( 1–110 aa ) were transduced with LKO-derived lentiviral vectors encoding shRNAs targeting CCDC137 . After 48 hr , cells were selected in 1 μg ml−1 puromycin prior to FACS analysis , 48 hr later . For the experiment in Figure 6 , U2OS cells expressing doxycycline-inducible CCDC137 were generated by transduction with a LKO-derived lentiviral vector ( Busnadiego et al . , 2014 ) followed by selection in 1 μg ml−1 puromycin . Cells were plated at the density of 2 . 5 × 105 in 6-well plates in the presence of doxycycline and the next day were infected with V1/HA-Vpr at an MOI of 2 or 0 . 5 . At 12 hr post-infection , doxycycline was replenished and at 48 hr post-infection , cells were harvested for Western blot analysis ( for cells infected at high MOI ) or cell cycle analysis ( for cells infected at low MOI ) . U2OS cells expressing V5-tagged CCDC137 were seeded on 3 . 5 cm , glass-bottomed dishes coated with poly-L-Lysine ( MatTek ) . At 48 hr after infection with V1/δ-Vpr or V1/HA-Vpr , cells were then fixed with 4% paraformaldehyde , permeabilized with 0 . 1% Triton X-100 and incubated with mouse anti-V5 monoclonal antibody ( Invitrogen , R960-25 ) and rabbit anti-Fibrillarin polyclonal antibody ( abcam , ab5821 ) followed by goat anti-mouse IgG Alexa fluor-594 conjugate and goat anti-rabbit IgG Alexa fluor-488 conjugate ( Invitrogen ) . Images were captured using an DeltaVision OMX SR imaging system ( GE Healthcare ) . Endogenous CCDC137 in 293 T cells or in macrophages was stained with rabbit anti-CCDC137 polyclonal antibody ( abcam , ab185368 ) using the same procedure . For detection of γ-H2AX foci , U2OS cells were transduced with lentiviruses encoding CCDC137-targeting shRNAs and , after selection with puromycin , cells were seeded on 3 . 5 cm , glass-bottomed dishes coated with poly-L-Lysine ( MatTek ) . Nuclear foci were visualized by immunostaining with rabbit anti-γ-H2AX ( abcam , ab11174 ) followed by a goat anti-rabbit IgG Alexa Fluor-594 conjugate ( Invitrogen ) . A Z-series of images were acquired using an DeltaVision OMX SR imaging system ( GE Healthcare ) . At 48 hr postinfection , macrophages in 8-well chamber slides were washed with PBS ( Ambion ) , fixed with 4% formaldehyde ( Thermo Fisher ) in PBS and permeabilized with 70% ethanol . Then , the cells were washed with Stellaris RNA FISH wash buffer A ( Biosearch Technologies ) and GFP RNA detected by incubation using 0 . 125 μM Cy5-labeled probes in Stellaris RNA FISH hybridization buffer ( Biosearch Technologies ) for 18 hr at 37°C . The cells were then washed twice in Stellaris RNA FISH wash buffer A ( Biosearch Technologies ) with the presence of Hoechst stain during the second wash . Then the cells were washed briefly with Stellaris RNA FISH wash buffer B ( Biosearch Technologies ) , rinsed three times with PBS and subject to imaging by deconvolution microscopy ( DeltaVision OMX SR imaging system ) . All images were generated by maximum intensity projection using the Z project function in ImageJ 1 . 52b . Corrected total cell fluorescence ( CTCF ) was calculated as the following: CTCF = Integrated Density - ( Area of selected cell x Mean fluorescence of background readings ) . The 28 probes against GFP ( cggtgaacagctcctcgc , gaccaggatgggcaccac , gtttacgtcgccgtccag , acacgctgaacttgtggc , gccggtggtgcagatgaa , ggtggtcacgagggtggg , actgcacgccgtaggtca , tcggggtagcggctgaag , agtcgtgctgcttcatgt , ggcatggcggacttgaag , ctcctggacgtagccttc , gccgtcgtccttgaagaa , tcggcgcgggtcttgtag , tgtcgccctcgaacttca , ctcgatgcggttcaccag , tgaagtcgatgcccttca , caggatgttgccgtcctc , cgttgtggctgttgtagt , gcttgtcggccatgatat , gtcctcgatgttgtggcg , gtagtggtcggcgagctg , cgatgggggtgttctgct , ttgtcgggcagcagcacg , gactgggtgctcaggtag , ttggggtctttgctcagg , catgtgatcgcgcttctc , ggtcacgaactccagcag , cttgtacagctcgtccat ) were designed using the Stellaris Probe Designer , version 2 . 0 ( Biosearch Technologies ) . Macrophages ( 6 × 105 ) were infected with V1/sh carrying shRNA targeting luciferase or CCDC137 ( II ) , and 48 hr later RNA was isolated using the NucleoSpin RNA Kit ( Macherey-Nagel ) . RNA levels were determined with Power SYBR Green RNA-to-CT 1-Step Kit using a StepOne Plus Real-Time PCR system ( Applied Biosystems ) . For relative quantification , samples were normalized to actin . The sequences of primers are GAGCGCACCATCTTCTTCAA ( GFP forward ) , TCCTTGAAGTCGATGCCCTT ( GFP reverse ) , GAGCTAGAACGATTCGCAGTTA ( Gag forward ) , CTGTCTGAAGGGATGGTTGTAG ( Gag reverse ) CATGTACGTTGCTATCCAGGC ( actin forward ) and CTCCTTAATGTCACGCACGAT ( actin reverse ) . Relative GFP and Gag expression was calculated as the value of 2^-[ΔCt ( GFP ) - ΔCt ( actin ) ] . To monitor cell cycle and HIV-1 ( V1 ) gene expression infection in living cells , U2OS cells expressing mClover-hGeminin ( 1–110 aa ) or primary macrophages were transferred into glass-bottom dishes and time-lapse microscopy was performed using a VivaView FL incubator microscope ( Olympus ) . In some experiments , cells were transduced with lentiviruses containing shRNA targeting CCDC137 , 36 hr prior to imaging . In some experiments , cells were infected with V1/δ-Vpr or V1/HA-Vpr expressing mCherry or GFP 12 to 24 hr prior to imaging . Images were captured every 30 min using GFP , mRFP and DIC filter sets for up to 72 hr . Preparation of movies was done using MetaMorph software ( Molecular Devices ) as previously described ( Holmes et al . , 2015 ) . Images had a depth of 12 bits , that is , an intensity range of 0–4095 . All data is plotted raw , that is individual values for each individual quantitative determination is plotted . The exception to this is CCDC137/Vpr western blot data in Figure 1B , in which the mean of two independent experiments is plotted , with error bars representing the range of the duplicate raw values . Statistical comparisons between groups in Figures 6C , 8H and 9E , F , G . were done using Graphpad Prism software , and p-values were calculated using a Welch’s t-test or a ratio t-test . | Like all viruses , the human immunodeficiency virus 1 ( HIV-1 ) cannot replicate on its own; to multiply , it needs to exploit the molecular machinery of a cell . A set of HIV-1 proteins is vital in this hijacking process , and they are required for the virus to make more of itself . However , HIV-1 also carries accessory proteins that are not absolutely necessary for the replication process , but which boost the growth of the virus by deactivating the defences of the infected cells . Amongst these proteins , the role of Viral Protein R ( Vpr for short ) has been particularly enigmatic . Previous experiments have shown that , in infected cells , Vpr is linked to several biological processes: it tags for destruction a large number of proteins , it causes the cells to stop dividing , and it encourages them to express the genetic information of the virus . How these different processes are connected and triggered by Vpr is still unknown . It particular , it remains unclear which protein is responsible for these changes when it is destroyed by Vpr . To investigate , Zhang and Bieniasz conducted a series of experiments to spot the proteins that interact with Vpr in human cells . This screening process highlighted a protein known as CCDC137 , which is depleted in cells infected by HIV-1 . To investigate the role of CCDC137 , Zhang and Bieniasz decreased the levels of the protein in human cells . This stopped the cells from dividing , just like during HIV-1 infection . Destroying CCDC137 also mimicked the effects of Vpr on HIV-1 gene expression , increasing the levels of virus proteins in infected cells . Finally , Zhang and Bieniasz made a mutant version of CCDC137 that Vpr could not destroy . When infected cells carried this mutant protein , they kept on dividing as normal . Taken together , these results suggest that Vpr works by triggering the destruction of the CCDC137 protein . Overall , this work represents the first step to understand the role of CCDC137 in both infected and healthy cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2020 | HIV-1 Vpr induces cell cycle arrest and enhances viral gene expression by depleting CCDC137 |
Solar irradiation including ultraviolet ( UV ) light causes tissue damage by generating reactive free radicals that can be electrophilic or nucleophilic due to unpaired electrons . Little is known about how free radicals induced by natural sunlight are rapidly detected and avoided by animals . We discover that Drosophila Transient Receptor Potential Ankyrin 1 ( TRPA1 ) , previously known only as an electrophile receptor , sensitively detects photochemically active sunlight through nucleophile sensitivity . Rapid light-dependent feeding deterrence in Drosophila was mediated only by the TRPA1 ( A ) isoform , despite the TRPA1 ( A ) and TRPA1 ( B ) isoforms having similar electrophile sensitivities . Such isoform dependence re-emerges in the detection of structurally varied nucleophilic compounds and nucleophilicity-accompanying hydrogen peroxide ( H2O2 ) . Furthermore , these isoform-dependent mechanisms require a common set of TRPA1 ( A ) -specific residues dispensable for electrophile detection . Collectively , TRPA1 ( A ) rapidly responds to natural sunlight intensities through its nucleophile sensitivity as a receptor of photochemically generated radicals , leading to an acute light-induced behavioral shift in Drosophila .
Highly reactive chemical compounds are dangerous because of their ability to covalently modify and incapacitate proteins and nucleic acids . Such reactive chemicals are often perceived as noxious and are avoided by animals through chemical nociception mechanisms , in which Transient Receptor Potential ( TRP ) channels play major roles ( Julius , 2013 ) . TRPA1 is the only conserved reactive chemical receptor in the bilateria from humans to flies ( Kang et al . , 2010 ) , and is activated rather than inactivated by covalent modification unlike most proteins ( Macpherson et al . , 2007 ) . Although chemical reactivity can be categorized into two opposite characters , electron-attracting electrophilicity and electron-donating nucleophilicity , only the former has been shown to provoke TRPA1-dependent nociception . Furthermore , there is no molecular mechanism attributed to the sensory detection of nucleophiles , while nucleophilic compounds are widespread in nature as antioxidant phytochemicals ( Lü et al . , 2010 ) and as decomposition gases of animal carcasses ( Dent et al . , 2004 ) , and strong nucleophiles , such as carbon monoxide and cyanide , can be fatal to animals ( Grut , 1954; Krahl and Clowes , 1940 ) . In insects , TRPA1 was originally thought to be a polymodal sensory receptor capable of detecting both temperature increases ( Viswanath et al . , 2003; Hamada et al . , 2008; Corfas and Vosshall , 2015 ) and chemical stimuli ( Kang et al . , 2010; Kwon et al . , 2010 ) . However , this polymodality would limit reliable detection of chemical stimuli when ambient temperature varies . In fact , the TrpA1 genes in D . melanogaster and malaria-transmitting Anopheles gambiae were recently found to produce two transcript variants with distinct 5’ exons containing individual start codons ( Kang et al . , 2012 ) . The two resulting TRPA1 channel isoforms , TRPA1 ( A ) and TRPA1 ( B ) , differ only in their N-termini , and share more than 90% of their primary structure . TRPA1 ( A ) , which is expressed in chemical-sensing neurons , is unable to confer thermal sensitivity to the sensory neurons , allowing TRPA1 ( A ) -positive cells to reliably detect reactive chemicals regardless of fluctuations in ambient temperature . In addition to the insufficient thermosensitivity , TRPA1 ( A ) has been under active investigations for its novel functions , such as the detection of citronellal ( Du et al . , 2015 ) , gut microbiome-controlling hypochlorous acid ( Du et al . , 2016 ) , and bacterial lipopolysaccharides ( Soldano et al . , 2016 ) . Although TRPA1 ( A ) and TRPA1 ( B ) are similarly sensitive to electrophiles ( Kang et al . , 2012 ) , the highly temperature-sensitive TRPA1 ( B ) is expressed in internal AC neurons that direct TrpA1-dependent long-term thermotaxis of the animal ( Hamada et al . , 2008; Ni et al . , 2013 ) , and is thereby inaccessible to reactive chemicals present in the environment . Thus , the functional segregation of TRPA1 isoforms into two distinct sensory circuits is critical for sensory discrimination between thermal and chemical inputs . Photochemical conversion of photonic to chemical energy greatly affects organisms , as is evident in vision , circadian rhythm , and photosynthesis . Low-wavelength solar radiation that reaches the surface of the Earth , generally in the range of ultraviolet ( UV ) to blue light , is a major driving force for such natural photochemical reactions . In contrast to the beneficial effects of photochemistry , the chemical reactivity of free radicals generated by low-wavelength light imposes DNA and tissue damage ( Murphy , 1975; Hannan et al . , 1984 ) and accelerates aging ( Fisher et al . , 1997; Gordon and Brieva , 2012 ) . TRPA1 has been characterized in the bilateria ( Kang et al . , 2010 ) as the molecular receptor for oxidative electrophilic reactivity , as reactive electrophilic compounds activate the non-selective cation channel through covalent modification of key cysteines in the ankyrin repeat domain ( Hinman et al . , 2006; Macpherson et al . , 2007 ) . Despite its electrophile sensitivity , mammalian TRPA1 requires an extremely high UV intensity ( 580 mW/cm2 ) for direct activation ( Hill and Schaefer , 2009 ) , which is at least 4-fold greater than the extraterrestrial solar constant ( SC: the total solar irradiation density measured by a satellite , 137 mW/cm2 [Gueymard , 2004] ) . The high UV intensity requirement for TRPA1 activation in mammals indicates that electrophilic sensitivity is inadequate for sensitive detection of photochemically-produced free radicals , although radicals are often regarded as inflicting electrophilic oxidative stress . However , Drosophila TRPA1 has been shown to readily respond to UV and H2O2 with the physiological significance and molecular basis of its enhanced sensitivity unknown ( Guntur , 2015 ) . Insects and birds are able to visualize upper-UV wavelengths ( above 320 nm ) via UV-specific rhodopsins ( Salcedo et al . , 2003; Ödeen and Håstad , 2013 ) . Visual detection of UV in this range by insects generally elicits attraction towards the UV source rather than avoidance ( Craig and Bernard , 1990; Washington , 2010 ) . At the same time , lower UV wavelengths , such as UVB ( 280–315 nm ) at natural intensities , have been known to decrease insect phytophagy ( Zavala et al . , 2001; Rousseaux et al . , 1998 ) via a direct effect on the animals that does not involve the visual system ( Mazza et al . , 1999 ) . However , the molecular mechanism of UV-induced feeding deterrence has yet to be unraveled . Here , using feeding assays combined with the Drosophila molecular genetics and electrophysiological analyses in in vivo neurons and heterologous Xenopus oocytes , we show that TRPA1 ( A ) is a nucleophile receptor , and that the ability to detect nucleophilicity enables TRPA1 ( A ) to detect light-evoked free radicals and mediate light-dependent feeding deterrence .
Insect herbivory is often reduced by solar UV radiation ( Mazza et al . , 1999 , 2002; Kuhlmann , 2009 ) , suggesting that UV radiation is responsible for acute control of insect feeding through a light-sensitive molecular mechanism . To examine whether UV radiation deters feeding through a direct impact on insect gustatory systems , we turned to the Drosophila model system . First , we tested if the aversive taste pathway responds to UV illumination using extracellular single sensillum recording , which monitors action potentials from Drosophila labellum taste neurons ( HODGSON et al . , 1955 ) . Aversion to bitter chemicals is in part coded in i-bristles ( Weiss et al . , 2011 ) , which house single bitter-tasting neurons ( Tanimura et al . , 2009 ) . Illumination of 295 nm UV light at an intensity of 5 . 2 mW/cm2 ( ~85% of the total UV intensity on the ground [6 . 1 mW/cm2] ) received by the fly labellum ( Figure 1—figure supplement 1a , b , d ) rapidly elicited firing of single taste neurons in i-a bristles which was sustained after illumination ( Figure 1a , b ) . Bitter-sensing taste cells in i-bristles also act as receptors for tissue-damaging chemicals through expression of the conserved reactive electrophile sensor TRPA1 ( Kang et al . , 2010; Kang et al . , 2012 ) . Because free radicals elicited by UV illumination are often regarded as oxidative electrophiles , we examined the i-bristles of the TrpA1insmutant flies , which lack a functional TrpA1 gene ( Rosenzweig et al . , 2008; Kang et al . , 2010; Kang et al . , 2012 ) . Interestingly , TrpA1ins showed an severely reduced UV response in i-bristles , suggesting the importance of TrpA1 for UV sensing in these sensilla ( Figure 1a , b ) . The cell viability of bristles without UV responses was confirmed with 1 mM berberine ( Figure 1—figure supplement 2 ) , a bitter chemical that selectively excites bitter-sensing neurons in i-a bristle sensilla ( Weiss et al . , 2011 ) . To assess whether the UV-dependent excitation of TrpA1–positive cells pertains to a reduction in the insects’ appetite , a modified capillary feeder ( Café ) assay ( Ja et al . , 2007 ) was used to appraise the effect of UV on feeding . Pairwise Café experiments were conducted in two groups of overnight-starved flies , which were allowed to drink 30 mM sucrose for 2 . 5 min from calibrated glass capillaries inserted into a fly vial with or without 312 nm UVB illumination reaching the inside of the vial at ~1 . 3 mW/cm2 ( Figure 1c , see Materials and methods for details ) , an irradiance less than 25% of the total UV intensity received on the surface of the Earth ( RReDC; Gueymard , 2004 ) . Avoidance indices were calculated to concisely compare the gustatory effect of UV across genotypes ( Figure 1c ) . UV illumination substantially decreased the feeding of wild-type flies ( WT: wcs ) , but did not disrupt the feeding of TrpA1ins mutants ( Figure 1d , e ) , suggesting that the TrpA1- and UV-dependent spikings in aversive taste neurons suppress food ingestion . 10 . 7554/eLife . 18425 . 003Figure 1 . UV-induced taste neuron firing and feeding deterrence require TrpA1 ( A ) but not TrpA1 ( B ) . ( a ) Representative UV responses from i-bristles of the indicated genotypes . wcs: white canton S ( wild type ) . TrpA1ins: TrpA1 knockout flies . Recording taken under 5 . 2 mW/cm2 UV illumination is marked by purple boxes . ( b ) Averaged data from a ( n = 4–5 ) . ( c ) Schematic illustration of modified Café assays used to test UV-induced feeding deterrence . ( d ) Ingestion amount/fly with or without 1 . 3 mW/cm2 312 nm UV illumination in wcs and TrpA1ins . ( e ) Suppression of feeding by UV illumination in the indicated genotypes , presented as an 'avoidance index' ( n = 5–6 ) . ( f ) Reintroduction of transcript variants TrpA1 ( A ) and TrpA1 ( B ) cDNAs differentially restores UV avoidance of TrpA1ins . Letters indicate statistically distinct groups ( p<0 . 05 , n = 8–12 , Tukey’s test ) . ( g ) Isoform-dependent rescue of UV-evoked neuronal responses in TrpA1ins . ( h ) Summary of g ( n = 5–6 ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , Tukey’s , Student’s t- or Mann-Whitney U tests . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 00310 . 7554/eLife . 18425 . 004Figure 1—figure supplement 1 . Setups for electrophysiological recordings of UV-evoked responses in in vivo taste neurons and Xenopus oocytes and estimation of light irradiance at the illuminated tissue . ( a ) Extracellular tip recording configured with the UV-emitting optical fiber cable . ( b ) Magnified image of the inset in ( a ) . ( c ) Two-electrode voltage clamping setup with the UV-emitting optical fiber cable . ( d ) An oval shape of the illumination surface resulting from the 45° angle between the light beam and the surface of the tissue was postulated to simplify estimation of the surface area of taste bristles contacting the light . NA values and inner diameters ( 2c ) of the fiber core were given by the manufacturers ( n = 1 for air in conversion of NA to the angle theta ) . Distance to the surface ( d ) was measured microscopically using a fine ruler . ( e ) Different considerations were used to calculate light intensity for TEVC , as the configuration between the fiber and oocytes is simple and invariant . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 00410 . 7554/eLife . 18425 . 005Figure 1—figure supplement 2 . Cell viability check for non-responders in extracellular recording experiments . The non-responding bristle sensilla in the experiments on UV responsiveness were stimulated by known tastants activating the corresponding cells . For i- and L-bristles , 1 mM berberine and 100 mM sucrose were used , respectively . In particular , non-responder cells expressing TRPA1 channels were tested with electrophilic NMM , which activates TRPA1s . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 00510 . 7554/eLife . 18425 . 006Figure 1—figure supplement 3 . Temperature changes in Café assays that measure UV-dependent feeding avoidance . ( a ) Temperature increases inside of vials with ( n = 21 ) or without ( n = 8 ) active air-cooling . ( b ) UV-induced feeding avoidance was temperature-dependent only for those expressing highly temperature-sensitive TRPA1 ( B ) ( n = 4–9 ) . ***p<0 . 001 , Student’s t-test or Mann-Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 006 UV irradiation imposes toxicity on biological tissues by generating free radicals , which are often regarded as oxidative electrophiles . TRPA1 ( A ) is similar to TRPA1 ( B ) in its responsiveness to reactive electrophiles , although the two isoforms have distinct thermal sensitivities ( Kang et al . , 2012 ) . For this reason , we anticipated that the two TRPA1 isoforms would be similar in UV responsiveness . However , reintroduction of either TrpA1 ( A ) and TrpA1 ( B ) cDNA to Gr66a-Gal4 bitter cells resulted in differential restoration of UV-induced feeding deterrence in TrpA1ins ( Figure 1f ) . Gr66a-Gal4 ( Dunipace et al . , 2001 ) drives expression in the bitter taste neurons of s- and i-bristle sensilla and has been successfully used in RNAi-knockdown and restoration of TrpA1 in feeding behavior experiments ( Kang et al . , 2010 , 2012 ) , indicating that it covers most of the TrpA1-positive taste neurons . In sensillum recordings , TrpA1 ( B ) expression in TrpA1ins i-bristle bitter neurons produced very few UV-evoked spikes , while TrpA1 ( A ) expression generated robust action potentials in response to UV illumination ( Figure 1g , h and Figure 1—figure supplement 2 ) . This implies that the feeding avoidance observed with TrpA1 ( B ) expression in Figure 1f may be due to TRPA1 ( B ) -related temperature sensitivity but not UV sensitivity . Indeed , the 2 . 5-minute-long UV illumination used in the feeding assays raised the temperature in the vials by 1 . 6 ± 0 . 07°C and 3 . 25 ± 0 . 13°C with and without active cooling , respectively , at ambient temperatures between 22 . 5 and 23°C ( Figure 1—figure supplement 3a ) . The greater extent of temperature increase in the vials without air-cooling resulted in higher avoidance in animals expressing TrpA1 ( B ) but not for TrpA1 ( A ) -expressing or WT animals ( Figure 1—figure supplement 3b ) . These data suggest that the mild UV-dependent feeding deterrence induced by TrpA1 ( B ) in TrpA1insanimals resulted from the high thermosensitivity of TRPA1 ( B ) , which sensed a temperature difference of 1 . 6°C . The contrasting UV sensitivity of the two TRPA1 isoforms may be related to the intracellular environment of bitter neurons , which contain only TRPA1 ( A ) ( Kang et al . , 2012 ) , rather than to their functional divergence . To test this possibility , the isoforms were ectopically expressed in the Gr5a-Gal4 ( Marella et al . , 2006 ) sweet-tasting neurons . Consistent with the previously observed lack of responsiveness to TRPA1 agonists ( Kang et al . , 2012 ) , the L-bristles of control animals failed to respond to 295 nm UV light , with the exception of a few mechanosensory responses occasionally caused by bristle deflection upon contact with the electrolyte ( Figure 2a , b , and Figure 1—figure supplement 2 ) . Similar to the results from Gr66a-Gal4 cells , Gr5a neurons expressing TrpA1 ( A ) but not those expressing TrpA1 ( B ) showed robust UV-induced firing ( Figure 2a , b ) , although , unlike bitter-sensing neurons , UV-evoked firing in Gr5a neurons was attenuated soon after the removal of illumination ( Figures 1g and 2a ) . Furthermore , Gr5a-Gal4 rescue with TrpA1 ( A ) and TrpA1 ( B ) enabled TrpA1ins flies to extend their probosces in response to UV and infrared ( IR ) light ( Figure 2c ) , demonstrating that expression of the two isoforms transforms sweet-tasting neurons into UV and IR receptors , respectively . These results suggest that TRPA1 ( A ) is capable of responding to UV light without the co-expression of other signaling factors such as GPCRs , as the intrinsic molecular thermosensor TRPA1 ( B ) independently reacts to IR , which warms TRPA1 ( B ) -expressing cells ( Kang et al . , 2012 ) . 10 . 7554/eLife . 18425 . 007Figure 2 . TRPA1 ( A ) expression is sufficient for UV responsiveness . ( a ) UV- and isoform-dependent action potentials in sweet-sensing neurons exogenously expressing TRPA1 isoforms . ( b ) Summary of a ( n = 4–5 ) . ( c ) Proboscis extension reflex ( PER ) to UV ( n = 24–25 ) and IR ( n = 22–24 ) in TrpA1ins flies ectopically rescued in sweet taste neurons . ( d-f ) Typical UV-evoked currents in Xenopus oocytes expressing the indicated isoforms . RR: 0 . 2 mM ruthenium red . NMM: 0 . 1 mM . Right , Current-voltage ( IV ) relationships at the indicated points in the Left panels . ( g ) Summary of d–f . UV responses normalized to NMM currents at +60 and −60 mV , respectively ( n = 4–5 ) . #: p<0 . 05 , ###: p<0 . 001 , ANOVA Repeated Measures test compared to the first response ( n ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , Tukey’s , Student’s t- or Mann-Whitney U tests . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 00710 . 7554/eLife . 18425 . 008Figure 2—figure supplement 1 . Human TRPA1 ( humTRPA1 ) is not activated by the same UV intensity as Drosophila TRPA1 ( A ) . Three serial 295 nm UV pulses at 5 . 2 mW/cm2 were administered to oocytes expressing the indicated TRPA1s . The UV-elicited current amplitudes at +60 or −60 mV were normalized to the 0 . 1 mM NMM-evoked current and averaged ( n = 4–8 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 00810 . 7554/eLife . 18425 . 009Figure 2—figure supplement 2 . TRPA1 ( A ) s from flies and mosquitoes do not need the cytosol of Xenopus oocytes for UV responsiveness . ( a–d ) Inside-out macropatches excised from Xenopus oocytes were subjected to electrophysiological recording under 295 nm UV illumination at the indicated intensity . Left: currents at -100 and +100 mV . Right: IV curves at the marked recorded points . ( e ) Current amplitudes were averaged and presented as bar graphs ( mean and SEM , n = 3–4 ) . **p<0 . 01 , ***p<0 . 001 , ANOVA Tukey’s test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 009 To further confirm that TRPA1 ( A ) serves as a molecular sensor of UV , we turned to Xenopus oocytes as a heterologous expression system . At an unnaturally high intensity of 350 nm UVA illumination , such as 580 mW/cm2 ( ~420% of SC and >9 , 000% of total UV intensity on the ground ) , mammalian TRPA1 was directly stimulated by UV illumination when heterologously expressed ( Hill and Schaefer , 2009 ) . In line with the observation that Drosophila TRPA1 ( A ) confers a low threshold of UV responsiveness in nonnative cells , heterologous Xenopus oocytes expressing TRPA1 ( A ) but not those expressing TRPA1 ( B ) showed TRPA1-dependent current increases in response to 295 nm UV light at ~62% of the total ground UV intensity , 3 . 8 mW/cm2 ( Figure 2d–f; for an estimation of UV irradiance received by oocytes see Figure 1—figure supplement 1e ) . The UV-induced current exhibited the reversal potential and outward rectification previously associated with currents recorded from fly TRPA1 ( Kang et al . , 2010 , 2012 ) . TRPA1 unselectively conducts cations , with reversal potentials close to 0 mV . The UV and NMM responses serially recorded from each cell showed similar reversal potentials of −5 . 6 ± 1 . 1 and −5 . 4 ± 1 . 0 mV , respectively ( not significantly different , paired t-test , n = 7 ) . Rectification was quantitated by calculating the ratio between the conductances at +60 to that at −60 mV . Moderate outward rectification was indicated by the ratios of the net UV and NMM responses ( Figure 2d , g ) , which were 1 . 8 ± 0 . 2 and 1 . 7 ± 0 . 3 ( not significantly different , paired t-test , n = 4 ) , respectively , when TRPA1 ( A ) showed a similar degree of activation in response to UV and NMM . Such UV-responsive ability of TRPA1 ( A ) in frog cells indicates the sufficiency of TRPA1 ( A ) as an autonomous UV receptor expressed in fly bitter-sensing neurons . In contrast , human TRPA1 ( humTRPA1 ) expressed in oocytes failed to yield current responses to the UV intensity of 3 . 8 mW/cm2 ( Figure 2—figure supplement 1 ) as expected from the excessive intensity required previously ( Hill and Schaefer , 2009 ) . Furthermore , inside-out macropatches from TRPA1-expressing oocytes also responded to UV light in an isoform-dependent manner ( Figure 2—figure supplement 2a , b , e ) . To exclude the possibility of leak current induced by UV illumination , we recorded from TRPA1 ( B ) -containing membranes over extended periods of time ( up to 350 s ) and did not observe a significant increase in current . Activation of TRPA1 ( A ) often showed a delayed onset before UV-evoked current responses , unlike TRPA1 ( A ) in the whole-cell configuration , suggesting that cytosolic reducing power aids in UV-dependent TRPA1 ( A ) activation . The ability to confer UV responsiveness to ectopic fly neurons and Xenopus oocytes strongly argues that TRPA1 ( A ) serves as the molecular UV receptor without other upstream signaling molecules or coreceptors . Next , we asked why TRPA1 ( A ) , but not TRPA1 ( B ) , can respond to UV light . The two isoforms differ in their N-termini which comprises less than 10% of the primary protein structure , but their reactive electrophile sensitivity is comparable ( Kang et al . , 2012 ) . We conducted conventional Café assays to confirm the similarity of sensitivity of the isoforms to the electrophile N-methyl maleimide ( NMM ) ( Figure 3a ) . WT but not TrpA1ins animals showed NMM-dependent feeding avoidance as previously reported ( Kang et al . , 2012 , 2010 ) . The reintroduction of either TrpA1 ( A ) or TrpA1 ( B ) cDNA similarly restored NMM-dependent feeding avoidance in TrpA1ins , demonstrating that the isoforms are similar in their ability to confer electrophile responsiveness in vivo . This raises the possibility that TRPA1 ( A ) detects a property of UV-generated free radicals other than oxidizing electrophilicity . Unpaired electrons in free radicals serve as both electrophiles and nucleophiles ( Domingo and Pérez , 2013 ) , as the lone electrons favor pairing by either accepting ( electrophilic ) or donating ( nucleophilic ) an electron . The primary oxyradical superoxide ( O2·- ) ( molecular oxygen that gained an electron ) , arising from UV illumination , is a well-known nucleophilic reductant ( Danen and Warner , 1977 ) . Also , hydrogen peroxide ( H2O2 ) , which can be derived from O2·- , is not only an oxidizing electrophile but also a reducing nucleophile owing to its two key chemical properties . First , when nucleophilic atoms , such as sulfur , nitrogen and oxygen , are adjacent to each other , the nucleophilicity of the compounds is dramatically increased ( the alpha effect [Edwards and Pearson , 1962] ) . H2O2 ( H-O-O-H ) contains two consecutive oxygen atoms , which supposedly renders it nucleophilic . Second , H2O2 , a weak acid , yields the hydroperoxide anion ( HOO- ) , a strong nucleophile ( Pearson and Edgington , 1962 ) . To examine if TRPA1 isoforms differentially respond to H2O2 , H2O2-dependent feeding avoidance was tested with Café assays . WT flies increasingly avoided ingestion of H2O2-containing food as the dose of H2O2 was increased from 10 to 100 mM , while TrpA1ins did not ( Figure 3b ) . The robust spiking response of bitter-sensing neurons in i-bristles to 100 mM H2O2required the TrpA1 gene ( Figure 3c , d , and Figure 3—figure supplement 1 ) . Like UV responses , feeding avoidance ( Figure 3e ) and neuronal responses ( Figure 3f , g and Figure 3—figure supplement 1 ) to H2O2 were preferentially rescued by TrpA1 ( A ) rather than TrpA1 ( B ) . Ectopic expression in Gr5a-Gal4 neurons recapitulated the isoform dependence observed in bitter-sensing cells ( Figure 3h , i and Figure 3—figure supplement 1 ) , indicating that the differential outcomes from expression of TrpA1 transcript variants are unrelated to cellular context . 10 . 7554/eLife . 18425 . 010Figure 3 . Responses of Drosophila TRPA1 to nucleophilicity-bearing H2O2are isoform-dependent . ( a ) Electrophilic NMM-induced feeding avoidance is not isoform-dependent ( n = 3–4 ) . Letters indicate statistically distinct groups ( p<0 . 01 , Tukey's test ) . ( b–d ) Feeding deterrence ( b ) and neuronal responses ( c and d ) to H2O2 are abolished in TrpA1ins ( b: n = 3–11 , d: n = 8–14 ) . ( e–i ) Isoform-dependent rescue of H2O2 feeding avoidance ( e , n = 4 ) and neuronal activation ( f–i , n = 6–29 ) in TrpA1ins . j and k , Typical H2O2 current recordings normalized to the maximum H2O2 response ( j ) and H2O2 dose-dependence ( k , n = 4–11 ) of TRPA1 isoforms in oocytes . Alternating colors represent increasing concentrations of H2O2 as indicated . **p<0 . 01 , ***p<0 . 001 , Tukey’s or Student’s t-tests . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01010 . 7554/eLife . 18425 . 011Figure 3—figure supplement 1 . Cell viability check for non-responders in extracellular recording experiments . The non-responding bristle sensilla in the experiments examining H2O2 responsiveness were stimulated by other tastants activating the corresponding cells . For i- and L-bristles , 1 mM berberine , and 100 mM sucrose were used , respectively . Non-responding cells expressing TRPA1 channels were tested with electrophilic NMM , which activates all TRPA1 channels . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 011 To date , H2O2-responding TRPs have been characterized as being indirectly stimulated and/or requiring high doses ( >1 mM ) of H2O2 to generate current under physiological conditions ( Yoshida et al . , 2006; Fonfria et al . , 2004 ) . In particular , extracellular Ca2+ is a requisite for the moderate H2O2 sensitivity ( EC50 =~ 230 μM ) of Ca2+-conducting mouse TRPA1 ( Andersson et al . , 2008 ) , which is activated directly by an elevation in intracellular [Ca2+] ( Wang et al . , 2008; Zurborg et al . , 2007 ) , providing evidence that H2O2 is a weak electrophilic oxidant compared to other electrophilic TRPA1 agonists . Interestingly , Drosophila TRPA1 ( A ) heterologously expressed in Xenopus oocytes was readily activated by H2O2 at concentrations as low as 100 nM ( Figure 3j , k , EC50 = 5 . 0±0 . 8 μM , and Supplementary file 1 ) . In contrast , the response of TRPA1 ( B ) was slow and required high H2O2 concentrations ( Figure 3j , k , EC50 = 0 . 9±0 . 2 mM ) , possibly because the response of TRPA1 ( B ) depends solely on the electrophilicity of H2O2 , similar to mammalian TRPA1s . The ~450-fold higher sensitivity of TRPA1 ( A ) than TRPA1 ( B ) in oocytes may account for the differential behavioral and neuronal H2O2 responses of the TRPA1 isoforms . Thus , H2O2 mimics UV in that feeding inhibitions by H2O2 and UV rely on TrpA1 ( A ) , suggesting that the nucleophilicity of H2O2 and UV-generated radicals is critical for activation of TRPA1 ( A ) . The high H2O2 concentration required for neuronal and behavioral responses compared to that required to evoke heterologous responses in oocytes may be due to enzymatic catalysis of H2O2 in Drosophila taste bristles , the activity of which may not be robust in Xenopus oocytes . In addition , chemical doses necessary for in vivo TRPA1 activation are usually much higher than those required for activation of TRPA1 that is heterologously expressed in oocytes according to previous studies ( Kang et al . , 2010 , 2012 ) . Yet , TRPA1 isoform dependence is consistent in in vivo and in vitro studies , which demonstrates that TRPA1 ( A ) is superior to TRPA1 ( B ) in sensing nucleophilicity-accompanying H2O2 in various contexts . A peculiar property of TRPA1 ( A ) is that its expression in oocytes effects small standing current at rest . This basal activity is little observed in cells expressing TRPA1 ( B ) or the mutants TRPA1 ( A ) C105A and TRPA1 ( A ) R113A/R116A ( Kang et al . , 2012 ) ( Figure 4b , c ) , in which the conserved Cys105 and Arg113/116 residues in the cytosolic N-terminus of TRPA1 ( A ) were replaced with Ala ( Figure 4a ) . This observation led to the hypothesis that the intracellular reducing/nucleophilic power for redox homeostasis partially opens TRPA1 ( A ) . To examine the idea , TRPA1 isoforms expressed in frog oocytes were subjected to perfusion buffer containing the well-known nucleophilic reductant dithiothreitol ( DTT ) . DTT contains two nucleophilic thiols and is a popular reductant used in the studies of protein biochemistry . Indeed , only the TRPA1 channel that produced the standing current showed dose-dependent responses to DTT in oocytes ( Figure 4d , EC50 =~92 . 8 μM and Figure 4—figure supplement 1 ) . The DTT response of TRPA1 ( B ) was little compared to that of TRPA1 ( A ) , revealing that detection of nucleophilic DTT by TRPA1 is also isoform-dependent . The current amplitude of TRPA1 ( A ) evoked by H2O2 is intermediate between those induced by DTT and NMM; the average maximal amplitudes of DTT- and H2O2-evoked currents were ~10% and ~30% of NMM responses , respectively ( Figure 4—figure supplement 2 ) , implying that H2O2 synergistically stimulates TRPA1 ( A ) through two distinct pathways . 10 . 7554/eLife . 18425 . 012Figure 4 . The sensitivity to nucleophilic dithiothreitol correlates with the UV and H2O2responses of TRPA1 ( A ) . ( a ) Sequence alignment of TRPA1 ( A ) N-terminal regions from indicated insects . Conserved residues that were substituted with alanine are indicated by arrows . ( b ) Representative current-voltage relationships in oocytes expressing TRPA1s . ( c ) Averaged basal activity of the indicated TRPA1 isoforms , normalized to NMM current ( n = 7–37 ) . ( d ) DTT dose dependence of TRPA1s , normalized to NMM current ( n = 4–10 ) . ( e ) Typical H2O2 responses of mutant TRPA1 ( A ) s and human TRPA1 ( humTRPA1 ) in comparison with WT TRPA1 ( A ) and TRPA1 ( B ) . ( f ) H2O2 dose dependence of TRPA1s in e ( n = 4–6 ) . ( g ) UV-evoked currents at +60 and −60 mV from WT and mutant TRPA1 ( A ) s normalized to NMM current ( n = 4 ) . *** or ###: comparison among first or second responses , respectively . ( h ) Typical extracellular recordings for UV or H2O2-induced action potentials from i-bristles expressing TrpA1 ( A ) R113A/R116A through Gr66a-Gal4 . Inset: NMM response from the UV-non-responder presented underneath . ( i ) Summary of h ( n = 6–12 ) . Response to the electrophile NMM was unimpaired despite severe attenuation of UV and H2O2 responses upon expression of TrpA1 ( A ) R113A/R116A . ( j ) Similarly to neuronal responses , feeding deterrence to UV and H2O2 was repressed by expression of TrpA1 ( A ) R113A/R116A ( n = 4–8 ) . **p<0 . 01 , *** or ###p<0 . 001 , Tukey’s test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01210 . 7554/eLife . 18425 . 013Figure 4—figure supplement 1 . Representative TRPA1 currents evoked by DTT in Xenopus oocytes . DTT-induced currents were compared with responses to 0 . 1 mM NMM for TRPA1s as indicated . ( a–e ) , Voltage-clamped at −60 mV . ( f ) A 300-msec voltage ramp between -60 and +60 mV was applied every second . Currents at either -60 or +60 mV were plotted as a function of time . RR: ruthenium red at 50 μM . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01310 . 7554/eLife . 18425 . 014Figure 4—figure supplement 2 . Maximal current amplitudes of TRPA1 ( A ) by H2O2 were intermediate between those of DTT and NMM . TRPA1 ( A ) currents in Xenopus oocytes that were evoked by either 10 mM DTT ( n = 5 ) or 0 . 1 mM H2O2 ( n = 10 ) were normalized by currents induced following 0 . 1 mM NMM application . The indicated concentrations for the chemicals yielded maximal current responses at −60 mV , which were used for estimation of the maximum current amplitudes . ***p<0 . 001 . Mann-Whitney U test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01410 . 7554/eLife . 18425 . 015Figure 4—figure supplement 3 . Expression of TrpA1 ( A ) C105A in bitter-tasting neurons failed to exhibit robust NMM-induced action potentials . Spiking frequencies of bitter neurons expressing either TrpA1 ( A ) C105A ( n = 6 ) or TrpA1 ( A ) R113A/R116A ( n = 7 ) in TrpA1-deficient mutant animals were plotted . ***p<0 . 001 , Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 015 As mentioned above , heterologously expressed TRPA1 ( A ) C105A and TRPA1 ( A ) R113A/R116A in oocytes appeared to lack the constitutive activity observed with TRPA1 ( A ) WT , suggesting that the mutants may be unable to respond to nucleophiles . Indeed , C105A and R113A/R116A substitutions compromised the DTT responsiveness of TRPA1 ( A ) such that it was indistinguishable from that of TRPA1 ( B ) . The NMM sensitivity of these mutants was previously shown to be very similar to that of TRPA1 ( A ) WT ( Kang et al . , 2012 ) , indicating that the mutations specifically impaired DTT-dependent activation . Consistent with a previous study in which high concentrations of DTT completely reversed the mammalian TRPA1 current provoked by reversible electrophilic agonists ( Macpherson et al . , 2007 ) , we found that cells expressing humTRPA1 seldom showed electrophysiological responses to DTT ( Figure 4d and Figure 4—figure supplement 1f , and Supplement file 1 ) . Notably , these DTT-insensitive mutants and humTRPA1 showed remarkably reduced responses to H2O2 ( Figure 4e , f ) ; the mutants and humTRPA1 were similar to TRPA1 ( B ) in H2O2 sensitivity ( Supplement file 1 ) and activation kinetics . These results indicate a strong structure-function association between the ability of TRPA1 ( A ) to respond to DTT and H2O2 ( Figure 4e , f ) . Furthermore , oocytes expressing either mutant failed to respond to 3 . 8 mW/cm2 295 nm UV irradiation ( Figure 4g ) , revealing the concomitant requirement of the conserved residues for DTT , H2O2 and UV responses . To demonstrate the in vivo implications of TRPA1 ( A ) nucleophile sensitivity in H2O2 and UV responsiveness , cDNAs encoding TrpA1 ( A ) C105A and TrpA1 ( A ) R113A/R116A were expressed in WT Gr66a-Gal4 neurons ( Figure 4h–j ) . While the TrpA1 ( A ) C105A transgene was not functionally expressed in Gr66a-Gal4 cells ( Figure 4—figure supplement 3 ) , expression of TrpA1 ( A ) R113A/R116A in Gr66a-Gal4 cells dramatically decreased both UV- and H2O2-dependent neuronal spiking responses and feeding avoidance but did not impair NMM responsiveness ( Figure 4h–j ) . These data demonstrate a specific dominant negative effect of the TrpA1 ( A ) R113A/R116A mutant on TRPA1 ( A ) -mediated H2O2 and UV detection in vivo , and strongly support the notion that the TrpA1-positive neurons are necessary for UV-dependent feeding avoidance . Taken together , the nucleophile-detecting ability , which is reliant on the isoform-specific N-terminus , allows gustatory TRPA1 ( A ) to sensitively respond to H2O2 and UV light . Thus , the TRPA1 ( A ) N-terminus in the cytosol offers a unique activation modality that is independent of the electrophile-sensing pathway involving cysteines in the ankyrin repeat domain , which are shared between isoforms . Next , we examined if nucleophile sensing of TRPA1 ( A ) can be a general underlying mechanism for the detection of UV illumination in insects . TRPA1 isoforms from malaria-transmitting mosquitoes , Anopheles gambiae ( agTRPA1 ) , were previously reported to be similar to their Drosophila counterparts in thermosensitivity ( Kang et al . , 2012 ) . The agTRPA1 isoforms were heterologously expressed in frog oocytes to investigate whether the reactions to nucleophiles , H2O2 and UV are shared characteristics of insect TRPA1 ( A ) s . Compared to Drosophila TRPA1 ( A ) , cells expressing agTRPA1 ( A ) exhibited considerably enhanced responsiveness to all three stimuli , while isoform dependencies existed , similar to Drosophila TRPA1 ( Figure 5 and Supplement file 1 ) . DTT more robustly activated agTRPA1 ( A ) with an order-lower EC50 of ~3 . 8 μM and ~10 times higher peak current amplitudes than Drosophila TRPA1 ( A ) when normalized to NMM responses ( Figure 5a , b and Figure 5—figure supplement 1a ) . Other structurally distinct nucleophiles , such as imidazole and benzyl thiocyanate ( BTC ) , also preferentially activated agTRPA1 ( A ) over agTRPA1 ( B ) ( Figure 5—figure supplement 2 ) , demonstrating that TRPA1 ( A ) responds to nucleophilicity and not the structures of the compounds . Oocytes microinjected with cRNA of agTrpA1 ( A ) but not of agTrpA1 ( B ) were often found to be of poor quality for electrophysiological characterization , presumably due to the large conductance resulting from agTRPA1 ( A ) activation by the nucleophilic reducing power in the cytosol . To avoid this problem , 3–4 hr after cRNA microinjection , the TRPA1 antagonist ruthenium red was added to the oocyte media at a concentration of 3 μM to block the agTRPA1 ( A ) activity induced by cytosolic reducing power and to yield cells appropriate for subsequent experiments . H2O2 elicited ~5 times larger NMM-normalized agTRPA1 ( A ) currents with a ~3 times lower EC50 of 1 . 7 ± 0 . 3 μM than did Drosophila TRPA1 ( A ) ( Figure 5c , d and Figure 5—figure supplement 1b , and Supplement file 1 ) . Furthermore , UV responses from agTRPA1 ( A ) were ~5 times higher than those from Drosophila TRPA1 ( A ) tested at the same settings ( Figure 5e , f ) . Inside-out macropatches from oocytes showed robust UV responses when expressing agTRPA1 ( A ) but not agTRPA1 ( B ) ( Figure 2—figure supplement 2c–e ) , which is indicative of direct UV detection . Consistent with its higher nucleophilic sensitivity and responsiveness than fly TRPA1 ( A ) , agTRPA1 ( A ) in excised membranes exhibited little latency and larger current amplitudes than did the fly ortholog . Thus , the functional study performed using agTRPA1 ( A ) conveys two important messages . First , detection of nucleophiles , H2O2 and UV is likely a common function of TRPA1 ( A ) in insects . Second , nucleophile sensitivity of insect TRPA1 ( A ) is tightly associated with the ability to rapidly detect H2O2 and UV illumination , as the sensitivities to the three stimuli are very well correlated with one another in experiments with agTRPA1 ( A ) as well as Drosophila TRPA1 ( A ) s . 10 . 7554/eLife . 18425 . 016Figure 5 . Concomitant natural variations in DTT , UV and H2O2 responsiveness between Anopheles gambiae and Drosophila melanogaster TRPA1 ( A ) s . ( a and b ) In Xenopus oocytes , agTRPA1s show isoform dependence to DTT as do their Drosophila counterparts , with larger response amplitudes ( n = 5–6 ) . Dose-dependency to DTT ( a ) and averaged peak current amplitudes evoked by DTT and NMM ( b ) are presented for the channels , as indicated . ( c and d ) The robust DTT receptor , agTRPA1 ( A ) , exhibits enhanced H2O2 responses compared to Drosophila TRPA1 ( A ) ( n = 4–5 ) . Dose-dependency to H2O2 ( c ) and averaged peak current amplitude ( d ) are compared between mosquito and fly TRPA1 isoforms . ( e and f ) agTRPA1 ( A ) responds more robustly to UV light than Drosophila TRPA1 ( A ) , while agTRPA1 ( B ) does not . A typical UV-evoked current response of agTRPA1 ( A ) is superimposed on the responses of agTRPA1 ( B ) and Drosophila TRPA1 ( A ) following normalization to the NMM response ( e ) . Normalized UV-elicited current amplitudes averaged for the indicated channels ( f , n = 4–12 ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , Tukey’s and Mann-Whitney U or Student’s t-tests . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01610 . 7554/eLife . 18425 . 017Figure 5—figure supplement 1 . Typical DTT ( a ) and H2O2 ( b ) responses of agTRPA1 ( A ) and agTRPA1 ( B ) heterologously expressed in Xenopus oocytes . RR: 50 μM ruthenium red . NMM: 0 . 1 mM N-methyl maleimide . DTT and H2O2 concentrations are indicated in the figure . Voltage clamped at −60 mV . The dashed line represents the baseline . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01710 . 7554/eLife . 18425 . 018Figure 5—figure supplement 2 . Nucleophiles other than DTT preferentially activate TRPA1 ( A ) over TRPA1 ( B ) . ( a–d ) Representative current traces from Xenopus oocytes expressing the indicated TRPA1 isoforms upon application of increasing imidazole concentrations ( voltage-clamped at −60 mV ) . Imidazole is the nucleophilic side chain of histidine , four of which form coordinate bonds with oxygen-binding Fe2+ in the heme . ( e ) Imidazole dose dependences of the indicated TRPA1 channels ( n = 3–5 ) . ( f–g ) Typical current recordings from cells expressing agTRPA1 isoforms in response to benzyl thiocyanate ( BTC ) . BTC is a naturally occurring plant nucleophile , while its isosteric compound benzyl isothiocyanate ( BITC ) activates TRPA1s because of its electrophilicity ( Hinman et al . , 2006 ) . ( h ) BTC dose dependence of agTRPA1s ( n = 5–12 ) . The agTRPA1 ( A ) isoform is much more sensitive to the nucleophile BTC than is agTRPA1 ( B ) . Letters indicate statistically distinct groups . ANOVA Dunn’s test for comparison between three or more groups . **p<0 . 01 , ***p<0 . 001 . Student’s t- or Mann-Whitney U tests between two groups . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 018 To evaluate the spectrum dependence of TrpA1-dependent feeding deterrence in fruit flies , monochromatic UVA light at a wavelength of 365 nm was used in the neuronal , behavioral and heterologous experiments , and the results from Xenopus oocytes were compared with those obtained using monochromatic UVB radiation ( Figure 6a , c , e ) . WT animals showed cellular and behavioral responses to UVA which relied on TrpA1 ( Figure 6a , c ) . For robust TrpA1-dependent gustatory neuronal spiking , UVA at 365 nm required a much greater intensity and a longer duration of irradiation , 42 . 1 mW/cm2 and ~1 min in total , respectively ( Figure 6a and Figure 6—figure supplement 1a ) . TrpA1insanimals were more appetitive under UVA , and consumed more sucrose than did controls , resulting in a negative avoidance index ( Figure 6c ) . The behavioral deficit of TrpA1ins was rescued by gustatory-specific Gr66a-Gal4 as well as the genomic rescue transgene ( Hamada et al . , 2008; Du et al . , 2016 ) . Note that wcs show a higher avoidance than do w+rescue flies . This is probably because the lack of eye pigments in wcs impairs the visual system , which is necessary for UVA attraction ( Figure 6—figure supplement 2c; wcs indicated by grey boxes ) . The attractive nature of UVA can also be observed in the feeding deterrence assay with visually intact mini-white-positive TrpA1ins ( Figure 6c ) , as the mutants show increased ingestion upon UVA illumination . To probe the possible role of photoreceptors in feeding deterrence , the chemical synaptic transmission of photoreceptors was inhibited by the tetanus toxin light chain ( TNT ) expressed under the control of GMR-Gal4 . This genetic perturbation insignificantly impaired UV-induced feeding deterrence ( Figure 6—figure supplement 2a ) , while the flies failed to show typical attraction responses to UVA at 365 nm ( Figure 6—figure supplement 2b , c ) . This result indicates that TrpA1-positive taste neurons are instrumental in avoidance , which is consistent with the suppression of feeding inhibition observed with gustatory expression of the dominant negative TrpA1 ( A ) transgene ( Figure 4j ) . To compare the innate sensitivity of TRPA1 isoforms to UVA and UVB light , isoforms heterologously expressed in oocytes were subjected to determination of dose dependence in response to changing light intensities ( Figure 6e , and Figure 6—figure supplement 1b ) . Consistent with the isoform dependence of nucleophile-associated stimuli , responses to UVA were observed when TRPA1 ( A ) but not with TRPA1 ( B ) was expressed . The half-maximal efficacy light irradiances ( EI50s ) of fly TRPA1 ( A ) to UVA and UVB were similar to each other ( 3 . 8 ± 2 . 2 and 2 . 7 ± 0 . 5 mW/cm2 at −60 mV , respectively ) , although the maximal response amplitudes elicited by UVA light were relatively lower than those elicited by UVB light . UV responses of agTRPA1 ( A ) were more robust in terms of the normalized maximal amplitude , but the EI50s ( 4 . 7 ± 2 . 7 and 3 . 0 ± 0 . 5 mW/cm2 at −60 mV for UVA and UVB , respectively ) were similar to those of fly TRPA1 ( A ) . 10 . 7554/eLife . 18425 . 019Figure 6 . White light activates TRPA1 ( A ) and deters feeding at natural intensities . Monochromatic UVA at 365 nm ( a , c and e ) and polychromatic white light ( b , d and f ) suppress feeding through TRPA1 ( A ) . ( a and b ) Illumination with 365 nm UVA light excites bitter-tasting neurons of i-a bristles at 42 . 1 mW/cm2 , and the response is dependent on TrpA1 ( n = 5–8 ) ( a ) . Polychromatic white light from a Xenon arc lamp stimulates TrpA1-dependent bitter-sensing neurons at 93 . 4 mW/cm2 , which is similar to natural solar intensity ( n = 5–9 ) ( b ) . Neuronal activation by white light requires UV , as it was abolished upon filtering out UV with the thin titanium dioxide-coated glass . * and #p<0 . 05 , *** and ###p<0 . 001 , Student’s t- or Tukey’s test for the first and second illuminations , respectively . ( c and d ) UVA ( c , n = 4–7 ) or white light illumination ( d , n = 4–7 ) hinders feeding depending on TrpA1 ( A ) . UV blocking from white light significantly reduces feeding avoidance ( orange bars ) . * and #p<0 . 05 , ** and ##p<0 . 01 , *** and ###p<0 . 001 , Tukey test for 2 . 5 and 5 min , respectively . § , §§ and §§§: p<0 . 05 , 0 . 01 and 0 . 001 , respectively , Student’s t-tests between illuminations with and without the UV filter . ( e and f ) UV ( e , n = 4–8 ) and white light ( f , n = 4–8 ) intensity dependences of fly ( upper ) and mosquito ( lower ) TRPA1 isoforms heterologously expressed in oocytes . UVB at 295 nm ( pale purple ) produced higher responses than UVA at 365 nm ( dark purple ) ( e ) . Blocking the UV component of white light significantly reduces the current elicited by white light illumination ( f ) . The half maximal efficacy intensities of UV and white light are given in the text and Supplement file 1 . Student’s t-test between illuminations with and without UV filter . * and # , ** and ## , and *** and ###p<0 . 05 , 0 . 01 , and 0 . 001 , respectively ( *: +60 and #: −60 mV ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 01910 . 7554/eLife . 18425 . 020Figure 6—figure supplement 1 . TRPA1 ( A ) -dependent neuronal and heterologous responses to UVA and white light . ( a ) Representative sensillum recording results following application of 365 nm UVA light . The right panels confirm the viability of non-responders . ( b ) Typical TEVC recording results with the indicated TRPA1 isoforms from flies ( upper ) and mosquitoes ( lower ) . Insets marked with ‘5x’ show current traces magnified five times for clear presentation , while scale bars are drawn for 1x traces . ( c ) Left , relative spectral intensities of the Xenon arc lamp ( black , DG4 , Sutter instruments , NJ , USA ) or sunlight on the ground ( grey ) . Right , light transmission through the UV filter used in Figure 6 . ~90% of light within the visible spectrum can pass through the filter . ( d ) Representative sensillum recordings with white light illumination from the indicated genotypes . The right panels confirm the viability of non-responders . ( e ) Typical TEVC recordings with the indicated TRPA1 isoforms from flies ( upper and middle ) and mosquitoes ( lower ) . TRPA1 ( A ) mutants that lack nucleophile sensitivity did not respond to white light ( middle ) . Colored boxes indicate the time windows of light illumination . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 02010 . 7554/eLife . 18425 . 021Figure 6—figure supplement 2 . Photoreceptors are important for UVA attraction but not for UVB-dependent feeding suppression . ( a ) UVB-illuminated Café assay results with animals in which chemical synaptic transmission of photoreceptors is silenced by the tetanus toxin light chain . NS: not significant . ( b ) Typical results showing UVA attraction of the indicated genotypes . The purple ramp illustrates the putative UVA gradient between the measured intensities at two internal extremes of the vial . Twenty seconds after illumination control flies mostly resided at the side of the UVA source , while flies with silenced neurons did not . ( c ) Summary of experiments in ( b ) and experiments with wcs ( grey ) . ***p<0 . 001 , Tukey’s test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 02110 . 7554/eLife . 18425 . 022Figure 6—figure supplement 3 . Blue but not green light is capable of activating taste neurons , which depends on TrpA1 . ( a ) Increasing intensities of blue light at 470 nm were administered as indicated to i-a bristles of WT and TrpA1ins . Green light did not elicit action potentials at a much higher intensity ( bottom ) . ( b ) Averaged data ( n = 5–8 ) . **p<0 . 01 , ***p<0 . 001 , Mann-Whitney test between WT and TrpA1ins . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 022 The total solar UV ( <400 nm ) intensity is ~6 . 1 mW/cm2 ( ~6 . 8% of total solar irradiance ) on the ground , and only ~0 . 08 mW/cm2 ( ~1 . 3% of total UV irradiance ) of UVB ( <315 nm ) reaches the ground ( RReDC ) . Accordingly , the requirement of UV irradiances for the TRPA1 ( A ) -dependent responses described above is much higher than the natural intensities of UVA or UVB light that insects receive . On the basis of this observation , it is conceivable that the TrpA1-dependent feeding deterrence is unlikely to occur in natural settings , although TRPA1 ( A ) is more sensitive by far than is humTRPA1 , which requires UVA intensities of ~580 mW/cm2 . Provided that the ability of nucleophile-detecting TRPA1 ( A ) s to sense free radicals is the mechanistic basis of the UV responsiveness of TRPA1 ( A ) s , we postulated that TRPA1 ( A ) might be capable of responding to polychromatic natural sunlight , as visible light with relatively short wavelengths such as violet and blue rays is also known to generate free radicals via photochemical reactions with essential organic compounds such as flavins ( Eichler et al . , 2005; Godley et al . , 2005 ) . To test this possibility , TrpA1 ( A ) -dependent responses were examined with white light from a Xenon arc lamp which produces a sunlight-simulating spectral output of the wavelengths higher than ~330 nm ( Figure 6—figure supplement 1c ) . Less than 2% of the total spectral intensity derived from a Xenon arc lamp is UV light from 330 to 400 nm . Indeed , an intensity of 93 . 4 mW/cm2 , which is comparable to natural sunlight irradiance on the ground , substantially increased action potentials in TrpA1-positive taste neurons ( Figure 6b , and Figure 6—figure supplement 1d ) . The increase in spiking was more apparent during the second 30 s illumination , while both the first and second 30 s responses to illumination required TrpA1 . In parallel with the critical role of UV light in TRPA1 ( A ) activation , blocking wavelengths below ~400 nm with a titanium-dioxide-coated glass filter ( Hossein Habibi et al . , 2010 ) ( Figure 6—figure supplement 1c , Right ) abolished the spiking responses to the level of those seen in the TrpA1ins neurons ( Figure 6b ) . Also , polychromatic light at an intensity of 57 . 1 mW/cm2 readily induced feeding inhibition that required TrpA1 , and UV filtering also significantly suppressed the feeding deterrence ( Figure 6d ) . In oocytes , TRPA1 ( A ) s but not TRPA1 ( B ) s showed current increases when subjected to a series of incrementing intensities of Xenon light ( Figure 6f , and Figure 6—figure supplement 1e ) . Fitting the data to the Hill equation yielded EI50s of 9 . 8 ± 4 . 1 and 2 . 5 ± 0 . 7 mW/cm2 for fly and mosquito TRPA1 ( A ) s , respectively , revealing that TRPA1 ( A ) s are sufficiently sensitive for detection of natural day light intensities . In terms of current amplitudes , agTRPA1 ( A ) generated ~6 times more robust light-induced currents at −60 mV than did the fly ortholog isoform at the highest light intensity used . The UV filter significantly decreased the current responses , indicating the importance of UV in TRPA1 ( A ) stimulation by white light . Furthermore , the nucleophilicity-specific mutants TRPA1 ( A ) C105A and TRPA1 ( A ) R113A/R116A expressed in oocytes behaved like the nucleophile-insensitive TRPA1 ( B ) isoform in response to white light ( Figure 6—figure supplement 1e ) . These results suggest that visible light with relatively short wavelengths can substantially contribute to the excitation of TrpA1 ( A ) -positive neurons , as white light from the Xenon arc lamp contains UV light at an intensity insufficient for robust activation of TrpA1 ( A ) -positive taste neurons . To test this possibility , the fly labellum was illuminated with 470 nm blue light at 10 s durations at doses that were sequentially increased from 33 to 186 mW/cm2 , and action potentials were registered from TrpA1-positive i-a bristles ( Figure 6—figure supplement 3 ) . The serial pulses of illumination elicited spikings above the intensity of 63 mW/cm2 in a TrpA1–dependent manner , indicating that blue light contributes to polychromatic TRPA1 ( A ) activation in assistance of UV . In contrast , 30 sec-long illumination with green light ( 540 nm ) rarely evoked spikings , even at a high intensity ( 362 mW/cm2 ) , demarcating the wavelengths capable of sufficient photochemical production of free radicals . Taken together , nucleophile sensitivity enables TRPA1 ( A ) to detect natural solar radiation , and thus suppress feeding behavior in flies . To corroborate the role of free radicals in light-induced TRPA1 ( A ) activation , we investigated whether UV-induced TRPA1 activation could be hindered by quenching either nucleophilicity or electrophilicity , as radicals are amphiphilic . Since electrophiles react with nucleophiles , electrophilic NMM and benzyl isothiocyanate ( BITC ) were used as nucleophile scavengers , while the nucleophiles DTT and BTC were used as electrophile scavengers ( BTC and BITC are isosteric but opposite in chemical reactivity ) . Because these compounds are TRPA1 ( A ) agonists , they are expected to increase rather than decrease TRPA1 ( A ) activity . The agonist concentrations used were selected to be lower than those that elicit fast activation of TRPA1 ( A ) ( Du et al . , 2015 ) . Interestingly , pre-application of each chemical to the i-a bristles via the recording electrode lowered the frequencies of UV-evoked action potentials , regardless of scavenging polarity ( Figure 7a , b ) . As Drosophila taste neurons may harbor multiple sensory signaling pathways , we suspected that the observed inhibition of neuronal excitation may have resulted from activation of inhibitory pathways in the bitter-tasting cells . To examine this possibility , scavenger efficacy was assessed in sweet-sensing Gr5a-Gal4 cells exogenously expressing TrpA1 ( A ) . Similar suppression of UV-induced TRPA1 ( A ) activation was observed when DTT and NMM were applied in these cells ( Figure 7c ) , supporting that mitigation of the TRPA1 UV responsiveness by the scavengers is unlikely to involve activation of inhibitory pathways . 10 . 7554/eLife . 18425 . 023Figure 7 . Nucleophilicity is required for UV or free radical-evoked TRPA1 ( A ) activation . ( a and b ) Nucleophile-scavenging electrophiles or electrophile-scavenging nucleophiles suppress UV responses in i-a bristles . Electrophiles are in red and nucleophiles are in blue . Typical results are presented in ( a ) , with mean and SEM values provided in ( b ) as bar graphs . *p<0 . 05 , ANOVA Dunn’s test . The numbers of conducted experiments are given at the bottom of each bar . ( c ) Ectopically expressed TRPA1 ( A ) in sugar-sensing cells of L-bristles shows a reduced UV response in the presence of DTT or NMM . The dashed box indicates the data set that is presented in the right panel as bar graphs . **p<0 . 01 , Tukey’s test . ( d ) Chemically generated free radicals activate agTRPA1 ( A ) in oocytes when ammonium persulfate ( APS ) and tetramethylethylenediamine ( TEMED ) were incubated for >30 min ( see also Figure 7—figure supplement 1 ) . Activation is abrogated by incubation with the nucleophile scavenger NMM . Left: a representative recording . Right: the averaged net effects of the APS/TEMED mixture on agTRPA1 ( A ) activation with or without NMM . **p<0 . 01 , Student’s t-test ( n = 4 and 6 and 6 ) . ( e ) NMM does not act as an antagonist of heterologous agTRPA1 ( A ) with either APS or TEMED alone . Left: a typical result with APS . Middle and Right: summary of APS ( n = 5 ) and TEMED ( n = 4–5 ) experiments , respectively . ###: p<0 . 001 , paired t-test . *p<0 . 05 , Student’s t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 02310 . 7554/eLife . 18425 . 024Figure 7—figure supplement 1 . TRPA1 ( A ) -specific activation by the APS+TEMED mixture is time-dependent . ( a ) The net current increases resulting from the APS and TEMED mixture ( 0 . 1 and 0 . 01 mM , respectively ) are binned into two groups based on incubation time as indicated and compared for their efficacy in activating agTRPA1 ( A ) expressed in frog oocytes ( n = 6 and 10 and 10 ) . ***p<0 . 001 , Student’s t-test . ( b ) agTRPA1 ( B ) did not respond to the indicated chemicals , while it generated large NMM current in oocytes . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 024 However , we cannot completely rule out that , by chance , both types of taste cell share inhibitory pathways that are activated by the scavengers . Therefore , the effect of the nucleophile scavenger NMM on free radical-induced TRPA1 ( A ) activation was tested in heterologous frog oocytes . Addition of tetramethylethylenediamine ( TEMED ) and ammonium persulfate ( APS ) initiates polymerization reactions , such as solidification of polyacrylamide gel , by generating free radicals ( Shirangi et al . , 2015 ) . To examine the responsiveness of TRPA1 ( A ) to free radicals , frog oocytes expressing agTRPA1 ( A ) were exposed to a mixture of 0 . 01 mM TEMED and 0 . 1 mM APS . APS alone activated agTPRA1 ( A ) but not agTRPA1 ( B ) ( Figure 7d , and Figure 7—figure supplement 1b ) , as persulfates , like peroxides , are also nucleophilic due to the alpha effect ( Edwards and Pearson , 1962 ) . To evaluate the net effect of radicals produced by the joint application of TEMED and APS , the cells were serially challenged in the order of 0 . 01 mM TEMED , 0 . 1 mM APS , and the TEMED and APS mixture ( 0 . 01 and 0 . 1 mM , respectively ) ( Figure 7d , Left ) . Beginning thirty minutes after mixing ( Figure 7—figure supplement 1a ) , the APS/TEMED mixture activated agTRPA1 ( A ) more robustly than did APS or TEMED alone . The 30 min latency in efficacy of the mixture is reminiscent of the incubation time necessary for solidification of a typical polyacrylamide gel after addition of APS/TEMED . Interestingly , the stimulatory effect of APS/TEMED co-incubation was abolished by adding nucleophile-scavenging NMM at 0 . 01 mM ( Figure 7d ) . To test if NMM suppresses the action of each chemical component , either APS or TEMED was mixed with NMM for 1 hr and then applied to agTRPA1 ( A ) -expressing cells . These experiments resulted in increases rather than decreases in the agTRPA1 ( A ) current ( Figure 7e ) , possibly reflecting the typical role of NMM as an electrophilic agonist of TRPA1 isoforms ( Kang et al . , 2012 ) . Therefore , it is conceivable that free radicals produced by incubation of APS and TEMED activate agTRPA1 ( A ) , which is readily antagonized by nucleophile-scavenging NMM . Thus , the nucleophilic nature of amphiphilic free radicals is critical for activation of TRPA1 ( A ) , providing the mechanistic basis of light-induced feeding deterrence .
It is well documented that insect phytophagy is increased when UVB light is filtered out ( Bothwell et al . , 1994; Rousseaux et al . , 1998; Zavala et al . , 2001 ) . The effect of UVB illumination can result from changes in plant physiology ( Kuhlmann , 2009 ) or direct detection by insect herbivores ( Mazza et al . , 1999 ) . We discovered that UV and visible light activate TRPA1 ( A ) via a photochemical reaction that generates free radicals , thus inhibiting food ingestion by fruit flies . TRPA1 ( A ) -expressing taste neurons appear to be responsible for feeding deterrence as light receptor cells , on the basis of three lines of evidence . First , TRPA1 ( A ) -expressing neurons fire robustly in response to UV illumination . Second , misexpression and heterologous expression of TRPA1 ( A ) confer light sensitivity to cells , suggesting that TRPA1 ( A ) expression is sufficient for light responsiveness . Third , expression of a dominant negative mutant TRPA1 ( A ) in bitter-sensing cells via Gr66a-Gal4 eliminates light sensitivity , as assessed by feeding suppression as well as electrophysiological recordings . Because many insect genomes contain exons encoding TRPA1 ( A ) ( Kang et al . , 2012 ) , it would be interesting to further investigate whether TRPA1 ( A ) expression is responsible for light sensitivity in other insects . The high responsiveness of agTRPA1 ( A ) observed in this study implies that TRPA1 ( A ) -dependent light detection might be a general function in insects . Our analyses of light irradiance required for Drosophila feeding deterrence revealed that feeding inhibition can readily occur in response not only to UV but also to strong white light , which is likely capable of inducing nucleophilic radicals in the intracellular environment . It is conceivable that the balance between attraction by the visual system and repulsion by TrpA1-dependent light sensors shapes overall behavioral outcomes in natural settings under illumination with polychromatic light and that strong solar irradiation , which produces a sufficient amount of free radicals for TRPA1 ( A ) activation , shifts the net behavioral outcomes towards repulsion . Light-induced feeding suppression is expected to occur in the middle of the day when insects are exposed to intense solar illumination . Indeed , the biting rhythm of mosquitoes is mostly out of the day time when solar irradiance is at its strongest ( Pates and Curtis , 2005 ) . In order to avoid harmful stimuli , animals need to overcome their urge to attractive stimuli , such as food . Feeding suppression may be a requisite for migration to shaded places , which suggests that flies may exhibit a negative phototaxis driven by light-induced TRPA1 ( A ) activation . Photochemical reactions underlie rhodopsin-mediated visual mechanisms , where photon-dependent actuation of retinal covalently bound to opsin triggers a biochemical signaling cascade and an electric potential shift in the photoreceptor . We found that UV and high energy visible light , which induces photochemical generation of free radicals in the biological tissues , can be sensed without the need of a cofactor like retinal , because the basic and shared property of the radicals , such as nucleophilicity , is sensed by TRPA1 ( A ) s . Detecting electrophilicity of reactive chemicals has been regarded as the key feature of the molecular chemical nociceptor TRPA1 in bilaterian animals ( Kang et al . , 2010 ) , probably because of evolution of bilaterians in oxygen-rich surroundings . Because strong nucleophilicity is short-lived in the oxidative environment on Earth , animals may not have had much opportunity to adapt to the need of nucleophile detection . However , small organisms could have been under greater evolutionary pressure to develop a sensitive nucleophile-sensing mechanism . Their small size likely predisposes such organisms to be vulnerable to the effects of photochemically active light because of their high surface area-to-volume ratios , which translates into more incoming UV toxicity for a given disintoxicating capacity . The solar energy embedded in the form of light induces nucleophilicity in the cytosol while passing through the oxidizing atmosphere . We found that insects can respond to photochemically induced nucleophilicity with TRPA1 ( A ) for sensitive and rapid detection of solar illumination . The domain for reception of nucleophilicity appears to reside in the cytoplasmic side of TRPA1 ( A ) , as the conserved residues in the cytosolic N-terminus are required for this function . Presumably , free radicals induced by photochemical reactions in the cytoplasm may remain nucleophilic longer than those in the extracellular oxidative environment due to the reducing environment within the cell . However , a receptor with a high level of nucleophile responsiveness would not well operate in this context . The cytosol is filled with reducing nucleophiles essential for redox homeostasis , which would keep the putative nucleophile receptor open and collapse the transmembrane cation gradients . Capable of synergy between the two opposing activation pathways ( Figure 8 ) and tuned to conduct a limited nucleophile-dependent current , Drosophila TRPA1 ( A ) is able to detect light-generated amphiphilic radicals without much disturbance from the cytosolic reducing power . 10 . 7554/eLife . 18425 . 025Figure 8 . Schematic illustration of the TRPA1 ( A ) activation mechanism in response to solar UV irradiation . Solar irradiation inflict both electrophilicity and nucleophilicity in the cytosol . The two opposing characteristics of radicals were sensed by two distinct domains of TRPA1 ( A ) . Electrophilicity is often neutralized by the cytosolic reducing power , but nucleophilicity is not interfered with . C and R in the circles respectively represent cysteine and arginine critical for each activation pathway . Chevrons depict amphiphilicity of free radicals with two opposing ends . DOI: http://dx . doi . org/10 . 7554/eLife . 18425 . 025 The high nucleophile responsiveness of agTRPA1 ( A ) suggests that mosquitoes were in more need of a sensitive mechanism for nucleophile detection and , thus , probably adopted a way to suppress basal activation of TRPA1 ( A ) by the cytosolic reducing power . In general , nucleophiles carrying extra electrons are able to form stable coordinate bonds with metal ions . Strong nucleophiles such as carbon monoxide ( CO ) and cyanide anions ( CN- ) mainly exert their fatal toxicity by masking Fe2+ , which is essential for the function of heme proteins such as hemoglobins and cytochromes ( GRUT , 1954; Krahl and Clowes , 1940 ) . Thus , the differential nucleophile responsiveness between TRPA1 ( A ) s may reflect the varying needs for avoidance due to divergent susceptibility of insects to these toxic compounds as well as strong solar irradiation . In addition , plants produce a wide variety of nucleophilic antioxidants such as phenolics , carotenoids and thiol compounds ( Pandey and Rizvi , 2009; Lü et al . , 2010 ) , which suggests that nucleophile sensitivity may represent the ecological relationship of an insect species with plants . While being nectarivorous , hematophagous mosquitoes are apparently less dependent on plants for reproduction than are phytosaprophagous fruit flies ( Markow and O’Grady , 2008 ) . It is also plausible that mosquitoes are equipped with a heightened nucleophile detection mechanism in order to avoid dead animals when searching for a fresh blood meal , as decomposing animal carcasses emit nucleophilic gases ( Dent et al . , 2004 ) . Therefore , the feeding niches of the species seem to be correlated with the nucleophile sensitivities of TRPA1 ( A ) s , although it has yet to be investigated if elevated TrpA1-dependent nucleophile sensitivity invariably accompanies hematophagy in other insect species . Conversely , the residual nucleophile sensitivity of fly TRPA1 ( A ) implies that the ability to detect free radical-producing light is critical to the animal , as the nucleophile responsiveness of TRPA1 ( A ) has been evolutionarily preserved , despite the close association of Drosophila with plants , ever since the nucleophile sensitivity evolved in a putative common ancestor of Drosophila and Anopheles . TRPA1 ( B ) has been widely used as a thermogenetic tool to remotely control neurons of interest ( Bernstein et al . , 2012 ) , and can respond to IR , which elevates the temperature of irradiated tissue ( Kang et al . , 2012 ) . On the other hand , TRPA1 ( A ) not only lacks thermal sensitivity for its devotion to a chemosensory role , but also detects photochemically active light such as UV light through its nucleophile sensitivity to radicals . Thus , the isoform divergence of insect TRPA1 extends its ability to perceive solar electromagnetic radiation to UV and IR , the two spectra flanking visible light . As TRPA1 ( A ) independently serves as a UV/light sensor in nonnative cells and TRP channels show much larger single channel conductance than does channelrhodopsin-2 ( Bernstein et al . , 2012 ) , TRPA1 ( A ) would make an excellent optogenetic actuator that can be directly stimulated by UV light without the need for additional other proteins or chemical cofactors . Although UV radiation may be toxic to the illuminated neurons , TRPA1 ( A ) requires a reasonable light intensity ( <5 . 2 mW/cm2: ~4% of SC and 85% of total ground UV intensity ) for substantial neuronal excitation compared to the blue light intensities ( 2 , 000 to 7 , 500 mW/cm2: 1 , 400 to 5 , 400% of SC ) ( Cardin et al . , 2010 ) used in typical optogenetic applications . Aside from the thermal stress that likely results from intense illumination , blue light irradiation used for excitation of channelrhodopsin-2 also generates free radicals in cells at intensities as low as 2 . 8 mW/cm2 ( Godley et al . , 2005 ) through ubiquitous and indispensable cellular compounds such as flavins ( Eichler et al . , 2005 ) , and prolonged illumination with blue light can result in organismal death ( Hori et al . , 2014 ) . Therefore , the use of TRPA1 ( A ) in combination with low-density UV illumination might be beneficial in that TRPA1 ( A ) may be sufficiently robust coupled with very weak but specific transcriptional promoters given its large single channel conductance . In conclusion , the nucleophile sensitivity of TRPA1 ( A ) not only aids insects in properly responding to solar irradiation and toxic nucleophiles , but also may potentially be used to develop a superior optogenetic tool for neural circuitry studies in various model systems .
The UAS-TrpA1 ( A ) and UAS-TrpA1 ( B ) transgenic lines and TrpA1ins were previously described ( Kang et al . , 2012; Rosenzweig et al . , 2008 ) . The UAS-TrpA1 ( A ) C105A and UAS-TrpA1 ( A ) R113A/R116A transgenic lines were generated by site-specific transgenesis ( Groth et al . , 2004 ) ( Rainbow Transgenic Flies , CA , USA ) , inserted to the attp16 site as the UAS-TrpA1 ( A ) and UAS-TrpA1 ( B ) lines . The Gr66a-Gal4 ( Dunipace et al . , 2001 ) and Gr5a-Gal4 ( Marella et al . , 2006 ) lines were gifts from Drs . Hubert Amrein and Kristin Scott , respectively . In vivo taste cell recordings were performed as detailed previously ( Kang et al . , 2012 ) . Briefly , bristles were identified based on the previously described sensillum map of the labellum ( Weiss et al . , 2011; Tanimura et al . , 2009 ) . The TrpA1-dependent response to NMM was observed in most i-bristles as reported previously ( Kang et al . , 2012 ) . Tricholine citrate ( TCC ) at 30 mM was used as an electrolyte in the glass recording electrodes . Chemicals were solubilized in the electrolyte solution , and then applied to taste neurons . Spiking frequencies to chemicals were calculated for entire recordings except for H2O2 recording in L bristles , for which spiking frequencies were calcuclated from the first 10 s . Spike amplitudes from Gr5a cells expressing TrpA1 ( A ) often gradually decreased to 0 mV within 20 s probably due to exhaustion of robustly firing cells . For the first 20 s of UV response recordings , the basal activity of neurons in the bristle was monitored , after which time UV illumination was administered to the sensilla for 20 s using optical fiber-coupled UV LEDs ( FCS-0295–000 , Mightex , CA , and UVTOP295 , Qphotonics , MI , USA for UVB at 295 nm and M365FP1 , Thorlabs , USA for UVA at 365 nm ) controlled by an SLA-series two-channel LED driver ( SLA-0100–2 , Mightex ) and a T-Cube LED driver ( LEDD1B , Thorlab , USA ) , respectively . The maximal optical fiber output of 295 nm UV was 0 . 063 mWusing a ball-lens type LED and that of 365 nm UV was 0 . 3 mW . These net power outputs at the tip of the optical fiber were measured with a photodiode sensor ( S120VC , Thorlabs , NJ , USA ) connected to a digital console ( PM100D , Thorlabs , NJ , USA ) . Illumination intensity was calculated by considering the size of illuminated area derived from the numerical aperture ( NA ) values of the optical fibers and the distance to the samples . Due to the complex shape of fly taste bristles on the labellum and various illumination angles between the light beam and tissue , we simplified the calculation by postulating a 45° angle and oval illumination area at a distance ( Figure 1—figure supplement 1d ) . For oocytes , circular areas were calculated ( Figure 1—figure supplement 1e ) . Blue and green light illumination was accomplished using a GFP or RFP excitation filter ( 470 or 540 nm with a bandpath of 50 , respectively ) equipped with a typical fluorescence microscope . The UV filter for experiments with white light consisted of glass deposited with nanolayers of titanium dioxide ( custom-made , Seoul Precision Optics , Seoul , Korea ) . Flies prepared for sensillum recording in response to light were used once to record from a single bristle , in order to test only naïve cells . The reference electrode containing hemolymph-like solution 3 . 1 ( HL3 . 1 ) ( Feng et al . , 2009 ) was inserted close to the labella taste neuron cell bodies from the back of the fly thorax , which held the proboscis in an extended configuration in order to minimize electrical noise stemming from movement of the live animal . Tasteprobe ( Syntech , Netherlands ) was used as a preamplifier to register the action potentials from the neurons , which were digitized with Powerlab ( ADI instruments , Australia ) . The obtained spiking frequencies were analyzed by Labchart ( ADI instruments , Australia ) . Non-responding bristles were re-tested with other agonists that activate the same neurons as indicated in the main text ( Figure 1—figure supplement 2 and Figure 3—figure supplement 1 ) . To quantitatively evaluate the impact of UV irradiation and chemicals on feeding deterrence , the capillary feeder ( Café ) assay ( Ja et al . , 2007 ) was used with minor modifications . In particular , feeding avoidance upon UV illumination was determined using two sibling populations of 16 hr starved flies . One population , consisting of a vial containing 20–23 flies 2–3 days of age was illuminated with 312 nm UV light with a UV lamp ( NB-UVB 311–313 nm , ATObeam , Goyang , Korea; UVB lamp , PL-S 9 W/01 , Phillips , Netherlands ) , 365 nm UV light ( LF-204 . LS UVlite ultraviolet lamps , UVITEC , Cambridge , UK ) , or with white light from a DG4 Xenon arc lamp ( Sutter , CA , USA ) at a distance of 2 . 5 cm from the standing vial , while the other group , which had a similar number of flies , was allowed to feed freely and was left untreated at the same time ( Figure 1c ) . Irradiance was measured as ~1 . 8 , 4 , and 57 . 1 mW/cm2for UVA , UVB , and white light , respectively , using an excised piece of a vial covering the photodiode probe ( S120VC , Thorlabs , NJ , USA ) to simulate internal irradiation . The vials were made of polypropylene , which has a low rate of UV transmission ( Kruenate et al . , 2004 ) , resulting in increased internal temperature , as described in Figure 1—figure supplement 3 . To minimize thermal accumulation , the UV-illuminated vial was actively cooled by fan-driven air flow while the internal temperature of a separately illuminated vial was concurrently monitored . After each feeding session , the change in the level of the menisci of 30 mM sucrose solutions in three calibrated glass capillary tubes ( #2920107 , Marienfeld , Lauda-Königshofen , Germany , 15 mm/μl ) was measured . Following measurement of the evaporated volume obtained from vials without flies , the distance readings were converted to volume measurements . The ingested volume per animal was then used to calculate an 'avoidance index' by dividing [ingested volume per fly in the sucrose-only vial minus ingested volume per fly in the UV-plus-sucrose vial] by the sum of ingestion volume per fly in either vial . For the Café assay for H2O2 , two capillaries containing the same solution were inserted into a vial together with two other capillaries with other tastants . The use of multiple capillaries for a single tastant mixture suppresses experimental variation , presumably owing to higher exposure of flies to tastants and an averaging effect between feeding amounts in separate tubes . To obtain an avoidance index , the volume of H2O2+sucrose consumption was subtracted from the volume of sucrose-only consumption , the result of which was in turn divided by total ingested volume . The proboscis extension reflex ( PER ) assay was performed with modifications as previously described ( Kang et al . , 2010; Kang et al . , 2012 ) . UV or IR-induced PER was monitored in TrpA1-deficient flies expressing either TrpA1 ( A ) or TrpA1 ( B ) in Gr5a-Gal4 cells . Flies that had been starved overnight were glued to glass slides , water-satiated , and illuminated with 254 nm UV light at an intensity of 0 . 28 mW/cm2 ( LF-204 . LS UVlite ultraviolet lamps , UVITEC Cambridge , UK ) for 2 min , during which time PER frequency was scored . When a fly fully extended its proboscis 10 times or more , a maximum score of one was given . The PER score of a fly that extended its proboscis fewer than 10 times was calculated by dividing the number of proboscis extensions by 10 . For IR-evoked PER , IR from a radiant heater ( 940 watt , JD07010-1002 , iSolar , Inchon , Korea ) was administered at a distance of 20 cm from the fly . UVA radiation at 365 nm was administered for 20 s from the bottom side of a horizontally placed vial ( Figure 6—figure supplement 2b ) that contained 3–4-day-old adult flies . Attraction indices were calculated by determining the fraction of the flies in the half of the vial close to the UVA source . TRPA1-dependent currents in Xenopus laevis oocytes induced by application of chemicals and light illumination were recorded by the two-electrode voltage clamping technique ( TEVC ) , as described previously ( Kang et al . , 2010; Kang et al . , 2012 ) . Briefly , ovaries were surgically prepared and subjected to digestion with 1 . 5 mg/ml collagenase for 1 . 5 hr . Subsequently , the follicular layer of the oocytes was manually removed . One day after microinjection of 50 nl of TrpA1 cRNA , oocytes were electrophysiologically examined while perfused with the recording solution ( 96 mM NaCl , 1 mM KCl , 1 mM MgCl2 , 5 mM HEPES , pH 7 . 6 ) . For UV illumination , the optical fiber terminal was mounted above the cell at a minimal distance to achieve the highest possible intensity ( Figure 1—figure supplement 1c ) . H2O2 ( HP1002 , GeorgiaChem , GA , USA ) and DTT ( 43819 Sigma Aldrich , MO , USA ) solutions were freshly prepared before use . For UV experiments , the initial voltage was −60 mV , and it was then changed in periods of 300 ms from −60 to +60 mV per second . For H2O2 and DTT responses , the voltage was held constant at −60 mV during recording . The current was amplified with a GeneClamp 500B amplifier ( Molecular Devices , CA , USA ) and registered by a digitizer ( Digidata 1440 A , Molecular Devices , CA , USA ) . Data from dose-dependence experiments were normalized with respect to 0 . 1 mM NMM currents recorded from the same cells , and fitted to the Hill equation using Sigmaplot12 . Patch-clamp recordings were carried out in an inside-out configuration using macropatches excised from Xenopus oocytes expressing TRPA1 . Currents were recorded with an EPC 10 patch-clamp amplifier ( HEKA Instruments , Germany ) controlled by Patchmaster ( HEKA Instruments , Germany ) . All current recordings were sampled at 10 kHz and filtered at 1 kHz . The patch pipettes were pulled from borosilicate capillaries ( Hilgenberg-GmbH , Germany ) using a Narishige puller ( PC-10 , Narishige , Tokyo , Japan ) . The patch pipettes had a resistance of 3 ~ 5 M when filled with pipette solution containing 130 mM NaOH , 3 mM HEPES , and 0 . 5 mM Na-EDTA adjusted to pH 7 . 6 with HCl . Cells were bath-perfused with a solution of 130 mM NaOH , 3 mM HEPES , and 1 mM MgCl2 , pH 7 . 6 , with HCl . An oocyte was shrunk in a hypertonic solution and the vitelline membrane was removed with forceps to access the plasma membrane . All recordings were carried out at room temperature . The currents from Xenopus oocytes were studied by holding the potential at 0 mV and ramped from -100 to +100 mV for 500 ms and then returned to 0 mV . Currents were analyzed and fitted using Patchmaster ( HEKA Instruments , Germany ) and Origin6 . 0 ( MicroCal , MA , USA ) . To compute proper sample sizes , we used the G power program available at www . gpower . hhu . de ( Faul , 2009 ) . To detect differences with 80% power between the mean values of two independent groups , four replicates in each group were necessary for a Student’s t-test with typical parameters ( alpha = 0 . 05 , effect size d = 3 ) . For ANOVA Tukey’s HSD tests with alpha = 0 . 05 and effect size f = 30 , three independent samples in each group were needed to compute a difference between the mean values of two independent groups in multiple comparisons . Student’s t-tests , ANOVA Tukey’s multiple comparison , ANOVA repeated measures , ANOVA Dunn’s test , and Mann-Whitney U tests were performed with Sigmaplot12 . Error bars indicate the standard error of mean ( SEM ) . Normality was tested by the Shapiro-Wilk method . When data failed to pass either normality or equal variance tests , they were analyzed by rank sum tests , such as Mann-Whitney U and ANOVA Dunn’s tests . Unless indicated otherwise , '* , ' '** , ' and '***' or '# , ' '## , ' and '###' represent p values of <0 . 05 , <0 . 01 , and <0 . 001 , respectively . Two groups of data were examined by Student’s t- or Mann-Whitney U tests . When comparing three or more groups , ANOVA tests were used . All averaged data points and error bars represent the mean±SEM , unless indicated otherwise . | Atoms are made up of a nucleus that contains protons and neutrons , which is orbited by electrons . The electrons orbit within shells that surround the nucleus and each shell can contain a specific number of electrons . A particle with an outer shell that is missing one or more electrons will be unstable and highly reactive . It will attempt to achieve a full outer shell either by sharing electrons with another particle , or by donating or stealing an electron . Particles that steal electrons are said to be “electrophilic” ( electron-loving ) while those that donate them are “nucleophilic” . Electrophilic and nucleophilic particles can damage DNA and proteins . In species from fruit flies to humans , electrophilic substances such as formaldehyde activate a type of ion channel called TRPA1 . These ion channels contribute to pain signaling , and their activation triggers unpleasant and painful sensations that deter animals from getting too close to electrophilic substances . However , it is not known if animals have an equivalent mechanism to help them avoid toxic nucleophilic compounds , like carbon monoxide and cyanide . Du , Ahn , Wen , Seo , Na et al . now show that fruit fly neurons produce two versions of the TRPA1 channel: one that is sensitive to electrophiles , plus a second that is sensitive to nucleophiles in addition to electrophiles . The existence of nucleophile-sensitive TRPA1 helps explain why fruit flies avoid feeding in strong sunlight . Ultraviolet radiation in sunlight triggers the production of reactive forms of oxygen that behave as strong nucleophiles . These reactive oxygen species – which can damage DNA – activate the nucleophile-sensitive TRPA1 and thereby trigger the fly’s avoidance behavior . Human TRPA1 responds only to electrophiles and not to nucleophiles . By targeting the nucleophile-sensitive version of insect TRPA1 , it may thus be possible to develop insect repellants that humans do not find aversive . Furthermore , TRPA1s from some insect species are more sensitive to nucleophiles than others , with a mosquitoes’ being more sensitive than the fruit flies’ . This means that insect repellants that target nucleophile-sensitive TRPA1 could potentially repel malaria-transmitting mosquitoes without affecting other insect species . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Nucleophile sensitivity of Drosophila TRPA1 underlies light-induced feeding deterrence |
Visceral adiposity confers significant risk for developing metabolic disease in obesity whereas preferential expansion of subcutaneous white adipose tissue ( WAT ) appears protective . Unlike subcutaneous WAT , visceral WAT is resistant to adopting a protective thermogenic phenotype characterized by the accumulation of Ucp1+ beige/BRITE adipocytes ( termed ‘browning’ ) . In this study , we investigated the physiological consequences of browning murine visceral WAT by selective genetic ablation of Zfp423 , a transcriptional suppressor of the adipocyte thermogenic program . Zfp423 deletion in fetal visceral adipose precursors ( Zfp423loxP/loxP; Wt1-Cre ) , or adult visceral white adipose precursors ( PdgfrbrtTA; TRE-Cre; Zfp423loxP/loxP ) , results in the accumulation of beige-like thermogenic adipocytes within multiple visceral adipose depots . Thermogenic visceral WAT improves cold tolerance and prevents and reverses insulin resistance in obesity . These data indicate that beneficial visceral WAT browning can be engineered by directing visceral white adipocyte precursors to a thermogenic adipocyte fate , and suggest a novel strategy to combat insulin resistance in obesity .
Adipocytes are critical regulators of energy balance and nutrient homeostasis . White adipocytes serve as the principal site for energy storage in mammals . These cells are characterized by a large unilocular lipid droplet and have the capacity to store or release energy depending on metabolic demand . White adipocytes also produce and secrete numerous cytokines and hormones that impact several aspects of physiology ( Rosen and Spiegelman , 2014 ) . Eutherian mammals also contain a second major class of adipocytes that function to catabolize stored lipids and produce heat . These thermogenic adipocytes , consisting of brown and beige/BRITE adipocytes , are characterized by their multilocular lipid droplet appearance , high mitochondrial content , and expression of uncoupling protein 1 ( Ucp1 ) ( Cohen and Spiegelman , 2015; Harms and Seale , 2013 ) . Brown/beige adipocytes have promising therapeutic potential as their activation in the setting of obesity has a profound impact on metabolic health . White adipose depots are broadly categorized as either subcutaneous or visceral adipose tissue , reflecting their anatomical location . Adipose tissue distribution is a strong predictor of metabolic outcome in obese individuals ( Karpe and Pinnick , 2015; Lee et al . , 2013 ) . Visceral adiposity strongly correlates with the development of insulin resistance , diabetes , and cardiovascular disease ( Kissebah et al . , 1982; Krotkiewski et al . , 1983; Ohlson et al . , 1985; Vague , 1956 ) . Preferential expansion of subcutaneous depots is associated with sustained insulin sensitivity ( Manolopoulos et al . , 2010 ) . Engineered rodent models highlight the protective role of subcutaneous adipose tissue . Transgenic animals overexpressing adiponectin or mitoNEET develop extreme subcutaneous obesity; however , these animals remain metabolically healthy ( Kim et al . , 2007; Kusminski et al . , 2012 ) . In humans , the location of visceral adipose tissue itself likely mediates some of its detrimental effects on energy metabolism; lipids , metabolites , and cytokines can drain directly into the portal circulation and affect liver function ( Rytka et al . , 2011 ) . Transplantation studies , cellular studies , and gene expression analyses , suggest that factors intrinsic to these depots may also determine their effect on nutrient homeostasis ( Tran et al . , 2008; Yamamoto et al . , 2010 ) . Anatomically distinct adipocytes are functionally unique , differing in their ability to undergo lipolysis , lipogenesis , and activate the thermogenic gene program ( Lee et al . , 2013; Macotela et al . , 2012; Morgan-Bathke et al . , 2015; Wu et al . , 2012 ) . Along these lines , lineage analyses reveal that anatomically distinct white adipocytes can originate from developmentally distinct precursor cells , and emerge at different times during development ( Billon and Dani , 2012; Chau et al . , 2014 ) . As such , it is now widely believed that visceral and subcutaneous adipocytes represent distinct subtypes of fat cells . In mice , visceral and subcutaneous white adipose depots differ remarkably in their ability to remodel under physiological conditions ( Hepler and Gupta , 2017 ) . Upon high-fat diet feeding , visceral adipose depots of mice expand by both adipocyte hypertrophy and through the formation of new adipocytes ( ‘adipogenesis’ ) ( Wang et al . , 2013 ) . Inguinal subcutaneous WAT expands predominantly through adipocyte hypertrophy . The differential capacity for adipogenesis is likely explained by factors present in the local microenvironment ( Jeffery et al . , 2016 ) . Another notable difference between the inguinal and visceral adipose depots in rodents is the capacity to adopt a thermogenic phenotype . Various stimuli , including β3-adrenergic receptor-agonism and cold exposure , drive the rapid accumulation of beige adipocytes in subcutaneous depots ( Vitali et al . , 2012; Wu et al . , 2012 ) . Genetic stimulation of subcutaneous beige adipogenesis renders mice resistant to high-fat diet induced obesity and/or diabetes ( Seale et al . , 2011; Shao et al . , 2016 ) , while inhibition of subcutaneous beiging leads to an earlier onset of insulin resistance during obesity ( Cohen et al . , 2014 ) . Most visceral depots in mice , particularly the gonadal and mesenteric adipose tissues , are relatively resistant to browning in response to physiological stimuli . With few exceptions ( Kiefer et al . , 2012 ) , most engineered mouse models of white adipose tissue browning exhibit beige cell accumulation preferentially in subcutaneous WAT depots ( Seale et al . , 2011; Stine et al . , 2016 ) . Visceral WAT depots may harbor mechanisms to suppress thermogenesis to ensure its function as a white , energy-storing , depot . We previously established a critical role for the transcription factor , Zfp423 , in the establishment and maintenance of the adipocyte lineage . Zfp423 is required for fetal differentiation of subcutaneous white adipocytes ( Gupta et al . , 2010; Shao et al . , 2017 ) . In adult mice , Zfp423 expression defines a subset of committed mural preadipocytes ( Gupta et al . , 2012; Vishvanath et al . , 2016 ) . Upon high-fat diet feeding , these mural cells undergo adipogenesis in visceral depots , contributing to adipocyte hyperplasia ( Vishvanath et al . , 2016 ) . Zfp423 is also expressed in nearly all mature adipocytes; however , its expression is more abundant in white adipocytes than brown adipocytes ( Shao et al . , 2016 ) . In the mature adipocyte , Zfp423 acts to maintain the energy-storing status of the white adipocyte through suppression of the thermogenic gene program ( Shao et al . , 2016 ) . Zfp423 likely exerts this function by serving as a transcriptional co-repressor of the brown/beige lineage determining transcription factor , Ebf2 . The loss of Zfp423 in mature white adipocytes triggers a robust conversion of white to beige adipocytes in subcutaneous WAT . Amongst the various adipose tissues , visceral WAT expresses the highest levels of Zfp423 . Importantly , we observed that visceral adipocytes lacking Zfp423 were also capable of inducing their thermogenic gene program when animals were stimulated pharmacologically with a β3 adrenergic receptor agonist ( Shao et al . , 2016 ) . This observation affords the possibility of examining whether the thermogenic capacity of visceral white adipose depots can be unlocked under physiological conditions , and whether thermogenic visceral WAT would be ultimately harmful or beneficial to systemic metabolic health . Here , we describe two mouse models of visceral adipose tissue browning derived through selective ablation of Zfp423 in visceral adipose precursors . We reveal that fetal visceral white preadipocytes can be redirected to a beige-like adipocyte fate through the loss of Zfp423 , leading to visceral depot mass reduction and fat redistribution towards subcutaneous depots . The browning of visceral depots improves cold tolerance and protects against the development of insulin resistance and hyperlipidemia in obesity . Moreover , we demonstrate that visceral mural preadipocytes in adult mice can also be directed to a thermogenic cell fate; this leads to beige-like , rather than white , adipocyte hyperplasia , in the expanding visceral WAT depots of diet-induced obese animals . Upon activation by β−3 adrenergic receptor agonism , these beige-like adipocytes can trigger an amelioration of insulin resistance in obese animals . Together , these data establish Zfp423 as a physiological suppressor of the thermogenic gene program in visceral WAT and provide proof of concept that the thermogenic capacity of visceral WAT can be induced under physiological conditions through removal of this molecular brake on the thermogenic gene program in adipose precursors . These data also highlight the potential of visceral WAT , much like subcutaneous WAT , to improve nutrient homeostasis in obesity when a thermogenic beige-like phenotype is induced .
We hypothesized that the thermogenic capacity of visceral white adipose depots can be unlocked under physiological conditions through removal of Zfp423 , and that engineered visceral thermogenic adipocytes can drive improvements in systemic metabolic health . To test this , we derived a mouse model in which Zfp423 is selectively inactivated in visceral , but not subcutaneous , WAT depots or classic brown adipose tissue depots ( BAT ) . These animals were generated by breeding transgenic mice expressing Cre recombinase under the control of the Wilms Tumor one locus ( Wt1-Cre ) to animals carrying the floxed Zfp423 alleles ( Zfp423loxP/loxP ) ( Figure 1A ) ( Zfp423loxP/loxP; Wt1-Cre animals , herein ‘Vis-KO’ mice ) . Wt1-Cre mice targets Wt1-expressing cells of the embryonic mesothelium , as well as descending cells , which include , adult mesothelial cells and intra-abdominal stromal cells within or surrounding visceral organs . Hastie and colleagues revealed that visceral white adipocytes , but not subcutaneous or classic brown adipocytes , descend from Wt1-expessing progenitors and are targeted by Wt1-Cre ( Chau et al . , 2014 ) . Specifically , Wt1-Cre targets the majority of gonadal adipocyte precursors and variable numbers of precursors in other visceral depots , including the mesenteric and retroperitoneal depots . Analysis of mRNA expression from multiple fat depots in Vis-KO mice confirmed the deletion of Zfp423 was selective to visceral WAT depots ( Figure 1B ) . 10 . 7554/eLife . 27669 . 003Figure 1 . Deletion of Zfp423 in fetal visceral white adipocyte precursors leads to the formation of thermogenic adipocytes in visceral depots . ( A ) A mouse model of visceral WAT selective ablation of Zfp423 ( Visceral-Knockout or ‘Vis-KO’ ) was derived by breeding animals expressing the gene encoding Cre recombinase under the control of the Wilms Tumor 1 ( WT1 ) promoter to animals carrying floxed Zfp423 alleles ( Zfp423loxP/loxP ) . Littermates carrying only Zfp423loxp/loxP alleles were used as control animals . ( B ) Fold change in mRNA levels within brown ( BAT ) , inguinal ( iWAT ) , gonadal ( gWAT ) , mesenteric ( mWAT ) , or retroperitoneal ( rWAT ) adipose tissue of control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6 mice . ( C ) Body weight of control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . n = 6 mice . ( D ) Fat pad weight ( normalized to body weight ) of control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6 mice . ( E– F ) Fold change in mRNA levels of mature adipocyte genes ( E ) and thermogenic genes ( F ) isolated from whole adipose tissue from control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6 mice . ( G ) mRNA levels ( normalized to Rps18 ) of Ucp1 isolated from whole adipose tissue from control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6 mice . ( H–K ) Relative basal oxygen consumption rates ( OCRs ) within diced gWAT ( H ) , mWAT ( I ) , rWAT ( J ) , and iWAT ( K ) from control ( white bar ) and Vis-KO ( red bar ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 independent replicates pooled from two mice . ( L–O ) Representative immunofluorescence staining of Perilipin ( green ) and DAPI ( blue ) in mWAT sections obtained from control ( L , M ) and Vis-KO ( N , O ) mice at 8 weeks of age . Scale bar , 200 μM . Panels M and O represent digital enhancements of the boxed regions shown in L and N , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 00310 . 7554/eLife . 27669 . 004Figure 1—figure supplement 1 . ( A–L ) Representative images of gonadal ( gWAT ) , retroperitoneal ( rWAT ) , and inguinal WAT ( iWAT ) from control and Vis-KO mice immunostained with antibodies raised against Perilipin ( green ) . Scale bar = 200 μM . Panels B , D , F , H , J , and L represent digital enhancements of the boxed regions shown in A , C , E , G , I , and K , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 004 Vis-KO mice were born in the expected Mendelian ratio and appeared grossly indistinguishable from controls . We did not observe a difference in body weight between eight weeks-old Vis-KO mice and littermate controls; however , we did observe a noticeable difference in body fat distribution ( Figure 1C ) . In comparison to control animals , Vis-KO mice had smaller gonadal WAT depots and larger subcutaneous inguinal WAT ( Figure 1D ) . mRNA levels of adipocyte-selective genes in visceral adipose depots from the Vis-KO were comparable to levels found in control animals ( Figure 1E ) ; these data suggest that visceral adipocyte differentiation per se does not depend on Zfp423 . However , in all visceral depots examined , there was a marked increase in the expression of key genes involved in adipocyte thermogenesis , including Ucp1 ( Figure 1F–G ) . This was accompanied by an increase in basal O2 consumption rates in explanted WAT depots ( Figure 1H–J ) . Furthermore , some visceral depots ( mesenteric and retroperitoneal ) lacking Zfp423 exhibit a multilocular appearance typical of thermogenic adipocytes ( Figure 1L–O , Figure 1—figure supplement 1A–L ) . Taken together , these results indicate that inactivation of Zfp423 in fetal visceral white adipocyte precursors leads to the widespread accumulation of beige-like thermogenic adipocytes within visceral adipose depots . Interestingly , despite the fact that Zfp423 was not inactivated in the subcutaneous inguinal WAT of Vis-KO animals , the thermogenic capacity of this depot was also enhanced ( Figure 1F , G , K ) . We next asked whether the browning of WAT depots in the Vis-KO animals may occur in a cell-autonomous manner . We isolated the adipose stromal vascular fraction ( SVF ) from control and Vis-KO mice and performed in vitro adipocyte differentiation assays . SVF cultures obtained from the inguinal WAT of control and Vis-KO animals differentiated to a similar degree , and no significant differences in levels of Zfp423 or thermogenic genes were observed ( Figure 2A–C ) . This is consistent with lack of Cre activity in this depot , and suggests that the increased inguinal WAT beiging in Vis-KO mice in vivo occurs secondary to inactivation of Zfp423 in the Wt1 lineage . 10 . 7554/eLife . 27669 . 005Figure 2 . Zfp423-deficient visceral adipose precursors differentiate into functional thermogenic adipocytes in vitro . ( A ) Oil Red O staining of adipocytes differentiated in vitro from inguinal and gonadal SVF of control and Vis-KO mice . ( B ) Fold change in mRNA levels of adipocyte-selective genes in inguinal adipocyte cultures differentiated as shown in A . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . ( C ) Fold change in mRNA levels of thermogenic genes in inguinal adipocyte cultures differentiated as shown in A and treated with vehicle or forskolin ( 10 µm ) for 3 hr . n = 4 replicates from two mice . ( D ) Fold change in mRNA levels of adipocyte-selective genes in gonadal adipocyte cultures as shown in A . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . ( E ) Fold change in mRNA levels of thermogenic genes in gonadal adipocyte cultures differentiated as shown in A and treated with vehicle or forskolin ( 10 µm ) for 3 hr . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . ( F ) Oxygen consumption rates ( OCRs ) within differentiated cultures of control ( white bars ) or Zfp423-deficient ( red bars ) gonadal adipocytes in the basal state , and in response to sequential additions of oligomycin ( ATP synthase inhibitor ) , FCCP ( chemical uncoupler ) , and rotenone/antimycin A ( complex I and complex III inhibitor ) . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . ( G ) Percent uncoupled respiration of differentiated cultures of control ( white bar ) or Zfp423-deficient ( red bar ) gonadal adipocytes as determined by basal and uncoupled ( oligomycin ) respiration data from Figure 2F . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . ( H ) Fold change in mRNA levels of white- , beige- , and brown-selective genes in differentiated cultures of control ( white bars ) or Zfp423-deficient ( red bars ) gonadal adipocytes . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 replicates from two mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 005 Cultures of gonadal WAT SVF from control and mutant animals also underwent adipogenesis to a similar degree under the differentiation conditions utilized ( dexamethasone , IMBX , insulin , and rosiglitazone ) . This was evident by comparable lipid accumulation and mRNA levels of adipocyte-selective genes in differentiated cultures ( Figure 2A , D ) . Importantly , Zfp423 mRNA was almost entirely absent from these cultures , reflecting the activity of the Wt1-Cre line in this depot . Cultures of gonadal white adipocytes lacking Zfp423 robustly activate their thermogenic gene program upon stimulation with forskolin ( Figure 2E ) . These changes were accompanied by increased basal , uncoupled ( oligomycin ) , and maximal ( FCCP ) respiration in differentiated gonadal adipocyte cultures from mutant animals ( Figure 2F , G ) . These data indicate that Zfp423-deficient visceral adipose precursors can differentiate into functional thermogenic adipocytes in a cell-autonomous manner . The cellular phenotype of Zfp423-deficient visceral adipocytes is reminiscent of the characterized phenotype of subcutaneous beige adipocyte cultures ( Wu et al . , 2012 ) ; therefore , we examined whether Zfp423-deficient visceral adipocyte cultures are enriched in the expression of beige-selective transcripts . Zfp423-deficient visceral adipocytes expressed lower levels of white adipocyte-selective genes; however , we did not observe a clear pattern of gene expression changes that would indicate a conversion of Zfp423-deficient visceral adipocytes to either subcutaneous beige or classic brown fat cells ( Figure 2H ) . Prior studies of the beige adipocyte determination factor , Prdm16 , revealed that genetic ablation of Prdm16 in adipose tissue leads to the loss of subcutaneous beige adipocytes , with inguinal WAT acquiring the molecular properties of visceral fat . Thus , we next asked whether the loss of Zfp423 reprograms visceral WAT into subcutaneous WAT . Alternatively , visceral WAT lacking Zfp423 may retain a global visceral phenotype but adopt a thermogenic phenotype reminiscent of classical brown or subcutaneous beige adipose tissues . In order to address this question we obtained and compared global gene expression profiles of adipose depots from control and Vis-KO animals through RNA sequencing ( RNA-seq ) . For this experiment , we treated all animals with a β−3 adrenergic receptor agonist ( CL316 , 243 ) , or vehicle , for 3 days at thermoneutrality . This treatment allowed us to capture the global gene expression profile of thermogenic adipose depots in their fully active state . CL316 , 243 treatment led to a much greater increase in mRNA levels of Ucp1 and most other thermogenic genes examined in visceral WAT depots of Vis-KO mice than in control animals ( Figure 3A , B ) . Levels of Ucp1 mRNA in the visceral depots lacking Zfp423 nearly reached levels of Ucp1 found in inguinal WAT following CL316 , 243 treatment ( i . e . subcutaneous beige adipose tissue ) ( Figure 3A ) . This was accompanied by the widespread appearance of multilocular cells in gonadal WAT with readily detectable Ucp1 protein expression ( Figure 3C–F ) . We again assayed the mRNA levels of genes commonly used as white , beige , and classic brown , adipocyte markers . Similar to results obtained from cultured Zfp423-deficient visceral adipocytes , the expression of some white adipocyte-selective genes examined were lower in Vis-KO gWAT; however , we once again did not observe a clear pattern of gene expression changes that would indicate a conversion of Zfp423-deficient gWAT to either subcutaneous beige or classic brown adipose tissue ( Figure 3G ) . 10 . 7554/eLife . 27669 . 006Figure 3 . Visceral adipose tissue deletion of Zfp423 leads to enhanced visceral browning upon β−3 adrenergic receptor agonism at thermoneutrality . ( A ) Control and Vis-KO mice were housed at room temperature ( 22°C ) until 6 weeks of age then transferred to thermoneutral housing conditions ( 30°C ) for 2 weeks . After 2 weeks at thermoneutrality , mice were treated with PBS ( vehicle ) or CL316 , 243 ( 1 mg/kg , intraperitoneal ) daily for 3 days . On the 4thday , tissues were harvested for analysis . mRNA levels ( relative to Rps18 ) of Ucp1 in iWAT , gWAT , and rWAT obtained from control and Vis-KO mice treated with vehicle or CL316 , 243 ( CL ) for 3 days at thermoneutrality . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( B ) Fold change in mRNA levels of thermogenic genes in gWAT obtained from control and Vis-KO mice treated with vehicle or CL316 , 243 ( CL ) for 3 days at thermoneutrality . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( C–F ) Representative brightfield images of Ucp1 immunoreactivity ( brown staining ) in gWAT sections obtained from control or Vis-KO mice treated with vehicle ( C , D ) or CL316 , 243 ( E , F ) . ( G ) Fold change in mRNA levels of white- , beige- , and brown-selective genes in gWAT obtained from control and Vis-KO mice treated with vehicle or CL316 , 243 ( CL ) for 3 days at thermoneutrality . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 006 Principal component analysis and unsupervised hierarchical clustering analysis of RNA-Seq data suggest that the global gene expression profile of Zfp423-deficient gWAT is distinct from subcutaneous beige adipose tissue and classic BAT . Instead , the Zfp423-deficient visceral WAT more closely resembles gWAT from control animals ( Figure 4A–B ) . We next compared the list of transcripts that are differentially expressed between β3-agonist treated Zfp423-deficient gWAT and β3-agonist treated gWAT of control animals ( i . e . genes whose expression reflect engineered thermogenic gWAT , Figure 4—source data 1 ) to the list of transcripts that are differentially regulated between β-agonist treated iWAT and vehicle-treated iWAT of control animals ( i . e . genes whose expression reflects beige inguinal adipose tissue , Figure 4—source data 2 ) . Zfp423-deficiency led to the differential expression of 1735 transcripts in gWAT ( FDR < 0 . 05 ) ( Figure 4C ) . Of these 1735 transcripts , 207 genes ( 12% ) overlap with genes differentially regulated following β3-agonist treatment of iWAT ( Figure 4—source data 3 ) . Functional classification of these 207 genes by gene ontology analysis indicated that nearly all of these genes are related to mitochondrial function and biogenesis , a hallmark of thermogenic adipocytes ( Figure 4D , E ) . These data suggest that Zfp423-deficient adipocytes present in Vis-KO animals are visceral adipocytes expressing at least a portion of the thermogenic gene program characteristic of subcutaneous beige adipocytes . 10 . 7554/eLife . 27669 . 007Figure 4 . Thermogenic Zfp423-deficient visceral adipose tissue is largely distinct from subcutaneous beige adipose tissue . ( A ) RNA-sequencing was performed on mRNA libraries derived from gonadal WAT ( gWAT ) of vehicle- and β3 agonist ( CL ) -treated control and Vis-KO mice , as well as from inguinal WAT and interscapular BAT of vehicle- and CL-treated control mice . Principle component ( PC ) analysis of sequencing data obtained from CL-treated animals . n = 3 sequenced libraries for each condition . ( B ) Unsupervised clustering dendogram of adipose depot samples . ( C ) Venn diagram depicting overlap in 1 ) transcripts that are differentially expressed between β3-agonist treated Zfp423-deficient gWAT and β3-agonist treated gWAT of control animals , and 2 ) transcripts that are differentially regulated between β-agonist treated iWAT and vehicle-treated iWAT of wild-type animals . See Figure 4—source data 1 , Figure 4—source data 2 , Figure 4—source data 3 . ( D ) Gene ontology analysis ( GOTERM_CC_FAT ) of the 207 overlapping genes shown in D . ( E ) Heatmap of top 50 induced genes of the 207 highlighted in C . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 00710 . 7554/eLife . 27669 . 008Figure 4—source data 1 . Genes whose expression is significantly altered in gonadal WAT of Vis-KO CL316 , 243 treated mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 00810 . 7554/eLife . 27669 . 009Figure 4—source data 2 . Genes whose expression is significantly altered in Inguinal WAT of CL316 , 243 treated mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 00910 . 7554/eLife . 27669 . 010Figure 4—source data 3 . List of 207 genes highlighted in Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 010 Subcutaneous beige adipocytes exert beneficial effects on nutrient homeostasis and can increase energy expenditure . The Vis-KO model affords the opportunity to explore the systemic benefits of unlocking the thermogenic phenotype of visceral adipose depots . To this end , we tested the ability of Vis-KO mice to maintain body temperature in response to cold challenge . Following 1 week of cold exposure , visceral WAT depots of the Vis-KO mice activated their thermogenic gene program ( Figure 5A , B ) and accumulated Ucp1+ multilocular adipocytes ( Figure 5C–J ) to a much greater extent than control animals . We did not observe a significant difference in core body temperature during an acute cold tolerance test of control and Vis-KO animals ( Figure 5K ) ; however , after four weeks of cold acclimation , the Vis-KO mice maintain higher core body temperature ( Figure 5L ) . Collectively , these data demonstrate that inactivation of Zfp423 is sufficient to unlock the thermogenic activity of visceral WAT depots under postnatal physiological conditions , and that browning of these depots can functionally enhance adaptive thermogenesis . 10 . 7554/eLife . 27669 . 011Figure 5 . Zfp423-deficiency enables cold-induced visceral WAT browning . ( A ) Control and Vis-KO mice were housed at room temperature ( 22°C ) until 6 weeks of age then transferred to thermoneutral housing ( 30°C ) for 2 weeks . After 2 weeks at thermoneutrality , animals were transferred to 6°C or left at thermoneutrality for seven full days . On the 8th day , tissues were harvested for analysis . mRNA levels ( relative to Rps18 ) of Ucp1 in iWAT , gWAT , and rWAT from control and Vis-KO mice cold exposed ( CE ) or housed at thermoneutrality ( TN ) for 7 days . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( B ) Fold change in mRNA levels of thermogenic genes in rWAT and gWAT from control and Vis-KO mice cold exposed ( CE ) or housed at thermoneutrality ( TN ) for 7 days . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( C–J ) Brightfield images of Ucp1 immunostaining ( brown ) within rWAT ( C–F ) and gWAT ( G–J ) sections obtained from control or Vis-KO mice following cold exposure ( CE ) or housing at thermoneutrality ( TN ) . ( K ) Body temperature at indicated time points following transfer of mice from thermoneutrality to cold ( 6°C ) . n = 4–5 mice . ( L ) Body temperature at indicated time points during a cold tolerance test after 4 weeks of acclimation to cold ( 6°C ) . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 011 We also asked whether the thermogenic visceral WAT present in Vis-KO animals confers protection against diet-induced obesity and/or impaired nutrient homeostasis . We administered 8 weeks-old male mice chow or high-fat diet ( HFD ) ( 60% calories from fat ) for 20 weeks . Over this period , we did not observe a significant difference in body weight between Vis-KO mice and controls ( Figure 6A ) . At the time of analysis ( 8 weeks of HFD feeding ) , we did not observe a significant difference in adiposity ( % body fat ) , lean mass , or fat distribution ( Figure 6B , C ) . Gene expression analysis of dissected tissues revealed that the loss of Zfp423 expression remained restricted to visceral depots ( Figure 6D ) . Levels of Ucp1 and other genes related to thermogenesis were significantly increased in visceral depots of obese Vis-KO mice ( Figure 6E , F ) ; however , mRNA levels of adipocyte-selective genes did not appear to be impacted by the loss of Zfp423 ( Figure 6G ) . The later suggests that Zfp423-deficiency did not impact de novo visceral adipocyte differentiation that occurs over this period of HFD feeding . The increase in thermogenic gene expression appeared functionally significant; rates of basal and maximal oxygen consumption were significantly elevated in the visceral depots of obese Vis-KO mice when compared to controls ( Figure 6H–J ) . We again observed a slight increase in the expression of thermogenic genes in the inguinal WAT of mutant animals; however , the metabolic activity of the inguinal WAT depot did not significantly differ from inguinal WAT of obese control mice ( Figure 6K ) . 10 . 7554/eLife . 27669 . 012Figure 6 . Obese mice lacking Zfp423 in visceral adipose tissue display enhanced visceral adipose tissue thermogenesis . ( A ) Weekly measurements of body weights of 8 weeks-old control and Vis-KO mice fed a standard chow diet ( Chow ) or high fat diet ( HFD ) for 20 weeks . n = 6–7 mice per group . ( B ) Total fat mass , lean mass , and water mass ( normalized to body weight ) of control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . n = 6–7 mice . ( C ) Fat depot mass ( normalized to body weight ) of control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . n = 6–7 mice . ( D ) Relative mRNA levels of Zfp423 in BAT , iWAT , gWAT , mWAT , and rWAT isolated from control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice per group . ( E ) mRNA levels ( normalized to Rps18 ) of Ucp1 within adipose depots of control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of chow or HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( F–G ) Relative mRNA levels of thermogenic genes ( F ) and adipocyte-selective genes ( G ) in iWAT , gWAT , mWAT , and rWAT isolated from control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( H–K ) Relative basal oxygen consumption rates ( OCRs ) within gWAT ( H ) , mWAT ( I ) , rWAT ( J ) , and iWAT ( K ) isolated from control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of chow or HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 independent replicates pooled from two mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 012 We also examined whether the observed increase in visceral WAT oxygen consumption impacts whole-body levels of energy expenditure . We assessed energy expenditure and food intake in control and knockout animals fed HFD for 8 weeks . At this point , body weights and body composition were comparable between the two groups of animals ( Figure 7A , B ) ; however , we observed a modest , but statistically significant , increase in O2 consumption , heat production , and CO2 production , in the Zfp423-deficient mice ( Figure 7C–F ) . We observed a slight increase in food intake over the dark cycle; however , these differences did not reach statistical significance ( Figure 7G ) . Moreover , overall locomotor activity measurements and respiratory exchange ratios were similar between control and knockout animals ( Figure 7H , I ) . These new data indicate that visceral WAT deletion of Zfp423 , and subsequent visceral browning , enhances energy expenditure , albeit not a strong enough degree to drive a robust difference in body weight . 10 . 7554/eLife . 27669 . 013Figure 7 . Browning of visceral adipose tissues is associated with increased energy expenditure . ( A ) Body weight of control ( white bar ) and Vis-KO ( red bar ) mice after 8 weeks of high fat diet feeding . n = 4–5 mice . ( B ) Total fat mass , lean mass , and water mass ( normalized to body weight ) of control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . n = 4–5 mice . ( C ) O2 consumption during two complete 12 hr light-dark cycles of control and Vis-KO mice following 8 weeks of high fat diet feeding . n = 4–5 mice . ( D–G ) Average O2 consumption ( D ) , heat production ( E ) , CO2 production ( F ) , and food intake ( G ) during the 24 hr , day , and night cycle in control ( white bars ) and Vis-KO ( red bars ) mice following 8 weeks of high fat diet . Bars represent averages from over the course of the 5 day measurement . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( H ) Average 24 hr X-beam , Y-beam , and Z-beam breaks of control ( white bars ) and Vis-KO ( red bars ) mice following 8 weeks of high fat diet . Bars represent averages from over the course of the 5 day measurement . n = 4–5 mice . ( I ) Average RER during the 24 hr , day , and night cycle in control ( white bars ) and Vis-KO ( red bars ) mice following 8 weeks of high fat diet . Bars represent averages from over the course of the 5 day measurement . n = 4–5 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 013 Further metabolic phenotyping of control and mutant animals revealed that the beige-like phenotype of visceral WAT in diet-induced obese Vis-KO mice was associated with striking improvements in nutrient homeostasis . Obese Vis-KO mice exhibited significantly better glucose tolerance test than obese control animals; excursions in blood glucose and serum insulin following glucose challenge were substantially lower in Vis-KO animals ( Figure 8A , B ) . Insulin tolerance tests suggest greater systemic insulin sensitivity in the Vis-KO mice ( Figure 8C ) . We further explored the impact of visceral WAT browning on systemic glucose metabolism by performing hypersulinemic-euglycemic clamp assays . The glucose infusion rate needed to maintain euglycemia ( ~138 mg/dl ) was increased in Vis-KO mice ( Figure 8D ) . This demonstrates an increase in whole-body insulin sensitivity , consistent with the aforementioned insulin tolerance tests . Importantly , endogenous glucose output was suppressed much more efficiently during the basal and clamped states in Vis-KO mice , likely reflecting improved insulin sensitivity at the level of the liver ( Figure 8E ) . These data are supported by liver gene expression analysis from a separate cohort of animals . The mRNA levels of key gluconeogenic genes in the Vis-KO livers are significantly lower than corresponding levels in livers from control animals after 8 weeks of HFD feeding ( Figure 8F ) . Furthermore , visceral adipose-selective browning phenotype is associated with lower serum triglyceride levels ( Figure 8G ) . 14C-2-deoxyglucose tracing revealed that the rate of whole-body glucose disposal did not significantly differ between control and Vis-KO mice ( Figure 8H ) ; however , glucose uptake was enhanced in the visceral rWAT depot , correlating with where the largest degree of visceral beiging is observed in this model ( Figure 8I ) . All together , these data demonstrate visceral deletion of Zfp423 leads to enhanced insulin sensitivity and reduced hepatic glucose output . Moreover , these data suggest that engineered thermogenic visceral adipose tissue , much like subcutaneous beige adipose tissue , can defend against the development of impaired glucose and lipid homeostasis in obesity . 10 . 7554/eLife . 27669 . 014Figure 8 . Mice lacking visceral adipose Zfp423 are protected against insulin resistance in obesity . ( A ) Intraperitoneal glucose tolerance tests of control and Vis-KO mice after 8 weeks of chow or HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( B ) Serum insulin levels of control and Vis-KO mice after 8 weeks of HFD feeding during the glucose tolerance test shown in A . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( C ) Insulin tolerance tests of control and Vis-KO mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( D–E ) Glucose infusion rate ( D ) and basal and clamped hepatic glucose production ( E ) during hyperinsulinemic-euglycemic clamp experiments performed on conscious , unrestrained control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 5–6 mice . ( F ) Relative mRNA levels of gluconeogenic genes in the livers of control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( G ) Serum triglyceride levels in control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–7 mice . ( H ) Glucose disappearance rate during hyperinsulinemic-euglycemic clamp experiments performed on conscious , unrestrained control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . n = 5–6 mice . ( I ) 2-deoxyglucose uptake quantification of iWAT , gWAT , rWAT , and mWAT during hyperinsulinemic-euglycemic clamp experiments performed on conscious , unrestrained control ( white bars ) and Vis-KO ( red bars ) mice after 8 weeks of HFD feeding . * denotes p<0 . 05 from unpaired Student’s t-test . n = 5–6 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 014 We previously demonstrated that adipocytes emerging in expanding visceral , but not subcutaneous , WAT depots of diet-induced obese mice arise through de novo adipocyte differentiation from mural progenitors expressing Pdgfrb ( Hepler et al . , 2017; Vishvanath et al . , 2016 ) . Whether these preadipocytes can be redirected to a thermogenic adipocyte fate , thereby driving beige-like adipocyte hyperplasia rather than white adipocyte hyperplasia , has been unclear . To address this , we generated a model that allows for doxycycline-inducible inactivation of Zfp423 in Pdgfrb-expressing mural cells ( inducible mural cell Zfp423 knockout , or ‘iMural-KO’ ) ( Figure 9A ) . The iMural-KO model was achieved by breeding PdgfrbrtTA transgenic mice to animals expressing Cre recombinase under the control of a tetracycline responsive element ( TRE-Cre ) and carrying floxed Zfp423 alleles ( Zfp423loxP/loxP ) . We also bred the Cre-dependent Rosa26R loxP-mtdTomato-loxP-mGFP ( mT/mG ) fluorescent reporter allele into the model; this allowed for the fate-mapping of targeted mural cells . Treatment of adult animals with doxycycline leads to Cre dependent inactivation of mural cell Zfp423 ( Figure 9B ) , and permanent fluorescent-tagging ( mGFP ) of Pdgfrβ+ cells . Importantly , mature adipocytes present at the time of doxycycline treatment are not targeted ( Figure 9B ) . Only those adipocytes that descend from mural cells following HFD feeding are Zfp423-deficient and will express mGFP . 10 . 7554/eLife . 27669 . 015Figure 9 . Adult mural cell deletion of Zfp423 does not impact visceral white adipogenesis associated with high fat diet feeding . ( A ) Doxycycline ( Dox ) inducible deletion of Zfp423 in adult mural cells ( inducible-Mural-Knockout or ‘iMural-KO’ ) is achieved by breeding PdgfrbrtTA transgenic mice to animals expressing Cre recombinase under the control of a tetracycline response element ( TRE-Cre ) , the Rosa26R loxP-mtdTomato-loxP-mGFP ( mT/mG ) fluorescent reporter allele , and the floxed Zfp423 alleles ( Zfp423loxP/loxP ) . Animals carrying only PdgfrbrtTA , TRE-Cre , and Rosa26RmT/mG alleles were used as controls . ( B ) Validation of the iMural-KO model . Relative mRNA levels of Zfp423 in purified adipocytes and purified Pdgfrβ+ cells from WAT obtained from control ( white bars ) and iMural-KO ( red bars ) mice at 8 weeks of age . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4–5 mice . ( C ) Mice were maintained at room temperature and fed standard chow diet until 6 weeks of age . Animals were then switched to Dox-containing chow diet for 9 days in order to induce Zfp423 deletion and mGFP expression in mural cells . Animals were then maintained on a Dox-containing high-fat diet ( HFD ) for 8 weeks then sacrificed for analysis . ( D ) Weekly body weight measurements of control and iMural-KO mice during HFD feeding . n = 6–8 mice . ( E–J ) Confocal images of immunostained gWAT obtained from control ( E–G ) and iMural-KO ( H–J ) mice fed high fat diet for 8 weeks . ( K ) Quantification of mGFP-expressing adipocytes ( mGFP+; perlipin+ ) observed in randomly chosen 10X magnification fields of gWAT sections obtained from control ( black circles ) and iMural-KO ( red squares ) mice following 8 weeks of HFD feeding . n = 6 mice . ( L ) Glucose tolerance tests of control and iMural-KO mice after 8 weeks of HFD feeding . n = 6–8 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 015 Following doxycycline treatment , we fed control and iMural-KO mice a HFD for 8 weeks ( Figure 9C ) . After 8 weeks of HFD feeding , the body weights of mutant animals were indistinguishable from controls ( Figure 9D ) . De novo white adipocyte differentiation occurred in gWAT of both control and mutant animals; newly derived adipocytes were indicated by the expression of mGFP ( Figure 9E–J ) . However , we did not observe a statistically significant difference in the number of mGFP+ adipocytes between control and iMural-KO mice ( Figure 9K ) . Overall , ~5–15% of gonadal adipocytes present in the gWAT of these obese animals descended from mural cells . Despite the fact that Zfp423 expression identifies mural adipose progenitors , these data suggest that adult mural progenitors do not require Zfp423 in vivo for their ability to undergo adipocyte differentiation in response to HFD feeding . Zfp423-deficient adipocytes originating from mural progenitors in the obese iMural-KO model were readily identified by mGFP expression; however , these cells were not multilocular ( Figure 9H , I , J ) . Moreover , the glucose tolerance of obese iMural-KO mice was similar to that of control animals ( Figure 9L ) . Our previous study revealed that the thermogenic activity of Zfp423-deficient adipocytes was dependent on active β-adrenergic signaling ( Shao et al . , 2016 ) ; it is well appreciated that rodent obesity is associated with augmented β-adrenergic receptor signaling ( Collins et al . , 1999; Collins and Surwit , 2001 ) . Therefore , we reasoned that the Zfp423-deficient adipocytes present in these in obese mice would require a stimulus to fully activate their thermogenic function in this setting . To test this , we utilized osmotic pumps to deliver the β3-adrenergic receptor agonist , CL316 , 243 , daily at a dose of 1 mg/kg/day for four continuous weeks . The agonist was given to obese control and iMural-KO mice after 8 weeks of HFD feeding ( Figure 10A ) . During the 4 weeks of the β3-agonist treatment , both control and iMural-KO animals exhibited a similarly mild decrease in body weight ( Figure 10B ) . However , after four weeks of β3-agonist treatment , the gonadal adipose depot mass of iMural-KO mice was significantly smaller than the gWAT mass of treated control mice ( Figure 10C ) . Immunohistochemistry for mGFP expression revealed that β3-agonist treatment did not induce the formation of any additional gonadal or inguinal adipocytes from the Pdgfrb lineage ( Figure 10D–H , Figure 10—figure supplement 1A , B ) . We again did not observe a difference in the numbers of mGFP-labelled adipocytes between control and knockout animals ( Figure 10—figure supplement 1A , B ) . Mural-cell derived adipocytes in the β3-agonist treated control animals remained unilocular; however , most mGFP+ adipocytes present in β3-agonist treated knockout mice were now multilocular ( Figure 10D–H ) . On average , ~6% of gonadal adipocytes appeared multilocular in the iMural-KO animals; very few , if any , gonadal multilocular adipocytes were present in β3-agonist treated control mice ( Figure 10—figure supplement 1C ) . Levels of Ucp1 mRNA were strongly elevated in visceral depots of the β3-agonist treated iMural-KO mice ( Figure 10—figure supplement 1E ) . On the other hand , the percentage of inguinal adipocytes appearing multilocular appeared low ( ~3% ) and comparable between control and knockout animals ( Figure 10—figure supplement 1D ) . However , despite the low and comparable levels of mural cell adipogenesis , the iWAT from β3-agonist treated knockout animals had relatively higher levels of Ucp1 mRNA in comparison to controls ( Figure 10—figure supplement 1E ) . The recruitment of this relatively low number of beige-like adipocytes appeared sufficient to drive an increase in basal respiration of adipose tissues ( Figure 10—figure supplement 1F ) . This active thermogenic phenotype of the visceral WAT correlated with markedly improved glucose tolerance and insulin sensitivity observed in β3-agonist iMural-KO mice ( Figure 10I–K ) . All together , these data indicate that visceral mural preadipocytes in adult mice can be directed to a dormant beige-like phenotype in the expanding visceral WAT depots of diet-induced obese animals . Upon activation by β3 adrenergic receptor agonism , these beige-like adipocytes appear to drive an improvement in insulin sensitivity in obese animals . 10 . 7554/eLife . 27669 . 016Figure 10 . Inactivation of Zfp423 in adult mural cells leads to beige , rather than white , adipocyte hyperplasia in diet-induced obesity . ( A ) Chow-fed 6 weeks-old animals were first administered Dox-containing chow for 9 days in order to inactive Zfp423 and induce permanent mGFP expression in Pdgfrb+ cells . Mice were then switched to a high-fat diet ( HFD ) containing Dox . After 8 weeks of Dox-HFD , the mice were treated daily with CL316 , 243 ( 1 mg/kg/24 hr ) or vehicle ( PBS ) for 4 weeks . ( B ) Weekly body weight measurements of obese control and iMural-KO mice following vehicle or CL316 , 243 administration . n = 6–8 mice . ( C ) Fat depot weights ( normalized to body weight ) of control and iMural-KO mice after vehicle or CL316 , 243 administration . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–8 mice . ( D–G ) Representative confocal images of Perilipin ( red ) , mGFP ( green ) and DAPI ( blue ) immunostaining of gWAT sections obtained from control and iMural-KO mice administered vehicle ( D , E ) , or CL316 , 243 ( F , G ) . ( H ) Digital enhancement of region outlined in ( G ) . ( I ) Glucose tolerance tests of control and iMural-KO mice following vehicle or CL316 , 243 administration . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–8 mice . ( J ) Serum insulin levels measured during intraperitoneal glucose tolerance tests of control and iMural-KO mice following vehicle or CL316 , 243 administration . n = 6–8 mice . ( K ) Insulin tolerance tests of control and iMural-KO mice following vehicle or CL316 , 243 administration . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–8 mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 01610 . 7554/eLife . 27669 . 017Figure 10—figure supplement 1 . Quantification of beige adipocyte hyperplasia in iMural-KO mice . ( A–B ) Quantification of mGFP-expressing adipocytes ( mGFP+; perlipin+ ) observed in randomly chosen 10X magnification fields of gWAT ( A ) or iWAT ( B ) sections of obese control and iMural-KO mice following vehicle or CL316 , 243 administration . n = 7 mice . ( C ) Percentage of all perilipin+ adipocytes counted in ( A ) that are multilocular . * denotes p<0 . 05 from unpaired Student’s t-test . n = 7 mice . ( D ) Percentage of all perilipin+ adipocytes counted in ( B ) that are multilocular . n = 7 mice . ( E ) Relative mRNA levels of Ucp1 in indicated WAT depots obtained from control and iMural-KO mice following vehicle or CL316 , 243 administration . * denotes p<0 . 05 from unpaired Student’s t-test . n = 6–8 mice . ( F ) Relative basal oxygen consumption rates ( OCRs ) of isolated WAT depots from control and iMural-KO mice . * denotes p<0 . 05 from unpaired Student’s t-test . n = 4 independent replicates pooled from two mice . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 017
It is now certain that adult humans have appreciable amounts of thermogenic adipose tissue consisting of brown and beige adipocytes ( Cypess et al . , 2013; Shinoda et al . , 2015; van Marken Lichtenbelt et al . , 2009; Virtanen et al . , 2009 ) . Upon activation , thermogenic adipose tissue in lean adults can impact glucose and lipid homeostasis ( Chondronikola et al . , 2014 , 2016; Cypess et al . , 2015 ) ; however , it still remains unclear as to whether sufficient amounts of thermogenic adipose tissue are present in obese individuals to exert beneficial therapeutic effects , even when fully activated . As such , there is tremendous interest in identifying strategies to increase the mass of functional thermogenic adipose tissue in obese patients with metabolic syndrome . To this end , a number of studies have now identified pathways/factors that can drive the natural formation of subcutaneous beige adipocytes or classic brown adipocytes ( Harms and Seale , 2013 ) . The negative impacts of visceral expansion during obesity on metabolic health make visceral fat a prime target for therapeutic intervention; however , effective strategies to improve the health of visceral WAT health have yet to emerge . In particular , whether browning of visceral depots , much like the browning of subcutaneous adipose tissue , would exert beneficial metabolic effects has been unclear . We recently reported that Zfp423 is a suppressor of the thermogenic gene program in adipocytes . Using this discovery as a tool , we reveal here that the thermogenic potential of visceral adipose depots in mice can be unlocked through removal of Zfp423 . Importantly , these data provide proof of concept that white adipose precursors in adult animals can be redirected to a beige-like adipocyte fate and improve insulin sensitivity in obesity . Overall , we observed beneficial effects of visceral WAT browning on nutrient homeostasis and cold tolerance . Nevertheless , we cannot exclude the possibility that inducing the thermogenic capacity of these depots may have detrimental effects under other physiological conditions not examined . Elevated expression of Zfp423 in visceral white adipocytes provides one explanation as to how visceral WAT depots resist adopting a thermogenic phenotype; however , it remains unclear as to why visceral WAT depots would adopt these anti-thermogenic mechanisms . Inactivation of Zfp423 in visceral WAT gives rise to thermogenic adipocytes that share properties of subcutaneous beige adipocytes and classic brown adipocytes . In particular , Zfp423 deletion in visceral WAT unlocks a functionally significant portion of the thermogenic gene program characteristic of activated ( β3 adrenergic receptor activated ) beige adipose tissue . Nevertheless , global gene expression profiling suggests that Zfp423-deficient visceral WAT largely retains a visceral WAT molecular signature , rather than adopting the global molecular program characteristic of subcutaneous WAT . Additional gene expression studies of isolated visceral and inguinal Ucp1+ adipocytes will be needed to fully characterize the similarities and differences between the anatomically distinct thermogenic fat cells . During preparation of this manuscript , Kirichok and colleagues reported the existence of two distinct types of thermogenic beige adipocytes present in visceral depots of mice stimulated with the β3-adrenergic receptor agonist for 10 days ( Bertholet et al . , 2017 ) . In particular , Bertholet et al . revealed that most thermogenic adipocytes in visceral WAT are devoid of Ucp1 protein and instead employ futile creatine cycling for thermogenesis . Zfp423-deficient visceral adipocytes express Ucp1 protein; however , the precise contribution of Ucp1-mediated uncoupling vs . creatine cycling in these cells remains unclear . Our data here highlight the ability of thermogenic adipocytes , even within the visceral compartment , to drive vast improvements in nutrient homeostasis in obese mice , without impacting body weight . Moreover , the mural cell knockout model of Zfp423 even suggests that a relatively low frequency of activated multilocular Zfp423-deficient visceral adipocytes ( 5–10% of adipocytes in β3-agonist treated obese iMural-KO mice ) can lead to improved insulin sensitivity . This is surprising given the opinion that such small numbers of beige cells are not likely to confer significant benefit ( Kalinovich et al . , 2017 ) . We , of course , cannot rule out the possibility that Zfp423 deficiency may impact the visceral adipose phenotype in a multitude of ways that influence energy metabolism . It is notable that in the Vis-KO model , the data largely point to the liver as a major site of improved insulin sensitivity . Previously reported work on Prdm16 and subcutaneous WAT beiging suggest a strong connection between subcutaneous beige cells and liver health ( Cohen et al . , 2014 ) . For the field at large , it remains an open question as to how brown/beige adipocytes exert beneficial effects on systemic metabolism , independent of their impact on body weight . This unique model may serve as a tool to explore the mechanisms of how visceral WAT browning leads to improved glucose metabolism . We consistently observed a modest , yet statistically significant , induction of the thermogenic gene program in the inguinal WAT depots of both genetic models . There is a limit to our ability to interpret these results from the iMural-KO mice . Inguinal adipocyte differentiation is not impacted in this model; however , Zfp423 is still inactivated in mural cells of this depot . It is not possible to exclude additional roles for Zfp423 in mural cells of iWAT or other tissues . Nevertheless , the dependency of the improved insulin sensitivity in obese iMural-KO mice on β3-adrenergic receptor activation does suggest that the increased numbers of thermogenic visceral adipocytes , at least in part , contribute to the phenotypes observed here . In the Vis-KO model , inactivation of Zfp423 is limited to visceral WAT depots . The subcutaneous WAT phenotype in this model is thus not cell autonomous and appears secondary to the visceral phenotype . It is possible that Zfp423-deficient visceral adipocytes produce circulating adipokines that influence systemic metabolism and thermogenesis . In fact , more and more ‘Batokines’ have recently emerged ( Lynes et al . , 2017; Svensson et al . , 2016; Thomou et al . , 2017; Villarroya et al . , 2017 ) . It is also possible that secondary browning of subcutaneous WAT may be triggered through central effects mediated by a visceral WAT to brain neural relay . A multitude of studies now support the idea that visceral adipocytes are developmentally , molecularly , and functionally , distinct from subcutaneous adipocytes . Our prior work revealed a critical role for Zfp423 in 3T3-L1 adipogenesis and in the regulation of Pparg and the fetal formation of subcutaneous white adipocytes in vivo ( Gupta et al . , 2010; Shao et al . , 2017 ) . Our work here reveals that Zfp423 is dispensable for the differentiation of visceral adipocytes within those visceral depots examined . Thus , Zfp423 expression defines mural white preadipocytes within visceral depots of adult mice; however , other factors can compensate for its absence in the initial stages of adipocyte differentiation , but not in suppressing the thermogenic gene program . The lack of impact on adipocyte differentiation was unexpected; however , this result is perhaps not surprising in light of recent studies of another pro-adipogenic transcription factor , Cebpa . In vitro , Cebpa is amongst the most critical adipogenesis factors; however , in vivo , it appears essential for postnatal , but not fetal , terminal differentiation of fat cells ( Wang et al . , 2015 ) . Thus , despite the expression of Cebpa throughout the adipose lineage , there appears to be temporal requirements for this particular factor . These data , along with other studies , highlight the emerging concept that distinct transcriptional regulatory mechanisms govern fetal vs . adult adipogenesis ( Jeffery et al . , 2015; Wang et al . , 2015 ) . Our data here further highlight the complexity in the regulation of adipogenesis in vivo by demonstrating the depot-specific requirements of pro-adipogenic transcription factors . Zfp423 expression in the adipose lineage appears highly regulated . Levels of Zfp423 in white adipocytes are reduced following cold exposure and β-adrenergic receptor agonism , and increased in brown adipose depots undergoing a ‘whitening’ transformation with age or in obesity ( Shao et al . , 2016 ) . The dynamic expression of Zfp423 , along with the loss of function data from our genetic mouse models , reveals Zfp423 as a critical physiological suppressor of the adipocyte thermogenic gene program . Future studies into the regulation of Zfp423 expression in adipocytes and/or mural adipose precursors may lead to novel strategies to unlock the thermogenic potential of visceral adipose tissue and combat the chronic metabolic defects associated with visceral obesity .
All animal experiments were performed according to procedures approved by the UTSW Animal Care and Use Committee . PdgfrbrtTA transgenic mice ( C57BL/6-Tg ( Pdgfrb-rtTA ) 58Gpta/J; JAX 028570; RRID:IMSR_JAX:028570 ) were previously described ( Vishvanath et al . , 2016 ) . TRE-Cre ( B6 . Cg-Tg ( tetO-cre ) 1Jaw/J; JAX 006234; RRID:IMSR_JAX:006234 ) , Rosa26RmT/mG ( B6 . 129 ( Cg ) -Gt ( ROSA ) 26Sortm4 ( ACTB-tdTomato , -EGFP ) Luo/J; JAX 007676; RRID:IMSR_JAX:007676 ) , and WT1-Cre ( Wt1tm1 ( EGFP/cre ) Wtp/J; JAX 010911; RRID:IMSR_JAX:010911 ) mice were obtained from Jackson Laboratories . Zfp423loxP/loxP mice were a gift from Dr . S . Warming ( Genentech ) ( Warming et al . , 2006 ) . Mice were maintained on a 12-hr light/dark cycle in a temperature-controlled environment ( room temperature , 22°C; thermoneutrality , 30°C; cold exposure , 6°C ) . Mice were given free access to food and water , and maintained on a standard chow diet , Dox-containing chow ( 600 mg/kg doxycycline , Bio-Serv , S4107 ) , or a dox-containing HFD ( 600 mg/kg dox , 60% kcal% fat , BioServ , S5867 ) . For acute β−3 adrenergic agonist administration , mice were transferred to 30°C chambers for two weeks then injected intraperitoneally with vehicle or CL316243 ( 1 mg/kg/day ) for 3 days . For chronic β−3 adrenergic agonist administration , mice were anesthetized by 2% isoflurane , and Alzet osmotic minipumps filled with vehicle ( PBS ) or CL 316243 ( 1mg/kg/24 hr ) were implanted subcutaneously in the dorsal region of the animals . Adipose tissues were harvested from perfused ( 4% paraformaldehyde ) adult mice . Paraffin processing and embedding was performed by the Molecular Pathology Core Facility at UTSW . Indirect immunofluorescence was performed as previously described ( Vishvanath et al . , 2016 ) . Antibodies used for immunofluorescence include: anti-GFP 1:700 ( Abcam ab13970 , RRID:AB_300798 ) , anti-perilipin 1:1500 ( Fitzgerald 20R-PP004 , RRID:AB_1288416 ) , anti-chicken Alexa 488 1:200 ( Invitrogen , RRID:AB_142924 ) , anti-guinea pig Alexa 647 1:200 ( Invitrogen , RRID:AB_141882 ) , and anti-guinea pig Alexa 488 1:200 ( Invitrogen , RRID:AB_142018 ) . For Ucp1 immunohistochemistry , paraffin-embedded sections were incubated with anti-Ucp1 ( Abcam ab10983; 1:500 , RRID:AB_2241462 ) , followed by secondary and tertiary signal amplification and detection using biotinylated anti-rabbit secondary ( Vector BA-1100 , RRID:AB_2336201 ) , HRP-conjugated streptavidin ( Dako ) , and DAB substrate ( Thermo ) . For quantification of adipocyte hyperplasia , paraffin sections were stained with perilipin and GFP by indirect immunofluorescence . The number of GFP+ perilipin+ and GFP- perilipin+ adipocytes were counted on 8–10 randomly selected 10X images of stained WAT depots . A total of 3 , 000–5 , 000 perilipin+ adipocytes were counted from each mouse . Each data point represents the percentage of GFP+ perilipin+ adipocytes from one mouse . SVF was isolated as previously described ( Shao et al . , 2016 ) . Briefly , minced adipose tissue was placed in digestion buffer ( 100 mM HEPES pH 7 . 4 , 120 mM NaCl , 50 mM KCl , 5 mM glucose , 1 m CaCl2 , 1 . 5% BSA , and 1 mg/mL collagenase D ( Roche 11088882001 ) and incubated in a 37°C shaking water bath for 2 hr . The mixture was then passed sequentially through a 100 µm cell strainer then a 40 µm cell strainer . Cells were blocked in 2% FBS/PBS containing anti-mouse CD16/CD32 Fc Block ( clone 2 . 4G2; 1:200; RRID:AB_394657 ) , then incubated with primary antibodies ( anti-CD31 clone 390 1:200 , RRID:AB_312903; anti-CD45 clone 30-F11 1:200 , RRID:AB_312971; anti-CD140b clone APB5 3:200 , RRID:AB_2268091 ) . Cells were sorted using a FACSAriaTM flow cytometer ( UTSW Flow Cytometry Core Facility ) . Adipose tissue SVF was isolated as described above . Cells were plated onto collagen-coated dishes and incubated at 10% CO2 . Gonadal SVF was maintained in growth media ( 60% pH7–7 . 4 low glucose DMEM , 40% pH 7 . 25 MCDB201 ( Sigma M6770 ) ) , supplemented with 2% FBS ( Fisher Scientific 03-600-511 Lot FB-002 ) 1% ITS premix ( Insulin-Transferrin-Selenium ) ( BD Bioscience 354352 ) , 0 . 1 mM L-ascorbic acid-2-2phosphate ( Sigma A8960-5G ) , 10 ng/mL FGF basic ( R&D systems 3139-FB-025/CF ) , Pen/Strep , and gentamicin . Inguinal SVF was maintained in DMEM/F12 ( Invitrogen ) supplemented with Glutamax , 10% FBS , Pen/Strep , and gentamicin . Upon reaching confluence , cultures were incubated with the adipogenesis induction cocktail ( growth media supplemented with 5 mg/ml insulin , 1 μM dexamethasone , 0 . 5 mM isobutylmethyxanthine , and 1 μM rosiglitazone ) for 48 hr . After 48 hr , the cells were maintained in growth media supplemented with 5 mg/ml insulin and 1 μM rosiglitazone until harvest . Differentiated cells were fixed in 4% PFA for 15 min at room temperature then washed twice with water . Cells were incubated in Oil Red O working solution ( 2 g Oil red O in 60% isopropanol ) for 10 min to stain accumulated lipids . Cells were then washed three times with water before bright field images were acquired . Relative mRNA levels were determined by quantitative PCR using SYBR Green chemistry . Values were normalized to Rps18 levels using the ΔΔ-Ct method . Unpaired Student's t-test was used to evaluate statistical significance . All primer sequences are listed in Table 1 . mRNA library preparation and RNA-sequencing was performed by the McDermott Center Sequencing Core at UT Southwestern . Total RNA used for library preparation was extracted from gonadal , inguinal , and brown adipose tissues of Control and Vis-KO mice after 3 days of treatment with CL-316 , 243 at thermoneutrality , as described above . Sequencing was performed on an Illumina HiSeq 2500 and reads were mapped to the mouse genome ( mm10 ) . Analysis was performed by the UT Southwestern McDermott Bioinformatics Core using Cufflinks/Cuffdiff software . Genes with an FDR < 0 . 05 were considered significantly differentially expressed between groups compared . Heatmaps were generated using the Pheatmap package in RStudio ( v3 . 3 ) . The cluster dendogram was generated using Hierarchical Cluster Analysis in RStudio ( v3 . 3 ) . Gene Ontology analysis was performed using the DAVID Functional Annotation tool ( https://david . ncifcrf . gov/ ) on differentially expressed genes between groups compared . Functional annotation for gene ontology using the GOTERM_CC_FAT category was selected and biological processes were assessed for statistical significance . All raw sequencing data has been deposited to Gene Expression Omnibus ( https://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE98132 ) . 10 . 7554/eLife . 27669 . 018Table 1 . qPCR primer sequences utilized for gene expression analysisDOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 018GeneForward 5'−3'Reverse 5'−3'AdipoqAGATGGCACTCCTGGAGAGAATTCTCCAGGCTCTCCTTTCCTAdipsinCTACATGGCTTCCGTGCAAGTAGTCGTCATCCGTCACTCCATCideaTCCTATGCTGCACAGATGACGTGCTCTTCTGTATCGCCCAGTCited1AACCTTGGAGTGAAGGATCGCGTAGGAGAGCCTATTGGAGATGTDio2CATTGATGAGGCTCACCCTTCGGTTCCGGTGCTTCTTAACCTEar2CCACAAAGCAGACAGGGAAACGCATGAGGCAAGCATTAGGACElovl3GTGTGCTTTGCCATCTACACGCTCCCAGTTCAACAACCTTGCFabp4GATGAAATCACCGCAGACGACATTCCACCACCAGCTTGTCACFoxo1GGCACTCCAAAACAGGACTTGAAGAAATGGCAGAGGGAGGAGG6pcGGGCTGTTTGAGGAAAGTGTGTATCCGACAGGAGGCTGGTAAGsta3AGATCGACGGGATGAAACTGGCAGATCCGCCACTCCTTCTHoxa9CCCCGACTTCAGTCCTTGCGATGCACGTAGGGGTGGTGLhx8GAGCTCGGACCAGCTTCATTGTTGTCCTGAGCGAACTGPck1TGTCTGTCCCATTGTCCACAGAAGGTAAGGAAGGGCGGTGTAPgc1αGCACCAGAAAACAGCTCCAAGCGTCAAACACAGCTTGACAGCPparg2GCATGGTGCCTTCGCTGATGGCATCTCTGTGTCAACCATGPrdm16ACACGCCAGTTCTCCAACCTGTTGCTTGTTGAGGGAGGAGGTAResistinAAGAACCTTTCATTTCCCCTCCTGTCCAGCAATTTAAGCCAATGTTRps18CATGCAAACCCACGACAGTACCTCACGCAGCTTGTTGTCTATcf21CCCTGAAAGTGGACTCCAACAGCTGAGCGGGCTTTTCTTAGTTle3GAGACTGAACACAATCCTAGCCGGAGTCCACGTACCCCGATTmem26AGGGGCTTCCTTAGGGTTTTCCCGTCTTGGATGAAGAAGCTGUcp1TCTCAGCCGGCTTAATGACTGGGCTTGCATTCTGACCTTCACWt1ATAGGCCAGGGCATGTGTATGCTGGTGCCTTGCTCTCTGATTZfp423CAGGCCCACAAGAAGAACAAGGTATCCTCGCAGTAGTCGCACAZic1CTGTTGTGGGAGACACGATGCCTCTTCTCAGGGCTCACAG Metabolic cage studies were conducted using TSE Phenomaster cages ( TSE Systems , Chesterfield , Missouri ) at the USTW Metabolic Phenotyping Core . Mice were acclimated in the metabolic chambers for 5 days before the start of the experiments . Food intake , movement , and CO2 and O2 levels were measured every 60 min for each mouse over a period of 5 days . For glucose tolerance tests , mice were fasted overnight and then administered glucose by intraperitoneal injection ( 1 g/kg body weight , Sigma ) . For insulin tolerance tests , mice were fasted for 4 hr and then administered insulin by intraperitoneal injection ( 0 . 75 U/kg body weight human insulin , Eli Lilly ) . At the indicated time-points , tail blood was collected . Hyperinsulinemic euglycemic clamps were performed on conscious , unrestrained mice as previously described ( Holland et al . , 2011 ) . Blood glucose was measured using Bayer Contour glucometers . Serum triglycerides were measured using Infinity Triglycerides Reagent ( Thermo Fisher Scientific ) . Serum insulin was measured using Ultra Sensitive Mouse Insulin ELISA ( Crystal Chem ) . Mice 8 weeks of age were transferred to a 30°C chamber ( thermoneutrality ) for two weeks . One week before cold exposure , IPTT-300 temperature transponders ( Bio Medic Data Systems ) were implanted subcutaneously in the dorsal region of the mice . Body temperature was assessed using a DAS-7006/7 s reader ( Bio Medic Data Systems ) . The mice were then transferred to the cold chamber ( 6°C ) or maintained at thermoneutrality . For the acute cold tolerance test , food was removed from the cages upon transfer to the cold chamber and body temperature was measured at the indicated time points . For the acclimated cold exposure test , food was removed from the cages after 4 weeks of acclimation to cold and body temperature was measured at the indicated time points . Adipose tissue fragments or cultured adipocytes were assayed for oxygen consumption rate ( OCR ) using an XF24 Extracellular Flux Analyzer ( Seahorse Bioscience , MA ) . Assays of mitochondrial function and respiration rates in whole adipose tissues or cells were performed as previously described ( Shao et al . , 2016 ) . In brief , adipose tissue was cut into 5–10 mg pieces and locked into an XF24 islet-capture Microplate ( Seahorse Bioscience ) . Adipose tissues or cultured adipocytes were equilibrated for 1 hr at 37°C in a CO2-free incubator in XF Assay Medium ( Modified DMEM , 0 mM Glucose; Seahorse Bioscience ) ( pH 7 . 4 ) , supplemented with 1 mM sodium pyruvate , 1 mM L- Glutamine and 7 mM glucose . Tissues or cells were subjected to a 10 min equilibration period and three assay cycles to measure the basal rate , comprising a 3 min mix , a 2 min wait and a 3 min measure period each . For cells , compounds were then added by automatic pneumatic injection followed by assay cycles after each , comprising of 3 min mix , 2 min wait and a 3 min measure period . OCR measurements were obtained following sequential additions of oligomyin ( 3 µM final concentration ) , FCCP ( 9 µM ) and antimycinA/rotenone ( 3 µM/30 nM ) . OCR measurements were recorded at set interval time points . All compounds and materials above were obtained from Sigma-Aldrich . All data were expressed as the mean ± SEM . We used GraphPad Prism 7 . 0 ( GraphPad Software , Inc . , La Jolla , CA , USA ) to perform the statistical analyses . Each experiment was performed at least twice and representative data are shown . For comparisons between two independent groups , a Student’s t-test was used and p<0 . 05 was considered statistically significant . The sample size estimation was determined as described below . All the detailed sample sizes , statistical test methods , and p-values are listed in Table 2 . Longitudinal metabolic cohorts were designed to detect a 25% improvement of glucose tolerance or a similar magnitude of insulin resistance with an assumed 15% standard deviation of the group means at a power of 80% and an alpha of 0 . 05 . This predicted approximately six animals per test group . All animals in the cohort were subsequently used for downstream assays ( gene expression , IHC , western blot ) to eliminate selection bias . We estimated the approximate effect size based on independent preliminary studies . Studies designed to characterize an in vitro difference in metabolic flux were estimated to have a slightly larger effect size of 30% with assumed 15% standard deviation of group means . To detect this difference at a power of 80% and an alpha of 0 . 05 , we predicted we would need four independent replicates per group . We estimated this effect size based on independent preliminary studies . 10 . 7554/eLife . 27669 . 019Table 2 . Statistical information . DOI: http://dx . doi . org/10 . 7554/eLife . 27669 . 019FigureN ( sample size ) Statistical test methodDescriptionp-value1BControl n = 6Unpaired Student’s t testgWAT4 . 56E-08Vis-KO n = 6mWAT0 . 0019rWAT0 . 00021DControl n = 6Unpaired Student’s t testiWAT0 . 0043Vis-KO n = 6gWAT0 . 00511EControl n = 6Unpaired Student’s t testiWATAdipoq0 . 0111Vis-KO n = 6gWATAdipsin0 . 0041FControl n = 6Unpaired Student’s t testiWATCidea0 . 012678087Vis-KO n = 6Dio20 . 015258263Elovl30 . 035976522Prdm160 . 043549713gWATCidea6 . 91848E-06Dio20 . 017046004Prdm169 . 5932E-05mWATCidea0 . 008754783Dio20 . 013198702Prdm160 . 018964447rWATCidea4 . 6109E-07Dio20 . 000934576Prdm160 . 0002606611GControl n = 6Unpaired Student’s t testiWAT0 . 019621248Vis-KO n = 6gWAT0 . 002534473mWAT0 . 000189507rWAT7 . 14636E-061HControl n = 4Unpaired Student’s t testgWATBasal0 . 02847Vis-KO n = 41IControl n = 4Unpaired Student’s t testmWATBasal0 . 00457Vis-KO n = 41JControl n = 4Unpaired Student’s t testrWATBasal0 . 04901Vis-KO n = 41KControl n = 4Unpaired Student’s t testiWATBasal0 . 03254Vis-KO n = 42BControl n = 4Unpaired Student’s t testInguinalAdipoq0 . 00251Vis-KO n = 42DControl n = 4Unpaired Student’s t testGonadalZfp4232 . 1E-06Vis-KO n = 42EControl n = 4Unpaired Student’s t testGonadalCidea ( red bar ) 3 . 1573E-06Vis-KO n = 4Cidea ( blue bar ) 0 . 028888184Dio20 . 039314815Ppargc1a0 . 00022784Ucp1 ( red bar ) 0 . 004431674Ucp1 ( blue bar ) 0 . 0003561762FControl n = 4Unpaired Student’s t testGonadalBasal6 . 73733E-05Vis-KO n = 4Oligo . 0 . 000488801FCCP0 . 029096962GControl n = 4Unpaired Student’s t testGonadal0 . 03356Vis-KO n = 42HControl n = 4Unpaired Student’s t testGonadalGsta30 . 019517318Vis-KO n = 4Tcf210 . 014978261Wt10 . 000632618Ear20 . 009978141Hoxa90 . 000427928Tmem260 . 004299176Zic10 . 0347839573AControl n = 5Unpaired Student’s t testgWAT0 . 000759641Vis-KO n = 4rWAT0 . 0017390733BControl n = 5Cidea0 . 000155141Vis-KO n = 4Dio20 . 009600651Prdm16 ( red bar ) 0 . 002457629Prdm16 ( blue bar ) 7 . 9394E-053GControl n = 5Unpaired Student’s t testgWATGsta30 . 044793954Vis-KO n = 4Resistin0 . 002421899Tmem260 . 0116519785AControl n = 5Unpaired Student’s t testgWATUcp10 . 011438323Vis-KO n = 4rWATUcp10 . 0011886245BControl n = 5Unpaired Student’s t testrWATCidea0 . 000780591Vis-KO n = 4Dio20 . 002682038Elovl30 . 01642267Prdm16 ( red bar ) 0 . 042213109Prdm16 ( blue bar ) 0 . 000810897gWATDio20 . 0221Prdm160 . 01275LControl n = 5Unpaired Student’s t testt = 4 hr0 . 014972958Vis-KO n = 4t = 5 hr0 . 0351388266DControl n = 7Unpaired Student’s t testgWATZfp4230 . 002716658Vis-KO n = 6mWATZfp4230 . 042346922rWATZfp4230 . 0185686086EControl n = 7Unpaired Student’s t testiWATUcp10 . 002716658Vis-KO n = 6gWATUcp10 . 042346922rWATUcp10 . 0185686086FControl n = 7Unpaired Student’s t testiWATCidea0 . 044864901Vis-KO n = 6gWATCidea0 . 011872021Prdm160 . 0463436636HControl n = 4Unpaired Student’s t testgWAT0 . 00648Vis-KO n = 46IControl n = 4Unpaired Student’s t testmWAT0 . 03908Vis-KO n = 46JControl n = 4Unpaired Student’s t testrWAT0 . 03982Vis-KO n = 47DControl n = 4Unpaired Student’s t testO2 Consumption24 hr0 . 020700477Vis-KO n = 5Dark0 . 0060343337EControl n = 4Unpaired Student’s t testHeat Production24 hr0 . 022957861Vis-KO n = 5Dark0 . 0073782417FControl n = 4Unpaired Student’s t testCO2 Production24 hr0 . 041576891Vis-KO n = 5Dark0 . 0179642818AControl n = 7Unpaired Student’s t testt = 0 min0 . 043779825Vis-KO n = 6t = 15 min0 . 004769743t = 30 min0 . 012903721t = 60 min0 . 0149156848BControl n = 7Unpaired Student’s t testt = 0 min0 . 009873322Vis-KO n = 6t = 30 min0 . 015838998CControl n = 7Unpaired Student’s t testt = 30 min0 . 0426Vis-KO n = 68DControl n = 6Unpaired Student’s t testGlucose Infusion Rate0 . 04689Vis-KO n = 58EControl n = 6Unpaired Student’s t testHepatic Glucose Prod . Basal0 . 039627397Vis-KO n = 5Clamped0 . 03469538FControl n = 7Unpaired Student’s t testLiverFoxo10 . 046418578Vis-KO n = 6G6Pc0 . 024438785Pck10 . 0476980798GControl n = 7Unpaired Student’s t testTAG0 . 02449Vis-KO n = 68IControl n = 4Unpaired Student’s t test2-DG UptakerWAT0 . 04023Vis-KO n = 59BControl n = 4Unpaired Student’s t testPdgfrβ cells1 . 1E-05iMural-KO n = 510CControl n = 8Unpaired Student’s t testgWAT0 . 00276iMural-KO n = 610IControl n = 8Unpaired Student’s t testt = 30 min0 . 003328841iMural-KO n = 6t = 60 min0 . 04891949110KControl n = 8Unpaired Student’s t testt = 15 min0 . 03752iMural-KO n = 610-S1CControl n = 7Unpaired Student’s t testgWAT7 . 8E-08iMural-KO n = 710-S1EControl n = 8Unpaired Student’s t testiWATUcp10 . 015960549iMural-KO n = 6gWATUcp10 . 005703268rWATUcp10 . 01188325110-S1FControl n = 5Unpaired Student’s t testiWAT0 . 00019253Vis-KO n = 5gWAT0 . 042623514mWAT0 . 017424031rWAT0 . 00228672 | Mammals have different types of fat cells in their bodies . White fat cells store energy for later use , and brown and beige fat cells burn energy to help keep the body warm . Individuals who are obese typically have too many white fat cells in and around their belly . This belly fat , also called visceral fat , accumulates around the organs and is believed to contribute to metabolic diseases , such as diabetes and heart disease . Individuals who are obese also have relatively few brown and beige energy-burning fat cells . Boosting the amount of brown and beige fat in individuals who are obese has been proposed as a potential way to reduce their risk of metabolic disease . One way to do this would be to encourage white visceral fat cells to become more like energy-burning beige or brown fat cells . Recent research has shown that white fat cells contain higher amounts of a protein called Zfp423 than brown or beige fat cells . This protein turns off the genes that fat cells use to burn energy and so keeps white fat cells in an energy-storing state . Now , Hepler et al . show that genetically modifying mice to turn off the gene that produces Zfp423 specifically in the precursor cells that become white fat cells causes more energy-burning beige cells to appear in their visceral fat . The genetically modified mice were better able to tolerate cold than normal mice . When placed on a high-fat diet , the modified mice were also less likely to become resistant to the effects of the hormone insulin – a process that can lead to the development of type 2 diabetes and may be linked to heart disease . This suggests that treatments that prevent Zfp423 from working in fat cells could help to treat or prevent diabetes and heart disease in people who are obese . Before such treatments can be developed , further work is needed to investigate how Zfp423 works in more detail , and to confirm that it has the same effects in human fat cells as it does in mice . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine"
] | 2017 | Directing visceral white adipocyte precursors to a thermogenic adipocyte fate improves insulin sensitivity in obese mice |
Segmented RNA viruses are ubiquitous pathogens , which include influenza viruses and rotaviruses . A major challenge in understanding their assembly is the combinatorial problem of a non-random selection of a full genomic set of distinct RNAs . This process involves complex RNA-RNA and protein-RNA interactions , which are often obscured by non-specific binding at concentrations approaching in vivo assembly conditions . Here , we present direct experimental evidence of sequence-specific inter-segment interactions between rotavirus RNAs , taking place in a complex RNA- and protein-rich milieu . We show that binding of the rotavirus-encoded non-structural protein NSP2 to viral ssRNAs results in the remodeling of RNA , which is conducive to formation of stable inter-segment contacts . To identify the sites of these interactions , we have developed an RNA-RNA SELEX approach for mapping the sequences involved in inter-segment base-pairing . Our findings elucidate the molecular basis underlying inter-segment interactions in rotaviruses , paving the way for delineating similar RNA-RNA interactions that govern assembly of other segmented RNA viruses .
Genomes of rotaviruses ( RVs ) , and other pathogens of the Reoviridae family , comprise nine to twelve double-stranded ( ds ) RNA segments , co-packaged into each infectious virion . The RNA assortment process , during which a single distinct positive-sense ssRNA copy of each of the genomic segments is selected for packaging and replication , occurs within cytoplasmic inclusion bodies termed viroplasms ( Patton and Spencer , 2000; Trask et al . , 2012 ) . Multiple copies of viral RNAs and proteins accumulate in viroplasms , of which the non-structural ssRNA-binding proteins NSP2 & NSP5 are essential components ( Patton and Spencer , 2000; Trask et al . , 2012; Taraporewala and Patton , 2004 ) . Current views of the mechanisms of segment assortment are based on the idea that selection of the correct genomic segments is governed by inter-segment RNA-RNA interactions ( Gavazzi et al . , 2013a; Gavazzi et al . , 2013b; Fajardo et al . , 2015; McDonald et al . , 2016 ) . However , the analysis of such interactions remains a particularly challenging task due to a vast number of RNA-RNA contacts present in multiple folding intermediates of large RNAs ( Solomatin et al . , 2010; Woodson , 2010a; Woodson , 2010b; Laing and Schlick , 2011 ) . This problem is further confounded by non-specific RNA self-association and aggregation , particularly in the presence of RNA-binding proteins , at concentrations mimicking in vivo assembly conditions ( Borodavka et al . , 2012 ) . Recently , several tools for investigating RNA-RNA interactions have been developed , based on proximity-ligation of interacting RNAs ( Helwak and Tollervey , 2014 ) and psoralen-mediated crosslinking of RNA duplexes ( Engreitz et al . , 2014; Lu et al . , 2016; Sharma et al . , 2016; Aw et al . , 2016 ) . While these powerful methods are well suited for detection of RNA duplexes in cells , further experimental validation of the identified RNA-RNA contacts is often required ( Weidmann et al . , 2016 ) . Moreover , such techniques give little insight into the dynamics and stability of the observed RNA-RNA interactions , particularly when they are involved in the assembly of macromolecular complexes . To address these challenges , we have developed an experimental framework for interrogating RNA-RNA interactions by taking advantage of two-colour fluorescence cross-correlation spectroscopy combined with pulsed interleaved excitation ( PIE-FCCS ) ( Müller et al . , 2005; Schrimpf et al . , 2017 ) . Such assays offer unprecedented capacity for detection of stable , sequence-specific interactions between labeled RNAs in complex mixtures of RNAs and proteins . Using PIE-FCCS , we show that incubation of a full set of eleven genomic ssRNAs with the non-structural protein NSP2 results in de novo formation of specific inter-segment RNA-RNA contacts . By testing pairwise RNA-RNA interactions between segment 11 ssRNA ( S11 ) and other RV ssRNAs , we demonstrate that S11 preferentially binds to a subset of genomic ssRNAs . We show that NSP2 binding to S11 RNA results in its structural reorganization , concomitant with the exposure of single-stranded areas , required for stabilization of new inter-segment interactions . The weakest detected pairwise interaction between S11 and S10 RNAs is significantly enhanced in the presence of a full set of eleven ssRNAs , suggesting formation of a complex RNA interaction network , stabilized by NSP2 . To gain further insights into specific inter-segment contacts of RV RNAs , we introduce an RNA-RNA SELEX approach for mapping the genomic sequences , mediating such interactions . Using FCCS methodology combined with RNA mutagenesis studies , we validate the sites of inter-segment RNA-RNA interactions , identified via RNA-RNA SELEX . This integrated approach provides unique insights into the stability of macromolecular complexes , containing multiple RNAs , and it can be universally applied for investigating assembly of other segmented RNA viruses and ribonucleoproteins .
Unlike most ensemble methods previously used for detecting inter-segment interactions in segmented RNA viruses ( Gavazzi et al . , 2013b; Fajardo et al . , 2015; Fournier et al . , 2012 ) , fluorescence correlation spectroscopy ( FCS ) allows probing of such interactions in extremely dilute solutions , effectively eliminating self-association and aggregation of RNAs ( Borodavka et al . , 2012; Borodavka et al . , 2016 ) . We employed this technique to identify specific inter-molecular RNA-RNA interactions , which remain stable at low sub-nanomolar concentrations . We used a dual-color extension of FCS , fluorescence cross-correlation spectroscopy ( FCCS , Materials and methods and Figure 1—figure supplement 1 ) for detecting interactions of differently labeled RNAs in the presence of unlabeled molecules , which may be required for stabilization of such RNA-RNA contacts . We first investigated whether the RV ssRNAs could spontaneously associate into larger RNA complexes in vitro . We examined binding of S1-S10 ssRNAs ( Figure 1—figure supplement 2 ) to the smallest genomic RNA S11 , for which there is a complete secondary structure model available ( Li et al . , 2010 ) . As described in Materials and methods , ATTO647-labeled S11 and ATTO565-labeled S10 ( + ) ssRNAs were incubated with unlabeled ( + ) ssRNAs S1 to S9 . After incubation , RNA samples were diluted ( 1 nM each ) and examined by two-color FCCS . Analysis of the cross-correlation function ( CCF ) between dye-labeled S11 and S10 RNAs suggests that the two RNAs do not interact with each other in the presence of a full set of eleven genomic ssRNA segment precursors ( Figure 1A , zero amplitude of CCF , shown as dashed magenta line ) . Upon incubation with eleven genomic ssRNAs , the apparent hydrodynamic radius , Rh of the S11 RNA remained unchanged , consistent with lack of RNA oligomerization under those conditions ( Supplementary file 1 ) . Because RNA segment assortment is likely to occur in the presence of the non-structural proteins NSP5 and NSP2 ( 3 , 24 ) , the latter capable of helix-unwinding and strand-annealing reactions in vitro ( Taraporewala and Patton , 2001; Borodavka et al . , 2015 ) , we first examined inter-segment RNA interactions in the presence of NSP2 . Incubation of eleven distinct ssRNAs S1-S11 with a 100-fold molar excess of NSP2 , that is , enough NSP2 to promote strand-annealing ( Borodavka et al . , 2015 ) , both labeled S11 and S10 formed a stable complex ( Figure 1A , CCF in blue ) . Analysis of the resulting ACFs and the CCF confirms that both NSP2-bound RNAs have similar hydrodynamic radii , Rh ~30 nm , significantly larger than the protein-free S11 and S10 ssRNAs ( Supplementary file 1 ) . Having established that the S11:S10 interaction is formed in the presence of a full genomic set of eleven ssRNAs upon incubation with NSP2 , we then explored whether the remaining S1-S9 RNAs are involved in stabilization of the detected interaction . Omitting S1-S9 RNAs from the reaction significantly decreased the apparent CCF amplitude between S11 and S10 RNAs ( Figure 1 , cf . panels A and B ) . This result suggests that the observed S11:S10 interaction involves additional RNAs that may be required to form a larger supramolecular complex ( Trask et al . , 2012; Gavazzi et al . , 2013a; Gavazzi et al . , 2013b ) . Alternatively , binding of one or more RNAs to either S11 or S10 may result in stabilization of the RNA conformations , which favor stable base-pairing between the two RNAs . Either interaction model predicts that some of the remaining segment ssRNAs ( S1-S9 ) should strongly interact with S11 RNA . We then examined the remaining pairwise interactions between S11 and S1-S9 RNAs , after incubating an equimolar mixture of the two differently labeled RNAs with NSP2 ( Figure 1B ) . The strongest pairwise interactions of S11 were observed for segments S3 , S6 and S5 , with CCF amplitudes similar to those , observed for the S11:S10 complex , incubated in the presence of S1-S9 RNAs ( CCFs ~ 0 . 5 , Figure 1A ) . The apparent CCF amplitudes between S11 and any other segment RNA are not proportional to the length or the size of the interacting RNA partner ( Supplementary file 1 ) , suggesting that it is the RNA sequence , rather than its size , that is important for S11 binding . We also examined the effect of NSP5 on ssRNA binding and inter-segment RNA complex formation . Both NSP2 and NSP5 were capable of binding a short 18-nt long ssRNA with nM affinity , yielding the apparent hydrodynamic radii , Rh , closely matching the sizes of NSP2 octamers ( Schuck et al . , 2001 ) , or NSP5 decamers ( Martin et al . , 2011 ) , respectively ( Figure 1—figure supplement 3A ) . Despite its ssRNA-binding activity , incubation of NSP5 with ssRNAs S5 and S11 at saturating amounts did not yield any detectable cross-correlation ( Figure 1—figure supplement 3B ) , indicating that NSP5 binding to ssRNAs did not promote the formation of a stable S5:S11 RNA complex . Remarkably , co-incubation of NSP2 in the presence of sub-micromolar amounts of NSP5 resulted in a marked decrease of the S5:S11 CCF amplitude , compared to the NSP2 incubation alone ( Figure 1—figure supplement 3C ) . Further addition of NSP5 resulted in severe aggregation of NSP2 , as previously reported by Jiang et al . ( 2006 ) . Given the NTPase and autophosphorylation activities of NSP2 ( 30–33 ) , we also investigated the effect of ATP on the formation of RNA-RNA contacts in the presence of NSP2 . Addition of 1 mM ATP did not affect RNA binding by NSP2 octamers ( Figure 1—figure supplement 4A ) , nor had it any effect on the apparent S11:S5 CCF amplitude ( Figure 1—figure supplement 4B ) . Collectively these data suggest that inter-segment RNA contact formation is not coupled to ATP hydrolysis and thus most likely to be dependent on the ssRNA-binding activity of NSP2 . To further investigate this , we examined the interactions between S5 and S11 RNAs co-incubated with a C-terminally truncated NSP2 mutant ( ΔC-NSP2 ) with a significantly reduced affinity for ssRNA ( Hu et al . , 2012 ) . The analysis of S11 and S5 RNAs ACFs reveals that both interacting RNAs recruited multiple copies of the full-length NSP2 or its deletion mutant ΔC-NSP2 , respectively , yielding similar hydrodynamic sizes of S5 ( Figure 1—figure supplement 5A , light and dark green ACFs ) . However , the apparent CCF amplitude between S5 and S11 was significantly reduced when the full-length NSP2 was replaced with ΔC-NSP2 ( Figure 1—figure supplement 5A , CCFs in blue and magenta ) . A drastic reduction of the CCF amplitude between S5 and S11 ssRNAs was also observed when rotavirus group A NSP2 was substituted with NSP2 from rotavirus group C ( Figure 1—figure supplement 5B ) , despite the apparent structural resemblance and similar enzymatic activities of both proteins ( Hu et al . , 2012 ) . Together , these results strongly suggest that the formation of stable inter-segment RNA-RNA contacts requires selective binding of RV group A NSP2 to ssRNAs , but not ssRNA-binding proteins NSP5 or RV group C NSP2 . Given that NSP2 has a capacity to bind multiple RNAs ( Schuck et al . , 2001 ) , the observed cross-correlation may reflect its simultaneous binding to differently labeled segment ssRNAs . The differences in relative affinities between S11 and other ssRNAs ( Figure 1B ) strongly suggest that the detected RNA interactions are not due to non-specific aggregation , resulting from NSP2 binding . Moreover , the observed inter-segment interactions remained stable in highly dilute solutions ( <1 nM ) , while the apparent affinity of NSP2 for ssRNA is nanomolar ( Hu et al . , 2012 ) . Thus , we hypothesized that the high stability of these interactions could arise from strong , specific inter-molecular RNA contacts , formed upon incubation with NSP2 . To test this , we removed NSP2 from the preformed stable RNA-RNA complex S11:S5 by digesting it with proteinase K , and examined the sample by FCCS . The interacting S5 and S11 RNAs had lower Rh upon removal of NSP2 ( Table 1 and Figure 1—figure supplement 6 ) , suggesting that both RNAs were protein-free . These highly diluted NSP2-free RNA-RNA complexes remained stable at 37°C for extended periods of time ( >30 min , Figure 1—figure supplement 6A ) . As a further control , we made a protein-free RNA-RNA complex S11:S5 by heat-annealing a mixture of fluorescently labeled RNAs S11 and S5 ( Materials and methods ) , and examined it by FCCS . Both heat-annealed RNAs strongly interacted ( Figure 1—figure supplement 6B ) , yielding an apparent Rh ~13–17 nm typical for a proteinase K-digested complex S11:S5 ( Table 1 ) . Together , these data strongly support the view that inter-segment interactions mediated by binding of NSP2 remain stable after removal of NSP2 due to the formation of new , specific inter-molecular RNA contacts . Having established that S11 preferentially binds to S3 , S5 and S6 RNAs , we devised an RNA-RNA SELEX methodology to identify the S11 RNA sequences that can form stable inter-molecular contacts with other segment ssRNAs . We used this approach to analyze the distribution of the RNA-RNA interaction sites in S11 alone , and in complex with NSP2 ( Figure 2A ) . As described in Materials and methods , we used a naïve RNA library comprising ~3 . 6×1013 unique 30-mer ssRNA sequences . Biotinylated S11 RNA was incubated with the naïve RNA library ( Figure 2A ) , and sequences strongly interacting with immobilized S11 RNA were separated from excess NSP2 and low affinity 30-mers through multiple rounds of washes . After several rounds of sequence enrichment , the resulting pool of 30-mers was subjected to high-throughput sequencing , yielding 4 . 9 × 106 unique sequences . To identify the sequences that strongly interact with S11 RNA , we aligned the reverse complements of all individual SELEX-enriched sequences to S11 using a probabilistic score-based approach ( Materials and methods ) . The resulting histogram indicates the areas , in which multiple SELEX-enriched sequences align to S11 RNA ( the peaks shown as a black dashed line , Figure 2B ) . The histogram peaks are correlated with highly exposed areas of S11 RNA that strongly interact with ssRNAs during the RNA-RNA SELEX procedure . Similarly , S11 areas with poor accessibilities , e . g . , stable helices , yielded the lowest number of sequence alignments . As the test of the ability of our approach to report on secondary structure of S11 RNA , we computed its structure using the constraints , derived from the RNA-RNA SELEX experiment ( Materials and methods ) . The analysis enabled us to precisely map the conserved helices H1-H3 ( Figure 2B and Figure 2—figure supplement 1 ) , consistent with the structure probing data and covariation analyses ( Li et al . , 2010 ) . Similarly , multiple histogram peaks aligned well to various loop regions that had been previously identified in S11 RNA ( Figure 2—figure supplement 1 ) . Overall , the RNA-RNA SELEX data reproduced multiple features of the target RNA structure , demonstrating that our approach does not perturb its solution conformations . More importantly , RNA-RNA SELEX allows direct identification of the RNA areas capable of forming stable inter-molecular contacts . Because the folded structures of interacting RNAs in isolation would present a substantial energy barrier for forming inter-molecular base pairs between them , RNA-RNA SELEX favors selection of unstructured RNA sequences . We reasoned that binding of NSP2 to ssRNAs would alter their structures and , thus , accessibilities of the RNA-binding sites . Analysis of the RNA-RNA SELEX data revealed significant enrichment of the RNAs interacting with S11 in the presence of NSP2 ( Figure 2B , solid red line ) , consistent with its strand-annealing activity ( Taraporewala and Patton , 2001; Borodavka et al . , 2015 ) . A comparison between the two SELEX experiments uncovered a number of additional RNA-binding sites in S11 RNA after its incubation with NSP2 ( Figure 2B and Figure 2—figure supplement 2A ) . The overall architecture of S11 , including stable helices H1-H3 , was not affected by NSP2 binding ( Figure 2B and C ) . Interestingly , the majority of the newly formed RNA-binding sites were located near the loop regions and helical junctions of the RNA ( Figure 2C ) , potentially reflecting the sites of preferential binding of NSP2 to S11 RNA . To further confirm that NSP2 binding to S11 results in conformational change of the RNA , we monitored the circular dichroism ( CD ) spectrum of S11 RNA in the presence of NSP2 , as described in Materials and methods . The protein-free S11 RNA spectrum is characteristic of an A-type helical conformation ( Figure 2—figure supplement 2B ) , typical for a folded RNA ( Borodavka et al . , 2015 ) . A positive 260–265 nm band decreases in response to increasing amounts of NSP2 ( Figure 2—figure supplement 2B , shown in blue ) , or thermal melting ( shown in red ) , indicating that the global helical RNA fold is destabilized by NSP2 binding , consistent with the structural rearrangements due to NSP2-mediated RNA unfolding . Having established that NSP2 binding is a prerequisite for inter-segment RNA-RNA interactions , we aligned the SELEX-enriched sequences to the S1-S10 RNAs ( Figure 3 ) . The analysis revealed multiple sequences within S1-S10 genome segments with a potential to interact with S11 RNA ( Figure 3 , black peaks ) . Remarkably , the distribution of the identified peaks changes in the presence of NSP2 ( Figure 3 , peaks shown in red ) , reflecting the observed conformational rearrangement of the S11 RNA . The identified sequences appear in the RV RNA segments with high statistical significance ( >200 hits using a statistical score of 14 or above for sequences , enriched in the presence of NSP2 ) , compared to the non-rotaviral control sequence ( Non-RV , Materials and methods and Figure 3—figure supplement 1A ) . This result suggests that the RV segment precursors may contain multiple sites that have the potential to interact with S11 RNA in the presence of NSP2 . Next , we compared the number of the hits resulting from RNA-RNA SELEX experiments , with the relative amplitudes of CCFs , determined in pairwise RNA-RNA interaction assays . There is a positive correlation between the abundance of S11-interacting sequences present in S1-S10 and the respective CCF amplitudes ( Figure 1B and Figure 3—figure supplement 1B ) . Because RNA-RNA SELEX provides the information about the accessibility of a target RNA , but not its interacting partners , we analyzed the sequences of segments S3 , S5 and S6 , which strongly interacted with S11 in the presence of NSP2 . We selected the peaks with significant sequence enrichment , capable of stable base-pairing with S11 RNA in the presence of NSP2 ( Figure 3 , peaks highlighted in magenta boxes ) . The identified sequences within S3 , S5 and S6 RNAs can stably base-pair with S11 RNA ( ΔG = −15 . 3 to −19 . 1 kcal/mol , Figure 3 , insets and Figure 3—figure supplement 2 ) . Overall , the stabilities of the identified inter-segment RNA-RNA duplexes correlate well with the apparent CCF amplitudes . We carried out similar analyses for RNAs S1 , S4 and S10 , which also contain sequences that have a potential of base pairing with S11 ( Figure 3 , highlighted in blue boxes ) . Despite the observed enrichment of the sequences interacting with S11 RNA , identified in S1 , S4 and S10 , the apparent CCF amplitudes determined in pairwise RNA-RNA interaction assays are low . This suggests that the identified RNA-RNA interaction sites may be sequestered by local secondary structures . To investigate this further , we performed FCCS with ATTO647-labeled S10 RNA and a Cy3-labeled ssRNA probe , representing region 84–100 of S11 RNA ( complementary to the region 272–288 of S10 RNA that corresponds to the peak shown in a blue box in Figure 3 ) . No significant cross-correlation between S10 and a complementary RNA sequence was observed after incubation with NSP2 ( Figure 3—figure supplement 3 ) . In contrast , heat-annealing of a mixture of the two RNAs resulted in a high CCF amplitude ( Figure 3—figure supplement 3B and C ) . This result strongly suggests that the identified sequence within S10 RNA is sequestered , and the free energy change of opening the intramolecular base pairs by NSP2 binding is too high . This idea is further corroborated by the analysis of secondary structure of S10 RNA ( Figure 3—figure supplement 4G ) . Similar analyses of S1 and S4 sequences ( Figure 3—figure supplement 4 ) suggest the identified S11 RNA-binding sites may be obstructed by local secondary structures that prevent interactions between S11 and those RNAs . As a further control , we investigated the effects of mutations disrupting inter-molecular base-pairing between S11 and the interacting RNAs . We examined interactions between S11 and S5 by introducing nucleotide substitutions within the identified RNA-RNA interaction sites ( nts 307–322 in S5 and 51–66 in S11 , Figure 4A ) . Substitution of either RNAs with its mutated counterpart ( S11mut or S5mut ) resulted in a dramatic reduction of the CCF amplitude ( Figure 4A , red and blue dotted lines ) , while the wild-type S5 and S11 RNAs formed stable complexes in the presence of NSP2 ( Figure 4A , solid green line ) . Similarly , incubation of S11 with a non-RV control RNA resulted in low CCF amplitude ( Figure 4A , dotted magenta line ) , reflecting its low propensity to form stable base-pairs with S11 RNA ( Figure 3—figure supplement 1 ) . Both combinations of S11 with S5mut and non-RV RNAs have similarly low CCF amplitudes , suggesting that the interacting RNAs did not stably interact . To further test this , we computer-generated a scrambled S11 RNA sequence with low propensity to form extensive base-pairing with S5 RNA ( Materials and methods ) . Using FCCS , we probed the interactions between the ‘scrambled’ RNA and S5 RNA . The resulting apparent CCF amplitude was lower compared to the S5:S11mut combination ( Figure 4A ) , suggesting that the observed weaker interactions are mediated by non-specific residual base-pairing between the mutated RNAs . We then used the two-color FCCS approach to probe the stability of S5:S11 RNA complex . The FCCS analysis of the preformed S5:S11 RNA complex , incubated at elevated temperatures , yielded the estimated melting temperature , Tm ~ 47°C ( Figure 4B ) . This value is in good agreement with the Tm of the identified RNA-RNA duplex , formed between S5 and S11 ( Figure 3 , inset , Tm = 46°C , calculated as described in Materials and methods ) . We also introduced compensatory nucleotide substitutions in the putative interacting sequences of S11 and S5mut in order to test if the disrupted inter-segment contacts could be rescued by restoring the stable base-pairing between the mutated RNAs . The resulting mutated trans-complementary RNAs S5mut and comp_mutS11 strongly interacted with each other upon incubation with NSP2 ( Figure 4—figure supplement 1A and B ) . Moreover , the CCF amplitude between the trans-complementary mutant RNAs was higher than that of the interacting wild-type RNA sequences , consistent with the increased stability of the rescued RNA-RNA contacts ( Figure 4—figure supplements 1B , –18 . 8 kcal/mol for the S5:S11 complex and −23 . 7 kcal/mol for S5mut:S11comp ) . Collectively , these data demonstrate the importance of the identified sequences in the formation of stable and sequence-specific RNA-RNA contacts between S5 and S11 . We also carried out similar mutagenesis analyses for segment precursors S3 and S6 ( Figure 4—figure supplement 2 ) , confirming the identified RNA-RNA interaction sites in S11 RNA . All identified sequences ( Figure 3 , peaks highlighted in magenta ) were capable of forming stable duplexes with S11 RNA ( insets in Figure 3 and Figure 3—figure supplement 2 ) , consistent with the high CCF amplitudes observed for the S3:S11 and S6:S11 complexes . Collectively , our data provide extensive experimental evidence that sequence-specific inter-molecular base-pairing mediated by NSP2 binding to viral ssRNAs , governs inter-segment interactions in RVs .
Several models of segmented genome encapsidation have been proposed to explain selective packaging of eleven RNAs in RVs ( Patton and Spencer , 2000; Trask et al . , 2012; McDonald et al . , 2016 ) . Here , we provide direct evidence of specific , inter-segment RNA-RNA interactions in RVs , by establishing experimental assays based on a combination of two-colour FCCS and an RNA-RNA SELEX . The FCCS-based assays allow rapid quantitative analyses of the stability of RNA-RNA and ribonucleoprotein complexes , providing additional information about the stoichiometry of the RNAs within such complexes . To delineate the sequences that mediate contacts between the RV ssRNAs , we have employed the RNA-RNA SELEX approach for identifying stable , specific inter-segment RNA-RNA interactions . The data presented here collectively support the model in which RV ssRNAs can specifically interact with each other as a result of RNA remodeling brought about by the ssRNA-binding protein NSP2 ( Figure 5 ) . The results of RNA-RNA SELEX against S11 reveal multiple areas of the RNA undergoing conformational rearrangements upon NSP2 binding , while the most stable S11 intra-molecular helices H1-H3 remain largely inaccessible . This model is consistent with the high affinity of NSP2 for ssRNA , but not dsRNA , which may underlie the protein’s ability to alter the kinetics of RNA-RNA hybridization by accelerating the breakdown of weaker intra-molecular helices , concomitant with the stabilization of the alternative , stable inter-molecular contacts . Substitution of NSP2 with its mutant ΔC-NSP2 with significantly lower affinity for ssRNA reduces the efficiency of formation of inter-segment contacts , further supporting the proposed model of NSP2-mediated remodeling of RV ( + ) ssRNAs . Such a mechanism would account for NSP2-facilitated selection of thermodynamically favorable inter-segment interactions that may not always follow strict pairing rules . Because of the high affinity of NSP2 for ssRNA it is expected that NSP2 may remain associated with the oligomeric ssRNA complex prior its replication . Given that NSP2 inhibits the initiation of replication , but not the elongation stage ( Vende et al . , 2003 ) , it is possible that the inner core protein VP2 binding to the ssRNA-NSP2 complex ( Guglielmi et al . , 2010; Estrozi et al . , 2013; Patton , 1996 ) would result in the initial displacement of the NSP2 during the core assembly nucleation , followed by further eviction of NSP2 during the elongation . In this model , the assembly of the core is nucleated by multiple copies of the RV polymerase VP1 bound to eleven pre-selected ( + ) ssRNAs ( Patton , 1996 ) , arranged within the ssRNA-NSP2 complex at defined spatial locations . The highly basic N-terminal ‘arms’ of VP2 would interact with VP1 and ssRNA ( Zeng et al . , 1998; Guglielmi et al . , 2010; Estrozi et al . , 2013 ) within the ribonucleoprotein complex , which could also facilitate the formation of the VP2 lattice around it . The proposed model would account for incorporation of multiple assorted ( + ) ssRNAs , each associated with its own copy of the VP1 polymerase ( Guglielmi et al . , 2010; Patton , 1996; Periz et al . , 2013 ) . Interestingly , at high concentrations VP2 self-assembles into empty core-like particles ( Crawford et al . , 1994 ) . This implies that VP2 nucleation must be tightly regulated to prevent early assembly to ensure specific packaging of eleven ( + ) ssRNAs ( Desselberger et al . , 2013; Viskovska et al . , 2014 ) . It is possible that nucleation of core assembly by ( + ) ssRNA complexes may occur at VP2 concentrations significantly lower , than those required for formation of empty cores ( Borodavka et al . , 2012 ) . Binding of NSP2 may also contribute to the regulation of over-nucleation of VP2 on the RNA substrate ( Viskovska et al . , 2014 ) . Early biochemical studies of RV assembly intermediates suggest that ( + ) ssRNA replication may take place within the pre-core intermediates lacking a complete VP2 shell ( Gallegos and Patton , 1989 ) . The proposed model would explain the absence of NSP2 inside the assembled virions , despite its abundance in viroplasms and high affinity for ssRNAs ( Figure 5 ) . Our studies also reveal several parallels in the ways , by which segmented RNA viruses , including RVs , Bluetongue virus ( BTV ) and influenza A viruses ( IAVs ) , may control selection of the correct RNA segments ( Fajardo et al . , 2015; Gerber et al . , 2014 ) . In BTV , multiple sites of the smallest segment S10 RNA have been shown to be involved in assortment of other genomic segments ( Fajardo et al . , 2015 ) . In IAVs , genome segment packaging into virions is coordinated via interactions among the different viral RNA segments , while the sites of inter-segment interactions overlap with the coding regions of segmental RNAs ( Gavazzi et al . , 2013b; Gerber et al . , 2014 ) . Our data suggest that the majority of RNA-RNA interaction sites are also located within the coding regions of the RV genomic segments ( Figure 4—figure supplement 3 ) . The IAV nucleoprotein ( NP ) has also been shown to regulate selective packaging of the viral genomic segments ( Moreira et al . , 2016 ) , revealing the distinct amino acid residues of NP that are required for packaging of specific RNA segments . Interestingly , the RNA-RNA interaction assays conducted in the presence of group C rotavirus NSP2 did not yield stable inter-segment RNA contacts normally formed upon co-incubating interacting ssRNAs with group A rotavirus NSP2 ( Figure 1—figure supplement 5B ) . Previously , using a NSP2 complementation system Taraporewala et al . reported that group C NSP2 could not rescue replication in NSP2-deficient cells infected with group A RV ( Taraporewala et al . , 2006 ) . Our assays present additional evidence that despite multiple structural similarities and analogous enzymatic activities , group C RV NSP2 not only fails to substitute group A RV NSP2 during assembly of viroplasms , but is also incompetent in promoting the formation of stable RNA-RNA contacts between segmental ssRNAs . Interestingly , the IAV segment ssRNAs were shown to form higher order species in vitro; however most of these interactions were only detectable upon co-incubation of the RNAs at 55°C ( Gavazzi et al . , 2013a ) . Similarly , BTV ssRNAs would also interact with each other , albeit with extremely low efficiencies , unless they were co-transcribed in vitro ( Fajardo et al . , 2015 ) . Using the FCCS-based approach , we did not detect inter-segment interactions prior the incubation with NSP2 , consistent with previously reported lack of association of the RNAs , emerging from the transcribing rotavirions ( Periz et al . , 2013 ) . Although the precise mechanism of the selective packaging of eleven distinct RNAs in RVs is still largely unknown , our results pave the way for delineating the mechanistic details and the RNA sequences essential for this process . Combining our approach with a recently established reverse genetics system for RVs ( Kanai et al . , 2017 ) will ultimately move us one step closer to understanding segment selection and encapsidation in these important pathogens . However , currently the analysis of such interactions in vivo is unfeasible , mostly due to the low rescue efficiency of the rotavirus reverse genetics system , through which only a very limited number of mutated genomic sequences have been rescued at this point ( T . Kobayashi , personal communication ) . In conclusion , the experimental approach reported here allows identification of RNA-RNA interactions without the need for cross-linking and intercalating agents , which may perturb dynamic RNA structures . It takes into account the kinetic effects of structure formation , e . g . , when RNA binding is concomitant with structural changes due to binding of other RNAs , or proteins . The FCCS-based assays presented here provide instant quantitative validation of RNA-RNA interactions and probe their stabilities . This approach can be used for high-throughput screening of the conditions that favor stabilization of RNA-RNA contacts , thus providing an additional dimension to the panoply of techniques for detecting base-paired RNA nucleotides . The combination of FCCS and RNA-RNA SELEX-based approaches holds promise for validation and quantitative analyses of the interacting viral ssRNAs and miRNAs . We anticipate that these methods will become useful tools in the armory of instruments for gaining novel mechanistic insights into the dynamics of genomic RNA-RNA interactions in RNA viruses .
Genome segment precursors S1-S11 ( GenBank IDs are listed in Supplementary file 2 ) of bovine rotavirus strain RF ( G6P6[1] ) were obtained as pUC19 cDNA clones ( Richards et al . , 2013 ) from Dr Ulrich Desselberger and Professor Andrew Lever ( University of Cambridge , UK ) . Mutant RNA sequences were produced from the original pUC19 plasmids harboring S11 , S5 and S6 cDNA clones using the Q5 site-directed mutagenesis kit ( NEB ) and the oligonucleotides listed in Supplementary file 2 . A DNA sequence was designed to make an RNA template with a minimal propensity to stably form base pairs ( longer than 8 bp ) with S5 and S6 RNAs . S11 RNA was used as an input sequence . It was mutated into a sequence with a minimal number of substitutions required to obtain a ‘scrambled’ sequence , which has only short ( maximum of 8 nt ) regions of complementarity with the sequences of S5 and S6 RNAs . The ‘scrambled’ 667-nt long RNA sequence , lacking extended regions of complementarity with S5 and S6 ( Supplementary file 2 ) , was cloned into a pUC19 vector , as previously described ( Borodavka et al . , 2016 ) . The non-rotavirus control RNA ( Non-RV ) was transcribed as a 1221-nt long Satellite Tobacco Necrosis Virus ( STNV ) genomic RNA from the DNA template as previously described ( Borodavka et al . , 2012 ) . cDNA clone of gene 8 of rotavirus A ( strain RF ) was used to PCR-amplify NSP2 ORF ( Supplementary file 2 ) with NcoI and XhoI restriction sites used for ligating the resulting double-digested NSP2-coding fragment into a linearized pET-28b vector . The resulting pET-28b-NSP2 construct was verified by sequencing and used for protein expression in BL21 ( DE3 ) E . coli as previously described ( Schuck et al . , 2001 ) . NTA-affinity purified NSP2 fractions were further purified over a HiTrap SP cation-exchange column ( Schuck et al . , 2001 ) . The concentrated peak fractions were resolved on a Superdex 200 10 × 300 GL column and pre-equilibrated with RNAse-free SEC buffer ( 25 mM HEPES-Na , pH 7 . 5 , 150 mM NaCl ) to ensure high purity and homogeneity of the preparation . Rotavirus group C NSP2 ( strain Bristol ) was expressed from pQE60g8C construct and purified , as described in Taraporewala et al . ( 2006 ) . Bovine rotavirus A ( strain RF ) protein NSP5 was expressed and purified as previously described in ( Martin et al . , 2011 ) . A plasmid for expression of the C-terminally truncated NSP2 variant ΔC-NSP2 was constructed using pET-28b-NSP2 vector by removing C-terminal residues 295–317 ( Hu et al . , 2012 ) . The resulting ΔC-NSP2 variant was expressed and purified following the purification procedures for a full-length NSP2 , as described above . RNA transcripts were produced by in vitro transcription of linearized DNA templates ( Supplementary file 2 ) using the HiScribe T7 RNA synthesis kit ( NEB ) . Transcribed RNAs were purified using RNeasy columns ( QIAGEN ) . All in vitro transcription reactions were carried out using the T7 RNA transcription kit ( NEB ) , following the manufacturer’s protocol , except for the fluorescently labeled or biotinylated RNA samples . RNA transcripts were labeled by introducing 5′-aminoallyl-uridine-5′-triphosphate ( 5-AA-UTP , Thermo Fisher ) during the in vitro transcription . Similar uridine derivatives with substituents at the fifth carbon have been previously shown to support the replication of RVs ( Silvestri et al . , 2004 ) . This minimally perturbing RNA labeling approach results in only few 5-AA-uridines incorporated per each transcript . Such a labeling strategy minimizes the influence of fluorescence quenching or Förster Resonance Energy Transfer ( FRET ) between the reporter dyes on the apparent cross-correlation amplitudes ( Foo et al . , 2012 ) . For labeling , amine-modified RNAs were produced by incorporating 5′-aminoallyl-uridine-5′-triphosphate ( 5-AA-UTP ) ( 3:1 ratio UTP:5-AA-UTP ) during T7 transcription ( Borodavka et al . , 2012 ) . Amine-modified RNAs ( 1–2 μM ) , in a total volume of 100 μL , RNAse-free 100 mM sodium borate buffer ( pH 8 ) , were reacted with 1 mM ATTO647- or ATTO565-NHS ester for 2 hr at 4°C . Under physiological conditions , the chosen dyes are characterized by zero net charge , relatively high hydrophilicity and high photo- and thermal stabilities , important for RNA melting experiments . Fluorescently labeled and biotinylated RNAs were purified using RNeasy columns , as previously described ( Borodavka et al . , 2012 ) . RNA labeling efficiencies were routinely examined spectrophotometrically ( Borodavka et al . , 2012 ) . Based on the estimated labeling efficiencies , typically 85% of all RNA molecules contained 1–6 dye molecules and less than 2% of RNAs were unlabeled . RNA biotinylation was carried out under similar conditions except EZLink Sulfo-NHS-LC-LC-biotin was used in lieu of NHS-esters of ATTO dyes . Further purification of biotinylated RNAs was carried out following the purification protocol used for dye-labeled RNAs ( see above ) . All RNA samples were routinely examined on denaturing formaldehyde agarose gels to ensure their integrity ( Figure 2—figure supplement 2B and C ) . For FCCS , to minimize any potential RNA self-association ( Figure 2—figure supplement 2D ) , all RNA samples were heat-annealed for 5 min at 70°C in 10 mM HEPES-Na , pH 7 . 4 , slowly cooled and diluted in an assay buffer ( 100 nM – 1 μM RNA in 20 mM HEPES-Na , pH 7 . 4 , 1 mM MgCl2 , 150 mM NaCl , 0 . 05% Tween 20 , 1 mM DTT ) . Reactions were set up with equimolar amounts of ATTO647-dye labeled S11 RNA or ‘scrambled’ or mutated S11 RNA sequence and ATTO565-dye labeled RNA Sx ( where x is any other interacting RNA partner ) . RNA samples were mixed in reaction buffer ( 0 . 1–2 . 5 μM total concentration ) and allowed to interact for 30 min at 37°C . For NSP2-mediated RNA-RNA-interaction assays , labeled RNA samples ( 150 nM total RNA concentration ) were incubated with 2 . 5–10 μM NSP2 ( or 2 . 5–10 µM BSA for control reactions ) . Reactions were allowed to proceed at 37°C for 30 min before they were stopped by diluting RNA samples into assay buffer , as discussed above , to achieve a final concentration of labeled RNAs of 1 nM ( each strand ) . These samples were further allowed to equilibrate at 25°C for 15 min prior to FCCS measurements . Proteinase K digestion of NSP2 was performed by adding 40 μg of proteinase K ( 800 U/ml ) to the NSP2-RNA samples and incubating them for 30 min at 37°C , as previously described ( Taraporewala and Patton , 2001 ) , prior to dilution into assay buffer as described above . Temperature melting experiments were performed by measuring the FCCS amplitudes of the pre-formed S5:S11 RNA-RNA complex ( 1 nM each RNA strand ) , treated with proteinase K . S5:S11 RNA-RNA complex was diluted into an assay buffer without MgCl2 . The diluted RNA samples ( 1 nM of each strand ) were incubated at 37-70°C for 15 min , after which they were slowly cooled down to 25°C prior to FCCS measurements . S10 RNA oligonucleotide hybridization assay was performed by heat-annealing RNA substrates ( 200 nM of ATTO647-labeled S10 RNA and Cy3-labelled RNA oligonucleotide Seq11_84_100 , as described in Supplementary file 2 ) for 5 min at 70°C in 100 mM NaCl , 20 mM HEPES-Na buffer , pH 7 . 4 , and slowly cooling down and diluting the complexes in assay buffer as described above . Dual-color fluorescence cross-correlation spectroscopy is a technique that allows sensitive detection of interactions of differently labeled molecules at low concentrations , typically with single-molecule sensitivity ( Figure 1—figure supplement 1 ) . FCCS detects coincident fluctuations in both the green and red channels . The amplitude of the cross-correlation function ( CCF ) is proportional to the number of double-labeled complexes and inversely proportional to the average number of labeled species in each individual channel ( i . e . , red and green ) . Therefore , the fraction of double-labeled species in a sample can be extracted from the ratio of the amplitudes of the cross-correlation ( CCF ) to the auto-correlation functions ( ACF ) . Conventionally , in dual-color FCCS , the signal from differently labeled molecules is divided into channels by emission ( e . g . , via dichroic mirrors and filters ) ( Schwille et al . , 1997; Bacia and Schwille , 2007 ) . Due to spectral crosstalk , an artificial cross-correlation signal can be observed between the two channels , even when the different species do not interact . When using pulsed interleaved excitation ( PIE ) in combination with the high temporal resolution of time-correlated single photon counting ( TCSPC ) , the signals can be further divided by the excitation source . This removes the influence of the crosstalk and greatly enhances both the sensitivity and specificity of this technique ( Müller et al . , 2005 ) . FCCS measurements of ATTO647 , and ATTO565 or Cy3-dye labeled RNAs were performed on a custom-built confocal microscope with time-correlated single photon counting ( TCSPC ) detection designed for pulsed interleaved excitation ( PIE ) , as previously described ( Müller et al . , 2005 ) . Two pulsed lasers at 561 nm ( frequency-doubled fiber laser , Toptica Photonics ) and 635 nm ( diode laser LDH-P-C-635b , Picoquant ) with a fixed repetition rate of 27 . 4 MHz were used for excitation with the laser power set to 10 μW , as measured before the objective . The emission of the red laser was delayed electronically by 20 ns with respect to the yellow laser to achieve pulsed interleaved excitation . A 60x , 1 . 27 NA objective ( Plan Apo IR 60 × WI , Nikon ) was used both to focus the excitation light and to collect the fluorescence . The fluorescence was focused on an 80 μm pinhole to remove out-of-focus light . Green and red emission signals were spectrally separated and focused on two single photon avalanche photodiodes ( APDs ) . The photon detection signals of each APDs are timed and recorded with separate TCSPC cards ( SPC-150 , Becker and Hickl ) . Data analysis and processing were performed using a custom-written Matlab-based data processing platform PAM ( PIE Analysis with Matlab , https://gitlab . com/PAM-PIE/PAM; a copy archived at https://github . com/elifesciences-publications/PAM-PIE-PAM ) , which includes tools for processing single point data collected using PIE ( Schrimpf et al . , 2017 ) . The ACFs and CCFs were fit with a one-component model showing dark state dynamics ( Bacia and Schwille , 2007 ) :G ( τ ) =γN⋅ ( 1+T1−T⋅e−τtT ) ⋅11+4Dτωr2⋅11+4Dτωz2+G ( ∞ ) Here , N is the number of individually diffusing molecules in the focus volume , D is the diffusion coefficient , and ωr and ωz are the distances from the center of the focus to the point where the fluorescence intensity has decayed to 1/e2 of the maximum intensity for the lateral and axial dimensions , respectively . T and tT denote the dark state fraction and the correlation time , respectively . G ( ∞ ) is a constant offset accounting for slow signal fluctuations . The geometric factor , γ , accounts for gradual intensity decrease of the assumed 3D Gaussian shape of the observation volume and is 0 . 35355 . The CCF amplitudes were normalized by the N , measured for the ATTO647 labeled S11 RNA ( or its mutated variants , including the ‘Scrambled’ RNA sequence ) . These normalized CCFs yield the fraction of the interacting ATTO565- and ATTO647-dye labeled ssRNAs . Thus , the fraction of interacting species is only an estimate of the absolute CCF value , as the stochastic labeling of the RNAs leads to the non-trivial weighting of a different number of fluorophores . In the case of complete binding with 1:1 stoichiometry , the ratio of the CCF/ACF amplitudes is expected to be close to one ( Foo et al . , 2012 ) . The lower absolute CCF amplitudes are attributed to small differences in the observation volumes for red and green fluorophores . RNA samples were measured at least three times , and the photon data were acquired for 10–15 min ( or up to 45 min when the stability of the preformed RNA-RNA complexes were investigated ) . The resulting ACF and CCF amplitudes were averaged , yielding mean CCF amplitudes reported here . A naïve 30-mer RNA library ( N30 ) consisting of ~3 . 6×1013 N30 random sequences flanked by primer sequences for RT-PCR amplification and transcription was produced by in vitro T7 transcription from a DNA template ( Supplementary file 2 ) . It was purified on Agencourt AMPure beads ( Beckman ) following the manufacturer’s protocol . Biotinylated S11 RNA target was prepared as described above . RNA target ( 200 nM ) was heat-annealed for 5 min at 70°C in a low ionic strength buffer ( 10 mM HEPES-Na , pH 7 . 4 ) , and slowly cooled to room temperature to minimize its oligomerization ( Li et al . , 2010 ) . A naïve N30 RNA library ( 10 μM ) , re-suspended in SB buffer ( 10 mM HEPES-Na , pH 7 . 2 , 150 mM NaCl , 1 mM DTT , 1 U/μL murine RNAse inhibitor ( NEB ) , 1 mM MgCl2 , 0 . 1 mg/ml bovine serum albumin , BSA ) was incubated with the S11 RNA target for 15 min at 37°C , either in the absence ( RNA alone selection ) or presence of 10 μM NSP2 ( NSP2 +selection ) . 20 μL of RNAse-free Dynabeads MyOne Streptavidin T1 beads ( Invitrogen ) were re-suspended in SB buffer and added to the S11 RNA samples followed by further incubation at 37°C for 15 min . Captured RNA targets with bound 30-mer RNAs were washed with SB buffer ( 200 μl ) 4 times for 10 min at 37°C during the first 3 rounds of selections , and the stringency of selection was increased after round 3 by increasing the number of washes up to 8 per round . Negative selections were carried out at every second round of SELEX using streptavidin-coated beads . S11-bound 30-mers were recovered by heat-elution ( 95°C , 5 min ) of the RNAs from carrier beads into 30 μl of RNAse-free water . Eluted RNA libraries were reverse-transcribed using Superscript III Reverse Transcriptase ( Thermo Scientific ) and primer P1 , and further PCR-amplified using primers P1 and P2 ( Supplementary file 2 ) . PCR-amplified SELEX-enriched cDNA libraries were analyzed by native PAGE after each group of two rounds of selection to confirm the presence of selected products for the next round of in vitro transcription and selection . In the final selection round , the resulting PCR products were prepared for Illumina MiSeq sequencing ( LIMM DNA Sequencing Facility , University of Leeds ) . We identified 4 , 857 , 793 unique sequences from the RNA alone SELEX data and 3 , 308 , 107 sequences from the NSP2+-enriched data . The lower number of the unique sequences from the NSP2 +selection is due to selective enrichment of certain RNA sequences in the presence of NSP2 . Each sequence from both data sets was aligned against all 10 segments ( S1-10 ) and the single best alignment was identified using a probabilistic score ( the Bernoulli score ) , which benchmarks the probability of a non-contiguous alignment to that of a contiguous alignment of N nucleotides ( Shakeel et al . , 2017 ) . Thus , a 30-nt long RNA sequence , which aligns to the genomic segment with a total of N non-contiguous matches ( N ≤ 30 ) , will have a Bernoulli score of N . The score N is equivalent to the probability that a short sequence of N nucleotides would match the sequence of the genomic segment ( Shakeel et al . , 2017 ) . We used the alignments with the Bernoulli score of 12 or above to analyze sequences that strongly interact with S11 RNA to identify highly accessible areas in S11 RNA . Higher scores ( 14 or above ) increase the stringency of the analysis ( see Figure 3—figure supplement 1 ) . We used RNA sequences with Bernoulli scores of 12 and above ( for alignments against S11 RNA target ) or 14 and above ( for other RNA segments ) to construct histogram plots for each genomic segment S1-S10 . The histogram bins are associated with individual nucleotides in each segment . Each sequence aligning to a nucleotide increments the histogram bin by one , so that the peak height corresponds to the number of sequences aligning to it . This approach allows identification of multiple RNA sequences that strongly interact with S11 RNA , including RNAs not fully matching the S1-S10 genomic sequences to account for nucleotide mismatches and possible gapped regions . We used the areas of high sequence alignments to identify regions that have a high probability of being single-stranded . These data were used as constraints for computing secondary structures of S11 RNA , with and without NSP2 , using Mfold ( Zuker , 2003 ) . CD spectra were acquired between 240 nm and 320 nm in a 1 cm-long path cell , as previously described ( 26 ) . CD spectra were recorded for the S11 RNA ( 200 nM in 10 mM HEPES-Na , pH 7 . 6 ) before and after incubation with either 10 µM NSP2 or RNAse-free acetylated bovine serum albumin ( BSA , New England Biolabs ) for 15 min . RNA secondary structure transitions upon thermal melting were also monitored at various temperatures up to 90oC . RNA secondary structures were predicted using minimum free energy ( MFE ) modeling in RNAfold to compute the thermodynamic ensemble of secondary structures and base-pairing probabilities . The centroid structures showing minimal base-pair distances to all other secondary structures in the Boltzmann ensemble were analyzed in order to identify regions of high and low accessibilities within S1-S10 RNAs . Similarly , the MFE structure of S11 RNA was calculated in Mfold ( RRID:SCR_008543 ) with the folding restraints derived from the RNA-RNA SELEX experiments . Differences in target RNA accessibilities were expressed as a normalized number of sequence alignments in the presence of NSP2 , after subtracting alignments , identified in the S11 structure alone . These values were used to calculate a color map , applied to the S11 RNA structure , shown in Figure 2C . RNA structures were visualized in VARNA ( Darty et al . , 2009 ) , and jVizRNA 2 . 0 was used for generating secondary structure circular plots ( Wiese et al . , 2005 ) . The free energies of hybridization of the interacting sequences were calculated using IntaRNA ( Freiburg RNA tools ) with the ensemble free energy calculations realized in Vienna RNA library ( Lorenz et al . , 2011; Wright et al . , 2014 ) . Duplex melting temperature ( Tm ) calculations were performed using nearest-neighbor parameters implemented in MELTING v . 5 . 0 ( 56 ) , estimated for 1 nM RNA strand concentration using the parameters matching the low-salt buffer conditions , used for FCCS measurements , as described above . | Rotavirus is a highly infectious virus that affects children worldwide , causing severe diarrhoea . Despite the introduction of several highly effective vaccines , more than 200 , 000 children still die from rotavirus each year . There are currently no drugs that can combat this disease once a child has been infected . Viruses carry the instructions that determine their properties and behavior in molecules of DNA or RNA . Unlike many other viruses , which typically have a single molecule of DNA or RNA , rotavirus has 11 distinct “RNA segments” . After invading a cell the virus begins to replicate itself . During replication , the RNA segments ( which consist of two strands of RNA paired together ) are copied many times . It is not clear how rotaviruses ‘count’ up to 11 so that each new virus acquires a single copy of each segment . Previous biochemical and structural studies of rotavirus replication suggest that selecting 11 distinct RNA segments must involve the RNAs forming complex interactions with proteins and other RNA molecules . Using a highly sensitive fluorescence-based approach , termed fluorescence cross-correlation spectroscopy , Borodavka et al . now present direct experimental evidence of interactions between the RNA segments that occur via single strands of the rotavirus RNA . These RNA-RNA interactions require the binding of a rotavirus protein NSP2 to the RNA strands , which results in the remodeling of the RNA; this remodeling is required to form stable contacts between different RNA segments . Furthermore , a new experimental approach ( called RNA-RNA SELEX ) developed by Borodavka et al . identified the parts of the RNA segments that may take part in these interactions . The results presented by Borodavka et al . pave the way for identifying the RNA-RNA interactions that govern how other segmented RNA viruses can package their genetic material . Further work to uncover the entire RNA interaction network in rotaviruses would also accelerate the design of new vaccines and may help us to develop antiviral drugs to treat infections . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"microbiology",
"and",
"infectious",
"disease"
] | 2017 | Protein-mediated RNA folding governs sequence-specific interactions between rotavirus genome segments |
Despite vast diversity in metabolites and the matching substrate specificity of their transporters , little is known about how evolution of transporter substrate specificities is linked to emergence of substrates via evolution of biosynthetic pathways . Transporter specificity towards the recently evolved glucosinolates characteristic of Brassicales is shown to evolve prior to emergence of glucosinolate biosynthesis . Furthermore , we show that glucosinolate transporters belonging to the ubiquitous NRT1/PTR FAMILY ( NPF ) likely evolved from transporters of the ancestral cyanogenic glucosides found across more than 2500 species outside of the Brassicales . Biochemical characterization of orthologs along the phylogenetic lineage from cassava to A . thaliana , suggests that alterations in the electrogenicity of the transporters accompanied changes in substrate specificity . Linking the evolutionary path of transporter substrate specificities to that of the biosynthetic pathways , exemplify how transporter substrate specificities originate and evolve as new biosynthesis pathways emerge .
Phospholipid-based cell membranes are the foundation for extant cellular life and with them arose the need for carrier proteins to shuttle metabolites across the semi-permeable membranes ( Mansy et al . , 2008; Deamer and Dworkin , 2005 ) . New biosynthesis pathways continuously emerged throughout evolution , resulting in a vast diversity in metabolite chemical structures ( >200 . 000 structures tentatively identified in the plant kingdom alone ) , where some are restricted to certain taxa and others are found broadly ( Weng et al . , 2012 ) . Matching the vast structural diversity of metabolites , approximately 10% of coding sequences of contemporary genomes encode transport proteins with diverse substrate specificities ( Saier and Ren , 2006 ) that enable transport of metabolites and ions into and out of cells . However , the evolutionary path that leads to the rise of new transporter substrate specificity upon emergence of new metabolites is unknown . Classical evolution theory ( Ohno , 2013 ) and several studies ( e . g . Fani and Fondi , 2009; Prasad et al . , 2012 ) support the hypothesis that new enzyme functions arise in duplicated genes if they are subject to unique selection pressure , - alternatively they rapidly become pseudogenes . A key example is found in the evolution of mineral corticoid and glucocorticoid receptors found in vertebrates ( Bridgham et al . , 2006; Carroll et al . , 2008 ) . These two receptors evolved post duplication of a dual specificity aldosterone and cortisol receptor basal to the jawed vertebrate lineage ( Bridgham et al . , 2006 ) . However , aldosterone biosynthesis did not arise before the advent of tetrapods suggesting that the ancestral receptor evolved affinity towards aldosterone before the hormone was present , possibly as a by-product of the receptors’ affinity towards chemically similar ligands ( Bridgham et al . , 2006; Carroll et al . , 2008 ) . Thus , it appears that selection pressure enforced upon related but distinct ligands can drive the emergence of receptors’ affinity towards a new substrate . In comparison , it is not clear if new transporters evolve de novo with emergence of new substrates , or whether gene duplications allow ancestral multifunctional proteins to take on greater specificity ( Khersonsky and Tawfik , 2010 ) . To answer this question , it is necessary to use a model system where the evolution of the biosynthetic pathway is known and where transporters have been identified . As a model system , we used the Brassicales-specific glucosinolate defense compounds with a biosynthetic pathway that diversified from the ancestral cyanogenic glucoside pathway found in more than 2500 plant species ( Sønderby et al . , 2010a; Halkier and Gershenzon , 2006; Bak et al . , 1998; Clausen et al . , 2015; Mithen et al . , 2010 ) . The two pathways share the initial enzymatic amino acid to oxime conversion , but produce structurally different end products ( Clausen et al . , 2015 ) . Through an arms race between plants and interacting organisms ( Bidart-Bouzat and Kliebenstein , 2008; Züst et al . , 2012; Kliebenstein et al . , 2005; de Vos et al . , 2008; Sanchez-Vallet et al . , 2010; Prasad et al . , 2012; Newton et al . , 2009; Fahey et al . , 2001; Agerbirk and Olsen , 2012 ) , the glucosinolate pathway evolved to produce >130 glucosinolate structures with diverse amino acid-derived side chains ( Fahey et al . , 2001; Agerbirk and Olsen , 2012 ) . Also , two H+/glucosinolate symporters , GTR1 and GTR2 , belonging to the NPF family ( Léran et al . , 2014 ) and with broad glucosinolate specificity ( i . e . no discrimination against amino acid side chain ) were identified in Arabidopsis thaliana ( Nour-Eldin et al . , 2012 ) that predominantly produces aliphatic and indole glucosinolates ( Mithen et al . , 2010; Brown et al . , 2003 ) . Although transporters for cyanogenic glucosides are yet to be identified ( Jørgensen et al . , 2005 ) , we set out to investigate whether the evolution of a new biosynthetic pathway ( here glucosinolates from cyanogenic glucosides ) promoted the co-evolution of transporter specificity , i . e . did glucosinolate transporters originate from cyanogenic glucoside transporters in the NPF family ? Furthermore , in planta studies suggest the existence of an additional glucosinolate transporter with narrow specificity for the recently evolved indole glucosinolates ( Andersen et al . , 2013 ) that are essential for innate immune responses ( Sanchez-Vallet et al . , 2010; Clay et al . , 2009; Bednarek et al . , 2009 ) . We therefore investigated if evolution within a biosynthetic pathway ( here emergence of indole glucosinolates ) is accompanied by evolution in transporter substrate specificity . Here we identify the first cyanogenic glucoside transporter in cassava and the first indole-specific glucosinolate transporter in A . thaliana . By characterizing substrate specificity and electrogenicity in orthologs along the phylogenetic lineage from cassava to A . thaliana , we provide a model for the evolutionary path of the substrate specificity of a plant specialized metabolite transporter . Surprisingly , we show that glucosinolate transport capacity likely occurred prior to the emergence of glucosinolate biosynthesis in dual-specificity transporters of cyanogenic glucosides and glucosinolates . With the emergence of glucosinolate biosynthesis , the transporters lost the capacity to transport cyanogenic glucosides . Moreover , we show that the first glucosinolate transporters had broad specificity and later subfunctionalized towards specific classes of glucosinolates . Our data suggests that changes in electrogenicity have accompanied the evolutionary changes in substrate specificity . Our results exemplify how new transporter substrate specificities evolve when new metabolites arise .
To assess the evolutionary path of GTR transporters , we first set out to identify the putative indole-specific glucosinolate transporter . In a previous study ( Nour-Eldin et al . , 2012 ) , we found that the glucosinolate transport capability of the NPF family is confined to the NPF2 . 8–2 . 14 transporters that cluster closely with AtGTR1 ( NPF2 . 10 ) and AtGTR2 ( NPF2 . 11 ) . We thus hypothesized that the indole-specific glucosinolate transporter could be found in this NPF subclade in A . thaliana ( Figure 1A ) . Via heterologous expression in Xenopus laevis oocytes , we screened six of the seven members within this subclade for transport of indol-3-yl-methyl glucosinolate ( I3M , the simplest indole glucosinolate ) and 4-methylthiobutyl glucosinolate ( 4MTB ) – representing a highly abundant aliphatic glucosinolate in A . thaliana ( Figure 1A–B ) . NPF2 . 9 ( At1g18880 , hereafter GTR3 ) - the closest homolog of GTR1 and GTR2 - transported I3M effectively ( Figure 1C ) . Two Electrode Voltage Clamp ( TEVC ) electrophysiology and time-course uptake assays showed that I3M , but not 4MTB , induces negative currents in GTR3-expressing oocytes ( Figure 2A–B ) and that GTR3 can over-accumulate I3M , but not 4MTB , against a concentration gradient ( Figure 2C–D and Figure 2—figure supplement 1A–B ) . Alternatively , an un-coupled conductance may accompany 4MTB transport in GTR3 resulting in non-electrogenic transport or transport rates may be below the electrophysiological detection level . In comparison , GTR1 over-accumulated both 4MTB and I3M ( Figure 2C–D and Figure 2—figure supplement 1A–B ) and elicited negative currents of similar amplitude for both glucosinolates ( Figure 2A–B ) . Plotting currents at −60 mV as a function of increasing I3M concentrations yielded a saturation curve best fitted by a Michaelis-Menten equation with Km towards I3M <25 uM for GTR1 , GTR2 and GTR3 ( Figure 2E–F and Figure 2—figure supplement 2 ) . Through competition assays we show that GTR3-mediated 4MTB uptake is strongly inhibited by 10% I3M , whereas 10-fold excess 4MTB does not affect I3M uptake ( Figure 3A–D ) . In contrast , GTR1 transports 4MTB and I3M to the same ratio as applied in the assay media ( Figure 3A–D ) . In accordance with previous characterization ( Wang and Tsay , 2011 ) , GTR3 imports nitrate into oocytes ( Figure 3G ) . Nitrate at concentrations 100-fold in excess of I3M or 4MTB did not outcompete uptake of neither glucosinolate , indicating that the two substrates are not mutually exclusive ( Figure 3E–G ) . In conclusion , our biochemical characterization shows that GTR3 is an electrogenic transporter with a high apparent affinity and strong preference for indole glucosinolates . We investigated the physiological relevance of GTR3´s role as an indole glucosinolate transporter in planta . GTR3 is strongly expressed in the plasma membrane of root phloem companion cells ( Wang and Tsay , 2011 ) and is co-expressed with GTR1 and GTR2 at the tissue level in other tissues according to publicly available translatome data ( Mustroph et al . , 2009 ) ( Figure 4—figure supplement 1 ) . A . thaliana gtr3 mutants accumulate significantly lower concentrations of indole glucosinolates in roots compared to wild type ( Figure 4A ) . This shift was increased in gtr1 gtr2 gtr3 triple knock-out ( tko ) , but was not seen in gtr1 gtr2 double knock-out ( dko ) ( Figure 4A–B , [Andersen et al . , 2013] ) . In the rosette , there is a trend , but no statistically significant increase in indole glucosinolates of the gtr3 mutant when compared to wild type ( Figure 4B ) . The gtr1 gtr2 dko shows a statistically significant increase in the rosette levels of indole glucosinolate that is further increased to 4 fold when also knocking out GTR3 ( gtr1 gtr2 gtr3 tko ) ( Figure 4B ) . This suggests that GTR1 , GTR2 and GTR3 all contribute to distributing indole glucosinolates between root and shoot . We used micro-grafting to further investigate the role of GTR1 , GTR2 and GTR3 in the source-sink relationship for indole glucosinolates between root and rosette . As MYB28 and MYB29 – key regulators of aliphatic glucosinolate biosynthesis ( Sønderby et al . , 2010b ) – are not necessary for expression of GTR1 , GTR2 and GTR3 ( Müller et al . , 2010 ) , we could use the glucosinolate biosynthetic null mutant - myb28/myb29 cyp79b2/cyp79b3 quadruple knockout ( qko ) - in micro-grafting experiments . By micro-grafting four-day-old A . thaliana seedlings of qko , gtr1 gtr2 gtr3 tko and wild type plants we created reciprocal grafts of roots and rosettes from all genotypes and analyzed glucosinolate content in root and rosette of three-week-old grafted plants . Based on substrate-specificity and overlapping expression of GTR1 , GTR2 and GTR3 we would expect that the distribution of aliphatic glucosinolates in a gtr1 gtr2 gtr3 tko would resemble the pattern in gtr1 gtr2 dko plants . In agreement , distribution of aliphatic glucosinolates in heterografts of gtr1 gtr2 gtr3 tko with wild type and qko plants , respectively , showed similar changes in distribution pattern for aliphatic glucosinolates as previously reported for heterografts of gtr1 gtr2 dko with wild type and qko plants , respectively ( Figure 4—figure supplement 2 and [Andersen et al . , 2013] ) . Furthermore , the grafting procedure does not influence the glucosinolate distribution as evidenced by homografts of wild type and gtr1 gtr2 gtr3 tko plants showing a similar distribution of indole glucosinolates as seen for non-grafted plants ( Figure 4C ) , and by homografts of qko plants being devoid of indole glucosinolates ( Figure 4C and [Andersen et al . , 2013] ) . Analysis of heterografted plants with no glucosinolate biosynthesis in the root showed that only small amounts of indole glucosinolates are transported from rosette to root . Similarly , it was evident from the ( qko/wt ) heterografts that when all three GTRs are expressed in the root , the root to rosette transport is below detection levels ( Figure 4C ) . However , when all three GTRs are knocked out in roots , we see a dramatic increase in the rosette indole glucosinolate content ( Figure 4 ) . In combination , this suggests that GTR3 ( along with GTR1 and GTR2 ) has a role in retaining indole glucosinolates in the root , presumably by importing indole glucosinolates into storage cells . From an evolutionary perspective , our findings propose two models for how substrate specificity evolved for the glucosinolate transporters . Either glucosinolate transport first arose with narrow specificity for indole glucosinolates followed by a broadening of the substrate specificity , or the reverse . To address this question and potentially determine when glucosinolate transport capability arose , we performed a phylogenetic analysis of NPF transporters from glucosinolate-producing species ( A . thaliana , Brassica rapa and Carica papaya ) and non-producing species ( Theobroma cacao ( cacao ) , Manihot esculenta ( cassava ) , Glycine max , Gossypium raimondi and Solanum lycopersicum ) . The phylogenetic analyses revealed three well-defined AtGTR1- , AtGTR2- and AtGTR3-containing subclades with NPF sequences exclusively from glucosinolate-producing species of the Brassicaceae ( A . thaliana and B . rapa ) ( Figure 5A and Figure 5—figure supplement 1 ) . Additionally , the analyses revealed a subclade , which grouped basal to the three GTR1–3 subclades . This subclade , which we name the GTR-like subclade contained GTR homologs from C . papaya , the most basal glucosinolate-producing species in Brassicales with a sequenced genome ( Mithen et al . , 2010; Goodstein et al . , 2012 ) , and from non-producing species ( M . truncatula , G . raimondii , S . lycopersicum , M . esculenta and T . cacao ) . The four subclades grouped in a larger clade , which we name the GTR-clade . To track the rise and evolution of glucosinolate substrate specificity we tested a range of transporters for glucosinolate transport activity via expression in X . laevis oocytes . All transporters described below were codon optimized for expression in X . laevis oocytes and tested for transport activity in their native form ( i . e . without tag ) . Additionally , we fused each gene to YFP in the C-terminus and confirmed its expression and localization to the plasmamembrane via confocal microscopy ( Figure 5—figure supplement 2 ) . In the following , lack of transport can therefore likely be attributed to lack of activity rather than lack of expression . Within the GTR1 and GTR3 subclades , we tested one of the respective orthologs from B . rapa and showed a strong preference for I3M by the tested BrGTR3 ortholog ( BrH02396 ) , whereas the tested BrGTR1 ortholog ( BrF01711 ) transported 4MTB and I3M with similar efficiency ( Figure 5B ) . Thus , the high preference for indole glucosinolates appears typical for GTR3 orthologs within the Brassicales-specific GTR3 subclade . Oocytes expressing GTR-like transporters from C . papaya over-accumulate both 4MTB and I3M relative to the assay media concentration and we named them CARICA PAPAYA Glucosinolate Transporter LIKE-1 ( CpGTRL1 ) and −2 ( CpGTRL2 ) , respectively ( Figure 5B ) . The ability of CpGTRL1 and −2 to transport both 4MTB and I3M is surprising as indole glucosinolates are found in A . thaliana and B . rapa , but not in C . papaya ( Mithen et al . , 2010 ) . Interestingly , CpGTRL2 from C . papaya transported 4MTB with a km of 85 ± 12 µM at −60 mV ( Figure 5D ) . This indicates that the high affinity of GTRs towards glucosinolates evolved before the diversification of the Brassicaceae and Caricaceae . Moreover , the data imply that the common ancestor of the GTR transporters was originally broad-specific and that GTR3 lost the ability to over-accumulate aliphatic glucosinolates after the divergence of C . papaya and the ancestor of Arabidopsis and Brassica ( ~72 . 1 MYA , median of 8 studies [Hedges et al . , 2006] ) . This suggests that preference for indole glucosinolates evolved as a subfunctionalization of ancestral , broad-specific glucosinolate transporters . To track the rise of transport capacity towards glucosinolates , we tested for glucosinolate uptake in the closest GTR homologs from cassava , which do not produce glucosinolates but produces the evolutionary related and ancestral cyanogenic glucosides ( McMohan et al . , 1995 ) . As a control , we included GTR homologs from cacao , which produces neither compound class ( Bjerg et al . , 1987; Seigler , 2005 ) . Oocytes expressing Me14G074000 from cassava over-accumulated both 4MTB and I3M relative to external media , while the expressed GTR homolog from cacao did not transport any glucosinolates ( Figure 5B , Figure 5—figure supplement 3 and Figure 5—figure supplement 2 ) . Hence , as cassava does not synthesize glucosinolates , the ability to transport glucosinolate appears to have arisen in the NPF family prior to the evolution of the glucosinolate biosynthetic pathway . Furthermore , the characterization of this potentially ancestral form of the glucosinolate transporters support that they first evolved with broad specificity towards aliphatic and indole glucosinolates . Glucosinolates and cyanogenic glucosides are structurally distinct ( Figure 1B and Figure 6A ) , yet they share functional moieties ( glucose moiety and amino acid-derived side chains ) ( Bak et al . , 1998; Clausen et al . , 2015 ) . Based on this and the existence of a transporter in cassava with glucosinolate transport capacity , we explored whether the glucosinolate NPF transporters may have evolved from cyanogenic glucoside transporters within the NPF family , much like glucosinolates evolved from cyanogenic glucoside biosynthesis ( Bak et al . , 1998; Clausen et al . , 2015; Mithen et al . , 2010 ) . To test this hypothesis we investigated the ability of selected GTRs and GTR-like homologs to transport representative cyanogenic glucosides , namely phenylalanine-derived prunasin and valine-derived linamarin ( Figure 6A ) . Notably , cassava produces only cyanogenic glucosides whereas C . papaya produces both cyanogenic glucosides and glucosinolates ( derived from phenylalanine ) . Oocytes expressing the glucosinolate transporters from C . papaya , B . rapa and A . thaliana did not accumulate the tested cyanogenic glucosides above trace amounts ( Figure 6A ) . In comparison , oocytes expressing Me14G074000 from cassava accumulated prunasin , but not linamarin , to levels equivalent to media ( Figure 6A and Figure 6—figure supplement 1 ) . This indicates that the substrate-binding cavity of this NPF transporter can accommodate both cyanogenic glucosides and glucosinolates . As Me14G074000 transports both compound classes , we propose that this transporter may represent a transition phase where specificity for cyanogenic glucosides is partially lost in favor of glucosinolate transport . This suggests that glucosinolate transporters evolved from those of cyanogenic glucosides . If Me14G074000 represents such a ‘transition’ transporter between cyanogenic glucoside-specific to glucosinolate-specific transporter , we hypothesized that the genome of the cyanogenic glucoside-producing cassava also encodes a GTR-like NPF transporter that is specific for cyanogenic glucosides . We tested this hypothesis by measuring transport activity of the six closest homologs of Me14G074000 from cassava ( Figure 6B–C and Figure 6—figure supplement 2 ) . The closest homolog , Me14G074100 , appears truncated ( data not shown ) with only five transmembrane-spanning domains but was nevertheless included in our analysis . All six transporters were tested for transport activity in their native form . Additionally , we fused YFP to the C-terminus of each gene to validate expression . Native Me14G074100 , Me01G191900 , Me09G097200 , and Me17G124600 did not result in uptake of 4MTB , I3M , prunasin or linamarin . Among the YFP-tagged transporters , only Me14G074100 and Me09G097200 did not express in the oocytes ( Figure 6—figure supplement 1 ) , and hence we cannot conclude whether these two transporters are inactive . Uptake of prunasin was detected in oocytes expressing Me15G176100 at levels similar to or slightly lower than the suggested ‘transition’ transporter ( Me14G074000 ) ( Figure 6C ) . In contrast , Me15G180400 strongly over-accumulated prunasin and linamarin to more than 12 and 8 times the media level ( Figure 6C ) while uptake of both aliphatic and indole glucosinolates by this transporter was negligible ( Figure 6—figure supplement 3 ) . Thus , Me15G180400 is specific towards cyanogenic glucosides . We named Me15G180400 MANIHOT ESCULENTA CYANOGENIC GLUCOSIDE TRANSPORTER-1 ( MeCGTR1 ) and to the best of our knowledge , it represents the first identification of an importer of cyanogenic glucosides . TEVC electrophysiology assays showed that prunasin and linamarin induce negative currents in MeCGTR1-expressing oocytes ( Figure 6D ) . Kinetic analysis of MeCGTR1 showed that this protein transports prunasin and linamarin with a Km of 80 ± 7 µM and 262 ± 15 µM , respectively , measured at a membrane potential clamped to −60 mV ( Figure 6E–F ) . This indicates that MeCGTR1 is a high-affinity , cyanogenic glucoside-specific transporter and shows that it is capable of over-accumulating against a concentration gradient . The existence of MeCGTR1 supports our hypothesis that the dual-specific Me14G074000 represents a ‘transition’ transporter evolutionarily positioned between cyanogenic glucoside-specific ( MeCGTR1 ) and glucosinolate-specific transporters ( GTR1–3 and GTRL1–2 ) . The identification and close phylogenetic relationship of glucosinolate-specific , dual-specific and cyanogenic glucoside-specific transporters within the NPF supports that glucosinolate transporters evolved from cyanogenic glucoside NPF transporters . Most characterized members of the SLC15/PepT/POT/NPF family are symporters that function by an electrogenic proton-coupled transport mechanism ( Nour-Eldin et al . , 2012; Parker and Newstead , 2014; Fei et al . , 1999; Chen et al . , 1999; Steel et al . , 1997; Mackenzie et al . , 1996; Fei et al . , 1994; Doki et al . , 2013; Solcan et al . , 2012; Chiang et al . , 2004 ) , that is , symport of protons generates a net influx of positive charge that can be measured as a negative current by TEVC . Characterization of the transporters identified in this study by both LCMS- and TEVC-based transport assays enabled us to investigate the evolution of electrogenicity of glucosinolate and cyanogenic glucoside transporters . Previously , we showed that AtGTR1 and −2 mediated transport of 4MTB induces negative currents as a result of net inward movement of protons during transport ( Nour-Eldin et al . , 2012 ) . In this study , we show that negative currents are also induced by both AtGTR1 and −2 when exposed to I3M ( Figure 2B and Figure 2—figure supplement 2 ) . This indicates that AtGTR1 and −2 transport 4MTB and I3M – two negatively charged glucosinolates with different amino acid side chains- via a similar electrogenic transport mechanism . Similarly , the tested GTR1 ortholog from B . rapa ( BrGTR1 ) also induced negative currents when exposed to 4MTB or I3M ( Figure 5C ) . In comparison , AtGTR3 and the GTR3 ortholog from B . rapa ( BrGTR3 ) only induced currents when exposed to I3M ( Figure 5C ) . No detectable currents were induced by 4MTB in neither AtGTR3 nor BrGTR3 ( Figure 2A–B and Figure 5C ) . Thus , it appears that electrogenic transport of - and the ability to upconcentrate - 4MTB is a property that distinguishes the GTR1 clade from the GTR3 clade . Similarly , exposure to cyanogenic glucosides did not induce currents in the putative transition transporter , Me14G074000; ( Figure 6D ) . In comparison , MeCGTR1 induced negative currents when exposed to the non-charged prunasin and linamarin ( Figure 6D ) . This suggests that transport of these two cyanogenic glucosides by MeCGTR1 is coupled to a net influx of cationic species and that transport of cyanogenic glucosides by MeCGTR1 and Me14G074000 appears to differ with respect to electrogenicity . Identification of the phylogenetically more basal glucosinolate transporters in C . papaya and cassava allowed us to investigate when electrogenic glucosinolate transport may have evolved . From the GTR-like clade , the glucosinolate transporting CpGTRL1 and CpGTRL2 induced currents when exposed to 4MTB or I3M ( Figure 5C ) whereas Me14G074000 from cassava did not induce detectable currents ( Figure 5—figure supplement 3C ) . All three transporters were able to upconcentrate both glucosinolates against their respective concentration gradient ( Figure 5B ) . Thus , our data suggest that the glucosinolate transport mechanism first arose as a non-electrogenic mechanism that later evolved to become electrogenic . Moreover , electrogenic transport appears not to be a prerequisite for the ability to over-accumulate glucosinolates . Previously , we showed that negative currents induced by AtGTR1 and AtGTR2 when exposed to the negatively charged glucosinolates , reflect a glucosinolate to proton stoichiometry of 1 ≤ 2 ( Nour-Eldin et al . , 2012 ) . Substrate-dependent variation in transport coupling stoichiometry between substrate and protons has been shown to depend on the length of the oligo-peptide substrate for PepTSo ( Parker and Newstead , 2014 ) . The non-electrogenic transport by Me14G074000 could indicate a different proton to glucosinolate stoichiometry compared to that of the A . thaliana orthologs . For example , the lack of detectable currents may be caused by an equal amount of negative and positive charges moving across the membrane during the transport cycle . This would suggest that changes in transporter substrate specificity for a given substrate are accompanied by changes in transporter electrogenicity . However , we cannot exclude that the lack of currents for Me14G074000 is caused by currents below detection limits or that co-transport of other ions may be ´masking´ the coupled transport by Me14G074000 . Nevertheless , the absence of induced currents by Me14G074000 indicates that transport of glucosinolates became electrogenic after the divergence of cassava and C . papaya . We believe that these genes provide a suitable model system for future studies that will investigate if the observed changes in transporter electrogenicities are caused by changes in coupling stoichiometry . This will lead to a mechanistic understanding of how substrate specificity and coupling stoichiometries co-evolve within the NPF family . Based on our findings we propose a model for the evolutionary path of glucosinolate transporter substrate specificity in the NPF family ( Figure 7 ) . By tracking the evolution of GTR transporter specificity towards glucosinolates , we propose that a duplication event introduced permissive mutations in a high affinity , electrogenic transporter of the ancestral cyanogenic glucosides ( represented here by MeCGTR1 ) to generate a ‘transition’ transporter with broad , non-electrogenic glucosinolate specificity and non-electrogenic cyanogenic glucoside specificity ( represented by Me14G074000 ) ( Figure 7 ) . With the advent of glucosinolate biosynthesis and through further duplication and evolutionary divergence , our data suggests that the dual-specificity transporter lost its cyanogenic glucoside transport capacity and became a high affinity , electrogenic broad-specific glucosinolate transporter ( represented by BrGTR1 , AtGTRs , CpGTRL1 and CpGTRL2 ) . The retainment of Me14G074000 in the cassava genome indicates that it may fulfill an important role in transport of cyanogenic glucosides despite its inferior transport properties compared to Me15G180400 . Alternatively , its retainment could be explained by specificity towards other yet unidentified substrates . Presently , we can also not exclude that Me14G074000 represents an ancestral non-electrogenic , multi-specificity transporter , which through duplication and subfunctionalization gave rise to the electrogenic transporters with high affinity for cyanogenic glucosides ( represented by MeCGTR1 ) and later for glucosinolates ( represented by CpGTRL1 , CpGTRL2 , BrGTR1 and AtGTRs ) , respectively ( Figure 7 ) . Further subfunctionalization within the GTR clade led to the evolution of the GTR3 subclade identified as transporters with preference and high affinity for indole glucosinolate . Thus , the subfunctionalization within the GTRs from broad to narrow specificity is contrary to the evolutionary dynamics proposed previously for substrate-transport evolution , where progenitor transporters had a narrow substrate specificity that expanded during evolution to become increasingly broad ( Lionarons et al . , 2012 ) . The large plant NPF family is homologous to the SLC15/PepT/PTR/POT families in bacteria and animals ( Léran et al . , 2014; Daniel et al . , 2006 ) . Several bacterial POTs ( Doki et al . , 2013; Solcan et al . , 2012; Newstead et al . , 2011 ) and one plant NPF homolog ( AtNPF6 . 3 ) ( Parker and Newstead , 2014; Sun et al . , 2014 ) have been crystalized along with their substrates . Hence , it is possible to discuss the substrate specificities determined in the present study in a structural context by analysing the amino acid residues that are key for substrate interaction . We constructed an alignment comprising the sequence of AtNPF6 . 3 , the crystalized bacterial POTs and the glucosinolate and cyanogenic glucoside transporters investigated in this study ( Figure 8—figure supplement 1 ) . From the structural studies on bacterial POTs and NPF6 . 3 ( Doki et al . , 2013; Solcan et al . , 2012; Parker and Newstead , 2014; Sun et al . , 2014; Aduri et al . , 2015 ) , the key substrate interacting amino acid residues were compiled , numbered P1-P13 ( Table 1 ) and located in the alignment ( Figure 8—figure supplement 1 ) . In addition , we constructed homology models using the recent structure of NPF6 . 3 as a template and depicted amino acid positions P1-P13 within the models ( Figure 8 inserts ) . Analysis of AtGTR1 , AtGTR3 , Me14g074000 and MeCGTR1 homology models showed the P1-P13 residues to be exposed to the central substrate binding cavity of the transporters ( Figure 8 ) and therefore to constitute candidates for substrate specificity determining residues . Five residues ( P1-P3 , P8 and P13 ) were conserved in all genes in the alignment . These constitute the EXXE[R/K] motif that has an indispensable role in proton coupling ( Table 1 ) ( Solcan et al . , 2012; Aduri et al . , 2015; Jørgensen et al . , 2015 ) . This suggests that the proton coupling mechanism is conserved regardless of the substrate specificity of a NPF transporter . Amino acid positions P7 and P9 are conserved respectively as asparagine and threonine across the GTR and cyanogenic glucoside transporters , whereas they are glycine and asparagine residues in POT transporters and phenylalanine and asparagine residues in NPF6 . 3 . The only moiety shared between glucosinolates and cyanogenic glucosides is the glucose moiety . Thus , P7 and P9 could be involved in the interaction with the glucose moiety ( Table 1 , Figure 1 and Figure 6A ) . Amino acid positions P10-P12 are not conserved across the GTRs , NPF6 . 3 and the POT transporters and the mutational pattern does not correlate to the changes we have seen in substrate specificity . Consequently , the role of these residues remains unclear . In contrast , amino acid positions P4 , P5 and P6 ( Figure 8 inserts ) show a conservation pattern that is consistent with the changes in substrate specificity shown in this study . This indicates that position P4 , P5 and P6 may contribute to determining the substrate specificity of glucosinolate and cyanogenic glucoside transporters . Despite recent advances in understanding substrate specificity of peptide transporters ( as outlined above ) , nothing is currently known about amino acid changes that determine transporter substrate specificity from an evolutionary perspective . Our work provides a framework for future studies to determine the amino acid changes that leads to substrate specificity changes during evolution of metabolite transporters . With the dawning of cellular life , primitive membrane structures leading to today’s complex phospholipid membranes necessitated membrane proteins to facilitate movement of structurally diverse compounds across cellular membranes ( Mansy et al . , 2008 ) . Towards understanding the evolutionary paths that lead to new transporter substrate specificities , we show that before a substrate emerges , transporter specificity for the substrate may be present in transporters of chemically similar , more ancient substrates . As new substrates emerges ( in this case glucosinolates ) , gene duplications allows such multifunctional transporters to diversify through subfunctionalization into transporters with greater specificity for the new substrate . Thus , our findings support one model ( Bridgham et al . , 2006 ) to the problem posed by the classical evolution model – about how a new function ( here transporter substrate specificity ) can be selected for unless the substrate is there . We propose that redundant ancestral transporters created by gene duplication remain active due to a multifunctional specificity . When the new substrate emerges these ancestral transporters can be recruited and evolve into transporters with greater specificity for the new compound . Moreover , from a mechanistic perspective , our data suggests that the evolution of new substrate specificities in coupled secondary transporters is accompanied by changes in the electrogenic properties of the transport mechanism . Unraveling the structural determinants underpinning stoichiometry and binding of substrate constitutes a new frontier for understanding the birth and development of new transporter substrate specificities at the molecular level . This not only applies to the universal NPF transporters , including their drug-delivering mammalian counterparts ( Brandsch , 2013 ) , but for transporters in general . From an agro-biotech perspective , the identification of a cyanogenic glucoside transporter with high apparent affinity supports a prominent role for NPF transporters in specialized metabolism ( Nour-Eldin et al . , 2012; Nour-Eldin and Halkier , 2013 ) and opens new possibilities for controlling cyanogenic glucoside content in edible parts of crops such as bitter almond ( Dicenta et al . , 2002 ) , barley ( Erb et al . , 1979 ) and cassava ( McMohan et al . , 1995 ) through transport engineering strategies ( Nour-Eldin and Halkier , 2013 ) .
The genes cloned and tested in this study are named as follows: ( BrH02396 ) , BrF01711 , CpGTRL2 ( Phytozome ID: evm . TU . supercontig_17 . 190 ) , CpGTRL1 ( Phytozome ID: evm . model . supercontig_17 . 189 ) , Cp17 . 188 ( Phytozome ID: evm . model . supercontig_17 . 188 ) , Tc1EG013796 ( Phytozome ID: Thecc1EG013796 ) , Me14G074000 ( Phytozome ID: Manes . 14G074000 , cassava4 . 1_004026m ) , Me15G176100 ( Phytozome ID: Manes . 15G176100 , cassava4 . 1_004213m ) , Me09G097200 ( Phytozome ID: Manes . 09G097200 , cassava4 . 1_034015m ) , Me01G191900 ( Phytozome ID: Manes . 01G191900 , cassava4 . 1_025742m ) , Me14G074100 ( Phytozome ID: Manes . 14G074100 cassava4 . 1_034466m ) , MeCGTR1 ( Phytozome ID: Manes . 15G180400 , cassava4 . 1_004125m ) and Me17G124600 ( Phytozome ID: Manes . 17G124600 , cassava4 . 1_029616m ) . Design and direct assembly of synthesized uracil-containing non-clonal DNA fragments into vectors by USER cloning is described in more detail at Bio-protocol ( Jørgensen et al . , 2017b ) . All coding sequences were codon optimized for expression in X . laevis ( NCBI Taxon: 8355 ) oocytes and synthesized as linear uracil containing DNA fragments ( uStrings ) by ThermoFisher Scientific Geneart . Each coding sequence was surrounded by the 8 bp USER tails that enable insertion into the USER compatible X . laevis expression vector pNB1u ( Geu-Flores et al . , 2007; Nour-Eldin et al . , 2006 ) . Each fragment contained a uracil at the appropriate location in each USER tail . The uracil was incorporated during synthesis . Thus , uStrings are mixed directly with the digested pNB1u vector without prior PCR amplification with uracil containing primers . Briefly , each uString was diluted to 100 ng/ul in H20 . The USER-compatible pNB1u X . laevis oocyte expression vector was digested with PacI/Nt . BbvCI overnight , gel purified and diluted to a concentration of ~50 ng/ul ( as previously described [Nour-Eldin et al . , 2006; MacAulay et al . , 2001] ) . For the USER reaction , 100 ng uStrings was mixed with 50 ng digested pNB1u , 1 unit USER enzyme ( NEB-M5505S ) , 2 µl 5xPCR reaction buffer and 5 µl H20 . The reaction was incubated at 37°C for 25 min , followed by 25 min at room temperature . The reaction mixture was used to transform chemically competent E . coli cells , plated on carbinicilin-containing LB plates . Selected colonies were grown overnight and extracted plasmids sequenced . All uStrings were inserted successfully into the pNB1u vector and out of the 13 genes synthesized and cloned , we had to sequence a second colony for only one of the genes . The fidelity and efficiency of cloning uStrings directly by USER cloning is satisfactory . For fluorophore tagging , coding sequences were PCR amplified from the expression constructs without the stop codon using uracil containing primers ( see Materials and methods list of USER primers ) . The PCR fragments were USER cloned ( as described previously [Nour-Eldin et al . , 2006] into an oocyte expression vector ( pNB1u variant , pLIFE22 [Jørgensen et al . , 2015] ) that translationally fuses the inserted coding sequence to a C-terminal YFP fluorophore , which is contained in the vector . PrimerNameSequenceForwardCpGTRL1GGCTTAAUATGGAAAGGGCTGCCATGGCReverse no stopCpGTRL1GGTTTAAUCCTCTGGACTCTTCGTTCACTTCGForwardThecc1EG013796GGCTTAAUATGGAAAAGAACGACAAAGAAGCCReverse no stopThecc1EG013796GGTTTAAUCCAACGAAGCTCTTGTCGCTCTForwardBrGTR3GGCTTAAUATGGAAGTGGAAAAGACCCAGGAAReverse no stopBrGTR3GGTTTAAUCCAACGGACACCTTGTCGAACTCGForwardMe15g176100GGCTTAAUATGGAAGATAAGGAAGAGAAGTCCReverse no stopMe15g176100GGTTTAAUCCCACAAGGTGTTTCTGAGACTGCTGForwardMe14g074100GGCTTAAUATGGAAGTGGAACAGAGCGTGGReverse no stopMe14g074100GGTTTAAUCCCTGCACCACTTCCAGAATCTTTGTForwardMe15g18400/MeCGTR1GGCTTAAUATGGAAAACGGCAACGATCACGReverse no stopMe15g18400/MeCGTR1GGTTTAAUCCCACGTGGTGCTTCACGCTAGForwardMe17g124600GGCTTAAUATGGAAAACAAAAAGCAGGAAACAReverse no stopMe17g124600GGTTTAAUCCCAGGTCGCTTGGGATGAAAGACForwardMe09g097200GGCTTAAUATGGAAAACATGATTATCGCCAGCReverse no stopMe09g097200GGTTTAAUCCGGCTGTAGCCTTCAGTTCCAGAForwardBrGTR1GGCTTAAUATGGAAAGAAAGCCCTTCGAGGTReverse no stopBrGTR1GGTTTAAUCCAACGCTGTTCTTAGCCTGCTTForwardMe14g074000GGCTTAAUATGGCCACAGGCGAGACAATCReverse no stopMe14g074000GGTTTAAUCCGGCCTTGATTGGCTTAACCTGCForwardCpGTRL2GGCTTAAUATGGAAATGGACGGCAAAGAGCReverse no stopCpGTRL2GGTTTAAUCCAACGTGGATGTTCTGCTTTTTCTTForwardCp17 . 188GGCTTAAUATGGCCTTCCTGCTGACCGReverse no stopCp17 . 188GGTTTAAUCCGATATCGCTCTGCTTGGTGCTForwardAtGTR1GGCTTAAUATGAAGAGCAGAGTCATTReverse no stopAtGTR1GGTTTAAUCCGACAGAGTTCTTGTCForwardAtGTR3GGCTTAAUATGGAGGTTGAGAAGACAGAGAAGReverse no stopAtGTR3GGTTTAAUCCCACTGACACCTTATCAAACTCAGC Oocyte bioimaging was performed essentially as previously described ( Geiger et al . , 2011 ) , with the addition that oocytes expressing YFP-tagged transporters were mounted on a glass slide and a Kulori ( 90 mM NaCl , 1 mM KCl , 1 mM MgCl2 , 10 mM MES adjusted to pH7 . 4 ) solution with 20 µM FM4-64fx was added 1 min prior to bioimaging by confocal scanning microscopy using a SPX5-X Point-scanning Confocal from Leica Microsystems . X . laevis oocytes ( stages V-VI ) were purchased as defolliculated oocytes ( stages V-VI ) from Ecocyte Biosciences ( Germany ) . Injection of 50 nl cRNA ( 500 ng/µl ) into X . laevis oocytes was done using a Drummond NANOJECT II ( Drummond scientific company , Bromall Pennsylvania ) . Oocytes were incubated for 3 days at 17°C in Kulori ( 90 mM NaCl , 1 mM KCl , 1 mM MgCl2 , 10 mM MES ) pH7 . 4 prior to assaying . NPF homologs ( also called SLC15/PepT/PTR/POT [Léran et al . , 2014; Daniel et al . , 2006] ) from Arabidopsis thaliana ( At ) , Brassica rapa ( Br ) , Carica papaya ( Cp ) , Theobroma cacao ( Tc ) , Manihot esculenta ( Me ) , Glycine max ( Gm ) , Gossypium raimondii ( Gr ) , Medicago truncatula ( Mt ) and Solanum lycopersicum ( Sl ) were retrieved from phytozome ( http://www . phytozome . net ) by searching for sequences classified as oligopeptide transporters ( PFAM:PF00854 and Panther: PTHR11654 ) . To remove pseudogenes , we predicted the number of transmembrane helices using TMHMM server V . 2 . 0 ( Sonnhammer et al . , 1998 ) and removed sequences with <6 transmembrane helices and fewer than 300 amino acids . Genes were renamed according to the following guidelines . evm . model . supercontig_139 . 55 from C . papaya , was renamed to ‘Cp139 . 55’ . Solyc10g024490 . 1 from S . lycopersicum was renamed to Sl10g024490 . 1 . Thecc1EG035998 from T . cacao was renamed to Tc1EG035998 . Glyma . 18G097800 from G . max was renamed to Gm18G097800 . Brara . C02073 from B . rapa was renamed to BrC02073 . Medtr4g107510 from M . truncatula was renamed to Mt4g107510 . Manes . 16G072300 from M . esculenta was renamed to Me16G072300 . Gorai . 012G121800 from G . raimondii was renamed to Gr012G121800 . Sequences were aligned using MUSCLE ( Edgar , 2004 ) with a gap open penalty of −2 . 9 , gap extend of 0 and hydrophobicity multiplier of 1 . 2 . Poorly aligned regions were trimmed manually . Prottest v3 . 4 . 2 ( Darriba et al . , 2011 ) was used with final multiple sequence alignments to identify the appropriate LG-based phylogenetic models to use for subsequent work . The best fit ( for phylogenies in Figure 1A , Figure 1—figure supplement 1 and Figure 6—figure supplement 2 ) was LG+I+G+F that use a general amino acid replacement matrix ( Le and Gascuel , 2008 ) with a proportion of invariable sites ( +I ) ( Reeves , 1992 ) , a gamma distribution for modelling the rate heterogeneity ( +G ) ( Yang , 1993 ) , and empirical amino acid frequencies ( +F ) ( Cao et al . , 1994 ) . The best fit ( for phylogenies in Figure 5A , Figure 6B and Figure 5—figure supplement 1 ) was LG+G+F . Bayesian inference trees were calculated using MrBayes 3 . 2 . 6 ( Huelsenbeck and Ronquist , 2001 ) until convergence was reached ( ‘average standard deviation of split frequencies' <0 . 01 ) . The temperature heating parameter was set to 0 . 05 ( temp = 0 . 05 ) to increase the chain swap acceptance rates , thereby reducing the chances of Markov chains to get stuck at local high-probability peaks . Burn-in was set to 25% ( burninfrac = 0 . 25 ) and the number of Markov chains was set to 8 ( nchains = 8 ) . Maximum likelihood trees were produced with RAxML 8 . 2 . 3 using the LG PROTGAMMA model and 500 bootstrap replicates ( Stamatakis , 2014 ) . RAxML bootstrap values were portrayed on the MrBayes generated consensus tree . NRT1 . 1 from Chlamydomonas reinhardtii ( Phytozome ID: Cre04 . g224700 ) was used as the out-group . All analyses were run in MPI via the CIPRES SCIENCE GATEWAY ( Miller et al . , 2011 ) at the San Diego Supercomputer Center ( SDSC ) . Trees were visualized in figtree ( http://tree . bio . ed . ac . uk/software/figtree/ ) and annotated with Adobe Illustrator . ‘Biological replicates’ denote replicated measurements using different biological cases , whereas ‘technical replicates’ use the same biological cases . Through pilot experiments average variation between biological replicates have been determined . Sample sizes for this study were decided upon as the best compromise between average power and experimental constraints . Uptake assays in Xenopus laevis oocytes using liquid chromatography–mass spectrometry to detect transport activity is described in more detail at Bio-protocol ( Jørgensen et al . , 2017a ) . X . laevis uptake assays were carried out as follows: Oocytes were preincubated in Kulori pH 5 for 5 min , transferred to Kulori pH five with substrate for 60 min incubation , followed by four washes and transferred to Eppendorf tubes ( one oocyte per tube ) . Excess washing buffer was removed and oocytes were busted in 50 μl of 50% MeOH ( with sinigrin as internal standard ) and the homogenate was left in the freezer for 2 hr to precipitate proteins . This was followed by centrifugation at 20 , 000 x g for 15 min to pellet remaining proteins . The supernatant was transferred to new tubes and diluted with 60 μl H2O . The diluted samples were filtered through a 0 . 45 μm PVDF based filter plate ( MSHVN4550 , Merck Millipore ) and subsequently analyzed by analytical Liquid Chromatography – Mass Spectrometry . 4-methylsulfinylbutyl glucosinolate ( 4MTB ) and 3-indolylmethylglucosinolate ( I3M ) were obtained from C2 Bioengineering ( http://www . glucosinolates . com/ ) and CFM Oskar Tropitzsch GmbH , Marktredwitz ( http://www . cfmot . de/ ) , respectively . Cyanogenic glucoside prunasin was synthesized by MSM as previously described ( Møller et al . , 2016 ) . Cyanogenic glucoside linamarin was purchased from Santa cruz biotechnology . ESI-LC-MS analysis of desulfo glucosinolates from X . laevis uptake assays were performed as described before ( Nour-Eldin et al . , 2012 ) . Extracts from uptake assays ( see above ) were directly analyzed by LC-MS/MS . Chromatography was performed on an Advance UHPLC system ( Bruker , Bremen , Germany ) . Separation was achieved on a Kinetex 1 . 7u XB-C18 column ( 100 × 2 . 1 mm , 1 . 7 μm , 100 Å , Phenomenex , Torrance , CA , USA ) . Formic acid ( 0 . 05% ) in water and acetonitrile ( supplied with 0 . 05% formic acid ) were employed as mobile phases A and B respectively . The elution profile was: 0–0 . 2 min , 2% B; 0 . 2–1 . 8 min , 2–30% B; 1 . 8–2 . 5 min 30–100% B , 2 . 5–2 . 8 min 100% B; 2 . 8–2 . 9 min 100–2% B and 2 . 9–4 . 0 min 2% B . The mobile phase flow rate was 400 μl min−1 . The column temperature was maintained at 40°C . The liquid chromatography was coupled to an EVOQ EliteTripleQuad mass spectrometer ( Bruker , Bremen , Germany ) equipped with an electrospray ion source ( ESI ) operated in combined positive and negative ionization mode . The instrument parameters were optimized by infusion experiments with pure standards . The ion spray voltage was maintained at +5000 V or −4000 V for cyanogenic glucoside and glucosinolate analysis , respectively . Cone temperature was set to 300°C and cone gas to 20 psi . Heated probe temperature was set to 180°C and probe gas flow to 50 psi . Nebulizing gas was set to 60 psi and collision gas to 1 . 6 mTorr . Nitrogen was used as probe and nebulizing gas and argon as collision gas . Active exhaust was constantly on . Multiple reaction monitoring ( MRM ) was used to monitor analyte parent ion → product ion transitions: MRM transitions were chosen based on direct infusion experiments . Detailed values for mass transitions is found in Materials and methods list of primers . MRM transitions for intact glucosinolates and cyanogenic glucosides determined by LC-MS/MS QQuantifier ion used for quantification of the compounds . Additional MRM transitions were used for compound identification . IS = internal Standard . CompoundQ1Q3CE [eV]Internal standardResponse factorSIN ( Sinigrin . 2-propenyl-GLS ) = IS358 . 097 . 0Q22n . a . n . a . 358 . 075 . 030358 . 0259 . 0204MTB420 . 097 . 0Q23Sinigrin0 . 99420 . 075 . 030420 . 0259 . 023I3M447 . 197 . 0Q10Sinigrin13 . 34447 . 1259 . 010447 . 1205 . 010Linamarin248 . 285 . 2Q−15Sinigrin163 . 9248 . 297 . 319248 . 2163 . 1−5Prunasin296 . 1163 . 1Q−4Sinigrin97 . 7296 . 185 . 2−17296 . 197 . 3−21 Both Q1 and Q3 quadrupoles were maintained at unit resolution . Bruker MS Workstation software ( Version 8 . 2 , Bruker , Bremen , Germany ) was used for data acquisition and processing . Linearity in ionization efficiencies was verified by analyzing dilution series of standard mixtures . Quantification of all compounds was achieved by use of sinigrin as internal standard . The concentration of imported substrate was calculated based on previous reports determining the water content of oocytes to be ~70% of total volume ( de Laat et al . , 1974 ) and an oocyte diameter of 1 . 5 mm . Assuming this , the oocyte cytosolic volume was estimated to be 1 μl allowing us to calculate the up-concentration of substrate . Nitrate uptake assays were carried out as follows: Oocytes were preincubated in Kulori pH 5 for 5 min , transferred to Kulori pH 5 with 15N-labelled KNO3 ( Sigma Aldrich , 335134 ) at the indicated concentration for 60 min . Subsequently , oocytes were washed 4 times in H2O and transferred to tin capsules prior to stable isotope analysis by IRMS ( Isotope Ratio Mass Spectrometry ) . Stable isotope ratio analysis of nitrogen were conducted as described by Laursen et al . ( 2013 ) . In brief , analysis were conducted using a Europa Scientific ANCA-SL Elemental Analyser coupled to a Europa Scientific 20–20 Tracermass mass spectrometer ( Sercon Ltd . , Crewe , UK ) . Quality control ( accuracy and precision ) was performed by analysis of standards and certified reference materials from the International Atomic Energy Agency , IAEA , Vienna , Austria and Iso-Analytical Limited , IA , Crewe , UK . All measurements were performed with a Two Electrode Voltage-Clamp system ( TEVC ) composed of an NPI TEC-03X amplifier ( NPI electronic GmbH , Germany ) connected to a PC with pCLAMP10 software ( Molecular devices , USA ) via an Axon Digidata 1440a digitizer ( Molecular devices , USA ) . Oocytes were placed in the recording chamber and perfused with a standard Kulori-based solution ( 90 mM NaCl , 1 mM KCl , 1 mM CaCl2 , 1 mM MgCl2 , 1 mM LaCl3 and 10 mM MES pH 5 ) . TEVC data was analysed in excel after extraction from pCLAMP10 software as a Microsoft Excel compatible worksheet . Substrate-dependent currents were calculated by subtracting currents before addition of substrate from currents after addition of substrate . Visualization and curve fitting to the Michaelis-Menten equation ( Equation 1 ) to calculate the apparent Km value was done using SigmaPlot version 12 . 5/13 . 0 ( Systat software , USA ) . Equation 1 - Michaelis-Menten equation . I is the current , Imax is the maximal current achieved by the transporter at saturating concentrations of substrate . ( 1 ) I=Imax*[ substrate ][ substrate ]+Km SigmaPlot version 12 . 3 ( Systat software , USA ) was used for statistical analysis and data plotting . Arabidopsis thaliana ecotype Columbia-0 ( Col-0 ) ( NCBI Taxon: 3702 ) lines and three insertion mutants gtr1 ( SAIL_801_G03 ) , gtr2 ( SAIL_20_B07 ) and nrt1 . 9–2/gtr3 ( GK-099B01 ) were obtained from NASC . To construct double and triple mutants , gtr3 homozygous were crossed to a previously characterized and published gtr1 gtr2 dko ( Nour-Eldin et al . , 2012 ) . F1 progeny of those crosses were all phenotypically normal . The resulting F2 progeny ( 160 plants ) were screened by PCR , and homozygous mutants , ( gtr1 gtr3 , gtr2 , gtr3 ) were obtained . Seeds from self-pollination of gtr1+/-gtr2 gtr3 plants were collected and allowed to self-pollinate . The following F3 progeny were screened by PCR and homozygous gtr1 gtr2 gtr3 mutants were identified ( Figure 4—figure supplement 3 ) . Plants used for rosette and root analysis were grown from sterilized A . thaliana seeds put onto 0 . 5 mL PCR tubes , which had been filled with agar ( 1% ( w/v ) sucrose ) and cut at the bottom . A total of 48 tubes were placed into a yellow pipette tip box filled with nutrient solution ( one-half strength Murashige and Skoog basal medium ) . Afterwards , the box was sealed with and incubated in a growth chamber under cultivated at 12 hr days with 70% relative humidity , and a light intensity of 100 mE . Upon root emergence into the growth media , PCR tubes containing the seedlings were transferred on to a perforated screw cap of 50 mL Falcon tubes . The cap was screwed onto the tube and the tubebottom was cut off to allow the root to grow into the media . A total of 30 bottles were arrayed in a closed plastic box filled with MS media . Air was pumped into the media by four tubes from holes at four corners of the box . The box was put into the same growth chamber and plant material was harvested for glucosinolate analysis after two weeks . For micro-grafting , seeds were surface-sterilized by washing in 70% ( v/v ) ethanol containing 0 . 05% ( v/v ) Triton X100 for 5 min followed by washing in water , left on sterile filter paper , and sown on half-strength MS agar plates . The plates were cold-stratified for 2 days followed by vertical growth for 3 to 5 d under long-day conditions ( light: 16 hr , 20°C; darkness: 8 hr , 16°C ) . Micro-grafting of A . thaliana seedlings was performed in a laminar flow cabinet using a dissection microscope as described ( Andersen et al . , 2014 ) . Briefly , Arabidopsis seedlings grown on MS-containing agar ( without sugar ) for 4 days were transferred to a sterile one layer thick wet nitrocellulose filter ( Whatman NC 45 ST ) and two layers of filter paper in a sterile petri dish . The cotyledons were removed and incisions were made on the hypocotyl close to the shoot using a sapphire knife . The root stocks and scions of seedlings were joined using sterile forceps . The plates were sealed using Micropore tape ( 3M ) and incubated vertically under long-day conditions ( light: 16 hr , 20°C; darkness: 8 hr , 16°C ) . Successfully joined seedlings were transferred to MS agar and kept under long-day conditions until the age of 3 weeks and analyzed by LC-MS . Glucosinolates were analyzed as desulfo-glucosinolates by UHPLC/TripleQuad-MS . Chromatography was performed on an Advance UHPLCTM system ( Bruker , Bremen , Germany ) equipped with a C-18 reversed phase column ( Kinetex 1 . 7 u XB-C18 , 10 cm x 2 . 1 mm , 1 . 7 µm particle size , Phenomenex , Torrance , CA , USA ) by using a 0 . 05% formic acid in water ( v/v ) ( solvent A ) −0 . 05% formic acid in acetonitrile ( v/v ) ( solvent B ) gradient at a flow rate of 0 . 4 ml*min−1 . The column temperature was maintained at 40°C . The gradient applied was as follows: 2% B ( 0 . 5 min ) , 2–30% ( 0 . 7 min ) , 30–100% ( 0 . 8 min ) , 100% B ( 0 . 5 min ) , 100–2% B ( 0 . 1 min ) , and 2% B ( 1 . 4 min ) . The liquid chromatography was coupled to an EVOQ Elite TripleQuad mass spectrometer ( Bruker , Bremen , Germany ) equipped with an electrospray ion source ( ESI ) operated in positive ionization mode . The ion spray voltage was maintained at +3500 V . Cone temperature was set to 300°C and cone gas to 20 psi ( arbitrary units ) . Heated probe temperature was set to 400°C and probe gas flow set to 40 psi . Nebulizing gas was set to 60 psi and collision gas to 1 . 6 mTorr . Desulfo-glucosinolates were monitored based on the following Multiple reaction monitoring ( MRM ) analyte parent ion → product ion transitions [Collision energy]: 3-methylthiopropyl ( 3mtp , m/z 328 → 166 [5V] ) ; 3-methylsulfinyl ( 3msp , m/z 344 → 182 [10V] ) ; 2-propenyl ( 2-prop , m/z 280→ 118 [5V] ) ; 3-hydroxypropyl ( 3ohp , m/z 298 → 118 [15V] ) ; 3-benzoyloxy ( 3bzo , m/z 402 collision gas to 1 . 6 mTorr . Desulfo-glucosinolates were monitored based on the following Multiple reaction monitoring ( MRM ) analym/z 294 → 132 [15V] ) ; ( R/S ) −2-hydroxy-3-butenyl , m/z 310 → 130 [15V]; 4-hydroxybutyl ( 4ohb , m/z 312 ision gas to 1 . 6 mTorr . Desulfo-glucosinolates were monitored based on the following Multiple reaction monitoring ( MRM ) analym/z 294 → 132 [15V] ) ; ( R/S ) −2-h-methylsulfinylheptyl ( 7msh , m/z 400 → 238 [7V] ) ; 8-methylthiooctyl ( 8mto , m/z 398 → 236 [5V] ) ; 8-methylsulfinyloctyl ( 8mso , m/z 414 → 252 [5V] ) ; indol-3-ylmethyl ( I3M , m/z 369 → 207 [10V] ) ; N-methoxy-indol-3-ylmethyl ( NMOI3M , m/z 399 → 237 [10V] ) ; 4-methoxy-indol-3-ylmethyl ( 4MOI3M , m/z 399 → 237 [10V] ) ; p-hydroxybenzyl ( pOHB , m/z 346 346 6 OHB , m/z 346 r . Desulfo-glucosinolates were monitored based on the following Multiple reaction monitoring ( MRM ) relative to the internal standard pOHB calculated from standard curves in control extracts . Homology models for AtGTR1 , AtGTR3 , Me14G074000 and MeCGTR1 was build using NPF6 . 3 ( PDB: 4OH3 ) as template ( Sun et al . , 2014 ) . Transporter homology models were built and optimized using Prime ( Jacobson et al . , 2004 ) included in the Schrödinger suite ( www . schrodinger . com ) . All homology models were validated using PROCHECK ( Laskowski et al . , 1993 ) . Homology models were embedded into a pre-equilibrated phosphatidyl oleoyl phosphatidylcholine ( POPC ) bilayer in a periodic boundary condition box with pre-equilibrated Simple Point Charge ( SPC ) water molecules in addition to Na+ and Cl- ions corresponding to a 150 mM buffer . Each system was subjected to a conjugate gradient energy minimization and relaxed by short molecular dynamics simulations ( MDs ) using the default ‘Relax model system’ protocol implemented in Desmond ( Bowers et al . , 2006 ) followed by 20 ns of MDs with periodic boundary conditions . A restriction was applied to the secondary structure of the transporters using a spring constant force of 0 . 5 kcal × mol−1 × Å−2 . The simulation temperature was set to 300K , and both temperature and pressure were kept constant during the MDs ( NPT ensemble simulation ) using the Nose-Hoover chain thermostat method ( Martyna et al . , 1992 ) and the Martyna-Tobias-Klein barostat method ( Martyna et al . , 1994 ) . Coordinates were stored every 2 fs . The MDs were run on a GPU computing cluster at the University of Talca , Chile , using 1 GPU GeForce GTX 980 for each simulation . The root-mean square deviations ( RMSD ) of the position for all backbone atoms of the models from their initial configuration as a function of simulation time are illustrated Figure 9 . All models were equilibrated after 4 ns of MDs ( except AtGTR3 , which reached equilibrium around 7 ns ) . The RMSD values remain within 4 Å for all models , demonstrating the conformational stability of the transporter structures . To determine the intracellular cavity/channel we used 3V ( Voss and Gerstein , 2010 ) via the web interface found at http://3vee . molmovdb . org . VMD ( Humphrey et al . , 1996 ) was used for visualizing and displaying homology models and cavity/channel . | All living cells are surrounded by membranes that protect them from the external environment . The membrane contains proteins called transporters , which move nutrients and other molecules ( known as substrates ) across the membrane . A variety of transporters have evolved to move the hundreds of thousands of different substrates found in nature . Plant cells make many different compounds to protect themselves from pests and diseases . A group of transporters known as the NPF family move some of these compounds across the cells outer membrane . The types of substrates they transport vary in different plants . In cassava , for example , NPF transporters move compounds called cyanogenic glucosides , which are poisonous to humans and other animals . On the other hand , NPF transporters in another plant called Arabidopsis thaliana can move bitter-tasting compounds called glucosinolates . The process that makes glucosinolates in plants evolved from the process that makes cyanogenic glucosides . Can transporters evolve the ability to move a new substrate before or after that substrate first appears ? To answer this question , Jørgensen et al . studied the NPF family in A . thaliana , cassava and another plant called papaya that makes both cyanogenic glucosides and glucosinolates . The experiments suggest that NPF transporters able to move both cyanogenic glucosides and glucosinolates evolved before plants evolved the ability to make glucosinolates . Later in evolution , these multi-specific transporters specialized to only move glucosinolates . Jørgensen et al . also show that early glucosinolate transporters could move a broad variety of glucosinolates but later evolved to only transport particular types . These findings show how transporters and the processes that make compounds in cells may evolve together . A future challenge will be to understand the molecular changes in a transporter that make it specific for a certain substrate . This may help researchers to develop new ways of controlling the amount of toxic compounds in crops we eat by manipulating how the compounds are transported . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"biochemistry",
"and",
"chemical",
"biology"
] | 2017 | Origin and evolution of transporter substrate specificity within the NPF family |
Microorganisms often exhibit a history-dependent phenotypic response after exposure to a stimulus which can be imperative for proper function . However , cells frequently experience unexpected environmental perturbations that might induce phenotypic switching . How cells maintain phenotypic states in the face of environmental fluctuations remains an open question . Here , we use environmental perturbations to characterize the resilience of phenotypic states in a synthetic gene network near a critical transition . We find that far from the critical transition an environmental perturbation may induce little to no phenotypic switching , whereas close to the critical transition the same perturbation can cause many cells to switch phenotypic states . This loss of resilience was observed for perturbations that interact directly with the gene circuit as well as for a variety of generic perturbations-such as salt , ethanol , or temperature shocks-that alter the state of the cell more broadly . We obtain qualitatively similar findings in natural gene circuits , such as the yeast GAL network . Our findings illustrate how phenotypic memory can become destabilized by environmental variability near a critical transition .
Microbes such as yeast and bacteria often adopt a specific phenotype in response to an environmental stimulus , and in many cases this phenotype can be retained even after the stimulus has been removed ( Ozbudak et al . , 2004; Acar et al . , 2005; Hu et al . , 2012; Williams et al . , 2013 ) . Stably storing such a phenotype can be critical for survival during adverse environmental changes ( Gasch et al . , 2000; Gasch and Werner-Washburne , 2002; Chen et al . , 2003; Brauer et al . , 2008 ) . Several well-studied natural gene networks can be found in multiple phenotypic states depending on the history that the cells have experienced , notably including the yeast galactose utilization network and sporulation commitment in Bacillus subtilis ( Acar et al . , 2005; Igoshin et al . , 2006 ) . Additionally , synthetic biology aims to create de novo phenotypic memory storage devices , which could be used in applications such as tracking dynamics of the gut microbiome ( Ajo-Franklin et al . , 2007; Kotula et al . , 2014 ) . In spite of the importance of phenotypic memory in natural and synthetic gene circuits , little is known about how environmental perturbations might disrupt a cell's ability to stably adopt a specific phenotype . Given that phenotypic memory often results from feedback loops within the cell ( Ferrell and Machleder , 1998; Pomerening et al . , 2003; Xiong and Ferrell , 2003; Angeli et al . , 2004; Ozbudak et al . , 2004 ) , it is possible that this phenotypic memory will display characteristics of other complex systems that exhibit bistability , memory , and associated critical transitions or ‘tipping points’ that lead to sudden changes in the state of the system in response to small changes in the environment . A phenomenon that may be especially relevant to cellular memory is that complex systems near a critical transition experience a loss of resilience of their stable states to external perturbations ( Dai et al . , 2012 ) . In particular , a system far from a critical transition may return to its original state following a perturbation , whereas closer to the critical transition the same environmental perturbation may cause the system to switch to an alternative stable state ( e . g . , collapse of a population [Dai et al . , 2012] ) . The dynamics in the vicinity of the critical point are very slow on both sides of the critical point . Therefore , it is worth noting that a short lived perturbation that pushes the system past the critical point may also not cause switching if the perturbation is short enough , if it pushes it to a position that is close to the critical point , or both . This aforementioned loss of resilience near a critical transition results from a shrinking basin of attraction in the stability landscape ( Scheffer et al . , 2009 ) . In the context of a gene network , this loss of resilience would manifest as an increasing sensitivity of phenotypic memory against environmental perturbations approaching the environmental condition in which cells would ( deterministically ) switch to a different phenotype . Our initial goal was to observe memory in a model gene circuit and then characterize the resilience of the phenotypic state against perturbations in the extracellular environment .
To study this predicted loss of resilience , we first employed a synthetic genetic switch in budding yeast that has previously been shown to exhibit hysteresis and bistability ( Gardner et al . , 2000; Blake et al . , 2006; Ellis et al . , 2009; Wu et al . , 2013 ) . This toggle switch is composed of two mutually inhibitory transcription factors , LacI and TetR ( Figure 1A ) . To enable tracking of the state of the cell , different color fluorescent proteins are expressed depending upon which of these transcriptional factors is highly expressed ( mCherry and eGFP , hereafter referred to as ‘RFP’ and ‘GFP’ ) . As external knobs to control the state of the cell , the inducer IPTG modulates the strength of repression of LacI , whereas ATc modulates the strength of repression of TetR . 10 . 7554/eLife . 07935 . 003Figure 1 . A toggle switch in yeast exhibits hysteresis and bistability . ( A ) A toggle switch consists of two mutually inhibitory transcription factors , two fluorescent readouts of the system state , and two small molecule inhibitors of the transcription factors . ( B ) Following growth in one of two histories , cells are then diluted into a range of ATc concentrations and propagated in culture for several days . Histogram counts are binned logarithmically . ( C ) The intensity of GFP fluorescence is plotted as a function of ( ATc ) for 11 different conditions for the high GFP history ( green triangles ) and high RFP history ( red circles ) after 92 hr of growth . The distributions are offset for ease of viewing . 20 , 000 events are collected , and then a narrow gate is drawn to select several hundred cells of roughly equal size . From this narrow gate , 50 cells at random are plotted . The region of memory is shaded in yellow . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 00310 . 7554/eLife . 07935 . 004Figure 1—figure supplement 1 . Fraction switched after 92 hr of growth , a proxy for the instability of the state , increases approaching a phenotypic switch . Yeast cells were pre-grown in the GFP state ( green triangles ) or RFP state ( red circles ) and then transferred to a range of ATc concentrations . After 92 hr of growth , the fraction of cells that have switched from their history state to the alternative state is plotted vs ATc . Error bars represent the standard error from three different samples of Forward Scatter Area vs Side Scatter Area ( FSC-A vs SSC-A ) . See Figure 2—figure supplement 2 for details of how a gate is drawn . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 00410 . 7554/eLife . 07935 . 005Figure 1—figure supplement 2 . The switching kinetics are non-exponential . Cells were pre-grown in the high RFP state and then transferred to 40 μM IPTG and 2 ng/ml ATc . The fraction of cells remaining in a high RFP state is plotted as a function of time . The switching kinetics are more complicated than what one would expect from first-order kinetics . Error bars represent the standard error from three different samples of FSC-A vs SSC-A . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 005 As a demonstration of how cellular memory operates in this gene network , one population of cells was pre-grown in IPTG to initialize the cells in a high GFP state ( Figure 1B ) . A separate population of cells was pre-grown in ATc to initialize the cells in a high RFP state . We then transferred the populations to a range of ATc concentrations and monitored the dynamics over several days using flow cytometry . All cultures were grown with a fixed concentration of IPTG ( 40 μM ) , which served as an orthogonal control variable for later experiments . For high and low concentrations of ATc , cells converged to a state that was independent of the history they were pre-grown in ( Figure 1B , C ) . However , for intermediate concentrations of ATc there was little to no switching of the phenotypic state of the cells even after 260 hr of growth and over 80 cell divisions . The mutual repression in this toggle switch therefore indeed allows for remarkably stable cellular memory for intermediate ATc concentrations . As expected ( Acar et al . , 2005 ) , the lifetime of the states decreases approaching the critical transition at which switching is deterministic ( hereafter , we refer to this point as the ‘critical transition’ or ‘phenotypic switch’ ) , although interestingly we observe non-exponential switching kinetics , potentially indicating the presence of metastable states in the gene network ( Figure 1—figure supplements 1 , 2 ) . This decrease in lifetime is one manifestation of deteriorating cellular memory approaching the critical transition , but it is not obvious whether in ATc concentrations with long lifetimes ( i . e . , strong memory ) the cells are also able to retain their memory in the face of environmental perturbations . In analogy to other complex systems , we hypothesized that phenotypic states in this gene network might become increasingly sensitive to perturbations near a critical transition ( Van Nes and Scheffer , 2007 ) . This expectation arises because the state's basin of attraction shrinks as the stable and unstable fixed points approach one another ( Figure 2A ) . Brief environmental perturbations will push the system out of equilibrium . If the system is far from the critical transition it will return to its original state after the perturbation is removed ( Figure 2B ) . However , close to the critical transition , the same perturbation might cause the system to cross the basin boundary , thus causing the cell to switch phenotypic states . Therefore , phenotypic states are expected to lose resilience to environmental perturbations near a critical transition . 10 . 7554/eLife . 07935 . 006Figure 2 . Cellular memory of the high GFP history in the toggle switch loses resilience to directional . ( A ) A schematic of how the effective potential changes and the basin of attraction shrinks approaching the critical transition . The size of the basin of attraction is determined by the distance between the stable and unstable fixed points . ( B ) Far from the critical transition , a perturbation temporarily depresses the value of GFP; the system recovers to its initial state after the perturbation is removed . Close to the critical transition , the same perturbation causes the system to cross the basin boundary into the alternative state . ( C ) 92 hr after history washout , cells at different distances from the phenotypic switch were exposed to a reduction in ( IPTG ) from 40 μM to 0 . 1 μM . Cells grew for 24 hr in this new condition . IPTG was then restored to 40 μM and cells were allowed to recover for 24 hr . Control cells were propagated with ( IPTG ) held fixed at 40 μM . ( D ) The fraction of cells that switched into a high RFP state in response to the perturbation is plotted as a function of distance from the tipping point . Two different strength perturbations , a weak ( 10 μM ) and a strong ( 0 . 1 μM ) are plotted . Error bars in D represent the standard error of three different samplings from forward scatter area ( FSC-A ) vs side scatter area ( SSC-A ) ( see Figure 2—figure supplement 2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 00610 . 7554/eLife . 07935 . 007Figure 2—figure supplement 1 . Cellular memory of the high RFP phenotypic state of the toggle switch loses resilience to directional perturbations . Cells were pre-grown in the RFP state . 92 hr after history washout , cells at different distances from the phenotypic switch were exposed to an increase in ( IPTG ) from 40 μM to 360 μM . Cells grew for 24 hr in this new condition . IPTG was then restored to 40 μM and cells were allowed to recover for 24 hr . The fraction of cells that switched into a high GFP state in response to the perturbation is plotted as a function of ( ATc ) . Control cells were propagated with ( IPTG ) held fixed at 40 μM . Error bars represent the standard error of three different samplings from forward scatter area ( FSC-A ) vs side scatter area ( SSC-A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 00710 . 7554/eLife . 07935 . 008Figure 2—figure supplement 2 . To control for cell size , tight gates on FSC-A vs SSC-A are selected for analysis . 10 , 000 events are collected on the flow cytometer , and the forward scatter area and side scatter area are plotted for each event . A narrow gate ( shown in black dashed lines ) is drawn to select a subset of cells ( approximately 200 ) for analysis . This effectively decouples fluorescence from cell size , so that differences in fluorescence are due to differences in expression of fluorescent proteins . Error bars in most plots are determined by analyzing three randomly chosen gates and calculating the standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 008 To test this theoretically proposed loss of resilience we perturbed cells at different distances from the critical transition ( i . e . , different ATc concentrations ) and measured how likely the cells were to switch into the alternative phenotypic state as a result of environmental perturbations . We examined two classes of perturbations , ‘directional’ ( in which the perturbation increased the probability of switching into the alternative state by inhibiting one or another of the two transcription factors that form the toggle switch ) and ‘generic’ ( in which the interaction between the toggle switch and the perturbation was not immediately obvious ) . To explore the directional perturbation we prepared cells in the GFP state by pre-growing them in IPTG . We then transferred the populations to multiple ATc concentrations for 4 days , leaving the cells in the high GFP state but at varying distances from the critical transition . At the end of the fourth day , the cells were then perturbed by decreasing the concentration of IPTG for 1 day before returning the IPTG concentration to its original value . Far from the critical transition ( 2 ng/ml ATc ) , the IPTG perturbation caused almost no cells to switch their phenotypic state ( Figure 2C ) . However , close to the critical transition ( 8 ng/ml ATc ) , the perturbation caused a great majority of the cells to switch into a high RFP state . Even after the perturbation was removed and the initial conditions were restored , many cells remained in a high RFP state . As expected , the severity of the environmental perturbation ( i . e . , the magnitude of reduction in IPTG ) correlated with the fraction of cells that switched states in response to the perturbation ( Figure 2D ) . Importantly , there was negligible switching ( ∼0 . 5% ) into the high RFP state for control cells grown in constant IPTG concentrations , demonstrating that the phenotypic switching was indeed caused by the perturbation . Moreover , the lack of phenotypic switching in the absence of the perturbation also indicates that in all of these conditions the traditional measure of cellular memory—the lifetime of the state—would classify all of these conditions as being stable with a high degree of cellular memory . Similar results were observed in the other direction , as cells pre-grown in the RFP state approach the critical transition associated with sudden switching to the GFP state ( Figure 2—figure supplement 1 ) . Thus , phenotypic states in the toggle switch lose resilience to directional perturbations near a critical transition . Cellular memory in development and cell cycle progression must be resilient against a wide range of different environmental perturbations . Given this , we wanted to explore whether memory in the toggle switch would lose resilience against generic perturbations approaching a critical transition . Cells from a high GFP history at different distances from the phenotypic switch ( 2 , 4 , and 8 ng/ml ATc ) were perturbed in several different ways for 24 hr: heat stress , osmotic stress with sodium chloride , ethanol stress , and a glucose pulse . Remarkably , we observed a loss of resilience against all four of these generic perturbations ( Figure 3A , B ) . Far from the critical transition ( 2 ng/ml ATc ) , there was little to no phenotypic switching in response to any of these ‘generic’ perturbations . However , close to the critical transition ( 8 ng/ml ATc ) we observed nearly complete switching in all perturbations , despite the fact that there was essentially no switching ( ∼0 . 5% ) in the absence of the perturbations . At a given distance from the phenotypic switch , increasing the strength of a generic perturbation increased the probability that cells would switch into the alternative state ( Figure 3—figure supplement 1 ) . The switching induced by the glucose perturbation can perhaps be understood by the fact that glucose shuts down expression of the entire system ( LacI , TetR , GFP , and RFP ) via catabolite repression of a GAL1 upstream activation sequence , thus pushing the cells toward a low GFP and low RFP state ( Gardner et al . , 2000; Ellis et al . , 2009 ) . The other three perturbations have much broader effects on the cell with no obvious connection to the toggle switch network being probed in our experiments . Cellular memory can therefore degrade near a critical transition for a wide range of different environmental perturbations ( Figure 3C ) . 10 . 7554/eLife . 07935 . 009Figure 3 . Cellular memory of the high GFP history in the toggle switch loses resilience to generic perturbations . ( A ) Cells were pre-grown in the high GFP state . 92 hr after history washout , cells at 2 , 4 , and 8 ng/ml ATc were exposed to an osmotic stress ( 600 mM NaCl ) . Cells grew for 24 hr in this new condition . The osmotic stress was then removed and cells were allowed to recover for 24 hr . Control cells were propagated with NaCl held constant throughout the whole time course . Growth media contains trace NaCl ( 2 mM ) . ( B ) The fraction of cells that switched into a high RFP state is plotted as a function of [ATc] . During the 24 hr perturbation period , cells were exposed to 6% ethanol ( pink ) , 600 mM NaCl ( peach ) , 37°C ( violet ) , 0 . 2% glucose ( blue ) , or no perturbation ( teal ) . Error bars in B represent the standard error of three different samplings from FSC-A vs SSC-A . All results were replicated in a second independent experiment several weeks later ( see Figure 3—figure supplement 2 ) . ( C ) A schematic of the key findings from the perturbation experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 00910 . 7554/eLife . 07935 . 010Figure 3—figure supplement 1 . Increasing the strength of a generic perturbation increases the probability that cells will switch into the alternative phenotypic state . Yeast cells expressing the toggle switch were pre-grown in a high GFP state and then transferred to an environmental condition that is close to the phenotypic switch ( [ATc] = 8 ng/ml ) . Cells were perturbed for 24 hr with a salt ( upper panel ) or an ethanol pulse ( lower panel ) and then allowed to recover for 24 hr . The fraction of cells that switch into the high RFP state in response to the perturbation is plotted as a function of perturbation intensity . ATc and IPTG were held fixed throughout the perturbation and recovery periods . Control cells were propagated with no supplemental salt or ethanol for comparison . Error bars represent the standard error of measurements from three different gatings on FSC-A vs SSC-A . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01010 . 7554/eLife . 07935 . 011Figure 3—figure supplement 2 . A loss of resilience to generic perturbations was confirmed in a second independent experiment . Same experiment as Figure 3B . Yeast cells expressing the toggle switch were pre-grown in the high GFP state and then transferred to different distances from the critical transition ( 2 , 4 , or 8 ng/ml ATc ) . After 92 hr of growth , the cells were perturbed for 24 hr and then allowed to recover for 24 hr . The fraction of cells that switched into a high RFP state is plotted as a function of ( ATc ) . During the perturbation period , cells were exposed to 6% ethanol ( yellow ) , 600 mM NaCl ( peach ) , 37°C ( violet ) , 0 . 2% glucose ( blue ) , or no perturbation ( teal ) . Error bars represent the standard error of three different samplings from FSC-A vs SSC-A . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01110 . 7554/eLife . 07935 . 012Figure 3—figure supplement 3 . Cellular memory of the RFP state of the toggle switch does not lose resilience to generic perturbations because salt , ethanol , and heat shocks stabilize the high RFP state . ( Left Panel ) Yeast cells were pre-grown in the high RFP state . 92 hr after history washout , cells at 2 , 4 , and 8 ng/ml ATc were exposed to 6% ethanol ( pink ) , 600 mM NaCl ( peach ) , 37°C ( violet ) , 0 . 2% glucose ( blue ) , or no perturbation ( teal ) . Cells grew for 24 hr in this new condition . The perturbation was then removed and cells were allowed to recover for 24 hr . Control cells were propagated in 2% galactose and 30°C with no salt or ethanol . The fraction of cells that switched into a high GFP state is plotted as a function of ( ATc ) . ( Right Panel ) The cells were pre-grown in a high GFP state ( green ) or high RFP state ( red ) and then transferred to an intermediate distance from the phenotypic switch ( 4 ng/ml ATc ) . 92 hr after history washout , the cells were perturbed as described above . The mean fluorescence shifts from its pre-perturbation location ( red or green dot ) to the tip of the perturbation arrow . Error bars in the left panel represent the Laplacian-corrected binomial counting error . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 012 Complex dynamic systems near critical transitions are predicted to lose resilience to a specific class of perturbations: those that push the system toward the alternative stable state . However , the effect of a salt shock or a heat shock on the GFP output of the toggle switch is challenging to predict a priori . Notably , it is possible that some generic perturbations will stabilize the state of the cells and reduce their probability of switching . Indeed , we found that none of the four generic perturbations led to a loss of resilience in the transition from the high RFP state ( Figure 3—figure supplement 3 ) . This result is unsurprising given that the perturbations did not strongly push the genetic network toward the high GFP state ( Figure 3—figure supplement 3 ) . For the salt , ethanol , and heat stresses , the perturbation increased the RFP output of the cells , thereby stabilizing the state . Thus , phenotypic states in the toggle switch only loss resilience to perturbations that push the system toward the alternative state . We reasoned that our experimentally-observed loss of resilience could be caused by a shrinking basin of attraction of the phenotypic state . To test this hypothesis , we used our perturbation experiments to estimate the location of the basin boundary in GFP-RFP space and see how it shifted for different concentrations of ATc ( Figure 4A ) . For a given concentration of ATc and a chosen perturbation , we measured the fraction of cells that switched into the alternative state at the end of the recovery period . Analysis of the GFP-RFP distribution at the end of the perturbation and before the recovery period revealed that many cells were in an indeterminate state between the two stable phenotypic states . We ranked the cells according to the ratio of GFP expression to RFP expression and assumed that a cell would switch phenotypic states when its ratio of GFP expression to RFP expression fell below a critical threshold . This is equivalent to assuming that the basin boundary between the two phenotypic states takes the form of a line on a log–log plot . To justify this assumption , we used a simple mathematical model of the toggle switch ( see Figure 4—figure supplement 1 and ‘Materials and methods’ ) . Our modeling indicated that this assumption would hold as long as the promoter strengths of the two transcription factors in the toggle switch were approximately equal . 10 . 7554/eLife . 07935 . 013Figure 4 . The loss of resilience is due to a shrinking basin of attraction of the phenotypic state . ( A ) By examining the RFP-GFP distribution at the end of the recovery period and comparing it to the end of the perturbation period , the basin boundary can be estimated . A cell is assumed to switch when its ratio of GFP to RFP expression falls below some threshold α , so the separatrix is a line with slope 1 and intercept α on a log–log plot . A simple model of the toggle switch supports this assumption ( see Figure 4—figure supplement 1 ) . For each [ATc] , eight perturbations ( 10 μM IPTG , 0 . 1 μM IPTG , 37°C , 200 mM NaCl , 600 mM NaCl , 2% ethanol , 6% ethanol , and 0 . 2% glucose ) were used to estimate α by minimizing the mean-squared deviation between the estimated and measured fractions . ( B ) The estimated fraction is compared to the measured fraction for [ATc] = 2 ng/ml ( □ ) , 4 ng/ml ( ∆ ) , and 8 ng/ml ( ○ ) . ( C ) The unperturbed GFP-RFP distribution for cells at 0 ng/ml ( green ) and 128 ng/ml ( red ) is overlaid with the estimated separatrix from 2 , 4 , and 8 ng/ml . μ ± σ is shaded for each separatrix . ( D ) The location of the high GFP stable fixed point ( green ) , unstable fixed point ( purple ) , and low GFP stable fixed point ( red ) are plotted as a function of ATc . The system is bistable for intermediate ATc and monostable at low and high ATc . We assume that switching follows a line in log-space connecting the centroids of the two distributions in Figure 4C ( see Figure 4—figure supplement 2 ) . Error bars in all plots represent the standard error from three samplings from the FSC-A vs SSC-A distribution . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01310 . 7554/eLife . 07935 . 014Figure 4—figure supplement 1 . A simple model of the toggle switch justifies the assumption that the basin boundary can be approximated as a line in LacI-TetR space . A deterministic model features cooperative binding of repressor proteins to their promoters ( see ‘Materials and methods’ section for details ) . The system was initialized with a wide range of initial concentrations of LacI and TetR . The system then evolved in time until it reached one of the two stable fixed points . Initial conditions leading to a high LacI state are shaded in green , and initial conditions leading to a high TetR state are shaded in red . The stable fixed points are represented by black circles . The basin boundary is the interface where the red and the green areas meet . Top: the promoters have equal strengths and bottom , the promoters have asymmetric strengths . See also the grey line in figure S7 of reference ( Wu et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01410 . 7554/eLife . 07935 . 015Figure 4—figure supplement 2 . Cell switching paths approximately follow a line on a log–log plot connecting the stable fixed points . Yeast cells in a high GFP condition ( 2 ng/ml ATc ) are plotted in green and cells in a high RFP condition ( 128 ng/ml ATc ) are plotted in red . The line connecting the centroids of the distribution is plotted in black . Overlaid in purple is the distribution of cells that were initially in a high GFP state and were then perturbed for 24 hr with a glucose pulse . The unstable fixed point in Figure 4D is estimated by finding the y-coordinate of the intersection of the basin boundary in Figure 4C with the black line above . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01510 . 7554/eLife . 07935 . 016Figure 4—figure supplement 3 . Increasing the duration of the perturbation increases the fraction of cells that switch phenotypes in a Gillespie simulation of the toggle switch . Here , p = 50 , K = 15 , and γ = 0 . 5 . 100 cells are initialized in a high Lac state and equilibrate for ten generations . Suddenly , K is increased to 100 for a variable time period ( ranging from 0 . 2 to 20 generations ) . K is then restored to 15 , and the cells are allowed to equilibrate for several more generations . At the end of the recovery , we determine what fraction of the cells remains in a high Lac state . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01610 . 7554/eLife . 07935 . 017Figure 4—figure supplement 4 . Not all perturbations induce switching in a Gillespie simulation of the toggle switch . Here , p = 50 , K = 15 , and γ = 0 . 5 . 100 cells are initialized in a high Lac state and equilibrate for ten generations . Suddenly , K is decreased to 1 for a variable time period ( ranging from 0 . 2 to 20 generations ) . K is then restored to 15 , and the cells equilibrate for several more generations . At the end of the recovery , we determine what fraction of the cells remains in a high Lac state . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 017 Using the assumption described above , we estimated the location of the basin boundary consistent with our experimentally observed fractions at the end of the recovery period . Encouragingly , we find that for a given concentration of ATc , the switching fraction for the eight different perturbations ( weak and strong IPTG , weak and strong ethanol , weak and strong salt , heat , glucose ) is well fit by a single separatrix ( Figure 4B ) . Moreover , as ATc increases from 2 to 8 ng/ml , the location of the basin boundary gets closer to the location of the high GFP stable fixed point ( Figure 4C ) . Knowing the location of the basin boundary allowed us to map the bifurcation ( Figure 4D ) . Thus , the loss of resilience appears to be driven by a shrinking basin of attraction of the phenotypic state near the critical transition . Given that the phenotypic states associated with our toggle switch lose resilience to perturbations near the critical transition , a natural question is whether , just as in other examples of critical transitions in complex systems such as ecosystem collapse , it is possible to develop warning indicators that this loss of resilience is taking place ( Dai et al . , 2012 ) . For example , one might expect that the mean GFP of cells in the high GFP state would decrease with increasing ATc , thus potentially signaling that the critical transition is approaching ( Isaacs et al . , 2003 ) . However , we find that the mean GFP of unswitched cells is approximately constant over the range of ATc concentrations in which we observe a loss of resilience ( Figure 1B and Figure 5A ) . Researchers in a number of fields have explored early warning indicators based on a loss of stability near a critical transition ( broadening of the effective potential as illustrated in Figure 2A ) ( Scheffer et al . , 2009 ) . This loss of stability would manifest in our experiments as an increase in the variation of GFP fluorescence among the population of ( unswitched ) cells . However , we find experimentally that there is no increase in variation within the population approaching the critical transition , even very close to the transition where the GFP state is metastable ( Figure 5B , C ) . The theoretically proposed early warning indicators based on local stability therefore fail to predict the critical transition in this gene network ( Menck et al . , 2013 ) . 10 . 7554/eLife . 07935 . 018Figure 5 . No significant change in mean fluorescence of the state and no significant increase in coefficient of variation approaching the critical transition . Yeast cells expressing the toggle switch were pre-grown in the high GFP state ( green triangles ) or the high RFP state ( red circles ) and then transferred to a range of ATc concentrations for 92 hr . ( A ) The mean GFP fluorescence of cells that have not switched from their pre-growth state is plotted against [ATc] . Above 16 ng/ml ( for the GFP history ) and below 1 ng/ml ( for the RFP history ) , all of the cells have switched into the alternative state . To quantify population variability , the standard deviation normalized to the mean ( coefficient of variation , i . e . , ‘CV’ ) is plotted for the B , log-transformed and C , linear values of fluorescence . When calculating variation , only cells that remain in the state they were pre-grown in are analyzed ( see inset in B ) . To minimize the effect of instrument noise in B and C , variation in RFP fluorescence is measured for the RFP history ( similarly , variation in GFP is measured for the GFP history ) . Error bars in A represent the standard error of three samplings from FSC-A vs SSC-A . Error bars in B and C are standard errors from 200 bootstrap resamplings of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 018 To explore the generality of our results , we chose to study the yeast GAL network , which displays an ‘all-or-none’ response in some sugar environments containing galactose ( Biggar and Crabtree , 2001; Song et al . , 2010 ) . However , the wild-type GAL network exhibits only weak memory ( Song et al . , 2010 ) . To expand the range of galactose concentrations for which the system has memory ( Acar et al . , 2005 ) , we used a strain of yeast constitutively expressing the repressor GAL80 ( which codes for a transcriptional repressor involved in one of the many feedback loops that stabilize memory in this network ) . To assess the state of the network , yellow fluorescent protein ( YFP ) expression was driven by a GAL1 promoter . Cells were pre-grown in high galactose ( GAL ON ) and then grown for a day in a range of galactose concentrations ( Figure 6A ) . We then examined the resilience of the GAL network to perturbations at different distances from its critical transition , similar to our experiments with the toggle switch . For the GAL network , a glucose pulse served as a directional perturbation ( due to catabolite repression [Gancedo , 1998] ) , and we again used heat , salt , and ethanol as generic perturbations . We found that the ‘GAL ON’ phenotypic state lost resilience to glucose and ethanol perturbations but not salt or heat ( Figure 6B ) . Similarly , we performed experiments probing the resilience of the ON state in the Escherichia coli lac operon , where we again observed a loss of resilience to both directional perturbations and to the generic perturbation ethanol ( Figure 6—figure supplement 1 ) . These three genetic switches have widely different architectures , and operate in two different species . Our results thus indicate that a loss of resilience approaching a phenotypic switch could be a general property of multistable gene networks and that ‘generic’ or global perturbations , such as temperature or salt shocks , can cause widespread loss of cellular memory . 10 . 7554/eLife . 07935 . 019Figure 6 . The yeast galactose network loses resilience to directional perturbations and the generic perturbation ethanol . ( A ) 25 hr after history washout , a gal80-inducible strain shows strong memory above 0 . 05% galactose . The two histories are offset for ease of viewing , and the region of memory is shaded in yellow . 50 cells at random are plotted from a tight gate on FSC-A vs SSC-A . ( B ) 25 hr after history washout , cells at 0 . 4% , 0 . 1% , and 0 . 08% galactose were exposed to 6% ethanol ( pink ) , 600 mM NaCl ( peach ) , 37°C ( violet ) , 0 . 1% glucose ( blue ) , or no perturbation ( teal ) . Cells grew for 12 hr in this new condition . The perturbation was then removed and cells were allowed to recover for 12 hr . Control cells were propagated with fixed glucose , galactose , and temperature . The fraction of cells that switched into a low YFP state after the perturbation is plotted as a function of [galactose] . Error bars are standard errors obtained by bootstrap with 200 resamplings of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 01910 . 7554/eLife . 07935 . 020Figure 6—figure supplement 1 . The Escherichia coli lactose network loses resilience to directional perturbations as well as to the generic perturbation ethanol . ( Upper panel ) The cells were pre-grown in either 0 or 100 μM TMG ( ‘LAC OFF’ and ‘LAC ON , ’ respectively ) for 20 hr and then diluted into a range of TMG concentrations . After 21 hr of growth , the system exhibits strong hysteresis . A constitutively expressed mCherry allows for discrimination between OFF cells and noise . The region of memory is shaded in gray , and the two histories are offset for ease of viewing . Fifty cells at random are plotted . ( Lower panel ) 21 hr after pre-growth washout , the cells from the ON history were exposed to several perturbations for 12 hr ( blue: 0 . 01% glucose; pink: 6% ethanol; peach: 1 M NaCl; violet: 43°C ) . The perturbations were then removed , and the cells were allowed to recover for 12 hr . The fraction of cells that switched into the OFF state is plotted as a function of [TMG] . Control cells ( green ) were grown at 37°C with no added salt , glucose , or ethanol . Error bars represent the standard error from 200 bootstrap resamplings of the data . DOI: http://dx . doi . org/10 . 7554/eLife . 07935 . 020
Here we have shown that brief perturbations in the extracellular environment can dramatically increase the rate of phenotypic switching from a highly stable memory state into an alternative state . We argue that this loss of resilience to perturbations near the bifurcation can be explained by a shrinking of the basin of attraction of the phenotypic state . By observing the GFP-RFP distribution at the end of the perturbation period and applying a simple threshold , we were able to accurately predict the fraction of cells that would return to the pre-perturbation memory state at the end of the recovery period . At a given distance from the bifurcation , the same basin boundary was able to accurately predict the switching fraction in response to eight different perturbations . The predictive success of this threshold is particularly impressive given that the different perturbations have dramatically different effects on the GFP-RFP distribution ( see Figure 4A ) . There is an interesting question of the precise mechanism by which the perturbations act on the system to push the cells to a new location in GFP-RFP space . Part of the answer is certainly that the perturbations cause a shift in the stability landscape underlying the phenotypic states . It is also possible that the perturbations cause an increase in noisy gene expression . Understanding what is occurring during the perturbation period , both biochemically and from a dynamical systems perspective , will be an area of investigation for future research . The perturbations ( salt stress , heat stress , etc ) lasted for a significant duration: 24 hr , or roughly 3 to 8 cell divisions . It is interesting to ask whether we would have seen similar results if we had perturbed the system for a shorter duration . To address this question , we performed Gillespie simulations using a very simple phenomenological model of the toggle switch ( see Source code 1 ) . Similar to our test tube experiments , we initialized the system in a high Lac state . The system was allowed to equilibrate for several generations , before we suddenly changed one of the parameters of the system ( either the disassociation constant between the transcription factor and the DNA or the promoter strength ) . We performed this perturbation for a variable duration ranging from 0 . 2 generations to 20 generations . The system was then allowed to recover ( by restoring the parameter to its initial value ) for several generations , and we then determined what fraction of cells remained in a high Lac state . Two key observations emerged . ( 1 ) The degree of switching is set by both the intensity and duration of the perturbation . For dramatic perturbations , the minimum duration of perturbation to induce switching is approximately set by the cell division time ( Figure 4—figure supplement 3 ) . ( 2 ) Not all perturbations induced phenotypic switching . Of note , perturbations which stabilized the high Lac state induced negligible switching into the high Tet state ( Figure 4—figure supplement 4 ) . These observations held regardless of whether the perturbation was achieved by changing the disassociation constant or the promoter strength . Understanding the stability of phenotypic states in gene networks remains an important challenge in biology . Here we have demonstrated that for robust cellular memory in natural contexts subject to environmental noise , low rates of stochastic switching from the phenotypic state is not sufficient . This is because the cellular memory must also be robust against environmental fluctuations and perturbations that are impossible to avoid . Here we found experimentally that several phenotypic states lost resilience to multiple environmental perturbations near a critical transition . Given that over time there will often be multiple kinds of environmental fluctuations and perturbations , our results argue that many forms of cellular memory will become destabilized near a critical transition . Our results provide a roadmap for exploring cellular memory and phenotypic switching in other contexts , from development to cancer progression .
Toggle switch strains have been previously characterized ( Ellis et al . , 2009; Wu et al . , 2013 ) . Gal80-inducible strains have been previously described ( Acar et al . , 2005 ) . LacY-YFP fusion strains have been previously described ( Choi et al . , 2008 ) . Cells were grown in synthetic media ( YNB and CSM—Trp—Leu; Sunrise Science , San Diego , CA , United States ) containing 2% galactose and ATc/IPTG ( Sigma Aldrich , St . Louis , MO , United States ) as described in the text . Cells were grown in 3 ml cultures in 14 ml VWR culture tubes ( VWR , Radnor , PA , United States ) and diluted daily to prevent saturation . Cells were pre-grown for 24 hr in either 1 mM IPTG or 250 ng/ml ATc and then transferred to a range of ATc concentrations as described in the text . 20 μl of cells were harvested daily at OD 0 . 5 , diluted in 180 μl PBS , and run immediately on a Miltenyi MACSQuant VYB flow cytometer ( Miltenyi Biotec , San Diego , CA , United States ) . After 4 days of serial transfer , cells were transferred to the perturbation environment . Cells grew for 24 hr in this environment before being characterized via flow cytometry and diluted into the original environment they were in before the perturbation ( hereafter , the ‘recovery environment’ ) . Cells grew for 24 hr in the recovery environment and then were characterized with flow cytometry . The basin boundary was estimated by assuming that cells switched phenotypes when the ratio of GFP to RFP expression fell below a critical threshold . All results were verified in two independent experiments carried out several weeks apart . Cells were pre-grown for 24 hr in either 0 . 5% galactose or 0 . 01% glucose ( GAL ON and GAL OFF , respectively ) plus 0 . 05 μg/ml doxycycline ( Sigma Aldrich ) . Cells were then diluted into 0 . 01% glucose , 0 . 05 μg/ml doxycycline , and galactose concentrations ranging from 0 to 2% . Cells were diluted every 12 hr and the OD was kept below 0 . 01 to minimize consumption of the sugars . Cells were harvested every 12 hr , concentrated via centrifugation , and characterized immediately via flow cytometry . Cells were transferred to the perturbation environment 24 hr after history removal . They were then grown for 12 hr , characterized via flow cytometry and diluted into the recovery environment , grown for a further 12 hr , and characterized via flow cytometry . Cells were grown in M9 media ( Sigma Aldrich ) with 0 . 1% succinic acid as a carbon source . Cells were first pre-grown in either 0 or 100 µM TMG ( ‘Lac OFF’ and ‘Lac ON , ’ respectively ) for 20 hr . The history condition was then washed out , and the two populations were then separately transferred to a range of TMG concentrations . Cells grew for 21 hr . YFP expression was assayed using flow cytometry as a proxy for the state of the Lac network . A constitutively expressed RFP enabled for discrimination between cells and noise . 21 hr after history washout , cells were diluted into the perturbation environment and grown for 12 hr . After 12 hr of growth , YFP expression was again assayed using flow cytometry . At the same time , the cells were diluted into the recovery environment and grown for a further 12 hr . At the end of the recovery period , YFP expression was again assayed using flow cytometry . Flow cytometry data was analyzed utilizing the Gore lab's flow cytometry tool kit , which can be accessed at http://gorelab . bitbucket . org/flowcytometrytools/ . The code for the Gillespie simulation of the toggle switch has been uploaded as a supplementary file . The rate of production of the two repressors is described by the following equations: ( 1 ) [LacI]˙=PTet1+ ( [TetR]KTet ) 2− γ[LacI] , ( 2 ) [TetR]˙=PLac1+ ( [LacI]KLac ) 2− γ[TetR] . The promoter strength is determined by the DNA sequence and has been previously characterized ( Ellis et al . , 2009 ) . As a simple approximation , we treat ATc and IPTG as modulating KTet and KLac . We assume that the proteins are stable , so the rate of transcription factor degradation is set by dilution via cell division . Our primary goal for the model was to predict the shape of the basin boundary between the two phenotypic states , particularly in a system where the promoter strengths are asymmetric . Figure 4—figure supplement 1 was generated by picking values of KTet and KLac so that the system exhibited bistability over a range of [LacI] and [TetR] . We initialized the system with many different [LacI] and [Tet] and allowed the system to evolve according to Equations 1 , 2 until the system reached a steady state . For the equal promoter strengths , the following parameters were used:PLac = PTet = 50KLac = KTet = 15γLac = γTet = 0 . 5 For the unequal promoter strengths , the following parameters were used:PLac = 30PTet = 35KLac = KTet = 15γLac = γTet = 0 . 5 For the Gillespie simulations to investigate the effect of perturbation duration on switching fraction , we again used:PLac = PTet = 50KLac = KTet = 15γLac = γTet = 0 . 5 Similar to our test tube experiments , we initialized the system in a high Lac state . The system was allowed to equilibrate for several generations , before we suddenly changed one of the parameters of the system ( either the disassociation constant between the transcription factor and the DNA or the promoter strength ) . We performed this perturbation for a variable duration ranging from 0 . 2 generations to 20 generations . The system was then allowed to recover ( by restoring the parameter to its initial value ) for several generations , and we then determined what fraction of cells remained in a high Lac state . | All organisms need to be able to react to the challenges thrown at them by their changing environment . Yeast , bacteria and other microbes have networks of genes that can give rise to many different traits and characteristics , which can also be referred to as phenotypes . A change in the environment can alter the activities' of the genes so that the microbes display a different phenotype . The point at which a small change in the environment can lead to a sudden switch in the phenotype is called a ‘critical transition’ . An individual microbe's history can influence the phenotype that it presents . However , it is not clear how microbes ‘remember’ their history , or how fluctuations in the environment might cause the microbe to lose the ability to store this memory and present a different phenotype instead . Here , Axelrod et al . studied phenotype memory in yeast cells grown in the laboratory . The experiments used cells that had been genetically modified to glow red in the presence of a molecule called anhydrotetracycline ( or ATc ) and to glow green in the absence of the molecule . Axelrod et al . examined what effect altering the levels of this molecule would have on the phenotype produced by the cells . First , the cells were grown with no ATc present for several generations so that the cells glowed green . Next , Axelrod et al . added different amounts of ATc were added . For moderate levels of ATc the cells continued to glow green , illustrating that they ‘remembered’ their prior growth condition . However , cells exposed to higher levels of ATc lost this memory and changed color . Next , Axelrod et al . carried out further experiments on cells exposed to ATc levels that were close to , or further away from the critical transition . At high levels of ATc ( that is , close to the critical transition ) , many cells switched from green to red when exposed to high temperatures , salt and other changes in the environment . On the other hand , very few of the cells grown in low levels of ATc—and therefore further away from the critical transition—changed color in response to the same fluctuations in their environment . These finding reveal that phenotype memory is less stable when yeast experience fluctuations in their environment close to a critical transition . Future work will seek to find out how salt or high temperatures can abolish phenotype memory . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"physics",
"of",
"living",
"systems",
"computational",
"and",
"systems",
"biology"
] | 2015 | Phenotypic states become increasingly sensitive to perturbations near a bifurcation in a synthetic gene network |
Understanding how bacteria affect plant health is crucial for developing sustainable crop production systems . We coupled ecological sampling and genome sequencing to characterize the population genetic history of Rhodococcus and the distribution patterns of virulence plasmids in isolates from nurseries . Analysis of chromosome sequences shows that plants host multiple lineages of Rhodococcus , and suggested that these bacteria are transmitted due to independent introductions , reservoir populations , and point source outbreaks . We demonstrate that isolates lacking virulence genes promote beneficial plant growth , and that the acquisition of a virulence plasmid is sufficient to transition beneficial symbionts to phytopathogens . This evolutionary transition , along with the distribution patterns of plasmids , reveals the impact of horizontal gene transfer in rapidly generating new pathogenic lineages and provides an alternative explanation for pathogen transmission patterns . Results also uncovered a misdiagnosed epidemic that implicated beneficial Rhodococcus bacteria as pathogens of pistachio . The misdiagnosis perpetuated the unnecessary removal of trees and exacerbated economic losses .
Symbioses are persistent and intimate interactions between organisms . In pathogenic interactions , one partner benefits at the expense of the other . In mutualistic symbioses , specific partners interact and reciprocally benefit . Associative symbioses are a variation of mutualism in which there is lower specificity between interacting partners ( Drogue et al . , 2012 ) . In agricultural systems , practices are employed to limit pathogens , to introduce nitrogen-fixing mutualistic rhizobia and to restore associative symbionts such as plant growth-promoting bacteria ( PGPB ) . PGPB can directly promote the growth of plants and protect against pathogens ( Barea et al . , 2005; Pieterse et al . , 2014 ) . The beneficial or parasitic outcomes of symbioses , especially those involving environmentally acquired partners , are often not guaranteed . The health of the host , location of the symbiont on the host , or unregulated proliferation of the symbiont can lead to alternative outcomes ( Lin and Koskella , 2015 ) . The genotype of the symbiont is also a critical factor , as horizontal gene transfer ( HGT ) can lead to the acquisition of new genes that innovate genomes , driving evolutionary transitions and establishing new lineages of beneficial or pathogenic symbionts ( Soucy et al . , 2015 ) . In some pathogenic symbionts , however , HGT does not bestow the genome with innovative functions , nor do these genomes exhibit substantive changes . Rather , the few horizontally acquired genes encode products that reprogram core genes , thereby co-opting the genome for virulence ( Letek et al . , 2010 ) . Rhodococcus is a genus of Gram-positive bacteria with members that persist in a variety of terrestrial and aquatic ecosystems ( de Carvalho et al . , 2014; Larkin et al . , 2005 ) . Rhodococcus includes taxonomic groups with members that have been repeatedly recovered from leaf and root tissues of various species of plants ( Bai et al . , 2015; Bodenhausen et al . , 2013; Bulgarelli et al . , 2012; Hong et al . , 2015 , 2016; Lebeis et al . , 2015; Lundberg et al . , 2012; Qin et al . , 2009 , 2011; Salam et al . , 2017 ) . It has been suggested that hosts enrich for members of Rhodococcus because of the beneficial traits of the bacteria ( Hong et al . , 2015 , 2016 ) . Plant-associated Rhodococcus species are better known as pathogens ( Putnam and Miller , 2007 ) . Two clades of Rhodococcus include members that can cause disease in over 100 genera of plants ( Creason et al . , 2014a2014a; Putnam and Miller , 2007 ) . Herbaceous plants are the most commonly affected whereas woody plants are less frequently infected . Disease symptoms include leafy galls , witches’-brooms , and other disfiguring growths . Pathogenic isolates of Rhodococcus , typified by the most-studied isolate D188 , require three virulence loci that are most frequently found clustered on virulence plasmids ( Crespi et al . , 1992; Stes et al . , 2011 ) . These plasmids are approximately 200 kb in length and are linear replicons ( Francis et al . , 2012; Creason et al . , 2014b ) . Some of the fas ( fasciation ) genes are necessary for disease and encode proteins that synthesize and modify cytokinins , which are predicted to be secreted effectors ( Crespi et al . , 1994; Pertry et al . , 2009 ) . The fasR gene , predicted to be a transcriptional regulator , is also necessary for pathogenicity ( Temmerman et al . , 2000 ) . att ( attenuation ) mutants are reportedly attenuated in disease and thus implicated in virulence ( Crespi et al . , 1992; Maes et al . , 2001 ) . The vicA gene , which encodes malate synthase , an enzyme in the glyoxylate cycle , is the only locus encoded on the chromosome implicated in virulence ( Vereecke et al . , 2002 ) . Pathogenic Rhodococcus are particularly problematic in agricultural settings that produce plants for their aesthetic value . A variety of biotic and abiotic stresses , some induced by anthropogenic practices , cause symptoms that are confused with those caused by Rhodococcus ( Putnam and Miller , 2007 ) . Furthermore , isolates of Rhodococcus that lack virulence genes are often cultured from symptomatic tissues ( Creason et al . , 2014b; Nikolaeva et al . , 2009 ) . Multiple tests are used to confirm that a plant is infected by pathogenic Rhodococcus . Bacteria must exhibit the proper morphology on selective media and be taxonomically assigned to Rhodococcus . The bacteria must have virulence genes and must cause disease symptoms in susceptible indicator plants . Because the fasR gene and some of the fas genes are necessary for pathogenicity of the bacteria , their detection is sufficient to confirm pathogenicity of Rhodococcus isolates . In 2011 , populations of micropropagated pistachio UCB-1 ( Pistacia atlantica × Pistacia integerrima ) rootstocks planted in commercial fields began showing an unusual phenotype ( Stamler et al . , 2015a , 2015b ) . Aerial phenotypes of ‘pistachio bushy top syndrome’ include shortened internodes , loss of apical dominance , stem galls , and reduced grafting success . Estimates suggest that more than 1 million trees grown on 25 , 000–30 , 000 acres were affected . Rhodococcus isolates were cultured from symptomatic plants , and when inoculated onto UCB-1 , they caused morphological changes to the hosts ( Stamler et al . , 2015b ) . This was the first report of pistachio being susceptible to Rhodococcus . Subsequent release of the genome sequences for PBTS1 and PBTS2 , the reported outbreak strains , indicated they lack virulence loci ( Stamler et al . , 2016 ) . The detection of vicA was used as evidence for pathogenic Rhodococcus bacteria and to guide management practices , the most extreme and costly being the removal of entire orchards . A second incidence of pistachio bushy top syndrome occurred in 2016 , resulting in the destruction of 1 . 5 million nursery trees . We determined and analyzed genome sequences from over 80 isolates of Rhodococcus , mostly collected from symptomatic herbaceous plants grown in production settings . Analysis of chromosomal sequences shows that plants host multiple lineages of Rhodococcus . Isolates that lack virulence plasmids can promote changes to the architecture of roots , but if a virulence plasmid is acquired , the isolates transition to being pathogenic . The analysis of chromosomal sequences of pathogens revealed the potential for multiple infections and reservoir populations at nursery sites , as well as for point source outbreaks . However , the distribution patterns of virulence plasmids suggested that agricultural systems can be locations that promote evolutionary transitions and the rapid generation of new lineages of pathogens , providing an alternative route for the spread of pathogens . Last , our results challenge previous conclusions that Rhodococcus isolates lacking virulence genes are causative agents of pistachio bushy top syndrome , suggesting that the pistachio syndrome was likely misdiagnosed ( Stamler et al . , 2015a , 2015b ) .
We used a genomic epidemiological approach to study the transmission patterns of Rhodococcus in a plant agricultural system . Sixty isolates were collected hierarchically across space and time . Multiple isolates were cultured from the same symptomatic tissues or from different plants grown at the same production site ( Supplementary file 1A ) . Previously sequenced isolates , many also collected from production sites , were included ( Supplementary file 1A; Creason et al . , 2014b ) . The nursery sources of the isolates were anonymized . Phylogenetic analysis placed the 60 isolates within Rhodococcus Clades I and II , which also included the 15 previously confirmed pathogenic isolates ( Figure 1; Creason et al . , 2014b ) . The two clades are sisters to Clades III and IV . Clade III consists of isolates that were cultured from microbiota of the model plantArabidopsis thaliana ( Bai et al . , 2015; Bodenhausen et al . , 2013; Bulgarelli et al . , 2012; Lebeis et al . , 2015; Lundberg et al . , 2012 ) . No member of Clades III or IV have virulence genes . The four clades are distinct , and separated by a long branch , from the other species of Rhodococcus ( Figure 1—figure supplement 1 ) . The isolates in the four clades were operationally classified into 17 species , indicated with a lowercase letter , of which 14 have at least one isolate cultured from a plant ( Supplementary file 1B ) . Fifty-one of the newly sequenced isolates encode virulence genes . We inspected their genome assemblies as well as those from previously sequenced pathogenic isolates ( Creason et al . , 2014b ) . All but four of the 66 genome assemblies had att , fasR , and fas on the same contig as pFi_009 . The fas locus is present in a region that is conserved in pFiD188 , the virulence plasmid of isolate D188 , suggested to be necessary for plasmid replication and maintenance ( Francis et al . , 2012 ) . The pFi_009 gene is predicted to encode a telomere-associated protein hypothesized to be necessary for the replication of the linear virulence plasmid . The genome assemblies of isolates 06-469-3-2 and 05-2254-6 had contig breaks that disrupted the linkages between virulence and plasmid-associated loci , but these contigs were nonetheless similar in composition and are co-linear to the reference plasmid sequence . The average sequencing coverage of plasmids relative to that of corresponding chromosomes was 1 . 89 ± 0 . 56 . Only three assemblies had coverages less than 1 . 0 but all three had virulence loci on the same contig as pFi_009 . A21d2 and A25f , previously sequenced , are exceptional because the virulence loci are encoded in their chromosomes ( Creason et al . , 2014b ) . Therefore , of the 66 isolates confirmed or inferred to be pathogenic , 64 carry a virulence plasmid . Pathogenic isolates were assigned to eight different species ( Figure 1 ) . We identified single nucleotide polymorphisms ( SNPs ) for 82 isolates and used these SNPs to define genotypes and to assemble two clade-specific minimum spanning networks ( Supplementary files 1C and 1D; Figure 2 ) . The genotypes show pairwise differences in between 220 and 11 , 714 SNPs . We mapped nursery information onto the network , which provided information on potential transmission patterns ( Figure 2 ) . In this figure , the ‘a’ identification is associated with nurseries in which we identified evidence of multiple and independent infections . Plants from nursery N15 , indicated with ‘a1’ , were infected by five genotypes belonging to Clade I and two belonging to Clade II . Nursery N8 ( ‘a2’ ) and several others were also associated with multiple genotypes . However , nursery N8 had a single host plant that was infected by at least three genotypes that represented six of the cultured isolates . The three isolates within one of these genotypes differ by up to 20 pairwise SNPs , whereas the two isolates in one of the other genotypes have no differences ( Supplementary file 1C ) . The third genotype is separated from the other two by 231 and 1414 pairwise SNPs ( Figure 2E ) . Epidemiological link ‘b’ was detected in nursery N15 . The corresponding genotype includes seven isolates from Clade IIa , sampled four years apart from Campanula plants ( Supplementary file 1A ) . These isolates are separated by 0–2 pairwise SNPs . Epidemiological links designated as ‘c’ were also made between isolates collected from geographically separated nurseries . Nurseries N1 , N12 , and N13 ( ‘c1’ ) had isolates of the associated genotype ( Clade IIb ) that are separated by 1–4 pairwise SNPs and were collected up to 13 years apart from Leucanthemum and Geranium plants . Nurseries N7 , N11 , and N14 ( ‘c2’ ) had isolates of the associated genotype ( Clade Ia ) that were separated by 12–20 pairwise SNPs . The isolates were collected nine years apart , and from Veronica plants . The virulence plasmids of 64 plant pathogenic Rhodococcus isolates were categorized on the basis of the phylogenetic analysis of 123 genes that are present in at least 95% of the virulence plasmids , and sub-categorized on the basis of patterns of gene presence/absence ( Figure 3; Figure 3—figure supplement 1; Supplementary file 1E ) . The phylogeny has strong support for two major plasmid types , as well as one unique type only carried by isolate LMG3616 ( Figure 3—figure supplement 1 ) . The plasmid sequences within the major clades are conserved and provided few informative nucleotide polymorphisms for sub-categorizing plasmids . The greatest contribution to plasmid diversity is gene gain and loss , and the relative clustering of plasmids based on the presence/absence of genes allowed subgrouping into INDEL variants , as indicated with uppercase letters in Figure 3 . At the level of plasmid type , presence/absence categorization was identical to the phylogeny ( Figure 3—figure supplement 1 ) . The genes that define each INDEL variant are often present in largely contiguous regions in the plasmid , and could have been acquired as blocks ( Francis et al . , 2012 ) . We next characterized the distribution of plasmids in epidemiologically linked isolates and showed that the patterns of plasmids were inconsistent with those expected of transmission between isolates of a chromosomal genotype ( Figures 2C–G and 3 ) . One genotype associated with ‘a1’ ( the most left node of network; Figure 2C ) includes isolates 05-339-1 and 05-339-2 , which differ by seven pairwise SNPs and which carry Type IA and Type IB plasmids , respectively . The genotype associated with ‘b’ includes isolate 02–815 , which carries a Type IA plasmid ( Figure 2D ) . This isolate is epidemiologically linked to six isolates ( 0–2 pairwise SNPs ) , but the isolates collected four years later carry a Type IB virulence plasmid . Likewise , the genotype associated with ‘c1’ includes isolate A22b , which carries a Type IA virulence plasmid and is linked to two isolates carrying Type IB plasmids ( Figure 2F ) . The most striking deviation is found in the genotype associated with ‘c2’ ( Figure 2G ) . Isolate 06-1477-1A carries a Type IB plasmid and is epidemiologically linked to two isolates that carry a Type IIA plasmid . Even the two Type IIA plasmids are dissimilar as one has a block of genes that , along with other gene INDELs , distinguishes it from the other . The diverse genotypes of Rhodococcus isolated from nurseries can carry similar plasmid types and these types are not taxonomically restricted . Isolates 06-412-2B of Clade Ib , 05-340-1 of Clade Ie and 06-156-3c of Clade IIa , all of group ‘a1’ detected in nursery N15 , have Type IB plasmids that differ by only a few gene INDELs ( Figure 2C and D ) . The phylogeny shows that there are cases in which isolates from the same clade , such as RS1C4 and 06-1059B-a ( solid black lines ) , carry different types of plasmids ( Figure 3—figure supplement 2; solid black lines ) . It was also observed that distantly related Rhodococcus isolates carry the same type of plasmid . For example , 15-649-1-2 of Clade II and LMG 3623 of Clade I both carry Type II plasmids ( Figure 3—figure supplement 2; solid red lines ) . Two pathogenic isolates were excluded from the analysis because the virulence loci are present in their chromosomes ( Creason et al . , 2014b ) . A25f was recovered from nursery N12 , whereas A21d2 was recovered from nursery N15 ( Supplementary file 1A ) . In this dataset , we identified nine isolates of Rhodococcus that lack virulence genes . We also identified five such isolates in a previous study , and often fail to detect virulence genes while diagnosing Rhodococcus cultured from diseased plants ( Creason et al . , 2014b ) . Others have implied that these are strains that have lost the plasmid ( Nikolaeva et al . , 2009 ) . The genetic diversity of co-existing isolates described here suggests otherwise ( Figure 1; Supplementary files 1A-D ) . Pathogenic 06-1477-1A and virulence-gene-lacking 06-1477-1B were isolated from a symptomatic Veronica plant , belong to Clades Id and Ia , respectively , and have 1521 pairwise SNPs . Pathogenic isolate 14-2483-1-1 ( Clade Ib ) was cultured from the same symptomatic plant as virulence-gene-lacking isolate 14-2483-1-2 ( Clade Ie ) , and the two differ by 2820 SNPs . Interestingly , 14-2483-1-2 has the attR and attX virulence genes , as well as 228 nucleotides of the attA virulence gene , on a contig that is dissimilar in sequence and greater in length than the virulence plasmids . The first two coding sequences and the intergenic regions have ≥88% nucleotide identity to corresponding sequences in D188 . For attA , only the first 65 nucleotides are identical to its homolog in D188 . Results from PCR confirmed that the structure of the locus was not a result of misassembly . Two virulence-gene-lacking isolates , 14-2470-1a and 14-2470-1b , were cultured from a symptomatic plant . These two isolates are in Clades IIa and IIb , respectively , and differ by 13 , 858 SNPs . Of the isolates cultured and tested from this plant , no pathogenic isolate was detected . An alternative explanation for the presence of genetically diverse , virulence-gene-lacking isolates of Rhodococcus is that such isolates are beneficial and enriched for by plants . We therefore compared the symbiosis phenotype of seven virulence-plasmid-lacking isolates that represent the four clades against four virulence-gene-carrying isolates , D188 , A44a , A25f , and A21d2 ( Supplementary file 1F; Creason et al . , 2014b; Desomer et al . , 1988; Lundberg et al . , 2012; Miteva et al . , 2004 ) . The virulence-gene-carrying isolates were previously determined to be pathogenic , and were selected on the basis of having variations in the structure and sequence of their virulence loci ( Creason et al . , 2014b ) . PBTS1 and PBTS2 , implicated as outbreak strains and cultured from the leaf endophytic compartment of pistachio , were also included ( Stamler et al . , 2015b ) . The seven virulence-gene-lacking isolates , as well as PBTS1 and PBTS2 , failed to cause disease when inoculated onto the meristems of mature Nicotiana benthamiana . The four pathogenic isolates caused leafy galls ( Figure 4A ) . These four were also the only ones tested that caused significant inhibition of root elongation , thickening of the stem , and terminal arrest at the cotyledon stage ( no primary growth or development of lateral roots ) , when assayed on seedlings of N . benthamiana ( Figure 4B and C ) . We examined plants up to two months after inoculation , and seedlings remained terminally arrested . Most isolates reduced the vertical growth of roots , compared to that of mock-inoculated plants , but inhibition by virulence-gene-lacking isolates was more variable and not as severe as that measured in the roots of pathogen-inoculated seedlings ( p-values were <0 . 0001 except for the treatment with PBTS2 [p-value = 0 . 1153] ) . Importantly , seedlings inoculated with virulence-gene-lacking isolates did not show thickening of the stem or terminal arrest at the cotyledon stage , morphological changes associated with disease ( Figure 4 ) . Instead , we noticed that all virulence-gene-lacking isolates caused changes to the architecture of the roots ( Figure 4B ) . Relative to mock-inoculated plants , there were proliferations in root hairs and the plants had more lateral roots or earlier development of lateral roots . The former change was quantified in seedlings that were inoculated with members from a subset of the virulence-gene-lacking isolates . There were significant increases in the number of root hairs ( averages ranging from 166 . 1 to 217 . 3; p-values were all ≤0 . 0045 ) , compared to mock-inoculated plants ( average of 89 . 8; Figure 5A and B ) . The isolates varied in their effect , with PBTS2 having the strongest measurable effect ( 217 . 3; p-value<0 . 0001 ) . GIC26 provoked the most visually striking proliferation of root hairs and its extreme effect challenged our ability to count and measure root hairs accurately ( Figure 5A; p-value=0 . 0003 ) . The average length of the root hairs was significantly longer following inoculation with the four isolates of Rhodococcus ( averages ranged from 0 . 02286 to 0 . 03604 mm vs 0 . 01186 mm in mock-inoculated plants; p-values were all <0 . 0001; Figure 5C ) . Pathogenic isolate D188 was included as a control , but the root hairs of plants that were inoculated with this isolate were too sparse in number to warrant quantification ( Figure 5A ) . Five additional isolates , including 14-2483-1-2 which has part of the att locus , were tested and shown to cause changes to the architecture of the roots of plants ( Figure 5—figure supplement 1 ) . We analyzed the genome sequences to identify genes that are potentially involved in providing beneficial traits ( Figure 5—figure supplement 2; Bruto et al . , 2014; Glick , 2012; Sparacino-Watkins et al . , 2014 ) . Few homologs or pathways were associated with beneficial traits or an endophytic lifestyle . There was also no obvious correlation with symbiosis phenotype . Some isolates have a homolog predicted to encode 1-aminocyclopropane-1-carboxylate ( ACC ) deaminase , but this sequence had a strong phylogenetic signal , and is predominantly in members of Clade Ia ( Glick , 2014 ) . No complete tryptophan-dependent auxin biosynthetic pathway was identified ( Spaepen et al . , 2007 ) . The most highly represented classes of carbohydrate active enzymes ( CAZYmes ) are members of carbohydrate esterase groups CE1 and CE10 . CE1 includes xylanases , which degrade hemicellulose ( Lombard et al . , 2014 ) . CE10 consists of cholinoesterases , a group of enzymes that act on non-carbohydrate sources . In addition , we identified cutinases ( C5 ) and pectate lyases ( PL22 ) . These classes of enzymes may contribute to the endophytic lifestyle of Rhodococcus . There are 108 genes that are enriched in the genomes of isolates from the four plant-associated clades ( Supplementary file 1G ) . Forty genes are annotated as hypothetical and many others lack sufficient information in their annotations . AntiSMASH analysis identified anywhere from 1 to ~ 20 loci that may be involved in the production of secondary metabolites ( Figure 5—figure supplement 2; Blin et al . , 2017 ) . Few had sufficient similarities to previously characterized loci to allow the inference of the identity of the metabolite . Whether and how the functions of these genes contribute to the plant-associated lifestyle are unknown . The pFiD188Δatt virulence plasmid was successfully conjugated into a subset of the Rhodococcus isolates that originally lacked virulence genes . This plasmid variant encodes fasR and fasA-F , but has a kanamycin resistance gene that disrupts attR , attX , and attA-G ( Maes et al . , 2001 ) . Regardless , plasmid pFiD188Δatt is sufficient for isolate D188 to cause disease in mature plants and seedlings , and plants that were treated with a strain containing this plasmid were no different from those infected with D188 ( p-value>0 . 9999; Figure 6—figure supplement 1; Maes et al . , 2001 ) . Despite repeated attempts , we were not able to conjugate the plasmid into isolates outside of Clade I successfully . Each of the pFiD188Δatt-carrying isolates caused leafy galls on plants ( Figure 6A ) . In addition , these isolates were no longer capable of causing increases in the growth of root hairs , unlike their corresponding near-isogenic genotypes ( Figure 6B ) . Instead , the isolates carrying pFiD188Δatt , when compared to their near-isogenic plasmid-lacking genotypes , caused disease symptoms and significantly inhibited the growth of seedlings ( Figure 6B–C; p-values were all <0 . 0001 ) . The inverse transition was also demonstrated . We isolated a variant of D188 , D188ΔpFiD188 , which lacks the virulence plasmid . When inoculated onto roots of N . benthamiana seedlings , D188ΔpFiD188 caused changes to the architecture of root systems and no longer caused disease to mature plants ( Figure 6B; Figure 6—figure supplement 2 ) . We had to generate a new plasmid-lacking strain because the previously generated D188-5 is compromised in in vitro growth ( Figure 1; Figure 6C; Figure 6—figure supplement 2; Desomer et al . , 1988 ) . Analysis of its genome sequence revealed a significant deletion of 25 . 4 kb from the chromosome ( Supplementary file 1H ) . Most of the affected 25 genes have annotated functions implicated in housekeeping functions . Sequencing of D188ΔpFiD188 confirmed that it only lacked the virulence plasmid . This isolate also grew similarly to D188 and had no measurable fitness defects ( Figure 6C; Figure 6—figure supplement 2 ) . Only three loci on pFiD188 have been implicated in virulence . We have not been able to repeat results showing that the deletion mutant of att is attenuated in virulence ( Figure 6—figure supplement 1; Crespi et al . , 1992; Maes et al . , 2001 ) . But when constitutively expressing attR , a homolog of the LysR transcriptional regulator necessary for att gene expression , D188 caused unusual leafy galls on N . benthamiana ( Figure 6—figure supplement 3 ) . Unlike normal galls that terminate primary growth , those caused by the attR-overexpressing strain regained meristematic activity . When inoculated onto roots , the symptoms were more variable , but nonetheless similar to those caused by D188 . The effects were significantly different relative to those seen in mock-inoculated seedlings ( p-value<0 . 0001 ) or in those inoculated with D188 ( p-value<0 . 0001 ) . The fas locus is predicted to be necessary for Rhodococcus to produce and secrete a mix of cytokinins ( Pertry et al . , 2009 ) . Approximately 0 . 1 μM of the synthetic cytokinin 6-benzylaminopurine ( BA ) was equivalent to a starting inoculum of only ~2 . 5×103 colony-forming units ( cfu ) of D188 ( Figure 7A ) . However , regardless of the amount of BA in the medium , the exogenously applied cytokinins only inhibited root elongation and did not provoke the thickening of stems or arrest plant growth . Our results show that PBTS1 and PBTS2 are not pathogenic on N . benthamiana ( Figures 4 and 5 ) . We next tested whether altering the dose influences the outcome of interaction between N . benthamiana and PBTS2 . As inoculum levels of PBTS2 were increased , there was a greater reduction in root length ( Figure 7B ) , but the effect was never to the same robustness and degree as that measured in seedlings infected with D188 . In addition , PBTS2 did not cause thickening of stems or terminal arrest in the growth of the plant . At 28 days post-inoculation ( dpi ) , the leaves of seedlings inoculated with the highest tested levels of PBTS2 had developed to the same stage as those of mock-inoculated seedlings , whereas D188-inoculated seedlings remained arrested in growth ( Figure 7C ) . The roots of seedlings inoculated with PBTS2 also formed lateral roots . As inoculum levels of PBTS2 were decreased , there was less reduction of root length , and at the lowest dose tested , roots were significantly longer ( 1 . 161 cm; p-value = 0 . 0043 ) than those of mock-treated plants ( 1 . 007 cm; Figure 7B ) . A pathogenic PBTS2 strain carrying pFiD188Δatt showed a dose effect similar to that seen for D188 ( Figure 7B; p-values were all >0 . 3584 for all within-dose comparisons ) . To exclude the possibility that these results are due to incompatibility between PBTS1 and PBTS2 and N . benthamiana , other species of plants were tested . We used pea , an indicator species for confirming pathogenic Rhodococcus , and UCB-1 pistachio , reportedly the host of the epidemic . Both plant species failed to show disease symptoms , regardless of whether isolates were tested individually or in combination ( Figure 7—figure supplement 1 ) . Nine additional pistachio isolates that lack virulence genes also failed to cause disease . Four isolates , 14-687 , 14-688 , 14-694 , and 14-700 , were cultured from asymptomatic pistachio plants while five , SR18 , AGD2M , AGD3B , AGD6D , and AGD6H , were cultured from symptomatic pistachio plants . Even pathogenic isolates D188 and A44a failed to cause disease symptoms in UCB-1 pistachio . Another aspect of the diagnosis of pistachio bushy top syndrome was the use of vicA to confirm pathogenic Rhodococcus ( Stamler et al . , 2015a , 2015b ) . Primers designed for fasA and fasD specifically amplified a product of expected size from pathogenic isolates D188 , A44a , and A25f , and failed to amplify a product from any of the tested beneficial strains ( Figure 8A; Supplementary file 1I; Nikolaeva et al . , 2012; Serdani et al . , 2013 ) . The primers for fasA , and fasD failed to yield a product from A21d2 because this isolate carries an analog of the fas locus ( Creason et al . , 2014b ) . The molecular detection of fasR using a loop-mediated isothermal amplification ( LAMP ) -based assay specifically distinguished all tested pathogenic isolates from beneficial isolates . The detection of vicA did not follow a pattern consistent with the pathogenicity phenotype ( Figure 8A ) . It has a high false-positive rate and detected several , but not all , beneficial isolates . Our repeated attempts to amplify vicA from PBTS1 were unsuccessful; there are four and six mismatches between the two primers used for PCR and the vicA sequence from PBTS1 . Homologs of vicA are predicted to be present in nearly all members of the Actinobacteria , including in all 407 Rhodococcus isolates for which genome sequences are available . The topologies of the Rhodococcus genus and the vicA trees are largely congruent , indicating that this locus is mostly vertically inherited , but with some evidence of recombination ( Figure 8B ) . To address the need for on-site molecular tools to distinguish pathogenic from beneficial genotypes , we used a new molecular detection method that is rapid , robust , and sensitive ( Piepenburg et al . , 2006 ) . We targeted attE and attG because they are the most unique relative to all sequences in the databases and are conserved among the pathogenic isolates that we have examined ( Creason et al . , 2014b ) . The use of the primers for attE and attG in standard PCR and recombinase polymerase amplification ( RPA ) basic successfully amplified products of expected size from DNA of pathogenic strains , including A21d2 and A25f ( Figure 8C ) . No products were detected when DNAs from beneficial strains were used as templates . An additional oligonucleotide probe that anneals within the amplified fragment was designed for RPA nfo , and when coupled with modified amplification primers , this probe was successful in detecting a product via lateral flow . This method was specific and discriminated between pathogenic and beneficial Rhodococcus . Moreover , RPA nfo can be completed in the absence of specialized equipment , and can yield results in just 30 min .
Whole-genome-enabled epidemiological studies have revealed local , global , and historical patterns for the transmission of human pathogens and have informed on health care ( Comas et al . , 2013; Croucher et al . , 2011; Harris et al . , 2010 , 2013; Mutreja et al . , 2011; Parkhill and Wren , 2011; Walker et al . , 2013 ) . Ours is a case study for using genomic epidemiology to uncover and explain the transmission patterns of phytopathogens in agricultural systems . The investigation of plasmid distribution highlighted the significant role of HGT in shaping the population structure of pathogenic bacteria and revealed challenges in modeling their transmission . Our analysis of chromosomal SNPs suggested that nurseries experience multiple and independent introductions of pathogenic Rhodococcus , exemplified by ‘a’ isolates observed in nurseries N15 and N8 ( Figure 2 ) . The link between isolates collected across time ( ‘b’ ) is indicative of a reservoir population that has a pathogenic genotype . The epidemiological links ( ‘c1’ and ‘c2’ ) of isolates from different nurseries support the possibility of point source outbreaks and suggest that the sources have reservoir populations . However , the distribution of plasmids also indicated that alternative processes may be occurring ( Figures 3 and 9 ) . First , two different plasmid types are associated with ‘c2’ . This is not expected from an outbreak and is more consistent with different members of a lineage acquiring plasmids separately . Second , different plasmid variants are carried by isolates that are defined by genotypes associated with ‘a1’ , ‘b’ , and ‘c1’ ( Figure 2 ) . These are not expected patterns and probably reflect separate acquisitions of plasmids by different members of a lineage or the rapid and independent evolution of plasmids . Third , genetically distinct lineages of Rhodococcus at nursery N15 carry the same variant of plasmid . This is best explained by multiple lineages acquiring plasmids from a common donor population . Nurseries often produce a large variety of perennials and clonally propagated plants that are frequently handled and intensely managed in multiple production settings , and are often in regions that produce many different agricultural commodities . These are prime locations for different genotypes of Rhodococcus to interact and for plasmids to be transferred , switched , and evolved . We also demonstrated that plants in agricultural systems are hosts to isolates of Rhodococcus that are probably beneficial associative symbionts ( Supplementary file 1A; Vacheron et al . , 2013 ) . Root hairs are extensions that increase the surface area of roots , forming an extensive interface between plant and soil . Virulence-gene-lacking isolates of Rhodococcus caused significant increases in the number and length of root hairs , which may enable plants to be more efficient in the uptake of water and dissolved nutrients ( Figure 5 ) . This potentially beneficial growth-promoting phenotype is consistent with the finding of a number of reports identifying Rhodococcus within endophytic compartments and the rhizosphere of plants , and with the suggestions that the bacteria are enriched for by plants because of their beneficial traits ( Bai et al . , 2015; Bodenhausen et al . , 2013; Bulgarelli et al . , 2012; Francis et al . , 2016; Hong et al . , 2015 , 2016; Lundberg et al . , 2012; Qin et al . , 2009 , 2011; Salam et al . , 2017 ) . The associative symbionts are genetically diverse and represent 14 different species circumscribed by four sister clades ( Figure 1; Supplementary file 1B ) . The mechanism by which isolates of Rhodococcus cause growth changes to plants is unknown , as genome mining efforts suggest that these traits are potentially novel ( Figure 5—figure supplement 2 ) . Acquisition of a virulence plasmid by isolates representing eight species of Clades I and II is sufficient to drive an evolutionary transition ( Figures 1 and 6 ) . Loss of the plasmid reverted Rhodococcus to being beneficial , consistent with the hypothesis that virulence genes function irrespective of genomic background ( Figure 6—figure supplement 2 ) . In addition , we could not identify any chromosome-located genes that are enriched in pathogenic isolates , in comparison to non-pathogenic isolates , that could be potential candidate virulence genes . Evolutionary transitions , such as the switch from being a free-living , non-pathogenic lineage to being a pathogenic lineage , have been detected frequently ( for example , by Bruto et al . , 2017 ) . The transition described for Rhodococcus is seamless and the mutualist to pathogen transition has been described only rarely ( Sachs et al . , 2011 ) . It is possible that focus on characterizing binary outcomes in symbioses have obscured the true fluidity of symbioses . Similar transitions may have occurred in Agrobacterium/Rhizobium , a group of bacteria that express plasmid-mediated traits of significance to plant agriculture ( de Lajudie et al . , 1999; Glaeser et al . , 2016; Hao et al . , 2012; Lacroix and Citovsky , 2016; Wang et al . , 2006 ) . Transitions and the rapid generation of new lineages of pathogens could occur in agricultural systems , where plants are frequently host to multiple Rhodococcus isolates of different genotypes . Pathogenic isolate 05-2254-6 , which is associated with ‘a2’ , is most similar to virulence-gene-lacking isolate , 14-1411-2a , also collected from N8 . The 220 pairwise SNPs that differ between these two isolates exceeded the threshold used to define genotypes , but the two isolates are nonetheless closely related ( Figures 1 and 2; Supplementary file 1C ) . This is not unique , as several pathogenic genotypes are related to virulence-gene-lacking genotypes , as indicated by their intermingling in the phylogeny and their connectivity in the network . The patterns involving distinct genotypes or plasmid types , such as ‘a1’ and ‘c2’ , respectively , are also consistent with evolutionary transitions ( Figures 2 and 3 ) . An example that is similar to ‘a1’ is the genotype of D188 and LMG3605 , both of which have different plasmid variants . The limited genetic diversity in virulence plasmids within our dataset was unexpected and suggests that their common ancestor evolved recently ( Figure 3—figure supplement 1 ) . This is in dramatic contrast to the virulence plasmids of Agrobacterium species of bacteria , where the Ti and Ri plasmids form different types that can be easily distinguished on the basis of the phylogeny of a single , conserved virulence gene ( Fuller et al . , 2017 ) . It is also remarkable that the virulence plasmid , which carries only three virulence loci , is sufficient for pathogenicity across eight genetically diverse species of Rhodococcus ( Figures 1 and 9; Creason et al . , 2014b; Francis et al . , 2012; Letek et al . , 2010 ) . FasR is a predicted transcriptional regulator , and is probably key for reprogramming the genome to transition bacteria to pathogens . The roles of the other two loci are still unclear . Results presented here suggest that att contributes to the maintenance of disease symptoms , but the mechanism is unknown ( Figure 6—figure supplement 3 ) . The Fas-produced mix of cytokinins are predicted to be secreted into plants and necessary to cause disease symptoms , but the existing data are not consistent with this hypothesis ( Pertry et al . , 2009 ) . Exceedingly low amounts of cytokinins are detected in culture-grown bacteria and plants have a variety of cytokinin-buffering mechanisms ( Creason et al . , 2014b; Kieber and Schaller , 2014; Pertry et al . , 2009 , 2010 ) . A miniscule amount of starting bacterial inoculum is sufficient to provoke disease symptoms , and exogenous applications of cytokinins fail to phenocopy the effects of pathogenic Rhodococcus ( Figure 7 ) . Last , the cytokinin mixture model is challenged by the revelation that the fitness of D188-5 , a key isolate used to develop the model , is severely compromised by a 25-kb deletion ( Figure 6—figure supplement 2; Supplementary file 1H; Desomer et al . , 1988; Maes et al . , 2001; Pertry et al . , 2009 , 2010; Stes et al . , 2011; Temmerman et al . , 2001 ) . This study highlights the importance of understanding the genetic and phenotypic characteristics of an organism and the consequences of prematurely drawing conclusions from incomplete data . There is no evidence to suggest that HGT or evolutionary transitions confounded the diagnosis of pistachio , and we could not detect pathogenic Rhodococcus from symptomatic tissues of pistachio . We were unsuccessful in repeating results showing that PBTS1 and PBTS2 , or other isolates cultured from pistachio , cause disease symptoms on plants ( Figure 4; Figure 7—figure supplement 1 ) . We could not amplify virulence genes from PBTS1 or PBTS2 , but when a virulence plasmid was introduced into PBTS2 , it was sufficient to transition PBTS2 to a pathogen of a plant species that is demonstrably susceptible to Rhodococcus ( Figures 6 and 7B ) . Whether pistachio is even a host for pathogenic Rhodococcus is unresolved ( Figure 7—figure supplement 1 ) . Nevertheless , we recognize the insurmountable challenge in showing that there is no possibility that pathogenic Rhodococcus causes pistachio bushy top syndrome . The results from this work prompted us to examine previous studies retrospectively ( Figure 9; [Stamler et al . , 2015a , 2015b] ) . There was a targeted search for Rhodococcus , the justification for which is unfounded because the symptoms on pistachio are unlike any produced by pathogenic Rhodococcus in any of the 100+ known hosts ( Putnam and Miller , 2007 ) . Bacteria were inexplicably cultured from asymptomatic leaves distal to symptomatic stem tissues . The key study did not include control strains or reproduce galling and graft failure , the most defining disease symptoms observed in field settings . A high inoculum of Rhodococcus was used and high doses of even beneficial bacteria can have costs ( Figure 7 ) . For example , some human diseases are caused by dysbiosis in which an imbalance of gut microbiotia causes disease ( Bloom et al . , 2011 ) . In addition , hosts often employ mechanisms to regulate or ensure nonpersistent interactions with beneficial bacteria ( Gutjahr and Parniske , 2013; Magori et al . , 2009; Nyholm and McFall-Ngai , 2004; Reid et al . , 2011; Wopereis et al . , 2000 ) . Results based on molecular detection were similarly tenuous . The use of the vicA locus was misleading and led to conclusions regarding the pathogenicity of Rhodococcus ( Figure 8 ) . The reported amplification of fragments of fasA and fasD ( GenBank accession numbers KP274062 and KP274064 ) from PBTS1 as well as fasD ( KP274067 ) from PBTS2 cannot be reconciled with the absence of the genes from the genome sequences ( Stamler et al . , 2015b , 2016 ) . We used a different technology to re-sequence independently prepared DNA from PBTS2 ( Supplementary file 1F ) . The assembly is co-linear with the publically available sequence and lacks the virulence plasmid and virulence genes ( Stamler et al . , 2016 ) . The possibility that virulence plasmids are unstable in populations grown outside of plant environments is not supported by the data . DNA extracted from bacteria grown in culture was used as a template in both PCR and whole-genome sequencing ( Stamler et al . , 2015b , 2016 ) . We successfully sequenced plasmids from 64 culture-grown pathogenic isolates , which had undergone multiple transfers , and repeatedly and successfully detected pathogenic isolates cultured from symptomatic tissues ( Creason et al . , 2014b ) . Some of the results of molecular detection were clearly artifacts . The two fragments reported to correspond to fasD are identical in sequence , each consisting of two short fragments of 147 and 194 nucleotides long that are artificially joined together by 280 ‘Ns’ . In our dataset , A21d2 is the only sequenced pathogenic isolate that lacks fasA , but it also lacks a homolog of fasD . It is thus not expected that PBTS2 should have only fasD but no fasA ( Stamler et al . , 2015b ) . The most egregious artifact was the reported amplification of vicA from PBTS1 ( GenBank accession number KP274063 ) , which we could not reproduce ( Figure 8 ) . When the sequence of the reportedly amplified fragment was used as a query to BLAST search the PBTS1 genome sequence directly , we failed to identify a homologous region . When used in searches against publically available databases , the top hits other than KP274063 were NM-J PBTS ( KR153287; 100% identity ) , DMS3-9 ( KJ677035; 97% identity ) , D188 complete genome ( CP015235; 96% identity ) , and PBTS2 complete genome ( CP015220; 96% identity ) . Another study posited that the virulence plasmids are unstable in populations persisting within plant environments ( Nikolaeva et al . , 2009 ) . Genome sequences showed instead that Rhodococcus isolates are from genetically distinct lineages . This observation further emphasizes the need to use appropriate molecular diagnostic tools that discriminate pathogenic from non-pathogenic bacteria . LAMP to detect fasR or RPA nfo to detect attE or attG are both sensitive methods that have the additional benefit of being rapid and less dependent on specialized equipment ( Figure 8; Serdani et al . , 2013 ) . The data are consistent with the possibility that previous conclusions rest on a misdiagnosis of pistachio bushy top syndrome ( Stamler et al . , 2015a , 2015b ) . If so , Rhodococcus was implicated as pathogenic , irrespective of genotype and in disregard of the genetic and phenotypic diversity of the genus . This mindset conflates all bacteria as causative agents of disease and demotes the importance of bacteria in promoting the health of their hosts ( Figure 9 ) . This potential misdiagnosis of pistachio bushy top syndrome could be responsible for catastrophic effects . An estimated 2 . 5 million trees have been destroyed , resulting in tremendous economic loss . Efforts to identify the true cause were decelerated . Accusations regarding the source of Rhodococcus have introduced conflict into an industry struggling with a considerable and unfamiliar problem . Considering the data described here , previous conclusions should , at the very least , be tempered and recommendations for managing plants in which Rhodococcus bacteria have been detected , reexamined . Ideally , these findings will renew efforts to identify the true nature of the syndrome afflicting UCB-1 pistachio rootstocks . Actinobacteria are prominent members of plant-associated communities , but the mechanisms and evolution of traits that are important for Gram-positive bacteria to reside in microbial communities and influence plant health are not well understood ( Figure 9 ) . Members of the Rhodococcus genus are excellent models for addressing this knowledge gap . A wealth of associated resources , such as an extensive and diverse collection of genotypes , associated genome sequences , and genetic tools have been developed for Rhodococcus . The members of this taxon can also interact with and benefit genetically tractable plant species . Last , the members of Rhodococcus are model organisms for characterizing evolutionary transitions between alternative symbiotic states .
Rhodococcus isolates used in this study are listed in Supplementary files 1A and F . Bacteria were maintained on solid LB medium at 28oC or grown overnight in LB medium at 28oC with shaking . Prior to conjugation , streptomycin-resistant Rhodococcus bacteria were selected for each of the recipient genotypes . Conjugations were done as previously described ( Desomer et al . , 1988 ) . Donor and recipient strains were grown in yeast extract buffer ( YEB ) and shaken at 28°C . Each genotype was mixed at a ratio of 1:1 and filtered through a nitrocellulose filter ( pore size , 0 . 45 μm; diameter , 25 mm; MilliporeSigma , Temecula , CA , USA ) . The filters were incubated on YEB agar plates for 24 to 28 h at 28°C . The cells were washed from the filter with 5 ml of a buffer containing 10 mM Tris-HCl pH 7 . 5 and 10 mM MgSO4 and then diluted and plated on YEB medium containing the appropriate antibiotic . Escherichia coli was grown on LB medium at 37oC . When appropriate , the medium was amended with 50 µg/ml of antibiotic kanamycin for Rhodococcus or E . coli . For growth curves , cultures of overnight-grown Rhodococcus were pelleted , washed , and resuspended at OD600 = 0 . 5 in a final volume of 200 µl of LB medium in 96-well flat-bottomed plates . The bacteria were grown for a period of 14 hr at 28oC with shaking in a Tecan Spark 10 m plate reader . Optical density ( OD600 ) measurements were taken every hour . Three technical replicates were included for each isolate , and the experiment was repeated at least three times with similar results . The Wizard genomic prep kit ( Promega , Fitchburg , WI , USA ) was used to extract genomic DNA from Rhodococcus . Directions for Gram-positive bacteria were followed . DNA was quantified with a Nanodrop spectrophotometer and adjusted to 50 ng/µl . Total genomic DNA was used to prepare Nextera XT libraries , and the resulting multiplexed libraries were sequenced on an Illumina HiSeq 3000 to generate 250mer paired end sequencing reads ( Center for Genome Research and Biocomputing [CGRB] , Oregon State University ) . Reads were processed as follows . FastQC was used to assess sequencing reads for quality ( Andrews , 2014 ) . BBduk v . 35 . 82 , with the parameters ‘ktrim = r k = 23 mink = 9 hdist = 1 minlength = 100 tpe tbo’ , was used to remove adapter sequences ( Bushnell , 2014 ) . SPAdes v . 3 . 1 . 1 , with the parameters ‘--careful -k 21 , 33 , 55 , 77 , 99’ was used to correct errors and to de novo assemble the reads into contigs ( Bankevich et al . , 2012 ) . Blobtools was used to assess assemblies and guide elimination of contigs likely to be derived from contaminating bacteria ( based on combined GC content , coverage , and contig annotation ) ( Kumar et al . , 2013 ) . Prokka was used to annotate the assembled genome sequences ( Seemann , 2014 ) . Sequences for the maximum likelihood multi-locus sequence analysis ( MLSA ) tree were acquired using the autoMLSA tool ( Davis Ii et al . , 2016 ) . The sequences for genes , ftsY ( ABG98302 . 1 ) , infH ( ABG98417 . 1 ) , rpoB ( ABG93773 . 1 ) , rsmA ( ABG97450 . 1 ) , secY ( ABG97930 . 1 ) , tsaD ( ABG97962 . 1 ) , and ychF ( ABG97656 . 1 ) from the genome sequence of Rhodococcus jostii RHA1 were translated and used as queries in TBLASTN v . 2 . 2 . 31 searches against the assembled genome sequences and the NCBI nt database , masked to Rhodococcus ( Adékambi et al . , 2011; accessed 12/2016 ) . Of those from NCBI nt , eight strains lacking all seven sequences and/or duplicate results were removed from the analysis . The sequences were aligned using MAFFT v . 6 . 864b with default settings ( Katoh and Standley , 2013 ) . A RAxML accessory script was used to determine the best-fitting protein model for each protein sequence alignment . Phylogenetic trees ( 100 ML searches , ‘autoMRE’ criterion bootstrap replicates ) were generated using RAxML v . 8 . 1 . 17 with a partitioned alignment of the MLSA protein sequences ( Stamatakis , 2014 ) . A similar analysis using R . fascians D188 vicA ( AMY55488 . 1 ) as a query was used to acquire and assemble a phylogeny of 162 malate synthase gene sequences from the NCBI nr database masked to Rhodococcus ( accessed 04/2017 ) . A cophylo plot of the MLSA and vicA trees was generated using the R package phytools ( Revell , 2012 ) . For the genes present in 95% of the virulence plasmids , sequences were concatenated using the R package EvobiR SuperMatrix function prior to constructing phylogenies ( Blackmon and Adams , 2015 ) . Only bootstrap values greater than 50 are shown . Bowtie2 v . 2 . 2 . 3 , with the option ‘--local’ , was used to align reads to the chromosome references sequences of D188 ( CP015235 . 1 ) or A44a ( GCF_000760735 . 1 ) , based on the clade assignment of the corresponding isolates ( Langmead et al . , 2009 ) . Alignments were converted to bam format using samtools v . 0 . 1 . 18 and read groups were added using Picard tools v . 2 . 0 . 1 ( Li et al . , 2009; Picard Tools , 2015 ) . GATK v . 3 . 7 HaplotypeCaller and the options ‘-ERC GVCF -ploidy 1’ were used to call variants for each isolate , and the data were then combined using GenotypeGVCFs ( McKenna et al . , 2010 ) . Variants were filtered using the R package vcfR with depth filtering using quantile probabilities of 0 . 25 and 0 . 75 as cutoffs and a minimum of four reads , as well as a missing data cutoff of 20% ( Knaus and Grünwald , 2017 ) . Variants were converted into a fasta alignment using bcftools v . 1 . 3–14-ge0890a1 vcf-to-tab and the perl script vcftab-to-fasta ( Li et al . , 2009; Chen , 2012 ) . Genotypes were called based on a threshold of 25 SNPs , and bitwise distances , using the R package poppr , were used to assemble minimum spanning networks ( Kamvar et al . , 2014 ) . Pairwise average nucleotide identity ( ANI ) between Rhodococcus isolates was calculated using autoANI ( Davis Ii et al . , 2016 ) . Get_homologues v . 20170418 with MCL clustering was used to cluster genes from 206 Rhodococcus genomes into orthologous groups ( Contreras-Moreira and Vinuesa , 2013 ) . The parse_pangenome_matrix . pl script of get_homologues was used to identify genes enriched ( with a 95% threshold ) in genomes of Rhodococcus in the four plant-associated clades . To identify pathogenicity loci in sequenced isolates , fasR , fas and att were used as queries in TBLASTN searches against genome assemblies . CONTIGuator was used to map assembled contigs to the reference strain D188 to search for sequences corresponding to pFiD188 or pFID188-like plasmids ( Galardini et al . , 2011 ) . Get_homologues v . 20170418 was used to cluster genes from each of the virulence plasmids as well as the other Rhodococcus linear plasmids ( Francis et al . , 2012 ) . Plasmids were clustered on the basis of gene presence/absence using binary distances and Ward's method for clustering ( ward . D2 ) . dbCAN HMMs 5 . 0 was downloaded and used with ad hoc scripts to identify CAZYmes from translated genome sequences ( Yin et al . , 2012 ) . The antiSMASH database was downloaded on 06/2017 and analyzed using antiSMASH ver . 4 . 0 ( Blin et al . , 2017 ) . Queries used in TBLASTN searches were gene sequences ascertained from searching the literature ( Bruto et al . , 2014; Glick , 2012; Sparacino-Watkins et al . , 2014 ) . HISAT2 was used to align sequencing reads from strain D188-5 to the D188 reference genome sequence . Variants ( SNPs ) were called using freebayes , filtered to those with quality score greater than 20 using vcffilter , and annotated using SNPdat v . 1 . 0 . 5 ( Doran and Creevey , 2013 ) . Seedling root inhibition assays were performed as described previously , with the exception that after bacteria were adjusted to OD600 = 0 . 5 , they were sometimes diluted or concentrated ( Creason et al . , 2014b ) . At least 100 seedlings were assayed per treatment . Images were taken at 7 days post inoculation ( dpi ) and data were analyzed . For root hair quantification , a dissecting microscope , equipped with a camera , was used to capture images at 10 and 25 dpi . Root hairs within a 1 cm segment , 1 cm below the stem were quantified using ImageJ ( Schneider et al . , 2012 ) . For cytokinin inhibition assays , three-day-old germinated seedlings were transplanted to MS ( half-strength MS , 0 . 5M MES ) medium containing DMSO ( control ) or 0 . 01–10 . 0 µM 6-benzylaminopurine ( BA ) and then grown and quantified in the same manner . Leafy galls were induced using the decapitation method on four-week-old N . benthamiana plants ( Creason et al . , 2014b ) . Images were taken 28 dpi . Pisum sativum ‘Alaska’ seeds were surface-sterilized in 70% ethanol for 1 min , 10% bleach for 10 min , and washed three times with sterile water . Seeds were soaked in sterile water for 60 min and then plated on water agar ( 15 g agar/L ) . Plates were incubated at 23°C until radicles were approximately 5 mm . Germinated seeds were soaked in suspension of Rhodococcus isolates ( OD600 = 0 . 2 ) or 10 mM MgCl2 buffer for 45 min . Ten seeds per treatment were included in each experiment . Inoculated seeds were placed in sterile test tubes containing 5 ml of Hoagland’s nutrient agar . Samples were incubated for 14 days at 23°C with a 16/8 light/dark cycle . Stem number and length were quantified at 14 dpi . Infection of UCB-1 pistachio was done , with minor modifications , according to previously described protocols ( Stamler et al . , 2015b ) . Briefly , control isolates , PBTS1 , PBTS2 , and a 1:1 mixture of PBTS1 and PBTS2 were suspended in 10 mM MgCl2 ( final OD600 = 0 . 7 ) . Treatment groups consisting of 15 seedlings were spray-inoculated with 200 ml of bacterial suspension . A mock-inoculated control group was sprayed with 200 ml of 10 mM MgCl2 . Inoculated seedlings were placed in humidity chambers for 14 days and the plants were maintained in a greenhouse for seven months . Tree height and internode length were measured at 30 day intervals , and the final measurements were recorded at 210 dpi . Unless indicated , all experiments were repeated at least three times with similar results . For all data sets , outliers were identified using the ROUT method ( Q = 1% ) and removed . Data were analyzed using One-way or Two-way ANOVA followed by Tukey’s multiple comparisons test ( GraphPad Prism v . 7 , GraphPad Software , La Jolla , CA , USA ) . Box and whisker plots were generated using the Tukey method; colored dots indicate outliers . Means are indicated by + . The attR gene was PCR-amplified from D188 genomic DNA and subcloned downstream to the L5 bacteriophage promoter in vector pJDC165 ( Jeff Cirillo , Texas A and M ) . The L5::attR construct was verified via Sanger sequencing . Rhodococcus competent cells were prepared from overnight-grown 3 ml cultures . Cells were pelleted and washed twice with sterile , cold dH2O , followed by one wash with sterile , cold 10% glycerol . Cells were resuspended in 50 µl 10% glycerol . Plasmid DNA ( 0 . 5–1 µg ) was added to the cells . After 30 min of incubation on ice , the cells were electroporated in 1 mm gap cuvettes at 2 . 2 kV . Cells were resuspended in 250 µl SOC medium and incubated at 28°C with shaking for 16 hr prior to plating on LB medium with appropriate antibiotics . For PCR , the following were used: 1x ThermoPol reaction buffer ( New England Biolab , Ipswich , MA , USA ) , 200 μM dNTPs , 0 . 2 μM of each primer , 50 ng genomic DNA template , 0 . 625 units Taq DNA polymerase ( New England Biolab , Ipswich , MA , USA ) , in a final volume of 25 μl . PCR conditions were 95°C , 3 min; 30 cycles of 95°C for 30 s , 55°C for 30 s , 72°C for 1 min; 72°C for 10 min; 16°C hold . Reactions with water , instead of a DNA template , were used as a negative control . For LAMP , the reaction mixture was as follows: 0 . 5 ng DNA template , 1x ThermoPol reaction buffer ( New England Biolab , Ipswich , MA , USA ) , 5 mM MgSO4 , 140 µM dNTPs , 146 µM hydoxynaphthol blue ( HNB ) , 1 . 6 µM each 16FIP and 16BIP primers , 0 . 2 µM each 16F3 and 16B3 primers , and 12 U Bst polymerase in a final volume of 25 µl . Reactions were incubated at 64°C for 60 min and then cooled to 4°C . Tubes were centrifuged briefly at 8000 rpm . RPA was done per the manufacturer’s instructions ( TwistAmp Basic , TwistDx Limited , Cambridge , UK ) . Reactions consist of 0 . 48 μM per primer , 29 . 5 μl rehydration buffer , 12 . 2 μl water , and 1 . 0 μl genomic DNA . A volume of 2 . 5 μl 280 mM magnesium acetate ( MgAc ) was added to initiate the reaction . The reaction was incubated at 37°C for 30 min ( Fuller et al . , 2017 ) . Products were purified using the QIAquick PCR purification kit ( Qiagen , Germany ) , run out on a 2 . 0% agarose gel , stained with ethidium bromide , and visualized under UV light . Products were verified via Sanger sequencing . RPA reactions coupled to lateral flow detection were comprised of 0 . 42 μM forward primer , 0 . 42 μM biotin-labeled reverse primer , 0 . 12 μM probe , 29 . 5 μM rehydration buffer , 12 . 2 μl water , and 1 . 0 μl of 25 ng/μl genomic DNA . Reactions were added to a freeze-dried pellet provided by the manufacturer ( TwistAmp nfo , TwistDx Limited , Cambridge , UK ) with the subsequent addition of 2 . 5 μl of 280 mM MgAc to initiate the reaction . Following subsequent incubation at 37°C for 30 min , the dual-labelled amplicon was visualized using a lateral flow dipstick ( Milenia Biotec GMbH , Germany ) . One microliter of the RPA product was diluted in 49 μl 1 . 0x PBST and 10 μl of the dilution were applied to the base of the dipstick , which was subsequently submerged in 100 μl 1 . 0 PBST at room temperature until the visualization of the positive control band , typically lasting two minutes . Sequences of primers and probes , and their modifications , are described in Supplementary file 1I . | All organisms live in a world teeming with bacteria . Some bacteria are beneficial and , for example , provide their hosts with nutrients . Others cause harm , for example , by stealing nutrients and causing disease . Many bacteria can also gain DNA from other bacteria , and the genes encoded within the new DNA can help them to live with other organisms . This can start the bacteria on an evolutionary path to becoming beneficial or harmful . Rhodococcus are bacteria that live in association with many species of plants , including trees . Most are harmless but some cause disease . Plants infected with harmful Rhodococcus can show deformed growth , which causes major losses to the nursery industry . Savory , Fuller , Weisberg et al . set out to understand how disease-causing Rhodococcus are introduced into nurseries , if they are transferred between nurseries , whether they persist in nurseries , and how to limit their spread . It turns out that harmless Rhodococcus are beneficial to plants . However , if these harmless bacteria gain a certain DNA molecule – called a virulence plasmid – they can convert into harmful bacteria . Further analysis showed that some nurseries repeatedly acquired the harmful bacteria . The pattern of affected nurseries suggested that some might have purchased diseased plants from a common provider . In other cases , the sources remained a mystery . Savory et al . also report that , contrary to previous findings , there is no evidence to support the diagnosis that Rhodococcus without a virulence plasmid are responsible for an unusual growth problem that has plagued the pistachio industry . In recent years , this incorrect diagnosis led to trees being unnecessarily destroyed , worsening the economic losses . These findings suggest that genes moving between bacteria can dramatically change how those bacteria interact with the organisms in which they live . It needs to be shown whether this is an exceptional process , unique to only certain groups of bacteria , or if it is more widespread in nature . These findings could inform future disease management strategies to better protect agricultural systems . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"epidemiology",
"and",
"global",
"health"
] | 2017 | Evolutionary transitions between beneficial and phytopathogenic Rhodococcus challenge disease management |
Spontaneous fluctuations of neural activity may explain why sensory responses vary across repeated presentations of the same physical stimulus . To test this hypothesis , we recorded electroencephalography in humans during stimulation with identical visual stimuli and analyzed how prestimulus neural oscillations modulate different stages of sensory processing reflected by distinct components of the event-related potential ( ERP ) . We found that strong prestimulus alpha- and beta-band power resulted in a suppression of early ERP components ( C1 and N150 ) and in an amplification of late components ( after 0 . 4 s ) , even after controlling for fluctuations in 1/f aperiodic signal and sleepiness . Whereas functional inhibition of sensory processing underlies the reduction of early ERP responses , we found that the modulation of non-zero-mean oscillations ( baseline shift ) accounted for the amplification of late responses . Distinguishing between these two mechanisms is crucial for understanding how internal brain states modulate the processing of incoming sensory information .
The brain generates complex patterns of neural activity even in the absence of sensory input or tasks . This activity is referred to as ‘spontaneous’ , ‘endogenous’ , or ‘prestimulus’ , as opposed to activity triggered by and , thus , following sensory events . Numerous studies have shown that such spontaneous neural activity can explain a substantial amount of the trial-by-trial variability in perceptual and cognitive performance ( e . g . , Haegens et al . , 2011; Myers et al . , 2014; Iemi et al . , 2017 ) and that abnormalities in spontaneous neural activity serve as biomarkers for neuropathologies ( e . g . , in schizophrenia , Parkison’s disease , and Autism Spectrum Disorder; Uhlhaas and Singer , 2010; McCarthy et al . , 2011; Simon and Wallace , 2016 ) and aging ( Voytek et al . , 2015 ) . Yet , the mechanisms by which spontaneous neural activity impacts the processing of sensory information remain unknown . This study aims to clarify how spontaneous fluctuations of prestimulus brain states , reflected by the power of low-frequency oscillations in the α- and β-bands ( 8–30 Hz ) , affect the trial-by-trial variability in the amplitude of sensory event-related potentials ( ERPs ) . The mechanisms underlying the effect of prestimulus power on ERP amplitudes are currently unknown , partly because previous studies have been inconsistent regarding the latency and even the directionality of this effect . Specifically , several studies found that prestimulus α-band power suppresses the amplitude of early ERP components ( <0 . 200 s: Rahn and Başar , 1993; Roberts et al . , 2014; Becker et al . , 2008; Başar and Stampfer , 1985; Jasiukaitis and Hakerem , 1988 ) , whereas other studies found that prestimulus α-band power enhances the amplitude of late ERP components ( >0 . 200 s: Dockree et al . , 2007; Becker et al . , 2008; Roberts et al . , 2014; Başar and Stampfer , 1985; Jasiukaitis and Hakerem , 1988; Barry et al . , 2000 ) . In this study , we addressed this issue by considering how prestimulus power affects the mechanisms of ERP generation at different latencies: namely , additive and baseline-shift mechanisms . ERP components occurring during the early time window ( <0 . 200 s ) are thought to be generated by an activation of sensory brain areas adding on top of ongoing activity ( additive mechanism: Bijma et al . , 2003; Shah et al . , 2004; Mäkinen et al . , 2005; Mazaheri and Jensen , 2006 ) . Invasive studies in non-human primates demonstrated that early ERP components are associated with an increase in the magnitude of multi-unit activity ( MUA ) in sensory areas ( Kraut et al . , 1985; Schroeder et al . , 1990; Schroeder et al . , 1991; Schroeder et al . , 1998; Shah et al . , 2004; Lakatos et al . , 2007 ) , presumably as a result of membrane depolarization due to excitatory synaptic activation ( Schroeder et al . , 1995 ) . Non-invasive studies in humans showed that early ERP components ( e . g . , C1 ) are associated with an increase in the hemodynamic fMRI signal in visual areas ( Di Russo et al . , 2002 ) , which may reflect an additive sensory response . Low-frequency neural oscillations are thought to set the state of the neural system for information processing ( Klimesch et al . , 2007; Jensen and Mazaheri , 2010; Mathewson et al . , 2011; Spitzer and Haegens , 2017 ) , which in turn may modulate the generation of additive ERP components . In particular , numerous studies have demonstrated that states of strong ongoing α- and β-band oscillations reflect a state of functional inhibition , indexed by a reduction of neuronal excitability ( e . g . , single-unit activity: Haegens et al . , 2011; Watson et al . , 2018; MUA: van Kerkoerle et al . , 2014; Becker et al . , 2015; ongoing γ-band power: Spaak et al . , 2012; hemodynamic fMRI signal: Goldman et al . , 2002; Becker et al . , 2011; Scheeringa et al . , 2011; Harvey et al . , 2013; Mayhew et al . , 2013 ) and of subjective perception ( e . g . , conservative perceptual bias: Limbach and Corballis , 2016; Iemi et al . , 2017; Craddock et al . , 2017; Iemi and Busch , 2018; lower perceptual confidence: Samaha et al . , 2017b; lower visibility ratings: Benwell et al . , 2017 ) . Accordingly , prestimulus low-frequency oscillations in the α- and β-bands may exert an inhibitory effect on the additive mechanism of ERP generation: that is , states of strong prestimulus power may suppress the activation of sensory areas , resulting in an attenuation of the amplitude of additive ERP components during the early time window ( Figure 1 ) . While early ERP components are likely to be generated primarily through an additive mechanism , late ERP components can have contributions from both additive and baseline-shift mechanisms ( where ‘baseline’ denotes the mean offset of the signal , rather than prestimulus activity ) . According to the baseline-shift account , the slow ERP component , which becomes visible during the late time window ( >0 . 200 s ) , is generated by a modulation of ongoing oscillatory power , rather than by an additive response ( Nikulin et al . , 2007; Nikulin et al . , 2010a; Mazaheri and Jensen , 2008; Mazaheri and Jensen , 2010; van Dijk et al . , 2010 ) . The effect of the baseline-shift mechanism on the relationship between prestimulus power and ERP amplitude has never been tested . In fact , it is generally assumed ( and not even questioned ) that neural oscillations are symmetrical around the zero line of the signal . Accordingly , trial averaging is expected to eliminate non-phase-locked oscillations due to phase cancellation ( assuming a random phase distribution over trials ) , thereby resulting in a signal baseline with a zero mean . It follows that a modulation of zero-mean oscillations by stimuli/tasks leaves the signal baseline unaffected ( Figure 1A ) . In contrast to this traditional view , recent studies ( Nikulin et al . , 2007; Nikulin et al . , 2010a; Mazaheri and Jensen , 2008; van Dijk et al . , 2010; Schalk , 2015; Cole and Voytek , 2017 ) proposed that neural oscillations do not vary symmetrically around the zero line of the signal , but rather around a non-zero offset/mean ( Figure 1B ) . Accordingly , trial averaging does not eliminate non-phase-locked oscillations with a non-zero mean . As a consequence , any amplitude modulation of oscillations with a non-zero mean is expected to change the signal baseline ( baseline shift ) , and will therefore affect the ERP amplitude . Specifically , during the event-related desynchronization ( ERD ) of low-frequency oscillations , the power suppression is expected to cause a slow shift of the signal baseline toward the zero line . Subtracting the prestimulus non-zero baseline from the post-stimulus signal creates a slow shift , which mirrors the spatio-temporal profile of the ERD . In particular , an ERD of oscillations with a negative non-zero-mean is expected to generate an upward slow shift of the ERP ( and viceversa ) . The idea that the ERD contributes to the generation of the slow ERP component , implies that the larger the ERD , the stronger the slow ERP component . Accordingly , we predicted that states of strong prestimulus power would yield a strong ERD ( Min et al . , 2007; Becker et al . , 2008; Tenke et al . , 2015; Benwell et al . , 2018 ) , resulting in an enhancement of the slow ERP component during the late time window . To summarize , states of strong prestimulus power are expected: ( 1 ) to suppress the amplitude of additive ERP components during the early time window ( via functional inhibition ) ; and ( 2 ) to amplify the late ERP component , generated by an event-related modulation of non-zero-mean oscillations ( via baseline shift ) . To test these predictions , we recorded electroencephalography ( EEG ) in human participants during rest and during stimulation with identical high-contrast checkerboard stimuli and analyzed the relationship between ERPs , ongoing and event-related oscillations . We find that the effects of prestimulus power on early and late ERP components are consistent with the functional inhibition and baseline-shift accounts , respectively . Taken together , these results largely resolve apparent inconsistencies in previous literature and specify two distinct mechanisms by which prestimulus neural oscillations modulate visual ERP components .
The experiment included stimulation trials with high-contrast checkerboard stimuli presented in the lower ( LVF; Figure 2A , left panel ) or upper ( UVF; Figure 2A , middle panel ) visual field with equal probability , and fixation-only trials without any checkerboard stimulus ( Fix; Figure 2A , right panel ) . All trials included a change of the central fixation mark at the time of stimulus presentation ( see Materials and methods for details ) . For each participant we quantified the ERP at the electrode with peak activity between 0 . 055 and 0 . 090 s after stimulus onset , reflecting the C1 component which indicates initial afferent activity in primary visual cortex ( Di Russo et al . , 2002 ) . On stimulation trials , the C1 component peaked on average at 0 . 079 s ( SEM = 0 . 001 ) and at 0 . 078 s ( SEM = 0 . 001 ) for LVF and UVF stimuli respectively , with a maximum at occipito-parietal electrodes ( Figure 2B , left and middle panels ) . The comparison of C1 amplitudes at individual peak electrodes on LVF ( M = 10 . 157 μV ; SEM = 0 . 918 ) and UVF ( M = −10 . 567 μV; SEM = 1 . 058 ) trials revealed the expected polarity reversal , confirming that this component originates from initial afferent activity in early visual areas ( Di Russo et al . , 2002; Di Russo et al . , 2003; Bao et al . , 2010 ) . Following the C1 , we observed a N150 component peaking between 0 . 100 and 0 . 200 s relative to stimulus onset , with an occipital topography and a consistently negative polarity for both LVF and UVF stimuli . The N150 was followed by a late deflection in the time range between 0 . 200 and 0 . 600 s relative to stimulus onset , with a parietal topography and consistent positive polarity for both LVF and UVF stimuli . As expected , on Fix trials no C1 and N150 components were detected in the ERP at individual C1-peak electrodes for LVF and UVF stimuli ( mean amplitude in the C1 time window: −0 . 044 μV , SEM = 0 . 196 ) . Fix trials showed a late positive deflection with similar timing and topography as on stimulation trials ( Figure 2B , right panel ) . For each participant we estimated the event-related synchronization ( ERS ) and desynchronization ( ERD ) at frequencies between 5 and 30 Hz and at each electrode and time point of the post-stimulus window ( 0–0 . 900 s ) . For the group-level statistical analysis , we used cluster permutation tests to determine at which time , frequency and electrode the ERS/ERD was significantly different from 0 across participants . On LVF trials , the statistical test revealed a negative cluster ( p<0 . 001 ) indicating ERD , spanning time points from 0 to 0 . 900 s relative to stimulus onset , frequencies between 6 and 30 Hz , and all 64 electrodes ( Figure 2C , left panel ) . The most negative t-statistic was found at 20 Hz , 0 . 234 s , and at electrode P7 . A positive cluster ( p=0 . 041 ) indicating ERS spanned time points from 0 to 0 . 900 s relative to stimulus onset , frequencies between 5 and 8 Hz , and all 64 electrodes . The most positive t-statistic was found at 5 Hz , 0 . 097 s , and at electrode P7 . On UVF trials , the statistical test revealed a negative cluster ( p<0 . 001 ) indicating ERD , spanning time points from 0 to 0 . 900 s relative to stimulus onset , frequencies between 6 and 30 Hz , and all 64 electrodes ( Figure 2C , middle panel ) . The most negative t-statistic was found at 20 Hz , 0 . 234 s , and at electrode P7 . This test also found two positive clusters indicating ERS . The first positive cluster ( p=0 . 032 ) spanned time points from 0 to 0 . 900 s relative to stimulus onset , frequencies between 5 and 8 Hz , and all 64 electrodes . Within this cluster , the most positive t-statistic was found at 5 Hz , 0 . 152 s , and at electrode P2 . The second positive cluster ( p=0 . 034 ) spanned time points from 0 . 49 to 0 . 900 s relative to stimulus onset , frequencies between 13 and 30 Hz , and 62 electrodes . Within this cluster , the most positive t-statistic was found at 17 Hz , 0 . 648 s , and at electrode FT7 . For Fix trials , the statistical test revealed one negative cluster ( p<0 . 001 ) indicating ERD , spanning time points from 0 to 0 . 900 s relative to fixation-target onset , frequencies between 5 and 30 Hz , and all 64 electrodes ( Figure 2C , right panel ) . On Fix trials , the most negative t-statistic was found at 20 Hz , 0 . 214 s , and at electrode PO8 . The functional inhibition account predicts that states of strong prestimulus power reflect neural inhibition , resulting in reduced amplitude specifically of early ERP components generated by the additive mechanism . To test for this mechanism , we classified trials in five bins based on single-trial estimates of oscillatory power at each electrode and each frequency between 5 and 30 Hz averaged over the 0 . 500 s prestimulus window ( i . e . , total-band power , Figure 3C ) and compared the ERP amplitude between the strongest and weakest power bin . These total-band power estimates reflect a mixture of both periodic ( i . e . , oscillations ) and aperiodic signals ( i . e . , 1/f ‘background’ noise; Voytek et al . , 2015 , see Figure 3C/D and Appendix 1—figure 1 ) . Therefore , to determine whether ERP differences between total-band power bins were indeed due to oscillatory activity , we repeated the binning analysis using single-trial estimates of the periodic signal ( i . e . , aperiodic-adjusted power; see Materials and methods for details ) . On LVF trials , a statistical test comparing ERP amplitudes during the early time window ( <0 . 200 s ) on trials with strong vs weak prestimulus power found a significant negative cluster ( p=0 . 015 ) . This indicates that the ERP amplitude in a time range containing the C1 ( 0 . 043 to 0 . 121 s ) was weaker ( i . e . , less positive ) on trials with strong prestimulus power between 8 and 28 Hz , and at all 64 electrodes ( Figure 3A , left panel ) . The most negative t-statistic was found at 10 Hz , 0 . 078 s , and at electrode P4 ( t ( 23 ) =−6 . 030 ) . At this time-frequency-electrode point , this effect was corroborated by the aperiodic-adjusted analysis ( t ( 23 ) =−3 . 634; FDR-corrected p=0 . 003 ) , demonstrating that the ERP amplitude was indeed modulated by oscillatory power ( Appendix 1—figure 1C ) . Furthermore , the statistical test revealed a significant positive cluster ( p<0 . 001 ) , indicating that the ERP amplitude in a time range containing the N150 ( 0 . 090 to 0 . 200 s ) was weaker ( i . e . , less negative ) on trials with strong prestimulus power between 5 and 24 Hz , and at all 64 electrodes ( Figure 3A , left panel ) . The most positive t-statistic was found at 9 Hz , 0 . 145 ms , and at electrode P1 ( total-band power: t ( 23 ) =7 . 940; aperiodic-adjusted power: t ( 23 ) =5 . 381; FDR-corrected p<0 . 001 , Appendix 1—figure 1C ) . On UVF trials , the statistical test during the early time window revealed two significant positive clusters . The first cluster ( p<0 . 001 ) indicated that the ERP amplitude in a time range containing the C1 ( 0 . 02 to 0 . 113 s ) was weaker ( i . e . , less negative ) on trials with strong prestimulus power between 5 and 22 Hz , and at all 64 electrodes ( Figure 3A , middle panel ) . The most positive t-statistic was found at 13 Hz , 0 . 082 ms , and at electrode PO4 ( total-band power: t ( 23 ) =8 . 365; aperiodic-adjusted power: t ( 23 ) =2 . 315; FDR-corrected p=0 . 045 , Appendix 1—figure 1C ) . The second cluster indicated that the ERP amplitude in a time range containing the N150 ( 0 . 125 to 0 . 200 s ) was weaker ( i . e . , less negative ) on trials with strong prestimulus power between 5 and 25 Hz , and at all 64 electrodes . The most positive t-statistic was found at 10 Hz , 0 . 168 ms , and at electrode F2 ( total-band power: t ( 23 ) =8 . 544; aperiodic-adjusted power: t ( 23 ) =5 . 785; FDR-corrected p<0 . 001 , Appendix 1—figure 1C ) . For Fix trials , the statistical test during the early time window found no significant clusters ( <0 . 200 s , Figure 3A , right panel ) . Taken together , the results on early ERP components show that ERP amplitude on stimulation trials is attenuated during states of strong prestimulus power , regardless of the polarity of the components . This provides evidence for the functional inhibition mechanism underlying the modulatory effect of prestimulus power on the early ERP components . The baseline-shift account predicts that states of strong prestimulus oscillations with a non-zero mean are followed by strong post-stimulus power suppression ( ERD ) , resulting in a greater ERP amplitude specifically during the late time window . To demonstrate that the late ERP component was generated by a baseline shift , it is necessary to establish that: ( 1 ) the ongoing oscillations have a non-zero mean; ( 2 ) the non-zero mean and the late ERP component have opposite polarity; and that ( 3 ) the ERD magnitude is associated with the amplitude of the late ERP component . To demonstrate the non-zero mean property of ongoing oscillations , we computed the Baseline Shift Index ( BSI: Nikulin et al . , 2007; Nikulin et al . , 2010a ) and the Amplitude Fluctuation Asymmetry Index ( AFAI: Mazaheri and Jensen , 2008 ) . For each participant we estimated BSI and AFAI from resting-state oscillations for each electrode and frequency between 5 and 30 Hz , and then tested whether these indices were significantly different from 0 across participants using cluster permutation tests . For AFAI , we found a significant negative cluster ( p-value < 0 . 001 ) between 5 and 30 Hz with a parietal , occipital and central topography ( Figure 4A/B ) , indicating a stronger modulation of the troughs relative to the peaks , resulting in a negative mean . For BSI , we found a significant negative cluster ( p-value < 0 . 001 ) between 5 and 21 Hz with a parietal , occipital and central topography ( Figure 4A/B ) , similar to AFAI . A negative BSI indicates that strong oscillatory power corresponds to a more negative value of the low-pass filtered signal , as expected in the presence of oscillations with a negative mean . Taken together , the results on BSI and AFAI provide evidence for a non-zero ( negative ) mean of resting-state low-frequency oscillations . It is important to note that the late ERP component had a positive polarity in all trial types ( Figure 1B ) , which is expected as a result of ERD of oscillations with a negative mean ( Nikulin et al . , 2007; Nikulin et al . , 2010a; Mazaheri and Jensen , 2008 ) . Next , we analyzed the relationship between the ERD magnitude and the ERP amplitude during the late time window ( >0 . 200 s ) . We compared the amplitude of the late ERP between groups of trials of weak and strong ERD estimated at each frequency and electrode . For the group-level statistical analysis , we used cluster permutation tests to determine significant ERP differences across ERP time points , and ERD electrodes and frequencies . The statistical test on LVF trials revealed one significant positive cluster ( p<0 . 001 ) , indicating that the late ERP ( 0 . 200–0 . 900 s ) was greater during states of stronger ERD at frequencies between 5 and 30 Hz , and at all 64 electrodes ( Figure 5A/B , left panel ) . The most positive t-statistic was found at 8 Hz , 0 . 266 s , and at electrode POz . The statistical test on UVF trials revealed two significant positive clusters , indicating that the late ERP ( cluster 1: 0 . 336–0 . 900 s; cluster 2: 0 . 200–0 . 328 s ) was greater during states of strong ERD at frequencies between 5 and 30 Hz , and at all 64 electrodes ( Figure 5A/B , middle panel ) . The most positive t-statistic was found at 19 Hz , 0 . 258 s , and at T8 electrode . The statistical test on Fix trials revealed one significant positive cluster ( p<0 . 001 ) , indicating that the late ERP ( 0 . 488–0 . 900 s ) was greater during states of strong ERD at frequencies between 5 and 30 Hz , and at all 64 electrodes ( Figure 5A/B , right panel ) . The most positive t-statistic value was found at 13 Hz , 0 . 637 s , and at electrode PO8 . Taken together , these results show that states of stronger ERD were associated with a more positive deflection of the late ERP component , consistent with the baseline-shift account . After demonstrating that the ERD magnitude correlates with the late ERP amplitude , we determined whether the ERD magnitude was , in turn , related to prestimulus power , as predicted by the baseline-shift account . To this end , we compared the ERD magnitude ( at the subject-specific C1 electrode ) between groups of trials of weak and strong prestimulus power estimated for each frequency and electrode , separately for each trial type . We found that strong low-frequency prestimulus oscillations were associated with strong ERD in all trial types ( Figure 6 ) . Note that this result is expected due to the circularity in estimating ERD and pre-stimulus power . Interestingly , we found that the poststimulus power is similar across different prestimulus α-band bins ( Figure 6C ) , suggesting that the stimulus suppressed α-band oscillations of different magnitudes to approximately the same level . After demonstrating that the late ERP amplitude correlates with the ERD magnitude , and that the ERD magnitude in turn correlates with prestimulus power , we tested whether prestimulus power was directly correlated with the amplitude of the late ERP component . To this end , we compared the late ERP amplitude between groups of trials with weak and strong prestimulus power estimated for each frequency and electrode . For the group-level statistical analysis , we used cluster permutation tests to determine significant differences across ERP time points , prestimulus-power frequencies , and electrodes . The statistical test during the late time window revealed a significant , sustained , and positive cluster in each trial type , indicating that the late ERP component was amplified during states of strong prestimulus power . On LVF trials , the significant positive cluster ( p<0 . 001 ) spanned time points from 0 . 402 to 0 . 900 s , frequencies between 5 and 25 Hz , and all 64 electrodes ( Figure 3A/B , left panel ) . The most positive t-statistic was found at 11 Hz , 0 . 676 s , and at electrode POz ( total-band power: t ( 23 ) =6 . 769; aperiodic-adjusted power: t ( 23 ) =3 . 004; FDR-corrected p=0 . 014 , Appendix 1—figure 1C ) . On UVF trials , the significant positive cluster ( p=0 . 004 ) spanned time points from 0 . 445 to 0 . 900 s relative to stimulus onset , frequencies between 5 and 15 Hz , and all 64 electrodes ( Figure 3A/B , middle panel ) . The most positive t-statistic was found at 5 Hz , 0 . 648 s , and at electrode CP1 ( total-band power: t ( 23 ) =7 . 600; aperiodic-adjusted power: t ( 23 ) =2 . 528; FDR-corrected p=0 . 037 , Appendix 1—figure 1C ) . On Fix trials , the significant positive cluster ( p<0 . 001 ) spanned time points from 0 . 484 to 0 . 900 s relative to fixation-target onset , frequencies between 5 and 23 Hz , and all 64 electrodes ( Figure 3A/B , right panel ) . The most positive t-statistic was found at 7 Hz , 0 . 781 s , and at electrode POz ( total-band power: t ( 23 ) =7 . 528; aperiodic-adjusted power: t ( 23 ) =1 . 881; FDR-corrected p=0 . 109 , Appendix 1—figure 1C ) . Taken together , these results show that: ( 1 ) the late ERP component is generated by a baseline shift during the ERD of non-zero mean oscillations; ( 2 ) states of strong prestimulus power are followed by strong ERD , which manifests as an enhancement of the late ERP component . To ensure that the relationship between prestimulus power and ERP amplitude was not simply an epiphenomenon of time-varying variables such as sleepiness , we analyzed the scores of a subjective sleepiness questionnaire that participants filled in at the end of every block ( Karolinska Sleepiness Scale , KSS: Kaida et al . , 2006 ) . First , at the single-subject level , we computed the correlation between prestimulus oscillatory power and KSS rating . At the group-level , we tested whether these correlations were significantly different from 0 . We found significant positive clusters for frequencies below 18 Hz and with a widespread topography in each trial type ( Appendix 1—figure 2A ) . This result indicates that the stronger the prestimulus power , the higher the subjective sleepiness . Within each participant we removed the contribution of sleepiness to the trial-by-trial estimates of oscillatory power and repeated the power-ERP analysis with these corrected power estimates . The results of this re-analysis ( Appendix 1—figure 2B ) were virtually identical to the ones obtained with raw power estimates ( Figure 3A ) , suggesting that the effects we observed were not confounded by sleepiness .
The results on the early ERP components in this study ( <0 . 200 s: C1/N150 ) confirm and extend findings from past literature in the visual and auditory modalities . Specifically , previous studies in the visual modality found a negative relationship between prestimulus α-band power and the amplitude of the visual N1P2 ( i . e . , amplitude difference between N1 and P2 components: Rahn and Başar , 1993 ) , N1 ( Roberts et al . , 2014 ) and N175 components ( Becker et al . , 2008 ) . A similar pattern of results was found for the N100 in the auditory modality ( Başar and Stampfer , 1985; Jasiukaitis and Hakerem , 1988 ) . It is important to note that previous results ( e . g . , Başar and Stampfer , 1985; Becker et al . , 2008 ) that have been used to support the functional inhibition account could actually have been caused by a baseline shift . In the current study , we leverage the fact that the C1 has a well-known polarity reversal as a function of the visual field of the stimulus . By showing that the absolute amplitude of the C1 component is diminished by stronger prestimulus power , regardless of polarity , we can rule out a baseline shift which would affect both polarities in the same direction ( e . g , a net increase or decrease of voltage ) . Additionally , previous results may have also have been due to ( 1 ) fluctuations of the 1/f aperiodic signal ( which affect total-band power estimates , see Appendix 1—figure 1 ) , or ( 2 ) fluctuations of sleepiness ( which affect both oscillatory power and ERP amplitude , see Appendix 1—figure 2 ) . In the current study , we confirmed that the early ERP amplitude was indeed reduced by oscillatory power , rather than just the 1/f aperiodic signal , and that this effect was not an epiphenomenon due to sleepiness . Taken together , these findings provide the first conclusive evidence for the functional inhibition effect of prestimulus oscillations on the early ERP amplitude . Unlike in the visual and auditory modality , the relationship between prestimulus power and early ERP components in the somatosensory modality ( e . g . , N1 ) may be non-linear ( inverted U-shaped: Zhang and Ding , 2010; Ai and Ro , 2014; Forschack et al . , 2017 ) or vary across early components ( i . e . , negative for M50 and positive for M70 , P35 and P60: Jones et al . , 2009; Nikouline et al . , 2000 ) . Similarly , the relationship between prestimulus power and somatosensory perceptual performance has been found to have an inverted U-shape ( Linkenkaer-Hansen et al . , 2004 ) , or to be linear ( Haegens et al . , 2011; Craddock et al . , 2017 ) . Taken together , these findings suggest that in the somatosensory domain distinct functional mechanisms may map onto low-frequency oscillations . Importantly , several studies report a positive relationship between prestimulus α-band power and the amplitude of the visual N100 ( Jansen and Brandt , 1991; Brandt , 1997 ) , N1P2 ( Brandt et al . , 1991; Brandt and Jansen , 1991; Barry et al . , 2000 ) and P200 ( Jansen and Brandt , 1991 ) . Thus , these studies appear inconsistent with the current results and other studies in the visual and auditory modality . However , a direct comparison is difficult for several reasons . First , some of these studies delivered visual stimuli to participants with eyes closed ( Brandt and Jansen , 1991; Brandt and Jansen , 1991; Brandt , 1997; Jansen and Brandt , 1991 ) . Instead , the majority of previous studies ( including ours ) delivered visual stimuli to participants with eyes open . It is known that oscillatory power in low frequencies has different spectral ( Barry et al . , 2007 ) and functional ( Kaida et al . , 2006 ) properties depending on whether subjects’ eyes are open or closed; thus , these inconsistencies may be due to the eyes-open/closed difference . Second , unlike our study , which analyzed a broad frequency band , 64 electrodes , and an extensive post-stimulus time window ( 0–0 . 900 s ) , most previous studies only analyzed a narrow frequency band , few electrodes and a single time point . Therefore , it is possible that the inconsistent effects in previous studies were due to this selective ( and under-sampled ) analysis of EEG data . Third , some previous studies lack: ( 1 ) sufficient description of the EEG analysis ( e . g . , Brandt , 1997 ) , ( 2 ) adequate statistical power ( due to low number of participants or trials: e . g Brandt and Jansen , 1991; Brandt and Jansen , 1991; Brandt , 1997 ) , and ( 3 ) quantitative statistical testing ( Brandt , 1997 ) . Consequently , this makes it difficult to compare these studies to the current one . The present results have implications for the role of low-frequency oscillations in perceptual decision-making and in the top-down control over sensory processing ( e . g . , by spatial attention ) . In fact , numerous studies have found that weak prestimulus α-band power increases observers’ hit rates for near-threshold stimuli ( Ergenoglu et al . , 2004; Zhang and Ding , 2010; Chaumon and Busch , 2014 ) . More recently , studies have demonstrated that this effect is not due to more accurate perception , but rather to a more liberal detection bias ( Limbach and Corballis , 2016; Iemi et al . , 2017; Craddock et al . , 2017; Iemi and Busch , 2018 ) and a concomitant increase in confidence ( Samaha et al . , 2017b ) and subjective visibility ( Benwell et al . , 2017 ) . Unfortunately , conventional signal detection theory cannot be used to distinguish between alternative kinds of bias ( Morgan et al . , 2013; Witt et al . , 2015 ) . Specifically , a change in bias could be due to a change in the observer’s deliberate decision strategy without any change in sensory processing ( decision bias ) . Alternatively , a change in bias could be due to a change in the subjective appearance of stimuli ( perceptual bias ) : liberal perceptual bias during states of weak prestimulus power could result from increased neural excitability amplifying both neural responses to sensory stimuli ( thereby increasing hit rates ) and responses to noise ( thereby increasing false alarm rates ) . Interestingly , the present finding that the C1 is amplified during states of weak prestimulus power , indicates that even the earliest visual evoked responses are modulated by prestimulus oscillations . Even though we could not study an equivalent amplification of responses to noise using the present paradigm , this finding supports a perceptual bias mechanism more than a decision bias mechanism . Furthermore , many experiments have noted a relationship between the topography of α-band power and the focus of covert spatial attention ( e . g . , Samaha et al . , 2016 ) . However , considerable debate exists as to whether this preparatory α-band modulation ( and hence spatial attention ) is capable of modulating feed-forward visual input ( e . g . , the C1 component ) . Our results show a clear impact of spontaneous α and β-band power on C1 amplitudes , supporting the idea that attention-related low-frequency modulation can affect the earliest stages of sensory processing . However , it is possible that attention-related and spontaneous oscillations have different effects on the amplitude of the C1 component . This question is a candidate for future investigation , ideally by using stimuli such as those employed here , which generated robust C1 responses . In this study we demonstrated that the late component of the visual ERP was generated by a modulation of non-zero mean oscillations via baseline shift . There are four requirements to demonstrate the baseline-shift mechanism . First , the ongoing oscillations must have a non-zero mean . To this end , we estimated the non-zero-mean property of resting-state oscillations using AFAI and BSI . This analysis revealed that α- and β-band oscillations were characterized by a negative non-zero-mean . The frequencies and electrodes of the significant cluster for AFAI were more extended relative to the cluster for BSI . This could be due to the fact that , unlike BSI , AFAI is biased by harmonics and thus it reflects both non-zero mean oscillations and the ‘comb-shape’ of oscillations , which may yield amplitude asymmetries even when the signal has a zero mean ( Nikulin et al . , 2010a; Nikulin et al . , 2010b ) . Thus , AFAI is expected to be susceptible to more asymmetry-related features with larger spatial and spectral distribution compared to BSI . Second , sensory stimuli must modulate the amplitude of ongoing oscillations . To test this requirement , we estimated the power modulation in the post-stimulus window relative to a prestimulus baseline ( i . e . , event-related oscillations: ERD/ERS ) . We observed a strong ERD in frequencies between 6 and 30 Hz in all three trial types . On Fix trials there were no robust early evoked components due to the lack of strong visual input , yet we observed an ERD following the same spatio-temporal dynamics as on stimulation trials ( though of a lesser magnitude ) . In addition to the ERD , we also observed a strong ERS below 8 Hz on stimulation trials ( but not on Fix trials ) possibly reflecting a leakage from the robust evoked components measured during the early time window . Third , the non-zero mean and the late ERP must have opposite polarity in case of ERD . Consistent with this requirement , our results showed that oscillations with a negative non-zero mean were associated with a late ERP component of positive polarity . Fourth , ERD magnitude must correlate with the amplitude of the late ERP component . Our results indicated that strong ERD of non-zero mean oscillations was associated with enhanced ERP amplitude during the late time window . Importantly , the late ERP component was characterized by a topography and time-course similar to the ones of the ERD , consistent with Mazaheri and Jensen ( 2008 ) . In sum , these findings confirm the four requirements necessary to demonstrate the baseline-shift mechanism for the generation of the late ERP component . The baseline-shift account predicts that stronger ERD occurs during states of stronger prestimulus power , which generates a greater baseline shift . In the case of negative non-zero-mean oscillations , this process results in an enhancement of the late ERP component with positive polarity . To test this prediction , we analyzed how prestimulus power is related to the ERD magnitude , and in turn to the amplitude of the late ERP component . We found that trials with strong prestimulus power were related to strong ERD magnitude , consistent with previous studies ( Min et al . , 2007; Becker et al . , 2008; Tenke et al . , 2015; Benwell et al . , 2018 ) . Due to circularity in these measures ( i . e . , ERD is computed with prestimulus power ) , the statistical estimates of this relationship are inflated . However , this pattern of results corroborates the prediction of the baseline-shift account . Specifically , we found that α-band power is reduced to approximately the same level regardless of prestimulus power ( Figure 6C ) . Accordingly , whereas the average prestimulus voltage is expected to differ between different prestimulus power bins due to the non-zero-mean property of neural oscillations , the average post-stimulus voltage in the late window is expected to be the same regardless of prestimulus power ( Figure 1B , upper panel ) . The baseline-shift account predicts that subtracting a stronger prestimulus signal ( strong power bin ) yields a stronger shift of the EEG signal from the prestimulus baseline , and thus a stronger late ERP component ( Figure 1B , lower panel ) . Consistent with our prediction , we also found a positive relationship between prestimulus power and the amplitude of the late ERP component . These results confirm and extend previous findings in visual and auditory modalities . Specifically , in the visual modality prestimulus α-band power was found to be positively correlated with the ERP amplitude in a late time window starting from 0 . 200 s relative to stimulus onset ( 0 . 550–0 . 800 s: Dockree et al . , 2007; 0 . 220–0 . 310 s: Becker et al . , 2008; 0 . 400 s: Roberts et al . , 2014 ) . A similar pattern of results was found on late ERP components in the auditory modality ( 0 . 250–0 . 800 s: Jasiukaitis and Hakerem , 1988; 0 . 400 s: Başar and Stampfer , 1985; 0 . 200–0 . 500 s: Barry et al . , 2000 ) . Previous studies ( e . g . , Barry et al . , 2000 ) were unable to explain the positive relationship between α-power and ERP amplitude , which appeared inconsistent with the functional inhibition account ( Haegens et al . , 2011 ) . Therefore , this study resolves this apparent inconsistency in previous literature , by demonstrating that this positive relationship can be accounted for by the baseline-shift mechanism , rather than functional inhibition . It may seem surprising that the effects of prestimulus power on the late ERP occurred after the peak of the classically-defined slow component at approximately 0 . 300 s relative to stimulus onset ( Nikulin et al . , 2007; Mazaheri and Jensen , 2008 ) . While early ERP components are likely generated primarily through the additive mechanism ( because ERD is negligible in this time window ) , late ERP components can have a contribution from both additive and baseline-shift mechanisms . Functional inhibition of additive components in the initial part of the late time window might have canceled the amplification effect due to the baseline shift . This cancellation might explain the lack of a significant effect at the peak of the late component . In contrast , the ERP during the later time window ( >0 . 400 s ) is more likely to show primarily baseline-shift-generated components and thus is more susceptible to the amplification effect of prestimulus power . We conclude that the positive modulation of the late ERP component is directly produced by the modulation of ERD magnitude as a function of prestimulus power . It is important to note that previous results on the late ERP component ( e . g . , Dockree et al . , 2007; Becker et al . , 2008 ) may have been influenced by ( 1 ) fluctuations of the 1/f aperiodic signal ( which affect total-band power estimates , see Appendix 1—figure 1 ) , or ( 2 ) fluctuations of sleepiness ( which affect both oscillatory power and ERP amplitude , see Appendix 1—figure 2 ) . In the current study , we confirmed that the late ERP amplitude was indeed amplified by oscillatory power , rather than just the 1/f aperiodic signal , and that this effect was not an epiphenomenon due to sleepiness . This provides the first evidence that the effect of prestimulus oscillations on the late ERP component is due to the mechanism of baseline shift . It is important to highlight that , while the functional inhibition account describes a ( proposed ) physiological mechanism , the baseline-shift account describes an effect that is largely the consequence of specific properties of the signal and the way we analyze it . That is , baseline shift is a result of a combination of preconditions including signal properties ( non-zero mean ) , the occurrence of an ERD , as well as conventional signal processing procedures ( i . e . , baseline correction ) . The modulation of late responses predicted by a baseline shift can only exist if these preconditions are met , while the functional inhibition account generalizes to cases involving zero-mean oscillations , does not depend on the presence of an ERD , and can be established using different measures of brain activity ( i . e . , not limited to ERPs ) . The results of this study demonstrate a modulatory role of low-frequency oscillations on ERP amplitude . Both effects of prestimulus oscillations on early and late ERP components were characterized by a broad frequency range spanning the α- and β-band . Likewise , the ERD and the non-zero-mean property of oscillations were found for both the α-band and β-band . Specifically , α-band ERD was sustained in time while β-band ERD was more transient , consistent with previous studies ( e . g . , Salenius et al . , 1997 ) . This suggests that β-band ERD may also reflect the generation of the late ERP component . One possible explanation for this multi-band effect can be the non-sinusoidal nature of neural oscillations ( e . g . , ‘comb-shape’: Cole and Voytek , 2017 ) , which applies to both α- and β-bands . In this case the event-related power modulation would similarly affect α- and β-band activity . Because of such comodulation , baseline-shifts associated with α-band oscillations would also appear for β-band oscillations , resulting in similar BSI and AFAI for both frequency bands ( Nikulin et al . , 2010a ) . The β-band effect may seem surprising given that the majority of past literature focused solely on the α-band due to its high signal-to-noise ratio compared to other frequencies . However , the broad frequency range of the effects reported in this study is in line with studies reporting a temporal and spatial co-modulation of α- and β-band oscillations ( Bastos et al . , 2015; Lakatos et al . , 2016; Michalareas et al . , 2016 ) . It is also consistent with recent studies reporting a similar relationship between α- and β-band prestimulus power , perceptual reports ( Benwell et al . , 2017; Iemi et al . , 2017; Samaha et al . , 2017a; Samaha et al . , 2017b; Iemi and Busch , 2018 ) and firing rate ( Watson et al . , 2018 ) . Accordingly , it has been proposed that β-band oscillations exert an inhibitory function , similar to α-band oscillations ( Spitzer and Haegens , 2017; Shin et al . , 2017; Kilavik et al . , 2013 ) . In this study we considered the additive ( Bijma et al . , 2003; Shah et al . , 2004; Mäkinen et al . , 2005; Mazaheri and Jensen , 2006 ) and baseline-shift ( Nikulin et al . , 2007; Mazaheri and Jensen , 2008 ) mechanisms for the generation of early and late ERP components ( Bijma et al . , 2003; Shah et al . , 2004; Mäkinen et al . , 2005; Mazaheri and Jensen , 2006 ) , respectively . In addition to these mechanisms , some studies have proposed that the ERP can be generated by a reorganization of ongoing oscillations via phase reset ( Sayers et al . , 1974; Makeig et al . , 2002; Klimesch et al . , 2004; Gruber et al . , 2005; Fell et al . , 2004; Fuentemilla et al . , 2006; Hanslmayr et al . , 2007; Sauseng et al . , 2007 ) . According to the phase-reset account , the phases of ongoing oscillations are aligned ( i . e . , phase-reset ) by the stimulus; as a consequence , averaging these phase-locked oscillations across trials does not lead to phase cancellation in the post-stimulus window , resulting in the generation of ERP components . Specifically , the phase reset of an oscillation at a particular frequency is expected to generate a component with similar frequency characteristics: for example , the α-band phase reset is thought to generate early ERP components of the same polarity with an inter-peak latency at approximately 100 ms ( as the C1 and N150 on UVF trials , see Figure 2B , middle panel ) ( Hanslmayr et al . , 2007; Sauseng et al . , 2007 ) . Several studies therefore proposed that ERP components at different latencies ( with different frequency characteristics ) are generated by a phase reset of α- and β-band oscillations ( P1 and N1: Klimesch et al . , 2004; Makeig et al . , 2002; Gruber et al . , 2005 ) or δ- and θ-band oscillations ( P300: Fell et al . , 2004 ) . However , please note that , while phase reset of α-band oscillations may explain the generation of the early ERP components with positive polarity on UVF trials , it cannot account for the opposite polarity of the C1 component on LVF trials ( see Figure 2B ) . Although our experiment was not designed to test the phase-reset hypothesis , we see two possible predictions that a phase-reset account could make on the relationship between prestimulus oscillatory power and ERP amplitude ( Hanslmayr et al . , 2007; Sauseng et al . , 2007 ) . On the one hand , it has been argued that phase reset in response to a stimulus can only occur if the oscillation already exists prior to the reset ( i . e . , during the prestimulus window ) . It follows that any ERP component generated by phase reset is likely to be absent during desynchronized states ( i . e . , weakest power bin ) ( Shah et al . , 2004; Sauseng et al . , 2007 ) . This suggests that trials with weakest prestimulus power may result in less prominent phase reset , which would manifest as a reduction of the ERP amplitude . Accordingly , we would expect a positive relationship between prestimulus oscillations in the α and β bands and the amplitude of the C1 and N150 components on UVF trials , which occur with an inter-peak latency of approximately 70–80 ms . Contrary to this prediction , our study showed that the C1 and N150 amplitudes on UVF trials were negatively correlated with prestimulus α- and β-band oscillations . In addition , we found that the amplitude of the slow ERP component was positively correlated with prestimulus α- and β-band oscillations . However , this correlation is unlikely to be accounted for by phase reset of α- and β-band oscillations because these rhythms are much faster than the one reflected in the slow ERP component ( i . e . , δ rhythm ) . On the other hand , one can argue that strong oscillations represent a state with pronounced neuronal synchronization that is not easily affected by weak sensory inputs , as also shown in previous modeling work ( Hansel and Sompolinsky , 1996 ) . Thus , during states of strong ongoing oscillations , phase-reset may be harder to be achieved and , consequently , is unlikely to result in ERP generation . Accordingly , this predicts an ERP attenuation by prestimulus power , consistent with our results during the early time window . It is worth noting that , in this study , it is difficult to distinguish whether this attenuation affects ERP components generated by additive or phase-reset mechanisms; invasive electrophysiological recordings allowing for higher spatial resolution might be required to address this particular question ( Hanslmayr et al . , 2007; Telenczuk et al . , 2010 ) . Regardless of the underlying mechanisms of ERP generation , our results during the early time window can be explained by the functional inhibition account . This study demonstrates that spontaneous fluctuations of oscillatory brain activity modulate the amplitude of visual ERP via two distinct mechanisms: ( 1 ) functional inhibition of the early additive ERP components and ( 2 ) baseline shift affecting the late ERP component . Therefore , these findings show that neural oscillations have concurrent opposing effects on ERP generation . Distinguishing between these effects is crucial for understanding how neural oscillations control the processing of incoming sensory information in the brain .
Previous studies on the relationship between neural oscillations and ERPs have typically reported samples of 7–19 participants ( e . g . , Jasiukaitis and Hakerem , 1988; Brandt and Jansen , 1991; Rahn and Başar , 1993; Nikulin et al . , 2007; Mazaheri and Jensen , 2008; Becker et al . , 2008; van Dijk et al . , 2010; Rajagovindan and Ding , 2011 ) . To ensure a robust estimate of our neurophysiological effect and account for potential missing data ( e . g . , due to artifacts ) , we recruited a larger sample of 27 participants ( mean age: 26 . 33 , SEM = 0 . 616; 14 females; three left-handed ) . All participants had normal or corrected-to-normal vision and no history of neurological disorders . Prior to the experiment , written informed consent was obtained from all participants . All experimental procedures were approved by the ethics committee of the German Psychological Society . Two participants were excluded before EEG preprocessing because of excessive artifacts . One participant was excluded after preprocessing because no C1 component could be detected , unlike the rest of the sample . A total of 24 participants were included in the analysis . The experiment was written in MATLAB ( RRID:SCR_001622 ) using the Psychophysics toolbox 3 ( RRID:SCR_002881: Brainard , 1997; Pelli , 1997 ) . The experiment included a resting-state session and a stimulation session , lasting approximately 1 . 5 hr including self-paced breaks . The resting-state session was divided in two recording blocks , each of which lasted 5 . 5 min , separated by a short self-paced break . In this session participants were required to keep their eyes open and fixate on a mark located at the center of the screen , to avoid movements and not to think of anything in particular . In the stimulation session , participants were presented with visual stimuli specifically designed to elicit a robust C1 component of the visual ERP . The C1 is described as the earliest component of the visual ERP with a peak latency between 0 . 055 and 0 . 09 s and an occipital topography . The C1 component is regarded as an index of initial afferent activity in primary visual cortex , because of its early latency and polarity reversal with reference to V1 anatomy ( Di Russo et al . , 2002; Di Russo et al . , 2003 ) . The stimuli consisted of full-contrast bilateral black-and-white checkerboard wedges . Each wedge corresponded to a radial segment of an annular checkerboard ( spatial frequency = 5 cycles per degree ) with inner and outer circle of 3 and 10° of eccentricity relative to a central fixation point , respectively . Each wedge covered 3 . 125% of the area of the annular checkerboard and spanned 11 . 25° of visual angle ( Vanegas et al . , 2013 ) . In each stimulation trial , a pair of wedges was presented bilaterally either in the UVF or LVF with equal probability . UVF and LVF stimulus positions were located at polar angles of 25° above and 45° below the horizontal meridian , respectively . These asymmetrical positions for UVF and LVF stimuli ensure a stimulation of primarily lower and upper banks of the calcarine fissure , respectively ( Aine et al . , 1996; Clark et al . , 1994; Di Russo et al . , 2002 ) , resulting in a polarity reversal of scalp potentials . A positive C1 component is obtained by LVF stimulation , while a negative C1 component is obtained by UVF stimulation ( Di Russo et al . , 2002; Di Russo et al . , 2003 ) . The stimuli were presented for a duration of 0 . 100 s ( Fu et al . , 2010; Ding et al . , 2014; Kelly et al . , 2008 ) at full contrast ( Hansen et al . , 2016; Vanegas et al . , 2013 ) on a gray background that was isoluminant relative to the stimuli’s mean luminance . The stimuli were presented at a viewing distance of 52 cm , on a cathode ray tube monitor operated at 100 Hz , situated in a dark , radio-frequency-interference shielded room . Throughout the experiment , fixation distance and head alignment were held constant using a chin rest . For each participant the stimulation session included 810 trials , divided into nine recording blocks . In each block , 60 trials contained stimuli in either LVF or UVF with equal probability ( stimulation trials ) , while 30 trials were stimulus-absent ( fixation-only trials ) . Trial type and stimulation field were randomized across trials within each block . To ensure that the participants maintained the gaze to the center , we included a discrimination task at the central fixation mark , similar to previous studies ( Di Russo et al . , 2002; Chen et al . , 2016 ) . On stimulation trials the central fixation mark turned into either one of two equally probable targets ( ‘>' or ‘<' ) during stimulus presentation for a duration of 0 . 100 s . On Fix trials , the change of the central fixation mark into the target occurred during a 0 . 100 s window between 1 . 8 and 2 . 4 s relative to trial onset . Note that , while the targets might have caused an involuntary shift of lateral attention , these effects would have cancelled out across trials because of the fully randomized presentation of the targets ( each recording block included 50% ‘<' targets and 50% ‘>' targets ) . Discrimination performance at or close to ceiling ( i . e . , 100% correct responses ) was expected if gaze was maintained on the central fixation mark . This task also ensured that the participants remained alert throughout the experiment . Mean accuracy in the fixation task was 94 . 85% ( SEM = 0 . 0109 ) and did not significantly differ between trial types ( p>0 . 05 ) , indicating that participants were able to maintain central fixation . Incorrect trials were discarded from further analysis ( mean = 41; SEM = 8 . 620 ) . On average we analyzed 761 ( SEM = 9 . 824 ) trials per participant . After target offset , the fixation mark was restored for a duration of 0 . 100 s . After this delay , the fixation mark turned into a question mark , which instructed the participants to deliver a response via a button press with their dominant hand . After the button press , the fixation mark was displayed again and a new trial started . The following stimulus presentation or fixation task occurred after a variable delay chosen from a uniform distribution between 1 . 8 and 2 . 4 s . In addition to the fixation task , to further prevent eye movements , all participants were trained prior to EEG recording to maintain fixation on the central mark and their fixation ability was monitored throughout the experiment using the electro-oculogram ( EOG ) . Moreover , we used a shape of the fixation mark specifically designed to maximize stable fixation ( Thaler et al . , 2013 ) . To control for an effect of sleepiness on the level of ongoing low-frequency power , we asked participants to report their level of sleepiness at the end of each block during resting-state and stimulation session . We used the Karolinska Sleepiness Scale ( KSS ) , which has been validated as an indicator of objective sleepiness ( Kaida et al . , 2006 ) . The KSS scale consists of a nine-point Likert-type scale ranging from 1 ( extremely alert ) to 9 ( very sleepy ) that represents the sleepiness level during the immediately preceding 5 minutes . The scale was presented on the screen at the end of every block and participants were instructed to report how alert they felt during the immediately preceding block by pressing the corresponding number on the keyboard ( 1–9 ) . After the button press , participants could take a self-paced break and the following block was initiated via button-press . EEG was recorded with a 64-channel Biosemi ActiveTwo system at a sampling rate of 1024 Hz . Electrodes were placed according to the international 10–10 system ( electrode locations can be found on the Biosemi website: https://www . biosemi . com/download/Cap_coords_all . xls ) . The horizontal and vertical electro-oculograms were recorded by additional electrodes at the lateral canthi of both eyes and below the eyes , respectively . As per the BioSemi system design , the Common Mode Sense ( CMS ) and Driven Right Leg ( DRL ) electrodes served as the ground . All scalp electrodes were referenced online to the CMS-DRL ground electrodes during recording . Electrode impedance was kept below 20 mV . The raw data was recorded with ActiView ( version 6 . 05 ) . The EEGLAB toolbox version 13 , running on MATLAB ( R2017b ) , was used to process and analyze the data ( Delorme and Makeig , 2004 ) . In both resting-state and stimulation sessions , the data were re-referenced to the mastoids and down-sampled to 256 Hz . In the resting-state session , data were epoched from 0 to 330 s relative to the start of the recording block . In the stimulation session , data were epoched from −1 . 6 to 1 . 3 s relative to the onset of the stimulus presentation on stimulation trials or to the onset of the fixation task on Fix trials . In both sessions , the data were then filtered using an acausal band-pass filter between 0 . 25 and 50 Hz . We manually removed gross artifacts such as eye blinks and noisy data segments . In the stimulation session , we discarded entire trials when a blink occurred within a critical 0 . 500 s time window preceding stimulus/fixation-target onset , to ensure that participants’ eyes were open at stimulus onset . Furthermore , we manually selected noisy channels on a trial-by-trial basis for spherical spline interpolation ( Perrin et al . , 1989 ) . We interpolated on average 8 channels ( SEM = 0 . 96 ) in 35 trials ( SEM = 6 . 71 ) . No channels were interpolated in the resting-state session . In both sessions , we transformed the EEG data using independent component analysis ( ICA ) , and then we used SASICA ( Semi-Automated Selection of Independent Components of the electroencephalogram for Artifact correction ) ( Chaumon et al . , 2015 ) to guide the exclusion of IC related to noisy channels and muscular contractions , as well as blinks and eye movements . On average , we removed 7 . 9 ( SEM 0 . 46 ) and 7 . 8 ( SEM 0 . 60 ) out of 72 ICs in the resting-state and stimulation session , respectively . The aim of this study was to examine the influence of prestimulus oscillatory power on ERP amplitude . We used visual stimuli to specifically elicit a robust C1 component of the visual ERP , which reflects initial afferent activity of the primary visual cortex ( Di Russo et al . , 2002 ) . For each participant we identified the electrode and time point with peak activity between 0 . 055 and 0 . 090 s after stimulus onset ( peak C1 activity: Di Russo et al . , 2002; Bao et al . , 2010 ) , separately for LVF and UVF trials . On Fix trials , no C1 component of the visual ERP is expected . To quantify the ERP for this trial type and to enable comparison with the stimulation trials , we averaged the EEG data across the subject-specific electrodes with peak C1-activity in the stimulation trials . We baseline corrected single-trial ERP estimates by subtracting the prestimulus signal baseline averaged across a 0 . 500 s prestimulus window . We used time-frequency analysis to obtain a measure of ongoing oscillatory power and to estimate event-related oscillations ( ERO ) . We first computed the stimulus-evoked , phase-locked activity ( ERP ) by averaging the EEG signal across trials . Then , we subtracted the average ERP from single-trial EEG signal . We applied this procedure separately for LVF , UVF , and Fix trials . This procedure ensures that the resulting ERO estimate does not contain stimulus-evoked activity ( Kalcher and Pfurtscheller , 1995 ) . Next , we applied a wavelet transform ( Morlet wavelets , 26 frequencies , frequency range: 5–30 Hz , number of cycles increasing linearly from 3 to 8 , time window: −1–1 . 3 s relative to stimulus onset ) to the EEG signal . This procedure was performed separately for each electrode and trial type ( LVF , UVF , and Fix ) . We then quantified ERO as follows: ( 1 ) ERO=Ppost-μ ( Ppre ) where Ppost is a the time course of post-stimulus oscillatory activity and μ ( Ppre ) is the average ongoing power in a prestimulus window between −0 . 600 and −0 . 100 s relative to stimulus onset . This window for baseline correction was chosen based on Mazaheri and Jensen ( 2008 ) to circumvent the temporal smearing due to the wavelet convolution . ERO<0 indicates the presence of an event-related desynchronization ( ERD ) , indicating stimulus-induced power attenuation . ERO>0 indicates the presence of an event-related synchronization ( ERS ) , indicating stimulus-induced power enhancement . This procedure was performed separately for each frequency , electrode and participant . In this study , we tested two potential mechanisms underlying the modulation of ERP amplitude by prestimulus power: namely , functional inhibition and baseline shift . Functional inhibition implies that the generation of early , additive ERP components is inhibited if stimulation occurs during a state of strong prestimulus activity . In other words , positive and negative early ERP components are expected to become less positive and less negative , respectively , during states of strong prestimulus power . The baseline-shift mechanism implies that states of strong prestimulus oscillations with a non-zero mean are followed by strong power suppression ( ERD ) , which in turn results in an enhancement of the late ERP component . Previous studies ( Nikulin et al . , 2007; Mazaheri and Jensen , 2008 ) proposed the following prerequisites for linking ERO to ERP generation ( baseline-shift mechanism ) : ( 1 ) the ongoing oscillations must have a non-zero mean; ( 2 ) sensory stimuli must modulate ERO magnitude; ( 3 ) the non-zero mean and the late ERP component must have opposite polarity in case of ERD; ( 4 ) ERO magnitude is associated with the amplitude of the late ERP component . The aim of the resting-state session was to estimate the non-zero-mean property of ongoing oscillations , which is known to be a critical requirement for generation of the late ERP via baseline shift ( Nikulin et al . , 2007; Mazaheri and Jensen , 2008 ) . To this end , we used two analytical methods: namely , the Baseline Shift Index ( Nikulin et al . , 2007 ) and the Amplitude Fluctuation Asymmetry Index ( Mazaheri and Jensen , 2008 ) . In each participant , we estimated BSI and AFAI for resting-state oscillations for each electrode and frequency between 5 and 30 Hz . Following Nikulin et al . ( 2007 ) , to quantify BSI we first band-pass filtered the EEG signal using a 4th-order Butterworth filter centered at each frequency of interest ±1 Hz . Then , we extracted oscillatory power using the Hilbert transform . Additionally , we low-pass filtered the EEG signal using a 4th-order Butterworth filter with a 3 Hz cut-off frequency . The baseline shifts are low-frequency components , because the amplitude modulation of 8–30 Hz frequency oscillation can be detected only at frequencies considerably lower than 8 Hz . Thus , the low-frequency components are extracted by low-pass filtering the artifact-cleaned data at 3 Hz , based on previous studies ( Nikulin et al . , 2007; Nikulin et al . , 2010a ) . We quantified the BSI as the Spearman correlation coefficient ( ρ ) between the low-passed EEG signal and the band-passed power , separately at each frequency and electrode . Accordingly , BSI=0 indicates no relationship between oscillatory power and low-passed signal , as expected for zero-mean oscillations . BSI>0 indicates that strong oscillatory power is correlated with an increase of the low-passed signal , as expected for positive-mean oscillations; instead , BSI<0 indicates that strong oscillatory power is correlated with a decrease of the low-passed signal , as expected for negative-mean oscillations . The amplitude modulation of oscillations with a non-zero mean affects the amplitude of peaks and troughs differently . If the peaks are larger than the troughs relative to the zero line , the ERD will make the electric field go to zero and thus reduce the peaks more strongly than the troughs . It follows that , in this case , any amplitude modulation is expected to produce larger variance for peaks than troughs . The different modulation of peaks and troughs can be captured by AFAI . Following Mazaheri and Jensen ( 2008 ) , to quantify AFAI we first band-pass filtered the EEG signal using a 4th-order Butterworth filter centered at each frequency of interest ±1 Hz , similarly to BSI computation . Then , we identified the time points of peaks and troughs in the band-passed data . These time points were then used to obtain the signal values of peaks and troughs in the non-band-passed ( broadband ) signal . We quantified the AFAI as the normalized difference between the variance of the peaks and troughs of the signal as follows: ( 2 ) AFAI=Var ( Sp ) -Var ( St ) Var ( Sp ) +Var ( St ) where Sp and St refer to the peak and trough values , respectively , estimated in the broadband signal , based on the band-passed signal at a specific frequency . Accordingly , an AFAI=0 indicates that the peaks and troughs are equally modulated ( as for a signal that is symmetric relative to the zero line ) , as expected for zero-mean oscillations . An AFAI≠0 indicates amplitude asymmetry: namely , positive values indicate a stronger modulation of the peaks relative to the troughs ( i . e . , positive amplitude asymmetry or positive mean ) and negative values indicate a stronger modulation of the troughs relative to the peaks ( i . e . , negative amplitude asymmetry or negative mean ) . Note that the sign of the BSI and AFAI predicts the polarity of the late ERP component: specifically , if the sign is negative ( oscillations with a negative mean ) and positive ( oscillations with a positive mean ) , event-related power suppression ( ERD ) will lead to a positive and negative deflection in the ERP , respectively ( Nikulin et al . , 2007; Nikulin et al . , 2010a; Mazaheri and Jensen , 2008 ) . To provide evidence that the baseline-shift mechanism generates the late ERP component , we analyzed the relationship between ERP and ERO ( ERS/ERD ) across trials , as proposed by Mazaheri and Jensen ( 2010 ) . According to the baseline-shift account , states of strong ERD should result in an enhanced late ERP component . To this end , we first identified trials with particularly weak and strong ERO , and then tested how these trials differed in ERP amplitude during the late time window ( >0 . 200 s ) . Specifically , we computed a trial-by-trial estimate of ERO magnitude at each electrode and frequency , averaged across the post-stimulus time window ( 0–0 . 900 s; see Event-related oscillations ) . We also computed a trial-by-trial estimate of the late ERP component at the subject-specific C1-peak electrode ( >0 . 200 s; see Event-related potentials ) . We baseline corrected single-trial ERP estimates by subtracting the prestimulus signal baseline averaged across a 0 . 500 s prestimulus window . Then , for each frequency , and electrode , trials were sorted from weak to strong ERO , divided into five bins ( Linkenkaer-Hansen et al . , 2004; Lange et al . , 2012; Baumgarten et al . , 2016; Iemi et al . , 2017 ) , and the amplitude of the late ERP component was calculated for each bin . The binning was done separately for each trial type ( LVF , UVF , and Fix ) and participant . Furthermore , to enable a comparison of the late ERP component across bins in each participant , the number of trials in each bin was equated by removing the trials recorded at the end of the experiment . To test the hypothesis , we then compared the amplitude of the late ERP component between strongest and weakest ERO bins ( see Statistical Testing for more details ) . We analyzed how prestimulus oscillatory activity influences ERP and ERO across trials . In this analysis , we identified trials with particularly weak and strong prestimulus oscillations , and then tested how these trials differed in the amplitude of the early and late ERP components and in the ERO magnitude . Specifically , we first computed a trial-by-trial estimate of oscillatory power with a Fast Fourier Transform ( FFT ) during a 0 . 500 s prestimulus window for each electrode and frequency . The FFT is advantageous here because , unlike wavelet convolution , the results of an FFT computed over the prestimulus period cannot be influenced by signals occurring in the post-stimulus window . We also computed a trial-by-trial estimate of ERP components during the early time window and ERP components during the late time window at the subject-specific electrode of C1 peak activity ( see Event-related potentials ) . We baseline corrected single-trial ERP estimates by subtracting the ERP averaged across a 0 . 500 s prestimulus window . In addition , we computed a trial-by-trial estimate of the ERO ( ERD/ERS ) at the subject-specific electrode of C1 peak activity and at each frequency and time point in the post-stimulus time window ( 0–0 . 900 s; see Event-related oscillations ) . In this analysis , both ERP and ERO were quantified at the subject-specific electrode of C1 peak activity to enable comparison between the effects of prestimulus power on ERP and ERO . Then , for each frequency , and electrode , trials were sorted from weak to strong prestimulus power and divided into five bins ( Linkenkaer-Hansen et al . , 2004; Lange et al . , 2012; Baumgarten et al . , 2016; Iemi et al . , 2017 ) . For each bin we calculated the ERO magnitude and the amplitude of the early and late ERP components . The binning was done for each trial type and participant . Furthermore , to enable a comparison of ERO and ERP across bins in each participant , the number of trials in each bin was equated by removing the trials recorded at the end of the experiment . We then compared the ERO magnitude and the amplitude of the early and late ERP components between bins of strongest and weakest prestimulus power ( see Statistical Testing for more details ) . Because the estimates of prestimulus power were computed with an FFT , they reflect a mixture of periodic ( i . e . , oscillations ) and aperiodic signals ( i . e , 1/f ‘background’ noise: Podvalny et al . , 2015; Voytek et al . , 2015 ) , referred to as total-band power . Therefore , we set out to determine whether ERP amplitude differed between bins of strongest and weakest periodic signal . To this end , we quantified a single-trial measure of the prestimulus power spectrum for the electrodes of maximal statistical effects separately for each component ( C1 , N150 , and LATE ) , and for each trial type ( LVF/UVF/Fix ) . Next , we parameterized the prestimulus power spectrum ( 1–48 Hz ) into periodic and aperiodic signals ( toolbox fooof: Haller et al . , 2018; Voytek et al . , 2015 ) . First , we fitted the power spectrum with an aperiodic function defined by a slope and an offset . Then , to obtain a measure of the periodic signal , we subtracted this aperiodic function from the original power spectrum , resulting in an aperiodic-adjusted power spectrum . Following the same procedure described above , we classified trials in five bins based on single-trial estimates of aperiodic-adjusted power at the frequencies of maximal statistical effects , and quantified the ERP amplitude ( at the subject-specific channels of maximal C1 response and at the time points of maximal statistical effects ) for each bin . Then , we compared the ERP amplitude between bins of strongest and weakest aperiodic-adjusted power ( using paired-sample t-tests , corrected for multiple comparisons using FDR; Benjamini and Hochberg , 1995 ) . Because ongoing oscillatory activity ( Kaida et al . , 2006; Zhang and Ding , 2010; Mathewson et al . , 2009; Benwell et al . , 2017 ) and ERP amplitude ( Megela and Teyler , 1979; Budd et al . , 1998; Truccolo et al . , 2002; de Munck et al . , 2004 ) may co-vary over the course of an experiment as a function of time-varying variables such as sleepiness , their correlation could be epiphenomenal . To rule this out , we asked participants to report their level of sleepiness at the end of each experimental block using the KSS questionnaire ( Kaida et al . , 2006 ) ; see Stimuli and Experimental Design for more details ) . We then estimated how prestimulus power was related to KSS ratings throughout the stimulation session . Specifically , we computed a trial-by-trial estimate of prestimulus power for each electrode and frequency using an FFT on the 0 . 500 s prestimulus window . We obtained a trial-by-trial estimate of KSS scores by assigning each trial within a block with the KSS score collected at the end of the block . We then used Generalized Linear Modeling ( GLM ) to predict KSS ratings from prestimulus power at the single-trial level . For each participant , electrode and frequency , we fit a regression model of the following form: ( 3 ) KSS=β0+β1P+εwhere KSS is the subjective sleepiness ratings obtained with the KSS questionnaire , P the prestimulus power at each frequency and electrode , β1 the estimated correlation coefficient indicating the contribution of P in explaining variability in KSS , and ε the residual errors . To remove the sleepiness-related time-varying changes in ongoing power , we recomputed a trial-by-trial measure of prestimulus power as follows: ( 4 ) PKSSC=P-β1KSSwhere P is the raw power estimates and β1 the estimated GLM coefficient reflecting the sleepiness-power relationship . We then repeated the binning analysis on the early and late ERP amplitudes described above , with this new trial-by-trial estimate of power where sleepiness-related time-varying changes were ruled out ( PKSSC ) . If the relationship between prestimulus power and ERP is not determined by sleepiness affecting both variables , this new binning analysis would replicate the results of the analysis performed on raw power estimates . In the resting-state session , within each subject , we first computed the AFAI and BSI at each frequency and electrode . For the group-level statistical inference , we then tested whether the AFAI and BSI were significantly different from 0 across the sample of participants . In the stimulation session , within each subject , we first computed: ( 1 ) the difference in the late ERP component between the weakest and strongest ERO bins , ( 2 ) the difference in the ERD magnitude between the weakest and strongest prestimulus power bins , and ( 3 ) the difference in ERP between the weakest and strongest prestimulus power bins ( separately for the early and late time window ) . For the group-level statistical inference , we computed the t-statistics of these differences ( ΔV ) against the null hypothesis that there was no difference between the bins . No significance testing was run for the analysis of the relationship between ERO and prestimulus power due to the circularity of these measures . For all other analyses , a non-parametric cluster based permutation test was used to determine significant effects ( Maris and Oostenveld , 2007 ) . By clustering neighboring samples ( i . e . , based on temporal-spectral-spacial adjacency ) , that show the same effect , this test deals with the multiple comparison problem while taking into account the dependency of the data . We obtained a distribution of the variables of interest ( i . e . , AFAI/BSI for the resting-state session and ΔV for the stimulation session ) under the null hypothesis by randomly permuting their signs 1000 times across participants . On each iteration , we tested the resulting variables with a two-tailed t-test against zero and computed the sum of the t-values within the largest contiguous cluster of significant frequency-electrode ( in the resting-state session ) or time-frequency-electrode points ( in the stimulation session ) that exceeded an a priori threshold ( cluster alpha = 0 . 05 ) , resulting in a distribution of t-sums expected under the null hypothesis . A final p-value was calculated as the proportion of t-sums under the null hypothesis larger than the sum of t-values within clusters in the observed data . Because the cluster permutation test is based on sampling all time points , and the ERP signal comprises early , fast components and late , slow components having different statistical properties , we decided to test for significant effects separately during the early ( <0 . 200 s ) and late ( >0 . 200 s ) time windows . To correct for multiple comparisons due to running the test twice on the same time series , we divided the final permutation alpha by 2 ( final alpha = 0 . 025 , bonferroni corrected ) and considered effects significant only if their p-values were below this threshold . We performed the statistical test separately for positive and negative clusters as recommended by Maris and Oostenveld ( 2007 ) for a two-sided cluster permutation test . We focused the statistical analysis on all electrodes , on frequencies from 5 to 30 Hz and between 0 and 0 . 900 s relative to stimulus onset . | Give a computer the same input and you should get back the same response every time . But give a human brain the same sensory input and you will see a range of different responses . This is because the brain’s response to sensory input depends not only on the properties of the input , but also on its own internal state at the time when the input is processed . Even in the absence of any input , the brain generates complex patterns of spontaneous activity . Fluctuations in this activity affect how the brain responds to the outside world . The electrical activity in the brain – both spontaneous and in response to sensory input – can be measured using electrodes close to the scalp: this measurement is referred to as electroencephalography , or EEG . Spontaneous brain activity takes the form of rhythmic waves , also known as oscillations . In a person who is awake and relaxed , the EEG consists mainly of slow oscillations called alpha and beta waves . Sensory input , such as an image or a sound , triggers changes in brain activity that can be seen in the EEG . This EEG response is called an event-related potential , or ERP , and consists of a characteristic pattern of peaks and troughs in the EEG . To find out how spontaneous brain activity affects ERPs , Iemi et al . presented images of black and white checkerboards to healthy volunteers . The results showed that the ERP looked different if the stimulus occurred during strong alpha and beta waves . The early part of the ERP – which occurs between 80 and 200 milliseconds after the onset of the stimulus – decreased in size , presumably because it was inhibited by strong alpha and beta waves . In contrast , the later part of the ERP – which occurs more than 400 milliseconds after stimulus onset – increased in size . This paradox is accounted for by a newly recognized feature of the oscillations , namely that they fluctuate around a non-zero value of the EEG . Thus , two different mechanisms contributed to these opposite changes . The findings add to our understanding of how spontaneous brain activity influences how we perceive the world around us . Furthermore , spontaneous brain activity differs in a number of disorders , including schizophrenia and autism . Understanding how spontaneous neural oscillations affect how the brain processes information from the senses could provide new insights into these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Multiple mechanisms link prestimulus neural oscillations to sensory responses |
According to the World Health Organization ( WHO ) , in 2018 , an estimated 228 million malaria cases occurred worldwide with most cases occurring in sub-Saharan Africa . Scale-up of vector control tools coupled with increased access to diagnosis and effective treatment has resulted in a large decline in malaria prevalence in some areas , but other areas have seen little change . Although interventional studies demonstrate that preventing malaria during pregnancy can reduce the rate of low birth weight ( i . e . child’s birth weight <2500 g ) , it remains unknown whether natural changes in parasite transmission and malaria burden can improve birth outcomes . We conducted an observational study of the effect of changing malaria burden on low birth weight using data from 18 , 112 births in 19 countries in sub-Saharan African countries during the years 2000–2015 . Specifically , we conducted a difference-in-differences study via a pair-of-pairs matching approach using the fact that some sub-Saharan areas experienced sharp drops in malaria prevalence and some experienced little change . A malaria prevalence decline from a high rate ( Plasmodium falciparum parasite rate in children aged 2-up-to-10 ( i . e . PfPR2-10 ) > 0 . 4 ) to a low rate ( PfPR2-10 < 0 . 2 ) is estimated to reduce the rate of low birth weight by 1 . 48 percentage points ( 95% confidence interval: 3 . 70 percentage points reduction , 0 . 74 percentage points increase ) , which is a 17% reduction in the low birth weight rate compared to the average ( 8 . 6% ) in our study population with observed birth weight records ( 1 . 48/8 . 6 ≈ 17% ) . When focusing on first pregnancies , a decline in malaria prevalence from high to low is estimated to have a greater impact on the low birth weight rate than for all births: 3 . 73 percentage points ( 95% confidence interval: 9 . 11 percentage points reduction , 1 . 64 percentage points increase ) . Although the confidence intervals cannot rule out the possibility of no effect at the 95% confidence level , the concurrence between our primary analysis , secondary analyses , and sensitivity analyses , and the magnitude of the effect size , contribute to the weight of the evidence suggesting that declining malaria burden can potentially substantially reduce the low birth weight rate at the community level in sub-Saharan Africa , particularly among firstborns . The novel statistical methodology developed in this article–a pair-of-pairs approach to a difference-in-differences study–could be useful for many settings in which different units are observed at different times . Ryan A . Simmons is supported by National Center for Advancing Translational Sciences ( UL1TR002553 ) . The funder had no role in study design , data collection and interpretation , or the decision to submit the work for publication .
In 2018 , according to the WHO , 2019 published by the WHO , an estimated 228 million malaria cases occurred worldwide , with an estimated 405 , 000 deaths from malaria globally ( WHO , 2019 ) . Dellicour et al . , 2010 estimated that around 85 million pregnancies occurred in 2007 in areas with stable Plasmodium falciparum ( one of the most prevalent malaria parasites ) transmission and therefore were at risk of malaria . Pregnant women are particularly susceptible to malaria , even if they have developed immunity from childhood infections , in part because parasitized cells in the placenta express unique variant surface antigens ( Rogerson et al . , 2007 ) . Women who are infected during pregnancy may or may not experience symptoms , but the presence of parasites has grave consequences for both mother and unborn baby . Parasites exacerbate maternal anemia and they also sequester in the placenta , leading to intrauterine growth restriction , low birth weight ( i . e . birth weight <2500 g ) , preterm delivery and even stillbirth and neonatal death . Preventing malaria during pregnancy with drugs or insecticide treated nets has a significant impact on pregnancy outcomes ( Eisele et al . , 2012; Kayentao et al . , 2013; Radeva-Petrova et al . , 2014 ) . Observational and interventional studies of malaria in pregnant women are complicated by the difficulty of enrolling women early in their pregnancy . However , in one study , early exposure to Plasmodium falciparum ( before 120 days gestation ) , prior to initiating malaria prevention measures , was associated with a reduction in birth weight of more than 200 g and reduced average gestational age of nearly 1 week ( Schmiegelow et al . , 2017 ) . For other representative studies on the negative influence of malaria infection during early pregnancy on birth outcomes , see Menendez et al . , 2000 , Ross and Smith , 2006 , Huynh et al . , 2011 , Valea et al . , 2012 , Walker et al . , 2014 , and Huynh et al . , 2015 . These results suggest the impact of malaria infection on stillbirths , perinatal , and neonatal mortality may be substantial and needs more careful examination ( Fowkes et al . , 2020; Gething et al . , 2020 ) . In the last few decades , malaria burden has declined in many parts of the world . Although the magnitude of the decline is difficult to estimate precisely , some reports suggest that the global cases of malaria declined by an estimated 41% between 2000 and 2015 ( WHO , 2016 ) and the clinical cases of Plasmodium falciparum malaria declined by 40% in Africa between 2000 and 2015 ( Bhatt et al . , 2015 ) . However , estimates of changing morbidity and mortality do not account for the effects of malaria in pregnancy . In the context of global reductions in malaria transmission , we expect fewer pregnancies are being exposed to infection and/or exposed less frequently . This should result in a significant reduction in preterm delivery , low birth weight and stillbirths . However , declining transmission will also lead to reductions in maternal immunity to malaria . Maternal immunity is important in mitigating the effects of malaria infection during pregnancy as is evidenced by the reduced impact of malaria exposure on the second , third and subsequent pregnancies . Thus , we anticipate a complex relationship between declining exposure and pregnancy outcomes that depends on both current transmission and historical transmission and community-level immunity ( Mayor et al . , 2015 ) . Understanding the potential causal effect of a reduction in malaria burden on the low birth weight rate is crucial as low birth weight is strongly associated with poor cognitive and physical development of children ( McCormick et al . , 1992; Avchen et al . , 2001; Guyatt and Snow , 2004 ) . Although we know from previous interventional studies that preventing malaria in pregnancy is associated with higher birth weight ( Eisele et al . , 2012; Radeva-Petrova et al . , 2014 ) , we do not know whether natural changes in malaria transmission intensity are similarly associated with improved birth outcomes . To address this question , we make use of the fact that while the overall prevalence of malaria has declined in sub-Saharan Africa , the decline has been uneven , with some malaria-endemic areas experiencing sharp drops and others experiencing little change . We use this heterogeneity to assess whether reductions in malaria prevalence reduce the proportion of infants born with low birth weight in sub-Saharan African countries . Our approach conducts a difference-in-differences study ( Card and Krueger , 2000; Angrist and Pischke , 2008; St . Clair and Cook , 2015 ) by leveraging recent developments in matching , a nonparametric statistical analysis approach that can make studies more robust to bias that can arise from statistical model misspecification ( Rubin , 1973; Rubin , 1979; Hansen , 2004; Ho et al . , 2007 ) .
In this analysis , we combine two rich data sources: ( 1 ) rasters of annual malaria prevalence means ( Bhatt et al . , 2015 ) and ( 2 ) the Demographic and Health Surveys ( DHS ) ( ICF , 2019 ) , and we marry two powerful statistical analysis methods of adjusting for covariates – difference-in-differences ( Card and Krueger , 2000; Abadie , 2005; Athey and Imbens , 2006; Angrist and Pischke , 2008; Dimick and Ryan , 2014; St . Clair and Cook , 2015 ) and matching ( Rubin , 1973; Rubin , 2006; Rosenbaum , 2002; Hansen , 2004; Stuart , 2010; Zubizarreta , 2012; Pimentel et al . , 2015 ) . We match geographically proximal DHS clusters that were collected in different time periods ( early vs . late ) and then identify pairs of early/late clusters that have either maintained high malaria transmission intensity or experienced substantial declines in malaria transmission intensity . We then match pairs of clusters that differ in their malaria transmission intensity ( maintained high vs . declined ) but are similar in other key characteristics . Once these quadruples ( pairs of pairs ) have been formed , our analysis moves to the individual births within these clusters . We use multiple imputation to categorize missing children’s birth weight records as either low birth weight or not , relying on the size of the child at birth reported subjectively by the mother and other demographic characteristics of the mother . Finally , we estimate the effect of declined malaria transmission intensity on the low birth weight rate by looking at the coefficient of the malaria prevalence indicator ( low vs . high ) contributing to the low birth weight rate in a mixed-effects linear probability model adjusted for covariates that are potential confounding variables , the group indicator ( individual being within a cluster with declined vs . maintained high malaria transmission intensity ) , and the time indicator ( late vs . early ) . The data we use in this work comes from the following three sources: In this article , we set the study period to be the years 2000–2015 , and correspondingly , all the results and conclusions obtained in this article are limited to the years 2000–2015 . We set the year 2000 as the starting point of the study period for two reasons . First , the year 2000 is the earliest year in which the estimated annual 𝑃𝑓PR2-10 is published by MAP , 2020 . Second , according to Bhatt et al . , 2015 , ‘the year 2000 marked a turning point in multilateral commitment to malaria control in sub-Saharan Africa , catalysed by the Roll Back Malaria initiative and the wider development agenda around the United Nations Millennium Development Goals' . We set the year 2015 as the ending point based on two considerations . First , when we designed our study in the year 2017 , the year 2015 was the latest year in which the estimated annual 𝑃𝑓PR2-10 was available to us . We became aware after starting our outcome analysis that MAP has published some post-2015 estimated annual 𝑃𝑓PR2-10 data since then , but , following Rubin , 2007’s advice to design observational studies before seeing and analyzing the outcome data , we felt it was best to stick with the design of our original study for this report and consider the additional data in a later report . Second , the year 2015 was set as a target year by a series of international goals on malaria control . For example , the United Nations Millennium Development Goals set a goal to ‘halt by 2015 and begin to reverse the incidence of malaria’ and `the more ambitious target defined later by the World Health Organization ( WHO ) of reducing case incidence by 75% relative to 2000 levels . ’ ( WHO , 2008; Bhatt et al . , 2015 ) . It is worth emphasizing that although we set the years 2000–2015 as the study period and did not investigate any post-2015 MAP data because of the above considerations , those published or upcoming post-2015 MAP data should be considered or leveraged for future related research or follow-up studies . After selecting 2000–2015 as our study period , we take the middle point years 2007 and 2008 as the cut-off and define the years 2000–2007 as the ‘early years’ and the years 2008–2015 as the ‘late years . ’ We include all the sub-Saharan countries that satisfy the following two criteria: ( 1 ) The rasters of estimated annual 𝑃𝑓PR2-10 between 2000 and 2015 created by the Malaria Atlas Project include that country . ( 2 ) For that country , IPUMS-DHS contains at least one standard DHS between 2000 and 2007 ( ‘early year’ ) and at least one standard DHS between 2008 and 2015 ( ‘late year’ ) , and both surveys include the cluster GPS coordinates . If there is more than one early ( late ) years for which the above data are all available , we chose the earliest early year ( latest late year ) . This choice was made to maximize the time interval for the decrease of malaria prevalence , if any , to have an effect on the birth weight of infants . For those countries that have at least one standard DHS with available cluster GPS data in the late year ( 2008–2015 ) , but no available standard DHS or GPS data in the early year ( 2000–2007 ) , we still include them if they have a standard DHS along with its GPS data for the year 1999 ( with a possible overlap into 1998 ) . In this case , we assign MAP annual 𝑃𝑓PR2-10 estimates from 2000 to the 1999 DHS data . This allows us to include two more countries , Cote d’Ivoire and Tanzania . The 19 sub-Saharan African countries that meet the above eligibility criteria are listed in Table 1 . From Table 1 , we can see that among the 19 countries , only two countries ( Congo Democratic Republic and Zambia ) happen to take the margin year 2007 as the early year and no countries take the margin year 2008 as the late year . This implies that our study is relatively insensitive to our way of defining the early years ( 2000–2007 ) and the late years ( 2008–2015 ) as most of the selected early years and late years in Table 1 do not fall near the margin years 2007 and 2008 .
We first evaluate the performance of the first-step matching where we focus on the geographical closeness of paired early year and late year clusters from the following three perspectives: ( 1 ) the geographic proximity of the early year and the late year clusters within each pair , which is evaluated through the mean distance of two paired clusters , the within-pair longitude’s correlation and latitude’s correlation between the paired early year and late year clusters , and the mean values of the longitudes and the latitudes of the paired early year and late year clusters; ( 2 ) the closeness of the mean annual malaria prevalence ( 𝑃𝑓PR2-10 ) of the early year and late year clusters at the early year ( i . e . the early malaria prevalence year in Table 1 ) ; ( 3 ) the closeness of the mean annual malaria prevalence of the early year and the late year clusters at the late year ( i . e . the late malaria prevalence year in Table 1 ) . We report the results in Table 4 , which indicate that the first step of our matching produced pairs of clusters which are close geographically and in their malaria prevalence at a given time . Of note , the mean Haversine distance of the early year clusters and late year clusters is 24 . 1 km among the 219 high-low pairs of clusters , and 28 . 7 km among the 219 high-high pairs of clusters . The within-pair longitudes’ and latitudes’ correlations between the paired early year and late year clusters among the high-low and high-high pairs are all nearly one . We then evaluate the performance of the second-step matching , where we focus on the closeness of the sociodemographic status of paired high-low and high-high pairs of clusters , by examining the balance of each covariate among high-low and high-high pairs of early year and late year clusters before and after matching . Recall that for each cluster , we calculate the six cluster-level covariates ( i . e . urban or rural , toilet facility , floor facility , electricity , mother’s education level , contraception indicator ) by averaging over all available individual-level records in that cluster . In each high-low or high-high pair of clusters , there are 12 associated covariates , six for the early year cluster in that pair and six for the late year cluster in that pair . Table 5 reports the mean of each covariate among high-low pairs of clusters and high-high pairs of clusters before and after matching , along with the absolute standardized differences before and after matching . From Table 5 , we can see that before matching , the high-high pairs are quite different from the high-low pairs , all absolute standardized differences are greater than 0 . 2 . The high-low pairs tend to be sociodemographically better off than the high-high pairs ( higher prevalence of improved toilet facilities and floor material facilities , higher prevalence of domestic electricity , higher levels of mother’s education , higher rate of contraceptive use , and more urban households ) . To reduce the bias from these observed covariates , we leverage optimal cardinality matching , as described above , to pair a high-low pair of clusters with a high-high pair and throw away the pairs of clusters for which the associated covariates cannot be balanced well . After matching , we can see that all 12 covariates are balanced well – all absolute standardized differences after matching are less than 0 . 1 . Appendix 1—table 3 summarizes the low malaria prevalence indicators , the time indicators , the group indicators , the covariates , and the birth weights of the 18 , 112 births in the matched clusters . Table 6 reports the estimated causal effect of reduced malaria burden ( low vs . high malaria prevalence ) on the rate of births with low birth weight , which is represented as the coefficient on the malaria prevalence indicator ( diagnostics for the multiple imputation that was used in generating the estimates in Table 6 are shown in Appendix 2—table 1 ) . We estimate that a decline in malaria prevalence from 𝑃𝑓PR2-10 > 0 . 40 to less than 0 . 20 reduces the rate of low birth weight by 1 . 48 percentage points ( 95% confidence interval: 3 . 70 percentage points reduction , 0 . 74 percentage points increase ) . A reduction in the low birth weight rate of 1 . 48 percentage points is substantial; recall that among the study individuals with nonmissing birth weight , the low birth weight rate was 8 . 6% , so a 1 . 48 percentage points reduction corresponds to a 17% reduction in the low birth weight rate . The results in Table 6 also show that there is strong evidence that mother’s age , child’s birth order , mother’s education level and child’s sex are also associated with the low birth weight rate . For example , mothers with higher education level are less likely to deliver a child with low birth weight , and boys are less likely to have low birth weight than girls , which agrees with the previous literature ( e . g . Brooke et al . , 1989; Valero De Bernabé et al . , 2004; Zeka et al . , 2008 ) . Our estimated reduction in the low birth weight rate of 1 . 48 percentage points from reducing malaria prevalence from high to low is similar to that from a naive difference-in-differences estimator that ignores covariates and missingness of birth weight records . The observed low birth weight rates among the records with observed birth weight within the early year clusters in high-low pairs is 9 . 33% , in the late year clusters in high-low pairs is 7 . 52% , in the early year clusters in high-high pairs is 9 . 18% , and in the late year clusters in high-high pairs is 9 . 06% . Therefore , the naive difference-in-differences estimator for the effect of reduced malaria burden without adjusting for covariates and missingness of birth weight records is ( 7 . 52% − 9 . 33% ) − ( 9 . 06% − 9 . 18% ) = − 1 . 69% ( i . e . 1 . 69 percentage points reduction on the low birth weight rate ) . Among all the high-low pairs of clusters in our sample , there has been a decrease in the low birth weight rate from the early years to the late years of 1 . 81 percentage points ( from 9 . 33% to 7 . 52% ) for records with observed birth weight and an estimated decrease of 2 . 04 percentage points ( from 10 . 48% to 8 . 44% ) when multiple imputation is used to impute missing birth weight records . We now explore how much of this decrease can be attributed to reduced malaria burden over time . The estimated effect in Table 6 of the time indicator ( late year vs . early year ) is a 0 . 06 percentage points reduction , which is much less than that of the low malaria prevalence indicator . Moreover , the estimated change in the low birth weight rate over time among high-low pairs that comes from changes in the covariates over time is a 0 . 52 percentage points reduction . This is calculated by looking at the difference between β^T𝐱¯early and β^T𝐱¯late , where β^T is the estimated coefficients of the covariate regressors listed in Table 6 , and 𝐱¯early and 𝐱¯late are the average values in high-low pairs of the covariate regressors of the individuals within the early year clusters and those within the late year clusters respectively . These results suggest that after adjusting for the observed covariates listed in Table 6 and missingness of birth weight records , the observed decrease in the low birth weight rate over time in high-low pairs comes mainly from reduced malaria burden over time instead of changes over time in the low birth weight rate that affect both high-low and high-high pairs of clusters . To illustrate this point and further verify the potentially substantial effect of reduced malaria burden on the low birth weight rate , we also plot the estimated low birth weight rate of each cluster among the high-high pairs and high-low pairs in our study sample in Figure 3 . From Figure 3 , we can see that although in general , for both high-high pairs and high-low pairs , the birth weight rates of the late year clusters are lower than those of the early year clusters , it is clear that the reductions in low birth weight rate from early year to late year among the high-low pairs are considerably greater than those among high-high pairs , suggesting that reducing community-level malaria burden can potentially substantially reduce the low birth weight rate . The results of our secondary analyses support the interpretation of our primary analysis: Recall that in the ‘Sensitivity analyses’ section and Appendix 3 , our sensitivity analyses consider a hypothetical unobserved covariate U that is correlated with both the low malaria prevalence indicator and the low birth weight indicator . For various values of the sensitivity parameters ( p1 , p2 ) , we report the corresponding point estimates and 95% CIs of the estimated treatment effect ( i . e . the coefficient of the low malaria prevalence indicator contributing to the low birth weight rate ) in Appendix 3—table 1 . The results from Appendix 3—table 1 show that the estimated treatment effect ranges from 1 . 13 percentage points reduction to 1 . 83 percentage points reduction ( on the low birth weight rate ) if both p1 and p2 are between −10 and 10 . Recall that p1 ( or p2 ) equals 10 ( or −10 ) means that the probability of the U taking value one increases ( or decreases ) by 10 percentage points if the individual’s low malaria prevalence indicator ( or the low birth weight rate indicator ) equals 1 . That is , allowing both the magnitude of p1 and the magnitude of p2 can be up to 10 means that we allow the existence of a nontrivial magnitude of unmeasured confounding in our sensitivity analyses . Therefore , the estimated treatment effect ranging from 1 . 13 percentage points reduction to 1 . 83 percentage points reduction when both p1 and p2 are between −10 and 10 means that the magnitude of the estimated treatment effect is still evident ( no less than 1 . 13 percentage points ) even if the magnitude of unmeasured confounding is nontrivial ( both |p1| and |p2| can be up to 10 ) . See Appendix 3 for the detailed results and interpretations of the sensitivity analyses . To conclude , although the confidence intervals of the coefficient of the low malaria prevalence indicator on the low birth weight rate presented in the ‘Results’ section cannot exclude a possibility of no effect at level 95% based on our proposed study sample selection procedure and statistical approach , the results and the corresponding interpretations of the primary analysis , the secondary analyses , and the sensitivity analyses have contributed to the weight of the evidence that reduced malaria burden has an important influence on the low birth weight rate in sub-Saharan Africa at the community level .
We have developed a pair-of-pairs matching approach to conduct a difference-in-differences study to examine the causal effect of a reduction in malaria prevalence on the low birth weight rate in sub-Saharan Africa during the years 2000–2015 . Although we cannot rule out no effect at a 95% confidence level , the magnitude of the estimated effect of a reduction from high malaria prevalence to low malaria prevalence on the low birth weight rate ( 1 . 48 percentage points ) is even greater than the estimated effect of a factor thought to be important , antenatal care during pregnancy ( 0 . 96 percentage points ) . In a secondary analysis , we find that reduction in malaria burden from high to low is estimated to be especially crucial for reducing the low birth weight rate of first born children , reducing it by 3 . 73 percentage points ( 95% CI: 9 . 11 percentage points reduction , 1 . 64 percentage points increase ) . This agrees with previous studies which demonstrate that the effects of malaria on birth outcomes are most pronounced in the first pregnancy ( e . g . McGregor et al . , 1983 ) . Previous studies have shown that individual malaria prevention during pregnancy reduces the chances of the woman’s baby having low birth weight ( Kayentao et al . , 2013 ) . In this paper , we examine the community-level effect of reductions in malaria on pregnancy outcomes as opposed to the individual-level effect of malaria prevention interventions during pregnancy . Our results support extrapolation of studies of antenatal malaria interventions on birth weight to populations experiencing declining malaria burden . Furthermore , we conclude that reports of declining malaria mortality underestimate the contribution of reduced malaria exposure during pregnancy on pregnancy outcomes and neonatal survival . Although some studies have documented higher rates of adverse pregnancy outcomes in malaria-infected women with declining antimalarial immunity , such as may be seen in communities with declining malaria exposure ( Mayor et al . , 2015 ) , our study demonstrates that overall reduction in exposure to infection , including during pregnancy , outweighs these individual changes in risk once infected . Strengths of our study include that we use state-of-the-art causal inference methods on a large representative data set . We develop a novel pair-of-pairs matching approach to conduct a difference-in-differences study to estimate the real world effectiveness of public health interventions by combining DHS data with other data sources . There are two major difficulties when using the DHS data to conduct a difference-in-differences study . The first major difficulty is that within each country the DHS samples different locations ( clusters ) over different survey years . Our first-step matching handles this difficulty through using optimal matching to pair the early year DHS clusters and the late year DHS clusters within the same country based on the geographic proximity of their locations . Then each formed pair of clusters can mimic a single cluster measured twice in two different survey years , which serves as the foundation of a difference-in-differences study . The second major difficulty is that although an advantage of the DHS data is that they contain many potentially important cluster-level and individual-level covariates , it may be difficult to come up with a statistical model that is both efficient and robust to adjust for these covariates . A traditional approach to estimating the real world effectiveness of an intervention in such settings is to run a regression of an outcome of interest on a measure of adherence to the treatment ( zero if in the period before the intervention was available and ranging from 0 to 1 after the intervention was available ) , covariates ( individual-level and cluster-level covariates ) and a random effect for the cluster ( Goetgeluk and Vansteelandt , 2008 ) . This regression approach relies heavily on correct specification of the model by which the covariates affect the outcome ( e . g . linear , quadratic , cubic ) , therefore the result can be severely biased by model misspecification ( Rubin , 1973; Hansen , 2004; Ho et al . , 2007 ) . We instead use a second-step matching to first optimally select and match the treated units ( i . e . high-low pairs of clusters ) and control units ( i . e . high-high pairs of clusters ) to ensure that they have balanced distributions of covariates across time and then run the regression with the dummy variables for the matched sets . Such a nonparametric data preprocessing step before running a regression can potentially reduce bias due to model misspecification ( Rubin , 1973; Hansen , 2004; Ho et al . , 2007 ) . Our merged study data set makes use of two aspects of the richness of the relevant data resources . First , from the perspective of sample size and length of time span , the data set includes over 18 , 000 births in 19 countries in sub-Saharan Africa and describes changes in the low birth weight rate over a 15-year period . Some of the studied regions had substantial changes in malaria parasite prevalence during this time period , whereas others did not , which provides us ample heterogeneity necessary for conducting a difference-in-differences study . Second , from the perspective of the comprehensiveness of information , our merged data set includes various types of information: from cluster-level to individual-level records; from geographic to sociodemographic characteristics; from surveyed data to predicted data . Some potential limitations of our study should be considered . First , we discretized the mean malaria prevalence ( i . e . 𝑃𝑓PR2-10 from 0 to 1 ) into high ( 𝑃𝑓PR2-10 > 0 . 4 ) , medium ( 𝑃𝑓PR2-10 lies in [0 . 2 , 0 . 4] ) , and low ( 𝑃𝑓PR2-10 < 0 . 2 ) , which means that the magnitude of the estimated causal effect depends on how we define these cut-offs . Our primary analysis suggests that reducing the malaria burden from high to low may substantially help control the low birth weight rate , and our secondary analyses suggest that a more dramatic reduction in malaria prevalence can lead to a more dramatic drop in the low birth weight rate . More research needs to be done on the minimum magnitude of the reduction in malaria prevalence that is needed to cause a substantial drop in the low birth weight rate . Second , we assigned the malaria prevalence ( i . e . 𝑃𝑓PR2-10 ) data to children’s records based on the DHS survey years which may not be exactly the same years as children’s actual birth years . For example , a child whose age is three years at the corresponding DHS survey year should have been born three years earlier before that DHS survey year , in which case we might have assigned the wrong 𝑃𝑓PR2-10 to that child’s gestational period . We examined this issue via SA1 and the result suggested that this did not induce much bias to the results of our primary analysis . The novel design-based causal inference approach developed in this work , a pair-of-pairs matching approach to conduct a difference-in-differences study ( i . e . the two-step matching procedure to form matched pairs of pairs as a nonparametric data preprocessing step in a difference-in-differences study ) , is potentially useful for researchers who would like to reduce the estimation bias due to potential model misspecification in the traditional difference-in-differences approach . Moreover , the general statistical methodology developed in this work can be applied beyond the malaria settings to handle the heterogeneity of survey time points and locations in data sets such as the Demographic and Health Surveys ( DHS ) . In summary , the contribution of malaria to stillbirth and neonatal mortality , for which low birth weight is a proxy , are currently not accounted for in global estimates of malaria mortality . Using a large representative data set and innovative statistical evidence , we found point estimates that suggested that reductions in malaria burden at the community level substantially reduce the low birth weight rate . To our knowledge , this is the first study of its kind to evaluate the causal effects of malaria control on birth outcomes using a causal inference framework . Although our confidence intervals do include a possibility of no effect , the evidence from our primary analysis and secondary analyses is strong enough to merit further study and motivate further investments in mitigating the intolerable burden of malaria . | Malaria infects around 230 million people each year , mostly in sub-Saharan Africa , and causes more than 400 , 000 deaths . Pregnant women are particularly susceptible to malaria . The parasite that causes malaria can sap the mother’s iron stores and may starve the baby of nutrients . Babies born to infected mothers often have low birth weights , which can have lasting effects on their health and brain development . Previous studies suggest that preventing malaria in pregnant women using insecticide-treated bed nets or medications may improve birth outcomes . Successful efforts to prevent malaria have led to substantially fewer infections in sub-Saharan Africa . But success has been uneven with some communities continuing to have high rates of infection . These differences may allow scientists to better understand the community-level impact of falling rates of malaria on pregnancy outcomes in the real world . Heng et al . estimated that reducing malaria transmission minimises the number of infants born with low birth weights in communities in sub-Saharan Africa . In an observational study , they used data on more than 18 , 000 births in 19 countries in this region between 2000 and 2015 to assess the effects of declining malaria rates on birth weights . They found that a decrease of malaria prevalence is estimated to reduce the rate of low birth weight by 1 . 48% , which is a 17% reduction in the number of observed newborns with low birth weight in the study population . First-born infants appeared to benefit the most . This highlights that malaria interventions are beneficial for pregnant women and their newborns . Most analyses of the impact and cost-benefit of malaria control have ignored the potential advantages of malaria control on birth weight , and may thus undermine the benefits of malaria control . The approach used by Heng et al . may further be useful for other epidemiologists studying global health . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2021 | Relationship between changing malaria burden and low birth weight in sub-Saharan Africa: A difference-in-differences study via a pair-of-pairs approach |
The Escherichia coli SMC complex , MukBEF , acts in chromosome segregation . MukBEF shares the distinctive architecture of other SMC complexes , with one prominent difference; unlike other kleisins , MukF forms dimers through its N-terminal domain . We show that a 4-helix bundle adjacent to the MukF dimerisation domain interacts functionally with the MukB coiled-coiled ‘neck’ adjacent to the ATPase head . We propose that this interaction leads to an asymmetric tripartite complex , as in other SMC complexes . Since MukF dimerisation is preserved during this interaction , MukF directs the formation of dimer of dimer MukBEF complexes , observed previously in vivo . The MukF N- and C-terminal domains stimulate MukB ATPase independently and additively . We demonstrate that impairment of the MukF interaction with MukB in vivo leads to ATP hydrolysis-dependent release of MukBEF complexes from chromosomes .
SMC ( Structural Maintenance of Chromosomes ) complexes have important roles in managing and processing chromosomes in all domains of life ( Gligoris and Löwe , 2016; Nolivos and Sherratt , 2014; Uhlmann , 2016 ) . The distinctive architecture of SMC proteins is conserved with the N-and C-terminal globular domains coming together to form an ATPase head and the intervening polypeptide folding upon itself to form ~50 nm long intramolecular coiled-coil arms , with a dimerisation hinge distal from the head ( Figure 1A ) . Upon ATP binding , the heads of SMC dimers engage to generate two ATPase active sites ( Haering et al . , 2002; Lammens et al . , 2004 ) . In eukaryotes , SMC complexes are exclusively heterodimeric , whilst those in bacteria are homodimers . Nevertheless , the distinctive SMC architecture is conserved , with a kleisin protein linking the two ATPase heads of an SMC dimer , thereby forming a large tripartite proteinaceous ring ( Figure 1 inset ) . Essential accessory ‘kite’ ( kleisin interacting winged-helix tandem elements ) or ‘hawk’ ( HEAT repeat subunits containing proteins associated with kleisins ) bind the kleisin ( Palecek and Gruber , 2015; Wells et al . , 2017 ) . Hawk proteins are present in cohesins and condensins , while kites are present in bacterial SMC complexes , including MukBEF , and eukaryote SMC5/6 complexes . This suggests that the bacterial complexes are more evolutionarily related to the SMC5/6 complexes of eukaryotes than to eukaryote cohesins and condensins . A substantial body of work has led to the hypothesis that DNA segments are topologically entrapped within these tripartite rings . ATP binding and hydrolysis are required for the entrapping of DNA within the rings , loading , and for DNA release , unloading ( Arumugam et al . , 2003; Çamdere et al . , 2015; Cuylen et al . , 2011; Gruber et al . , 2003; Haering et al . , 2002; Haering et al . , 2008; Hu et al . , 2011; Kanno et al . , 2015; ; Nasmyth , 2011; Murayama and Uhlmann , 2014; Uhlmann , 2016; Wilhelm et al . , 2015 ) . E . coli and its closest γ-proteobacterial relatives , encode an apparently distant SMC relative , MukBEF , with little primary sequence homology to other SMCs ( Nolivos and Sherratt , 2014 ) . Organisms encoding MukBEF have co-evolved a number of other distinctive proteins , some of which interact with MukB physically and/or functionally; specifically , topoisomerase IV and MatP both interact with MukB in vitro and in vivo ( BrezellecBrézellec et al . , 2006; Hayama and Marians , 2010; Hayama et al . , 2013; Li et al . , 2010; Nicolas et al . , 2014; Nolivos et al . , 2016; Vos et al . , 2013 ) . MukB forms SMC homodimers , whereas MukF is the kleisin and MukE the kite protein that binds MukF ( Palecek and Gruber , 2015 ) . All three proteins of the MukBEF complex are required for function , and their impairment leads to defects in chromosome segregation , manifested by impairment of segregation of newly replicated origins ( ori ) and mis-orientation of chromosomes with respect to their genetic map within cells ( Danilova et al . , 2007 ) . In rich media , this leads to inviability at higher temperatures and formation of anucleate cells during permissive low-temperature growth ( Niki et al . , 1991; Yamanaka et al . , 1996 ) . Where characterised , most SMC complexes bind their cognate monomeric kleisins asymmetrically , with their N-terminal regions binding the SMC ‘neck’ adjacent to the ATPase head of the molecule distal to the molecule binding the kleisin C-terminus , thereby-forming the tripartite protein ring ( Figure 1 inset ) ( Bürmann et al . , 2013; Gligoris et al . , 2014; Gligoris and Löwe , 2016; Gruber et al . , 2003; Haering et al . , 2004; Huis in’t Veld et al . , 2014 ) . MukF is an atypical kleisin , in that it forms stable dimers through interacting N-terminal winged-helix domains ( WHD ) , while its C-terminal domain interacts with the MukB head at the cap , as is the case for characterised SMC kleisins ( Fennell-Fezzie et al . , 2005; Woo et al . , 2009 ) . Therefore , one MukB dimer is expected to bind a MukF dimer and two MukE dimers in the absence of ATP ( Figure 1A ) ; ATP binding leads to head dimerisation , accompanied by steric expulsion of one of the two MukF C-terminal domains and head engagement ( Woo et al . , 2009 , Figure 1B left panel ) . Here , we reveal that MukF , like other characterised kleisins , interacts functionally with the MukB neck , through a 4-helix bundle in its N-terminal domain , while its C-terminal domain interacts with the MukB head at the cap . We show that this interaction with the MukB neck is required for MukBEF function in vivo , and infer that this interaction is established and broken during cycles of ATP binding and hydrolysis . Impairment of this interaction in vivo leads to ATP hydrolysis-dependent release of MukBEF clusters from chromosomes . Interactions of the MukF N-terminal domain with the MukB neck and MukF C-terminal domain with the MukB head , activate MukB ATPase independently and additively in vitro , with the addition of both fragments restoring wild type MukB ATPase levels . Each of these ATPase activities was inhibited by MukE , with the inhibition being relieved in the presence of DNA . We show that interaction of the MukF N-terminal domain with the MukB neck did not compromise MukF dimerisation . Therefore , MukF dimerisation in heads-engaged MukBEF complexes directs formation of dimers of MukBEF dimers , thereby explaining the stoichiometry observed in vivo ( Badrinarayanan et al . , 2012a , Figure 1B right panel ) .
Because of the intriguing distinction between dimeric MukF and monomeric kleisins ( Figure 1 ) , we set out to test , if the MukF N-terminal domain would interact with the MukB neck , thereby exhibiting the architecture of other SMC dimers and their cognate kleisins . In order to undertake an initial characterisation of MukB-MukF interactions , C- and N-terminal MukF Flag-tagged truncations , immobilised on anti-Flag resin , were analysed for binding to intact MukB and its truncated derivatives containing just the ATPase head ( MukBH ) , or the head plus approximately a third of the adjacent coiled-coil region ( MukBHN ) . The latter variant would be expected to retain any ‘neck’ interaction determinants for MukF , based on similarity with kleisin interacting ‘necks’ , adjacent to the SMC heads of other SMC complexes ( Figure 2A; Bürmann et al . , 2013; Gligoris et al . , 2014 ) . MukF C-terminal derivatives , FC1 and FC2 , interacted with MukB , and all of its derivatives , as expected , because the MukB ATPase head participates in this interaction ( Figure 2B; Figure 1; Woo et al . , 2009 ) . FN2 , containing the N-terminal dimerisation WHD and an adjacent 4-helix bundle , interacted strongly with intact MukB and MukBHN , but not with MukBH , consistent with FN2 interacting with the MukB neck ( Figure 2B ) . Since we reproducibly recovered low levels of MukBH in pulldowns with FN2 , the MukB head might also bind FN2 weakly , although we could not substantiate this by further biochemical analyses ( below ) . We detected no interactions of FN1 , containing just the WHD involved in MukF dimerisation , with the MukB derivatives . In contrast , FN3 , containing the 4-helix bundle ( helices 6–9 ) , and FN4 , carrying only helices 8 and 9 , interacted with MukB and MukBHN , but not MukBH ( Figure 2B ) . Consistent with this , FN6 lacking helices 8 and 9 failed to show an interaction in size exclusion chromatography-multi-angle light scattering ( SEC-MALS ) assays ( below ) and FN7 , lacking helix 9 failed to interact with MukBHN ( Figure 2B bottom right panel ) . We conclude that while helices 8 and 9 of the 4-helix bundle are sufficient for interaction with the MukB neck , helix 9 is essential . To confirm these observations , and to determine the molecular mass of the complexes , we used SEC-MALS ( Figure 3 ) . MukBHN was monomeric in solution , while FN2 was dimeric , as expected from structural analyses ( Fennell-Fezzie et al . , 2005; Woo et al . , 2009 ) . When mixed at a molar ratio of 1 MukBHN monomer:1 . 25 2FN2 ( in the figures , we refer to FN2 dimers as 2FN2 , to reflect their dimeric state ) , two additional peaks of masses 165 kDa and 284 kDa were evident in addition to the MukBHN monomers ( Figure 3A left panel ) . We interpret these as complexes in which either one or two MukBHN molecules bound independently to a single FN2 dimer . Consistent with this interpretation , more of the larger complexes were observed at higher MukBHN to 2FN2 ratios ( 3:1; Figure 3—figure supplement 1 ) . Therefore , the interaction between the MukF N-terminal domain and the MukB neck does not compromise MukF dimerisation . The relatively low proportion of complexes of stoichiometry MukBHN-2FN2-MukBHN as compared to MukBHN-2FN2 in the presence of a large excess of MukBHN , indicates that binding of the second MukBHN to MukBHN-2FN2 complex may be less favourable than binding of the first MukBHN to FN2 . In agreement with the Flag-MukF-MukB interaction assays , FC2 , but not FN2 , formed complexes with MukBH ( Figure 3A middle and right panels ) . FN6 , which lacks the two C-terminal helices , 8 and 9 , of the 4-helix bundle , failed to form complexes with MukBHN ( Figure 3B ) . Addition of ATP did not significantly alter the nature or abundance of complexes containing MukBHN and FN2 or FC2 ( Figure 3—figure supplement 2 ) . This is consistent with MukBHN , which is a monomer in solution , being unable to form stable heads-engaged dimers with either FN2 or FC2 in the presence of ATP . We next tested whether monomers of MukBHN can simultaneously bind both FN2 and FC2 . SEC analysis ( Figure 3C ) showed that mixtures of MukBHN , FN2 and FC2 yielded larger complexes ( olive green trace ) than those formed with MukBHN and FN2 alone ( dark green trace ) , consistent with binding of both FN2 and FC2 to a single monomer of MukBHN . Nevertheless , it was not possible to assign precise masses to these by light scattering , because of the dynamic nature of the complexes and an inability to completely resolve them under a range of SEC conditions . Therefore , a complex containing a MukB dimer with unengaged heads , bound to a MukF dimer may be stabilised by MukF interactions to both the MukB head and neck . An equivalent result was observed with B . subtilis SMC complexes , with both head and neck of a single SMC molecule being bound simultaneously by kleisin N-and C-terminal domains ( Bürmann et al . , 2013 ) . To characterise further the interaction of MukF N- and C-terminal domains to MukB , we determined the binding affinities of fluorescently labelled FN2 , FN10 , FN3 and FC2 using Fluorescence Correlation Spectroscopy ( FCS ) and Fluorescence Polarization Anisotropy ( FPA ) . Both domains bound to MukB with similar affinities , with Kds in the 9–26 nM range , suggesting that interactions of the N-terminal and C-terminal MukF domains with the MukB neck and head , respectively , are similarly strong ( Figure 3—figure supplement 3 ) . FN10 , which in addition to the N-terminal domain also carries the MukF middle region , bound more tightly to MukB than FN2 , consistent with the MukF middle region interacting directly with MukB ( Woo et al . , 2009 ) MukB dimers alone had negligible ATPase activity ( Figure 4 ) , in agreement with previous reports ( Petrushenko et . al . , 2005; Woo et al . , 2009 ) . Addition of MukF kleisin led to robust MukB ATPase . The steady state ATPase rate was ~21 ATP molecules hydrolysed/min/MukB dimer , under conditions of MukF excess ( Figure 4—figure supplement 1A ) . MukF alone did not exhibit ATPase activity . To dissect the MukF requirements for MukB ATPase , we assayed two MukF truncations , containing either the N-terminal domain ( FN2 ) , or the C-terminal domain and the middle region ( FC2 ) ( Figure 2A ) . Both variants stimulated MukB ATPase ( Figure 4 ) . Saturating FC2 , at a 2 . 5-fold molar excess , gave 60% of the maximal ATPase obtained with MukF , while saturating FN2 ( at a 2 . 5-fold molar excess ) gave 33% of maximum ATPase ( Figure 4—figure supplement 1BC ) , while a truncation equivalent to FN2 plus the MukF middle region raised this level to ~50% ( later ) . Addition of FN2 and FC2 together restored ATPase to the level observed with wild type MukF . FN6 , lacking helices 8 and 9 , did not stimulate MukB ATPase , consistent with its failure to interact with the MukB neck ( Figure 4—figure supplement 1D ) . Taken together , the results show that MukB ATPase is activated additively and independently by the N-and C-terminal domains of MukF , with each domain being able to activate ~50% of maximal MukB ATPase . To gain further insight into the interaction of the MukF 4-helix bundle and the MukB neck , variants altered in the MukB neck and MukF helix 9 were analysed for their activity and binding . The mutagenesis strategy was informed by structures of comparable kleisin and SMC neck interactions in yeast cohesin , and B . subtilis SMC complexes ( Gligoris et al . , 2014; Huis in’t Veld et al , 2014 , Bürmann et al . , 2013 ) ; see Materials and methods for the mutagenesis strategy ) . Three variants with triple substitutions in helix 9 of FN2 exhibited an impaired ability to activate MukB ATPase . FN2m2 ( substitutions R279E K283A R286A ) displayed a ~10 fold reduction in the ability to activate MukB ATPase , FN2m3 ( D261K S265K Q268A ) , showed a ~2 fold reduction , while FN2m1 ( D272K I275K R279D ) was reduced by about a third ( Figure 5A ) . Moreover , SEC analysis showed that FN2m1 and FN2m2 failed to interact detectably with MukBHN ( Figure 5—figure supplement 1 ) . Consistent with these results , functional in vivo complementation analysis of the ability of full length MukF variants , containing these sets of mutations , showed that neither MukFm2 nor MukFm3 could complement the temperature-sensitivity of ΔmukF cells , while MukFm1 exhibited partial complementation ( Figure 5—figure supplement 2 ) . In addition , we analysed MukB variants carrying three double amino acid substitutions in the neck , located near MukB C-terminal head domain . They were designed to be at different locations on the putative candidate coiled-coil helix , that protrudes from the C-terminal subdomain of the MukB head , in positions that were predicted to point towards the MukF 4-helix bundle ( Figure 5B ) . MukBm3 ( L1219K L1226K ) had about 35% of wild type ATPase activity when activated by full length MukF; FN2 was unable to activate its ATPase , while FC2 activated it as efficiently as wild type MukF , consistent with its defect in interaction with FN2 ( Figure 5—figure supplement 3; Figure 5—figure supplement 4 ) . MukBm2 ( E1216A E1230A ) , showed no reduction in ATPase , while MukBm1 ( M1215K L1222K ) had <10% of MukF-stimulated MukB ATPase ( Figure 5B ) , indicating that both the MukF N- and C-terminal fragments failed to activate the MukB ATPase of this variant ( Figure 5—figure supplement 3 ) . Although we believe that this protein can fold correctly , the failure to have its ATPase activated by the MukF C-terminal domain is not yet understood; perhaps substitutions at these residues result in an alteration of the coiled-coil structure adjacent to the head , thereby compromising head engagement . MukBm1 and MukBm3 failed to bind FN2 in FPA assays ( Figure 5—figure supplement 4 ) . Consistent with these results , expression in vivo of MukBm1 failed to complement the temperature-sensitive growth defect of ∆mukB cells , while MukBm3 exhibited partial complementation . MukBm2 expression fully complemented the Muk- phenotype ( Figure 5—figure supplement 5 ) . These data suggest that the MukBm3 altered residues ( L1219K and L1226K ) could either be directly involved in the interaction with MukF 4-helix bundle , or that their replacement by charged lysine residues , interferes with a normal interaction interface . Using the E . coli MukEF crystal structure ( pdb , 3EUH; Woo et al . , 2009 ) , along with the structure of the engaged MukB heads , we modelled a FN2 dimer bound by two monomers of MukBHN . This indicated that unless a major conformational change within FN2 dimer takes place upon MukBHN binding , the arrangement of the heads , imposed by interaction of their necks with the 4-helix bundles of the dimer , would be very different from the one revealed by the structure of the engaged MukB heads complex ( Figure 5—figure supplement 6A ) . The motifs that compose the two ATPase active sites in each head monomer would be distant and rotated away from each other . Therefore , if simultaneous binding of the two necks within the intact MukB dimer by the two N-terminal domains of MukF dimer is possible , it would produce a complex whose heads would not be able to engage in ATP binding . Whether such a complex is generated at any stage of the MukBEF activity cycle remains to be determined . In conclusion , the functional interaction between the MukF N-terminal helix 9 and the neck region of MukB coiled-coil revealed and characterised here is equivalent to the similar interaction in other characterised SMC complexes ( Gligoris et al . ( 2014 ) ; Huis in’t Veld et al , 2014 , Bürmann et al . , 2013 ) . MukE inhibited MukF-stimulated MukB ATPase in steady-state assays ( Figure 6A; Bahng et al . , 2016 ) . This inhibition was MukE concentration-dependent ( Figure 6—figure supplement 1 ) . We then tested whether MukE could equally inhibit the ATPase activated by the isolated C- and N-terminal domains of MukF . The incorporation of MukE into a MukBF complex depends on the asymmetric binding of a MukE dimer to the MukF middle region , which also interacts with MukB head in the engaged MukB heads complex ( Shin et al . , 2009; Woo et al . , 2009; Figure 7A and Figure 1B left panel ) . In the absence of MukE , the N- and C-terminal variants of MukF carrying the entire middle region , FN10 and FC2 , respectively , showed 50–60% of wild type MukF activation activity , whereas variants lacking the middle region , FN2 and FC5 , showed 25–33% of activation activity , thereby implicating the middle region , whether it be specified by the C- or N-terminal domains , in stabilising or directing , a conformation that optimises ATP hydrolysis . Both the MukF C-terminal head binding fragment ( Figure 7A , ‘Hb’ ) and the MukE binding segment of the MukF middle region ( Figure 7A , ‘E1a , E1b , E2’ ) contributed to the optimal activation activity ( Figure 7B ) . The molecular basis underlying the role of this middle region segment in maximising steady state MukB ATPase remains unclear; there are no structural data available to inform how the MukF middle region interacts with the MukB head in the absence of MukE . MukE inhibited MukB ATPase activated by the MukF N- and C-terminal domain variants that carried complete MukE dimer binding sites ( FN9 , FN10 and FC2 ) , with a ~4 fold greater inhibition of activation by FN10 , as compared to FC2 ( Figure 7B ) . MukE was unable to inhibit ATPase stimulated by FN2 and FC5 , both of which were lacking MukE binding sites . The effect of MukE on FC4 , lacking the N-terminal part of the MukE1 binding site ( E1A ) was to partially inhibit ATPase activation . These data demonstrate that each of the MukB ATPase activities , stimulated independently by the N- and C-terminal domains of MukF can be inhibited by MukE binding to MukBF . Previous reports have shown no effect of DNA on the ATPase of MukBEF ( Chen et al . , 2008 , Petrushenko et al . , 2006; Woo et al . , 2009 ) , whereas B . subtilis SMC ATPase was reported to be stimulated modestly by DNA ( Hirano and Hirano , 2004 ) . We confirmed that MukB ATPase is independent of the presence of DNA ( Figure 6B ) ; addition of 53 bp ds linear DNA at 20-fold excess ( 10 μM ) over MukB ( 0 . 5 μM ) , did not influence MukBF ATPase activity . MukBF ATPase was not dependent on residual DNA contamination of the proteins as judged by the observation that extensive DNase treatment of MukBF did not influence the ATPase level . DNA alleviated the MukE-mediated inhibition of MukB ATPase . At 5–10-fold excess of DNA over MukB , the ATPase level was restored to ~50% of the level in the absence of MukE ( Figure 6B ) . A similar restoration of activity was observed for most other MukF variants , although FC4 exhibited similar MukE-inhibited ATPase activities in the presence and absence of DNA ( Figure 7B ) . MukF derivatives , FN2 and FC5 , lacking MukE binding sites were not inhibited by MukE and did not respond to DNA . The position of the DNA binding interface on MukB heads , defined by structure-informed mutational analysis ( Figure 7C; Woo et al . , 2009 ) , indicated that DNA binding to this interface could clash with MukE dimer binding to the MukF middle region in a heads-engaged MukBEF complex . Therefore , it seems possible that relief of MukE inhibition by DNA might reflect a competition between MukE and DNA for binding to the MukBF head complex , consistent with the demonstration that MukFE can disrupt MukB-DNA interactions ( Petrushenko et al . , 2006b ) . Since N-terminal and C-terminal domains of MukF could independently and additively bind MukB at the neck and cap , and independently stimulate MukB ATPase activity , we analysed the consequences of disruption of the interactions of endogenous MukF with MukB neck and cap regions in live cells . To this end , we over-expressed either FN2 or FC5 polypeptides from the inducible arabinose promoter on a multicopy plasmid in cells expressing chromosomal MukBmYPetEF . We assessed MukBEF function by analysing the presence and behaviour MukBmYPetEF clusters observed as fluorescent foci associated with the replication origin ( ori ) ( Badrinarayanan et al . , 2012a , 2012b; Nolivos et al . , 2016 ) . Induced over-expression of FN2 led to a rapid loss of MukBEF foci ( half-life of loss ~10 min ) , whereas FC5 over-expression had a lesser effect on focus loss ( half-life of loss ~45 min ) ( Figure 8 , Figure 8—figure supplement 1 ) . In both cases , residual MukBEF clusters remained ori-associated . We then tested if normal cycles of ATP binding and hydrolysis are responsible for the inferred turnover of MukF within functional MukBEF complexes , by testing the effect of fragment production on MukBEQEF complexes that are impaired in ATP hydrolysis and form clusters that turn over very slowly at the replication terminus ( ter ) rather than at ori ( Badrinarayanan et al . , 2012a ) . Over-expression of either FN2 or FC5 had little effect on ter-associated fluorescent MukBEQmYPetEF clusters , consistent with the observation in FRAP experiments that there was little turnover of these complexes , presumably as a consequence of their impaired ATP hydrolysis ( Badrinarayanan et al . , 2012a ) . Nevertheless , we cannot exclude the possibility that the failure to lose MukBEQEF complexes on FN2 or FC5 over-expression is a consequence of the altered cellular localisation of MukBEQEF complexes rather than their impaired ATP hydrolysis . Analysis of the protein composition in MukB+ and MukBEQ extracts , verified comparable high levels of induced expression of FN2 and FC5 in mukB and mukBEQ cells ( >100 fold excess over endogenous MukF for both FN2 and FC5 , when judged by Western blots; Figure 8—figure supplement 2 ) . These observations are consistent with the hypothesis that the MukF interaction with MukB breaks and reforms during cycles of ATP binding and hydrolysis , and that impairment of this interaction leads to loss of functional MukBEF clusters from the chromosome . We have also considered the possibility that the loss of MukBEF clusters from ori after over-expression of FN2 results from the disruption of MukF dimers , rather than an opening of a ring interface . If this were the case , we would have expected the same result in mukBEQ cells . The relatively low turnover of this interaction as compared to the dwell time of MukBEF complexes in vivo ( ~50 s ) and the rates of ATPase measured in vitro could be a consequence of the chelate effect arising from the fact that when the N- or C-terminal domain is released from the MukB neck , or cap , respectively , the reminder of MukF remains associated with MukB through its other interactions , thereby giving a high re-binding rate . FN2 over-expression led to a ~4 fold more efficient displacement of labelled MukBEF complexes from DNA , than over-expression of FC5 . These observations of ATP hydrolysis-dependent release are consistent with the interface between the MukF N-terminal domain and the neck disengaging more frequently than the interface between MukF C-terminal domain and the cap during the activity cycles of MukBEF . Alternatively , disruption of the MukF-neck interaction could lead to more imminent release of DNA from the complex . We note that at least one MukF C-terminal domain-MukB cap interface in a dimeric MukBEF complex has to be broken in each activity cycle to allow the formation of an asymmetric heads-engaged dimeric complex in which only one MukB monomer can bind a MukF C-terminus because of steric occlusion ( Shin et al . , 2009; Woo et al . , 2009 ) . Rebinding of a second MukF C-terminal domain to a MukB cap that becomes available after ATP hydrolysis might be necessary to initiate the next cycle of MukBEF activity . Interfaces between the kleisin N-terminus and the SMC neck in yeast , drosophila and human cohesin complexes have been proposed previously to function as DNA exit gates dependent on SMC ATP hydrolysis ( Beckouët et al . , 2016; Buheitel and Stemmann , 2013; Chan et al . , 2012; Eichinger et al . , 2013; Huis in‘t Veld et al . , 2014 ) .
The work here has revealed two important and related new insights into the action of the E . coli SMC complex , MukBEF . First , the demonstration that the MukF N-terminal domain interacts functionally with the MukB neck , with impairment of this interaction leading to MukBEF complex release from chromosomes and a Muk- phenotype . Second , the observation that the MukF C- and N-terminal domains activate MukB ATPase independently and additively , with each domain contributing ~50% of the maximal activity in steady state assays . We propose that both of these properties relate to the formation of an asymmetric complex between a MukB dimer and MukEF after ATP binding and consequent MukB head engagement . A crystal structure of a complex between a heads-engaged Haemophilus ducreyi MukB dimer bound by MukFE revealed this asymmetry ( Figure 1B; pdb 3EUK; Woo et al . , 2009; Figure 9A ) . In the asymmetric structure , only one MukB head was bound by a MukF C-terminal domain , with the adjacent MukF middle region binding to the second MukB monomer of the MukB dimer , thereby sterically occluding the binding of a second MukF C-terminal domain . This asymmetry induced by MukB head engagement was also observed in solution ( Woo et al . , 2009 ) . Because the MukF N-terminal domain was absent in the variant used , the interaction uncovered here between the MukF N-terminal domain and the MukB neck was not evident . This view of an asymmetric MukBEF complex when heads are engaged is supported by biochemical and in vivo studies ( Shin et al . , 2009; Badrinarayanan et al . , 2012a ) . Furthermore , such an asymmetry directed by interaction of the C- and N-terminal domains of kleisin with the head and neck of SMC dimers appears to be functionally conserved , regardless of whether they form SMC homodimers or heterodimers ( Bürmann et al . , 2013; Gligoris , 2014; Huis in‘t Veld et al . 2014 ) . However , unlike other kleisins , which are apparently monomeric , MukF is a stable dimer and its dimerisation domain is adjacent to the 4-helix bundle , to which helices 8 and 9 belong . Nevertheless , binding of these helices to the MukB neck did not interfere with MukF dimerisation , a result consistent with our previous analysis inferring the existence of MukBEF dimers of dimers in vivo ( Badrinarayanan et al . , 2012a; Figure 9A; top panel; right ) . Note that an asymmetric heads-engaged MukBEF dimer can potentially form a symmetric dimer of dimers as a consequence of MukF dimerisation . We propose that in an engaged-heads dimeric MukBEF complex that the MukF C- and N-terminal domains contained within a single MukF molecule bind separate MukB monomers in the dimer ( Figure 9A; top panel-trans ) , as demonstrated for other SMC complexes . Nevertheless , we cannot exclude the possibility that the asymmetric complex has the MukF C- and N-terminal domains bound to only one of the two MukB molecules , since we have shown that a single monomer of MukBHN can interact with separated MukF N- and C-terminal domains . If this were the case , an additional asymmetric interaction of the MukF middle region with the MukB monomer that is not bound by the MukF C- and N-terminal domains would be present ( Figure 9A; top panel; cis ) . Examination of the heads-engaged asymmetric MukBEF crystal structure , in which the MukF N-terminal is absent , does not allow us to distinguish between these possibilities . Interaction between the MukF N- terminal domain and MukB neck is not only essential for MukBEF function in vivo , but also appears to be broken and reformed during cycles of ATP binding and hydrolysis . Additionally , interaction between the MukF C-terminal domain and the MukB cap may also break and reform during ATP binding and hydrolysis cycles . The observation that impairment of the normal MukF-MukB interactions leads to ATP hydrolysis-dependent loss of MukBEF clusters from chromosomes indicates that by opening of at least one MukB-MukF interface , DNA can be released from the ‘bottom ring chamber’ formed by a kleisin bridging a MukB head and the MukB neck of a partner molecule . This result provides further support for a mechanism in which ATP hydrolysis is required to release MukBEF and other SMC complexes from chromosomes ( Murayama and Uhlmann , 2015; Nolivos et al . , 2016 ) . Equivalent interfaces between the kleisin and SMC3 neck in the yeast , drosophila and human cohesin complexes have also been proposed to act as DNA exit gates and it has been proposed that this interaction , which is not required for loading onto chromosomes , turns-over in response to ATP binding and hydrolysis ( Beckouët et al . , 2016; Buheitel and Stemmann , 2013; Chan et al . , 2012; Eichinger et al . , 2013; Elbatsh et al . , 2017 ) . Although , a DNA exit gate formed by the SMC coiled-coil neck-kleisin interaction appears to be conserved , we think it possible that other interfaces could additionally be used for DNA release . For example , the hinge dimerisation interface , which has been proposed to be a DNA entrance gate ( Buheitel and Stemmann , 2013; Gruber et al . , 2006 ) , might additionally function as an exit gate under some conditions ( Murayama and Uhlmann , 2013; Uhlmann , 2016 ) . Because there are two potential proteinaceous chambers in SMC complexes , the upper one formed by a heads-engaged SMC complex and the lower one by the kleisin bound to the SMC ( Diebold-Durand et al . , 2017; Uhlmann , 2016 ) , each of these chambers could have exit ( and entrance ) gates for DNA segments entrapped within each of them . In MukBEF , interaction of MatP-matS with the MukB hinge has been proposed to promote ATP hydrolysis-dependent release of MukBEF clusters from the ter region of the chromosome , suggestive of release through the dimerisation hinge ( Nolivos et al . , 2016 ) . Similarly , MukB-dependent stimulation of catalysis by TopoIV could arise as a consequence of DNA exiting the MukB hinge and being presented to the TopoIV entrance gate , which is in proximity to the MukB hinge ( Vos et al . , 2013; Zawadzki et al . , 2015 ) . We propose that the observed independent and additive regulation of MukB ATPase by the MukF C- and N-terminal domains , may reflect asymmetry in the two ATPases resulting from asymmetric heads-engaged MukBEF dimer complexes . One of these could be activated by the MukF N-terminal domain and the other by the C-terminal domain . Nevertheless , examination of the heads-engaged MukBEF crystal structure did not reveal any differences in the two ATPase active sites . How the C- and N-terminal MukF domains activate MukB ATPase remain unclear; but their independent and additive action through interaction with different MukB targets , most likely on separate MukB molecules , is in our opinion , most consistent with them activating separate ATPase sites in a MukB dimer . Whether these activations occur at the stage of ATP binding , head engagement , the actual catalytic step , or several of these , remains to be determined . Asymmetric ATPase mechanisms have been demonstrated for ABC transporters , which share the overall organisation of their ATPase heads with SMCs ( ter Beek et al . , 2014; Procko et al . , 2009; Zhou et al . , 2016 ) . Similarly , eukaryotic heterodimeric SMC complexes have active site region differences; for example , comparison of SMC1 sequences with those of SMC3 show protein family-specific differences , with their ATPases being differentially regulated , and in at least some cases , with independent functions ( Beckouët et al . , 2016; Çamdere et al . , 2015; Elbatsh et al . , 2017 , 2016 ) . We have inferred previously from in vivo experiments that ATP hydrolysis by MukBEF is required for both loading and unloading onto DNA ( Nolivos et al . , 2016 ) . Therefore , one plausible explanation of our data is that the MukB neck-kleisin interaction acts in DNA unloading , leaving the MukB head-kleisin interaction to function in loading onto chromosomes . Similarly , a functional asymmetry in yeast cohesion ATPase active sites has been proposed , with the one equivalent to the kleisin-neck interaction uncovered here being required for release from chromosomes , while both ATPases were implicated in loading onto chromosomes ( Çamdere et al . , 2015; Elbatsh et al . , 2017 ) . Our own data do not address whether the interaction with the MukB neck is also required for loading . The significance of how and why MukE inhibits the MukB ATPase activated by either the MukF C-terminal , or the N-terminal domains , remains unclear . It could arise simply from the fact that MukE binding stabilises a particular MukBF conformation , thereby leading to less turnover during the steady state multiple turnover ATPase assays . Alternatively , or additionally , this could reflect MukE playing a regulatory role during transitions between various stages of MukBEF activity cycle . Other in vitro and in vivo studies have postulated a regulatory role of MukE ( Gloyd et al . , 2011; She et al . , 2013 ) , although details of how this regulation is mediated have been unclear . Nevertheless , depletion of MukE in vivo mimics the ATP hydrolysis-impaired phenotype of a MukBEQ mutant , which loads slowly onto DNA in the ter region , but is unable to undergo the multiple cycles of ATP binding and hydrolysis required to target to ori ( Badrinarayanan et al . , 2012b; Nolivos et al . , 2016 ) . The ability of DNA to relieve MukE inhibition of MukB ATPase could result from MukE and DNA competing for binding to the MukB head . This is consistent with in vitro studies , which showed competition between MukEF and DNA for MukB binding and that MukEF inhibited MukB-mediated DNA condensation ( Cui et al . , 2008; Petrushenko et al . , 2006b ) . Furthermore , a patch of positively charged amino acid residues on the surface of MukB head , close to the base of the neck , was shown to be important for interaction with DNA ( Figure 7C; Woo et al . , 2009 ) . Projection of B-form DNA onto this patch highlights the potential competition of DNA- and MukE-binding to a MukBF complex , which may reflect alternative states during the MukBEF-DNA activity cycle . A range of structures , alongside extensive biochemical and functional analyses , leads to the conclusion that all SMC complexes , including MukBEF , share distinctive architectures and similarities in their likely molecular basic mechanisms of action on chromosomes . Central to the SMC complex mechanism is the ability to bind and hydrolyse ATP in a modulated fashion , which directs stable loading onto chromosomes , regulated release from chromosomes and autonomous rapid transport with respect to DNA , which is likely to depend on such loading and release ( Diebold-Durand et al . , 2017; Terkawa et al . , 2017; Wang et al . , 2017 ) . Any such transport must require at least two specific DNA-SMC complex attachment points on different conformational states of the complex , with coordinated transitions as transport proceeds . Our finding that MukF dimerisation is maintained during its interaction with the MukB neck , not only validates our demonstration of dimers of MukBEF dimers in active MukBEF clusters in vivo , but provides support for our previously proposed ‘rock ( or rope ) climber’ model for the transport of MukBEF dimer of dimers with respect to DNA ( Figure 9B; Badrinarayanan et al . , 2012a ) . This model assumes that dimers of MukBEF dimers are a minimal functional unit , in which coordinated capture and processing of DNA segments by each MukBEF dimer , similar to the action of a climber reaching out to ‘grab’ a rock/rope alternatively with each arm . The staggered cycles of ATP binding and hydrolysis , DNA trapping and release and associated conformational changes could effectively coordinate the activity within the partner dimers within MukBEF dimers . For SMC complexes that do not obviously form dimers of dimers , the type of loop-capture and fusion model proposed by Diebold-Durand et al . , 2017 has merits . In this , DNA loops captured in the upper SMC chamber are transferred to the lower chamber , where they fuse with a pre-existing loop , thereby meeting the basic requirements for ATP hydrolysis-driven transport .
MukB , MukBH , MukBHN , MukE , were 6xHis-tagged at the C-terminus ( pET21 ) , while MukF and its C- and N-terminal truncations were 6xHis-tagged at the N-terminus ( pET28 ) . MukB variants were expressed in strain FL01 , which is mukB 3xFLAG C3013I ( NEB ) . MukF variants and MukE 6xHis-tagged at the C-terminus were expressed from pET21 in C3013I cells ( NEB ) . 2L cultures were grown in LB with appropriate antibiotics at 37°C to A600 ~0 . 6 and induced by adding IPTG at final concentration of 0 . 4 mM . After 2 hr at 30°C , cells were harvested by centrifugation , re-suspended in 30 ml lysis buffer ( 50 mM HEPES pH 7 . 5 , 300 mM NaCl , 5%glycerol , 10 mM imidazole ) supplemented with 1 tablet of protease inhibitor ( PI ) , and homogenised . Cell debris was removed by centrifugation and clear cell lysates were mixed with 5 ml equilibrated TALON Superflow resin , poured into a column , then washed with 10 X volume of washing buffer ( 50 mM HEPES pH 7 . 5 , 300 mM NaCl , 5% glycerol , 25 mM imidazole , PI ) . Bound proteins were eluted in elution buffer ( 50 mM HEPES pH 7 . 5 , 300 mM NaCl , 5% glycerol , 250 mM imidazole ) . The fractions from TALON were diluted to 100 mM NaCl buffer and injected to HiTrapTM Heparin HP column ( GE Healthcare ) pre-equilibrated with Buffer A ( 50 mM HEPES pH 7 . 5 , 100 mM NaCl , 10% glycerol , 1 mM EDTA , 1 mM DTT ) , then the column was washed at 1 ml/min flow rate until constant A280 . Purified fractions were eluted with a gradient 100–1000 mM NaCl . For MukE and MukF purifications , fractions from Talon were diluted and injected onto a HiTrap DEAE FF column ( GE healthcare ) pre-equilibrated in Buffer A . Purified fractions were eluted with a gradient 100–1000 mM NaCl . Protein concentration was estimated by UV absorption at 280 nm on Nanodrop spectrophotometer , and protein purity and identity confirmed by electrospray ionisation mass-spectrometry and SDS PAGE . Proteins were aliquoted and stored at −20°C in a buffer containing 10% glycerol . ATP hydrolysis was analysed in steady state reactions using an ENZCheck Phosphate Assay Kit ( Life Technologies ) . 150 µL samples containing standard reaction buffer supplemented with 2 mM of ATP were assayed in a BMG Labtech PherAstar FS plate reader at 25°C . The data were analysed using MARS data analysis software . Quantitation of phosphate release was determined using the extinction coefficient of 11 , 200 M−1cm−1 for the phosphate-dependent reaction at A360 nm at pH 7 . 0 . Purified proteins were fractionated on a Superose 6 10/300 GL or a Superose 12 10/300 column equilibrated with 50 mM HEPES , pH 7 . 5 buffer containing 100 mM NaCl , 1 mMDTT , 1 mM EDTA , at flow rate of 0 . 5 ml/min . 500 µl samples containing analysed proteins were injected on the column and run at a flow rate of 0 . 5 ml/min . SEC-MALS analysis was performed at 20°C using a Shimadzu ( Kyoto , Japan ) chromatography system , connected in-line to a Heleos8+ multi angle light scattering detector and an Optilab T-rEX refractive index ( RI ) detector ( Wyatt Technologies , Goleta , CA ) . Protein samples in 50 mM HEPES pH 7 . 5 , 100 mM NaCl , 1 mM DTT , 1 mM EDTA , 10% glycerol , were injected in this system , and the resulting MALS , RI and UV traces processed in ASTRA 6 ( Wyatt Technologies ) . MukF FLAG-tagged fragments were expressed from pET DUET plasmids in C3013I cells ( NEB ) . 1L cultures were grown in LB with carbenicilin ( 100 µg/ml ) at 37°C to A600 ~0 . 6 and induced by adding IPTG to a final concentration 0 . 4 mM . After 2 hr at 30°C , cells were harvested by centrifugation , re-suspended in 30 ml lysis buffer ( 50 mM HEPES pH 7 . 5 , 300 mM NaCl , 5%glycerol , 10 mM imidazole ) supplemented with 1 tablet of protease inhibitor ( PI ) , and homogenised . Cell debris was removed by centrifugation and clear cell lysates were mixed with 150 µl Anti-FLAG M2 Affinity gel ( Sigma Aldrich ) , incubated for 1 hr at 4°C . The resin was then washed three times with the same buffer containing 250 mM NaCl , resuspended in 1 ml of buffer I ( 50 mM HEPES pH 7 . 5 , 100 mM NaCl ) , and purified MukB , MukBH or MukBHN were added . After 45 min incubation ( 4°C ) the resin was washed three times , re-suspended in 200 µl of protein loading buffer ( NEB ) and analysed on 4–20% gradient SDS PAGE . Our design of amino acid substitutions in the coiled-coil of MukB neck and MukF helix 9 was informed by the arrangement and interactions at the interface between Scc1 kleisin helices that interact with the coiled-coil of cohesin ( Gligoris et al . , 2014 ) . Both SMC coiled-coil helices , one protruding from N-terminal and the other from the C-terminal subdomain interact with Scc1 . We targeted the MukB C-terminal helix for mutagenesis because we could make better predictions for the orientation of this helix in the MukB neck . Three sets of double mutations were constructed; they mapped to the same side of the helix , but with each set on a slightly different face . The rationale for mutagenesis in MukF helix 9 was to mutate solvent exposed residues that were not predicted to interact with other MukF helices , or be obstructed by MukE binding . Three sets of triple mutations along the helix were constructed . All amino acid residues chosen for mutagenesis were either invariant or very highly conserved among bacterial species . Point mutations in plasmid-encoded genes were made using Q5 site-directed mutagenesis Kit ( NEB ) . Primers were designed with NEBase Changer . 10 ng of the template was taken to the reaction . Plasmids were isolated and mutations confirmed by sequencing . The ability of leaky plasmid-encoded MukF or MukB expression from pET21 , in the absence of IPTG , to complement the temperature-sensitive growth defect of ∆mukF ( AB 233 ) or ∆mukB ( RRL149 ) cells , respectively , at 37°C in LB was assayed . Cells were transformed with pET21 carrying MukF or MukB , or their variants , and allowed to recover for 8 hr post transformation at permissive temperature then plated in duplicates on LB plates containing carbenicillin ( 100 µg/ml ) . One plate was incubated at non-permissive ( 37°C ) and the other one at permissive ( 20°C ) temperature . Colonies from plates incubated at permissive temperature were streaked in duplicate and grown at permissive and non-permissive temperature along with positive and negative controls . Strains were streaked onto LB plates with appropriate antibiotics . Single colonies were inoculated into M9 glycerol ( 0 . 2% ) and grown overnight at 37°C to A6000 . 4–0 . 6 , then diluted into fresh M9 and grown to A600 0 . 1 . Cells were spun and immobilised on agarose pads between two glass coverslips ( 1 . 5 thickness ) . 1% agarose pads were prepared by mixing low-fluorescence 2% agarose ( Bio-Rad ) in dH2O 1:1 with 2x growth medium . For analysis of MukBEF fluorescent clusters ( foci ) , strains carrying either functional MukBmYPet ( SN182 ) , or the ATP hydrolysis-impaired mutant MukBEQmYPet ( SN311 , Nolivos et al . , 2016 ) were used . Wide-field fluorescence microscopy used an Eclipse TE2000-U microscope ( Nikon ) , equipped with an 100x/NA1 . 4 oil PlanApo objective and a Cool-Snap HQ2 CCD , and using Metamorph software for image acquisition . Over-expression of FN2 and FC2 was from pBAD24 plasmids containing the appropriate arabinose-inducible MukF derivative . Strains were transformed with given plasmid and grown in M9 glycerol medium supplemented with 0 . 2% glucose to limit leaky expression from the arabinose promoter . Once cultures reached A600 ~0 . 1 , cells were centrifuged and re-suspended in M9 glycerol medium supplemented with 0 . 2% L-Arabinose and grown at 37°C . Every 20 min , cells from1 ml of culture were taken , centrifuged , placed on agarose pad and imaged . As a control , strain carrying empty pBAD24 vector was analysed . Cells were segmented from brightfield images using MicrobeTracker ( Sliusarenko et al . , 2011 ) . MukB foci were detected using ‘spotfinderM’ , available as part of the MicrobeTracker Suite . FCS was carried out on a ConfoCor 2 system ( Carl Zeiss ) . The 633 nm line of a HeNe laser was directed via a 488/561/633 dichroic mirror and focused with a Zeiss C-Apochromat 40 Å~NA 1 . 2 water immersion objective to excite experimental samples containing Cy5 . Fluorescence emission was collected using a 655 nm long pass filter and recorded by an avalanche photodiode . The pinhole diameter was adjusted to 83 μm ( one Airy unit ) , and the pinhole position was optimised with use of the automatic pinhole adjustment for Cy5 . All FCS experiments were carried out in Lab-Tek ( Nagle Nunc International ) eight-well chambered borosilicate glass plates at 22 ± 1°C . In the assay , diffusion of Cy5-labelled FN2 and FN10 fragments at fixed concentrations ( ~10 nM ) was measured in samples carrying MukB at a range of concentrations up to 160 μM . Since MukB is much larger than any of the fragments used , up to a 3-fold increase in diffusion time was observed . The intensity of fluorescence signal was measured and the autocorrelation function G ( t ) was determined for diffusing fluorescently labelled species present in the sample . If two species with different diffusional properties are present , the autocorrelation function G ( t ) can be described as a two-component model that allows analysis of the abundance of each species:G ( τ ) =[1−T+Texp ( −ττT ) ]N−1=[1−Y ( 1+ττsubstrate ) 1+r02z02ττsubstrate+Y ( 1+ττproduct ) 1+r02z02ττproduct]where T is the average fraction of dye molecules in the triplet state with the relaxation time τT , N is the average number of fluorescent molecules in the volume observed , Y is the relative fraction of fragment bound to MukB , τ substrate and τ product are the diffusion time constants of free protein ( labelled fragment as indicated for individual experiment and fragment bound to MukB ) , respectively , and r0 and z0 are the lateral and axial dimensions , respectively , of the observation volume . All calculations , including the evaluation of the autocorrelation curves , which was carried out with a Marquardt nonlinear least-square fitting procedure , were performed using the ConfoCor 2 instrument software . To obtain the % of bound and unbound fragments , the diffusion times for fluorescently labelled fragment were measured and fixed during data analysis . The diffusion time for the complex of a given fragment and MukB was estimated based on measured diffusion time for labelled MukB . No change in diffusion time for labelled MukB was observed when unlabelled fragment was added; therefore , the measured diffusion time for MukB was used as a fixed value during data analysis . Experiments were done on a BMG LABTECH PHERAstar FS next-generation microplate reader with an FP 590–50 675–50 optic module . Samples were measured in Corning black 96 well flat bottom half volume plates at 25°C . All sample volumes were 100 μL . Cy5 labelled FN3 and FC2 were used at 5 nM and 9 nM respectively . The concentration of MukB was varied from 0 . 1 nM to 1 µM . Samples were equilibrated for 40 min before measurement . Experiments were repeated thrice and standard deviations are reported . Data were plotted and analysed using Sigmaplot , where Kd and total receptor concentration were solved simultaneously . Binding reached saturation above 160 nM MukB . Binding of FN3 or FC2 with 1 µM MukB was used as a 100% bound reading . The fraction of FN3 or FC2 bound was determined using the equation:[1− ( Max value−Current valueMax value−Min value ) ]∗100% Data were plotted and the values of Kd and ‘total receptor’ concentration ( RT ) were simultaneously determined using Sigmaplot by solving the quadratic for fraction bound ( B ) below , B= ( MukBT + Kd + RT ) − ( ( −MukBT − Kd − RT ) 2 − 4MukBTRT ) 2 MukB+ ( SN182 ) and MukBEQ mutant ( SN311 ) cells were transformed with pBAD , pBAD-FN2 ( pKZ111 ) and pBAD-FC5 ( pZ103 ) plasmids . Cells were grown at 22°C , to A6000 . 4–0 . 6 , induced with 0 . 2% L-Arabinose for 3 hr . Cultures were spun down and cell pellets were resuspended in gel loading buffer and proteins were separated by a 4–20% gradient SDS PAGE followed by Western blots with mouse anti-MukF antibody as primary and goat anti-mouse as secondary antibody . In-solution trypsin digestion . Bacterial lysates were prepared from 50 ml cultures grown in M9 minimum media to A600 ~0 . 2 , with expression from pBAD- induced with 0 . 2% arabinose for 1 hr . Cells were centrifuged and the pellet was resuspended in 200 μL of 0 . 1% SDS in PBS , sonicated and incubated for 5 min at 100°C . After centrifugation , supernatant was collected and protein concentration was as assessed using the BCA ( Thermo Fisher Scientific , USA ) method . Then , 10 μg of protein extract was digested with 0 . 2 μg of sequencing-grade trypsin ( Promega , Mannheim , Germany ) overnight at 37°C Proteins were reduced with DTT and alkylated using iodoacetamide . Each sample was prepared for digestion in duplicate . NanoLC-MS/MS Analysis . For each run , 1 . 5 μg of the digested protein samples was injected onto an RP C18 precolumn ( Thermo Fisher Scientific , Waltham , MA , USA ) connected to a 75 μm i . d . ×25 cm RP C18 Acclaim PepMap column with a particle size of 2 μm and a pore size of 100 Å , using a Dionex UltiMate 3000 RSLCnano System ( Thermo Fisher Scientific ) . Every sample was injected in duplicate at random . Before analysis , the system was calibrated using Pierce LTQ ESI Positive Ion Calibration Solution ( Thermo Fisher Scientific ) . The following LC buffers were used: buffer A ( 0 . 1% ( v/v ) formic acid in Milli-Q water ) and buffer B ( 0 . 1% formic acid in 90% acetonitrile ) . The peptides were eluted from the column with a constant flow rate of 300 nL . min−1 with a linear gradient of buffer B from 5% to 65% for 120 min . At 100 min , the gradient was increased to 90% B and was held there for 10 min . Between 110 and 120 min , the gradient returned to 5% to re-equilibrate the column for the next injection . The peptides eluted from the column were analysed in the data-dependent MS/MS mode on a Q-Exactive Orbitrap mass spectrometer ( Thermo Fisher Scientific ) . The instrument settings were as follows: the resolution was set to 70 , 000 for MS scans , and 17 , 500 for the MS/MS scans to increase the acquisition rate . The MS scan range was from 300 to 2000 m/z . The MS AGC target was set to 1 × 106 counts , whereas the MS/MS AGC target was set to 5 × 104 . Dynamic exclusion was set with a duration of 20 s . The isolation window was set to 2 m/z . Analysis of proteomic data . After each LC-MS/MS run , the raw files were analysed by Proteome Discoverer , version 1 . 4 . 14 ( Thermo Fisher Scientific ) . The identification of proteins was performed using the MASCOT engine against the UniProt after adding to database sequences of recombinant proteins FN2 and FC5 . Analyses were completed using the following parameters: a tolerance level of 10 ppm for MS and 0 . 05 Da for MS/MS and with 1% FDR . Trypsin was used as the digesting enzyme , and one missed cleavage was allowed . Estimation of protein abundance was based on the emPAI parameter ( Ishihama et al . , 2005 ) . | Most DNA in a cell is arranged in structures called chromosomes . From bacteria to humans , chromosomes have to be compacted and highly organized to allow the cells to maintain and use their genetic information . In all organisms , large ring-shaped protein complexes play a crucial role in managing chromosomes . They transport and organize DNA thanks to reactions whose precise mechanism remains unknown . In bacteria , MukB and a type of kleisin called MukF are two examples of molecules involved in chromosome management . Two MukBs join at one end to form a hinge; at the other end , each MukB protein has a neck and a head . The two heads are linked by the kleisin to form a large protein ring , which can open to capture DNA . The MukB heads can trigger a biochemical reaction that creates the energy essential to trap and release DNA during DNA transport . Here , Zawadzka et al . study how the different components of the MukB-kleisin complex interact with each other to undergo the biochemical reactions that lead to DNA transport . The experiments show that the kleisin joins two MukB heads by attaching the base of one to the neck of the other , asymmetrically closing the ring . The separate interactions of different regions of the kleisin to the head and neck of MukB independently activate the two MukB heads , thereby controlling essential steps in the reactions with DNA . Two MukB-kleisin ring complexes are joined to each other because of a tight interaction between the two kleisin molecules . This leads Zawadzka et al . to suggest that DNA is sequentially grabbed and released from these two rings during DNA transport , similar to how a climbing rope is attached and released through carabiners . Cells cannot survive or be healthy without their chromosomes being accurately managed . It is still unclear how molecules such as MukBs and kleinsins drive this process . A better picture of their structure and interactions is an essential first step to understand these mechanisms . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2018 | MukB ATPases are regulated independently by the N- and C-terminal domains of MukF kleisin |
Planar cell polarity ( PCP ) signaling controls the polarization of cells within the plane of an epithelium . Two molecular modules composed of Fat ( Ft ) /Dachsous ( Ds ) /Four-jointed ( Fj ) and a ‘PCP-core’ including Frizzled ( Fz ) and Dishevelled ( Dsh ) contribute to polarization of individual cells . How polarity is globally coordinated with tissue axes is unresolved . Consistent with previous results , we find that the Ft/Ds/Fj-module has an effect on a MT-cytoskeleton . Here , we provide evidence for the model that the Ft/Ds/Fj-module provides directional information to the core-module through this MT organizing function . We show Ft/Ds/Fj-dependent initial polarization of the apical MT-cytoskeleton prior to global alignment of the core-module , reveal that the anchoring of apical non-centrosomal MTs at apical junctions is polarized , observe that directional trafficking of vesicles containing Dsh depends on Ft , and demonstrate the feasibility of this model by mathematical simulation . Together , these results support the hypothesis that Ft/Ds/Fj provides a signal to orient core PCP function via MT polarization .
In Drosophila and in vertebrates , six proteins constituting a ‘core’ PCP module acquire asymmetric distributions to polarize epithelial cells along a planar axis ( Goodrich and Strutt , 2011 ) . In the fly wing epithelium , three of the six proteins , Frizzled , Dishevelled and Diego ( Dgo ) , become enriched at the distal adherens junctions ( AJ ) , two , Van Gogh ( Vang ) and Prickle ( Pk ) localize to the proximal side , while Starry night/Flamingo ( Fmi ) localizes to both proximal and distal sides ( Axelrod , 2009 ) . Preferential interactions between Fmi/Fz and Fmi/Vang complexes across cell boundaries ( Lawrence et al . , 2004; Chen et al . , 2008; Strutt and Strutt , 2008 ) and intercellular feedback loops ( Tree et al . , 2002; Amonlirdviman et al . , 2005 ) can account for intracellular segregation of these complexes and coordinated alignment among neighboring cells . However , it remains unclear how this local polarity is globally oriented with respect to the tissue axes . It is proposed that the Ft/Ds/Fj system , comprising the atypical cadherins Ft ( Yang et al . , 2002 ) , Ds ( Adler et al . , 1998 ) and the Golgi-resident protein Fj ( Zeidler et al . , 1999 ) , acts as a ‘global’ PCP module , transducing tissue level directional cues encoded by opposing Ds and Fj expression gradients , to orient the core PCP module ( Yang et al . , 2002; Ma et al . , 2003 ) . Though a mechanism that might transmit a directional signal from the Ft/Ds/Fj module to the core module is suggested by existing observations , important additional data are needed to support the model . In the Drosophila pupal wing , apical non-centrosomal MTs are aligned along the proximal distal axis prior to the onset of hair growth ( Fristrom and Fristrom , 1975; Eaton et al . , 1996; Turner and Adler , 1998; Shimada et al . , 2006 ) . The Ft/Ds/Fj module plays an incompletely defined role in organization of these MTs ( Harumoto et al . , 2010 ) , and MT-associated vesicles containing Fz are observed to preferentially move in a plus-end directed fashion toward the distal cell cortex ( Shimada et al . , 2006 ) , leading to the hypothesis that Ft/Ds/Fj signals via these MTs to orient core PCP function . However , a comprehensive spatiotemporal correlation between Ds and Fj gradients , MT orientation and direction of core protein polarization has not been examined , nor have corresponding effects of global Ft/Ds/Fj perturbations on MTs and directional vesicle trafficking been examined . In this study , we provide additional evidence for this model in the Drosophila wing . We find that the apical microtubule ( MT ) cytoskeleton ( Eaton et al . , 1996; Shimada et al . , 2006; Harumoto et al . , 2010 ) shows strong spatial and temporal correlation with core protein asymmetry throughout wing development . We show that , in the developing wing , Ds and Fj signal through a PCP-specific domain of Ft , together with one or more partially redundant , additional signal ( s ) , to polarize these apical MTs . Ft coordinates association of MTs with apical intercellular junctions , suggesting that Ft and Ds spatially regulate capture and organization of the apical MT cytoskeleton . We show that , in addition to Fz , vesicles containing Dsh are transcytosed on these MTs , and that transcytosis is disrupted in ft or ds mutant tissue , suggesting that this trafficking provides directional bias for core protein localization . Together , our results support the hypothesis that global polarity information is provided by the Ft/Ds/Fj module and other signals to orient the apical MT network , which in turn orients polarization of the core PCP module .
Apical MT alignment and orientation of core PCP protein domains have been shown to correlate , but have only been examined in several small domains during late pupal wing development ( Shimada et al . , 2006; Harumoto et al . , 2010 ) ( between 14 and 30 hr after puparium formation [APF] ) ( See also Figure 1—figure supplement 1B–D ) . If MT alignment provides directional bias for core protein polarization , one should observe a spatiotemporal correlation across the entire wing throughout the time core PCP proteins are polarized . Core PCP protein polarization with respect to the tissue axes is first observed during larval wing development ( Classen et al . , 2005 ) . We therefore surveyed apical MT structure beginning in third instar . To facilitate this broad analysis , we used tubulin staining . While foregoing the ability to determine plus-end orientation as provided by analysis of EB1 comets , this approach enables analysis of vastly greater numbers of MTs than does the EB1 assay , and also provides the potential to distinguish a more stable , anchored population , though in fact we see a strong correlation between results from both methods ( Figure 1—figure supplement 2A–B′; Video 1 ) . Similarly , both anti-tubulin and anti-tyrosinated tubulin antibodies produce indistinguishable results ( Figure 1—figure supplement 2C ) . 10 . 7554/eLife . 02893 . 003Video 1 . Live in vivo imaging of EB1::GFP and Cherry::Jupiter . Live in vivo imaging of EB1::GFP comets and Jupiter::Cherry labeled MTs in 24 hAPF pupal wing . Refers to Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 003 The earliest evident apical tubulin staining was seen in early to mid third-instar , and revealed an asymmetrical accumulation of tubulin in mostly single ‘dots’ within each cell ( Figure 1A ) . Image analysis of the tubulin dots revealed a significant bias of dots localizing on the proximal side of the cell ( side closest to the hinge fold; Figure 1—figure supplement 1A–A″ ) . Shortly thereafter , MTs appear to fan out from these sites toward the center of the wing pounch ( the future distal side ) ( Figure 1B ) . In EM images , these early MTs appear as dense bundles anchored to proximal junctions ( Figure 1E ) . 10 . 7554/eLife . 02893 . 004Figure 1 . Apical junction-anchored MTs appear in early third-instar wing discs . ( A–C ) Tubulin staining in progressively older third-instar wing pouch . Note the asymmetric accumulation of tubulin on the proximal side of the cell in early third-instar wing pouch and growing MTs observed in mid-late third-instar wing pouch . White arrows in the first wing disc cartoon show the proximal distal axis , with distal in the center of the wing pouch , and proximal tissue forming a ring around the future wing blade . ( D ) MTs in 7 hAPF pupal wing . Scale bar: 5 μm . ( A′ ) A vertical cross-section of the early third-instar wing disc . Apical is at the top . ( C′ and D′ ) Orientations of MTs , in the regions shown , calculated using OrientationJ , aligned with the P-D axis at each stage and location ( arrows ) . ( E ) TEM micrograph of MTs in early third-instar wing pouch . Note anchored MTs at intercellular junction . ( F ) Schematic of organization of MTs , Ft and Ds in early third instar wing cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 00410 . 7554/eLife . 02893 . 005Figure 1—figure supplement 1 . Organization of the apical MT cytoskeleton in early third-instar wing discs . ( A ) Tubulin staining in third-instar wing pouch . Note the asymmetric accumulation of tubulin on the proximal side of the cell in early third-instar wing pouch . Scale bar , 2 . 5 Δm . White arrows show the proximal distal axis in the wing disc , with distal in the center of the wing pouch , and proximal tissue forming a ring around the future wing blade . ( A′ ) Correlation analysis of image ( A ) for tubulin and Dsh . Note the increasing overlap of signal when the image of tubulin is shifted leftward ( proximal ) relative to the Dsh image , resulting in a peak at distance A . A rightward shift ( distal ) initially shows a decrease in correlation of tubulin and Dsh signals , with a peak at larger distance B , indicating that , on average , the tubulin signals are closer to the proximal than to the distal sides of cells . Similar dorsal or ventral translocations show that tubulin signals are , on average , equidistant between dorsal and ventral sides of the cell . ( B–D ) Orientations of apical MTs correlate with core PCP protein polarization and Fj/Ds gradients throughout wing development . Spatiotemporal correlation of core PCP protein polarity ( Dsh::GFP ) with MT alignment ( TyrTUB ) at time points from 19 to 30 hAPF . Rose plots show distributions of MT orientation ( P-D axis , corresponding to the Fj and Ds gradients , is plotted as horizontal ( 90° ) ; Plots , derived from OrientationJ , are composed of 36 bins of 5° each ) . Note that orientation remains polarized until 30 hAPF , at which time orientations become randomized . Scale bars: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 00510 . 7554/eLife . 02893 . 006Figure 1—figure supplement 2 . Correlation between EB1 comets and MT orientation in Drosophila wing epithelium . ( A and B ) Overlays of 84 frames from time-lapse videos ( Video 1 ) in 24 hAPF pupal wing expressing EB1::GFP and Jupiter::Cherry ( Jupiter is a MT associated protein ) . ( A′ and B′ ) Orientations of MTs and EB1 comets ( panels on the left ) are color-coded using OrientationJ ( 18 ) . Scale bar: 5 μm . ( C ) MT staining with anti-α-tubulin antibody is showing identical pattern as staining with anti-Tyr-tubulin antibody . Scale bar: 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 006 Throughout wing development , image analysis software ( ‘Materials and methods’ ) applied to fluorescence images demonstrated that MTs are strongly aligned along the evolving P/D axis , and the orientation of MTs was reflected in the evolving pattern of polarized PCP proteins , from the radial pattern ( P/D polarity vectors from hinge fold toward center of wing pouch , resulting in concentric circles of P/D cell boundaries ) evident in third instar and early pupal stage ( Figure 1C′–D′ ) to the parallel pattern of 19–30 hAPF pupal wings that presages the hair polarity pattern ( Figure 1—figure supplement 1B″–D″; Strutt et al . , 2002; Ma et al . , 2003; Matakatsu and Blair , 2004; Classen et al . , 2005; Rogulja et al . , 2008; Aigouy et al . , 2010; Hogan et al . , 2011; Sagner et al . , 2012 ) . Importantly , at the time the MT dots appear , there is no evident core PCP protein asymmetry , whereas core PCP asymmetry becomes globally aligned along the P/D axis in slightly older discs only after apical junction-anchored MTs appear ( Figure 1A–C ) , consistent with a requirement of the apical MT cytoskeleton for core module alignment . It is important to note that one would not expect a perfect correlation between MT orientation and orientation of core PCP proteins . The core PCP mechanism , acting through feedback loops , is expected to optimize local alignment of core PCP proteins . This influence is stronger than the directional input produced by global directional signals , and is therefore expected produce the most locally coordinated possible alignment despite the possibility of discontinuities or irregularities in the underlying global biasing inputs ( Ma et al . , 2003 ) . Nonetheless , strong correlation was seen at all times and locations examined . Transmission Electron Microscopy ( TEM ) of 24 hr wings confirmed the previously described polarized organization of MTs that reach across the cell ( Eaton et al . , 1996; Shimada et al . , 2006; Harumoto et al . , 2010; Figure 2A ) , and also revealed that the previously observed associations of planar MTs with apical intercellular junctions ( Fristrom and Fristrom , 1975 ) form juxtaposed , intercellular structures with MT anchoring sites on adjacent membranes of neighboring cells ( Figure 2A , C–E ) . These anchoring sites were preferentially observed at P/D cell boundaries ( Figure 2B ) . The dense , mostly single , MT organizing centers in each cell observed at the ‘dot’ stage ( Figure 1E ) evidently evolve into an arrangement in which multiple organizing centers are distributed around the cell in an oriented and polarized arrangement ( Figure 2 ) . 10 . 7554/eLife . 02893 . 007Figure 2 . TEM micrographs showing apical MT organization in 24 hAPF pupal wings . ( A ) TEM micrograph of MTs in 24 hAPF pupal wing . Note MTs anchored at both sides of intercellular junctions ( arrow ) . ( B ) Graph depicting the localization of MT anchoring sites in the cell relative to the cell centroid . Plot is composed of 20 bins of 18° each . ( C ) Organization of the apical MT cytoskeleton traced from a single micrograph showing MTs spanning the cell preferentially in the P-D orientation . ( D ) TEM revealing planar MTs with anchoring sites on adjacent P-D cell membranes juxtaposed between two neighboring cells , with a sketch of the junctions and MTs . ( E ) Three additional images . The first is a lighter exposure of the image in D , to better reveal the structure of the junctions . Yellow arrowheads show associations of planar MTs with apical intercellular junctions . Scale bars , 100 nm . ( F ) TEM micrograpf showing no association of centrosome with MTs . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 00710 . 7554/eLife . 02893 . 008Figure 2—figure supplement 1 . Distributions of potential MT interacting proteins . ( A ) Patronin , ( B ) Par-1 and ( C ) α-catenin . Patronin is not membrane associated , and Par-1 and α-catenin are cortical but symmetric . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 008 We wished to determine how apical MTs are captured or nucleated at membrane junctions by staining for candidate proteins . EM images showed no association of centrosomes with MTs in non-dividing cells throughout wing development , suggesting that apical MTs are nucleated elsewhere ( Figure 2F ) . We detected the minus-end binding proteins γ-tubulin and Patronin ( Stearns and Kirschner , 1994; Goodwin and Vale , 2010 ) inside the cell , but not at the cell cortex , where they have been observed in other contexts ( Meng et al . , 2008; Feldman and Priess , 2012; Figure 2—figure supplement 1A ) . Consistent with this , patronin knockdown ( Mummery-Widmer et al . , 2009 ) shows no PCP phenotype . MT associated proteins β-catenin ( Armadillo ) ( Ligon et al . , 2001; McCartney et al . , 2001 ) , α-catenin ( McCartney et al . , 2001 ) and PAR-1 ( Doerflinger et al . , 2003; Harumoto et al . , 2010 ) are all present symmetrically at the cell cortex but did not show asymmetric localization at the AJs , suggesting they are not involved in apical MT anchoring ( Figure 1—figure supplement 1B , C ) . These results imply that there is an alternative mechanism that nucleates early , apical non-centrosomal MTs . What is the spatial signal that polarizes non-centrosomal MTs ? Prior evidence suggests that the Ft/Ds/Fj pathway plays a role in organization of apical MTs ( Harumoto et al . , 2010; Marcinkevicius and Zallen , 2013 ) . In the wing , as in other tissues , Ds and Fj are expressed in opposing gradients ( Strutt et al . , 2002; Ma et al . , 2003; Matakatsu and Blair , 2004; Rogulja et al . , 2008; Aigouy et al . , 2010; Hogan et al . , 2011; Sagner et al . , 2012 ) which are converted into subcellular asymmetries of Ft and Ds heterodimers ( Brittle et al . , 2012 ) . Biased subcellular orientations of asymmetric Ft-Ds heterodimers could play a role in polarization of the apical MT cytoskeleton . Consistent with this , the direction of MT growth between 14 hAPF and 30 hAPF have been shown to correlate with Ds and Fj gradient direction ( Strutt et al . , 2002; Ma et al . , 2003; Matakatsu and Blair , 2004; Hogan et al . , 2011 ) in the central part of the pupal wing ( Harumoto et al . , 2010 ) . The possibility that the Ft/Ds/Fj system may organize apical MTs is also supported by prior EB1 comet assays showing that apical MTs are abnormal in ds mutant pupal wings , and that Ds or Ft misexpression perturbs their orientation ( Harumoto et al . , 2010 ) . However , the reported assays were too limited to draw strong conclusions about overall architecture or evolution of the MT pattern ( Harumoto et al . , 2010 ) . Here , our analysis shows that MTs are generally aligned with Ds and Fj gradients from their first appearance in third instar discs ( Rogulja et al . , 2008 ) , when they emerge on proximal sides of the cell cortex ( Figure 1A ) . Throughout larval wing discs and pupal wings , MT orientation correlates with the direction of Ds and Fj gradients ( Figure 1A–D , Figure 1—figure supplement 1 ) . As noted previously ( Brittle et al . , 2012 ) , in imaginal discs , when the tissue is small and gradients appear to be steeper ( Figure 3—figure supplement 1C ) , marked subcellular asymmetry of Ft localization is observed that substantially overlaps core protein distribution ( Figure 3—figure supplement 1A , B ) . We detect a similar relationship in pupal wings ( Figure 3—figure supplements 2 and 3 ) . The steepest regions of the gradients correspond to the most polarized MTs ( compare Figure 1D to Figure 3—figure supplement 2A ) . Therefore , Ds and Fj gradients are appropriately aligned to polarize the MT cytoskeleton and thereby bias core PCP protein polarization from its earliest appearance . These data are consistent with the temporal requirement for Ds in the larval stage ( Matakatsu and Blair , 2004; Aigouy et al . , 2010 ) . Note that caution is required in deciphering the Ds and Fj gradients . Existing data for Fj expression all derive from Fj-LacZ expression , and should therefore be considered only approximate at best . For Ds , low magnification images can be deceptive , since excess cytoplasmic signal , in contrast to the relevant membrane pool , cannot be distinguished , and because smaller cells give the appearance of higher concentrations in low magnification even if membrane intensity is constant . Therefore , we have focused on analyzing subcellular asymmetric localization of Ds in high magnification images , and examples of these data are shown in Figure 3—figure supplements 1–4 . We observe that , for the most part , asymmetry of Ds localization is very similar to that of core protein localization . One exception is the posterior margin of the wing , where Ds often appears to be oriented more posteriorly than is Fmi , though overall levels of asymmetry are modest ( Figure 3—figure supplement 4 , box 5 ) . In this region , Ds and Fmi polarities therefore appear to be less tightly coupled . While we do not know the reason for this , we can speculatively suggest several possibilities . These include ( 1 ) the tendency for the core system to promote local alignment producing a more parallel arrangement than would a direct readout of the Ds pattern; ( 2 ) that oppositely oriented Ft-Ds heterodimers are unevenly distributed despite the even distribution of total Ds; ( 3 ) the tendency of MTs to align along the long axis of the cell may be stronger than the influence of Ft-Ds . Consistent with this , MT orientation correlates more strongly with the long axis of cells than with the Ds asymmetry ( Figure 3—figure supplement 4 , box 5 ) . Finally , ( 4 ) we cannot rule out the possibility that other unknown signals may also be acting on MT orientation ( see ‘An independent directional signal in the wing periphery ? ’ ) . It has been suggested that PCP defects in ds and ft mutants may be due to activation of the Hippo tumor suppressor pathway , which is also controlled by Ft/Ds/Fj , because ft mutant larval wing discs with rescued Hippo signaling show only weak PCP defects limited to the most proximal part of the wing despite the clonal PCP phenotypes observed in pupal and adult wings ( Brittle et al . , 2012 ) . To better assess whether the Ft/Ds/Fj pathway modulates the MT cytoskeleton independent of Hippo signaling , and to do so across the expanse of the wing , we analyzed ftnull mutant flies rescued with FtΔECDΔN-1 , a truncated form of Ft lacking a PCP signaling domain ( Matakatsu and Blair , 2012 ) . These flies are deficient for PCP signaling , but competent for Hippo pathway regulation . They showed PCP defects in the proximal-central part of the adult wing , displaying swirling patterns reminiscent of those in ft clones ( Ma et al . , 2003 ) ( Figure 3A; Matakatsu and Blair , 2012 ) . At 24 hAPF , MTs in the proximal and central part of the wing , where hair polarity is often disturbed , were randomized ( Figure 3A , A′ , Figure 3—figure supplement 5A , A′ ) . In contrast , the peripheral and distal regions of these wings had more coherent hair polarity , with hairs pointing more toward the wing margin than in wildtype wings ( Figure 3A ) , mirroring the orientation of core PCP protein domains ( Figure 3—figure supplement 5A , A″ ) . In these peripheral regions , MTs were ordered and oriented with the hairs and core PCP domains ( Figure 3A″ ) . Flies in which the Ds and Fj gradients were removed showed an essentially identical phenotype ( ds38k fjN7/dsUA071 fjd1; UAS-Ds/TubP-Gal4; Figure 3—figure supplement 5B–B″ ) . Finally , MT orientation was randomized in ft or ds clones in the same proximal part of the wing where it is disturbed in ft mutant wings rescued for Hippo signaling ( Figure 3—figure supplement 5C; see also [Ma et al . , 2008] ) . These data show that in the central part of pupal wings , MT orientation , core PCP protein polarity and adult polarity are strongly dependent on PCP signaling through Ft ( Figure 3—figure supplements 5 and 6 ) . They also suggest the existence of an additional signal , perhaps from the wing margin , that can orient MTs in the periphery of the wing . 10 . 7554/eLife . 02893 . 009Figure 3 . MTs are misaligned in ft and ds mutant wing . ( A–A″ ) Analysis of hair polarity and MT orientation in ftl ( 2 ) fd / ftGRV ActP-Gal4 UAS-FtΔECDΔN-1 mutant flies . ( A′ ) MTs in the proximal and central part of the wing , where hair polarity is disturbed , are randomized . ( A″ ) MTs in the distal/peripheral part of the wing are oriented with the hairs pointing toward the margin ( all plots are composed of 36 bins of 5° each ) . ( B and C ) Ft-dependent instructive re-organization of the MT cytoskeleton . Wildtype cells around ft clones only see Ds in their mutant neighbors , and therefore preferentially accumulate Ft at their junction with the mutant cell . The orientation of MT dots is changed in wildtype cells around ft clones in third instar ( B and B′ ) and 24 hr pupal wings ( C and C′ ) . White arrows show P-D axis and yellow arrowheads show MT spots initiating toward the clone rather than proximally ( B′ ) or MTs organized perpendicular to the clone boundary ( C′ ) rather than proximally-distally ( right panels are high magnification images of left panel boxes ) . ( C″ ) Orientations of MTs in the wildtype cells bordering ft clones calculated using OrientationJ . Scale bars: 50 μm ( B and C ) and 5 μm ( B′–C′ ) . The P-D axis is defined as a radius from the center of the wing disc to the circumference , as described in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 00910 . 7554/eLife . 02893 . 010Figure 3—figure supplement 1 . Orientations of Ft , Ds and core proteins correlate in wing disc . ( A ) Ft staining in third-instar wing pouch . Right panel is a high magnification image of the box in the left panel . ( B ) Dsh::GFP and Ft localization in a mid third-instar wing disc . Note asymmetric localization of Dsh and Ft that overlap on the P/D boundaries . Scale bar: 2 . 5 μm . ( C ) Image of Ds::GFP in third-instar wing disc . On the right and below are vertical section of wing pounch ( along red lines ) showing the gradient of Ds::GFP along the P/D axis . Scale bar: 2 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01010 . 7554/eLife . 02893 . 011Figure 3—figure supplement 2 . Orientations of Ds gradient and core proteins correlate in early pupal wing ( 7 hAPF ) . ( A ) Images of 7 hAPF pupal wings showing Ds staining and Dsh::GFP . Note the Ds staining projecting into the central part of the wing . High Ds expression retracts fully into the hinge around or shortly after 24 hAPF ( Figure 3—figure supplement 3A , B ) . ( B ) A low magnification image of Dsh::GFP at 7 hAPF with overall polarity derived from OrientationJ showing the radial ( P/D axis from wing hinge toward wing margin ) pattern . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01110 . 7554/eLife . 02893 . 012Figure 3—figure supplement 3 . Orientations of Ds gradient and core proteins correlate in late pupal wing ( 24 hAPF ) . ( A and B ) Images of 24 hAPF and 30 hAPF pupal wings showing Ds staining and Dsh::GFP . ( C ) Note the strong overlap of Fmi and Ds subcellular localization in the central part of the wing at 24 hAPF . ( D ) Surface plots of Ds staining at different pupal ages showing the shape of the overall gradient . Scale bars: 100 μm for top image and 5 μm for bottom . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01210 . 7554/eLife . 02893 . 013Figure 3—figure supplement 4 . Orientations of Ds gradient , MTs and core proteins correlate in late pupal wing . The mostly parallel pattern of Ds localization at or shortly after 24 hAPF generally correlates with both the MTs and core PCP polarity , although subtle divergence in the posterior suggests the possibility of other inputs that organize the MTs . ( A ) A low magnification image of Fmi staining and Ds::GFP at 26 hAPF with overall polarity derived from OrientationJ . Below are high magnification images from areas marked with red boxes ( 1–4 ) in top panels . Scale bars: 100 μm for top image and 7 . 5 μm for bottom . ( B ) Note the strong overlap of Fmi and Ds subcellular localization and the correspondence with MT orientation at 24 hAPF . The boxes in B correspond to the equivalent region in panel A—Box 2 and Box 5 . Scale bar: 7 . 5 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01310 . 7554/eLife . 02893 . 014Figure 3—figure supplement 5 . MTs in ft mutant wings are randomized . ( A ) Hair polarity in ftl ( 2 ) fd/ftGRV ActP-Gal4 UAS-FtΔECDΔN-1 adult wing and higher magnification images from the boxes showing hair polarity , Fmi staining pattern and tubulin orientation in 28 hAPF pupal wing . Note that MT orientation is randomized in the central proximal part of the wing ( A′ ) , but is more ordered and oriented toward the margin closer to the periphery ( A″ ) . ( C ) MTs in ds and ft clones are randomized . ( B ) Hair polarity in ds38k fjN7/dsUA071 fjd1; UAS-Ds/TubP-Gal4 adult wing and higher magnification images from the boxes showing hair polarity , Fmi staining pattern and tubulin orientation in 28 hAPF pupal wing . Note that MT orientation is randomized in the central proximal part of the wing ( B′ ) , but is more ordered and oriented toward the margin closer to the periphery ( B″ ) . ( C ) MTs in ds and ft clones are randomized . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01410 . 7554/eLife . 02893 . 015Figure 3—figure supplement 6 . Wing hair polarity in ft mutant wings reflects core PCP orientation in these wings . ( A ) Hair polarity in wildtype wing . ( B ) Hair polarity in ftl ( 2 ) fd/ftGRV ActP-Gal4 UAS-FtΔECDΔN-1 adult wing ( left top ) and core PCP polarity in 28 hAPF pupal wing ( middle ) in same genotype . Right panel is high magnification image of middle panel box . ( C ) Hair polarity in ds38k fjN7/dsUA071 fjd1; UAS-Ds/TubP-Gal4 adult wing ( left top ) and core PCP polarity in 28 hAPF pupal wing ( middle ) in same genotype . Right panel is high magnification image of middle panel box . Orientations of cell boundaries are color-coded using OrientationJ ( 18 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01510 . 7554/eLife . 02893 . 016Figure 3—figure supplement 7 . Overexpression of Wnt4 affects MT polarity . ( A ) Wing hair and MT polarity in wildtype wing . ( B ) Wing hair and MT polarity in en-Gal4 UAS-Wnt4 are disturbed in wildtype cells outside the expressing area of Wnt4 between L3 and L4 veins . Note that wing hairs are pointing toward posterior region where Wnt4 is expressed and that MTs are oriented along the Wnt4 gradient generated outside the expression area of Wnt4 . ( C ) Orientations of MTs in the wildtype cells between L3 and L4 veins calculated using OrientationJ . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 016 To determine whether Ft-Ds signaling regulates MT orientation in wing discs , and to determine whether it is instructive or merely necessary , we studied the boundaries of ft clones . In cells surrounding ft clones in wing discs , where unopposed Ds within the clone is expected to recruit excess Ft to the neighboring cell boundaries , nascent MT bundles are inappropriately polarized toward the cell border abutting the mutant cells ( Figure 3B , B′ ) . Similarly , in pupal wings MTs are perpendicular to the clone boundary ( Figure 3C , C″ ) , consistent with the reported non-autonomy resulting from manipulating the Ft/Ds/Fj system ( Brittle et al . , 2012 ) . To quantify this result , we applied our image analysis tool to cells bordering ( n = 51 ) ft clones in regions where MTs would otherwise be expected to run parallel to the clone border . Figure 3C″ shows that in these cells , MTs are reorganized predominantly perpendicular to the clone border . These results are consistent with the reported reversal of MT orientation in wings with an ectopic Ds gradient , although this was only examined late in the polarization process , during pupal development ( Harumoto et al . , 2010 ) . Therefore , Ft and Ds are both required and instructive for MT organization . Distally biased microtubule ( MT ) -dependent trafficking of Fz-containing vesicles has been shown to occur during polarization of the core PCP proteins , and both are sensitive to MT disruption , suggesting that transport is required for polarization ( Shimada et al . , 2006 ) . Since we have proposed that Dsh is the critical determinant that must be asymmetrically localized ( Amonlirdviman et al . , 2005; Axelrod , 2009 ) , we also examined Dsh::GFP vesicle movement in developing wings in the AJ plane , between 15 and 32 hr after puparium formation ( hAPF ) . The majority of vesicles ( 80% , n = 1192 ) moved along the P/D axis , and showed a significant though modest bias towards distal vs proximal transport ( Figure 4A , C–D; Video 2 ) . Dsh::GFP vesicles exhibited two distinct patterns of trafficking . First , and most commonly , vesicles emerged from one side of the cell and were transported directly across the length of the cell to be incorporated into the membrane of an opposing cellular face ( Figure 4A′; Video 2 ) . Movement was highly linear and processive , though occasional backtracking and zig-zagging was observed . Often , multiple vesicles followed similar paths in a given cell , with vesicle scission and fusion appearing to occur repeatedly at specific sites . Second , a minority of Dsh::GFP vesicles took staggered paths without directional bias , paused frequently , and often left the apical plane of the cell ( Figure 4B , B′; Video 3 ) . The former pattern likely reflects polarized transcytosis , resulting in net transport of Dsh to the distal membrane , while the latter reflects a recycling pathway . Consistently , in fixed specimens , only a minor fraction of Dsh vesicles co-stains with the early endosome marker Rab5 or exocyst protein Sec5 ( Figure 4—figure supplement 1A ) . Thus , a minority of vesicles moves through the recycling pathway while the majority of Dsh vesicles appears to be part of a transcytosis pathway . In contrast , we see no directionally biased trafficking of Vang::YFP vesicles ( Figure 4E ) . Note that biased directional transport of any one component of the core PCP proteins should be sufficient to provide an input bias; bulk transport to achieve the remainder of asymmetric localization is expected to occur by diffusion in combination with feedback at intercellular junctions . Together with prior data , our observations suggest that the ‘distal’ components Fz and Dsh , but not ‘proximal’ components , are subject to directional trafficking . 10 . 7554/eLife . 02893 . 017Figure 4 . Dsh::GFP vesicles move preferentially toward the distal side of the cell . ( A and B ) Timelapse images of Dsh::GFP vesicles in 24 hAPF pupal wing . The majority of Dsh::GFP vesicles displayed directed , fast transcytotic movement with a distal bias ( A ) . A minority of vesicles moved slowly along irregular paths ( B ) ( derived from Video 2 and Video 3 ) . ( A′ and B′ ) Overlays of 73 frames from timelapse videos in A and B . Proximal is to the left and anterior is at the top . ( C ) The ratios of transcytosing Dsh::GFP vesicles at different pupal ages ( n ≥ 1192 ) . ( D ) Ratio of Dsh::GFP vesicles moving toward proximal , distal , anterior or posterior , or not moving ( n = 50 ) . ( E ) Similar plot of Vang::YFP vesicles showing absence of directed trafficking ( n = 81 ) . ( F ) Dsh::GFP vesicle movement inside ft clones showing little net movement and no significant bias ( n = 42 ) . ( G ) Dsh1::GFP vesicles in dsh1 wings showing a bias to P-D vesicle movement among vesicles showing net movement ( n = 120 ) . Numbers were too small to test significance of a possible difference between P and D . Scale bars , 5 μm . Significant differences between proximal and distal at p ≤ 0 . 05 using the binomial test are marked with * . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01710 . 7554/eLife . 02893 . 018Figure 4—figure supplement 1 . Vesicle identification and tubulin staining . ( A ) 24 hAPF Dsh::GFP wing co-labeled for Sec5 and Rab5 . Many Dsh::GFP vesicles do not label with Sec5 or Rab5 ( pink arrowheads ) , while some stain for Sec5 ( yellow arrowhead ) or Rab5 ( blue arrowhead ) . ( B ) Tubulin staining in fz and dsh1 mutant flies . Compare to Figure 1—figure supplement 1C . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01810 . 7554/eLife . 02893 . 019Video 2 . Live in vivo imaging of Dsh::GFP . Live in vivo imaging of Dsh::GFP vesicles in 24 hAPF pupal wing . Refers to Figure 4A . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 01910 . 7554/eLife . 02893 . 020Video 3 . Live in vivo imaging of Dsh::GFP . Live in vivo imaging of Dsh::GFP vesicles in 24 hAPF pupal wing . Refers to Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 020 The observed spatiotemporal organization of the MT cytoskeleton ( Figure 2C ) is suitable for directing biased transport of Dsh and Fz vesicles across cells ( i . e . , transcytosis ) that could bias the direction of core PCP protein polarization . Furthermore , the repeated scission of vesicles from specific regions suggests that vesicle formation may be coupled to the locations of MT-associated junctional structures observed in EM images . Furthermore , the common directionality of these temporally clustered vesicular trafficking events suggests that individual MTs nucleated at or associated with a given junctional density , are likely to have the same polarity . Though the hypothesis that biased trafficking of Dsh and Fz depends on polarized MTs that are organized by Ft/Ds/Fj is appealing ( Harumoto et al . , 2010 ) , no data directly link Ft/Ds/Fj function to directed vesicle trafficking . To test this , we examined Dsh::GFP vesicle movement in proximal ft mutant pupal wing tissue . We observed that , in comparison to wildtype cells , vesicles moved much shorter distances ( or showed no net movement ) , without directional bias , and lacked processivity , instead taking random paths with frequent direction changes ( Figure 4F ) . Apical MTs are oriented independent of core PCP mutants fz ( Shimada et al . , 2006; Harumoto et al . , 2010 ) , vang ( Harumoto et al . , 2010 ) and dsh1 ( Figure 4—figure supplement 1B ) . However , Fz vesicle trafficking was not scored in a core mutant background because vesicle production depends on most or all core proteins ( Shimada et al . , 2006 ) . To verify that directed vesicle trafficking depends on Ft/Ds/Fj but not on core protein asymmetry , we measured movement of Dsh1::GFP in dsh1 flies . No core protein asymmetry is evident in dsh1 wings , but Dsh1::GFP vesicles are still produced , albeit at a lower frequency than in wildtype . We found that , unlike in ft mutant cells , most Dsh1::GFP vesicles move along the P-D axis ( Figure 4G; Video 4 ) , consistent with the presence of oriented MTs , and indicating that oriented trafficking does not depend on core protein asymmetry . We observed that Dsh1::GFP vesicles frequently fail to fuse with membranes , and compared to wildtype , more frequently disappear from the apical plane or do not exhibit overall net movement , perhaps reflecting defects specific to the Dsh1 allele . These results show that apical planar MTs that direct biased transcytosis of Dsh depend on the Ft/Ds/Fj pathway , but not on core module function . Furthermore , the movement of Dsh1 vesicles , the faster kinetics of transcytosing Dsh vesicles , and the greater processivity , compared to Fz vesicles all suggest that Dsh and Fz vesicles may be at least partially distinct populations . 10 . 7554/eLife . 02893 . 021Video 4 . Live in vivo imaging of Dsh1::GFP in dsh1 mutant flies . Live in vivo imaging of Dsh1::GFP vesicles in 24 hAPF dsh1 pupal wing . Refers to Figure 4G . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 021 Our results thus far suggest that gradients of Ds and Fj expression , by producing asymmetric orientation of Ft-Ds heterodimers , provide directional information to bias core protein polarization . Given the apparent variation in asymmetry of Ft-Ds dimers at different times and places in wing development , we wished to assess the potential consequences of this variation on core PCP protein asymmetry . We therefore simulated this mechanism by adapting our previously described mathematical model for PCP signaling ( Amonlirdviman et al . , 2005; Ma et al . , 2008 ) . The modified model establishes a MT network with polarity determined by the relative concentrations of Ft on any side of a cell . User-defined input gradients of Ds and Fj determine Ft concentrations in a manner consistent with the experimentally defined model . Dsh is then transported toward the plus ends of MTs , while still permitting bulk movement of all components by diffusion ( see Supplementary file 1 ) . We first validated the model by correctly reproducing the domineering non-autonomy ( or lack thereof ) surrounding clones of core PCP mutants ( Figure 5—figure supplement 1 ) . Furthermore , we confirmed that the model correctly simulates the ability of the core module to propagate polarization across small ft mutant clones ( Figure 5—figure supplement 1 ) . The model then allowed us to predict the response to different configurations of the Ds or Fj gradients . As discussed above , there is considerable ambiguity about the shape of the Ds gradient through wing development , but to a first approximation , it appears to undergo considerable change from larval wing discs , where there is a comparatively linear gradient , at least in the distal portion of the wing not hidden by tissue folds , to one with a steep drop and very shallow or flat portion in the 24–30 hr pupal wing ( Figure 3—figure supplements 1–4; Ma et al . , 2003; Matakatsu and Blair , 2004; Hogan et al . , 2011 ) . In third instar discs , gradients of Ds and Fj are roughly linear ( Figure 3—figure supplement 1C ) . In simulation , oppositely oriented linear gradients of Ds and Fj polarize a field of cells with similar kinetics and identical steady state levels of polarization across the entire field . Whereas in the larval wing disc , the Ds gradient may be gradual , in the pupal wing , the gradient of Ds approaches a step gradient as it rearranges first to a projection of high Ds in the central part of the pupal wing , and later to a very high proximal concentrations and a shallow or even flat distal distribution . Notably , simulation of a linear gradient , a step gradient , or a steep proximal Ds gradient and a shallow or flat distal gradient produces identical levels of steady state polarization across the field and similar proximal and distal kinetics , showing that the mechanism is not expected to be very sensitive to the precise shape of the Ds gradient ( Figure 5 ) . In the cases of a steep local Ds gradient , propagation of Ft-Ds polarity into an adjacent shallow or flat region is weak , and limited to two columns of cells ( data not shown ) , most likely due to absence of a robust feedback mechanism in our model . Similarly , propagation of Ft-Ds polarization in vivo is seen to be much weaker than that of the core PCP mechanism ( c . f . Ambegaonkar et al . , 2012; Brittle et al . , 2012 ) . Therefore , propagation of polarity through the shallow or flat region is primarily due to polarization and propagation of the core PCP system . 10 . 7554/eLife . 02893 . 022Figure 5 . Simulations of polarization kinetics with different shapes of Ds gradients or imposed MT structures , using a mathematical model incorporating a simple representation of the Ft/Ds/Fj system to polarize MTs . Input Ds gradients are shown on the left . Resulting ( or imposed , on the right side of the last example ) MT organization on the proximal or distal portion of the gradient are plotted ( center ) . Kinetics of polarization of cells in column 8 ( proximal = left ) , column 15 ( center ) or column 23 ( distal = right; see Figure 5—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 02210 . 7554/eLife . 02893 . 023Figure 5—figure supplement 1 . Simulation results . Simulations of clones of fz ( A ) , dsh ( B ) and ft ( C ) null clones validating that the mathematical model captures the domineering non-autonomy phenotypes as well as the ability of the core system to propagate through ft mutant tissue . The color scale represents the magnitude of the vector sum of Dsh in each cell . The length of the hair and its distance from the center of the cell is also plotted proportionally to the vector sum of Dsh in each cell . ( D ) A 6 × 30 cell grid used to simulate polarization kinetics . Cells with kinetics plotted in Figure 5 are marked with asterisks . Greyscales represent the combined quantity of Dsh on either side of each shared cell boundary , as would be seen with light microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 02893 . 023 In the distal part of the wing , MTs are oriented in the P-D direction , but have no detectable polarity bias . We therefore simulated several additional conditions . First , we simulated a steep proximal gradient with a distal zero level of Ds that produces randomized distal MT orientation in our model . Global directional input is therefore restricted to the proximal region . In this case , we see equivalent proximal and distal steady state polarization , but with a substantial delay in reaching steady state in the distal region , reflecting time needed to propagate polarity from proximal to distal through the core mechanism . To simulate MTs that are P-D oriented , but without a measurable plus-end bias , as are observed in pupal wings , we enforced this MT architecture in the distal wing , with a steep proximal Ds gradient . Simulation of this condition predicts modestly faster core PCP polarization compared to random distal MTs , but not to change steady state core PCP polarization . Therefore , the observed unbiased but oriented MTs in the distal wing might facilitate more rapid core PCP polarization and also maintenance of polarity in the face of perturbations . In similar simulations , we tested the response to different configurations of the Fj gradient . Again , we found that steady state polarization is insensitive to the gradient configuration , with only slight differences in kinetics ( data not shown ) . A similar result was obtained when simultaneously altering the shapes of both gradients ( data not shown ) . From these simulations , we conclude that so long as the Ds and Fj gradients are in the proper direction , their potentially evolving profiles are not expected to have a substantial effect on the resulting core PCP polarity . Distal ( peripheral ) polarity is independent of Ft function; polarity in this region may depend on a spatial signal perhaps originating from the wing margin . A candidate for this signal is the proposed redundant functions of Wnt4 and Wg . Both are expressed at the wing margin , combined loss-of-function produces a mild polarity phenotype , and overexpression of Wnt4 to a much greater extent than Wg perturbs hair polarity ( Lawrence et al . , 2002; Lim et al . , 2005; Wu et al . , 2013 ) . Based on a cell culture assay , these Wnts were suggested to impact core PCP function by blocking interactions between Fz and Vang . However , our observation that a Ft independent signal might polarize MTs near the wing margin suggests that other possible mechanisms should be considered . Notably Wnt4 ( but not Wg ) overexpression reorganizes MTs ( Figure 3—figure supplement 7 ) , suggesting that a different possible mechanism for Wnt4 function should be entertained .
Together , our findings support the hypothesis that a polarized MT cytoskeleton orients PCP throughout wing development by directing the trafficking of Dsh containing vesicles . Furthermore , they confirm that the Ft/Ds/Fj PCP module directs orientation of this apical MT cytoskeleton , at least in the proximal central portion of the wing . We infer that a second signal , acting near the wing margin and perhaps originating from the margin , can also organize MTs to orient the core PCP mechanism . The recent finding that Wnts expressed at the wing margin regulate PCP suggests a possible identity for this signal ( Wu et al . , 2013 ) . We propose that in third instar wings , when the tissue is smaller , the two signals are largely redundant , so that defects in Hippo-rescued ft mutants are limited to the most proximal regions ( Feng and Irvine , 2007; Brittle et al . , 2012; Matakatsu and Blair , 2012; Pan et al . , 2013 ) , whereas in larger pupal and adult wings , Hippo-rescued ft mutants show larger regions of disturbed polarity . One difficulty in understanding how a wing margin-based signal might contribute to polarization is that much of the anterior and posterior margin is parallel to the direction of polarization , while the distal portion of the margin is perpendicular . Additional studies will be needed to understand potential signals from the margin . Notably , even in regions well polarized by the presumed wing margin signal , ectopic Ds expression can reorganize polarity ( Matakatsu and Blair , 2004; Harumoto et al . , 2010 ) . Furthermore , the strong correspondence of MT orientation and core protein orientation throughout wing morphogenesis , their overall correspondence to Ds-Fj gradients , and the ability of altered Ft or Ds expression patterns to reconfigure MT orientation , suggest that these gradients provide instructional information for core PCP orientation , at least in the proximal and central region of the wing . This signal likely acts in conjunction with other molecular signals , particularly in the peripheral region of the wing , and perhaps with mechanical inputs such as cell flow and cell elongation ( Aigouy et al . , 2010 ) . Recently , it was shown that the tissue and compartment specific expression predominance of the Pk vs Spiny-legs isoforms of Pk determines the direction of Ds and Fj gradient interpretation ( Ayukawa et al . , 2014; Olofsson and Axelrod , 2014 ) . Thus , for example , polarization of the core module can occur in the same direction in the Anterior and Posterior compartments of the abdomen despite oppositely oriented Ds and Fj gradients in these compartments . In summary , we provide evidence favoring the model that the Ft/Ds/Fj global PCP module , together with a partially redundant and as yet unidentified peripheral wing signal , orients apical polarized microtubules , directing Dsh-vesicle transcytosis , and thereby imparting directional information to the core PCP module . To what extent other global signals may function in other tissues remains to be determined . The presence of polarized MTs suggests that a MT dependent global cue may also function in vertebrate PCP ( Vladar et al . , 2012 ) .
The following fly lines and mutant alleles were used:OreR , Dsh::GFP , actinP-Vang::EYFP , Ds::GFP ( Brittle et al . , 2012 ) , ciGal4/UAS-EB1::GFP , ubi-Jupiter::mCherry , armP-Fz::GFP , ftl ( 2 ) fd FRT40A Dsh::GFP/NLS::mRFP FRT40A; T155Gal4 UAS-FLP/+ , ftGRV FRT40A , ftGRV FRT40A/ftl ( 2 ) fd FRT40A; UAS-FtΔECDΔN1 ( Matakatsu and Blair , 2012 ) /ActP-Gal4 , ds38k fjN7/dsUA071 fjd1; UAS-Ds/TubP-Gal4 , hs-FLP; dsUA071 FRT40A/FRT40A Tub-Gal80; TubP-Gal4 UAS-mCD8GFP/+ , ftGr-V FRT40A/FRT40A Tub-Gal80; TubP-Gal4 UAS-mCD8GFP/+ , ftl ( 2 ) fd dGC13/ftl ( 2 ) fd d1 , enGal4/UAS-Wnt4 . Drosophila pupal wings were prepared for imaging as previously described ( Axelrod , 2001 ) . Primary antibodies were as follows: mouse anti-Fmi ( #74 , Developmental Studies Hybridoma Bank ) , mouse anti-Arm ( Developmental Studies Hybridoma Bank ) , rabbit anti-alpha Tubulin ( Abcam , Cambridge , UK ) , rat anti-tyrosinated Tubulin ( Abcam , Cambridge , UK ) , rat anti-Ds and rat anti-Ft ( Yang et al . , 2002 ) . Images were acquired on a Leica TCS SP5 AOBS confocal microscope using a 63x objective and processed with LAS AF ( Leica ) . For EM analysis , wing imaginal discs and pupal wings were fixed in a mixture of 4% paraformaldehyde and 2% glutaraldehyde in 0 . 1 M sodium cacodylate buffer , pH 7 . 3 overnight at RT . Samples were post-fixed in 1% osmium tetroxide in 0 . 1 M PBS for 1 hr at RT , stained with uranyl acetate , dehydrated with a graded ethanol series and embedded in EMbed-812 ( Electron Microscopy Sciences ) . Ultrathin sections were cut and analyzed with a JEOL JEM-1400 microscope using a Gatan Orius Camera . For live imaging , pupae were removed from incubation at 25°C 10 min prior to the desired time APF . Pupae were mounted on a small piece of double-sided tape and forceps were used to dissect open a small window in their pupal cases to provide visual access to the live pupal wing . Approximately 50 μl of halocarbon oil was placed over each dissected pupa to allow its release from the tape , following which the pupae were mounted on a VivaScience petriPERM 50 hydrophobic membrane disk in halocarbon oil between pieces of hydrated Whatman paper for in vivo confocal fluorescence microscopy . Videos showing Dsh::GFP and membrane RFP ( used to mark wildtype cells ) show that internalized Dsh::GFP is always seen to co-stain with RFP . Furthermore , Shimada et al . reported that in fixed images the majority of Dsh and Fz vesicles overlap ( Shimada et al . , 2006 ) . We are therefore confident that internalized Dsh::GFP is in vesicles . For analysis of MT orientation we used OrientationJ software ( Rezakhaniha et al . , 2012 ) . We analyzed an average of 20 images from 5 to 10 wing discs or pupal wings for each time point or genotype . The images were taken from appropriate parts of the wing as shown in figures and aligned along the P/D axis ( which is plotted as horizontal ) of the wing disc or pupal wing . Analysis of MT anchoring sites in 24 hAPF wings was done using ImageJ . To analyze the localization of apical tubulin in early third-instar wing pouch we used the cross correlation method as previously described ( Matis et al . , 2012 ) . For analysis of live imaging time-series , vesicles were only measured and quantified if they were visible in two or more consecutive frames at the level of the adherens junctions . Images were taken at 5 s intervals . We analyzed manually 1192 particle tracks to calculate the net direction of movement ( proximal , distal , anterior , posterior or ‘stuck’–no change in location between any two consecutive frames ) . A vector was taken between the first and last points of each track to calculate the net direction of movement . A mathematical model , incorporating the proposed mechanism of the Ft/Ds/Fj module to organize MTs , has been created based upon our previously published ODE model ( Ma et al . , 2008 ) . Details are described in Supplementary file 1 . Simulations of clones ( Figure 5—figure supplement 1 ) or of wildtype grids with user defined Fj and Ds gradients ( Figure 5 ) were performed to assess kinetics and quasi-steady state levels of polarization . | Almost all cells exhibit some sort of polarity: the epithelial cells that line the digestive tract , for example , have an apical domain , which faces out , and a basal domain , which faces the tissue underneath . Some epithelial cells also exhibit planar cell polarity: this involves key structures within the cell being oriented along an axis within the plane of an epithelium . Disruption of planar cell polarity is associated with various developmental defects . It is known that the planar polarity of epithelial cells relies on two molecular complexes—a ‘core’ complex and a signaling complex called the Ft/Ds/Fj system—working together . While each of these complexes contributes to whole tissues having the correct polarity , the way they interact to achieve this is not fully understood . Now , by studying epithelial cells in the wings of fruit flies , Matis et al . have provided evidence for a specific model for this interaction . The process starts with the Ft/Ds/Fj signaling complex , which orients structures called microtubules inside the cell . Microtubules are involved in providing structural support for cells , and also in the transport of organelles within cells . Once the microtubules are oriented in the correct direction , they help to orient the core complex by moving some of the proteins that make up this complex in a specified direction . An important future challenge will be to understand how the proteins in the Ft/Ds/Fj system interact with microtubules to give them their orientation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2014 | Microtubules provide directional information for core PCP function |
Axon injury triggers a series of changes in the axonal cytoskeleton that are prerequisites for effective axon regeneration . In Caenorhabditis elegans the signaling protein Exchange Factor for ARF-6 ( EFA-6 ) is a potent intrinsic inhibitor of axon regrowth . Here we show that axon injury triggers rapid EFA-6-dependent inhibition of axonal microtubule ( MT ) dynamics , concomitant with relocalization of EFA-6 . EFA-6 relocalization and axon regrowth inhibition require a conserved 18-aa motif in its otherwise intrinsically disordered N-terminal domain . The EFA-6 N-terminus binds the MT-associated proteins TAC-1/Transforming-Acidic-Coiled-Coil , and ZYG-8/Doublecortin-Like-Kinase , both of which are required for regenerative growth cone formation , and which act downstream of EFA-6 . After injury TAC-1 and EFA-6 transiently relocalize to sites marked by the MT minus end binding protein PTRN-1/Patronin . We propose that EFA-6 acts as a bifunctional injury-responsive regulator of axonal MT dynamics , acting at the cell cortex in the steady state and at MT minus ends after injury .
In mature nervous systems axons regenerate poorly after injury , leading to permanent functional deficits . Both the nature of the extracellular environment and the intrinsic growth competence of the neuron contribute to the extent of axon regeneration ( Case and Tessier-Lavigne , 2005 ) . The mammalian central nervous system ( CNS ) expresses a variety of environmental regeneration inhibitory factors , including myelin-associated proteins , chondroitin sulfate proteoglycans and glial scar tissue that functions as a physical barrier ( Schwab , 2004; Silver and Miller , 2004 ) . However genetic removal of these inhibitory factors results in only limited improvement in regeneration of severed axons ( Lee et al . , 2009 , 2010 ) . Recent studies have strongly supported the importance of cell-intrinsic determinants in axon regeneration . Loss of function in cell-intrinsic growth inhibitors such as Phosphatase and Tensin homolog , PTEN , and Suppressor Of Cytokine Signaling-3 , SOCS3 , can dramatically improve axon regrowth even in the inhibitory CNS environment ( Park et al . , 2008; Sun et al . , 2011 ) . Genetic and pharmacological manipulation of cell autonomous signaling pathways can dramatically improve regrowth of severed axons in various injury paradigms ( Moore et al . , 2009; Hellal et al . , 2011; Sengottuvel et al . , 2011; Shin et al . , 2012; Watkins et al . , 2013; Ruschel et al . , 2015 ) . During developmental axon outgrowth and in regenerative regrowth of mature neurons , the formation and extension of growth cones involve extensive remodeling of the microtubule ( MT ) cytoskeleton ( Bradke et al . , 2012; Chisholm , 2013 ) . Cellular compartments undergoing rapid morphological changes , such as axonal growth cones , are enriched in dynamic MTs ( Suter et al . , 2004 ) , while mature axons or dendrites contain predominantly stabilized MTs ( Baas et al . , 1993 ) . When an axon is injured , MTs are locally disassembled or severed , potentially creating free plus ends for new MT polymerization . Subsequently , the number of growing MTs increases , followed by more persistent MT growth correlated with formation of regenerative growth cone and axon extension ( Erez and Spira , 2008; Ghosh-Roy et al . , 2012 ) . Complete removal of an axon also leads to dramatic upregulation of MT dynamics in the soma and dendrites ( Stone et al . , 2010 ) . MT disorganization contributes to dystrophic end bulb formation after injury in CNS ( Ertürk et al . , 2007 ) . Moderate stabilization of MT dynamics by Taxol or other MT stabilizers can promote axon regrowth in vitro and in the mammalian CNS ( Usher et al . , 2010; Hellal et al . , 2011; Sengottuvel et al . , 2011; Ruschel et al . , 2015 ) ; the effects of Taxol in vivo have been partly replicated ( Popovich et al . , 2014; Ruschel et al . , 2015 ) . Thus , there is a critical need to define the endogenous regulators of MTs after injury . In a large-scale screen for genes affecting adult axon regeneration in Caenorhabditis elegans , we identified Exchange Factor for ARF-6 ( EFA-6 ) as an intrinsic inhibitor of axon regrowth ( Chen et al . , 2011 ) . The EFA-6/EFA6 protein family is conserved from yeast to mammals , and is defined by its Sec7 domain , which confers guanine exchange factor ( GEF ) activity for Arf6 GTPases ( Franco et al . , 1999 ) . Unexpectedly , the regrowth-inhibitory function of EFA-6 is independent of its GEF activity , and instead is mediated by its N-terminal domain ( Chen et al . , 2011 ) . The EFA-6 N-terminal domain inhibits MT growth at the cell cortex of C . elegans embryos via a conserved motif of 18 amino acids ( O'Rourke et al . , 2010 ) . Nonetheless , the mechanism by which EFA-6 regulates MT dynamics is unknown . Here we reveal that axon injury triggers rapid and transient relocalization of EFA-6 , concomitant with an initial downregulation of axonal MT dynamics . The N-terminal 18-aa motif is required for injury-induced relocalization and for inhibition of axonal MT growth . We show that the EFA-6 N-terminal domain interacts with MT associated proteins TAC-1 , a member of the transforming acidic coiled-coil ( TACC ) family , and ZYG-8 , an ortholog of doublecortin-like kinase ( DCLK ) . TAC-1 and ZYG-8 are required for initiation of axon regeneration , and their overexpression can promote regrowth . We further show that injury triggers relocalization of EFA-6 and TAC-1 to sites overlapping with the MT minus end binding protein Patronin/PTRN-1 . We propose that EFA-6 is a bifunctional injury-responsive regulator of MT dynamics , acting at the cell cortex in the steady state and at MT minus ends after axon injury .
In the one-cell stage embryo EFA-6 localizes to the plasma membrane via its C-terminal PH ( pleckstrin homology ) domain , and this plasma membrane localization of EFA-6 is necessary for it to inhibit cortical MT growth ( O'Rourke et al . , 2010 ) . To determine the subcellular location of EFA-6 in neurons , we expressed a series of GFP-tagged EFA-6 fusion proteins . Full-length EFA-6 tagged with GFP at the N- or C-termini , expressed at a range of concentrations , localized to the plasma membrane of the soma and processes of neurons ( Figure 1A , B , Figure 1—figure supplement 1A ) ; deletion of the PH domain ( FLΔPH ) resulted in cytosolic localization ( Figure 1C , upper panel ) . The first 150 residues of the EFA-6 N-terminus ( N150 ) , expression of which inhibits axon regrowth ( Chen et al . , 2011 ) , was localized to the cytosol similarly to FLΔPH ( Figure 1C , D ) . Conversely , EFA-6 proteins lacking the N-terminal 150 amino acids ( FLΔN150 ) showed plasma membrane localization resembling that of full-length EFA-6 ( Figure 1B , E ) . 10 . 7554/eLife . 08695 . 003Figure 1 . Axon injury triggers rapid relocalization of Exchange Factor for ARF-6 ( EFA-6 ) , mediated by its N-terminal domain . ( A ) Single focal plane images of PLM ( top ) and nerve ring ( bottom ) showing membrane localization of EFA-6 . Transgenes: Pmec-4-EFA-6::GFP ( juEx6467 ) ( top ) and Prgef-1-GFP::EFA-6 ( juEx6374 ) ( bottom ) . ( B ) Localization of full length EFA-6 ( Pmec-4-GFP::EFA-6 , juEx6160 ) before , 2 min after , and 1 hr post axotomy . Projections of confocal z stacks , inverted grayscale; enlargements in inserts . Bottom , fluorescence intensity along line scan . ( C–E ) Localization of GFP::EFA-6 fusion protein lacking the PH domain ( FLΔPH ) ( Pmec-4-GFP::EFA-6 FLΔPH , juEx6453 ) , EFA-6 N-terminal 150 aa ( N150 ) ( Pmec-4-GFP::EFA-6N150 , juEx3531 ) , and EFA-6 lacking the N-terminus ( FLΔN150 ) ( Pmec-4-GFP::EFA-6FLΔN150 , juEx6154 ) . ( F ) Colocalization of EFA-6FL and EFA-6N150 puncta after injury . Localization of EFA-6FL::GFP and EFA-6N150::mKate2 ( juEx6522 ) in PLM before and after axotomy . EFA-6 full length protein and N terminus relocalize to overlapping puncta . ( G , H ) Requirement for the 18-aa motif for relocalization of EFA-6FL and EFA-6N150 . Localization of GFP::EFA-6FLΔ18aa ( juEx6156 ) , and GFP::EFA-6N150Δ18aa ( juEx3535 ) in touch neurons before and 2 min after axotomy . ( I ) Quantitation of puncta before and 2 min after injury in axons expressing different EFA-6 fragments . Statistics , one-way ANOVA with Bonferroni post test; n = 5 for each bar; **p < 0 . 01 , ns , not significant . Scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 00310 . 7554/eLife . 08695 . 004Figure 1—figure supplement 1 . Injury induced relocalization of EFA-6 is independent of expression level or location of tag . ( A ) Pmec-4-GFP::EFA-6 transgenic lines generated at different concentrations of injected plasmid ( juEx2642 was generated at 30 ng/μl , juEx3188 was generated at 1 ng/μl ) displayed similar localization , before and 2 min post axotomy . ( B ) The relocalization of EFA-6 is independent of the site of GFP tagging . Transgenes: Pmec-4-GFP::EFA-6 ( juEx6160 ) . Pmec-4-EFA-6N150::GFP::EFA-6C ( juEx6463 ) ( GFP inserted between aa 363–364 ) and Pmec-4-EFA-6::GFP ( juEx6467 ) . All three transgenes cause premature PLM termination to similar extents ( ∼30% undershooting ) . Red arrows , site of axotomy . Scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 00410 . 7554/eLife . 08695 . 005Figure 1—figure supplement 2 . Full-length EFA-6 and N terminus relocalize to the same puncta after injury . ( A ) Colocalization of EFA-6FL::GFP and EFA-6N150::mKate2 in PLM before and after axotomy . Enlarged images of the regions in boxes ( 25 μm length ) are shown in Figure 1F . Graphs of line scans are shown below . EFA-6FL::GFP was primarily localized to the plasma membrane and EFA-6N150::mKate2 was diffuse in soma and axon before injury; both became punctate after injury , and these puncta co-localized . ( B ) Localization of EFA-6FL::GFP and EFA-6N150Δ18::mKate2 in PLM before and after axotomy . Enlarged images of the regions in boxes and graphs of line scans are shown below . EFA-6N150Δ18::mKate2 was diffuse before and after axotomy . ( C ) Velocity of relocalization spread for GFP::EFA-6FL and GFP::EFA-6N150 , calculated by measuring the distance between injury site and the boundary between punctate and even GFP distribution in the distal or proximal axon at 2 . 3 s ( 10 frames ) post axotomy . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 00510 . 7554/eLife . 08695 . 006Figure 1—figure supplement 3 . Injury induced relocalization of EFA-6 . ( A ) Representative images of GFP::EFA-6N150 before and post axotomy . Before axotomy , GFP::EFA-6N150 was diffuse in the axon; by 2 min post axotomy , it became punctate; 20 min post axotomy , GFP recovered to a diffuse pattern . ( B ) Line scan along axon at different time points . ( C ) Quantitation of GFP::EFA-6N150 puncta at different times ( before , 2 min and 20 min after axotomy ) . Transgene: Pmec-4-GFP::EFA-6N150 ( juEx3531 ) . Statistics: One-way ANOVA with Bonferroni post test . ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 . ( D ) GFP::EFA-6 expressed in motor neurons commissures showed similar relocalization response to injury . Transgene: Prgef-1-GFP::EFA-6 ( juEx6374 ) . ( E ) GFP::EFA-6N150 expressed in motor neurons displayed similar injury-induced relocalization . Transgene: Punc-25-GFP::EFA-6N150 ( juEx6229 ) . ( F ) Injury to the soma or dendrite of PLM led to similar EFA-6 relocalization . ( G ) ARF-6::GFP localization before and after axotomy . Transgene: Pmec-4-ARF-6::GFP ( juEx5906 ) . Images of PLM before injury and 2 min post injury are shown . Red arrows indicate axotomy sites . Scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 006 We next examined how axon injury affected EFA-6 localization . Within seconds of laser axotomy of the PLM axon , GFP::EFA-6FL redistributed from a generally even plasma membrane localization to more discrete puncta ( Figure 1B , Video 1 ) . This relocalization was also observed when GFP was tagged to the C terminus or in the middle of EFA-6 ( Figure 1—figure supplement 1B ) , and did not require the PH domain ( Figure 1C ) . GFP::EFA-6N150 , although not membrane associated , also became punctate after injury , whereas GFP::EFA-6FLΔN150 did not relocalize after axon injury ( Figure 1D , E ) . Full-length EFA-6 and the N terminal domain ( N150 ) appear to relocalize to the same puncta in response to injury , as shown by co-expressing EFA-6FL::GFP and EFA-6N150::mKate2 ( Figure 1F , Figure 1—figure supplement 2A ) . mKate2::EFA-6N150Δ18 ( which does not relocalize , see below ) did not co-localize with EFA-6FL::GFP after injury , suggesting EFA-6 puncta are not due to non-specific aggregation of proteins after injury ( Figure 1—figure supplement 2B ) . The region of relocalized EFA-6FL or EFA-6N150 expanded bidirectionally from the injury site at ∼8 μm s−1 , into the soma and dendrite ( Figure 1—figure supplement 2C , Video 1 ) . The density of injury-triggered puncta of EFA-6FL and EFA-6N150 gradually decreased over the next 20–60 min ( Figure 1B , Figure 1—figure supplement 3A–C ) . We observed similar re-localization of GFP::EFA-6 after injury of motor neuron axons ( Figure 1—figure supplement 3D , E ) , as well as when injury was delivered to the soma , distal axon , or posterior processes of mechanosensory neurons ( Figure 1—figure supplement 3F ) . These results indicate that injury-triggered EFA-6 relocalization occurs in multiple neuron types , irrespective of the site of injury and independent of the Sec7 or PH domains . The N-terminal 1–70 aa ( N70 ) was the smallest fragment tested that displayed relocalization after injury ( Figure 1I ) . We tested a variety of other neuronal proteins , including the MT plus-end binding proteins EBP-1 and EBP-2 , KLP-7/kinesin-13 ( Chen et al . , 2011; Ghosh-Roy et al . , 2012 ) , ARF-6 ( the presumed substrate for EFA-6's GEF activity ) , SAX-3 ( transmembrane receptor ) and synaptic vesicle proteins ( SNB-1/synaptobrevin , RAB-3/GTPase , UNC-57/endophilin ) , and found that none showed similar relocalization after axon injury ( Figure 1—figure supplement 3G , and data not shown ) . Thus , injury-triggered relocalization is specific to EFA-6 . 10 . 7554/eLife . 08695 . 007Video 1 . Injury-induced GFP::EFA-6 relocalization in neurons ( ALM ) . Transgene: Pmec-4-GFP::EFA-6 ( juEx6160 ) . The video is 103 s , taken at 1 s/frame . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 007 The punctate distribution of EFA-6 after injury suggested that EFA-6 might become sequestered to a subcellular compartment . To address whether injury alters EFA-6 mobility within the cell we performed FRAP ( Fluorescence Recovery After Photobleaching ) . In uninjured neurons , EFA-6FL::GFP recovered with t½ = 4 s and an immobile fraction of 25% . In contrast , EFA-6FL::GFP puncta 2 min after injury showed dramatically reduced recovery , with >85% of the protein in the immobile fraction ( Figure 2A , B ) . We were not able to calculate t½ due to the extremely low recovery rate . By 1 hr after injury , EFA-6FL::GFP partially returned to its steady state localization ( Figure 1B ) , and its recovery rate was partially restored , with an immobile fraction of 50% ( Figure 2A , B ) . GFP::EFA-6N150 gave similar FRAP results ( not shown ) . This analysis suggests that after injury EFA-6 is sequestered to subcellular structures . 10 . 7554/eLife . 08695 . 008Figure 2 . Injury-induced relocalization of EFA-6 correlates with its ability to regulate regrowth and microtubule ( MT ) dynamics . ( A ) FRAP of GFP::EFA-6 ( juEx6160 ) before and after axon injury; regions of interest indicated by red circles; green circles were used to calibrate baseline fluorescence intensity . ( B ) Normalized average fluorescence intensity after FRAP . ( C ) PLM termination defects in efa-6 ( lf ) mutants and EFA-6 overexpressing transgenic animals . n = 40 for each bar . See Figure 2—figure supplement 1 for definitions of PLM overshooting and undershooting . ( D ) Normalized axon regrowth of efa-6 ( lf ) mutants and EFA-6 overexpressors . n ≥ 10 . ( E ) Quantitation of EBP::GFP dynamics in intact axons from wt , efa-6 ( tm3124 ) and transgenic animals expressing different EFA-6 fragments under Pmec-4 promoter . n ≥ 10 . Statistics , one-way ANOVA with Bonferroni post test; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05; ns , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 00810 . 7554/eLife . 08695 . 009Figure 2—figure supplement 1 . EFA-6 relocalization correlates with protein function in axon termination . ( A ) Representative images of PLM development ( see quantitation in Figure 2D and strain information in Supplementary file 1C ) . Asterisk indicates ALM soma . A normal PLM axon usually terminates posterior to ALM soma . The terminus ( indicated by a red triangle ) of an overshooting PLM axon is anterior to the ALM soma . Scale , 50 μm . ( B ) Transgenic animals with relatively normal PLM morphology ( left panel ) were used for axon regeneration analysis ( right panel ) . Arrow: injury site . ( C ) Representative kymographs of EBP-GFP ( juIs338 ) in wild type , efa-6 ( tm3124 ) and in transgenic strains expressing EFA-6 fragments . See quantitation in Figure 2F and strain genotypes in Supplementary file 1C . ( D ) Locomotor defects due to pan-neural overexpression of EFA-6N150 or EFA-6N150Δ18aa . Overexpression of EFA-6N150 in all neurons results in small , uncoordinated animals; deletion of the 18 aa motif abolishes this effect . ( E ) The locomotion velocity of animals overexpressing EFA-6N150 ( rgef-1 promoter ) is reduced compared to wild type , whereas EFA-6NΔ18aa-overexpressing animals display normal locomotion; WormTracker analysis . Statistics: One-way ANOVA with Bonferroni post test . ***p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 00910 . 7554/eLife . 08695 . 010Figure 2—figure supplement 2 . The conserved 18-aa motif in the EFA-6 N-terminus is a region of local protein order . ( A ) Cartoons of EFA6 protein family domains , based on sequences from NCBI . The N-termini of EFA6 family members overall have low sequence complexity and lack recognizable domains , with the exception of a predicted PDZ domain in Drosophila EFA6 . A conserved 18-aa motif ( red box ) is found in the N termini of Caenorhabditis elegans and Drosophila EFA6 proteins . Drosophila EFA6 has an N-terminal PDZ domain ( brown box ) . ( B ) Plot of intrinsic protein disorder score for C . elegans EFA-6 using the metapredictor PONDR-FIT ( disprot . org ) ( Xue et al . , 2010 ) . The Sec7 , PH , and CC domains show low disorder probability , consistent with their defined tertiary structures . The EFA-6 N terminus has an overall high disorder probability except for the 18-aa motif . ( C ) C . elegans efa-6 intron-exon structure , with deletion mutations indicated as black boxes . Protein domains are colored as in ( A ) . ( D ) Pan-neural ( Prgef-1 ) expression of the EFA-6 N terminus from a single copy insertion transgene juSi86 rescues the enhanced regrowth of efa-6 ( ju1200 ) and efa-6 ( tm3124 ) . Statistics: One-way ANOVA with Bonferroni post test . ***p < 0 . 001 , **p < 0 . 01 , *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 010 efa-6 ( lf ) mutants display significantly increased PLM axon regrowth following laser axotomy in adults , as well as impenetrant developmental overshooting of PLM axons; conversely , overexpression of EFA-6 strongly inhibits axon regrowth after axotomy , and causes premature truncation of PLM axon growth ( ‘undershooting’ ) in development ( Chen et al . , 2011 ) ( Figure 2C , D , Figure 2—figure supplement 1A ) . We therefore tested if the capacity of EFA-6 protein fragments to relocalize correlated with their effects on PLM development and regrowth . We found that overexpression of the EFA-6 N terminal fragments that displayed injury-induced relocalization ( N150 , N100 , N70 ) all caused axons of PLM neurons to undershoot ( Figure 2C , Figure 2—figure supplement 1A; Table 1 ) . We performed axotomy on those axons that exhibited normal morphology in L4 larvae and found that these axons all showed significantly reduced regrowth ( Figure 2D , Figure 2—figure supplement 1B ) . In contrast , overexpression of smaller fragments ( N24 , N42 or N[43–70 aa] ) of the EFA-6 N-terminus that did not relocalize in response to axotomy did not significantly inhibit PLM outgrowth or regrowth ( Figure 2C , D and Table 1 ) . Interestingly , overexpression of EFA-6FLΔN150 , but not of EFA-6FL , caused PLM overshooting , suggesting the C-terminal region of EFA-6 may inhibit the activity of the N-terminus . Overexpression of EFA-6FL strongly inhibits axonal MT dynamics , as assessed using the EBP-GFP assay for growing MT plus ends ( Chen et al . , 2011 ) . Using live imaging and quantitative kymograph analysis of EBP-2::GFP comets , we found that only those EFA-6 fragments displaying injury-induced relocalization also inhibited steady-state axonal MT dynamics when overexpressed ( Figure 2E , Figure 2—figure supplement 1C ) . These observations suggest the ability of EFA-6 protein fragments to inhibit axon growth and MT dynamics closely correlates with their ability to relocalize in response to injury . 10 . 7554/eLife . 08695 . 011Table 1 . Localization and function of EFA-6 protein fragments in PLM neuronsDOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 011EFA-6 proteinsGFP fusion protein localizationInjury-induced re-localizationOverexpression effect on regrowthOverexpression effect on axon developmentFull length ( FL ) Cortical membraneyes51 . 2% ± 7 . 1% ***30% undershootingFLΔN150Cortical membraneno100 . 5% ± 6 . 6% ns22 . 5% overshootingN150Cytosolicyes28 . 5% ± 7 . 1% ***87 . 5% undershooting18 aaCytosolic + nuclearno65 . 3% ± 4 . 7% *10% mild undershootingN150Δ18aaCytosolic + nuclearno55 . 0% ± 4 . 7% ***5% mild undershootingN150 ( 33–38A ) Cytosolic + nuclearno57 . 6% ± 10 . 4% ***5% mild undershootingN150 ( 25–32A ) Cytosolic + nuclearno59% ± 5 . 7% ***7 . 5% mild undershootingN150 S33A , D35ACytosolic + nuclearno63% ± 7 . 8% **7 . 5% mild undershootingFLΔ18aaCortical membraneno77 . 9% ± 7 . 3% nswtN100Cytosolicyes26 . 6% ± 4 . 7% ***87 . 5% undershootingN70Cytosolicyes32 . 7% ± 4 . 5% ***85% undershootingN42Cytosolic + nuclearno66 . 2% ± 5 . 7% *12 . 5% undershootingN24Cytosolic + nuclearno80 . 7% ± 5 . 0% nswtMild and severe undershooting are defined as PLM termination anterior or posterior to the PVM soma respectively; ‘undershooting’ includes both mild and severe undershooting . See Figure 2—figure supplement 1 . All EFA6 family members contain extended N-termini of low sequence complexity and lack defined sequence motifs ( Figure 2—figure supplement 2A ) . The sole known region of sequence conservation among N-termini of some EFA6 family members is a motif of 18 aa ( residues 25–42 in C . elegans EFA-6 ) , necessary and partially sufficient for EFA-6's effects on cortical MT growth in the embryo ( O'Rourke et al . , 2010 ) . Analysis of EFA-6/EFA6 N-termini using algorithms that predict natively disordered protein regions ( Xue et al . , 2010 ) indicates that the EFA-6 N-terminus has a high probability of protein disorder , with the exception of the 18 aa motif ( Figure 2—figure supplement 2B ) . To address the role of the 18 aa motif in EFA-6's function in neurons , we expressed constructs in which the 18 aa motif was deleted or mutated . We found that full length EFA-6 lacking the motif , GFP::EFA-6FLΔ18 , localized to the plasma membrane but did not cause axonal developmental defects ( Figures 1G , 2C ) . GFP::EFA-6N150 was diffuse in the cytosol and excluded from the nucleus ( Figure 1D ) . Deletion of the 18 aa motif from the N-terminus ( N150Δ18 ) abolished this nuclear exclusion , largely resembling free GFP ( Figure 1H ) . Overexpression of EFA-6N150Δ18 also did not cause developmental abnormalities ( Figure 2C , Figure 2—figure supplement 1A ) . Pan-neural overexpression of EFA-6N150 , but not of EFA-6N150Δ18 , caused aberrant locomotion ( Figure 2—figure supplement 1D , E ) , suggesting that high levels of EFA-6 N-terminal domain perturb function of many neurons , in a manner dependent on the 18 aa motif . Moreover , mutating 2 or more residues within the 18 aa motif to alanines significantly reduced the activity of the N-terminus in multiple assays ( Table 1 ) , suggesting the sequence of the 18 aa motif is critical for the N-terminus to function . The 18 aa motif was essential for injury-triggered relocalization of EFA-6FL and EFA-6N150 ( Figure 1G , H ) , as well as for their inhibitory effects in axon regrowth ( Figure 2D ) . Expression of the 18 aa motif alone did not cause PLM developmental defects or confer injury-induced re-localization ( Figures 1I , 2C ) , but mildly inhibited axon regrowth ( Figure 2D ) , suggesting that injured axons are highly sensitive to the activity of this motif , but that the surrounding context is required for full activity of the N terminus . Two efa-6 loss of function alleles , tm3124 and ok3353 , delete genomic sequences that encode the Sec7 domain , and are predicted to cause frameshifts after the N-terminus ( Figure 2—figure supplement 2C ) . Both mutations cause embryonic phenotypes similar to efa-6 RNAi ( O'Rourke et al . , 2010 ) and display similarly enhanced axon regrowth ( Chen et al . , 2011 ) . As these mutations do not delete the N-terminus , it is possible that truncated proteins might be produced in these mutants . We therefore generated a targeted deletion , efa-6 ( ju1200 ) , that removes the genomic sequences encoding the 18 aa motif ( Figure 2—figure supplement 2C ) . The axon developmental and regrowth phenotypes of efa-6 ( ju1200 ) mutants were indistinguishable from those of efa-6 ( tm3124 ) ( Figure 2C , D ) . In addition , single copy transgene expression of EFA-6N150 ( juSi86 ) rescued the regeneration defects of both efa-6 ( tm3124 ) and efa-6 ( ju1200 ) to similar degrees ( Figure 2—figure supplement 2D ) , and also rescued developmental axon overgrowth ( not shown ) . Thus , the increased axon regrowth of efa-6 mutants reflects a complete loss of EFA-6 function , and the major activity of EFA-6 in axon growth resides in the N-terminus , dependent on the 18 aa motif . Below , we refer to efa-6 ( tm3124 ) as efa-6 ( 0 ) . To understand how EFA-6 inhibits axon regeneration , we next searched for EFA-6 interacting proteins using yeast two-hybrid screening . We identified two strong interactors , the MT-associated proteins ( MAPs ) TAC-1 and ZYG-8 . ZYG-8 is the C . elegans ortholog of mammalian DCLK , defined by an N-terminal doublecortin domain and a C-terminal kinase domain ( Gönczy et al . , 2001 ) . ZYG-8 is required for spindle positioning in embryos ( Gönczy et al . , 2001 ) , and for normal axonal MT architecture in post-mitotic neurons ( Bellanger et al . , 2012 ) . TAC-1 is the sole TACC protein in C . elegans ( Bellanger and Gönczy , 2003; Le Bot et al . , 2003; Srayko et al . , 2003 ) and can form a complex with ZYG-8 to regulate MT assembly in embryos ( Bellanger et al . , 2007 ) . In the yeast two-hybrid assay , we found that both TAC-1 and ZYG-8 interacted with EFA-6N150 , dependent on the 18 aa motif ( Figure 3A , Figure 3—figure supplement 1 ) . TAC-1 and ZYG-8 interacted by two-hybrid assay , and TAC-1 interacted strongly with ZYG-8ΔKD ( Figure 3—figure supplement 1 ) . To independently verify these interactions we transfected tagged proteins in HEK293 cells and performed co-immunoprecipitation . We found that TAC-1 co-immunoprecipitated with EFA-6N150 , but not with EFA-6N150Δ18 ( Figure 3B ) . Likewise , ZYG-8 and EFA-6N150 could be co-immunoprecipitated when coexpressed ( Figure 3C ) . These studies suggest that TAC-1 and ZYG-8 specifically interact with the EFA-6 N-terminus . We further tested binding of EFA-6 to TAC-1 and ZYG-8 in cells co-transfected with EFA-6N150 , TAC-1 and ZYG-8ΔKD . After immunoprecipitation of EFA-6N150 we could detect both TAC-1 and ZYG-8 , and the interactions between EFA-6 and TAC-1 ( or ZYG-8 ) were not affected by the presence of ZYG-8 ( or TAC-1 ) ( Figure 3D ) . This result suggests that EFA-6 , TAC-1 , and ZYG-8 might exist in the same ternary complex . 10 . 7554/eLife . 08695 . 012Figure 3 . EFA-6 interacts with the MT-associated proteins ( MAPs ) TAC-1 and ZYG-8 . ( A ) Summary of two-hybrid analyses . The N-terminus of EFA-6 ( N150 ) is necessary and sufficient for its interaction with TAC-1 and ZYG-8 . Deletion of the 18-aa motif from the N-terminus severely impairs binding to TAC-1 and ZYG-8 . The interaction between EFA-6 and ZYG-8 does not require the ZYG-8 kinase domain . EFA-6 did not interact with MEC-7/β-tubulin in the two-hybrid assay . ‘+++’ , ‘+’ , and ‘−’ indicate strong , weak , or undetectable interaction , respectively . ( B–D ) Co-immunoprecipitation ( Co-IP ) of EFA-6 and interactors in HEK293 cells . Indicated constructs were co-transfected into HEK293 cells at a 1:1 ratio . M2-FLAG conjugated magnetic beads were used for IP , and rabbit anti FLAG or anti HA antibodies used for western blotting ( WB ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 01210 . 7554/eLife . 08695 . 013Figure 3—figure supplement 1 . Two-hybrid analysis of the interactions between EFA-6 and TAC-1 or ZYG-8 . Pairs of constructs encoding an activation domain ( AD ) fusion protein and a DNA binding domain ( DBD ) fusion protein , as indicated , were co-transformed into yeast strain L40 . See Supplementary file 1B for details of plasmids . Transformed yeasts were grown on agar plates with SD medium ( synthetic minimal medium ) lacking leucine and tryptophan . Interactions were examined on agar plates with SD medium lacking leucine , tryptophan , and histidine ( KUWLH ) , with or without 1 mM 3-AT . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 013 TAC-1 and ZYG-8 are essential for embryonic cell division , and tac-1 and zyg-8 null mutants are maternal-effect embryonic lethal ( Gönczy et al . , 1999 ) . To examine the roles of these genes in axon regeneration we first used conditional ( temperature sensitive , ts ) alleles ( Gönczy et al . , 2001; Bellanger et al . , 2007 ) , here denoted lf . When shifted from permissive ( 15°C ) to restrictive ( 25°C ) temperature at L1 stage , zyg-8 ( lf ) mutants displayed normal touch axon morphology ( not shown ) . After axotomy in the L4 stage , these animals showed strongly reduced axon regrowth in both PLM and ALM neurons , compared to controls subjected to identical temperature shifts ( Figure 4A , Figure 4—figure supplement 1A , B ) . zyg-8 ( lf ) mutants also showed reduced axon regrowth when maintained at 15°C , even though PLM axon development was normal ( Figure 4—figure supplement 1C , D ) , suggesting that axon regrowth is highly sensitive to reduction of zyg-8 function . The axon regrowth defect in zyg-8 ( lf ) mutants was fully rescued by a single copy transgene expressing ZYG-8 driven by a touch neuron specific promoter ( Figure 4—figure supplement 1C , D ) , indicating a cell-autonomous function . Similarly , tac-1 ( lf ) mutants displayed reduced PLM axon regrowth when shifted from 15 to 25°C in the L1 stage , 20 hr before axotomy ( Figure 4A , B ) ; tac-1 ( lf ) animals raised at 15°C displayed normal regrowth ( not shown ) . As the extent to which these ts mutations impair gene function in post-mitotic cells is not known , we further examined PLM regrowth in tac-1 ( ok3305 ) null mutants ( here denoted tac-1 ( 0 ) ) using a genetic mosaic strategy ( Figure 4C ) . We rescued the maternal-effect embryonic lethality of tac-1 ( 0 ) mutants with a single copy insertion transgene expressing a floxed version of tac-1 ( + ) ( juSi148 or juSi162 ) ( see ‘Materials and methods’ ) . We then deleted tac-1 ( + ) specifically in touch neurons by expressing Cre recombinase under the control of the mec-7 promoter . Cre-mediated deletion of tac-1 ( + ) occurred in 8/8 transgenic ( juSi162; Pmec-7-Cre ) animals ( Figure 4—figure supplement 1E ) , suggesting Cre-mediated recombination was efficient . In animals with touch neuron specific deletion of tac-1 , the PLM axon developed normally , but axon regrowth was impaired to a degree similar to tac-1 ( lf ) mutants after L1 upshift ( Figure 4B ) , indicating that TAC-1 functions cell autonomously in PLM axon regrowth . 10 . 7554/eLife . 08695 . 014Figure 4 . TAC-1 and ZYG-8 promote axon regrowth downstream of EFA-6 . ( A ) Normalized PLM axon regrowth at 24 hr . Strains were maintained at 15°C , shifted to 25°C 20 hr before axotomy , and kept at 25°C after axotomy for all experiments with ts ( lf ) alleles . ( B ) Normalized PLM axon regrowth at 24 hr post axotomy . Loss of TAC-1 impairs axon regrowth in a cell-autonomous manner . ( C ) Strategy for neuron-specific deletion of tac-1 mutants with Mos-SCI single copy transgene of floxed tac-1 . ( D ) Representative images of axon regrowth at 6 hr post axotomy . WT regrowing axons usually displayed a regenerative growth cone ( arrow ) at 6 hr post-axotomy whereas zyg-8 ( lf ) and tac-1 ( lf ) mutant axons rarely display growth cones . ( E ) Quantitation of initial axon regrowth at 6 hr . ( F ) Percentage of axons with regenerative growth cones 6 hr post axotomy . Statistics , one-way ANOVA with Bonferroni post test; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05; ns , not significant . n ≥ 10 . Scale , 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 01410 . 7554/eLife . 08695 . 015Figure 4—figure supplement 1 . EFA-6 and its interactors regulate axon regeneration . ( A ) zyg-8 ( or484ts ) displays reduced ALM regrowth . Normalized ALM axon regrowth , 24 hr post axotomy . Strains were maintained at 15°C , shifted to 25°C 20 hr before axotomy , and kept at 25°C after axotomy . Statistics: Student's t-test . ***p < 0 . 001 . ( B ) Representative images for panel ( A ) ; scale , 10 μm . ( C ) PLM axon regrowth defect of zyg-8 ( or484 ) at permissive temperature ( 15°C ) and rescue by single copy insertion transgene Pmec-4-ZYG-8 ( juSi193 ) . Statistics: one-way ANOVA with Bonferroni post test; ***p < 0 . 001 . ( D ) Representative images for panel ( C ) ; Scale , 25 μm . ( E ) PCR assay for efficiency of the Cre-lox recombination in touch neurons . 8 animals of strain CZ20478 [tac-1 ( ok3305 ) ; juSi162 ( lox-tac-1-lox Mos-SCI ) V; juEx6042 ( Pmec-7-nCre;ttx-3-rfp ) ] were tested for the excision product ( red arrow on the gel image ) . Excision of the lox-flanked fragment occurred in 8/8 animals . No excision was seen in juSi162 animals lacking the Cre transgene . Individual animals expressing transgenic marker Pttx-3-RFP from the strain CZ20478 [tac-1 ( ok3305 ) ; juSi162 ( lox-tac-1-lox Mos-SCI ) ; juEx6042 ( Pmec-7-nCre;Pttx-3-RFP ) ] were genotyped for the Cre-dependent deletion of tac-1 single copy insertion ( juSi162 ) . Animals not expressing Pttx-3-RFP , that is , tac-1 ( ok3305 ) ; juSi162 ( lox-tac-1-lox Mos-SCI ) , were used as negative control . The following primers ( specific to juSi162 and not endogenous tac-1 ) were used in PCR: ttTi5605 homology arm: ACGCCCAGGAGAACACGTTAG ( left black arrow ) tac-1-5′ UTR: AGATCCACCCTCACCATCAC ( middle black arrow ) unc-119: TTCGCTGTCCTGTCACACTCG ( red arrow ) . Without Cre-dependent deletion , this PCR will produce 3472 and 860 bp fragments . After deletion , the PCR product will be 642 bp . tac-1 ( ok3305 ) is a deletion of 812 bp with an insertion of 23 bp of random sequence at the deletion . We designed primers in the 23 bp insertion to detect ok3305 , so that the tac-1 ( + ) transgene will not interfere with genotyping for ok3305 . ( F ) Overexpressing TAC-1 or ZYG-8 does not enhance axon regrowth of efa-6 ( lf ) , consistent with function in the same pathway . Normalized PLM axon regrowth of strains with indicated genotypes . Statistics: one-way ANOVA with Bonferroni post test; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 015 We next tested when ZYG-8 and TAC-1 are required in axon regeneration . We performed temperature upshifts 20 hr before axotomy and examined axon regrowth at 6 hr post axotomy ( hpa ) . zyg-8 ( lf ) and tac-1 ( lf ) mutants both displayed significantly reduced regrowth at 6 hpa , and significantly fewer regenerative growth cones ( defined as axonal tips containing filopodia and/or lamellipodia ) , compared to controls ( Figure 4D–F ) . tac-1 ( lf ) zyg-8 ( lf ) double mutants did not show further reduction in axon regrowth , compared to single mutants ( Figure 4A ) , suggesting TAC-1 and ZYG-8 could act in a common pathway . zyg-8 ( lf ) efa-6 ( 0 ) double mutants resembled zyg-8 ( lf ) single mutants , and tac-1 ( lf ) efa-6 ( 0 ) double mutants resembled tac-1 ( lf ) in axon regrowth ( Figure 4A ) , consistent with zyg-8 and tac-1 acting downstream of efa-6 . Conversely , overexpression of TAC-1 was sufficient to enhance PLM regrowth , and did not further enhance the regrowth of efa-6 ( 0 ) mutants ( Figure 4—figure supplement 1F ) . Our previous analysis showed that loss of EFA-6 function resulted in elevated axonal MT dynamics several hours after injury ( Chen et al . , 2011 ) . To test whether MT dynamics might be influenced by EFA-6 immediately after injury , we examined the acute effects of axon injury on MT dynamics . In wild type animals , axonal EBP-2::GFP comets were dramatically reduced within 50 s post axotomy , consistent with injury triggering rapid MT destabilization ( Figure 5A ) . In efa-6 ( tm3124 ) or efa-6 ( ju1200 ) mutants we observed slightly increased numbers of EBP-2::GFP comets in uninjured axons , compared to wild type; but these animals did not show a significant reduction in EBP-2::GFP comets immediately after injury ( Figure 5A and Figure 5—figure supplement 1A ) . The slight reduction in axonal comets after injury in efa-6 ( 0 ) suggests that some MTs can also be downregulated independent of EFA-6 . In efa-6 ( gf ) animals overexpressing EFA-6N150 , the total number of growing MT plus ends in uninjured axons was reduced compared to wild type , and was not further reduced after injury ( Figure 5A ) . tac-1 ( lf ) mutants had slightly reduced axonal MT dynamics in uninjured axons , and displayed injury-dependent downregulation , whereas zyg-8 ( lf ) mutants showed fewer dynamic MTs in uninjured axons , and did not further downregulate MTs after injury ( Figure 5A ) . The mild phenotype in tac-1 ( lf ) could be due to incomplete loss of function of this ts allele . We further tested the tac-1 ( ok3305 ) deletion allele using Cre-induced tissue-specific knockout ( ‘Materials and methods’ ) . Compared to control , tac-1 ( 0 ) touch neurons displayed reduced dynamics in uninjured axons similar to zyg-8 ( lf ) , and showed no further reduction after injury ( Figure 5—figure supplement 1B ) . As reported previously , dynamic axonal MTs are significantly increased at 3 hr post axotomy , and this is further enhanced in efa-6 ( 0 ) mutants ( Chen et al . , 2011 ) ( Figure 5B ) . Neither tac-1 ( lf ) nor zyg-8 ( lf ) mutants upregulated dynamic axonal MTs by 3 hr post injury ( Figure 5B ) . Moreover , MT dynamics in efa-6 ( 0 ) tac-1 ( lf ) or efa-6 ( 0 ) zyg-8 ( lf ) double mutants resembled tac-1 ( lf ) or zyg-8 ( lf ) single mutants ( Figure 5B ) , consistent with EFA-6 functioning upstream of ZYG-8 . Thus , axon injury causes an immediate inhibition in growing MTs , dependent on EFA-6 and correlating with its relocalization , followed by a more prolonged increase in growing MTs , dependent on the function of TAC-1 and ZYG-8 . 10 . 7554/eLife . 08695 . 016Figure 5 . Injury triggers rapid down-regulation of MT dynamics dependent on EFA-6 . ( A ) MT dynamics ( EBP-2::GFP ) before and immediately after injury . Kymographs were created from videos of 400 frames ( 0 . 23 s/frame ) , 200 frames before and 200 frames after axotomy . Lower panel: Quantitation of EBP-2::GFP tracks in proximal axon before and immediately after injury . ( B ) MT dynamics 3 hr post injury . Kymographs were created from videos of 200 frames ( 0 . 23 s/frame ) . Quantitation of EBP-2::GFP tracks in 40 μm of the proximal axon for 47 s 3 hr post injury . Red line represents time of axotomy; arrow indicates injury site . Strains were maintained at 15°C , shifted to 25°C 20 hr before axotomy . Alleles: efa-6 ( tm3124 ) , tac-1 ( or455ts ) , zyg-8 ( or484ts ) . Statistics , one-way ANOVA with Bonferroni post test; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05; ns , not significant . n ≥ 10 axons per condition . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 01610 . 7554/eLife . 08695 . 017Figure 5—figure supplement 1 . MT dynamics triggered by injury . ( A ) MT dynamics ( EBP-2::GFP ) before and immediately after injury in wild type and efa-6 ( ju1200 ) mutant . A significant reduction in MT growth ( defined as number of EBP-2::GFP tracks per kymograph ) was seen in wild type animals , but not in efa-6 ( ju1200 ) , similar to tm3124 . ( B ) MT dynamics before and after injury in the tissue-specific tac-1 deletion mutant and in control strains . Statistics: one-way ANOVA with Bonferroni post test; ***p < 0 . 001; **p < 0 . 01; *p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 017 We next assessed axonal localization of TAC-1 and ZYG-8 in the steady state and after injury . In wild type uninjured touch neurons , GFP::ZYG-8 was diffuse throughout axon ( Figure 6A ) ; GFP::TAC-1 or TAC-1::GFP localized to one or two perinuclear spots in the soma and was also diffusely distributed along the axon ( Figure 6B , Figure 6—figure supplement 1A ) . These patterns were not altered in efa-6 ( 0 ) ( Figure 6A , Figure 6—figure supplement 1A ) . Conversely , steady state localization of GFP::EFA-6FL or GFP::EFA-6N150 was normal in tac-1 ( 0 ) or zyg-8 ( 0 ) ( Figure 6C ) . Thus , in uninjured axons , EFA-6 , TAC-1 , and ZYG-8 appear to localize independently . 10 . 7554/eLife . 08695 . 018Figure 6 . TAC-1 relocalizes in response to injury to become co-localized with EFA-6 . ( A ) Localization of GFP::ZYG-8 in touch neurons before and 2 min after axotomy in wild type and efa-6 ( tm3124 ) . GFP::ZYG-8 localization is not affected by axon injury or loss of EFA-6 . Transgene: Pmec-4-GFP::ZYG-8 ( juEx5932 ) . ( B ) GFP::TAC-1 in PLM before and 2 min after axotomy at wild type and efa-6 ( ju1200 ) backgrounds . Injury triggered relocalization of TAC-1 was similar to EFA-6 and not dependent on EFA-6 . Pmec-4-GFP::TAC-1 ( juEx5759 ) . ( C ) GFP::EFA-6N150 ( juEx3531 ) localization in wild-type , tac-1 ( ok3305 ) and a putative zyg-8 null allele zyg-8 ( t1518 ) ( Gönczy et al . , 2001 ) . Relocalization of EFA-6N150 was not dependent on TAC-1 or ZYG-8 . ( D ) Localization of EFA-6FL::GFP and mKate2::TAC-1 before and after axotomy in a touch neuron . Before axotomy , TAC-1 was diffuse in soma and along the axon , and concentrated in a large perinuclear dot . EFA-6 was predominantly localized to the plasma membrane and also in the perinuclear dot marked by TAC-1 . After axotomy , both proteins became punctate and the puncta were partially co-localized; enlargements in small boxes below . Graphs of line scans along the axon are shown below the enlarged images . Arrow , injury site; scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 01810 . 7554/eLife . 08695 . 019Figure 6—figure supplement 1 . Injury triggered relocalization . ( A ) Localization of TAC-1::GFP before and 2 min after injury; TAC-1 with a C-terminal GFP tag relocalized similarly to N-terminally tagged TAC-1 . Localization of TAC-1::GFP in efa-6 ( ju1200 ) is similar to wild type background . Transgene: Pmec-4-TAC-1::GFP ( juEx6362 ) . ( B ) Co-localization of TAC-1::GFP and EFA-6N150::mKate2 in puncta after axotomy . Before axotomy , TAC-1 and EFA-6N150 were diffuse along the axon , and also localized in a perinuclear dot in the soma . After axotomy , both proteins were punctate and the puncta co-localized with each other . As the axonal signal was relatively dim , only soma images are shown . Line scan in soma below images . Arrow: injury site; scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 019 We asked whether TAC-1 and ZYG-8 also responded dynamically to injury . Whereas GFP::ZYG-8 remained diffuse after injury ( Figure 6A ) , TAC-1::GFP or GFP::TAC-1 relocalized rapidly to puncta along the axon and in the soma ( Figure 6B , Figure 6—figure supplement 1A ) . We next co-expressed EFA-6FL::GFP and mKate2::TAC-1 in touch neurons , and found that while EFA-6FL::GFP localized to the plasma membrane throughout the cell , the large perinuclear TAC-1 spots also recruited EFA-6 ( Figure 6D ) , suggesting the two proteins can interact in neurons . Localization of EFA-6 to large perinuclear spots was not observed when EFA-6FL::GFP was expressed alone ( Figure 1A ) . After injury axonal EFA-6FL::GFP and mKate2::TAC-1 relocalized to largely overlapping puncta ( Figure 6D ) ; co-expressed EFA-6N150::mKate2 and TAC-1::GFP displayed similar co-localization ( Figure 6—figure supplement 1B ) . The injury-induced re-localization of GFP::TAC-1 occurred normally in efa-6 ( ju1200 ) ( Figure 6B ) , as did re-localization of GFP::EFA-6N150 in tac-1 ( 0 ) or zyg-8 ( 0 ) ( Figure 6C ) , suggesting that these proteins relocalize independently of each other . TAC-1 , like other TACC family members , is thought to directly interact with MTs . We considered the possibility that , after injury , TAC-1 and EFA-6 relocalized to axonal MTs . As the punctate localization of TAC-1 and EFA-6 does not resemble that of EBP-2::GFP ( i . e . , growing MT plus ends ) , we tested whether TAC-1 and EFA-6 were becoming localized to MT minus ends . PTRN-1 is the C . elegans member of the Patronin/CAMSAP family , known to bind to and stabilize MT minus ends ( Meng et al . , 2008; Goodwin and Vale , 2010; Yau et al . , 2014 ) . In C . elegans neurons PTRN-1 localizes to axonal puncta , likely the sites of MT minus ends ( Chuang et al . , 2014; Marcette et al . , 2014; Richardson et al . , 2014 ) . GFP::PTRN-1 localization does not dramatically change after axon injury ( Chuang et al . , 2014 ) , while in the same time period EFA-6N150::mKate2 became punctate and partially colocalized with GFP::PTRN-1 , independent of the tagged reporters ( Figure 7A , Figure 7—figure supplement 1B ) . Similarly , after injury TAC-1::mKate2 became highly colocalized with GFP::PTRN-1 ( Figure 7—figure supplement 2 ) . These observations suggest that axon injury causes TAC-1 to relocalize to PTRN-1-containing puncta , and causes EFA-6 to relocalize to regions overlapping with or closely adjacent to the TAC-1/PTRN-1 puncta . 10 . 7554/eLife . 08695 . 020Figure 7 . EFA-6 and TAC-1 re-localize to puncta overlapping with the MT minus end-binding protein Patronin/PTRN-1 . ( A ) Localization of PTRN-1 and EFA-6N150 in PLM before and after axotomy . EFA-6N150::mKate2 was diffuse in soma and axon before injury , and became punctate after injury and these puncta co-localized to GFP::PTRN-1 . Enlarged images of the regions in boxes are shown below . Graphs of line scans along the axon and F ( in ) /F ( out ) ratio quantitation are shown below . Increased F ( in ) /F ( out ) ratio indicates higher degree of colocalization post axotomy; see Figure 7—figure supplement 1A and ‘Materials and methods’ for calculation of F ( in ) /F ( out ) . Statistics: Student's t-test . ***p < 0 . 001 . ( B ) Epistatic interactions between efa-6 ( 0 ) , ptrn-1 ( 0 ) , and tac-1 ( lf ) . Normalized PLM regrowth . Strains without or with temperature shift ( cultured at 15°C and upshifted to 25°C 20 hr before axotomy and kept at 25°C for 24 hr after axotomy ) were quantified separately . Statistics: one-way ANOVA with Bonferroni post test . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 02010 . 7554/eLife . 08695 . 021Figure 7—figure supplement 1 . Co-localization of EFA-6 with PTRN-1 after injury . ( A ) Cartoons illustrating quantitation of co-localization [F ( in ) / ( out ) ratio]; see ‘Materials and methods’ for details of the colocalization ratio calculation . ( B ) Localization of tagRFP::PTRN-1 and GFP::EFA-6N150 in PLM , before and after axotomy . GFP::EFA-6N150 became punctate after injury and co-localized with tagRFP::PTRN-1 puncta; 3 representative puncta are marked . Line scan and F ( in ) /F ( out ) ratio quantitation shown below the images . Statistics: Student's t-test . ***p < 0 . 001 . ( C ) Localization of GFP::EFA-6N150 in ptrn-1 ( lt1 ) , before and after axon injury . Localization of GFP::EFA-6N150 was not dependent on PTRN-1 . ptrn-1 ( lt1 ) is a MosDel-induced deletion that removes the entire ptrn-1 coding sequence ( Chuang et al . , 2014 ) . ( D ) Localization of GFP::EFA-6N150 in tac-1 ( lf ) ptrn-1 ( 0 ) double mutant is indistinguishable from wild type . ( E ) Localization of TAC-1::GFP in ptrn-1 ( 0 ) before and after axon injury . Localization of TAC-1::GFP was not dependent on PTRN-1 . Arrow: injury site; scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 02110 . 7554/eLife . 08695 . 022Figure 7—figure supplement 2 . TAC-1 relocalizes to PTRN-1 puncta after injury . Localization of GFP::PTRN-1 and mKate2::TAC-1 in distal PLM before and after axotomy; enlargements and graphs as in panel A . mKate2::TAC-1 was diffuse in the axon before injury , and became punctate after injury , colocalizing with GFP::PTRN-1 ( 3 representative puncta marked with lines ) . Arrow , injury site; Scale , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 08695 . 022 We further asked whether injury-induced relocalization of EFA-6 or TAC-1 required PTRN-1 . In a ptrn-1 null mutant GFP::EFA-6N150 or TAC-1::GFP relocalized after injury as in wild type ( Figure 7—figure supplement 1C , E ) . GFP::EFA-6N150 localization in the tac-1 ( 0 ) ptrn-1 ( 0 ) double mutant was normal either before or after injury ( Figure 7—figure supplement 1D ) . Thus , although TAC-1 and EFA-6 relocalize to PTRN-1-containing puncta and adjacent regions respectively , their recruitment does not absolutely require PTRN-1 . The relocalization of EFA-6 and TAC-1 may involve multiple redundant factors . ptrn-1 null mutants display largely normal PLM outgrowth and significantly reduced axon regeneration ( Chuang et al . , 2014 ) . We found that efa-6 ( 0 ) ptrn-1 ( 0 ) double mutants resembled ptrn-1 ( 0 ) single mutants in regeneration ( Figure 7B ) ; conversely , tac-1 ( lf ) ptrn-1 ( 0 ) double mutants were not further enhanced , compared to either single mutant ( Figure 7B ) . These double mutant analyses are consistent with PTRN-1 and TAC-1 acting in a common pathway in regeneration , with EFA-6 acting as a negative regulator of one or both proteins .
EFA-6 induces catastrophe or pausing of growth at MT plus ends at the cortex of embryonic cells ( O'Rourke et al . , 2010 ) , and our analysis is consistent with this model in steady-state ( uninjured ) axons . In mature neurons EFA-6 localizes to the cell membrane via its PH domain . efa-6 ( ( 0 ) ) mutant axons display elevated numbers of growing MTs in the steady state , and display impenetrant developmental overgrowth , indicating that in the absence of injury EFA-6 restrains axonal MT dynamics and mildly inhibits axon outgrowth . These steady-state roles of EFA-6 are mediated by the N-terminus , as they are fully rescued by expression of the N-terminal 150 aa . Overexpression of either full length EFA-6 or the N-terminus causes axons to undershoot , whereas overexpression of EFA-6ΔN causes axons to overshoot . The opposing effects of overexpression of the N-terminus and the C-terminus suggest that in the steady state the EFA-6 N-terminus might be inhibited by the remainder of the protein . After axon injury , EFA-6 displays a dramatic and specific transient relocalization to punctate structures associated with MT minus ends . Interestingly , this relocalization does not require membrane association , as EFA-6 fragments lacking the PH domain , or containing only the N-terminal 1–70 aa , were not membrane-localized , yet became relocalized to axonal puncta after injury . As relocalized full-length EFA-6 appears to remain membrane associated , it is possible that after injury EFA-6 localizes to the cytoskeleton via its N terminus while remaining membrane associated via its PH domain . Speculatively , injury signals may increase EFA-6 N-terminus activity by releasing it from inhibition by the EFA-6 C-terminus . The N-termini of EFA6 family members display low sequence complexity and minimal primary sequence similarity , with the exception of the 18 aa motif found in invertebrate family members ( O'Rourke et al . , 2010 ) . The N-termini of C . elegans , Drosophila and mammalian EFA6 proteins all have a high probability of intrinsic disorder . Intrinsically disordered proteins ( IDPs ) and disordered protein regions are increasingly recognized as having important biological activities ( Oldfield and Dunker , 2014 ) . Well-studied examples of IDPs in the nervous system include the MT-binding proteins Tau ( Schweers et al . , 1994 ) and stathmin ( Honnappa et al . , 2006 ) . Intrinsic disorder does not imply a lack of structure , but rather allows these regions to function as binding surfaces with multiple interacting partners; IDPs are often hubs in protein–protein interaction networks ( Cumberworth et al . , 2013 ) . In the EFA-6 N-terminus the 18 aa motif is predicted to have relative structural order , and might act as a molecular recognition feature as in other IDPs . The 18 aa motif is critical for EFA-6 relocalization and for the EFA-6 N-terminus to interact with ZYG-8 and TAC-1 . Moreover , as mutation of two or more residues within the 18 aa motif impairs its activity , both the exact sequence of the 18 aa motif and the surrounding extended intrinsically disordered domain appear to be critical for the ability of the EFA-6 N-terminus to regulate MT dynamics . In some assays , the requirement for the 18 aa motif in EFA-6 function was not all-or-none , suggesting it may facilitate the function of the surrounding interacting domain . Moreover , neither ZYG-8 nor TAC-1 may directly bind the 18 aa motif; the motif may be important for correct folding of the larger N-terminal interacting domain . Our analysis suggests the EFA-6 N-terminus regulates MT dynamics indirectly via the MAPs ZYG-8 and TAC-1 . Identification of TAC-1 and ZYG-8 as EFA-6 interactors was unexpected , as in embryonic cells TAC-1 is predominantly centrosomal or cytoplasmic ( Le Bot et al . , 2003 ) , and ZYG-8 localizes along MTs ( Bellanger et al . , 2007 ) , whereas EFA-6 is cortically localized ( O'Rourke et al . , 2010 ) . TAC-1 and ZYG-8 interact physically , but localize independently ( Bellanger et al . , 2007 ) . However EFA-6 , TAC-1 , and ZYG-8 are all present in axons , suggesting these proteins may interact directly in differentiated cells . TAC-1 and EFA-6 partly colocalize after injury , consistent with their interaction being regulated by injury . Although ZYG-8 localization appeared unaffected by injury , it is predicted to associate along the length of MTs , and so could also interact with EFA-6 after injury . Notably , loss of function of TAC-1 or ZYG-8 did not detectably affect developmental axon outgrowth , but strongly blocked axon regeneration , indicating regenerative regrowth is highly dependent on these MAPs . Neither TACCs nor ZYG-8/DCLK exclusively associate with MT plus ends . TACC proteins can act both at plus and minus ends of centrosomal MTs ( Lee et al . , 2001; Peset and Vernos , 2008 ) . Localization of human TACC3 to minus ends is regulated by Aurora A dependent phosphorylation ( Barros et al . , 2005; LeRoy et al . , 2007 ) . Conversely , Doublecortin ( DCX ) decorates the length of the MT lattice and stabilizes it ( Moores et al . , 2006 ) . DCX can also track MT plus ends , and acts as a nucleation factor , stabilizing polymerization intermediates ( Moores et al . , 2004; Bechstedt and Brouhard , 2012 ) . Like DCX , DCLK is thought to be able to interact with MTs along their length ( Burgess and Reiner , 2000 ) . Our results suggest that after axon injury , TAC-1 , EFA-6 , and possibly ZYG-8 may interact with one another at specific subregions of MTs at or adjacent to minus ends . As EFA-6 has opposite effects to TAC-1 and ZYG-8 , EFA-6 may transiently inhibit the activity of TAC-1 or ZYG-8 , resulting in the rapid disruption of axonal MT growth . Following this initial phase , EFA-6 returns to its steady state , relieving the inhibition of TAC-1 and ZYG-8 , which are then required for the later upregulation of axonal MT dynamics . CAMSAPs/Patronins directly bind MT minus ends and protect them from depolymerization by Kinesin-13 ( Tournebize et al . , 2000; Meng et al . , 2008; Goodwin and Vale , 2010; Hendershott and Vale , 2014; Jiang et al . , 2014 ) . The minus end protection activity of CAMSAP family proteins is important for their function in maintaining noncentrosomal MTs . In C . elegans axons PTRN-1 localizes to puncta that are thought to define sites of MT minus end anchoring or stabilization ( Marcette et al . , 2014; Richardson et al . , 2014 ) . Strikingly , EFA-6 and TAC-1 relocalize close to or at these sites after injury , suggesting MT minus ends may be an important site of regulation . As PTRN-1 itself is not required for EFA-6 or TAC-1 relocalization , other proteins may be involved in their targeting . Indeed , the near-normal development and behavior of ptrn-1 null mutants suggests additional factors can stabilize MT minus ends in noncentrosomal arrays . Like TAC-1 and ZYG-8 , PTRN-1 is required for axon regrowth . EFA-6 or its interactors might modulate the function of PTRN-1 in axon regrowth . MT dynamic instability , first studied as a function of tubulin concentration in vitro , is influenced by a wide array of positive and negative regulators in vivo . For example , in Xenopus egg extracts , MT dynamics are determined by a balance between the MT growth-promoting XMAP215 and MT-destabilizing XKCMI ( Tournebize et al . , 2000; Kinoshita et al . , 2001 ) . Our findings suggest that the initial stages in axon regeneration are also driven by a sequence of shifts in the balance between opposing activities of MT destabilizers such as EFA-6 and MT growth promoting factors such as TAC-1 and ZYG-8 . Axonal injury in C . elegans triggers a highly regulated sequence of changes in MT dynamics that correlate closely with changes in EFA-6 localization and activity . We did not detect significant down-regulation of MT growth in response to injury in the absence of EFA-6 , using two independent alleles . Although a small decrease in MT dynamics was observed in efa-6 mutants , this was not statistically significant , and calculations of statistical power suggest that changes of >15% should be detectable in experiments of the sample size used here . Nevertheless there may be EFA-6-dependent and EFA-6-independent effects on MT dynamics immediately after injury . Changes in MT dynamics within seconds of axon injury have been studied in Aplysia neurons ( Erez et al . , 2007; Erez and Spira , 2008 ) , which display rapid local MT depolymerization followed by repolymerization over several minutes . Drosophila neurons also display acute and chronic alterations in MT dynamics after injury ( Chen et al . , 2012; Lu et al . , 2015 ) . Recent in vivo imaging of mammalian axons found an acute increase in axonal MT dynamics after laser axotomy , followed by a sustained increase over several days ( Kleele et al . , 2014 ) . Thus , the exact sequence of MT dynamics changes after injury may vary between cell types and organisms . An important future goal will be to address the role of EFA6 family members in mammalian axon regeneration , and whether manipulation of this MT regulatory pathway can enhance regeneration in therapeutic settings .
We maintained C . elegans following standard methods . Transgenes were introduced into mutant backgrounds by crossing or injection; homozygosity for all mutations was confirmed by PCR or sequencing . We used the following published transgenes: Pmec-7-GFP ( muIs32 ) , Pmec-4-GFP ( zdIs5 ) , Pmec-4-EBP-2::GFP ( juIs338 ) ( Chuang et al . , 2014 ) . We made new plasmids by Gateway recombination ( Invitrogen / Life Technologies , Grand Island , NY ) or Gibson assembly , as listed in Supplementary file 1A; new transgenes are listed in Supplementary file 1C . We amplified cDNAs from existing clones or from total first-strand cDNA; all clones were sequenced . Mutations were introduced by Quikchange mutagenesis ( Agilent Technologies , Santa Clara , CA ) . We followed standard procedures to generate multicopy extrachromosomal transgenes; plasmids were injected at 1–30 ng/μl , and co-injection markers at 75 ng/μl . We analyzed 3 to 5 lines per construct . We made single copy insertions using Mos-SCI ( http://www . wormbuilder . org/ ) , on chromosomes IV ( strain EG8081 ) or V ( EG8083 ) . We collected fluorescence images on Zeiss LSM710 or LSM510 confocal microscopes . We performed laser axotomy as described ( Chen et al . , 2011 ) . We performed live imaging and analysis of EBP-2::GFP dynamics as described ( Ghosh-Roy et al . , 2012 ) ; in some experiments we immobilized animals in 30 mM muscimol on pads of 10% agarose in M9 . For quantitative analysis of protein localization and colocalization , animals were immobilized with either 0 . 7% phenoxypropanol or 30 mM muscimol . Confocal z-stacks were collected with 0 . 5 μm intervals . Typically 3–4 slices span an axon ( 1–2 μm diameter ) . Images were analyzed using Zeiss Zen and Metamorph ( Molecular Devices , Sunnyvale , CA ) . Briefly , we drew lines along the axon , starting at the injury site or soma , then used the line scan tool to measure the average fluorescence intensity of 8 pixels ( ∼0 . 7 μm ) surrounding the lines . For puncta number ( or peak # ) analysis , we counted any peak with intensity greater than the mean +1 SD as a punctum . To quantitate colocalization , we measured the F ( in ) /F ( out ) ratio ( see Figure 7—figure supplement 1A ) using Metamorph software . A small ROI ( region of interest ) was drawn to cover one punctum of GFP::PTRN-1 . Average intensity of mKate2 ( EFA-6N150 ) within the ROI was measured as F ( in ) . The ROI was then duplicated to cover a small region in the axon with no GFP::PTRN-1 puncta , and average intensity of mKate2 ( EFA-6N150 ) within this ROI was measured as F ( out ) . F ( in ) /F ( out ) was then calculated as [F ( in ) − background intensity]/[F ( out ) − background intensity] . Before injury , EFA-6N::mKate2 is evenly distributed in the axon , and mKate2 intensity inside and outside of the GFP::PTRN-1 puncta is similar in level , so the F ( in ) /F ( out ) ratio is close to 1 . 2 min post injury , EFA-6N is relocalized to PTRN-1 puncta , so mKate2 intensity within the PTRN-1 puncta is much higher than outside the puncta , resulting in a significantly higher F ( in ) /F ( out ) ratio . For FRAP we set circular regions of interest ( ROIs ) for acquisition and photobleaching , using 2% laser power for acquisition and 100% laser power ( 488 nm , Zeiss LSM710 ) for photobleaching . We acquired 5 and 25 images before and after photobleaching . We chose ROIs 1 μm diameter from regions of median initial intensity in the soma or axon . In videos of uninjured neurons , we placed ROIs in regions with diffuse or relatively enriched GFP signal; in injured neurons , ROIs were drawn around puncta . Average fluorescence intensity in each ROI at each frame was measured in Zen . To generate FRAP curves we normalized intensity to the average of the five frames prior to photobleaching . We calculated t1/2 following standard formulas; the immobile fraction was calculated by the Zen program . We performed two-hybrid screening as described ( Wang et al . , 2013 ) . We cloned EFA-6 full-length cDNA and fragments into a pMB27-Gal4-BD-gtwy vector , derived from the pPC97-Gal4-BD vector . We transformed baits into yeast strain Y8930 , and mated these to a pPC86-Gal4-AD prey library of mixed-stage C . elegans cDNAs in strain Y8800 . Plasmids for two-hybrid experiments are listed in Supplementary file 1B . We screened >2 × 106 independent colonies per bait , and identified interacting cDNAs by plasmid amplification and sequencing . To test specific interactions we cloned the appropriate full length or fragment cDNAs into the pACT2 ( Gal4 activation domain ) or pBTM116 ( LexA DNA-binding domain ) vectors ( Clontech , Mountain View , CA ) and co-transformed constructs into yeast strain L40 . We grew transformed yeasts on agar plates with SD medium ( synthetic minimal medium ) lacking leucine and tryptophan; interactions were examined on plates with SD medium lacking leucine , tryptophan , and histidine , with or without 3-AT . Plasmids used in co-immunoprecipitation experiments are listed in Supplementary file 1B . We co-transfected FLAG-tagged EFA-6N150 or EFA-6N150Δ18 , and HA-tagged TAC-1 or ZYG-8 into HEK293 cells using X-tremeGene 9 DNA Transfection Reagent ( Roche Diagnostics Corporation , Indianapolis , IN ) . 48 hr after transfection , cells were lysed using lysis buffer ( 25 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 1 mM EDTA , 1% NP-40 and 5% glycerol ) . Anti-FLAG M2 antibody conjugated magnetic beads ( Sigma M8823 , Sigma-Aldrich , St Louis , MO ) were used for IP; anti-HA ( rabbit ) ( Abcam ab9110 , Abcam , Cambridge , UK ) and anti-FLAG ( rabbit ) ( Sigma F7425 ) were used for western blotting . We used CRISPR based gene targeting ( Dickinson et al . , 2013 ) to delete the genomic region encoding the EFA-6 18 aa motif . Briefly , we obtained Cas9-sgRNA plasmid from the Goldstein lab and inserted an sgRNA sequence targeting efa-6 into the vector using Quikchange mutagenesis . The two sgRNA sequences were GGCGAGGGGCTCCATCAATGG and GATGCAACTGTGGTACCTGG , targeting efa-6 exon 1 and exon 2 respectively . 50 ng/μl of each Cas9-sgRNA plasmid and 20 ng/μl Psur-5-mCherry were co-injected into wild type animals . From 15 F1 progeny we found one animal heterozygous for efa-6 ( ju1200 ) , which deletes 500 bp of exon 1 , intron 1 , and exon 2 , and has a 26 bp insertion . mRNAs produced in efa-6 ( ju1200 ) encode polypeptides with a premature stop codon after amino acid 15 , eliminating the 18 aa motif and the rest of EFA-6 . We measured locomotion velocity using WormTracker 2 . 0 ( Brown et al . , 2013 ) NGM plates were seeded with OP50 bacteria 3 hr before experiments . Individual young adult worms were picked gently from the culture plate to a fresh tracking plate . 1 min later , the plate was placed on the worm tracker platform and locomotion recorded for 1 min at 10 frames per second for each animal . We used Prism ( GraphPad Software , La Jolla , CA ) for all statistical analysis . A two-tailed Student's t-test was used for comparisons of two groups . For multiple comparisons we used one-way ANOVA with Bonferroni post test . To compare variables such as growth cone percentage we used the Fisher exact test . | In the nervous system , cells called neurons carry information around the body . These cells have long thin projections called axons that allow the information to pass very quickly along the cell to junctions with other neurons . Neurons in adult mammals are limited in their ability to regenerate , so any damage to axons , for example , due to a stroke or a brain injury , tends to be permanent . Therefore , an important goal in neuroscience research is to discover the genes and proteins that are involved in regenerating axons as this may make it possible to develop new therapies . An internal scaffold called the cytoskeleton supports the three-dimensional shape of the axons . Changes in the cytoskeleton are required to allow neurons to regenerate axons after injury , and drugs that stabilize filaments called microtubules in the cytoskeleton can promote these changes . Chen et al . used a technique called laser microsurgery to sever individual axons in a roundworm known as C . elegans and then observed whether these axons could regenerate . The experiments reveal that a protein called EFA-6 blocks the regeneration of neurons by preventing rearrangements in the cytoskeleton . EFA-6 is normally found at the membrane that surrounds the neuron . However , Chen et al . show that when the axon is damaged , this protein rapidly moves to areas near the ends of microtubule filaments . EFA-6 interacts with two other proteins that are associated with microtubules and are required for axons to be able to regenerate . Chen et al . 's findings demonstrate that several proteins that regulate microtubule filaments play a key role in regenerating axons . All three of these proteins are found in humans and other animals so they have the potential to be targeted by drug therapies in future . The next challenge is to understand the details of how EFA-6 activity is affected by axon injury , and how this alters the cytoskeleton . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2015 | Axon injury triggers EFA-6 mediated destabilization of axonal microtubules via TACC and doublecortin like kinase |
Natural environments feature mixtures of odorants of diverse quantities , qualities and complexities . Olfactory receptor neurons ( ORNs ) are the first layer in the sensory pathway and transmit the olfactory signal to higher regions of the brain . Yet , the response of ORNs to mixtures is strongly non-additive , and exhibits antagonistic interactions among odorants . Here , we model the processing of mixtures by mammalian ORNs , focusing on the role of inhibitory mechanisms . We show how antagonism leads to an effective ‘normalization’ of the ensemble ORN response , that is , the distribution of responses of the ORN population induced by any mixture is largely independent of the number of components in the mixture . This property arises from a novel mechanism involving the distinct statistical properties of receptor binding and activation , without any recurrent neuronal circuitry . Normalization allows our encoding model to outperform non-interacting models in odor discrimination tasks , leads to experimentally testable predictions and explains several psychophysical experiments in humans .
The olfactory system , like other sensory modalities , is entrusted to perform certain basic computational tasks . Of primary importance is the specific identification of odors and the recognition of isolated sources or objects in an olfactory scene . A typical scene in a natural environment is complex: the olfactory landscape is determined by the chemical composition of odorants released by the objects , the stoichiometry of the mixture and the physical location of the objects relative to the observer . An efficient olfactory system is expected to eliminate irrelevant background components and de-mix contextually relevant components received as a blend ( Ache et al . , 2016; Cardé and Willis , 2008; Gottfried , 2010; Hopfield , 1999; Howard and Gottfried , 2014; Jinks and Laing , 1999; Knudsen et al . , 1993; Pentzek et al . , 2007; Raguso , 2008; Riffell , 2012; Riffell et al . , 2008; Riffell et al . , 2014; Rokni et al . , 2014; Stevenson and Wilson , 2007; Szyszka and Stierle , 2014; Thomas-Danguin et al . , 2014 ) . The importance of filtering a complex background is shared by the olfactory and the adaptive immune systems . In the latter , lymphocytes must quickly and accurately identify a small fraction of foreign ligands in a sea of native ligands ( Abbas et al . , 2014 ) . Inhibitory feedback plays a key role in meeting the challenge of a proper combination of rapidity , sensitivity and specificity ( François et al . , 2013 ) . Inhibitory interactions in the form of receptor antagonism have indeed been observed in experiments with olfactory receptor neurons ( ORNs ) ( Oka et al . , 2004; Takeuchi et al . , 2009; Kurahashi et al . , 1994 ) , although it has not been quantified systematically . For instance , the response of cells expressing the mOR-EG receptor is strongly suppressed when methyl isoeugenol is delivered together with the receptor’s cognate ligand , eugenol , at equal concentrations ( Oka et al . , 2004 ) . Further evidence of intensity suppression and overshadowing ( i . e . when one odorant makes another indiscernible ) in the perception of odorant mixtures comes from psychophysical observations ( Lawless , 1997; Keller and Vosshall , 2004; Doty and Laing , 2015; Thomas-Danguin et al . , 2014 ) . The importance of peripheral interactions in shaping mixture perception has been directly shown by electrophysiological and psychophysical measurements ( Bell et al . , 1987; Laing and Willcox , 1987; Chaput et al . , 2012 ) . However , the functional role , if any , of inhibition at the ORN level remains unknown . Each ORN expresses receptors of a particular type , which typically display broad sensitivities to different odorants , whereas each odorant binds promiscuously to receptors of many types . The axons of ORNs of a common type converge onto glomeruli , where the axon terminals form synaptic contacts with mitral and tufted ( M/T ) projection neurons leading to the cortex , as well as local periglomerular ( PG ) interneurons . The activation of an individual glomerulus therefore represents the activation of a single ORN type . Discriminatory computations are carried out by brain regions such as the olfactory cortex , which receive combinatorial information from the entire ORN ensemble . To achieve a quantitative description of ORN inhibitory effects , it is then imperative to take their global nature into account . In other words , it is necessary to address the knowledge gap between the mixture response properties of a single ORN , the ensemble glomerular response , and ultimately its influence on odor discrimination and perception , which constitutes the goal of the present work . Previous computational models that examined discrimination tasks have , for simplicity , assumed a linear summation model of mixture response at the ORN level ( Hopfield , 1999; Koulakov et al . , 2007; Zhang and Sharpee , 2016; Zwicker et al . , 2016; Mathis et al . , 2016; Grabska-Barwińska et al . , 2017 ) . Conversely , our emphasis is on explicitly characterizing the ORNs’ biophysical attributes , with a focus on mixture response properties . A key aspect of the model is that odorant-receptor interactions depend on two distinct features: the sensitivity to binding and the efficiency of activation after binding , respectively . Competitive antagonism occurs when a component in a mixture that binds strongly , activates the downstream transduction pathway less effectively compared to other components . While this might naively seem disadvantageous , we show how an antagonistic encoding model has inherent normalization properties , leading to superior performance in odor discrimination and identification tasks . Finally , we make an explicit connection between a variety of psychophysical observations related to the perception of odorant mixtures and inhibitory effects at the single ORN level , providing a potential neurobiological basis to perceptual phenomena .
Odorants in the nasal cavity are captured by G-protein-coupled-receptors located on the cilia of ORNs . The conversion of chemical binding events into transduction currents in the cilia leads to spike signals transmitted to the brain ( see Refs . [Pifferi et al . , 2010; Kleene , 2008] for reviews ) . The signal transduction pathway is complex; here , we build a chemical rate model of the mammalian ORN meant to capture its major response properties . The model complements and extends previous work on mixture interactions ( Rospars et al . , 2008; Cruz and Lowe , 2013 ) . In Figure 1A , we illustrate the transduction pathway as modeled , and present a summarized version below ( see Materials and methods for details ) . The binding of an odorant to a receptor induces the activation of the odor-receptor complex via a two-step process . For an odorant X , its interactions at the receptor level are then represented by a two-step process: ( 1 ) R+X⇌κ1RX⇌κ2RX∗ , where κ1 and κ2 are the ratios of backward to forward rates for the binding and the activation steps . R , RX and RX∗ represent unbound , bound inactive and bound activated receptors , respectively . Activated complexes convert ATP into cAMP molecules via adenylyl cyclase III . The rate of production of cAMP is assumed independent of the available ATP ( which is in excess ) , and is therefore proportional only to the number of activated receptors . The cAMP molecules diffuse locally and cooperatively open nearby cyclic-nucleotide-gated ( CNG ) channels permeable to Ca2+ and Na+ ions . Since CNG channels are distributed uniformly along the cilia ( Flannery et al . , 2006; Takeuchi and Kurahashi , 2005 ) , we ignore the time required for cAMP to diffuse from its production site to the CNG channel . CNG channels have four binding sites for cAMP that exhibit allosteric cooperativity , which leads to nonlinear response functions of the Hill form ( Segel , 1993 ) and constitutes the first stage of signal amplification in the ORN . The generated Ca2+ current is then proportional to the number of fully bound CNG channels and is exchanged out of the cell at a constant rate through Na+/Ca2+ exchangers . Ca2+ ions open Cl− channels further downstream , which produces an outward amplifying Cl− current ( Boccaccio and Menini , 2007 ) . Spike firing , in proportion to the current , follows . Lastly , ORN adaptation occurs due to the subsequent blocking of the CNG channels by a Ca2+-calmodulin complex . Despite the complexity of the transduction model , the specific identity of odorants plays a role only at the level of receptors ( except for masking agents , described below ) . Odorants are characterized by two parameters ( mathematically defined by ( Equation 15 ) in the Materials and methods ) : the sensitivity , κ−1=1+κ2κ1κ2 , which controls the affinity of X to the receptor , and the activation efficacy , η , which combines κ2 with parameters of downstream reaction steps to measure the current produced by X once bound . Numerically integrating the set of coupled rate equations presented in the Methods yields the temporal firing rate response of the ORN to pulses of odorant molecules and their mixtures at various concentrations ( Figure 1B–D ) . For single odorants , the model successfully captures the strongly non-linear peak response for different concentrations of the odorant , the latency in response ( Rospars et al . , 2003 ) , the quadratic rate of cAMP production ( Takeuchi and Kurahashi , 2017 ) , and calcium-based adaptation ( Kurahashi and Menini , 1997 ) ( Figure 1E–G ) . Since our focus in subsequent analysis will be on the peak response , its form is reproduced here ( see Materials and methods for details ) for a monomolecular odorant delivered at a concentration C for a fixed , short duration: ( 2 ) FC=Fmax1+ ( 1+C/κηC/κ ) n . Here , n is the Hill coefficient and Fmax is the maximum physiologically possible firing rate , which depends on parameters related to the transduction pathway downstream of receptor activation and can be rescaled to unity . The maximal response at saturating concentrations , F∞=Fmax/ ( 1+η−n ) is truncated below Fmax by η , which controls the equilibrium level of activated receptors . On stimulation with more than one chemical species , the different species bind and activate the ORN in distinct ways . Its peak response to a pair of odorant molecules A and B is a special case of the general formula ( Equation 14 ) derived in the Materials and methods , and reads: ( 3 ) FCA+CB=Fmax1+ ( 1+CA/κA+CB/κBηACA/κA+ηBCB/κB ) n . Strong amplification by the ion channels render the mixture response hyper-additive at concentrations close to the sensitivity threshold; at higher concentrations , as the receptors become saturated and the odorants compete for limited binding sites , the response turns hypo-additive . The reduction in response due to competitive antagonism is determined by the binding affinity of the weaker odorant ( with lower activation efficacy ) relative to the stronger one . Before further theoretical analysis , it is worth clarifying a few points . One may question the relevance of introducing a separate η parameter , since the saturating ORN response could instead be due to limiting factors downstream of receptor activation . Our assumption is motivated by observations made from spike recordings of single rat ORNs to odorant mixtures , where the responses of the same ORN to saturating concentrations of different odorants yield very different firing rates ( Rospars et al . , 2008 ) . Since odorant-ORN specific interactions occur only at the receptor , the differences at saturation must be due to differences in receptor activation efficacy . Further evidence comes from examining the minimum latency of spiking response to different odorants at saturating concentrations ( Figure 1E ) , which can range from a few hundred milliseconds to a few seconds depending on the odorant-ORN pair ( Rospars et al . , 2003 ) . The lifetime of odorants bound to a receptor is short ( of a few milliseconds ) and the probability of activation of a particular G-protein in a 50 ms interval at saturating concentrations is low ( Bhandawat et al . , 2005 ) . These observations strongly suggest that the latency arises due to a relatively slow build-up of cAMP over many low-probability activation events . The differences in latency at saturating concentrations , when all receptors are bound , is therefore most likely due to differing activation probabilities , in line with our assumptions on η . Note that even though both the sensitivity κ−1 and the activation efficacy η depend on κ2 ( Equation 15 ) , the low probability of activation , which is reflected in the limit κ2≫1 , implies κ−1 and η depend separately on κ1 and κ2 . While a purely competitive model of mixture interactions captured many cases from previous experiments on ORN mixture responses , a significant fraction showed discrepancies ( Rospars et al . , 2008 ) . Non-competitive interactions are particularly manifest in synergy or suppression , which correspond to the mixture response curve lying above or below the individual response curves for each odorant , respectively , neither of which is possible with pure competition . Below , we show how non-competitive antagonistic effects , namely masking , can generate those effects . Non-competitive inhibition due to PI3K-dependent antagonism ( Ukhanov et al . , 2010 ) is beyond the scope of this paper . Masking is the phenomenon of non-specific suppression of CNG channel currents ( Kurahashi et al . , 1994; Takeuchi and Kurahashi , 2017 ) . Experimental evidence suggests that masking agents disrupt the lipid bilayer on the cell membrane , and thereby alter the binding affinity of cAMP to the CNG channels ( Takeuchi et al . , 2009 ) . Masking agents can also be odorants ( like amyl acetate ) , that is , they also bind to receptors and excite the transduction pathway . We suppose that the agents bind to sites on the lipid bilayer , and that multiple masking agents compete for the available sites . Similar to ( Equation 1 ) , the effects of a masking agent are determined by its affinity KM for the masking binding sites , and a masking coefficient μ ( lying between 0 and 1 ) , which measures its inhibitory effects once bound ( see Materials and methods ) . The latter quantifies the lowered affinity of cAMP for the CNG channels when a masking agent is bound in the vicinity of the channel . Since the activation efficacy η that appears in ( Equation 2 ) quantifies the effective rate of signal transduction , the effect of a lowered affinity appears as a lowered value for η . Specifically , we show that η→ ( 1−μM~ ) η , where M~ is the fraction of masking sites occupied by the masking agent . It follows that the firing rate at saturating concentrations , which depends on η , is reduced ( see [Equation 19] in Materials and methods ) . The model above ( and detailed in the Materials and methods ) reproduces qualitative features of odorant suppression observed in experiments ( Kurahashi et al . , 1994; Takeuchi et al . , 2013 ) ( Figure 2A , B ) . We stress that the presented fits serve as qualitative consistency checks; more experimental data on masking agents is required to verify the specifics of the model . Importantly , we show that both suppression and , counterintuitively , synergy are possible mixture interactions that could arise due to masking ( Figure 2C , D ) , both of which have been observed in experiments with single ORNs ( Rospars et al . , 2008 ) . To define suppression and synergy , we remark that the response curve of the mixture , taking into account competitive binding alone , always lies between the response curves of the individual components . This property is a simple consequence of the fact that the number of activated receptors for a mixture ( at a particular total concentration ) can never be larger/smaller than the most/least effective odorant delivered alone at the same total concentration . Suppression can then be defined as the situation when the mixture response curve is lower than the lowest response curves among the components while synergy occurs when the mixture response is higher than the highest response curve . In our model , synergy is qualitatively due to the taming of suppressive effects: for instance , let us consider a component A that binds masking sites more weakly than B , yet it is a stronger suppressor . If A binds and activates the ORN more strongly than B , the net effect of mixing A and B is to reduce the suppressive masking effect of A and unmask its strong activation properties , which can exceed the individual response curves as in Figure 2C . Equipped with the biophysical model above , we now proceed to investigate the functional consequences of antagonism . To this end , we first define a model of olfactory encoding that focuses on competitive antagonism and introduce simplifying assumptions to highlight the main ideas . An odorant is defined by two N-dimensional vectors of sensitivities κ−1 and activation efficacies η across N distinct ORN receptor types . We take N=250 , which is large enough to generalize our results across species . Parameters for different odorants are drawn independently from log-normal probability distributions ( see Materials and methods ) , although our main conclusions below do not depend on their specific form . The width of the κ−1 distribution reflects the broad sensitivities of odorants , spanning about six orders of magnitude ( Saito et al . , 2009 ) . An odorant caught in a sniff elicits a response in the glomerular ensemble whose individual activations vary in magnitude and progress differently in time . We focus on the vector y , which represents the peak responses ( Equation 2 ) to the odorant for the different ORN types . The statistics of the glomerular response is encoded in the distribution of y , which is directly related to the distribution of η for saturating concentrations of the odorant . A straightforward generalization of the expression for the binary mixture response from ( Equation 3 ) allows us to write the mixture response for K components in terms of effective mixture parameters κmix and ηmix ( which replace κ and η in [Equation 2] ) : ( 4 ) κmix−1=∑i=1K βiκi−1 ; ηmix=κmix∑i=1K ηiβiκi−1where βi is the fraction of component i in the mixture . We first consider for simplicity an equiproportionate mixture , and then show below that this is not a limitation . The key observation made here is that , since sensitivities are broadly distributed , we typically have one term dominating the sum in the expression for κmix−1 ( see Figure 3B ) . If we suppose that this dominant term is κM−1 , we may write: ( 5 ) κmix−1≈βMκM−1 ; ηmix≈ηM , for typical values of ηM/ηi for i≠M . The expression for ηmix above follows from ( Equation 4 ) when we approximate the sum in the expression for ηmix as ηMβMκM−1; multiplying this sum with the expression for κmix in ( Equation 5 ) then yields the approximation . Relation ( Equation 5 ) still holds for complex mixtures as the broad width of the sensitivity distribution ensures the dominance of one of the κ’s even for relatively large numbers of components . In Figure 3C , we show that the approximations in Equation 5 result in a relative error in κmix−1 and ηmix of about ∼40% for equiproportionate mixtures with over a hundred components . In other words , it is highly likely that only one component in the mixture occupies the receptors of ORNs of a particular type , and thus the response from these ORNs is determined by the activation efficacy of that specific component . In order to measure the strength of competitive antagonism , we introduce the antagonistic factor ρ . To define ρ , we note that an odorant A competitively antagonizes odorant B ( at equal concentrations ) when its sensitivity exceeds that of B ( κA−1>κB−1 ) , yet its activation efficacy is lower than B ( ηA<ηB ) . A quantification of this relationship is the Pearson correlation coefficient between binding and activation strengths across the ORN ensemble: ( 6 ) ρ≡Corr ( logκ−1 , logη ) , where the logarithms conveniently account for the broad range of the two variables . Let us first consider the extreme case of ρ=1 , when there is no antagonism as the odorant that binds best also has the strongest activation . This corresponds to an ORN behaving as a logical OR gate , a feature shared by any additive model of mixture response . From Equation 5 , the consequence is that ηmix always takes the maximum value of η in the mixture , which significantly biases ηmix toward higher values . To see this , we define the sparsity p , that is , the fraction of glomerular responses above a certain threshold τ for an odorant at saturating concentrations; then , the fraction above τ for a mixture with K components is given by 1− ( 1−p ) K . Thus , for any additive model of mixture response , the sparsity quickly saturates to one as K increases and all information about individual components is lost . On the contrary , when κ−1 and η are independent , that is , ρ=0 , ηM is independent of the constraint κM−1≫κi−1 ( i≠M ) implicit in Equation 5 and the distribution of ηmix precisely matches the distribution of the single component ηM . The ρ=0 condition of decorrelation between κ−1 and η is important for the argument since , even though Equation 5 holds generally , the constraint that M is the most sensitive odorant biases the statistics of ηmix for ρ>0 . Since the entire distribution of activations across the ORN ensemble is invariant to the number of components in the mixture , we conclude that the statistics of activation is conserved as the complexity of the mixture increases , that is , the population response is ‘normalized’ ( Figure 4A–C , Figure 4—figure supplement 1 ) . Remarkably , such a normalization of the mixture response is a direct consequence of antagonism in receptor encoding , independent of any neural circuitry . The upshot is that a sparse representation for a single odorant typically remains sparse for a complex mixture , allowing for improved performance in the detection of individual components ( as quantified in the next Section ) . Notably , the above arguments rely solely on the broad distribution of the sensitivities , which enables the approximation in Equation 5 . Despite the ∼40% error in this approximation , our arguments for normalization presented above still hold for a mixture with over a hundred components , as shown in Figure 4A . An important consequence is that the approximation becomes even better when the concentrations of the components are allowed to be different , as any variation in the βi’s makes the distribution of each term βiκi−1 even broader . This is confirmed by the plots in Figure 3C . The result holds generally true for any distribution of η and any broad distribution of κ−1; their log-normal forms are used here only to simplify subsequent calculations ( Figure 4C ) . Structural constraints at the receptor level , however , are likely to hamper a perfect decorrelation ρ=0 . Nevertheless , normalization does not require an exact equality and its effects fade gradually as ρ increases ( see Figure 4A ) . The extent of the advantageous effects of normalization ( and the ρ value ) depends on the sparsity of activation for single odorants , the number of components in the mixture and their properties . Indeed , the effect of normalization on the detection of an odorant in the presence of a large number of other odorants depends on the balance between two opposing factors . On the one hand , normalization induces an advantageous effect of maintaining sparsity and preventing saturation of the bulb , leading to easier segmentation . On the other hand , the number of active glomeruli corresponding to each odorant is greatly reduced , which makes detection harder . To explore how our model performs in discrimination tasks , we next compute the performance of the antagonistic encoding model described above in detecting a known odorant from a large background of unknown odorants , that is , figure-ground segregation ( Rokni et al . , 2014 ) . The capacity of an optimal Bayesian decoder in the task depends on the mutual information I ( T;y ) ( in bits ) that the glomerular pattern y preserves about the presence ( T=1 ) or absence ( T=0 ) of the target . To simplify the calculation of I ( T;y ) , we convert the vector of continuous values y into a binary vector z by applying a threshold τ that partitions the glomeruli into two subsets , active and inactive glomeruli ( see Figure 3A ) . In general , any continuous read-out is demarcated into a few discrete , distinguishable states depending on the level of intrinsic noise in the system . Taking more graded states into account will not change the qualitative result of our calculation . Specifically , as we show below , the relative performance between an antagonistic and a non-antagonistic model is still dominated by the loss of information due to glomerular saturation , which occurs independently of the number of gradations in our read-out . Figure 5A demonstrates that an encoding model with significant antagonism ( ρ=0 ) contains more information than a non-antagonistic model ( ρ=1 ) as the background increases in complexity . The results are robust to the presence of significant internal variability in the transduction pathway of an ORN ( Figure 5—figure supplement 1 ) . When the number of odorants in the mixture is small , the glomerular pattern for the non-antagonistic model is not saturated and preserves information about the glomeruli activated by each component . For a specified sparsity p ( as defined in the previous section as the fraction of glomerular responses above the threshold τ at high concentrations ) , the non-antagonistic case is thus advantageous when the mixture complexity is less than ∼1/p . However , for mixtures of higher complexity , it is useful to introduce correspondingly higher levels of antagonism in order to prevent saturation and still maintain an ability to segment out different components . To further emphasize this point , for different values of p , we compute the level of antagonism that maximizes information transmission when the background varies both in composition and complexity ( see Materials and methods ) . We find that for experimentally observed levels ( 0 . 1–0 . 3 ) of sparsity ( Saito et al . , 2009; Lin et al . , 2006; Soucy et al . , 2009; Vincis et al . , 2012 ) , it is always advantageous to incorporate non-zero levels of antagonism into odorant encoding ( Figure 5B ) . We measure the performance of a linear classifier in component separation , the task of identifying several known components from a mixture , for different levels of antagonism . Component separation is qualitatively different as the information about the other known odorants can be recurrently exploited to extract more information about an odorant’s presence or absence ( Grabska-Barwińska et al . , 2017 ) . First , a linear classifier is trained to individually identify 500 known odorants in the presence of other odorants from the set . In the test phase , a mixture which contains 1 to 20 known components , uniformly chosen , is delivered . The hit rate measures the fraction of odorants that were correctly identified , while the false positives ( FPs ) is the number of odorants out of the 500 that were not actually present but were declared to be present . Generalized Receiver Operating Characteristics ( ROC ) curves are drawn by varying the detection threshold of the linear classifier for each odorant ( Figure 5C–D ) . We find again that antagonism in receptors yields superior performance , independent of sparsity . Psychophysical observations related to odor perception were the primary investigative tools before neurobiological studies became prominent in the last few decades . Formulating an explicit connection between the vast body of literature on olfactory psychophysics and recent discoveries in neurobiology remains a challenge , particularly since perception is influenced by interactions throughout the olfactory sensory pathway ( Jinks and Laing , 1999; Su et al . , 2009; Grossman et al . , 2008; Wilson and Sullivan , 2011 ) . Various direct and indirect measurements ( Bell et al . , 1987; Laing and Willcox , 1987; Chaput et al . , 2012 ) strongly hint at the role of receptor-level interactions , although a mechanistic explanation for how these effects may arise has not been proposed . Here , we examine the possible relation between antagonism and observations from psychophysical experiments on the perception of odor mixtures . The upshot is that the combination of competitive antagonism and masking supports the diverse range of well-established psychophysical effects enumerated hereafter . Specifically , the list of psychophysical effects relevant here is as follows ( see ( Lawless , 1997; Keller and Vosshall , 2004; Doty and Laing , 2015; Thomas-Danguin et al . , 2014; Takeuchi et al . , 2013 ) ) . ( 1 ) Inhibition and synergy: The former is the strong reduction of perceived intensities when two odorants are mixed , usually at high concentrations; synergy is occasionally observed , namely at low concentrations . ( 2 ) Masking: When the concentration of a masking agent , such as 2 , 4 , 6- trichloroanisole ( called cork taint ) , is increased , the perceived intensity of the odorant decreases . ( 3 ) Symmetric and asymmetric suppression: When two odorants of equal perceived intensities are added , they typically suppress each other in a striking reciprocal fashion so that the perceived intensity of both odorants is still equal but sharply lowered . Asymmetric suppression ( sometimes called counteracting ) , where the intensity of one of the odorant is lowered more than the other , is observed occasionally . ( 4 ) Overshadowing: The loss of perception of a less intense odorant when a more intense odorant is present in a mixture . To examine the prevalence of inhibition and synergy , we estimate the inferred concentration ( or perceived intensity ) of a component in the mixture as the concentration at which those precise number of glomeruli corresponding to the odorant would have been activated ( i . e . above a fixed threshold ) had that odorant been delivered alone . In Figure 6A , we show that this simple algorithm leads to an unbiased estimate of the concentration over a broad range of concentrations . Figure 6A further demonstrates that inhibition and synergy naturally arise from competitive antagonism . Stronger inhibition arises at higher concentrations as the glomeruli that are otherwise activated by an odorant when delivered alone are antagonized by the second odorant , leading to lower perceived intensity . At lower concentrations , competitive antagonism plays a limited role; instead , cooperative effects in the ORN transduction pathway result in hyper-additivity , which pushes a few glomeruli above the activation threshold and gives rise to a small synergistic effect . Masking is also readily explained by our model , where the inferred concentration ( based on the activated glomeruli ) is below the actual concentration as the masking agent’s concentration increases ( Figure 6B ) . To quantify suppression , we use the fraction of suppressed glomeruli , defined as the fraction of glomeruli which are inactive in the mixture of A and B , yet are activated in isolation by odorant A ( B ) and not activated in isolation by odorant B ( A ) . Figure 6C shows that the reciprocal suppression of intense binary mixtures is conspicuously absent for a non-antagonistic model . To see this , let us suppose that each odorant individually activates half of all glomeruli ( i . e . the sparsity p = 0 . 5 ) . Note , A is more sensitive than B to half ( on average ) of all the glomeruli that B activates . Then , for ρ=0 , when they are delivered together at equal concentrations , A binds better to half the otherwise active receptors of B , activating half of them and suppressing the other half . Since B has precisely the same effect on A , each odorant reciprocally suppresses the other . On the other hand , when there is no antagonism , A still binds better to half the active receptors of B , but now activates all of them , resulting in no suppression . A strong asymmetric suppressive effect is observed when one of the odorants has a capacity for masking . Finally , to quantify overshadowing , we train a logistic regressor to identify a set of known odorants as in Figure 5C–D . A weak odorant B is delivered along with a stronger odorant A at varying concentration ratios . The probability of presence of B ascomputed by the regressor is compared against the the ratio of concentrations of A and B . When the probability of presence goes below the detection threshold ( set at 0 . 5 ) , B is no longer detected and is ‘overshadowed’ . Figure 6D demonstrates that overshadowing for binary mixtures is intensified by antagonism , in spite of its superior discriminatory performance for more complex mixtures .
Natural smells are due to mixtures of many chemicals , yet the need for tight stimulus control in experiments often leads to a focus on individual molecular entities . In this paper , we have characterized mixture interactions with a realistic biophysical model . Importantly , we explored how these interactions can naturally lead to ‘normalization’ of the glomerular responses , improve the coding capacity of the olfactory system , and account for many observed perceptual phenomena . The odorant receptor dynamics in our model is based on a two-step activation process analogous to previous works in vertebrates ( Rospars et al . , 2008; Cruz and Lowe , 2013 ) and the fly ( Nagel and Wilson , 2011 ) . A key aspect of two-step activation is that it separates sensitivity of ligand binding from activation efficacy ( del Castillo and Katz , 1954 ) . At the structural level , this distinction is consistent with observations on the binding and activation of GPCRs ( Strange , 2008 ) . The common and parsimonious approximation of a single step ( parametrized by a Kd for affinity ) in binding models , appears too drastic . In contrast to earlier work , we explicitly model the pathway downstream of receptor activation ( Pifferi et al . , 2010; Kleene , 2008 ) , which features the successive steps of cAMP production , allosteric opening of CNG channels , and ultimately current fluxes . This provides a biophysical basis to the cooperativity effects that were previously introduced ad hoc . Moreover , this explicit formulation allows us to go beyond pure competitive antagonism , which was reported to explain about half the cases and thus requires generalizations ( Rospars et al . , 2008 ) . In particular , non-competitive antagonistic effects , such as our masking model for the non-specific suppression of the cyclic nucleotide-gated channels permit us to account for synergy and inhibition effects that are impossible for competitive antagonism . A major focus of our work is the functional role of antagonistic interactions . Antagonistic reduction of glomerular activation can be seen as a form of ‘normalization of activity’ . Normalization with increasing stimulus intensity or complexity is common in neural systems ( Carandini and Heeger , 2011 ) and has been thought of as a circuit property that involves inhibitory synaptic interactions ( Wachowiak et al . , 2002; Cleland et al . , 2007 ) . In the olfactory system , this was elegantly demonstrated in the Drosophila antennal lobe , where activation of increasing number of receptors ( or glomeruli ) proportionally increases inhibition provided to any one glomerular channel ( Olsen et al . , 2010 ) . Similarly , in fish and mouse olfactory bulbs , increasing stimulus intensity is thought to recruit populations of interneurons ( namely , short axon cells ) that inhibit principal cells , leading to blunted activity for higher stimulus intensities . In an extreme example of this , a mouse with a particular receptor forcibly expressed widely has remarkably similar activation of the mitral cell population despite the massive increase in the input when the cognate odorant is presented ( Roland et al . , 2016 ) . The key insight from our model is that normalization is granted at the level of receptors by purely statistical reasons , without any additional circuit burden . In the limit of full statistical decorrelation between ligands’ binding affinity and activation efficacy , the distribution of activations across the ORN ensemble for a mixture coincides with that of a single monomolecular odorant , a property which has been confirmed in the fly ( Stevens , 2016 ) . Why would we need normalization if the optimal way to preserve information is to simply copy the input signal , i . e . have ORNs functioning as pure relays ? Copying however requires an unrealistically broad dynamic range , especially for the processing of natural mixtures , where the concentration and the number of components can fluctuate wildly . Normalization at the first layer in the sensory pathway helps avoid early saturation effects that would confound the entire processing pathway . However , nonlinear distortions of the signal do lead to loss of information , and the balance between the two effects calls for their quantification . Our information theoretic calculations demonstrate that detection of a target odorant within a complex mixture is enhanced by antagonistic interactions , and that holds for a wide range of receptor tuning widths , that is , the average number of activated glomeruli per odorant . Some of the predictions from our analysis can be tested experimentally . Direct measurements from mammalian ORNs have been obtained in vitro in many biophysical studies studying signal transduction ( Lowe and Gold , 1995; Bhandawat et al . , 2005 ) . Although such preparations offer excellent access for measurement , there is significant uncertainty about mimicking the native conditions in terms of delivery of odors ( airborne vs solution ) as well as the ionic composition of the perfusion medium , which will affect response amplitudes . Electrical recordings from individual ORNs are difficult and have low yield , but have offered tantalizing hints on different nonlinear interactions ( Rospars et al . , 2008 ) . More extensive data will likely have to rely on glomerular imaging methods ( Bozza et al . , 2004; Rokni et al . , 2014; Mathis et al . , 2016 ) , which offer robust signals and extended recording times to obtain measurements at different concentrations and mixture ratios ( Soucy et al . , 2009 ) . Here too , there are some concerns including potential effects of feedback mediated by olfactory bulb neurons on ORN axon terminals , particularly through GABAb receptors ( McGann et al . , 2005; McGann , 2013 ) . Carefully controlled experiments that isolate feedforward sensory signals could reveal the prevalence of antagonistic interactions in mammalian ORNs . Extensive measurements at a large range of concentrations and mixture ratios will allow robust fitting of our model to obtain accurate estimates of the two key parameters , η and κ . A key empirical test of our theory will then rest on the relation between η and κ for each glomerulus ( or ORN ) . If these two parameters are highly correlated most of the time , then antagonistic interactions of the sort described in our theory will likely have only weak impact on olfaction . Even a modest decorrelation , on the other hand , will give rise to important effects on ensemble coding of odor mixtures even at the front end of the olfactory system . In addition to functional advantages , we showed that antagonistic effects are consistent with psychophysical effects observed in mixture perception . Experiments show that the perceived intensity level of an odorant is empirically related to the true concentration of the odorant as a power function , reflecting their proportional relationship on a single logarithmic scale ( Lawless , 1997; Wojcik and Sirotin , 2014 ) . The intensity level for binary mixtures perception is commonly described via a vector sum of the intensities of each component ( Berglund et al . , 1973; Laing et al . , 1984; Berglund and Olsson , 1993 ) . The vector model captures the level independent , symmetric nature of mixture suppression . The biophysical model presented here is consistent with an even broader range of perceptual phenomena , including level independency , synergy , symmetric and asymmetric suppression , masking and overshadowing . The bottomline is that global antagonistic interactions at the ORN level may play a major role in non-trivial perceptual phenomena . In conclusion , ORNs are far from simple relays , and their strong nonlinear interactions crucially affect olfactory processing . Non-competitive antagonistic mechanisms , such as masking effects discussed here , have not been widely studied in mice and they may only occur for selected odorants . While this experimentally necessitates an extensive experimental dataset , the non-competitive effects presented here make their future investigation particularly relevant . Finally , the generality of the potential relations highlighted here between ORN antagonism and psychophysical phenomena motivates their exploration in mice , where a broader arsenal of experimental techniques and manipulations can be leveraged .
Here , we present a phenomenological description of non-competitive masking processes . We suppose that masking agents bind sites on the lipid bilayer and compete for their limited number . The suppression timescale and off-timescale are smaller than a few hundred milliseconds ( Kurahashi et al . , 1994 ) , justifying the assumption of steady state . In steady state , the occupancy fraction of the ith masking agent with concentration Mi and binding affinity KMi is ( 18 ) M~i=KMiMi1+∑i KMiMi . The disruption of a CNG channel conformation due to agent i is supposed to alter the affinity of cAMP to one of its binding sites on the channel in the reactions ( 9 ) . The energy of the cAMP bound state is increased by Δϵi and its probability is reduced by the corresponding Gibbs factor e−βΔϵi , where β=1/kT is the inverse temperature . The resulting reduction in the opening of the channels is most conveniently accounted for by a mean-field approach where the channel opening rate k′G appearing in ( 10 ) is modified by the masking agents . In other words , k′G→χMk′G with the suppression factor χM<1 derived below . It follows from the definition ( 13 ) of η that a modification of k′G by χM carries over to η as η→χMη . Therefore , when saturating concentrations of excitatory odorants are presented together with masking agents that produce a masking coefficient χM , the maximal firing rate is reduced as ( 19 ) FM ( ∞ ) =Fmax1+1/ ( χMη ) n . which reflects the masking effect . The dependence of the suppression factor χM on the concentrations Mi of the masking agents is estimated as follows . Let us denote the radius of disruption of the channels by a single masking molecule on the lipid bilayer by r , and the surface density of masking binding sites by σ . The typical number of masking binding sites surrounding a given CNG channel is then λ=πσr2 . The number nmask of masking binding sites within distance r of a CNG channel is assumed to be Poisson distributed , that is P ( nmask ) =e−λλnmasknmask ! . For a given number of sites nmask , the vector of their occupancy numbers I= ( i1 , i2 , . . iK+1 ) is distributed following a multinomial distribution with probabilities given by Equation 18 , that is , P ( I ) = ( nmaski1 , i2 , … , iK+1 ) ∏k=1K+1 M~kik . The index K+1 corresponds to unoccupancy , M~K+1=1−∑i=1K M~i and iK+1=nmask−∑k ik . The probability of each I is proportional to its Gibbs factor e−βΔϵ ( I ) , where Δϵ ( I ) is the energy shift to the binding of masking agents . We consider the first step in Equation 10; similar arguments hold for successive ones . The unmodified a0kG=e−βϵ is the ratio between the probability for a channel to be cAMP bound or cAMP unbound , and ϵ is their energy difference . In the presence of masking , there are multiple cAMP bound and unbound states , which differ in their occupancy of the masking binding sites . The sum over all those states defines the probabilities Pb and Pu of cAMP bound and unbound , respectively . The suppression factor χM that modifies k′G→χMk′G is obtained as the ratio ( eβϵPb/Pu ) 1/n , where the 1/n power stems from the definition of k′G in ( 10 ) . The sum Pb is obtained by combining all the previous factors: ( 20 ) Pb=∑nmask=0∞ e−λλnmasknmask ! ∑i1 , i2 , . . iK+1nmask e−β ( ϵ+Δϵ ( I ) ) Z ( nmaski1 , i2 , … , iK+1 ) ∏k=1K+1 M~kik . where Z is a normalization factor . Assuming the masking sites do not affect the energy of the channels when cAMP is unbound , the sum Pu has a similar expression with ϵ+Δϵ=0 . It is then verified that Pu=1/Z . As for Pb , the simplest possible assumptions are that Δϵ ( i1 , i2 , … , iK ) =∑k=0K ikΔϵk is additive , and the masking binding sites are dilute , that is , λ is small . Equation ( 20 ) reduces then to ( 21 ) χMn=eβϵPbPu=e−λ∑k M~k ( 1−e−βΔϵk ) ≈1−∑k=1K μkM~k , where μk=λ ( 1−e−βΔϵk ) satisfy 0≤μk≤1 , and the same inequality holds for χM . In general , masking agents can affect multiple CNG channel subunits ( Chen et al . , 2006 ) . If a masking agent affects the binding of cAMP to j CNG subunits , the suppression effect is χM= ( 1−∑i μiM~i ) m with m=j/n . The ratio 1−FM ( ∞ ) F ( ∞ ) in ( Equation 19 ) is plotted in Figure 2B and compared to experimental data . In Figure 2C , D , the parameters for generating the response curves for odorants A and B are κA=1 , κB=1 , ηA=1 , ηB=5 . For synergy ( Figure 2C ) , the masking parameters are KM , A=10−5 , KM , B=10−1 , μA=0 , μB=0 . 7 , while the corresponding parameters for inhibition ( Figure 2D ) are KM , A=1 , KM , B=10−5 , μA=0 , μB=0 . 7 . The parameter m is chosen to be unity in both cases . Every odorant is defined by a vector of binding sensitivities κ−1 and a vector of activation efficacies η , each with dimensionality N , where N=250 is the chosen number of receptor types . An odorant’s binding sensitivity to a particular receptor type is drawn independently from a log-normal distribution logκ−1∼N ( 0 , σκ−1 ) , where its standard deviation σκ−1 is set to 4 to obtain a six orders of magnitude separation between the most sensitive and least sensitive receptor types ( Saito et al . , 2009 ) . The activation efficacies are similarly drawn independently for each receptor type such that logη∼N ( 0 , 1 ) . The measure of antagonism , ρ , is defined as the Pearson correlation coefficient between logκ−1 and logη: ( 22 ) ρ≡⟨logκ−1logη⟩−⟨logκ−1⟩⟨logη⟩σκ−1ση where σ2 denotes the variance of the random variables and the angular brackets denote expectation values . To generate an odorant-receptor pair , first logη is drawn from the standard normal distribution . Then , logκ−1 is generated with correlation ρ as logκ−1=σκ−1 ( ρlogη+1−ρ2ω ) , where ω is drawn from a standard normal distribution . At saturating concentrations of an odorant , the peak firing rate it elicits for a receptor type with activation efficacy η is F∞=11+η−n ( see ( Equation 16 ) , where Fmax can be chosen to be unity ) . The rescaled glomerular activation vector y ( each component is rescaled between 0 and 1 ) is given by y=11+η−n , where the transformation is performed on each component of the vector . The probability p that each component exceed a threshold τ is given by the probability that a random variable drawn from a standard normal distribution exceed 1nlog1−ττ . This probability p represents the sparsity of the glomerular activations z at saturating concentrations after thresholding . The sparsity is set by selecting τ=11+e−nΦ−1 ( 1−p ) , where Φ is the cumulative distribution function for a standard normal random variable . To quantify the performance of the encoding model in figure-ground segregation , we compute the mutual information between the absence ( T=0 ) or presence ( T=1 ) of the target odorant T and the glomerular activation pattern z . Noise is introduced due to the presence of background odorants of unknown sensitivities and activation efficacies . The mutual information controls the performance of an optimal Bayesian decoder in detecting a target odorant in a background by using the glomerular activation pattern as input . The mutual information is defined as ( 23 ) I ( T;z ) =H ( T ) −H ( T|z ) , where H ( T ) is the entropy of target presence or absence , which equals one bit ( since the target is present in half the trials ) . The second term on the right hand side H ( T|z ) ( in bits ) is given by ( 24 ) H ( T|z ) =−∑z Pr ( z ) {Pr ( T=1|z ) log2Pr ( T=1|z ) +Pr ( T=0|z ) log2Pr ( T=0|z ) } , where Bayes’ formula yields Pr ( T=1|z ) =Pr ( z|T=1 ) Pr ( z|T=0 ) +Pr ( z|T=1 ) , and Pr ( z ) =∑nb Pr ( z|nb ) Pr ( nb ) , where Pr ( nb ) is the distribution of the number of background odorants . H ( T|z ) is estimated numerically by using Monte Carlo sampling . The quantities Pr ( z|T=1 ) and Pr ( z|T=0 ) are also computed numerically by observing that Pr ( z|T=1 ) =∑nb Pr ( z|T=1 , nb ) Pr ( nb ) . Due to the independence of the receptor types and since the background odorants are independently drawn , the probability Pr ( z|T=1 , nb ) factorizes into N multiplicative terms , each of which can be pre-computed prior to Monte Carlo sampling . To obtain the results in Figure 5A , we choose p=0 . 5 . When the number of background odorants fluctuate ( as in Figure 5B ) , we draw nb from a truncated exponential distribution with a mean of 32 and truncated at a maximum of 128 background odorants . To show that internal noise does not affect our results , we compare the performances of antagonistic and non-antagonistic models in figure-ground segregation by including an additional noise term in the expression for ORN response ( Equation 16 ) . Specifically , the response of an ORN to a mixture is modified as ηmix→ ( 1+ϵ ) ηmix , where ϵ is an effective noise term that condenses the variability in signal transduction relative to the number of activated receptors . We train a linear classifier to identify a target odorant against a fixed number of background odors . The target odorant and background odorants ( of varying composition ) are delivered at concentrations drawn from a uniform distribution in the logscale over 3 orders of magnitude . The discrimination accuracy of the linear classifier is shown in Figure 5—figure supplement 1 for ρ=0 , 1 and ϵ=0 and 0 . 4 . The results show the superior performance of antagonism when the number of background odors is large even at noise levels of 40% . The higher performance for ϵ=0 . 4 compared to ϵ=0 for ρ=1 ( at >30 background odorants ) occurs because the noise can desaturate glomeruli that are otherwise saturated . For the component separation task , we use an ensemble of linear classifiers as our decoders . A linear classifier computes the probability of presence of an odorant from the glomerular activation pattern z as 11+exp ( − ( θ . z+b ) ) , where θ and b are the vector of learned weights and bias , respectively . First , linear classifiers are trained to identify odorants from a fixed set S of 500 odorants . During the training phase , each classifier is trained to identify the presence of its target against a background of one to ten odorants also chosen from S . In each trial of the test phase , 1 to 20 odorants are uniformly chosen from S and the component separation performance is measured using the fraction of correct identifications ( hit rate ) and the number of false positives ( FPs ) . An odorant is declared to be present if the probability of presence exceeds a detection threshold . The hit rate and the number of false positives are modulated by sliding the detection threshold of each linear classifier , yielding the generalized ROC curves in Figure 5C , D . To infer the concentration based on the glomerular profile for a single odorant , we first note that glomeruli are progressively recruited as the concentration of the odorant increases depending on their sensitivity to that odorant . Suppose a glomerulus gi corresponding to this odorant is first recruited at concentration ci , where ci increases with the index i . Then , the contribution of gi to the total inferred log concentration , logc , is taken to be logci−logci−1 for i>1 and logc1 for i=1 ( here c1 corresponds to the concentration at which the most sensitive receptor becomes active ) . This simple scheme to infer the concentration is accurate for a single odorant ( Figure 6A ) and can be easily shown to have a concentration invariant Weber ratio ∼logcmax/c1pN where N is the number of receptor types , p is the sparsity and cmax is the saturating concentration ( Koulakov et al . , 2007 ) . To obtain the results for mixtures when ρ=0 , 0 . 5 and 1 , for each concentration of odorant A , another odorant B is delivered at an equal concentration . The log concentration of A is then inferred by computing the sum of the contributions from the glomeruli corresponding to odorant A as described above . The curve corresponding to the masking agent is obtained similarly as above for ρ=0 with B having a masking coefficient μB=1 . To plot the curves in Figure 6B , odorant A is delivered at a fixed concentration ( red , dashed line ) and a masking agent is applied at increasing concentrations ( horizontal axis ) . Here , we use ρ=0 and we assume the masking agent does not bind to anyof the receptors . To show symmetric and asymmetric suppression in Figure 6C , the concentration of A is fixed at a saturating concentration and the concentration of B is varied to get different concentration ratios CA/CB ( horizontal axis ) . We define a suppressed glomerulus as one that is active when an odorant is delivered individually yet is inactive when A and B are delivered together . The fraction of suppressed glomeruli for each concentration ratio is averaged over many samplings of the two odorants A and B . To obtain the results for overshadowing , the different concentration ratios are generated similar to the method described above for Figure 6C . Logistic regressors that can detect odorants A and B are first trained to identify them when delivered alone at varying concentrations . A logistic regressor computes the probability of presence of an odorant given the glomerular pattern of activation . A detection threshold can then be applied on the probability of presence in order to declare the odorant present or absent . In Figure 6D , we show the probability of presence of odorant B as determined by the logistic regressor corresponding to B for different concentration ratios and values of ρ ( solid lines ) . Similar curves for odorant A are shown as dashed lines . Code for the modeling can be accessed at: https://github . com/greddy992/Odor-mixtures ( Reddy , 2018 ) ; copy archived at https://github . com/elifesciences-publications/Odor-mixtures ) . | When ordering in a coffee shop , you probably recognize and enjoy the aroma of freshly roasted coffee beans . But as well as coffee , you can also smell the croissants behind the counter and maybe even the perfume or cologne of the person next to you . Each of these scents consists of a collection of chemicals , or odorants . To distinguish between the aroma of coffee and that of croissants , your brain must group the odorants appropriately and then keep the groups separate from each other . This is not a trivial task . Odorants bind to proteins called odorant receptors found on the surface of cells in the nose called olfactory receptor neurons . But each odorant does not have its own dedicated receptor . Instead , a single odorant will bind to multiple types of odorant receptors , and thus , each olfactory receptor neuron may respond to multiple odorants . So how does the brain encode mixtures of odorants in a way that allows us to distinguish one aroma from another ? Reddy , Zak et al . have developed a computational model to explain how this process works . The model assumes that an odorant triggers a response in an olfactory receptor neuron via two steps . First , the odorant binds to an odorant receptor . Second , the bound odorant activates the receptor . But the odorant that binds most strongly to a receptor will not necessarily be the odorant that is best at activating that receptor . This allows a phenomenon called competitive antagonism to occur . This is when one odorant in a mixture binds more strongly to a receptor than the other odorants , but only weakly activates that receptor . In so doing , the strongly bound odorant prevents the other odorants from binding to and activating the receptor . This helps tame the dominating influence of background odors , which might otherwise saturate the responses of individual olfactory receptor neurons . Reddy , Zak et al . show that processes such as competitive antagonism enable olfactory receptor neurons to encode all of the odors within a mixture . The model can explain various phenomena observed in experiments and it adds to our understanding of how the brain generates our sense of smell . The model may also be relevant to other biological systems that must filter weak signals from a dominant background . These include the immune system , which must distinguish a small set of foreign proteins from the much larger number of proteins that make up our bodies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2018 | Antagonism in olfactory receptor neurons and its implications for the perception of odor mixtures |
Macrophage-mediated phagocytosis and cytokine production represent the front lines of resistance to bacterial invaders . A key feature of this pro-inflammatory response in mammals is the complex remodeling of cellular metabolism towards aerobic glycolysis . Although the function of bactericidal macrophages is highly conserved , the metabolic remodeling of insect macrophages remains poorly understood . Here , we used adults of the fruit fly Drosophila melanogaster to investigate the metabolic changes that occur in macrophages during the acute and resolution phases of Streptococcus-induced sepsis . Our studies revealed that orthologs of Hypoxia inducible factor 1α ( HIF1α ) and Lactate dehydrogenase ( LDH ) are required for macrophage activation , their bactericidal function , and resistance to infection , thus documenting the conservation of this cellular response between insects and mammals . Further , we show that macrophages employing aerobic glycolysis induce changes in systemic metabolism that are necessary to meet the biosynthetic and energetic demands of their function and resistance to bacterial infection .
Macrophages represent a highly specialized and versatile population of cells that occur in all animals and perform a diversity of functions ( Lim et al . , 2017 ) . In the absence of an activating stimulus , macrophages reside as quiescent sentinel cells that have minimal metabolic requirements ( Davies and Taylor , 2015 ) . In response to extracellular triggers , however , macrophages undergo a dramatic change in behavior that coincides with an enhanced metabolic rate and increased energy demands ( Pearce and Pearce , 2013 ) . In this regard , the manner by which macrophages mount a response is dictated by the activating stimuli , which include tissue-damage- , pathogen- or microbe-associated molecular patterns ( DAMPs , PAMPs and MAMPs , respectively ) , as well as signaling molecules that are secreted by other cells , such as cytokines . Each challenge requires the induction of specific metabolic and physiological processes that allow for an adequate immune response ( Kawai and Akira , 2011 ) – cellular changes that are collectively known as a polarization phenotype . Macrophages polarize into bactericidal ( M1 ) or healing ( M2 ) functional phenotypes characterized mainly by metabolism ( Mills et al . , 2000 ) . M1 and M2 polarization phenotypes utilize distinct ways of generating ATP ( glycolysis vs . oxidative phosphorylation ) and metabolizing arginine ( NO synthesis vs . the ornithine cycle ) ( O'Neill and Pearce , 2016 ) . Nowadays , the whole spectrum of polarization phenotypes corresponding to particular functions has been described ( Mosser and Edwards , 2008; Martinez and Gordon , 2014 ) . Perhaps the most dramatic change in macrophage metabolism associates with the M1 bactericidal phenotype , in which cells increase both glucose consumption and lactate production independently of oxygen concentration - a phenomenon known as aerobic glycolysis ( AG ) ( Warburg et al . , 1927; Warburg , 1956 ) . The resulting metabolic program promotes increased glucose catabolism , thus allowing M1 macrophages to generate enough of the ATP and glycolytic intermediates necessary for elevated phagocytic cell activity ( Liberti and Locasale , 2016 ) . This shift in cellular metabolism towards AG appears to be a determining factor in macrophage function and in the development of the pro-inflammatory phenotype ( Galván-Peña and O'Neill , 2014 ) . Hypoxia inducible factor 1α ( HIF1α ) is a key regulator of AG within macrophages . Although this transcription factor is normally degraded in the presence of oxygen , the triggering of either Toll-like receptor ( TLR ) or Tumor necrosis factor receptor ( TNFR ) signaling within macrophages activates Nuclear factor kappa-B ( NFĸB ) and stabilizes HIF1α independently of oxygen availability ( Siegert et al . , 2015; Jung et al . , 2003 ) . This normoxic HIF1α stabilization promotes the expression of genes that are under the control of hypoxia response elements ( HREs ) , many of which are involved in cellular metabolism , cell survival , proliferation , and cytokine signaling ( Dengler et al . , 2014 ) . In this regard , two of the key HIF1α target genes encode the enzymes pyruvate dehydrogenase kinase ( PDK ) and lactate dehydrogenase ( LDH ) , which together shunt pyruvate away from the mitochondria and maintain NAD+/NADH redox balance independently of oxidative phosphorylation . Inhibition of both HIF1α and LDH represents an efficient experimental strategy to direct cellular metabolism from AG to oxidative phosphorylation in both mice and Drosophila ( Allison et al . , 2014; Geeraerts et al . , 2017 ) , demonstrating the crucial role of these enzymes in this metabolic switch . Although pyruvate metabolism within the tricarboxylic acid ( TCA ) cycle is limited during AG , the TCA intermediates are essential for many cellular processes . Therefore , cells under AG rely on feeding the TCA cycle with glutamine , causing a TCA cycle to be ‘broken’ ( Langston et al . , 2017 ) . Such a dramatic change in mitochondrial metabolism leads to significant imbalances in the cytosolic accumulation of TCA metabolites ( such as NO , succinate , fumarate , L-2-hydroxyglutarate ) that further contribute to HIF1α stabilization ( Bailey and Nathan , 2018 ) . While this feedback maintains AG , it simultaneously makes it dependent on a sufficient supply of nutrients from the environment ( Iommarini et al . , 2017 ) . Macrophages employing AG must consume sufficient carbohydrates to support biosynthesis and growth . In order to ensure an adequate supply of sugar and other nutrients , these cells produce signaling molecules that affect systemic metabolism in order to secure enough energy for themselves – a concept recently defined as selfish immune theory ( Jeong et al . , 2003; Straub , 2014 ) . According to this theory , signaling molecules released by immune cells induce systemic metabolic changes such as hyperglycemia and systemic insulin resistance to increase the titer of nutrients that are available for the immune response and to limit their consumption by other tissues and processes ( Dolezal , 2015 ) . As many of these signaling molecules are direct HIF1α transcriptional targets , HIF1α stabilization directs the cellular metabolism while it simultaneously induces the expression of genes that have an impact on the whole systemic metabolism ( Peyssonnaux et al . , 2007; Imtiyaz and Simon , 2010 ) . Thus , macrophages not only phagocytose cells but they also regulate the systemic metabolism of an organism . Drosophila macrophages , like those of mammals , serve an essential role in the immune system and are capable of responding to a wide array of stimuli , ranging from pathogenic bacteria and fungi to the corpses of apoptotic cells ( Wood and Martin , 2017; Sears et al . , 2003; Govind , 2008 ) . The mechanism of the bactericidal function itself is highly conserved at the molecular level between Drosophila and mammalian macrophages . In both Drosophila and mammals , two central signaling pathways , Toll and Imd ( TLR and TNFR functional homologs ) are triggered in response to pathogenic stimuli ( Valanne et al . , 2011; Buchon et al . , 2014; Lemaitre et al . , 1996 ) . The Toll and Imd pathways induce the NFĸB signaling in Drosophila , so we can assume that the phagocytic role of macrophages could be accompanied by stabilization of the HIF1α ortholog , Similar ( Sima ) , hereafter referred to as Hif1α ( van Uden et al . , 2011 ) . Indeed , normoxic stabilization of Hif1α followed by its nuclear localization and increased expression of HRE-controlled genes can induce metabolic changes that are typical of AG ( Romero et al . , 2008; Li et al . , 2013; Liu et al . , 2006; Herranz and Cohen , 2017; Eichenlaub et al . , 2018 ) . Even though the HRE-controlled genes frequently appear in transcriptomic data for activated insect macrophages ( Irving et al . , 2005; Johansson et al . , 2005 ) , the direct role of Hif1α in the macrophages has not yet been tested . Considering that the molecular mechanisms that control macrophage activation are similar in both Drosophila and humans , it seems logical that the metabolic changes that occur within these cells would also be comparable , but the metabolism of insect macrophages remains poorly understood . Here , we address this question by analyzing in vivo metabolic and transcriptional changes in adult Drosophila phagocytic macrophages by employing a model of Streptococcus-pneumoniae-induced sepsis . The well-defined progress of this infection allowed us to distinguish three phases of the immune response according to the changing dynamics of bacterial growth ( acute , plateau , and resolution phase ) ( Figure 1A ) . The acute phase lasts for the first 24 hr , during which the streptococcal population is rapidly growing and its abundance must be limited by phagocytosis to avert early death ( Pham et al . , 2007; Bajgar and Dolezal , 2018 ) . The established equilibrium between continuous bacterial growth and host bacterial killing results in the plateau phase lasting for the next four days . At the end of this period , the immune system of the host surmounts the infection and clears the majority of the pathogens . The following resolution phase ( 120 hr post-infection ( hpi ) and later ) is essential for macrophage-mediated clearance of bacterial residues and for the reestablishment of homeostasis ( Bajgar and Dolezal , 2018; Chambers et al . , 2012 ) . To analyze processes that are characteristic of highly active phagocytic macrophages in Drosophila , we compared the attributes of acute phase macrophages ( APMФs ) with those of macrophages from uninfected individuals and resolution-phase macrophages ( RPMФs ) . Using a previously described hemolectin-driven GFP ( HmlGal4 >UAS eGFP ) ( Jung et al . , 2005 ) , we isolated Drosophila adult macrophages ( approximately 15 , 000 cells/replicate ) and analyzed the metabolic and transcriptional responses that are induced within these cells upon infection ( Figure 1B ) . Our approach revealed that Drosophila macrophages respond to the acute phase of bacterial infection by increasing glucose uptake , elevating glycolytic flux , and producing lactate . Moreover , as in mammals , the activation and maintenance of AG within Drosophila macrophages depend on Hif1α , and require elevated Ldh activity . We also demonstrate that the induction of AG within Drosophila macrophages leads to a change in systemic carbohydrate metabolism . Overall , our findings demonstrate that Drosophila macrophages must induce both autonomous and systemic changes in carbohydrate metabolism to mount a proper bactericidal function and to resist infection .
Since the bactericidal function of phagocytic cells is connected with AG in mice ( Mills et al . , 2000 ) , we analyzed Drosophila macrophages for the occurrence of AG hallmarks , such as increased glucose uptake , an increase in glycolytic flux , and the generation of an NADH pool facilitating the Ldh-mediated reduction of pyruvate to lactate ( Langston et al . , 2017 ) . The distribution of fluorescently labeled deoxyglucose ( NBDG ) in an organism , frequently used in cancer research , reflects the competitive potential of tissues in glucose internalization ( Cox et al . , 2015 ) . We tested the effect of immune response activation on glucose distribution among tissues in Drosophila by feeding the infected or control flies with NBDG during a 24-hr period before the signal detection . Infected flies displayed prominent NBDG accumulation in APMФs compared to other tissues , which is in contrast to the distribution of NBDG seen in uninfected controls or in flies fed during the resolution phase of infection , which displayed no such accumulation ( Figure 2A , B ) . These results indicate an increased potential of phagocytosing macrophages to consume glucose in direct competition with other tissues during the acute phase of bacterial infection . The increased NBDG uptake by macrophages was further supported by gene expression analysis , which revealed that the transcription of genes encoding both glycolytic enzymes and LDH , but not TCA cycle enzymes , was significantly upregulated in APMФs ( Figure 2C ) . Moreover , these changes in glycolytic genes were restricted to the acute phase of infection as most glycolytic genes returned to a basal level of expression during the resolution phase , whereas hexokinase and enolase ( similarly to all analyzed TCA cycle genes ) even showed decreased expression ( Figure 2C and Figure 2—figure supplement 1 ) , which can be ascribed to the global suppression of metabolism in these cells . Overall , these results indicate that macrophages specifically upregulate glucose metabolism in response to S . pneumoniae infection . Increased glucose uptake and expression of glycolytic genes , including Ldh , suggest an increased glycolytic flux and preferential reduction of pyruvate to lactate in APMФs . To confirm this , we measured the enzymatic activity of LDH , as an enzyme responsible for the diversion of pyruvate from TCA , and phosphoglucose isomerase ( Pgi ) , as a glycolytic enzyme representative . In agreement with the expression data , Pgi enzymatic activity was significantly increased in APMФs compared to control and compared to the situation observed during the resolution phase of infection ( Figure 2D ) . The activity of Ldh increased not only in APMФs but also in RPMФs ( Figure 2E ) . Moreover , the observed increase in Ldh activity was directly correlated with increased lactate production in vivo , as the hemolymph of infected individuals during both the acute and the resolution phases of infection contained significantly elevated lactate levels as compared to controls ( Figure 2G ) . Overall , our results demonstrate that Drosophila macrophages respond to S . pneumoniae infection by upregulating lactate production . The primary reason why cells produce lactate as a byproduct of AG is to maintain NAD+/NADH redox balance . High levels of glycolytic flux produce excess NADH as a result of glyceraldehyde-3-phosphate dehydrogenase 1 ( Gapdh1 ) activity ( Olenchock et al . , 2017 ) . Consistently , we observed that NADH levels were significantly increased in APMФs and , to a lesser extent , in RPMФs when compared with controls ( Figure 2F ) . When considered in the context of gene expression and enzyme activity assays , these results support a model in which activated Drosophila macrophages undergo a dramatic metabolic remodeling towards AG during bacterial infection . Since Hif1α can induce AG in both murine and Drosophila cells ( Peyssonnaux et al . , 2007; Herranz and Cohen , 2017; Eichenlaub et al . , 2018 ) , we examined the possibility that this transcription factor also promotes glucose catabolism within activated macrophages . Although Hif1α is known to be expressed continuously in almost all tissues and regulated predominantly at the post-translational level , we observed that Hif1α mRNA was significantly elevated in APMФs ( Figure 3D ) . To determine whether this increase correlates with the elevated expression of Hif1α target genes , we used a transgenic β-galactosidase reporter under the control of a HRE ( HRE-LacZ ) , which is primarily induced by HIF1α ( Lavista-Llanos et al . , 2002 ) although the involvement of other transcription factors cannot be entirely excluded . Although some cells exhibited HRE-LacZ expression in uninfected individuals , the number of β-galactosidase-positive macrophages rose dramatically in flies during the acute phase of infection ( Figure 3A ) . These results suggest that Hif1α activity is increased in APMФs and confirms the previously reported expression pattern of glycolytic genes ( see Figure 2—figure supplement 1A–F ) . As increased lactate production is a hallmark of AG , we examined Ldh expression in macrophages using a transgene that expresses a Ldh-mCherry fusion protein from an endogenous Ldh promoter . The expression of Ldh-mCherry in adult flies harboring the HmlGal4 >UAS eGFP reporter revealed that macrophages from uninfected adults expressed Ldh at levels that markedly exceeded the expression of this reporter in other tissues . This perhaps indicates that these cells are primed to generate lactate prior to infection ( Figure 3B ) , as the Ldh-mCherry pattern did not change significantly after infection ( data not shown ) . Ldh expression , however , was significantly upregulated in APMФs ( Figure 3C ) , further supporting our observation that S . pneumoniae induces Ldh activity ( Figure 2E ) , which is in agreement with elevated NADH levels ( Figure 2G ) . The regulation of Ldh expression by Hif1α in activated immune cells was verified by knocking down Hif1α expression in macrophages 24 hr before infection ( Hml >Hif1α[RNAi] ) . This strategy not only reduced Hif1α expression within APMФs ( Figure 4—figure supplement 1G ) , but also led to the loss of the ability to increase Ldh expression in APMФs , indicating that Hif1α is essential for the elevated Ldh activity in APMФs ( Figure 3E ) . To determine whether the observed increase in Hif1α activity is necessary to trigger AG in stimulated macrophages , we used Hml >Hif1α[RNAi] and examined the metabolic consequences . This treatment led to the abrogation of the metabolic changes associated with AG . Following infection , APMФs expressing Hif1α[RNAi] did not accumulate NBDG ( Figure 4A ) , and failed to show increased expression of glycolytic genes ( with the exception of Gpdh1 ) ( Figure 4B ) . Moreover , these Hml >Hif1α[RNAi]-expressing cells exhibited no increase in either Pgi or Ldh enzyme activity and displayed decreased NADH levels when compared with controls ( Figure 4D , E , F ) . These results indicate that Hif1α activity is essential for inducing AG in macrophages during the immune response . As a complement to these cell-specific studies of Hif1α , we also used Hml-Gal4 driving UAS-Ldh[RNAi] ( Hml > Ldh[RNAi] ) to reduce Ldh expression within APMФs . Intriguingly , although this approach successfully reduced Ldh activity in macrophages ( Figure 4G ) , the metabolic consequences were relatively mild . Within APMФs , Hml > Ldh[RNAi] did not disrupt NBDG uptake and Pgi activity remained elevated ( Figure 4C and I ) . Twenty-four hours after infection , however , we observed that NADH in Hml > Ldh[RNAi] macrophages failed to increase to the levels observed in infected controls ( Figure 4H ) , thus revealing that increased Ldh activity is required for full metabolic reprograming of Drosophila macrophages in response to bacterial infection . As we have shown previously ( Bajgar and Dolezal , 2018 ) , the systemic metabolic adaptation of carbohydrate metabolism is intimately linked to the effective function of the immune system during streptococcal infection . Therefore , we focused on the characterization of systemic carbohydrate metabolism during the acute phase of infection in Hml >Hif1α[RNAi] and Hml >Ldh[RNAi] flies ( Figure 5 ) . Both control genotypes underwent the expected metabolic response during the acute phase of streptococcal infection: a significantly raised level of circulating glucose was accompanied by a strong depletion of glycogen stores in tissues . The Hif1α silencing completely suppressed the infection-induced changes in carbohydrate metabolism , but infected Hml >Ldh[RNAi] flies still significantly increased circulating glucose , albeit to a lesser extent than in the infected controls ( Figure 5A ) . Although the glycogen stores appeared to be lowered in Hml >Ldh[RNAi] flies upon infection , the decrease was statistically insignificant ( Figure 5B ) . Importantly , the macrophage-specific knockdown of either Hif1α or Ldh suppressed the occurrence of an infection-induced increase in circulating lactate titer ( Figure 5C ) . These results show that APMФs are prominent lactate producers during the acute phase of the infection , and suggest that only full activation of APMФs with Hif1α-induced metabolic changes leads to reprograming of systemic carbohydrate metabolism . Our results suggest that Drosophila macrophages activate AG and systemic metabolic changes in order to mount a successful immune response . In support of this hypothesis , we observed a significant decrease in the viability of adult flies expressing either Hml >Hif1α[RNAi] or Hml >Ldh[RNAi] following S . pneumoniae infection . By 72 hr post infection , 25% of Hml >Hif1α[RNAi] flies died compared to 7% of controls , and the medium time to death ( MTD ) in Hml >Hif1α[RNAi] flies was 10 days compared to 23 days in controls ( Figure 6A ) . Moreover , pathogen load in Hml >Hif1α[RNAi] flies was substantially elevated when compared with that in controls at the second and third day post-infection ( Figure 6C ) . We observed similar effects in Hml >Ldh[RNAi] flies , in which S . pneumoniae infection resulted in a decreased survival rate , a MTD of 9 days relative to the 18 days observed in controls , and elevated bacterial load during days 2 and 3 post-infection ( Figure 6B and D ) . These results reveal that Hif1α and Ldh serve essential roles in both survival of infection and bacterial killing , and demonstrate how shift towards AG associated with systemic metabolic changes in activated macrophages is required to mount a successful immune response .
Mammalian macrophages stimulated by bacteria have been shown to rewire their metabolism temporarily towards AG in order to develop an adequate bactericidal response ( Olenchock et al . , 2017; Nonnenmacher and Hiller , 2018; Browne et al . , 2013 ) . Although well established in mammals , such metabolic adaptation has not been experimentally tested in insect macrophages to date . We show here that Drosophila macrophages that are activated by bacterial infection undergo a dramatic remodeling of cellular metabolism . We demonstrate that acute-phase macrophages exhibit hallmarks of AG , such as elevated uptake of glucose , increased expression and activity of glycolytic genes , elevation of NADH levels , and preferential LDH-mediated conversion of pyruvate to lactate . Through macrophage-specific gene knockdown , we identified Hif1α to be essential for the induction of increased glycolytic flux as well as for the increased activity of LDH . Both Hif1α and Ldh are necessary for the full development of infection-induced changes in systemic carbohydrate metabolism and for resistance to bacterial infection . A major takeaway of our work is that the cellular response to bacterial infection is an energetically challenging process that imposes significant metabolic demands upon the host . Our findings demonstrate that Drosophila macrophages meet these metabolic demands by inducing AG during the acute phase of S . pneumoniae infection , as evidenced by the increased expression of glycolytic enzyme genes and elevated NADH levels . This increase in LDH enzyme activity in the absence of elevated TCA cycle activity suggests that macrophages preferentially convert pyruvate to lactate and is consistent with the elevated concentration of lactate observed in hemolymph . However , we find that this metabolic adaptation is temporary , as AG is terminated during the resolution phase of infection . This latter observation is important because it reveals that macrophages temporally regulate metabolic flux throughout an infection and because it establishes Drosophila as a powerful model for exploring the molecular mechanisms that control immune cell metabolism . Our findings also extend the similarities between fly and mammalian models of macrophage polarization , as we identified Hif1α and Ldh as being crucial for the establishment and maintenance of AG in acute-phase macrophages . The importance of these factors is demonstrated by the macrophage-specific Hif1α knockdown experiment described above , in which many of the hallmark characteristics of AG , including expression of the Ldh gene , were abolished . This finding highlights the conserved and ancient role for Hif1α in regulating the switch between glycolytic and oxidative metabolism ( Webster , 2003 ) , and suggests that this function evolved as a means of allowing cells to adapt quickly to changing physiological conditions and cell-specific metabolic needs . The role of Hif1α in regulating this switch is of significant interest because , although this transcription factor is classically associated with the response to hypoxia , our study adds to the growing list of examples in which Hif1α remodels cellular metabolism in the context of cell proliferation , activation , and competition , even under normoxic conditions ( Miyazawa and Aulehla , 2018 ) . Moreover , our finding is particularly intriguing in light of the fact that Hif1α also serves a key role in promoting AG in neoplastic tumor cells ( Herranz and Cohen , 2017; Eichenlaub et al . , 2018; Wang et al . , 2016 ) . Therefore , our studies of fly macrophages provide a new in vivo system in which we can study how Hif1α promotes cell activity by modulating central carbon metabolism . While Hif1α drives AG in Drosophila macrophages via transcriptional regulation of target genes , the role of Ldh in these cells is more complicated . Although acute-phase macrophages still consume more glucose upon Ldh knockdown , these cells exhibit significantly lower Pgi activity and NADH levels , and the titer of circulating lactate also drops . Our results suggest that even though Ldh acts only at the last step of AG , its role is essential for full metabolic reprograming and efficient function of immune cells . Drosophila Ldh , like its mammalian ortholog , is responsible for the reduction of pyruvate to lactate , which is linked with the regeneration of NAD+ from NADH . However , this single reaction has an immense impact on cellular metabolism . Both the accumulation of pyruvate and the lack of NAD+ can become limiting in cells with high glycolytic flux ( Olenchock et al . , 2017 ) . In addition , Ldh-dependent removal of cytosolic pyruvate was recently found to be essential to prevent pyruvate entry into mitochondria and a subsequent change of TCA cycle course ( Eichenlaub et al . , 2018; Wang et al . , 2016 ) . Although not addressed in our study , changes in mitochondrial metabolism are also closely associated with AG and should be the focus of future studies of activated Drosophila macrophages . The interconnection between the transcriptional activity of Hif1α and the change of mitochondrial metabolism in Drosophila has recently been elucidated . The activation of several direct targets of Hif1α transcriptional activity leads to an inhibition of the classical course of the TCA cycle ( Wang et al . , 2016; Eichenlaub et al . , 2018 ) . One of the well-understood mechanisms is the prevention of pyruvate entry into the TCA cycle , which is caused by increased kinase activity of pyruvate dehydrogenase kinase 1 ( PDK1 ) . PDK1-mediated phosphorylation of pyruvate dehydrogenase ( PDH ) directly inhibits its enzymatic function , which is essential for pyruvate conversion to acetyl-CoA ( Wang et al . , 2016 ) . This event causes a cytoplasmic accumulation of TCA cycle intermediates and thus promotes a secondary wave of Hif1α stabilization through inhibition of prolyl hydroxylase dehydrogenase ( PHD ) under normoxic conditions ( Koivunen et al . , 2007; Freije et al . , 2012 ) . The change in the TCA cycle is further needed for mitochondrial production of ROS that are transferred to the phagolysosome for bacterial killing ( Forrester et al . , 2018; Williams and O'Neill , 2018 ) . We further demonstrate that both Hif1α and Ldh are crucial not only for full macrophage activation , but also for the bactericidal function of the immune cells , with the rearrangement of macrophage metabolism towards AG being essential for resistance to infection and host survival . An important aspect of AG is the functional dependence of macrophages on sufficient supply of external energy resources , as demonstrated in both mammalian and insect phagocytes ( Bajgar and Dolezal , 2018; Anderson et al . , 1973a; Anderson et al . , 1973b; Langston et al . , 2017; Newsholme et al . , 1986 ) and documented here by increased consumption of glucose . Immune cells therefore generate systemic factors to secure sufficient supply of nutrients by altering the function of other organs and by regulating systemic metabolism ( Bajgar et al . , 2015; Bajgar and Dolezal , 2018 ) . While the identification of specific signaling factors is beyond the scope of this work , there are several candidate molecules in Drosophila that are known to be produced by activated macrophages as a reaction to the metabolic state of the cell . Although it is likely that multiple factors will be involved in this process , we can presume that these factors will reflect the metabolic state of the cells ( e . g . , extracellular adenosine ) , or that they will be linked to the transcriptional program that causes the switch towards AG ( e . g . , Imaginal morphogenesis protein late 2 ( ImpL2 ) ) . In our previous work , we showed that the systemic metabolic switch upon infection depends on extracellular adenosine , which is produced by the activated immune cells ( Bajgar and Dolezal , 2018; Bajgar et al . , 2015 ) . The production of adenosine directly reflects a metabolic state of the cell , such as increased consumption of ATP ( Worku and Newby , 1983 ) , as well as the accelerated occurrence of methylation events ( German et al . , 1983; Wu et al . , 2005 ) . Expression of ImpL2 was shown to be regulated by Hif1α ( Li et al . , 2013 ) , and as ImpL2 was previously identified as a mediator of cancer-induced loss of energy reserves in flies due to its anti-insulin role ( Kwon et al . , 2015; Figueroa-Clarevega and Bilder , 2015 ) , it could represent another link between AG in macrophages and changes in systemic metabolism that ensure sufficient supply of energy resources . Finally , our findings raise an interesting question regarding the links between AG and the ability of immune cells to respond quickly to infection . Recent studies of mammalian macrophage metabolism revealed that AG is essential for the development of innate immune memory - called trained immunity ( Netea et al . , 2016 ) . The mechanism of trained immunity relies on chromatin remodeling by epigenetic factors that enable cells to react with higher efficiency in response to re-infection by a particular pathogen ( Kim et al . , 2019 ) . As many chromatin remodeling enzymes need cofactors ( such as acetyl-CoA , NAD+ , α-KG ) for the remodeling of the epigenetic landscape , their function can be influenced by the metabolic state of the cell . Induction of AG leads to the accumulation of many cofactors that are essential for a proper function of these enzymes ( Kim et al . , 2019; Baardman et al . , 2015 ) . The concept of trained immunity not only is valid for mammals , but is rather present in many invertebrate clades ( where it is called immune priming; Milutinović and Kurtz , 2016; Pham et al . , 2007 ) . Our observation of AG as a characteristic feature of activated Drosophila macrophages thus raises a question of its importance for the development of trained immunity in insects and other invertebrates . Taken together , our findings demonstrate how the molecular mechanisms that control AG induction in Drosophila macrophages exhibit a surprisingly high level of evolutionary conservation between mammals and insects , thus emphasizing that this metabolic switch is essential for survival of infection and hinting at the potential role for AG in the development of immune memory . In conclusion , we have shown that infection-induced systemic changes in carbohydrate metabolism are associated with changes of macrophage cellular metabolism , and that both can be affected by macrophage-specific Hif1α and Ldh knockdown . Our data thus link the metabolic state of macrophages with the systemic metabolic changes . On the basis of our previous research on the selfish nature of the immune system under challenge ( Straub , 2014 ) , we envision that the shift in the cellular metabolism of macrophages leads to the production of signals that alter the systemic metabolism , thereby securing the sufficient energy supply necessary to allow the macrophages to fight the infection . By linking the induction of macrophage polarization with systemic metabolism and systemic outcomes in vivo , our experimental system can aid future research towards better understanding of the immune system and of diseases related to its malfunction .
Flies were raised on a diet containing cornmeal ( 80 g/l ) , agar ( 10 g/l ) , yeast ( 40 g/l ) , saccharose ( 50 g/l ) and 10% methylparaben ( 16 . 7 mL/l ) and were kept in a controlled humidity environment with natural 12 hr/12 hr light/dark periods at 25°C , except for those used in temperature-controlled Gal80 experiments . Flies bearing Gal80 were transferred at 29°C 24 hr prior to infection in order to degrade temperature-sensitive Gal 80 protein . Prior to experiments , flies were kept in plastic vials on 0% glucose diet ( cornmeal 53 . 5 g/l , agar 6 . 2 g/l , yeast 28 . 2 g/l and 10% methylparaben 16 . 7 mL/l ) for 7 days and transferred into fresh vials every second day without CO2 in order to ensure good condition of the food . Infected flies were kept on 0% glucose diet in incubators at 29°C due to the temperature sensitivity of S . pneumoniae . Drosophila Stock Centre in Bloomington provided TRiP control and Ldh[RNAi] flies . Hif1α[RNAi] and KK control flies were obtained from Vienna Drosophila Resource Center . Ldh-mCherry strain was kindly provided by Jason Tennessen , HRE-LacZ by Pablo Wappner and Hml > GFP by Bruno Lemaitre . The w1118 strain has a genetic background based on CantonS . The S . pneumoniae strain EJ1 was stored at −80°C in Tryptic Soy Broth ( TSB ) media containing 16% glycerol . For the experiments , bacteria were streaked onto agar plates containing 3% TSB and 100 μg/mL streptomycin and subsequently incubated at 37°C + 5% CO2 overnight . Single colonies were inoculated into 3 mL of TSB liquid media with 100 μg/mL of streptomycin and 100 , 000 units of catalase and incubated at 37°C + 5% CO2 overnight . Bacterial density was measured after an additional 4 hr so that it reached an approximate 0 . 4 OD600 . Final bacterial cultures were centrifuged and dissolved in phosphate-buffered saline ( PBS ) so the final OD reached A = 2 . 4 . The S . pneumoniae culture was kept on ice prior to injection and during the injection itself . Seven-day-old males ( survival experiments , qPCR assays , measurement of metabolites and enzymatic activity ) or females ( X-gal staining , NBDG assay ) were anaesthetized with CO2 and injected with 50 nL culture containing 20 , 000 S . pneumoniae bacteria or 50 nL of mock buffer ( PBS ) into the ventrolateral side of the abdomen using an Eppendorf Femtojet Microinjector . Sixteen randomly chosen flies per genotype and treatment were anaesthetized with CO2 and individually homogenized in 200 µL PBS using a motorized plastic pestle . Serial dilutions were plated onto TSB agar plates and incubated at 37°C overnight . The number of colonies was counted at 0 , 24 , 48 and 72 hpi . Collected data were compared using Tukey's multiple comparisons test in Graphpad Prism software . Sidak's multiple comparison correction was performed . Injected flies were kept at 29°C in vials with approximately 30 individuals per vial and were transferred onto a fresh food every other day . Dead flies were counted daily . At least three independent experiments were performed and combined into one survival curve created in Graphpad Prism software; the individual experiments showed comparable results . Average number of individuals was more than 500 for each genotype . Data were analyzed by Log-rank and Grehan-Breslow-Wilcoxon tests ( which gave more weight to deaths at early time points ) . GFP-labeled hemocytes were isolated from HmlΔ-Gal4 UAS-eGFP male flies using fluorescence-activated cell sorting ( FACS ) . Approximately 200 flies were anaesthetized with CO2 , washed in PBS and homogenized in 600 μL of PBS using a pestle . Homogenate was sieved through a nylon cell strainer ( ⌀ 40 μm ) . This strainer was then additionally washed with 200 µL of PBS , which was added to the homogenate subsequently . Samples were centrifuged ( 3 min , 6°C , 3500 RPM ) and the supernatant was washed in ice cold PBS after each centrifugation ( 3x ) . Prior to sorting , samples were transferred to polystyrene FACS tubes using a disposable bacterial filter ( ⌀ 50 µm , Sysmex ) and sorted into 100 µL of TRIzol Reagent ( Invitrogen ) using a S3TM Cell Sorter ( BioRad ) . Sorted cells were verified by fluorescent microscopy and by differential interference contrast ( DIC ) . Sorted hemocytes were homogenized using a DEPC-treated pestle and RNA was extracted by TRIzol Reagent ( Invitrogen ) according to the manufacturer's protocol . Superscript III Reverse Transcriptase ( Invitrogen ) and oligo ( dT ) 20 primer was used for reverse transcription . Amounts of mRNA of particular genes were quantified on a CFX 1000 Touch Real-Time Cycler ( Bio-Rad ) using the TP 2x SYBR Master Mix ( Top-Bio ) in three technical replicates with the following conditions: initial denaturation for 3 min at 95°C , then amplification for 15 s at 94°C , 30 s at 54°C , 40 s at 72°C for 40 cycles and melting curve analysis at 65–85°C/step 0 . 5°C . Primer sequences are listed in the Key Resources Table . qPCR data were analyzed with double delta Ct analysis , and expressions or particular genes were normalized to the expression of Ribosomal protein 49 ( Rp49 ) in the same sample . Relative values ( fold change ) to control were compared and are shown in the graphs . Samples for gene expression analysis were collected from three independent experiments . Data were compared with Tukey's multiple comparisons test in Graphpad Prism software . Sidak's multiple comparison correction was performed . HmlΔ-Gal4 UAS-eGFP adults were placed on a cornmeal diet with an added 200 µL of 2-NBDG ( excitation/emission maxima of ~465/540 nm , 5 mg/mL stock ( used 10 , 000x diluted ) , Thermo-Fisher ) , which was soaked into the surface of food , immediately after infection ( flies analyzed at 24 hpi ) or 96 hpi ( flies analyzed at 120 hpi ) . After 1 day , flies were prepared for microscopy ( Olympus IX71 ) . Flies for glucose uptake analysis were collected from three independent experiments . X-gal staining was performed on infected HRE-HRE-CRE-LacZ females . Flies were dipped in 75% EtOH for 1 s in order to make their cuticle non-hydrophobic and dissected in PBS . Fixation was performed with 2 . 5% glutaraldehyde/PBS on a LabRoller rotator for 7 min at room temperature . Adults were then washed three times in PBS . Next , two washings were performed with a PT solution ( 1 mL 10xPBS ( Ambion ) , 100 µL 1M MgCl2 × 6H2O , 300 µL 10% Triton , 8 mL dH2O , 320 µL 0 . 1M K4[Fe ( CN ) 6] , 320 µL 0 . 1 M K3[Fe ( CN ) 6] ) for 10 min . Finally , PT solution with few grains of X-gal ( Sigma ) was added . Samples were placed in a thermoblock at 37°C and occasionally mixed , and the colorimetric reaction was monitored . The reaction was stopped with three PBS washings at the same time for all samples . Samples for HRE activation evaluation were collected from four independent experiments . Five flies were homogenized in 200 µL of PBS and centrifuged ( 3 min , 4°C , 8000 RPM ) for glycogen measurement . For lactate and glucose measurement , hemolymph was isolated from 25 adult males by centrifugation ( 14 , 000 RPM , 5 min ) through a silicagel filter into 50 µL PBS . Half of all samples were used for the quantification of proteins . Samples for glucose , glycogen and lactate measurement were denatured at 75°C for 10 min , whereas samples for protein quantification were stored in −80°C . Glucose was measured using a Glucose ( GO ) Assay ( GAGO-20 ) Kit ( Sigma ) according to the manufacturer’s protocol . Colorimetric reaction was measured at 540 nm . For glycogen quantification , sample was mixed with amyloglucosidase ( Sigma ) and incubated at 37°C for 30 min . A Bicinchoninic Acid Assay ( BCA ) Kit ( Sigma ) was used for protein quantification according to the supplier's protocol and the absorbance was measured at 595 nm . A Lactate Assay Kit ( Sigma ) was used for lactate concentration quantification according to the manufacturer's protocol . The absorbance was measured at 570 nm . Samples for metabolite concentration were collected from six independent experiments . Measured data were compared in Graphpad Prism using Tukey's multiple comparisons test . Sidak's multiple comparison correction was performed . The enzymatic activities of lactate dehydrogenase and phosphoglucose isomerase were measured using a Lacate Dehydrogenase Activity Assay Kit ( Sigma ) or a Phosphoglucose Isomerase Colorimetric Assay Kit ( Sigma ) , respectively , according to the supplier's protocol in 10 , 000 FACS-sorted hemocytes for each sample . Colorimetric reaction was measured at 450 nm . Samples for enzymatic activity detection were collected from six independent experiments . Measured values were compared in Graphpad Prism software using Tukey's multiple comparisons test . Sidak's multiple comparison correction was performed . | Macrophages are the immune system's first line of defense against infection . These immune cells can be found in all tissues and organs , watching for signs of disease-causing agents and targeting them for destruction . Maintaining macrophages costs energy , so to minimize waste , these cells spend most of their lives in 'low power mode' . When macrophages sense harmful bacteria , they rapidly awaken and trigger a series of immune events that protect the body from infection . However , to perform these protective tasks macrophages need a sudden surge in energy . In mammals , activated macrophages get their energy from aerobic glycolysis – a series of chemical reactions normally reserved for low oxygen environments . Switching on this metabolic process requires a protein called hypoxia inducible factor 1α ( HIF-1 α ) , which switches on the genes that macrophages need to generate energy as quickly as possible . Macrophages then maintain their energy supply by sending out chemical signals which divert glucose away from the rest of the body . Fruit flies are regularly used as a model system for studying human disease , as the mechanisms they use to defend themselves from infections are similar to human immune cells . However , it remains unclear whether their macrophages undergo the same metabolic changes during an infection . To address this question , Krejčová et al . isolated macrophages from fruit flies that had been infected with bacteria . Experiments studying the metabolism of these cells revealed that , just like human macrophages , they responded to bacteria by taking in more glucose and generating energy via aerobic glycolysis . The macrophages of these flies were also found to draw in energy from the rest of the body by raising blood sugar levels and depleting stores of glucose . Similar to human macrophages , these metabolic changes depended on HIF1α , and flies without this protein were unable to secure the level of energy needed to effectively fight off the bacteria . These findings suggest that this metabolic switch to aerobic glycolysis is a conserved mechanism that both insects and mammals use to fight off infections . This means in the future fruit flies could be used as a model organism for studying diseases associated with macrophage mis-activation , such as chronic inflammation and autoimmune diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2019 | Drosophila macrophages switch to aerobic glycolysis to mount effective antibacterial defense |
High-income countries are experiencing measles reemergence as the result of suboptimal vaccine uptake and marked immunity gaps among adults . In 2017 , the Italian Government introduced mandatory vaccination at school entry for ten infectious diseases , including measles . However , sustainable and effective vaccination strategies targeting adults are still lacking . We use a data-driven model of household demography to estimate the potential impact on future measles epidemiology of a novel immunization strategy , to be implemented on top of the 2017 regulation , which consists of offering measles vaccine to the parents of children who get vaccinated . Model simulations suggest that the current vaccination efforts in Italy would not be sufficient to interrupt measles transmission before 2045 because of the frequency of susceptible individuals between 17 and 44 years of age . The integration of the current policy with parental vaccination has the potential to reduce susceptible adults by 17–35% , increasing the chance of measles elimination before 2045 up to 78 . 9–96 . 5% .
The Global Measles and Rubella Strategic Plan 2012–2020 set the ambitious goal of eliminating measles in at least five World Health Organization ( WHO ) regions by 2020 . Two years before the deadline , only the Americas have achieved measles elimination . Measles is endemic in 14 countries of the WHO European Region , including high-income countries such as Germany , Belgium , France , and Italy ( World Health Organization Regional Office for Europe , 2016 ) , and it still represents a major concern for public health . In 2017 , Italy experienced one of the largest measles outbreaks of the past decade in the European Region with four deaths and 5098 cases , 4042 of which were confirmed by positive laboratory results ( Italian National Institute of Health , 2017; European Centre for Disease Prevention and Control , 2018 ) . The highest incidence was observed in infants under one year of age . About 70% of the reported cases were older than 20 years , with a median age of 27 years ( Italian National Institute of Health , 2017; Filia et al . , 2017 ) , suggesting that measles circulation in Italy is at least partially supported by transmission between adults . Significant immunity gaps in these age segments of the population have been highlighted by a serological screening of the population ( Rota et al . , 2008 ) and by recent modeling studies analyzing long-term processes that affect measles transmission dynamics in the Italian population ( Merler and Ajelli , 2014; Trentini et al . , 2017 ) . The high fraction of measles-susceptible individuals of between 15 and 45 years of age is the result of past suboptimal routine vaccination coverage and the absence of major nationwide epidemics in recent decades , which allowed adolescents to escape both vaccination and natural infection ( Filia et al . , 2017; Trentini et al . , 2017 ) . In Italy , the first measles national immunization program was setup in 1983 with a single dose of measles vaccine being administered at 9 months of age . A second dose program was introduced in 1999 . However , routine vaccination coverage remained below 80% until 2003 , the year of approval of the Italian National Plan for the elimination of Measles and Congenital Rubella . Thereafter , vaccine uptake levels have progressively increased , even though a decrease in coverage has been detected in most recent years , possibly associated with vaccine hesitancy ( Filia et al . , 2017; Merler and Ajelli , 2014; Giambi et al . , 2018 ) . As a matter of fact , the national coverage reached a peak of 91% in 2010 , which is well below the 95% threshold generally considered to be necessary for measles elimination ( Anderson and May , 1991 ) . In July 2017 , the Italian Government approved a regulation ( 119/2017 ) requiring parents to vaccinate their children before school entry against ten infectious diseases , including measles ( Signorelli et al . , 2018; D’Ancona et al . , 2018; Italian Ministry of Health , 2017 ) . Vaccination against measles is now free of charge and mandatory for all children under 16 years . Unvaccinated children are not allowed to attend kindergartens , and financial penalties are imposed on the parents of unvaccinated students attending higher school levels . This regulation has the potential to increase vaccine uptake in new birth cohorts and to immunize school-age children who have escaped routine vaccination ( Trentini et al . , 2019 ) . However , the new policy will not impact the existing immunity gaps in older age groups . In particular , the achievement and maintenance of high vaccination coverage among children may not be enough to avoid the reemergence of measles in the future ( Trentini et al . , 2017; Trentini et al . , 2019; Durrheim , 2017 ) . In order to progress towards measles elimination , it is thus crucial for Italy to identify feasible , sustainable , and effective strategies to reduce the number of susceptible individuals among those who have already left the school system ( Filia et al . , 2017; Trentini et al . , 2017; Durrheim , 2017; Thompson , 2017 ) . The aim of this work is to propose and investigate the effectiveness of a vaccination strategy to be introduced on top of the current policy . The proposed strategy consists of offering vaccination to the parents of all of the children who receive any measles vaccine dose .
We perform different sensitivity analyses to assess the robustness of the obtained estimates when considering: In Appendix 1 , we also report the results obtained when measles epidemiology is simulated by considering the vaccination strategy adopted in Italy before the introduction of mandatory vaccination at school entry in July 2017 , and present a sensitivity analysis to assess the robustness of the estimates of the exponential growth rate r associated with the 2017 epidemic when including possible underreporting of cases ( Ciofi Degli Atti et al . , 2002 ) .
The performed analysis shows that an improvement of vaccine uptake at school entry to achieve 99% coverage in the current program may anticipate the timing of measles elimination to 2039 . If vaccine uptake at school entry were to be 75% , measles elimination could be achieved in 2042 , which is comparable to what might be obtained by reaching 50% of eligible families with parental vaccination in the baseline analysis . By contrast , our simulations show that , under the most optimistic scenario of 99% of coverage both for parental vaccination and vaccination at school entry measles elimination could be achieved , on average , as early as 2023 . The assumption of a shorter or longer generation time would affect model estimates of the effective reproduction number over time . In particular , under parental vaccination at 50% of coverage , a shorter ( longer ) generation time would result in an anticipation ( delay ) of the timing of measles elimination , which is estimated to occur before 2045 in 99 . 2% ( 16 . 8% ) of model realizations . When a generation time lasting 18 days is considered , the current policy at current coverage levels was insufficient to achieve measles elimination by 2045 in 99 . 9% of model realizations . The obtained estimates are qualitatively robust when considering alternative age-specific mixing patterns for the Italian population , although the inclusion of contact matrices estimated through the modeling approach ( Fumanelli et al . , 2012 ) results in delayed measles elimination under all considered vaccination scenarios . On the other hand , under the ( hardly realistic ) scenario of a population that mixes fully at random ( i . e . , by assuming homogeneous mixing ) , neither the current policy nor its combination with parental vaccination would be sufficient to achieve measles elimination by 2045 . Qualitative temporal patterns in the evolution of the effective reproduction number estimated by exploring different levels of measles transmissibility during the prodromal and exanthema phase are generally robust . The largest quantitative difference can be detected when most secondary cases are generated in the prodromal phase . In this case , under the current policy , measles elimination is predicted to occur before 2045 in 73 . 9% of model realizations instead of the 12 . 0% of model realizations seen for the baseline analysis . Similarly , under parental vaccination at 50% of coverage , when most of secondary cases are generated in the prodromal phase , measles elimination is predicted to occur before 2045 in 98 . 7% of model realizations compared to the 78 . 9% of model realizations seen in the baseline analysis . Finally , when considering an extreme scenario in which only 25% of measles cases were reported during the 2017 outbreak , we estimate the exponential growth rate to be 0 . 29 ( 95% CI: 0 . 21–0 . 37 ) , similar to that obtained when only reported cases are used: 0 . 29 ( 95% CI: 0 . 25–0 . 33 ) . As estimates of the effective reproduction number depend only on the growth rate and on measles natural history , these results suggest that our findings are robust with respect to the reporting rate ( and size ) of the 2017 measles outbreak . Details on the performed sensitivity analyses are reported and discussed in Appendix 1 .
In July 2017 , the Italian Government approved a regulation requiring parents to vaccinate their children before school entry against ten infections , including measles . Recent estimates suggest that the new regulation allowed the vaccination of 50% of individuals who escaped routine vaccination ( Italian Ministry of Health , 2019; Italian National Institute of Health , 2019 ) . Our modeling study shows that the current policy would reduce measles susceptibility in the age segments of the population characterized by higher contact rates , resulting in a remarkable decrease in the infection transmission potential and making measles elimination a realistic target . However , if only 50% of unvaccinated children are vaccinated at school entry , disease elimination would probably be achieved only after 2045 . Offering vaccination to the parents of children who receive a measles vaccine dose could progressively reduce by 17–35% the immunity gaps among individuals who are between 18 and 45 years of age in 2018 . The implementation of this program would decrease the overall susceptibility of the population by 6 . 2–22 . 0% , and would increase the probability of measles elimination before 2045 from 12 . 0% ( estimated in the absence of this additional policy ) to 78 . 9–96 . 5% . The effectiveness of this strategy clearly depends on both the coverage achieved through childhood immunization ( routine programs and vaccination at school entry ) and on the willingness of parents to be vaccinated themselves . The obtained estimates are generally robust with respect to different assumptions on the duration of measles generation time and on the relative transmissibility of measles during the prodromal and exanthema phases . On the other hand , under the assumption of homogeneous mixing in the population , neither the current immunization program nor parental vaccination appear to be sufficient to eliminate measles before 2045 . The study presents a few limitations that should be carefully considered in order to achieve a better interpretation of the obtained results . In particular , our estimates of the effective reproduction number were obtained using measles cases reported during the 2017 outbreak . The current degree of measles underreporting in statutory notifications is unknown . However , our estimates of the effective reproduction number are stable with respect to the possible underreporting of cases during the outbreak ( Ciofi Degli Atti et al . , 2002 ) . The proposed analysis did not take into account potential geographical heterogeneities in measles immunity levels at the sub-national scale . Although the new regulation is expected to harmonize the vaccine offer and its uptake in Italy , significant regional differences in both immunization schedule and coverage have been reported in the past ( Bonanni et al . , 2015 ) . Regions characterized by a lower than national average vaccine uptake in the past may therefore experience a delay in measles elimination with respect to the results presented in this work . In our work , future measles susceptibility might have been overestimated , as we did not explicitly model measles transmission , thus disregarding the impact of future measles spread on the immunity profile of the Italian population . Although both the occurrence and magnitude of future measles epidemics are largely uncertain and difficult to predict ( Earn et al . , 2000 ) , changing patterns of measles transmission may affect both the number of susceptible adults and the incidence of severe disease in the coming years . However , the population infected during the 2017 Italian outbreak—one of the largest occurred in Europe in the last years—represented only 0 . 1% of the estimated susceptible population in the country ( Italian National Institute of Health , 2017; Trentini et al . , 2017 ) . This suggests that the explicit inclusion of measles transmission may have a limited impact on short- or medium-term estimates of the immunity profile and measles transmission potential . The proposed analysis relies on the simplifying assumption that parents decide whether to vaccinate their children regardless of past vaccination behavior , although it is likely that parents vaccinate either all or none of their children . All children receiving vaccination indirectly present their parents with the opportunity to vaccinate themselves . As the children receiving vaccination may be clustered in a smaller number of households than is the case in our model , we are probably overestimating the potential number of parents who are eligible for measles vaccination . In particular , in our simulations , 98 . 7% of families with children between 1 and 15 years of age are considered as eligible for parental vaccination , whereas in a perfectly clustered model , this percentage would be 88 . 1% . On the other hand , clustering of unvaccinated children may have a larger effect on measles transmission dynamics than on the number of parents eligible for vaccination . Finally , we assumed that routine vaccination coverage would not be affected by the implementation of the new national policy , and that the coverage of the catch-up campaign conducted in 2017–2018 was the same as that of vaccination at school entry ( both for pre-primary and for primary schools ) . However , data released in December 2018 by the Italian Ministry of Health suggest that the new regulation on mandatory vaccination at school entry may have indirectly affected the first-dose vaccine uptake for children under 3 years of age . In particular , the available records show that the first-dose vaccination coverage in the 2015 age cohort has increased from 91 . 4% in 2017 to 94 . 2% in 2018 , ( Italian Ministry of Health , 2019 ) although the fraction of unvaccinated children who were vaccinated thanks to the new regulation may vary depending on the age cohort considered ( D’Ancona et al . , 2018; Italian Ministry of Health , 2019 ) . According to the most recent estimates , measles vaccination coverage in the 2014 age cohort has increased from 87 . 3% in 2016 to 94 . 4% in 2018 , suggesting that the new regulation resulted in the vaccination of about 56% of unvaccinated children in this cohort ( Italian National Institute of Health , 2019 ) . In conclusion , our analysis shows that a marked increase in childhood immunization rates would not be sufficient to achieve measles elimination in the short- or medium-term in Italy . These results confirm the need for appropriate strategies to vaccinate individuals who have already left the school system in order to reduce critical immunity gaps in young adults ( Trentini et al . , 2019; Filia et al . , 2017; Trentini et al . , 2017; Durrheim , 2017; Thompson , 2017; Wise , 2018; Gidding et al . , 2007 ) . Attempts made to date in this direction either have only been partially effective or have required remarkable efforts in terms of the costs to and commitment of the public health authorities ( Gidding et al . , 2007; Morice et al . , 2003; Kelly et al . , 2007 ) . In Costa Rica , a measles-rubella vaccination campaign targeting adults aged 15–39 years was successfully conducted in 2000 , but it required huge efforts of communication , social mobilization , and the use of house-to-house vaccination teams ( Morice et al . , 2003 ) . In 2001–2002 , a vaccination campaign in Australia targeting young adults aged between 18 and 30 years who visited their general practitioner ( GP ) had little effect on the immunity gaps , probably because of a lack of promotion and central coordination ( Gidding et al . , 2007; Kelly et al . , 2007 ) . In Europe , beyond some local attempts to immunize adolescents and individuals before school leaving , which have only marginally affected the vaccine uptake ( Lashkari and El Bashir , 2010; Vazzoler et al . , 2014 ) , little has been done to reduce residual susceptibility in adults . Interventions recently set up include attempts to raise awareness in people attending social events that may represent potential hotspots for measles transmission ( Public Health England , 2018 ) . In this work , a new strategy is proposed , consisting of offering vaccination to parents of children who are being vaccinated against measles . Although the proposed policy can reach only a fraction of susceptible adults , that is those with children in the measles-vaccination age group , the obtained results suggest that this strategy may be both feasible and effective . In particular , our results suggest that vaccinating 50% of parents who agreed to vaccinate their children , and may therefore be inclined towards accepting vaccination , would promote measles elimination as well as reaching 50% of children who still escape measles vaccination despite the fact that vaccination is now mandatory in Italy ( i . e . , increasing vaccination coverage at school entry from 50% to 75% ) . The sustainability of the proposed strategy should be carefully evaluated by public health decision makers . However , a key advantage of this policy is that it does not require targeted activities to recruit parents , thus resulting in a relatively simple implementation protocol . Beyond parental vaccination , alternative immunization strategies aimed at reducing residual susceptibility in adults may also be considered . These may include the extension of mandatory vaccination at university entry – an intervention already implemented in different US states . Other immunization efforts may include the introduction of proof of immunity as a condition for the enrolment of health care workers ( HCWs ) , for whom measles vaccination is only recommended in most European countries ( Galanakis et al . , 2014; Maltezou et al . , 2019 ) . The need to improve vaccination coverage among HCWs is due to their potential to amplify measles outbreaks and their higher risk of exposure to the virus , as observed in the 2017 Italian outbreak in which 7% of cases were HCWs ( Maltezou et al . , 2019 ) . The achievement of measles elimination remains a global health priority . Actions may be also required to raise awareness and consensus about the benefits coming from vaccination and to increase the overall vaccine uptake . Country-specific policies should be identified and carefully evaluated by decision makers in order to anticipate the time of measles elimination as much as possible . | Measles is one of the world’s most contagious diseases causing thousands of deaths every year , despite a safe and effective vaccine being available since the 1960s . High rates of vaccination – about 95% of each age group – are required to eliminate measles , but national and global health agencies struggle to achieve high vaccination rates because some parents were and still are hesitant to vaccinate their children . As a result , large measles epidemics continue to occur even in countries with well-established vaccination programs . In Italy , low vaccination rates year after year have resulted in large numbers of unprotected youth and adults . The country has recently introduced mandatory measles vaccination at school entry to improve vaccination coverage among children . Yet a high proportion of measles cases in Italy continue to occur in people over 20 years old , a situation that could be improved by immunization programs targeting adults . One approach would be to take advantage of the compulsory vaccination of children by offering parents the vaccine at the same time . Marziano et al . used computer modeling to estimate how various vaccination scenarios would affect measles spread in Italy . Their models showed that current vaccination policies targeting school age children would be unlikely to eliminate measles before 2045 . On the other hand , if 50% of parents were also vaccinated , elimination could be achieved by 2042 , and as early as 2031 if 99% of parents agreed to vaccination . Marziano et al . show that a parental vaccination campaign could reduce the population of adults susceptible to measles in Italy and help the country stop the spread of the disease . However , more research is needed to assess how feasible and sustainable this policy would be . Additional policies to increase vaccination against measles in adults could also help , but parental vaccination has a key advantage: it does not require active targeting to recruit parents , since they are already immunizing their children . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2019 | Parental vaccination to reduce measles immunity gaps in Italy |
The identification of clinically viable strategies for overcoming resistance to platinum chemotherapy in lung adenocarcinoma has previously been hampered by inappropriately tailored in vitro assays of drug response . Therefore , using a pulse model that closely mimics the in vivo pharmacokinetics of platinum therapy , we profiled cisplatin-induced signalling , DNA-damage and apoptotic responses across a panel of human lung adenocarcinoma cell lines . By coupling this data to real-time , single-cell imaging of cell cycle and apoptosis we provide a fine-grained stratification of response , where a P70S6K-mediated signalling axis promotes resistance on a TP53 wildtype or null background , but not a mutant TP53 background . This finding highlights the value of in vitro models that match the physiological pharmacokinetics of drug exposure . Furthermore , it also demonstrates the importance of a mechanistic understanding of the interplay between somatic mutations and the signalling networks that govern drug response for the implementation of any consistently effective , patient-specific therapy .
Lung adenocarcinoma is the most common form of lung cancer , the leading cause of cancer-related death worldwide . Lung adenocarcinoma is typically diagnosed late , meaning that most patients require systemic chemotherapy ( Chen et al . , 2014 ) . Platinum-based chemotherapy is likely to remain an important treatment modality for these patients due to the emergence of resistance to targeted therapies in EGFR , ALK or ROS mutant tumours ( Lindeman et al . , 2018 ) , and the fact that most patients do not respond to single agent immunotherapy ( Kim et al . , 2019; Doroshow et al . , 2019 ) . Despite the use of platinum-based chemotherapy in lung adenocarcinoma for over four decades , response rates remain below 30% due to the prevalence of innate resistance ( Pilkington et al . , 2015; Bonanno et al . , 2014 ) . In addition , dose-related nephrotoxicity remains a challenge in many patients ( Pabla and Dong , 2008 ) . Strategies to improve platinum efficacy could therefore significantly improve outcomes for lung adenocarcinoma patients . However , unravelling platinum resistance in lung adenocarcinoma has proven challenging , as over 147 mechanisms of resistance have been proposed ( Stewart , 2007 ) , yet there remains a lack of viable clinical options to improve response rates . From an experimental viewpoint , discordance between the in vivo pharmacokinetics of platinum chemotherapies and their use within in vitro assays has likely contributed to the identification of putative resistance mechanisms and drug targets that have not ultimately translated to the clinic . Traditionally , in vitro methods for the investigation of drug response have involved culturing cancer cells in the continuous presence of high-dose chemotherapy over several days . This contrasts with pharmacokinetic studies in humans and rodents demonstrating that both cisplatin and carboplatin are rapidly cleared from the circulation , and the tumour , within 2–3 hr following administration ( Andersson et al . , 1996; Johansen et al . , 2002 ) . Therefore , in this study we have utilised a cisplatin pulse model , which more closely recapitulates these physiological pharmacokinetics , aiming to maintain the fidelity of the apoptotic mechanism mediated by cisplatin in vivo . We have previously shown that predictive computational models of drug-induced apoptotic signalling dynamics can be used as a prognostic indicator of neuroblastoma patient survival ( Fey et al . , 2015 ) . To move towards a similar concept to platinum resistance in lung adenocarcinoma , we now present an in-depth analysis of the dynamic signalling response to a pulse of platinum chemotherapy , describing the relationship between a number of key signalling nodes , the DNA damage response and platinum sensitivity . Importantly , we also propose a therapeutic strategy targeting P70S6K using the dual PI3K/mTOR inhibitor dactolisib , with the potential to improve the efficacy of current platinum-based treatment regimens .
In order to directly compare the response of lung adenocarcinoma cells to the continuous presence of cisplatin , or a pulse of cisplatin that mimics in vivo pharmacokinetics ( 2 hr , 5 μg/mL ) ( Figure 1—figure supplement 1A ) , we monitored the growth and apoptosis of the innately resistant A549 lung adenocarcinoma cell line ( Marini et al . , 2018 ) by both live cell imaging ( Figure 1—figure supplement 1B ) and a cell viability assay ( Figure 1—figure supplement 1C ) , under both conditions . This analysis demonstrated that while continuous exposure to cisplatin resulted in decreased cell number and increased apoptosis over 72 hr , a pulse of cisplatin only reduced the rate of cell proliferation and did not induce apoptosis . To further examine the differences between these two models , we used multiplexed , bead-based protein analysis to investigate the DNA damage , apoptotic and signalling response for key pathway components previously implicated in the response to continuous cisplatin exposure ( Stewart , 2007; Marini et al . , 2018; Supplementary file 1 , Figure 1—figure supplement 1D ) . As might be expected , the continuous exposure model resulted in a significantly elevated and sustained DNA damage response when compared to the pulse model , particularly for the phosphorylation of Chk2 ( Ser345 ) , p53 ( Ser15 and Ser46 ) , pH2A . X ( Ser139 - γH2A . X ) and expression of p21 and MDM2 ( Figure 1—figure supplement 2 ) . This heightened DNA damage response during the continuous exposure to cisplatin was also reflected in the increased activation of Caspase 3 , which was completely absent for the pulse model . Furthermore , while p38 and ERK activation were significantly increased in cells continuously exposed to cisplatin , the expression of MCL-1 and detection of MCL-1/Bak dimers only significantly increased in cells treated with a pulse of cisplatin . This finding demonstrates that not only does the continuous exposure model result in a DNA damage and apoptotic response that is incongruent with that observed following a pulse of cisplatin , the dynamics of key signalling pathways are fundamentally different between these two treatment models . Taken together , this data demonstrates that previous mechanisms of platinum resistance established using a continuous exposure model should be reconsidered in the light of these new findings regarding the response to physiological levels of drug exposure . To investigate potential mechanisms of platinum resistance using a model consistent with the physiological pharmacokinetics of platinum therapy ( Figure 1A ) , we applied this cisplatin pulse model to a panel of six lung adenocarcinoma cell lines with distinct TP53 mutation backgrounds ( two wildtype lines , two mutant TP53 lines and two TP53 null ) and measured the apoptotic response at 72 hr ( Figure 1B ) . Based upon this model we observed a range of sensitivity to cisplatin , from the most resistant A549 line ( ~3% apoptosis ) to the most responsive NCI-H1299 line ( ~32% apoptosis ) . However , these cell lines could not be stratified simply according to their TP53 mutation status , or other frequently observed genetic alterations ( Supplementary file 2 ) . As the action of drug-efflux pumps is another commonly proposed mechanism of resistance to platinum therapy ( Hoffmann and Lambert , 2014 ) , we performed fluorescence microscopy with an antibody towards cisplatin-induced DNA adducts at multiple time-points following a 2 hr cisplatin pulse ( Figure 1C ) . This analysis demonstrated that within this model , all six cell lines displayed significant nuclear localised cisplatin-DNA adducts following a 2 hr pulse ( Figure 1C , Figure 1—figure supplement 3 ) , suggesting that drug efflux is not associated with variations in the apoptotic response to a pulse of cisplatin in these lines . Furthermore , these cisplatin-DNA adducts progressively resolved over a 72 hr period in all cell lines ( Figure 1C ) , confirming that pathways responsible for facilitating the removal of cisplatin adducts are also functional across this panel . To gain an understanding of the wider signalling networks associated with resistance to platinum therapy , we utilised multiplexed , magnetic bead based assays to profile the signalling and DNA damage response at multiple time points following a 2 hr pulse of cisplatin , across all six cell lines ( Figure 1D ) . For this analysis , we tracked the dynamics of 47 different protein analytes ( Supplementary file 3 ) over a 72 hr period , focusing on elements of signalling network structures that we recently implicated in platinum chemoresistance ( Marini et al . , 2018 ) , including the MAPK , PI3K/mTOR , NF-κB and TGFβ pathways , as well as a number of key apoptotic mediators and DNA damage response proteins ( Figure 1D ) . From this dataset , a clear correlation can be seen between the TP53 mutation status of each cell line and the dynamics of the p53 pathway ( Figure 1—figure supplement 4 ) , which validates the fidelity of the multiplexed platform for this type of analysis . In line with the detection of cisplatin-DNA adducts ( Figure 1C ) , Chk1 ( Ser345 ) phosphorylation increased rapidly in all lines following the 2 hr pulse , followed by a slower wave of Chk2 ( Thr68 ) phosphorylation in all lines except SW-1573 . Within both the TP53 wildtype cell lines ( A549 and SW-1573 ) this was followed by phosphorylation of p53 ( Ser15 and Ser46 ) , the accumulation of total p53 and increased expression of the p53 transcriptional targets MDM2 and p21 . In the TP53 mutant cell lines ( NCI-H1573 and NCI-H1975 ) p53 phosphorylation still occurred , however in line with their loss of DNA binding capability , this did not result in increased expression of MDM2 or p21 . As would be expected for the TP53 null lines ( NCI-H358 and NCI-H1299 ) , p53 is absent and therefore not detected in either the total or phosphorylated form . Interestingly though , there was a significant increase in p21 and MDM2 expression in the NCI-H358 line in the absence of p53 expression . As we had already determined that TP53 status alone was not sufficient to explain resistance to cisplatin ( Figure 1B ) , we further analysed the whole dataset by performing a principal component analysis ( PCA ) ( Figure 2 ) . This form of dimensionality reduction can be used to identify correlative relationships between variables within a large dataset , and here we have used it to create a visual representation of the association between key signalling nodes and the response to cisplatin across the entire cell line panel . Using this multi-dimensional analysis , we were able to capture ~70% of variance in the dataset within the first four principal components ( Figure 2—figure supplement 1A ) . Unsurprisingly , plotting the first two principal components ( PC1 and PC2 ) against each other ( Figure 2A ) resulted in separation of the cell lines primarily according to their TP53 status . As might be expected , within PC1 and PC2 the TP53 wild-type A549 and SW-1573 lines associated with higher p21 and MDM2 expression , while the TP53 mutant lines separated from the TP53 null lines mostly on the basis of higher p53 expression and phosphorylation levels ( Figure 2A , B , C ) . However , plotting the third and fourth principal components ( PC3 and PC4 , Figure 2D , E ) created a clear delineation between the three most resistant cell lines ( A549 , NCI-H358 and NCI-H1573 ) , which cluster towards the left hand side of PC3 ( x-axis ) , and the three most sensitive lines ( SW-1573 , NCI-1975 and NCI-H1299 ) which move progressively along PC3 over the 72 hr timeframe . As platinum chemotherapies work by forming covalent DNA adducts , which distort the DNA helix and block replication , the progressive accumulation of single stranded and double stranded breaks is thought to induce apoptosis ( Jamieson and Lippard , 1999 ) . This is in line with the movement of cisplatin sensitive cell lines along PC3 ( x-axis ) towards higher levels of γH2A . X ( H2A . XS139 ) , cleaved caspase 3 and a stress associated MAPK signalling axis ( pATF2 , pJNK , pc-Jun ) ( Figure 2D , E ) . The association between this signalling state and increased sensitivity to cisplatin can also be clearly observed by overlaying an orthogonal readout of the apoptotic response onto this PCA plot ( Figure 3A ) . This real-time apoptosis data was generated using live-cell imaging with a fluorescent caspase substrate as an indicator of cell death across the cell line panel for 72 hr following the cisplatin pulse treatment . Using this approach , we now created a visual representation that both reflects the variance within the original dataset and demonstrates the key signalling nodes that are associated with differing degrees of platinum-induced apoptosis . While increasing levels of apoptosis are observed over time in the three sensitive cell lines ( SW-1573 , NCI-1975 and NCI-H1299 ) , the three resistant cell lines display significantly lower levels of apoptosis ( Figure 2—figure supplement 1B ) and do not move along PC3 towards the region of DNA damage and apoptosis . Instead the NCI-H358 and NCI-1573 lines remain towards the left hand side of PC3 , in a region characterised by higher phosphorylation of the mTOR pathway component P70S6K ( Thr389 ) , the MAPK components MEK1 ( Ser217/221 ) and P90RSK ( S380 ) , NF-κB ( S536 ) and IκBα ( Ser32/36 ) ( Figure 2D , E ) . The resistant A549 line also remains shifted towards the left of PC3 , although also moves up PC4 towards a region with higher expression of TGFβRII , cleaved caspase 9 , pAkt ( Ser473 ) and pSTAT3 ( Ser727 ) ( Figure 2D , E ) . The association of elevated pMEK ( Ser217/221 ) , pNF-κB ( S536 ) , pSTAT3 ( Ser727 ) and pP70S6K ( Thr389 ) within these resistant cell lines was also further confirmed by western blotting of independent samples ( Figure 2C ) . The clear separation of resistant and sensitive cell lines within this multi-dimensional analysis suggests an antagonistic relationship between apoptosis promoted through the progressive accumulation of DNA-damage following a cisplatin pulse , and elevated mTOR , MAPK , Akt , STAT or NF- κB signalling events . However , as a statistical process , relationships derived from a principal component analysis are purely correlative in nature and require a further degree of validation before any causative conclusions may be drawn . We therefore sought to identify effectors of platinum resistance by including specific inhibitors of these signalling pathway components during and after the 2 hr cisplatin pulse , followed by a measurement of apoptosis at 72 hr ( Figure 3B ) . Using the resistant A549 ( Figure 3C ) and NCI-H358 ( Figure 3D ) cell lines we observed that specific inhibitors of Akt ( MK2206 ) , STAT3 ( S3I-201 ) , MEK ( UO126 ) or NF- κB ( SC-75741 ) did not significantly increase apoptosis in either cell line , suggesting that while elevated activity of these signalling proteins may be present in one or both of these resistant lines ( Figure 2C ) , they are not causally associated with resistance to cisplatin . Instead , under these conditions , only the inhibition of P70S6K with the dual PI3K/mTOR inhibitor dactolisib resulted in a significant increase in cisplatin-induced apoptosis in both cell lines . P70S6K is a serine/threonine-specific protein kinase known to require phosphorylation by both PI3K and mTOR for activation ( Sunami et al . , 2010; Moser et al . , 1997 ) . While dactolisib will result in the inhibition of a number of substrates downstream of both PI3K and mTOR , the lack of sensitisation by an Akt inhibitor ( MK2206 ) suggests a specific role for P70S6K in mediating resistance to cisplatin . Taking the opposite approach , elevated levels of JNK pathway activity ( pJNK , pc-Jun ) were also observed following cisplatin treatment within the SW-1573 line ( Figure 2C , D , E ) . However , the inclusion of a JNK inhibitor ( JNK inhibitor VIII ) did not prevent cisplatin induced apoptosis in this cell line ( Figure 2—figure supplement 1C ) , suggesting that JNK activity was not promoting apoptosis in this context . Comparing the relative expression levels of phosphorylated P70S6K across our stratified panel of cell lines , higher expression was observed in the resistant NCI-H358 and NCI-H1573 lines ( Figure 2C ) . This expression pattern was also mirrored by the levels of total P70S6K ( Figure 2C ) , which seems to be primarily driving the observed levels of P70S6K phosphorylation . Interestingly , the resistant A549 line did not have elevated P70S6K expression or phosphorylation , although as reflected in the PCA plot ( Figure 2E ) this cell line did have high expression of the P70S6K substrate , Ribosomal Protein S6 , leading to a level of phosphorylation equivalent to that observed in the resistant NCI-H358 line ( Figure 2C ) . Utilising the highest expressing ( NCI-H358 ) and lowest expressing ( NCI-H1299 ) lines , which are also both TP53 null , we observed that cisplatin treatment resulted in greater caspase 3 cleavage and γH2A . X expression in the sensitive NCI-H1299 line ( Figure 3E ) , which is in line with the multiplexed signalling analysis ( Figure 1D ) . Crucially , treatment of the resistant NCI-H358 line with dactolisib during and after the cisplatin pulse increased both caspase 3 cleavage and γH2A . X expression to that observed in the sensitive NCI-H1299 line . Conceptually , this now mimics the movement of the resistant NCI-H358 line along PC3 of our PCA analysis , towards the region of DNA damage and apoptosis characterised by the sensitive cell lines ( Figures 2 and 3A ) . This finding demonstrates that while this form of multi-dimensional analysis of signalling networks creates a set of correlative relationships , this data can also be utilised to investigate causal effectors of downstream cellular behaviour . As mentioned above , dactolisib can efficiently inhibit the phosphorylation of P70S6K , but will also result in the inhibition of a number of other potential PI3K/mTOR substrates . To confirm that P70S6K is a key downstream component mediating platinum resistance in this context , we knocked down P70S6K using two independent siRNAs in the NCI-H358 line , prior to proceeding with a cisplatin pulse ( Figure 3F , Figure 3—figure supplement 1A ) . Under these conditions , cisplatin treatment resulted in significantly higher caspase 3 cleavage and γH2A . X expression in the P70S6K knockdown NCI-H358 cells ( Figure 3G ) , whilst P70S6K knockdown did not further sensitise the already cisplatin-sensitive NCI-H1299 cell line ( Figure 3—figure supplement 1B ) . Additionally , P70S6K knockdown with both siRNAs also significantly increased cisplatin-induced apoptosis in the NCI-H358 cells , to the same extent as that observed with dactolisib treatment ( Figure 3H , Figure 3—figure supplement 1C ) . To confirm the efficacy this combination therapy approach in vivo and validate the findings of our pulse model , we treated mice bearing NCI-H358 xenografts with either carboplatin , dactolisib , or a combination of both ( Figure 3I ) . For this model , dactolisib ( 45 mg/kg ) was delivered by oral gavage , prior to a one-off intraperitoneal injection of carboplatin ( 60 mg/kg ) , and throughout the course of the experiment . Under these conditions , carboplatin had no significant effect upon tumour growth , while dactolisib had a moderate effect as a single agent that was only significant at days 14 and 17 . However the combination of carboplatin and dactolisib completely halted tumour growth , even up to 28 days following the single pulse of carboplatin treatment . Furthermore , an analysis of tumour sections revealed that the carboplatin and dactolisib treated xenografts were not only smaller in size , but also had significantly larger regions of necrosis ( Figure 3J ) . These findings demonstrate that elevated P70S6K activity can specifically mediate resistance to platinum-based chemotherapy in lung adenocarcinoma , which can be effectively targeted with the dual PI3K/mTOR inhibitor dactolisib . Importantly , elevated P70S6K activity is frequently observed in several cancer subtypes , including lung cancer ( Chen et al . , 2017 ) , and its overexpression is commonly associated with aggressive malignant phenotypes and poor overall prognoses ( Ip and Wong , 2012 ) . Indeed , an analysis of TCGA data using cBioPortal ( Gao et al . , 2013; Cerami et al . , 2012; Hoadley et al . , 2018 ) reveals that RPS6KB1 and RPS6KB2 , the two isoforms of P70S6K , are amplified or over-expressed in 20% and 11% of lung adenocarcinoma cases , respectively ( Figure 4A ) . Importantly , while TP53 mutations or deletions also occur within 47% of lung adenocarcinomas , RPS6KB1 and RPS6KB2 amplification/over-expression occur on the background of either wildtype or mutant TP53 . Further analysis with KM plotter ( Győrffy et al . , 2013 ) demonstrated that the elevated mRNA expression of both isoforms was also significantly associated with poor overall survival of lung adenocarcinoma patients ( Figure 4B ) . We therefore investigated the association with survival and response to chemotherapy by performing IHC for total P70S6K in a cohort of 52 lung adenocarcinoma patients that all received a neoadjuvant chemotherapy regimen containing platinum-based chemotherapy ( Figure 4C , Supplementary file 4 ) . In this cohort , high expression of P70S6K was significantly associated with poor overall survival ( Figure 4D ) , while there was a non-significant trend towards higher P70S6K expression in patients with progressive disease ( Figure 4E ) and a later disease stage ( Figure 4F ) . In eight patients with matched diagnosis and relapse samples , the levels of P70S6K expression were also significantly increased upon relapse ( Figure 4G ) , further highlighting the functional role of P70S6K in the cellular response to platinum therapy . As P70S6K has a known role in cell cycle progression ( Lane et al . , 1993 ) , we performed live cell imaging of the FUCCI two-colour sensor of cell cycle progression ( Sakaue-Sawano et al . , 2008 ) across the resistant A549 , NCI-H1573 and NCI-H358 cell lines . This approach allowed us to track the cell cycle progression and fate of individual cells for 72 hr following treatment with a 2 hr pulse of cisplatin ( Figure 5A ) . In this assay , treatment with a single pulse of cisplatin caused several notable cell cycle responses . First , A549 ( TP53 wildtype ) and NCI-H358 ( TP53 null ) cells were significantly more likely to remain arrested in G1 , both before ( G1 arrest before mitosis; ABM ) and after undergoing mitosis ( G1 arrest after mitosis; AAM ) ( Figure 5B , C ) . In contrast , NCI-H1573 cells did not show any significant increase in G1 transit time , as may be expected from a TP53 mutant cell line . For all cell lines , cisplatin pulsing resulted in a significant increase in S/G2 transit time ( Figure 5B , C ) . Notably , the NCI-H1573 ( TP53 mutant ) and NCI-H358 ( TP53 null ) cells delayed for more time in G2 ( G2 arrest; mean 2443 min and 2555 min respectively ) compared to A549 cells ( mean 1426 min ) , after which the mutant and null cells often entered a prolonged aberrant mitosis ( Figure 5—figure supplement 1 ) , resulting in a small increase in cells dying during mitosis or in the following G1/S phase ( Figure 5C , death after mitosis; DAM ) . In contrast , A549 cells rarely entered into mitosis , instead many of these S/G2 arrested cells turned from green ( S/G2 ) back to red ( G1 ) without undergoing mitosis ( G2-exit ) ( Figure 5—figure supplement 1 ) . A similar G2-exit , senescence state has been reported to be dependent on p21 ( Gire and Dulic , 2015 ) and likely provides TP53 wildtype cells protection from death by preventing progression through an aberrant mitosis . Therefore , to confirm the specific role of p53 in this context , this assay was repeated with an siRNA-mediated knockdown of p53 in A549 cells ( Figure 5—figure supplement 2 ) . In line with this hypothesis , this orthogonal approach revealed a significant increase in G2-arrest in the p53 knockdown cells , along with a decrease in occurrence of G2-exit , effectively pheno-copying the observed difference between the A549 wildtype and NCI-H358 p53 null cell lines ( Figure 5C ) . Treatment of all cell lines with dactolisib as a single agent also caused a significant increase in G1 arrest both before ( G1 ABM ) and after mitosis ( G1 AAM ) , irrespective of TP53 status ( Figure 5B , C ) . Consequently , for all cell lines , dactolisib treatment of cells within early G1 phase during the cisplatin pulse resulted in many cells remaining arrested and viable in G1 ( G1 ABM ) , highlighting the importance of DNA replication for inducing cisplatin induced toxicity and killing . However , for the TP53 wildtype and null cells that were in late G1 or S phase at the time of cisplatin pulsing , dactolisib treatment significantly reduced S/G2 transit time ( Figure 5B ) and greatly increased the number of cells that underwent apoptosis before mitosis ( Death before mitosis; DBM ) ( Figure 5C , D ) . Notably , dactolisib treatment did not prevent cisplatin treated cells from entering into an aberrant mitosis , but did significantly increase the duration of mitotic arrest , independent of TP53 status ( Figure 5B ) , which correlated with an increase in death during or after mitosis ( DAM ) in TP53 wildtype and null , but not mutant cells ( Figure 5C , D ) . Surprisingly , dactolisib treatment did not result in any sensitisation of TP53 mutant H1573 cells to cisplatin ( Figure 5D ) , which also corresponded with a failure of dactolisib to significantly reduce S/G2 phase transit time ( Figure 5B ) . In summary , combination therapy with dactolisib sensitised actively cycling cells to cisplatin through distinct mechanisms dependent on TP53 status . In the TP53 wildtype A549 cells it induced pre-mitotic cell death and prevented cells from undergoing a protective G2-exit , which is likely partially dependent on the p53-p21 axis . Similarly , in the TP53 null line , dactolisib promoted both pre-mitotic cell death and the entry of cells into a prolonged deleterious aberrant mitosis , the latter likely due to a weakened p53-dependent G2 checkpoint ( Engeland , 2018 ) . While in TP53 mutant H1573 cells , there was no significant sensitisation due to a sustained S/G2 arrest in the presence of dactolisib . To further confirm the specific role of P70S6K in this cell cycle phenotype , this assay was repeated along with siRNA mediated P70S6K knockdown in the resistant NCI-H358 cell line ( Figure 5—figure supplement 3 ) . This approach demonstrated that specific ablation of P70S6K in this model resulted in a non-significant trend towards a shortened G2-arrest with both P70S6K siRNAs , and a greatly increased cell death both before ( DBM ) or after mitosis ( DAM ) . While the non-significant decrease in G2-arrest observed upon siRNA-mediated knockdown of P70S6K may have been limited by potential heterogeneity in the level of knockdown achieved , the increase in cell death further suggests that dactolisib mediated inhibition of P70S6K is responsible for the sensitisation of lung adenocarcinoma cells to cisplatin treatment . A notable observation arising from this cell cycle analysis was the ability of dactolisib to increase apoptosis by preventing an unexpected cell cycle arrest in the TP53 null NCI-H358 cell line . Interestingly , this association between cell cycle arrest and cisplatin resistance can also be observed from the original analysis of signalling dynamics ( Figure 1D , Figure 1—figure supplement 4 ) . While both the NCI-H1299 and NCI-H358 cell lines are TP53 null , there was an increased expression of p21 following cisplatin treatment in the resistant NCI-H358 line , but not the sensitive NCI-H1299 line . Importantly , the increased p21 expression by NCI-H358 cells was also observed by western blotting of independent samples and was efficiently inhibited by both treatment with dactolisib ( Figure 6A , B ) and the specific knockdown of P70S6K with siRNA ( Figure 6C ) . Taken together , this data suggests that P70S6K activity is necessary for promoting p21 expression in order to maintain cell cycle arrest following cisplatin treatment . However , in the absence of p53 , we observed that this may potentially be mediated by the related transcription factor p63 , which is elevated in the NCI-H358 line following treatment with cisplatin , and also inhibited by the addition of dactolisib ( Figure 6A , B ) or treatment with P70S6K siRNA ( Figure 6C ) . Given the potential role for p21 in mediating resistance to cisplatin in the NCI-H358 line , we investigated whether sensitisation by dactolisib would be generalizable across the original panel of 6 lung adenocarcinoma cell lines with differing TP53 status . In accordance with this hypothesis , and in line with the FUCCI single-cell imaging ( Figure 5A ) , both the TP53 wildtype cell lines were significantly sensitised to cisplatin by the addition dactolisib , as was the resistant TP53 null line , NCI-H358 ( Figures 2D , H and 6D ) . The sensitive TP53 null line , NCI-H1299 , was not further sensitised and appeared to already be at the upper limit of apoptosis within this model . Also , in line with the FUCCI analysis , dactolisib was not able to increase cisplatin-induced apoptosis in the TP53 mutant line NCI-H1573 , whilst it significantly antagonised cisplatin-induced apoptosis in the TP53 mutant NCI-H1975 line . While TP53 mutation status alone is not sufficient to determine sensitivity to cisplatin ( Figure 1B ) , this dependence upon TP53 mutation status for dactolisib-mediated sensitisation to cisplatin treatment is likely explained by considering the ability of p21 to either maintain a G2-arrest , or promote an aberrant G2-exit ( Gire and Dulic , 2015 ) , thereby preventing DNA damage induced apoptosis occurring during , or shortly after mitosis . In line with this hypothesis , the addition of dactolisib to the TP53 wildtype A549 cell line inhibited cisplatin-induced p53 and p21 accumulation , resulting in elevated markers of DNA damage ( γH2A . X expression ) and apoptosis ( caspase 3 cleavage ) following cisplatin treatment ( Figure 6E ) . However , in the TP53 mutant NCI-H1573 line , which was not sensitised by dactolisib , p21 expression was not significantly elevated following cisplatin treatment , and γH2A . X expression and caspase 3 cleavage were unchanged by the addition of dactolisib ( Figure 7 ) . To further validate this link between sensitisation by dactolisib and TP53 mutation status , we took another panel of 8 lung adenocarcinoma cell lines ( Supplementary file 5 ) and correlated their relative expression of phosphorylated P70S6K ( Figure 6F ) to their apoptotic response to cisplatin ( Figure 6G ) . This secondary analysis revealed a general trend towards higher levels of phosphorylated P70S6K and decreased apoptosis in response to cisplatin across the whole panel . However , the strength of this correlation was greatly increased when considering the TP53 status of the lines , with a strong correlation observed for wildtype TP53 lines ( r2 = 0 . 6092 ) but no correlation for TP53 mutant lines ( r2 = 0 . 1764 ) . In agreement with our finding that elevated P70S6K activity promotes resistance in a TP53 dependent manner , combination therapy with cisplatin and dactolisib significantly sensitised the high pP70S6K/TP53 wildtype NCI-H292 cell line , but not the high pP70S6K/TP53 mutant NCI-H2009 line . Therefore , this data demonstrates that P70S6K promotes an aberrant G2-exit or prolonged G2-arrest following cisplatin treatment , presumably through enhanced translation of p53/p21 in TP53 wildtype cell lines and p63/p21 in TP53 null lines ( Figure 7 ) . However , in the presence of mutant p53 , P70S6K is not able to influence this DNA damage induced signalling axis and is therefore not associated with sensitivity to cisplatin .
A number of studies have been performed investigating the potential mechanisms of resistance to platinum-based chemotherapies ( Stewart , 2007 ) , although this has yet to result in the identification of clinically successful combination therapies for lung adenocarcinoma . One potential explanation for the high number of in vitro findings that have not translated to the clinic is the use of experimental techniques that do not replicate the in vivo pharmacokinetics of platinum therapies ( Shen et al . , 2012 ) . To date , most in vitro approaches have involved continuously culturing cell lines in the presence of high doses of platinum chemotherapy for multiple days , potentially allowing for more extensive DNA damage and off-target effects that would not be seen in vivo . Here we have utilised a pulse model that more accurately models both the concentration and timing of cisplatin that would be observed clinically . This is an important consideration , as platinum therapies are known to act via the formation of DNA adducts , although they are also capable of bonding with proteins and RNA ( Jamieson and Lippard , 1999 ) . It is likely that the continuous culturing of cells in the presence of high doses of these drugs in vitro would result in the accumulation of numerous off-target adducts and the activation of a stress response incongruent with the actual in vivo apoptotic mechanism . Indeed , this effect may be apparent in Figure 1—figure supplement 1D/2 , where p38 activity is induced by the continuous exposure model , but not by the pulse model . A number of previous studies have implicated the activity of p38 in mediating resistance to cisplatin ( Sarin et al . , 2018; Pereira et al . , 2013; Galan-Moya et al . , 2011; Hernández Losa et al . , 2003 ) , however we did not observe highly elevated p38 signalling in any cell line following the cisplatin pulse ( Figure 1D ) , nor a correlation between p38 activity and the apoptotic response ( Figure 2 ) . Instead , through the use of this pulse model , and by overlaying real-time apoptosis data onto a multivariate signalling analysis across a panel of lung adenocarcinoma lines , we have now identified elevated P70S6K activity as an effector of inherent platinum chemoresistance . High levels of P70S6K expression have previously been associated with aggressive tumour behaviour in lung cancer ( Chen et al . , 2017 ) , as well as other cancer types such as breast ( Holz , 2012 ) , colorectal ( Nozawa et al . , 2007 ) and liver ( Sahin et al . , 2004 ) cancers . This is also supported by our analysis of P70S6K expression patient samples ( Figure 4 ) , where high mRNA and protein levels were both associated with poor overall survival , and P70S6K protein expression was significantly elevated in relapsed tumours . Importantly , we also propose that combination therapy with the dual PI3K/mTOR inhibitor dactolisib can sensitise lung adenocarcinomas to platinum chemotherapy , although only in the context TP53 WT or TP53 null tumours . This detailed mechanistic understanding is especially important given that P70S6K amplification can occur on a background of either mutant or wildtype TP53 ( Figure 4A ) . Functionally , dactolisib will result in the inhibition of a number of substrates downstream of PI3K and mTOR . However , as the specific knockdown of P70S6K with siRNA recapitulated the effects of dactolisib on p63 , p21 , γH2A . X , cleaved caspase 3 and apoptosis , it is likely that the elevated expression and activity of P70S6K in resistant lung adenocarcinoma cells is the functional target of dactolisib and a causal effector of platinum resistance in these cells . Further supporting this central role of P70S6K , previous research has demonstrated that P70S6K can regulate cell cycle progression via the enhanced translation of p21 mRNA ( Guégan et al . , 2014 ) , although our data also suggest that P70S6K may also play a role in the enhanced expression of both p53 and p63 in this context . The combination of cisplatin and dactolisib has also previously been proposed for osteosarcoma ( Huang et al . , 2018 ) , head and neck squamous cell carcinoma ( Hsu et al . , 2018 ) and bladder cancer ( Moon et al . , 2014 ) . However , our finding that this combination therapy is effective in TP53 WT and null , but not TP53 mutant tumours , has significant implications for the clinical application of these drugs and the design of potential clinical trials . This fine grained stratification of response is an important finding , as the failure of many clinical trials is frequently attributed to the lack of suitable biomarkers to stratify the patient cohort ( Hay et al . , 2014 ) , highlighting the value of our approach that is capable of dissecting the mechanism of variable drug response across differing genetic backgrounds . The inability of dactolisib to regulate p21 expression in TP53 mutant tumour cells following cisplatin treatment likely underlies the inability of dactolisib to sensitise these cells to cisplatin . Therefore , within the context of a potential clinical trial for this drug combination , the TP53 mutation status of each patient tumour would be a central consideration for identification of the most effective treatment strategy and an understanding of individual patient response . However , this finding is also widely applicable to the understanding of how cancer-related signalling networks govern drug response and chemoresistance , and how a thorough characterisation of these dynamics will be necessary for the proper design and implementation of any consistently effective , patient-specific therapy .
The cisplatin modified DNA ( ab103261 ) and γH2A . X ( ab26350 ) antibodies were from Abcam ( Cambridge , USA ) . The phospho-STAT3 S727 ( #9134 ) , phospho-Akt S473 ( #9271 ) , phospho-NF-κB S536 ( #3033 ) , pERK T202/Y204 ( #9101 ) , pATF2 T71 ( #5112 ) , ( P70S6K ( #2708 ) , phospho-P70S6K T389 ( #9205 ) , p21 ( #2947 ) , cleaved caspase 3 ( #9661 ) and cleaved caspase 7 ( #9491 ) antibodies were from Cell Signaling ( MA , United States ) . The γH2A . X ( S139 ) antibody ( AB26350 ) was from Abcam ( MA , USA ) . The p53 antibody ( sc-126 ) was from Santa Cruz Biotechnology ( TX , USA ) . The p63 antibody ( NB100-691 ) was from Novus Biologicals ( CO , USA ) . The actin monoclonal antibody ( AC-15 ) was from Sigma-Aldrich ( MO , United States ) . The SignalSilence p70/85 S6 Kinase siRNA was from Cell Signaling ( MA , United States ) . The plasmids for FUCCI live cell imaging , mVenus-hGeminin ( 1/110 ) and mCherry-hCdt1 ( 30/120 ) , were a kind gift from Dr Atsushi Miyawaki ( Riken , Japan ) . Dactolisib ( NZP-BEZ235 ) , MK2206 , S3I-201 , UO126 and SC75741 were all from Selleck Chem ( MA , USA ) . All lung adenocarcinoma cell lines have been previously described ( Marini et al . , 2018 ) . The lines were cultured in Advanced RPMI ( 12633012 , Gibco ) containing 1% FCS and 1% GlutaMAX ( 35050–061 , Gibco ) under standard tissue culture conditions ( 5% CO2 , 20% O2 ) . All cell lines were authenticated by short tandem repeat polymorphism , single-nucleotide polymorphism , and fingerprint analyses , passaged for less than 6 months . All cell lines were confirmed as negative for mycoplasma contamination using the MycoAlert luminescence detection kit ( Lonza , Switzerland ) . Stable cell lines expressing the FUCCI biosensor were generated as previously described ( Sakaue-Sawano et al . , 2008 ) using mVenus-hGeminin ( 1/110 ) and mCherry-hCdt1 ( 30/120 ) probes . Briefly , cells were first transduced with mVenus-hGeminin ( 1/110 ) lentiviral particles . Cells were FACS sorted based upon Venus fluorescence , and the resulting cell population transduced again with mCherry-hCdt1 ( 30/120 ) lentiviral particles . mCherry positive cells were FACS sorted , resulting in a double positive population used for live cell imaging . Samples for flow cytometry were fixed in −20°C ethanol overnight , and then resuspended in a DNA staining solution containing 10 mg/mL RNaseA and 1 mg/mL propidium iodide for 30 min before analysis . Flow cytometry was performed using BD FACS Canto II system . For the visualisation of cisplatin-DNA adducts following treatment with a pulse of cisplatin ( 5 μg/mL , 2 hr ) , cells were grown on Histogrip ( Life Technologies ) coated glass coverslips . Prior to fixation , cells were treated with 1% Triton X-100 for 1 min , fixed with ice-cold 100% Methanol and stored overnight at −20°C . The cells were then permeabilized with 0 . 5% Triton X-100 for 10 min , washed twice with PBS and incubated for 30 min with 2M HCl to denature the DNA . Cells were washed again and incubated in 1% H2O2 to quench the endogenous peroxidases , before blocking for 30 min and incubated with primary antibody ( 1:500 ) overnight at 4 C in blocking solution . The following day , cells were incubated with Secondary-Biotinylated Ab at RT , before incubating them with ABC solution ( Vectastain elite ABC HRP kit , Vector Laboratories ) . Cells were later incubated with TSA solution ( TSA plus Cyanine 3 System , PerkinElmer ) . DNA was stained with H33342 and images collected using a Leica DMI5500 ( 40x magnification ) . Images were quantified using Fiji Software . Briefly , color images were split into separate channels . H33342 channel was used to identify the nucleus and generate a mask that was placed on top of the CisPt-DNA channel to quantify the signal coming from the nuclear area . Cytoplasmic area was manually identified in each cell and signal quantified . Background signal was obtained and subtracted from the nuclear and cytoplasm signal . From this data , a nuclear/cytoplasmic ratio was obtained for between 100 and 200 cells at each time point . Multiplex analysis was performed using a Bio-Plex MAGPIX system ( #171015044 ) and Bio-Plex Pro-Wash Station ( Biorad ) . All cell lysates were prepared using standard cell lysis buffer ( 50 mM Tris HCl pH7 . 4 , 150 mM NaCl , 1 mM EDTA , 1% ( v/v ) TritonX-100 ) supplemented with protease and phosphatase inhibitors . Lysates were analysed on all kits , according to manufacturer’s instructions . Data was generated using Bio-Plex Manager MP and analysed on the Bio-Plex Manager 6 . 1 software . Lysates were analysed using the Milliplex map DNA Damage/Genotoxicity Magnetic Bead Panel ( 7-plex ) , Milliplex map Early Apoptosis Magnetic Bead ( 7-plex ) , Milliplex map TGFβ Signalling Pathway Magnetic Bead kit ( 6-plex ) , Bio-Plex Pro Cell Signalling Akt Panel ( 8-plex ) , Bio-Plex Pro Cell Signalling MAPK Panel ( 9-plex ) , Bio-Plex Pro RBM Apoptosis Panel two and Bio-Plex Pro RBM Apoptosis Panel 3 . Individual beads were also used to analyse NF-κB ( Ser536 ) , IκBα ( Ser32/36 ) , c-Jun ( Ser63 ) , total P53 and cleaved PARP ( Biorad ) . The data was normalized to the median value at the 0 hr time point for each analyte and a log transformation was conducted on the resulting dataset . The principal component analysis was performed using MATLAB and Statistics Toolbox Release 2019a ( The Mathworks , Inc , Natick , Massachusetts , United States of America ) . Live-cell imaging of apoptosis was performed using an Essen Bioscience IncuCyte ZOOM Live-Cell Analysis System and a Thermo Fisher Scientific HERAcell 240i CO2 Incubator . Cells were seeded into Corning Costar TC-treated 96-Well Plates and imaged over a 72 hr period at 2 hr intervals over 4 fields of view per well . Caspase activation was visualised using 1 µM NucView 488 Caspase-3 Enzyme Substrate ( Biotium ) . Cell viability was quantified using propidium iodide ( 1 μg/mL ) . The generated images were analysed using IncuCyte ZOOM Software Version 2016B . Accompanying cytotoxicity assays were performed using the CellTiter 96 aqueous non-radioactive cell proliferation assay , according to the manufacturer’s instructions ( Promega ) . Lysates for western blotting were prepared using standard lysis buffer ( 50 mM Tris HCl pH7 . 4 , 150 mM NaCl , 1 mM EDTA , 1% ( v/v ) Triton X-100 ) containing protease inhibitor ( p8340 , Sigma ) and 0 . 2 mM sodium orthovanadate . The NuPAGE SDS PAGE Gel System and NuPAGE Bis Tris Precast Gels ( 4–12% and 12% ) ( Life Technologies ) were used to perform gel electrophoresis . Western Lightning PLUS Enhanced Chemiluminescent Substrate ( PerkinElmer ) was used for imaging western blots on the Vilber Lourmat Fusion chemiluminescent imaging system . Quantitative western blotting was performed using multistrip western blotting , as performed previously ( Kennedy et al . , 2019 ) . NCI-H358 cells ( 2 × 106 ) resuspended in 100 μL PBS:Matrigel were injected subcutaneously into the flanks of nude mice . Tumour growth was assessed twice weekly by calliper measurement and mice were randomized to treatment arms when tumours reached 150 mm3 ( using the formula: width2 x length x 0 . 5 ) . Carboplatin ( 60 mg/kg ) was delivered by a single tail-vein injection . Dactolisib was prepared in 10% DMSO:90% PEG300 and administered twice-weekly at 45 mg/kg by oral gavage . All in vivo experiments , procedures and endpoints were approved by the Garvan Institute of Medical Research Animal Ethics Committee . Immunohistochemistry was performed on formalin fixed paraffin embedded sections using the Leica BOND RX ( Leica , Wetzlar , Germany ) . Slides were first dewaxed and rehydrated , followed by heat induced antigen retrieval performed with Epitope Retrieval Solution 1 BOND ( Leica , Wetzlar , Germany ) . Primary antibodies were diluted 1:600 ( P70S6K ) and 1:500 ( Ki67 ) in Leica antibody diluent and incubated for 60 min on slides . Antibody staining was completed using the Bond Polymer Refine IHC protocol and reagents ( Leica , Wetzlar , Germany ) . Slides were counterstained on the Leica Autostainer XL ( Leica , Wetzlar , Germany ) . Leica CV5030 Glass Coverslipper ( Leica , Wetzlar , Germany ) and brightfield images were taken on the Aperio CS2 Slide Scanner ( Leica , Wetzlar , Germany ) . Quantification of Ki67 staining was performed on three fields of view for each tumour section , and quantified using the particle analysis function of Image J ( v1 . 49 ) . For FUCCI live cell imaging , cells were seeded on 12 well plates and imaged using a Leica DMI6000 using a 20x NA 0 . 4 objective . Images were taken every 20 min for 72 hr . Before adding any drug ( cisplatin , dactolisib or both ) an image was taken in order to annotate the cell cycle phase before commencing treatment . Individual cells were tracked manually , with the colour of the nucleus annotated at each time point ( Red = G1; Yellow = G1/S , Green = S/G2 ) , the cells were also scored for nuclear envelope breakdown ( NEBD ) and early signs of anaphase . Mitotic length was calculated by the time period from commencement of NEBD to anaphase . Interphase length was calculated from anaphase to the next daughter cell NEBD . Only one daughter was followed and annotated . Tracking graphs were generated using Prism 7 . | Lung adenocarcinoma is the most common type of lung cancer , and it emerges because of a variety of harmful genetic changes , or mutations . Two lung cancer patients – or indeed , two different sets of cancerous cells within a patient – may therefore carry different damaging mutations . A group of drugs called platinum-based chemotherapies are currently the most effective way to treat lung adenocarcinoma . Yet , only 30% of patients actually respond to the therapy . Many studies conducted in laboratory settings have tried to understand why most cases are resistant to treatment , with limited success . Here , Hastings , Gonzalez-Rajal et al . propose that previous research has been inconclusive because studies done in the laboratory do not reflect how the treatment is actually administered . In patients , platinum-based drugs are cleared from the body within a few hours , but during experiments , the treatment is continually administered to cells growing in a dish . Hastings , Gonzalez-Rajal et al . therefore developed a laboratory method that mimics the way cells are exposed to platinum-based chemotherapy in the body . These experiments showed that the lung adenocarcinoma cells which resisted treatment also carried high levels of a protein known as P70S6K . Pairing platinum-based chemotherapy with a drug that blocks the activity of P70S6K killed these resistant cells . This combination also treated human lung adenocarcinoma tumours growing under the skin of mice . However , it was ineffective on cancerous cells that carry a mutation in a protein called p53 , which is often defective in cancers . Overall , this work demonstrates the need to refine how drugs are tested in the laboratory to better reflect real-life conditions . It also underlines the importance of personalizing drug combinations to the genetic background of each tumour , a concept that will be vital to consider in future clinical trials . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2020 | Analysis of pulsed cisplatin signalling dynamics identifies effectors of resistance in lung adenocarcinoma |
Synaptic plasticity , which underlies learning and memory , depends on calcium elevation in neurons , but the precise relationship between calcium and spatiotemporal patterns of synaptic inputs is unclear . Here , we develop a biologically realistic computational model of striatal spiny projection neurons with sophisticated calcium dynamics , based on data from rodents of both sexes , to investigate how spatiotemporally clustered and distributed excitatory and inhibitory inputs affect spine calcium . We demonstrate that coordinated excitatory synaptic inputs evoke enhanced calcium elevation specific to stimulated spines , with lower but physiologically relevant calcium elevation in nearby non-stimulated spines . Results further show a novel and important function of inhibition—to enhance the difference in calcium between stimulated and non-stimulated spines . These findings suggest that spine calcium dynamics encode synaptic input patterns and may serve as a signal for both stimulus-specific potentiation and heterosynaptic depression , maintaining balanced activity in a dendritic branch while inducing pattern-specific plasticity .
Neurons receive information from other neural cells in the form of patterns of activation of different synaptic inputs . Such input patterns differ in their locations on the dendritic tree ( spatial properties ) and their timing ( temporal properties ) ( London and Häusser , 2005 ) . As a given neuron may receive synaptic inputs from hundreds to thousands of other neurons , a critical question in neuroscience is how multiple synaptic inputs are integrated to produce neuronal output . Further , certain patterns of input can induce synaptic plasticity—neural activity-dependent changes in synaptic efficacy that underlie learning and memory . Yet , it remains unclear how spatiotemporal properties of synaptic input patterns may affect synaptic plasticity ( Destexhe and Marder , 2004; van Bommel and Mikhaylova , 2016 ) . Dendrites are capable of complex , non-linear forms of synaptic integration , which are sensitive to the spatiotemporal properties of synaptic inputs ( Stuart and Spruston , 2015 ) . For instance , in vitro studies have shown that near-simultaneous stimulation of a group of spatially clustered excitatory synapses on a thin dendritic branch can elicit supralinear , prolonged membrane depolarizations in the soma ( known as plateau potentials ) . These plateau potentials have been observed in pyramidal neurons of the cortex ( Larkum et al . , 2009; Schiller et al . , 2000 ) and hippocampus ( Golding et al . , 2002; Harnett et al . , 2012; Makara and Magee , 2013 ) , and also in spiny projection neurons of the striatum ( Du et al . , 2017; Mahfooz et al . , 2016; Oikonomou et al . , 2014; Plotkin et al . , 2011 ) . These non-linear responses to spatiotemporally clustered synaptic input can induce synaptic plasticity . Specifically , long-term potentiation ( LTP ) of synaptic inputs can be induced by stimulation of clustered synapses , independently of postsynaptic action potentials ( Brandalise et al . , 2016; Golding et al . , 2002; Gordon et al . , 2006; Losonczy et al . , 2008 ) . Calcium influx into neuronal dendrites and spines is a critical mechanism linking synaptic input patterns to synaptic plasticity , as calcium is required for most forms of neuronal plasticity throughout the brain ( Greer and Greenberg , 2008; Higley and Sabatini , 2008; Zucker , 1999 ) . The conjunction of synaptic inputs and postsynaptic depolarization produces calcium influx through the NMDA subtype of glutamate receptor ( NMDAR ) channels ( Bartol et al . , 2015; Schiller et al . , 1998; Sjöström and Nelson , 2002 ) . Activation of calcium-permeable ligand-gated or voltage-gated ion channels also yields calcium influx . The resulting elevation in intracellular calcium acts as a second messenger to initiate multiple signaling cascades that produce various forms of synaptic plasticity . Calcium , therefore , connects the electrical activity at the network or neuronal level to the subcellular level of biochemical signaling and plasticity . The relationship between calcium and plasticity is complex , as calcium elevation is required for both LTP and long-term depression ( LTD ) . Both experiments and theory propose that plasticity outcomes depend on the specific dynamics of intracellular calcium , including amplitude , duration , and location ( Evans and Blackwell , 2015; Graupner and Brunel , 2012 ) . Thus , determining how calcium dynamics in dendrites and spines depend on spatiotemporal patterns of synaptic input will advance our understanding of how those same patterns induce plasticity and ultimately influence learning and memory . Spatiotemporally clustered synaptic inputs that produce supralinear plateau potentials ( also called NMDA spikes ) also cause elevated dendritic calcium concentration localized to the stimulated dendritic branch ( Antic et al . , 2010; Larkum et al . , 2009; Major et al . , 2008; Schiller et al . , 2000 ) . In vivo , NMDAR-dependent calcium transients that are limited to specific dendritic branches and spines of pyramidal neurons correspond with spine-specific structural plasticity and behavioral learning ( Cichon and Gan , 2015 ) . In vitro , repeated synaptic stimulation of neighboring spines can result in supralinear spine calcium transients and LTP ( Weber et al . , 2016 ) , even in the absence of somatic plateau potentials . Thus , spatiotemporally clustered patterns of synaptic inputs are critical for information processing and plasticity , but it is unclear how distributed input patterns , which likely occur frequently in vivo , produce non-linear synaptic responses and affect synaptic plasticity . Although much of the literature has focused on synaptic integration in pyramidal neurons , NMDAR-dependent plateau potentials also have been observed in the spiny projection neurons ( SPNs ) of the striatum , which is the input nucleus of the basal ganglia ( Du et al . , 2017; Mahfooz et al . , 2016; Oikonomou et al . , 2014; Plotkin et al . , 2011 ) . The striatum integrates glutamatergic input from cortex and thalamus and dopaminergic input from substantia nigra to learn goal-directed actions , motor skills , and habits ( Kreitzer and Malenka , 2008 ) . Calcium elevation and dopamine are required for synaptic plasticity in SPNs ( Yagishita et al . , 2014 ) . It has been suggested that calcium elevation ( through downstream signaling events ) may generate a ‘synaptic eligibility trace’ that , when followed by dopamine stimulation , produces LTP ( Shindou et al . , 2018 ) ; thus calcium dynamics are critical even in brain regions that require dopamine for synaptic plasticity . Similar to pyramidal neurons , near-synchronous synaptic input to a cluster of 10 – 20 neighboring spines evokes NMDAR-dependent plateau potentials in SPNs ( Du et al . , 2017; Plotkin et al . , 2011 ) , but only when the cluster of spines is located distally , and not proximally . Although not yet demonstrated , this supralinearity may produce synaptic plasticity at these distal SPN synapses . SPNs and pyramidal neurons exhibit key differences which motivate the further study of SPNs . In contrast to pyramidal cells , SPNs lack the morphologically distinct apical , oblique , and basal dendritic branches characteristic of pyramidal neurons ( Gertler et al . , 2008; Spruston , 2008 ) . Whereas pyramidal neurons exhibit sodium- and calcium-spikes in addition to NMDA spikes ( Stuart and Spruston , 2015 ) , SPNs lack sodium channels in distal dendrites and dendritic calcium-spikes have not been measured in these neurons ( Day et al . , 2008; Plotkin et al . , 2011 ) . SPNs rest at more hyperpolarized membrane potentials than pyramidal neurons , in part because of differences in hyperpolarization-activated ion channels—SPNs strongly express inward rectifying potassium channels ( KIR ) and lack the hyperpolarization-activated cyclic nucleotide-gated ( HCN ) channels seen in pyramidal neurons ( Nisenbaum and Wilson , 1995 ) . SPNs also transition between hyperpolarized down-states and depolarized up-states in vivo ( Wilson and Kawaguchi , 1996 ) , and dendritic non-linearities may play an important role in driving these state transitions ( Du et al . , 2017; Plotkin et al . , 2011 ) . Lastly , cortical axons make only 1 – 3 synapses with a single SPN ( Kincaid et al . , 1998 ) , and the large number of synaptic inputs required to produce an upstate ( Blackwell et al . , 2003; Stern et al . , 1998 ) suggests that in SPNs spatiotemporally dispersed synaptic inputs occur frequently . The striatal microcircuit also differs significantly from that of the cortex or hippocampus . The local circuit of the striatum is almost entirely inhibitory ( Burke et al . , 2017 ) , including the SPN collaterals . Other sources of inhibition include fast spiking interneurons ( FSIs ) , low-threshold spiking interneurons ( LTSIs ) , and neurogliaform ( NGF ) interneurons ( Burke et al . , 2017; Ibáñez-Sandoval et al . , 2011; Kawaguchi et al . , 1995; Koos et al . , 2004; Straub et al . , 2016; Tepper and Bolam , 2004; Koós and Tepper , 1999 ) . Synaptic inputs onto SPNs from these sources exhibit distinct spatial and temporal properties . FSIs fire at high rates and are limited to the soma and proximal dendrites of SPNs . In contrast , SPN collaterals , LTSIs , and NGFs target distal dendrites of SPNs , and NGF synapses further display distinctly slow GABAA kinetics . These distinct sources of inhibition have been shown to regulate plateau potentials in somatic recordings ( Du et al . , 2017 ) . Yet , although inhibition is clearly important for striatal function , its role in regulating local calcium transients in dendritic spines is unknown . When clustered synaptic input produces a plateau potential , it is unclear how synapse-specificity is maintained . Calcium imaging experiments indicate that the entire dendritic branch at the site of clustered synaptic input experiences robust calcium elevation ( Plotkin et al . , 2011 ) , and the plateau potential likely propagates from the dendritic branch into neighboring non-stimulated dendritic spines ( Koch and Zador , 1993 ) , which could minimize synapse-specificity . However , neither experiments nor models have investigated the degree of synapse-specificity in the calcium response during supralinear plateau potentials , in any neuron type . Given the importance of synaptic activity patterns for information processing and plasticity , understanding the role of both spatiotemporally clustered and distributed synaptic input patterns on calcium dynamics is critical . However , because of experimental technical constraints , computational modeling is required to investigate the response to spatiotemporally dispersed inputs . We address these critical issues regarding the effect of spatiotemporal input patterns on calcium dynamics in stimulated spines , non-stimulated spines , and dendritic branches in a detailed computational SPN model . We show that both dispersed and clustered synaptic inputs can evoke supralinear calcium influx into stimulated spines with spatial specificity . Lastly , our most novel finding is that inhibition enhances spatial- and synapse-specificity of spine calcium transients during plateau potentials .
We developed a detailed biophysical SPN model to investigate the effect of spatially and temporally clustered and distributed synaptic inputs on spine calcium dynamics ( Figure 1A ) . Ion channel densities were tuned to reproduce SPN electrophysiological characteristics in response to current injection ( Figure 1B ) . The model exhibits the inward rectification and sag in response to hyperpolarizing current injection , latency to first action potential , shallow AHP amplitude , input resistance , and low firing frequency characteristic of SPN recordings ( Nisenbaum and Wilson , 1995 ) . Calcium dynamics were incorporated into the model to reproduce an array of experimental data ( Figure 1A , C ) . Voltage compartments were subdivided into smaller calcium compartments—either radial diffusion shells in the soma and dendrites , or axial diffusion slabs in the spines . Each shell or slab also had diffusible calcium buffers ( calmodulin and calbindin ) , a low affinity fixed buffer ( Matthews and Dietrich , 2015 ) , and plasma membrane calcium pumps ( for compartments adjacent to membranes ) . Maximal conductances of voltage-gated calcium channels ( VGCCs ) were tuned to reproduce experiments measuring the calcium concentration elevations in dendrites and spines in response to a back-propagating action potential ( bAP ) . Conductances of NMDAR and AMPAR channels were tuned to reproduce calcium imaging experiments showing spine calcium elevation during a single excitatory postsynaptic potential ( EPSP ) . Figure 1C shows that bAP-evoked calcium elevation is greater in proximal dendrites than in the soma and decreases with distance in tertiary dendrites , consistent with experimental reports ( Day et al . , 2008; Kerr and Plenz , 2002 ) . Also , peak calcium in a proximal spine from a bAP was 0 . 18 μM with a time constant of decay of 74 ms , and in response to a single EPSP peaked at 0 . 2 μM with a time constant of decay of 73 ms , similar to experimental results when simulated under similar calcium-indicator conditions ( Shindou et al . , 2011 ) . The relative contributions of specific VGCC types and synaptic calcium sources to spine calcium elevation were also tuned to reproduce experimental data ( Carter and Sabatini , 2004; Higley and Sabatini , 2010 ) . Blockade of NMDARs , AMPARs , and T-type , R-type , or L-type VGCCs reduced the spine calcium elevation in response to a single EPSP . Spine calcium elevation in response to a bAP or EPSP was evaluated for parameter variations of ±10% ( Figure 1—figure supplement 1 ) . These results demonstrate that the model peak spine calcium response is robust to parameter variations , with at most a 10% change in response to a bAP and a 30% change in response to an EPSP ( within reported experimental variability ) ( Shindou et al . , 2011 ) . To further assess model robustness , we evaluated the fates of calcium entering a dendritic spine during a single EPSP for 100 ms following synaptic stimulation . Calcium fates , which were calculated as the quantity ( moles ) of free calcium , buffered calcium , pumped calcium , and diffused calcium per timestep ( Figure 1—figure supplement 2 ) , exhibited similar dynamics to a published computational model with three-dimensional reaction-diffusion in reconstructed dendritic spines from pyramidal neurons ( Bartol et al . , 2015 ) . Together , the ability to reproduce multiple sources of both electrophysiology and calcium-imaging data suggest that the model is well-suited to investigate the effects of synaptic activity on calcium dynamics . SPNs exhibit plateau potentials in response to spatiotemporally clustered synaptic inputs to distal dendritic spines , but it is unclear how these plateau potentials affect spine calcium dynamics or synaptic plasticity . Further , the likelihood of spatial clustering of synaptic inputs occurring in vivo is unknown , and may be low , as individual cortical axons make few synaptic connections to a single SPN ( Kincaid et al . , 1998 ) . The ability of spatially distributed synaptic inputs to produce a plateau potential would increase the biological relevance of plateau potentials for in vivo function of SPNs . To investigate the role of spatial input patterns , we first verified that our model reproduces plateau potentials in response to spatiotemporally clustered synaptic inputs to distal dendritic spines , and we then investigated the effect of both spatially clustered and spatially dispersed excitatory synaptic inputs on spine calcium elevation . To first verify that plateau potentials occur in the model as in reported experiments , we simulated the model with synchronous synaptic input to 1–18 spines within an 18 μm dendritic segment located at increasing distances along a single terminal dendritic branch ( Figure 2A–B ) . The model SPN exhibits a non-linear plateau potential when a cluster of distal , but not proximal , dendritic spines are simultaneously stimulated ( we refer to stimulated synapses on dendritic spines as ‘stimulated spines’ ) ( Figure 2A ) , consistent with experimental observations ( Du et al . , 2017; Plotkin et al . , 2011 ) . Blockade of NMDARs abolished the plateau potential , whereas blockade of VGCCs attenuated the plateau potential , consistent with previous experiments ( Plotkin et al . , 2011 ) . Having confirmed that the electrical response is consistent with published results , we focused the remainder of our investigation on spine calcium concentration because of the importance of calcium for synaptic plasticity . Supralinear spine calcium transients occur during simultaneous synaptic input to spatially clustered spines located on the distal portion of a dendritic branch . ( Figure 2B ) . Stimulation at proximal locations shows a sublinear peak spine calcium response for up to 18 simultaneously stimulated spines . At distal locations , a sharp threshold emerges where the stimulation of a single additional spine produces a supralinear spine calcium elevation . Above threshold , stimulation of additional spines produces a graded increase in the magnitude of the spine calcium elevation . The threshold is distance-dependent , with fewer stimulated spines required to produce a supralinear calcium response closer to the terminal end of the dendritic branch . These results suggest that the distal dendritic spines of SPNs may be important sites for synaptic cooperativity—the ability of synapses to evoke supralinear responses when stimulated together . Furthermore , the supralinear spine calcium elevation evoked by clustered synaptic inputs may be an important mechanism for synaptic plasticity . A critical question is whether spatial clustering of cortical inputs is required , or if spatially dispersed inputs can still cooperate to produce plateau potentials and supralinear spine calcium elevation . To address this question , excitatory inputs were randomly distributed over the proximal ( 27 – 119 μm from soma ) or distal ( 135 – 225 μm from soma ) regions of a dendritic branch , or over the entire branch ( Figure 2C ) . Synaptic inputs that are spatially dispersed over the distal half of the branch still produce a supralinear response , although it is slightly smaller than the response to spatially clustered inputs . Too much spatial dispersion is not tolerated , as distributing 20 inputs over the entire branch no longer elicits supralinear spine calcium elevation . These results suggest that simultaneously stimulated synapses on distal dendritic spines still produce robust calcium influx when distributed within a 90 μm segment , indicating that close spatial clustering may not be a strong requirement for synaptic plasticity . In summary , the simulations of spatially dispersed synaptic inputs predict that SPNs produce plateau potentials and supralinear spine calcium elevation in response to the coordinated stimulation of ~16 excitatory synaptic inputs on a distal dendritic branch , and that this effect does not require precise spatial clustering . Synapse-specificity—that potentiation is limited to only those synapses which actively contribute to a postsynaptic response—is critical for synaptic plasticity to underlie learning and memory . Synapse-specific LTP requires that the elevated calcium concentration is confined to stimulated spines , as opposed to non-stimulated , neighboring spines , during a plateau potential . It is not clear how synapse-specificity is maintained during plateau potentials , when the entire dendritic branch becomes strongly depolarized and experiences calcium elevation as shown in calcium-imaging experiments ( Plotkin et al . , 2011 ) . The strong dendritic depolarization could lead to calcium influx through VGCCs in neighboring non-stimulated spines or diffusion from dendrite to spine , leading to heterosynaptic LTP and a loss of synapse-specificity . To evaluate the extent of synapse-specificity , we investigated whether the calcium response differs between stimulated and neighboring non-stimulated spines for both spatially clustered and distributed inputs to a single dendritic branch ( Figure 3 ) . To quantify synapse-specificity , we sampled calcium concentration from spines along the entire branch , both proximal and distal spines , including non-stimulated spines located between two stimulated spines for the spatially dispersed condition . In response to synaptic inputs to distal dendritic spines , peak calcium elevation in stimulated spines was an order of magnitude higher than in non-stimulated , neighboring spines . Similarly , when synaptic inputs were spatially dispersed , peak calcium elevation in stimulated spines was again an order of magnitude higher than in non-stimulated , interspersed spines ( Figure 3A ) . Thus , calcium elevation may provide a synapse-specific signal for potentiation of coordinated synaptic inputs to spines on the distal dendritic branch , whether the synaptic inputs are spatially clustered or dispersed . Although calcium elevation in non-stimulated spines was significantly smaller than stimulated spines , it was also significantly higher than baseline , although only for a subset of non-stimulated spines . Specifically , only the distally located non-stimulated spines exhibited a calcium elevation , with negligible calcium elevation in proximal non-stimulated spines . This suggests that heterosynaptic effects that may occur during a plateau potential would be limited to neighboring non-stimulated spines located on the distal dendritic branch . This distal-to-proximal gradient in calcium elevation is also observed in the dendritic shaft . Therefore , other calcium-dependent types of plasticity , such as homeostatic scaling or branch strength plasticity , also may occur in the distal dendritic shaft during plateau potentials . In summary , these results show three distinct , spatially specific calcium responses to coordinated distal synaptic inputs: a high , supralinear response in stimulated spines; an intermediate response in non-stimulated distal spines and in the distal dendritic shaft; and a negligible response in proximal non-stimulated spines . Understanding the biological mechanisms controlling differences in calcium dynamics between stimulated and neighboring non-stimulated spines may yield greater insights to synapse-specific signaling . Three different sources of spine calcium elevation—calcium permeable synaptic channels , voltage-gated calcium channels , and diffusion ( of calcium or calcium-bound buffers ) —could enhance or decrease differences in spine calcium concentration between stimulated and non-stimulated spines . To investigate which of these mechanisms distinguishes stimulated spines , we analyzed membrane potential , spine calcium concentration , and calcium channel currents in a stimulated spine and a neighboring non-stimulated spine during clustered distal synaptic stimulation . Additionally , we isolated VGCC-mediated and diffusion-mediated calcium elevations in non-stimulated spines by selectively blocking diffusion between non-stimulated spines and the dendritic shaft . Calcium influx through synaptic channels underlies the high spine calcium elevation specific to stimulated spines . Analysis of spine head membrane potential indicated little difference between stimulated and neighboring non-stimulated spines ( Figure 3B ) , indicating that VGCC-mediated calcium elevation would be similar in neighboring stimulated or non-stimulated spines . Analysis of VGCC calcium currents ( Figure 3D ) confirmed that VGCCs were activated similarly in stimulated and non-stimulated spines , whereas analysis of synaptic calcium currents in stimulated spines indicated orders of magnitude higher currents than VGCC currents in stimulated spines ( Figure 3E ) . VGCCs and diffusion both contribute to calcium elevations in non-stimulated neighboring spines . Blocking diffusion of calcium and calcium buffers between the dendritic shaft and non-stimulated spines reduced the later phase of calcium elevation in non-stimulated spines ( Figure 3C ) , indicating that diffusion may support calcium transient duration . In contrast , the early phase and peak amplitude of the calcium transient did not change when diffusion between the non-stimulated spines and dendritic shaft was blocked , indicating that VGCCs on non-stimulated spines underlie calcium transient peak amplitude . Together , these results indicate that the spatial specificity of calcium elevation results from synaptic calcium influx in stimulated spines , whereas both diffusion and VGCC influx increase calcium concentration in neighboring , non-stimulated spines . Consequently , calcium transients exhibit a robust synapse-specific signal in stimulated spines despite highly similar membrane potentials in neighboring spines during plateau potentials . Further , as membrane potential is sharply attenuated as depolarization propagates towards the soma , the VGCC-mediated elevation is limited to distal non-stimulated spines whereas proximal non-stimulated spines exhibit negligible calcium elevation . The distinct calcium transients and sources observed may therefore support stimulus-specific synaptic plasticity in stimulated spines and heterosynaptic plasticity in neighboring , non-stimulated spines . Synaptic inputs are likely to be spatially and temporally dispersed on multiple dendritic branches in vivo . This raises critical questions: to evoke supralinear spine calcium transients , do cortical inputs need to target a single dendritic branch , or can inputs be spatially dispersed on the entire neuron ? Further , would synaptic inputs on multiple branches act independently—such that the number of synaptic inputs per branch required to evoke supralinear calcium transients is independent of synaptic inputs to other branches—or would they interact ( cooperate ) to lower the threshold number of synaptic inputs per branch required for supralinear calcium transients ? To address these questions , we investigated the effect of spatial dispersion of synaptic inputs to multiple dendritic branches . Synaptic inputs were randomly distributed on two tertiary branches with a common secondary branch parent; on four tertiary branches with a common primary branch parent; or on eight tertiary dendritic branches . Additionally , we used both simultaneous synaptic inputs and temporal dispersion , created from random , exponentially distributed intervals . Simulations used average interstimulus intervals ( ISIs ) of 2 . 5 , 5 , or 10 ms between synaptic inputs on each branch ( e . g . with a 10 ms mean ISI per branch , when two branches are stimulated , the overall mean ISI is 5 ms for all inputs ) ( Figure 4 ) . Specifying the ISI per branch has the advantage that the total duration of synaptic stimulation is independent of the number of branches . The temporal order of stimulated spines was randomly selected and followed no spatial pattern . When simultaneous synaptic inputs are spatially dispersed across multiple branches , the total number of synaptic inputs required to elicit supralinear spine calcium transients is increased ( Figure 4B ) . This indicates a degree of dendritic branch independence . However , slightly fewer stimulated spines per branch are required to evoke supralinear spine calcium transients when more branches are stimulated , indicating a small degree of interaction between branches . Thus , supralinear calcium elevation in stimulated spines is dendritic branch-specific , although the required number of stimulated spines on a single branch may be slightly reduced by interaction with synaptic inputs to other dendritic branches . As the average ISI is increased , there is an overall reduction in supralinear spine calcium elevation . However , when multiple branches are stimulated , increasing the mean ISI up to 5 ms can increase spine calcium elevation , indicating that , with spatial dispersion , a small temporal dispersion increases cooperativity . The corresponding somatic voltage ( Figure 4C ) shows sustained depolarization , which may be causing the increased spine calcium elevation . Supralinearity in spine calcium elevation appears to be negligible for average ISIs greater than 10 ms . This suggests that there is a limited temporal window for evoking supralinear spine calcium transients in SPNs , although precisely synchronous inputs are not required . Altogether , results using spatiotemporally dispersed synaptic inputs suggest that each distal dendritic branch in SPNs may function as a relatively independent subunit for integrating synaptic inputs with spine-specific calcium responses . These results are consistent with the theory that individual dendritic branches serve as critical subunits for synaptic integration and plasticity ( Branco and Häusser , 2010 ) . The distinct calcium signals in proximal versus distal stimulated spines in response to coordinated synaptic inputs suggests that proximal and distal synapses may have distinct functions in SPNs . This raises the question of whether distal and proximal synapses can interact or instead function independently . Just as a somatic action potential can back-propagate to enhance spine calcium elevation , we investigated whether a cluster of synaptic inputs , in the absence of a somatic action potential , can enhance spine calcium elevation . To test whether a distally evoked plateau potential could interact with a stimulated proximal synapse on the same branch , we paired clustered distal stimulation and single proximal spine stimulation with varying temporal intervals and evaluated the calcium responses in the proximal spine ( Figure 5 ) . Our results show that distal clustered synaptic inputs can facilitate calcium elevation in response to synaptic input to proximal spines , dependent on timing . Simultaneous stimulation of a distal cluster and single proximal spine facilitated the proximal spine peak calcium concentration ( Figure 5A ) . The simultaneous stimulation produced higher proximal spine calcium elevation than the sum of the proximal and distal stimulations performed independently . We repeated the pairing at four different ISIs—proximal stimulus 25 or 50 ms before distal cluster stimulus , and proximal stimulus 25 or 50 ms after the distal cluster ( Figure 5B ) . We found that peak calcium elevation in the proximal spine was facilitated if the proximal stimulus came 25 ms after the distal stimuli . However , peak calcium elevation in the proximal spine was not facilitated when the proximal spine stimulus came 50 or 25 ms before the distal stimuli , nor when the proximal stimulus came 50 ms after the distal stimuli . Interestingly , the duration of the calcium transient in the proximal spine was prolonged when the proximal spine was stimulated prior to the distal cluster . To investigate the mechanism underlying this temporal dependence , we evaluated the voltage in the dendritic shaft at the base of the proximal spine . The enhancement of peak calcium in the proximal spine correlated with the amplitude of the depolarization propagating from the distally evoked plateau potential . Together , these results indicate that distally evoked plateau potentials can interact with proximal synaptic inputs within a 25 ms temporal window , and that the interaction shows asymmetrical dependence on timing . Thus , similar to spike-timing-dependent plasticity rules which depend on the order and timing of presynaptic activation and a postsynaptic spike , we predict that a plateau potential originating in the distal dendrite may facilitate plasticity when paired with stimulation of proximal synapses . Multiple sources of inhibitory , GABAergic synapses on SPNs may regulate synaptic integration and spine calcium transients . The ionotropic GABAA synapses exhibit distinct spatial organization , with fast spiking interneurons ( FSIs ) targeting proximal dendrites and low threshold spiking interneurons ( LTSIs ) and SPN collaterals targeting distal dendrites , and they exhibit distinct temporal kinetics , with neurogliaform ( NGF ) interneurons exhibiting much slower activation and inactivation time constants ( Ibáñez-Sandoval et al . , 2011; Straub et al . , 2016; Tepper et al . , 2010 ) . To make predictions about synaptic integration and plasticity that may be relevant in vivo , it is critical to include the effects of inhibition in our investigation , as inhibitory signaling is extensive in the striatum . Thus , we investigated the effects of GABAA kinetics , location , and timing relative to clustered glutamatergic input on regulating plateau potentials and spine calcium elevation ( Figure 6 ) . Our results demonstrate that non-linear spine calcium dynamics depend strongly on the location of simultaneously stimulated GABAA synapses ( Figure 6B ) . We paired stimulation of a single GABAergic synapse on the dendritic shaft with simultaneous stimulation of excitatory synapses on a cluster of distal dendritic spines , and repeated simulations while varying the location of the single GABAergic synapse . We ran the same set of simulations for both the slow and fast GABAA kinetics . Distal GABAergic inputs near the site of clustered excitatory synaptic inputs have the strongest inhibitory effect on spine calcium elevation , reducing the response by ~50% for slow GABAA kinetics , whereas proximally located GABAergic synapses have little effect on spine calcium . The effect was similarly distance-dependent but weaker for fast GABAA kinetics , with a 25% reduction in spine calcium elevation . These results suggest that distally located GABAergic synapses from LTSIs , SPN collaterals , or NGF interneurons likely regulate the occurrence of supralinear spine calcium influx in response to clustered glutamatergic stimulation , whereas FSIs with proximal synapses do not regulate distal spine calcium dynamics . Additionally , our results show that the timing of GABAergic stimulation relative to clustered glutamatergic stimulation affects the resulting spine calcium elevation ( Figure 6A ) . We paired stimulation of a single GABAergic synapse located on the distal dendritic shaft with stimulation of excitatory synapses on a cluster of distal dendritic spines at the same dendritic location , and repeated simulations varying the timing between GABAergic and glutamatergic stimulation . Again , we ran the same simulations for GABAergic synapses with either slow or fast GABAA kinetics . For both slow and fast GABAA kinetics , the timing of inhibition relative to excitation strongly affected spine calcium elevation . For the slow GABAA kinetics , GABAergic input from 100 ms before to 25 ms after clustered glutamatergic stimulation diminished spine calcium elevation , whereas the fast GABAA kinetics have a much narrower temporal window for inhibiting the calcium response . Together , these results indicate that inhibition may strongly regulate spine calcium dynamics when active prior to or concurrent with clustered glutamatergic input . In particular , GABAA synapses with slow kinetics ( i . e . from NGF interneurons ) may be particularly potent regulators of dendritic integration and synaptic interactions in SPNs . As FSIs provide strong proximal inhibition to SPNs in vivo , we also investigated whether a train of proximal GABAergic inputs ( as opposed to a single input ) would affect cooperativity among stimulated spines or between dendritic branches ( Figure 6C ) . A train of 20 GABAergic inputs ( ISI = 3 ms ) was applied to the proximal , primary dendritic branch while 16 or 32 glutamatergic inputs per branch were dispersed over the entire branch on one or two neighboring tertiary branches with an ISI of 2 . 5 ms per branch . The onset of the FSI input train and the glutamatergic stimulation was simultaneous . Interestingly , we found that a train of proximal GABAergic stimulation enhanced spine calcium responses when glutamatergic inputs were sub- or near-threshold ( 16 spines/branch on one or two branches ) , but not when above threshold ( 32 spines/branch on two branches ) for supralinear spine calcium elevation . We assessed whether the effect of FSI input trains on spine calcium elevation depended on distance of stimulated spines from the soma ( Figure 6—figure supplement 1 ) , and we found that distal spines , but not proximal spines , exhibit elevated calcium in response to proximal FSI inputs relative to the control ( no FSI ) condition . Our finding that proximal GABAergic stimulation may enhance spine calcium elevation is consistent with experiments demonstrating that when SPNs are in the hyperpolarized downstate , GABAergic stimulation produces depolarization ( Blackwell et al . , 2003; Bracci and Panzeri , 2006 ) . This depolarization may propagate to distal dendritic spines to enhance supralinear spine calcium elevation in stimulated spines when SPNs are in a hyperpolarized downstate . Notably , although we observed above that inhibition close to excitatory inputs can reduce the magnitude of spine calcium elevation in stimulated spines ( likely by lowering dendritic propagation of potentials ) , inhibition neither abolished the plateau potential nor fully blocked supralinear spine calcium elevation . This raised the intriguing possibility that the reduced local membrane resistance caused by inhibition may also influence non-stimulated spines . Thus , we evaluated the effect of inhibition by measuring the peak calcium response in non-stimulated spines relative to the peak calcium response in stimulated spines in the presence or absence of a co-located and simultaneously stimulated GABAergic synapse during simultaneous excitatory stimulation of a distal cluster of dendritic spines . The most significant functional effect of distal GABAergic synaptic input is to enhance spatial specificity ( Figure 7 ) . When compared to the control ( no-GABAergic input ) condition , stimulation of a single GABAergic synapse on the distal dendrite reduces the ratio of peak calcium elevation in non-stimulated spines relative to the peak in stimulated spines ( Figure 7A ) . Further , GABAergic stimulation narrows the spatial extent of calcium influx in non-stimulated spines and in the dendritic shaft ( Figure 7B ) , reducing the ratio of non-stimulated spine peak calcium to stimulated spine peak calcium much more for distant than adjacent spines and thereby limiting the spatial extent of heterosynaptic calcium elevation . As calcium elevation can lead to both LTP and LTD , with lower levels of calcium elevation associated with LTD and higher levels associated with LTP , inhibition may critically regulate heterosynaptic plasticity by reducing calcium elevation in non-stimulated spines from a level that could lead to LTP to a level that could lead to LTD , thereby preserving synapse-specific potentiation . It is critical in computational modeling to assess the robustness of results to variations in parameter values . To assess the robustness of the effect of inhibition on synapse-specificity , we systematically varied conductances of voltage-gated and synaptic channels , calcium buffer quantities and pump densities , and spine neck axial resistance by ±10% and 20% . For each condition , we computed the specificity ratio as the ratio of peak calcium of non-stimulated spines to peak calcium of stimulated spines , such that a smaller value indicates higher specificity . We then divided the specificity ratio observed with inhibition by the specificity ratio observed without inhibition ( called GABA/No GABA specificity ratio ) , such that a value <1 indicates that inhibition enhances specificity , while a value >1 indicates that inhibition reduces specificity . Inhibition consistently enhanced spatial specificity across all parameter variations , except for a 20% decrease in NMDAR conductance ( Figure 7—figure supplement 1 ) . The magnitude of the effect of inhibition was most sensitive to NMDAR , CaR , and GABAR conductances . As expected , a larger GABAR conductance enhanced the GABA/No GABA specificity ratio . Larger inward currents reduced specificity , and the large effect with NMDAR and CaR conductances confirms our previous results showing that these are critical parameters for the supralinear spine calcium response . Surprisingly , decreases in NMDAR or CaR conductances also reduced the GABA/No GABA specificity ratio . When either CaR or NMDAR conductance is lowered , stimulation is below threshold , and the reduced activation of VGCCs greatly enhances spatial specificity in the absence of inhibition . The sensitivity to CaR and NMDAR conductances suggests that neurons may regulate the balance of these channels; thus , we repeated simulations with an increase of CaR and a decrease of NMDAR ( and vice versa , no attempt was made to balance these changes ) . As predicted , sensitivity to the paired parameter change was smaller than sensitivity to a single parameter change . In summary , this parameter sensitivity analysis suggests that the balance of NMDAR and CaR channels may fine-tune spatial specificity of spine calcium transients . In addition to robustness to parameter variations , it is important that our main result does not depend on underlying assumptions in the modeling method . Specifically , our model of calcium dynamics assumes that deterministic , one-dimensional diffusion and reaction equations are sufficient to capture the dynamics of calcium elevation in dendritic spines . We evaluated whether the following variations affected the central finding that inhibition enhances synapse-specificity: ( 1 ) spatial and temporal discretization step size; ( 2 ) variations in calcium diffusion rates; ( 3 ) facilitated calcium diffusion between spine neck and dendritic shaft; and ( 4 ) variations in calcium influx and efflux caused by channels and pumps . Importantly , these findings did not qualitatively change the main result that inhibition enhances spatial specificity ( Figure 7—figure supplement 2 ) . A critical functional question is whether inhibition regulates synaptic specificity for spatiotemporally distributed inputs , which are likely to occur in vivo . To answer this question , we applied both excitatory ( n = 200 or 300 ) and inhibitory ( n = 50 ) synaptic inputs with random spatial dispersion , with each individual synaptic input consisting of an independent Poisson process with an average frequency of 2 . 5 Hz . We assessed spine calcium with histograms of integrated calcium for the 1 s simulation duration , as well as scatter plots of peak values versus normalized duration ( i . e . integrated calcium concentration/peak calcium concentration ) for each spine . At the lower ( n = 200 ) activity level , synaptic specificity was robust with or without inhibition , with little overlap in integrated calcium and no overlap for peak versus duration scatter points ( Figure 7C ) . However , with higher ( n = 300 ) activity and no inhibition , synaptic specificity was reduced , as overlap in both the histogram and scatter points occurred for stimulated and non-stimulated spines . Notably , when inhibition was included during higher activity , synaptic specificity was increased , indicating that inhibition may function in vivo to maintain spatial specificity of spine calcium transients ( Figure 7D ) . To quantify the extent to which inhibition increased separation between stimulated and non-stimulated spine calcium responses , we performed a cluster analysis using duration and peak calcium as parameters . For the cluster analysis , we performed K-means clustering on the unlabeled data , and then computed the confusion matrix—the number of correctly and incorrectly clustered data points in the unlabeled cluster analysis compared to the real , labeled dataset , where the label is whether the spine received synaptic input . We evaluated the number of incorrectly clustered spines in the confusion matrix and computed the distance between clusters as a metric for the effect of inhibition on synaptic specificity . For the higher activity level without inhibition , the cluster analysis incorrectly identified the stimulation status of 58 spines , whereas with inhibition , the stimulation status of only 19 spines was incorrectly identified ( out of 3280 total spines ) . For the lower activity level , no spines were incorrectly identified , but inhibition increased the distance between cluster centroids by 5 . 3% . Altogether , these results suggest that although inhibition reduces calcium elevation in stimulated spines , its most significant effect is to enhance stimulus-specific calcium signaling and it may therefore regulate heterosynaptic plasticity .
We demonstrated that spatial and temporal patterns of synaptic input to SPNs produce spatially specific and stimulus-specific spine calcium responses in a biologically detailed computational model with sophisticated calcium dynamics . A sufficient number of excitatory inputs to a distal dendritic branch can cooperate to evoke supralinear spine calcium responses in the synaptically stimulated spines , whether the stimulated spines are clustered or dispersed along the distal portion of the branch . The resulting spine calcium responses exhibit specificity for synaptically stimulated spines , with an order of magnitude higher calcium influx than non-stimulated neighboring spines , and this spatial specificity is further enhanced by nearby GABAergic synaptic stimulation . Synaptic cooperativity exhibits dendritic-branch specificity , as synaptic inputs to one branch produce higher spine calcium responses than the same number of inputs distributed to multiple branches . Additionally , the calcium elevation in a proximal spine can be facilitated when the synaptic input to that spine occurs synchronous with or shortly following distally evoked plateau potentials , indicating potential interactions between distal and proximal synaptic input . Together , our results suggest that spine calcium elevation is a stimulus-specific signal which can differentiate between synaptically stimulated and non-stimulated spines . We predict that these distinct calcium responses will enable both input-specific LTP and heterosynaptic LTD , and that inhibition critically regulates heterosynaptic activity . In key ways , our results extend other experimental and modeling studies which have found that striatal spiny projection neurons exhibit plateau potentials in response to clustered excitation of neighboring spines on a distal dendritic branch ( Du et al . , 2017; Plotkin et al . , 2011 ) . First , our focus is on spine calcium dynamics , whereas previous studies focused on somatic voltage . Plotkin et al . ( 2011 ) did find robust calcium elevation at the site of clustered synaptic stimulation; however , our results additionally show the spatial specificity of spine calcium elevation . Second , our model predicts that spatial clustering within the branch is not required for supralinear spine calcium elevation or plateau potentials , if enough inputs are located distally on the same dendritic branch . SPNs receive few ( one to three ) synapses from any individual cortical neuron and convergent inputs from thousands of cortical neurons ( Kincaid et al . , 1998; Zheng and Wilson , 2002 ) , and the likelihood of clustered , synchronous inputs to occur in vivo is unknown , so the ability for dispersed inputs to a distal branch to also evoke supralinear responses may increase the likelihood of supralinear spine calcium transients occurring in vivo . Our finding that coordinated synaptic inputs produce supralinear spine calcium elevation suggests a mechanism for ‘cooperative LTP’ ( LTP produced by stimulation of multiple synaptic inputs in the absence of a postsynaptic action potential ) in striatal SPNs . Although the induction of cooperative LTP by clustered synaptic input has not been demonstrated in SPNs , it has been observed in pyramidal neurons in cortex and hippocampus ( Brandalise et al . , 2016; Golding et al . , 2002; Gordon et al . , 2006; Harnett et al . , 2012; Larkum et al . , 2009; Losonczy et al . , 2008; Makara and Magee , 2013; Schiller et al . , 2000; Weber et al . , 2016 ) . In pyramidal neurons , cooperative LTP requires NMDAR activation and can be induced by synaptic stimulation of a small number of neighboring spines on a thin dendritic branch . Similarly , we observe that supralinear spine calcium responses in SPNs are NMDAR-dependent and are achieved by coordinated synaptic inputs to a ( thin ) distal branch . The dependence of supralinear spine calcium elevation on spatially and temporally distributed inputs also is similar to observations in pyramidal neurons . Spatially and temporally distributed inputs to a single distal dendritic branch in CA3 pyramidal neurons have been shown to evoke NMDA spikes , for inputs spatially distributed within a 60 micron length of dendritic branch or temporally distributed with intervals up to 2 ms ( Makara and Magee , 2013 ) . These results are consistent with our findings that inputs spatially dispersed over the distal half of a dendritic branch or temporally dispersed with a mean interval of 2 . 5 ms can still evoke supralinear spine calcium elevations . Although LTP induced by supralinear spine calcium elevation in response to coordinated synaptic inputs has not been experimentally demonstrated in the striatum , results from our model along with findings from other brain regions predict that striatal SPNs will exhibit cooperative LTP , and that strict spatial clustering and precise temporal synchrony of synaptic inputs may not be required . Our finding that non-stimulated neighboring spines exhibit intermediate levels of calcium elevation suggests that these spines may undergo heterosynaptic plasticity ( potentiation or depression ) or metaplasticity . For instance , synaptic stimulation and induction of plasticity can induce metaplastic changes at nearby , non-stimulated spines in hippocampal pyramidal neurons ( Govindarajan et al . , 2011; Harvey and Svoboda , 2007 ) . In contrast to metaplasticity , a recent hippocampal study found heterosynaptic depression of nearby non-stimulated spines when a cluster of synaptically stimulated spines underwent potentiation ( Oh et al . , 2015 ) . Our results suggest that coordinated synaptic stimulation of clustered spines causes sufficient depolarization of neighboring non-stimulated spines for calcium elevation via VGCCs , which may underlie heterosynaptic changes ( LTP or LTD ) in nearby spines . Calcium-dependent plasticity may also occur in the dendritic shaft . Dendritic spikes in pyramidal neurons can induce branch-specific potentiation of dendritic branch strength , via NMDAR-dependent regulation of dendritic potassium ( Kv4 . 2 ) channels ( Losonczy et al . , 2008 ) . Our results showing a moderate calcium elevation in the distal dendritic shaft suggest a mechanism linking NMDAR-dependent plateau potentials to dendritic branch strength plasticity . As the same ( Kv4 . 2 ) potassium channels are expressed in SPN dendrites , we predict that branch-strength plasticity may also occur in SPNs . Our most novel result is that inhibition enhances spatial specificity . Although inhibition also attenuates the supralinear response to coordinated synaptic inputs , the reduction in non-stimulated spines is more significant . Our finding that inhibitory synaptic input located distally on the dendritic branch attenuates the supralinear response to coordinated excitatory synaptic inputs is consistent with recent studies in the striatum ( Du et al . , 2017 ) and in cortical pyramidal neurons ( Doron et al . , 2017 ) . However , we additionally found that inhibition enhances the spatial specificity of spine calcium elevation in response to coordinated excitatory inputs , a novel finding with implications for calcium-dependent signaling pathways . The role of distal ( but not proximal ) inhibition in regulating dendritic and spine calcium transients is consistent with a study showing that distal inhibition regulates bAP-induced calcium influx in cortical pyramidal neurons ( Marlin and Carter , 2014 ) , and a study showing that distal inhibitory inputs exhibit stronger inhibition of excitatory potentials than proximal inhibitory inputs ( Gidon and Segev , 2012 ) . Our findings also are consistent with a model study showing that inhibition can regulate the direction of homosynaptic ( input-specific ) plasticity for dendritic shaft synapses ( Bar-Ilan et al . , 2012 ) , but we additionally show that inhibition can regulate both heterosynaptic and homosynaptic calcium signaling in dendritic spines . Our results suggest that the complex functional roles of inhibition may include regulating the balance of LTP and LTD within the dendritic branch by , for instance , changing heterosynaptic LTP to LTD; heterosynaptic LTD to no change in synaptic strength; or limiting the spatial extent in the dendrite of heterosynaptic plasticity . In agreement with Du et al . ( 2017 ) , inhibition located distally on the dendritic branch ( near the site of excitatory inputs ) within a limited time window relative to excitatory inputs exerted the strongest inhibitory effect . These results suggest that interneurons targeting distal dendrites , such as LTSIs or neuropeptide Y-expressing NGFs , as well as SPN collaterals , may regulate cooperative LTP to support cell assembly formation in the striatum ( Ibáñez-Sandoval et al . , 2011; Ponzi and Wickens , 2013; Straub et al . , 2016; Tepper et al . , 2010 ) . Importantly , our results are robust to parameter variations . The enhancement of spatial specificity by inhibition is most sensitive to NMDAR , CaR , and GABAR conductances , consistent with the NMDAR and CaR channels being the major sources of depolarization and calcium influx on synaptically stimulated and non-stimulated spines , respectively . The main finding that inhibition enhances spatial specificity was observed for all parameter variations ( except for a 20% decrease in NMDAR conductance ) , although the amount that inhibition enhanced spatial specificity did depend on parameter values . Interestingly , both increases and decreases in NMDAR and CaR conductance reduced the ability for inhibition to increase spatial specificity . In the case of increased NMDAR or CaR conductance , greater depolarization would spread into non-stimulated spines , and the ability of the inhibitory current to counter the depolarization would be reduced . In the case of decreased NMDAR or CaR conductance , synapse-specificity is increased in the control ( no inhibition ) condition because weaker depolarization leads to significantly lower calcium influx to non-stimulated spines; thus , the ability of inhibition to further enhance spatial specificity is reduced . The sensitivity to NMDAR and CaR conductances is consistent with these channels being critical for plateau potential generation and duration in SPNs ( Plotkin et al . , 2011 ) , and suggests that an optimal balance of inward currents may be necessary for spatial specificity . The degree to which inhibition enhances spatial-specificity also is quite sensitive to GABAR conductance; however , the qualitative outcome is consistent for the entire range of GABAR conductances we evaluated . Therefore , the strength of inhibitory synapses may critically regulate spatial specificity and synaptic plasticity . We predict that calcium-dependent plasticity of inhibitory synapses , as has been observed in pyramidal neurons ( Chiu et al . , 2018 ) , may regulate the degree of spatial specificity in a branch-specific manner . The computational modeling methods we employed assume that calcium dynamics in dendritic spines can be modeled deterministically with one-dimensional axial diffusion of calcium and mobile buffers . These assumptions may limit our ability to fully capture calcium dynamics in dendritic spines . However , the scope of the model , which includes the entire dendritic morphology with thousands of dendritic spines , precludes simulating stochastic reaction kinetics and diffusion in three-dimensions . Crucially , we verified that our conclusions did not depend on limitations of our modeling methodology . Specifically , our method may underestimate the calcium diffusion rate from narrow spine necks to the dendritic shaft ( Stiles et al . , 1996 ) . However , neither varying the calcium diffusion rate nor facilitating spine-neck diffusion affected our main result that spine calcium transients exhibit spatial specificity which is enhanced by inhibition . Our examination of mechanisms underlying spatial specificity showed that calcium permeable ligand-gated and voltage-gated ion channels on synaptically stimulated and non-stimulated spines , respectively , were the main contributors to the amplitude of the calcium transients , and that diffusion only had a limited effect on the duration of calcium elevation in non-stimulated spines . The calcium modeling method of the GENESIS simulator also does not account for the local radius of curvature of the cell membrane , which may affect fluxes and reactions ( i . e . calcium pumps ) at the membrane ( Bell et al . , 2018; Rangamani et al . , 2013 ) . However , our main result—that inhibition enhances spatial specificity—is robust to variations in parameters affecting calcium influx and efflux . Thus , although the exact magnitude of calcium concentration we report may be limited in accuracy because of methodological limitations , the key conclusions based on the ratios of calcium concentration in stimulated versus non-stimulated spines are qualitatively unchanged . Our conclusions are also supported by the model’s ability to reproduce observations from calcium imaging experiments for stimulation of individual spines . Further , the quantity of spine calcium that is free , bound , pumped or diffuses out during synaptic stimulation in our model exhibits similar dynamics to a detailed three-dimensional model of spine calcium dynamics ( Bartol et al . , 2015 ) . However , as our model does not include variations in spine geometry , it would be of interest for future work to consider the effects of realistic spine geometries ( including curvatures ) and three-dimensional reaction-diffusion modeling on spine calcium transients during dendritic plateau potentials . Together , our results have implications for striatal function and plasticity in vivo . With inputs likely to be spatiotemporally distributed , our results suggest that fewer than 10 spines per branch may produce supralinear spine calcium when a few hundred total synapses on the dendritic tree are stimulated , similar to estimates of the number of synaptic inputs driving striatal upstates ( Blackwell et al . , 2003 ) . Inhibitory inputs in vivo may control stimulus-specific and heterosynaptic plasticity , serving as a homeostatic mechanism to balance potentiation of excitatory inputs on a dendritic branch-specific level . Our recent work has shown that a calcium-based plasticity rule reproduces diverse in vitro plasticity protocols ( Jędrzejewska-Szmek et al . , 2017 ) . Qualitatively extending this calcium-based plasticity rule to our results using clustered synaptic inputs ( using the same threshold values reported ) would predict that , in the absence of inhibition , supralinear spine calcium transients would produce LTP in synaptically stimulated spines , but also that non-stimulated neighboring spines would be above the LTP threshold and undergo heterosynaptic LTP . In this scenario , inhibition could reduce spine calcium elevation in non-stimulated spines to be above the LTD threshold but below the LTP threshold , switching the response from heterosynaptic LTP to heterosynaptic LTD for many ( although not all ) non-stimulated spines , whereas the responses in synaptically stimulated spines would remain above the LTP threshold . In the more in vivo-like conditions we simulated with spatiotemporally distributed inputs ( Figure 7C–D ) , the spine calcium values we observed are more consistent with the plasticity rule thresholds , predicting LTP in synaptically stimulated spines and LTP , LTD , or no change in non-stimulated spines , with inhibition switching heterosynaptic LTP to LTD for non-stimulated spines . Extending this calcium-based plasticity rule to assess plasticity during repeated in vivo-like inputs may yield key predictions about how spatiotemporally distributed in vivo activity with trial-to-trial variability induces stable , synapse-specific plasticity .
A biophysically detailed SPN model we previously published ( Jędrzejewska-Szmek et al . , 2017 ) was modified for this study ( Figure 1A ) . The morphology consisted of a single cylindrical soma ( 11 . 3 µm length , 22 . 6 µm diameter ) with four primary dendrites ( 12 µm length , 2 . 25 µm diameter ) , each branching twice into a total of eight secondary dendrites ( 14 µm length , 1 . 4 µm diameter ) and 16 tertiary dendrites ( 198 µm length , tapered diameter from 0 . 89 µm proximally to 0 . 3 µm diameter distally [Wilson , 1992] ) . Tertiary dendritic branches were subdivided into 3 µm long compartments to accurately model interactions among neighboring dendritic spines ( Gulledge et al . , 2012 ) . Spines were explicitly modeled as a cylindrical head ( 0 . 5 µm diameter , 0 . 5 µm length ) and neck ( 0 . 12 µm diameter , 0 . 5 µm length ) and were distributed on secondary and tertiary dendritic branches with a density of 1 spine/µm , for a total of 3280 spines in the entire model . Membrane resistivity and capacitivity were set to 1 . 875 ohms-m2 and 0 . 01 Farads/m2 , respectively . Axial resistance was set to 1 . 25 ohm-m for all compartments except for spine neck compartments , which were set to 11 . 3 ohm-m to achieve a neck resistance of 500 MΩ , as estimated from experimental data ( Harnett et al . , 2012 ) . Passive parameters were determined by fitting the model to hyperpolarizing current injection ( Figure 1B ) . As described previously ( Jędrzejewska-Szmek et al . , 2017 ) , the model includes the following voltage-gated sodium and potassium ion channels ( Table 1 ) : A fast sodium channel ( NaF ) ( Ogata and Tatebayashi , 1990 ) ; fast ( Kaf/Kv4 . 2 ) ( Tkatch et al . , 2000 ) and slow ( Kas/Kv1 . 2 ) ( Shen et al . , 2004 ) A-type potassium channels; an inwardly rectifying potassium channel ( Kir ) ( Steephen and Manchanda , 2009 ) ; and a resistant persistent potassium channel ( Krp ) ( Nisenbaum and Wilson , 1995 ) . Additionally , the model includes a big conductance voltage- and calcium-activated potassium channel ( BK ) ( Berkefeld et al . , 2006 ) and a small conductance calcium-activated potassium channel ( SK ) ( Maylie et al . , 2004 ) . Six VGCCs are also included in the model ( Table 1 ) : CaR ( Brevi et al . , 2001; Foehring et al . , 2000 ) , CaN ( Cav2 . 2 ) ( Bargas et al . , 1994; Kasai and Neher , 1992; McNaughton and Randall , 1997 ) , CaL1 . 2 ( Cav1 . 2 ) ( Bargas et al . , 1994; Kasai and Neher , 1992; Tuckwell , 2012 ) , CaT3 . 2 ( Cav3 . 2/ α1H ) ( McRory et al . , 2001 ) , CaT3 . 3 ( Cav3 . 3/ α1I ) ( McRory et al . , 2001 ) , and CaL1 . 3 ( Cav1 . 3 ) ( Tuckwell , 2012 ) . Channel kinetic equations and parameters are similar to our previously reported model ( Jędrzejewska-Szmek et al . , 2017 ) , except we converted the previously nonspecific CaT channel to CaT3 . 3 and added a CaT3 . 2 channel with the following parameters: m vhalf = −43 . 15 mV; m vslope = −5 . 43 mV; h vhalf = −73 . 9 mV; h vslope = 2 . 76 mV; m tau alpha rate = 160 , 000/V/s; m tau alpha vhalf = 112; m tau alpha vslope = 11; m tau beta rate = 8500; m tau beta vslope = 12 . 5; m tau baseline offset = 0 . 0009 s; htau=22 . 25+0 . 0455e-Vm mV7 . 46 ( ms ) . Channel conductance values were tuned to reproduce electrophysiology recordings ( Table 1 ) . The soma and dendrites contain NaF , Kaf , Kas , Krp , and BK channels; SK channels are present in the soma and dendritic spines ( Higley and Sabatini , 2010 ) . Calcium currents are modeled with the Goldman-Hodgkin-Katz ( GHK ) current equation to accurately account for the calcium driving potential . Calcium-dependent inactivation ( CDI ) was implemented for CaR , CaN , CaL1 . 2 , and CaL1 . 3 channels ( Liang et al . , 2003 ) . CaT channels were located in spines and distal dendrites ( Carter and Sabatini , 2004; McRory et al . , 2001; Plotkin et al . , 2011 ) , but not soma or proximal dendrites ( Bargas et al . , 1994 ) . CaR , CaL1 . 2 , and CaL1 . 3 channels were located in soma , dendrites , and spines ( Carter and Sabatini , 2004; Higley and Sabatini , 2010 ) . CaN channels were restricted to the soma ( Carter and Sabatini , 2004 ) . Calcium channel densities were tuned to experimentally reported calcium imaging for synaptic activation of a single spine ( Higley and Sabatini , 2010; Shindou et al . , 2011 ) and back-propagating AP-induced calcium influx into dendrites ( Carter and Sabatini , 2004; Day et al . , 2008; Kerr and Plenz , 2002; Shindou et al . , 2011 ) and spines ( Carter and Sabatini , 2004; Shindou et al . , 2011 ) ( Figure 1B ) . Contribution of specific channels to calcium influx was tuned to experiments blocking specific channel types ( Carter and Sabatini , 2004; Higley and Sabatini , 2010 ) . Intracellular calcium concentration , diffusion , buffers , and pumps were modeled with the difshell object in GENESIS ( Bower and Beeman , 1998 ) . Calcium had a diffusion constant of 200 µm2/s ( Allbritton et al . , 1992 ) . One-dimensional radial diffusion was implemented in the dendrites and soma by subdividing each cylindrical electrical compartment into a series of concentric shells; the submembrane shell had a diameter of 0 . 1 µm , and successive shells doubled in diameter ( Anwar et al . , 2014 ) . One-dimensional axial diffusion was modeled in the spines and necks by subdividing the spine and neck electrical compartment into six cylindrical slabs , three for each compartment . Diffusion was also implemented between the spine neck and the submembrane shell of the dendrite . Calcium extrusion was implemented with Michaelis-Menten models of a plasma membrane calcium ATPase ( PMCA ) in the soma , dendrites , and spines , and a sodium-calcium exchanger ( NCX ) in the spines ( Table 2 ) . Calcium-permeable ion channels provide calcium influx to the spine head slabs and the dendritic/somatic submembrane shell . In the spine head , CaL1 . 3 ( Olson et al . , 2005 ) and calcium-permeable synaptic channels provide calcium influx to the outermost calcium slab ( the postsynaptic density ) , while CaL1 . 2 , CaR , and CaT provide calcium influx to the middle slab . Additionally , the SK channel in spines was dependent on calcium concentration of the middle slab . Calcium buffers ( Table 2 ) were modeled with the difbuffer object in GENESIS , which allows for buffering of calcium within difshells and diffusion of buffers ( calcium-bound or free ) between difshells . The model included the endogenous mobile buffers calbindin and calmodulin ( N and C terminals ) , as well as an endogenous immobile buffer that was required to avoid unrealistic calcium elevations ( Matthews et al . , 2013; Matthews and Dietrich , 2015 ) . The endogenous buffer quantities ( Table 2 ) give a buffer capacity ratio close to 90 , which is consistent with experimental estimates of buffer capacity in SPN spines and dendrites ( Carter and Sabatini , 2004 ) . Exogenous calcium buffers ( calcium indicator dyes ) were included in simulations when tuning to experimental calcium-imaging data . NMDAR and AMPAR synaptic channels were included on the spine heads and contributed calcium to the outermost spine head difshell . The fractional calcium currents were 5% of the total NMDAR current ( implemented with the GHK current equation ) and 0 . 1 % of the total AMPAR current . The AMPAR/NMDAR maximal conductance ratio was set to 1 . 0 , and the conductances were set to achieve a unitary somatic PSP of ~2 mV , similar to the uncaging evoked EPSPs in Plotkin et al . , 2011 . Calcium-dependent inactivation of the NMDAR channel was implemented based on equations in a published model ( Farinella et al . , 2014 ) . In simulations that included GABAA stimulation , GABAA synaptic channels were included on the dendritic shaft with a maximal conductance of 1 . 2 nS . GABAA kinetics were either fast , consistent with synapses from fast spiking interneurons , low-threshold spiking interneurons , or SPN collaterals ( Straub et al . , 2016 ) , or slow , consistent with NPY-neurogliaform synapses ( Ibáñez-Sandoval et al . , 2011 ) . Simulations were done with various spatiotemporal patterns of synaptic input as described in the results . In cases with asynchronous stimulation , the order of spine stimulation was randomly assigned . For simulations with random temporal dispersion , the ISI consisted of exponentially distributed intervals with an average ISI of 2 . 5 , 5 , or 10 ms per branch; for example the actual ISI for the 10 ms per branch case with two total branches stimulated was 5 ms , and with four total branches stimulated was 2 . 5 ms . This was done to make the total time of stimulation independent of number of branches stimulated , and to facilitate comparisons between simulations on a per-branches-stimulated basis ( i . e . Figure 4 ) . The minimum ISI values drawn from exponential distributions were unconstrained . The temporal order of asynchronously stimulated spines was randomly selected and followed no spatial pattern . The model was simulated in GENESIS ( Bower and Beeman , 1998 ) with a timestep of 0 . 01 ms . For simulations evaluating discretization , the timestep was increased to 0 . 001 ms , dendritic compartments were subdivided to 1 ( rather than the standard 3 ) micron length compartments , the number of calcium shells in dendrites was increased using a constant shell depth equal to the submembrane shell depth , and calcium slabs in the spine head and neck were increased from three to six slabs . For simulations evaluating the effect of a coupling surface area from the spine head to neck or the spine neck to dendritic shaft , the GENESIS source code was changed to calculate the surface area for diffusion between two difshells ( or difbuffers ) using the geometric average of the surface areas of the two shells , rather than the GENESIS default of the minimum of the two shell surface areas . Analysis was done in Python 2 . 7 , using the Numpy , SciPy , Pandas , and Matplotlib python packages . Model simulation and analysis files are available on ModelDB . For statistical analysis of simulus specificity in Figure 7C–D , a cluster analysis was performed in SAS9 . 4 , using duration and peak calcium as parameters to quantify the extent to which inhibition increased separation between stimulated and non-stimulated spines . The procedure FASTCLUS was used to create two clusters , and the output gave a measure of the distance between clusters of stimulated and non-stimulated spines . In addition , the procedure FREQ was applied to the output of the cluster analysis to generate the confusion matrices and identify the number of incorrectly labeled spines . | How do we form new memories ? The human brain contains almost 90 billion neurons , which communicate with one another at junctions called synapses . Each neuron has a shape a little like that of a tree , and is covered in branches called dendrites . Synapses typically form between the end of one neuron and a dendrite on another . Most scientists believe that the brain forms new memories by changing the strength of these synapses . But a number of questions remain about how this process works . There are two types of synapses: excitatory and inhibitory . When an excitatory synapse becomes active , calcium ions flow into the dendrite of the receiving neuron . The calcium ions then trigger processes inside the cell that are essential for changing the strength of the synapse , and thus forming a memory . But what happens when an inhibitory synapse becomes active ? How does this affect memory ? Additionally , each neuron forms synapses with thousands of others , with several synapses on a single dendrite . To form a memory about a specific experience , the brain must strengthen only the synapses that relate to that experience . How does the brain manage to target these synapses specifically ? Do the synapses need to be clustered on the same dendritic branch , or can they be spread apart ? And do all the synapses need to be active at exactly the same time ? Dorman et al . investigated these questions by developing a computer model of a neuron . Testing the model revealed that the synapses related to an experience do not all need to be active at exactly the same time to form a memory . Moreover , the synapses can be spread across multiple dendrites . Finally , the model showed that inhibitory synapses are critical for preventing calcium ions from spreading within dendritic branches and entering inactive synapses . This ensures that only the synapses active during a specific experience become stronger . Many brain disorders , including substance abuse and addiction , involve errors in the processes that underlie learning and memory . By increasing our understanding of how the structure of brain cells supports these processes , the current findings could one day lead to better treatments for these and other disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2018 | Inhibition enhances spatially-specific calcium encoding of synaptic input patterns in a biologically constrained model |
Transcriptional elongation by RNA polymerase ( Pol ) II is essential for gene expression during cell growth and differentiation . The positive transcription elongation factor b ( P-TEFb ) stimulates transcriptional elongation by phosphorylating Pol II and antagonizing negative elongation factors . A reservoir of P-TEFb is sequestered in the inactive 7SK snRNP where 7SK snRNA and the La-related protein LARP7 are required for the integrity of this complex . Here , we show that P-TEFb activity is important for the epithelial–mesenchymal transition ( EMT ) and breast cancer progression . Decreased levels of LARP7 and 7SK snRNA redistribute P-TEFb to the transcriptionally active super elongation complex , resulting in P-TEFb activation and increased transcription of EMT transcription factors , including Slug , FOXC2 , ZEB2 , and Twist1 , to promote breast cancer EMT , invasion , and metastasis . Our data provide the first demonstration that the transcription elongation machinery plays a key role in promoting breast cancer progression by directly controlling the expression of upstream EMT regulators .
Cancer cells have fundamentally altered gene expression profiles that drive their pathogenic features . In eukaryotes , gene transcription is mainly performed by RNA polymerase ( Pol ) II and can be controlled at multiple stages including pre-initiation , initiation , elongation , and termination ( Shilatifard et al . , 2003; Sims et al . , 2004 ) . Transcription elongation has recently been demonstrated to play a critical role in regulating cell growth and differentiation . In Drosophila and human embryonic stem cells , a large number of genes involved in cell growth , renewal , and differentiation are found to be controlled at the elongation stage ( Guenther et al . , 2007; Muse et al . , 2007; Zeitlinger et al . , 2007 ) . Moreover , in mammalian cells , the transcription elongation machinery has also been implicated in the regulation of cell proliferation and differentiation ( Zhou and Yik , 2006; Romano and Giordano , 2008 ) . Shortly after initiation of transcription , Pol II pauses near the transcription start site largely due to the actions of two negative transcription elongation factors NELF and DSIF ( Peterlin and Price , 2006 ) . The human positive transcription elongation factor b ( P-TEFb ) reverses this block and stimulates transcriptional elongation by phosphorylating the two negative elongation factors as well as the C-terminal domain ( CTD ) of the largest subunit of Pol II . These modification events antagonize the actions of the negative elongation factors and also promote co-transcriptional mRNA processing ( Zhou et al . , 2012 ) . P-TEFb is a heterodimer composed of cyclin-dependent kinase 9 ( CDK9 ) and its regulatory partner Cyclin T1 or T2 ( CycT1 or T2 ) . The activity of P-TEFb is stringently maintained in a functional equilibrium in cells to accommodate transcriptional demands for different biological activities ( Zhou and Yik , 2006 ) . Three P-TEFb-containing complexes have been identified , including the inhibitory 7SK snRNP complex and two active transcription complexes: the Brd4-P-TEFb complex and the super elongation complex ( SEC ) ( Zhou et al . , 2012; Lu et al . , 2013b ) . Under normal growth conditions , more than half of nuclear P-TEFb is sequestered in the catalytically inactive 7SK snRNP , which also contains the 7SK snRNA , HEXIM1 ( or the homologous HEXIM2 ) , MePCE , and LARP7 . This complex represents the major cellular reservoir of inactive P-TEFb ( Zhou and Yik , 2006; Zhou et al . , 2012 ) . Within 7SK snRNP , the 7SK snRNA serves as a central scaffold that coordinates key protein–protein interactions and allows HEXIM1/2 to inhibit CDK9 ( Yik et al . , 2003 ) . The La-related protein LARP7 binds to nearly all the nuclear 7SK snRNA via the 3′-UUU-OH sequence and protects it against exonuclease cleavage ( He et al . , 2008; Krueger et al . , 2008 ) . Stable RNAi-mediated knockdown of LARP7 causes an almost complete depletion of 7SK snRNA , leading to disruption of 7SK snRNP and a shift of P-TEFb toward the active state ( He et al . , 2008 ) . Considering that the 7SK snRNP is the primary source of suppressed P-TEFb , the level of LARP7 directly affects the amount of active P-TEFb , thereby playing a key role in controlling P-TEFb activity . In response to a number of conditions/agents that globally impact growth and differentiation , P-TEFb is released from the 7SK snRNP and recruited to chromatin templates by the bromodomain protein Brd4 , which binds to acetylated histones and the Mediator complex to promote transcription and cell cycle progression ( Yang et al . , 2005; Mochizuki et al . , 2008; Yang et al . , 2008 ) . In addition to Brd4 , P-TEFb released from 7SK snRNP can also be recruited by a number of gene-specific transcription factors such as the HIV-1 Tat protein and the mixed lineage leukemia ( MLL ) fusion proteins to form the multi-subunit SEC complex to specifically and efficiently activate their target genes ( He et al . , 2010; Sobhian et al . , 2010; Lu et al . , 2013a ) . Accumulating data have implicated the involvement of various components of P-TEFb-containing complexes in human cancer . For example , several components of the P-TEFb-containing SEC , such as AFF1 , AFF4 , ELL1 , AF9 , and ENL , are all translocation partners of MLL and are important for MLL-based leukemogenesis ( Lin et al . , 2010; Yokoyama et al . , 2010; Smith et al . , 2011 ) . In addition , Brd4 has also been considered a promising therapeutic target in acute myeloid leukemia ( AML ) because of its ability to sustain P-TEFb-dependent c-Myc expression ( Blobel et al . , 2011; Zuber et al . , 2011 ) . Finally , several lines of evidence have implicated the control of P-TEFb by the 7SK snRNP in human breast cancer . First of all , HEXIM1 has been proposed as an inhibitor of breast cell growth since its expression is downregulated by estrogens in breast tumors ( Wittmann et al . , 2003 ) . Moreover , microsatellite instability ( MSI ) -induced frameshift mutations in the LARP7 gene have been detected in a significant population of gastric cancer samples , implicating a potential tumor suppressor role of LARP7 in cancers ( Mori et al . , 2002 ) . Consistent with this result , we have previously shown that LARP7 knockdown in the mammary epithelial cell line MCF10A disrupts cell polarity and blocks morphological differentiation when cultured in the three-dimensional laminin-rich extracellular matrix ( 3D IrECM ) ( He et al . , 2008 ) . Despite these observations , virtually nothing is known about whether P-TEFb and its associated factors may play a key role during human cancer progression . In this study , we investigated the function of the P-TEFb functional equilibrium in controlling the epithelial–mesenchymal transition ( EMT ) , invasion , and metastasis of human breast cancer . By knocking down LARP7 , we released P-TEFb from the 7SK snRNP and stimulated the P-TEFb-dependent transcription of EMT-related genes , resulting in breast cancer EMT and enhanced invasion and metastasis . Our analyses have revealed a strong causative relationship between the invasive phenotypes of human breast cancer and P-TEFb activation by disrupting the 7SK snRNP . Our study has thus provided the first demonstration that the transcription elongation machinery and the P-TEFb network play critical roles in regulating tumor progression , EMT , and metastasis by directly controlling the expression of EMT/metastasis-related genes .
To investigate whether P-TEFb and its associated factors are involved in human breast cancer progression , we first examined their expression patterns in the publicly accessible Oncomine microarray database . Of the known components in the three major P-TEFb-containing complexes , the 7SK snRNP , the Brd4-bound complex , and the SEC , only LARP7 and HEXIM1 , two signature components of the 7SK snRNP , showed consistent alteration in human breast cancer tissues . In two independent clinical data sets containing LARP7 information ( Zhao et al . , 2004; Finak et al . , 2008 ) , LARP7 expression was markedly reduced in breast cancer tissues , especially in the invasive carcinoma , when compared with the matched normal tissues ( Figure 1A ) . As downregulation of HEXIM1 in human breast cancer has been reported previously ( Wittmann et al . , 2003 ) , we focused on LARP7 in this study . 10 . 7554/eLife . 02907 . 003Figure 1 . LARP7 is significantly downregulated in invasive human breast cancer tissues and cells . ( A ) Box plots show decreased levels of LARP7 in invasive breast carcinoma ( left ) , invasive ductal carcinoma , and lobular carcinoma ( right ) compared with normal breast tissues in two microarray data sets . **: the p values ( p<0 . 01 , compared with normal breast tissues ) were determined by the Student's t test . ( B ) Kaplan–Meier analysis of overall survival and recurrence-free survival of breast cancer patients stratified by the expression of LARP7 . The p values were calculated by the log-rank test . ( C ) Immunohistochemical staining of LARP7 in normal human mammary tissue ( n = 6 ) , ductal carcinoma in situ ( DCIS ) ( n = 14 ) , and invasive ductal carcinoma ( n = 120 ) . The intensity of LARP7 staining was quantified using ImageJ Plus and shown in the box plot below . Scale bars represent 40 μm . **: the p value ( p<0 . 01 , compared with normal breast and DCIS tissues ) was determined by the Student's t test . ( D ) Western blotting ( WB ) analysis of the levels of LARP7 , phospho-Ser2 ( pSer2 ) , total Pol II , CyclinT1 , CDK9 , and HEXIM1 in various breast cancer cell lines ( upper panels ) and Northern blotting ( NB ) analysis of 7SK snRNA levels ( lower panels ) . Tubulin , 28S and 18S RNAs were used as loading controls . Expression of LARP7 , pSer2 of Pol II , and 7SK RNA was quantified , normalized to that in EpH4 cells , and shown in the graph to the right . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 00310 . 7554/eLife . 02907 . 004Figure 1—figure supplement 1 . qRT-PCR analysis of LARP7 mRNA levels in untransformed MCF10A and various breast cancer cell lines . PCR values were normalized to the levels of β-actin . Data are presented as the mean ± SD from three independent measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 004 We further analyzed the NKI295 breast cancer microarray data set that contains information on the clinical outcomes of patients ( van de Vijver et al . , 2002 ) to investigate the correlation between LARP7 level and clinical characteristics . Statistical analysis revealed that downregulation of LARP7 correlated with features of advanced cancer progression including estrogen receptor ( ER ) status , tumor size , and metastasis ( Table 1 ) . In particular , higher LARP7 levels were observed in older ( p=0 . 003 ) and ER-positive ( p=0 . 002 ) patients , and lower LARP7 expression was shown in larger ( p=0 . 012 ) , poorly differentiated ( p=0 . 05 ) and metastatic ( p=0 . 001 ) tumors . More importantly , high levels of LARP7 were significantly associated with increased overall survival and recurrence-free survival ( Figure 1B ) . Thus , downregulation of LARP7 correlates with breast cancer progression , metastasis , and poor prognosis . 10 . 7554/eLife . 02907 . 005Table 1 . Association between clinical characteristics and LARP7 levels in breast cancerDOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 005CharacteristicLow levels of LARP7 ( n = 148 ) High levels of LARP7 ( n = 147 ) p valueAge <45 yr95700 . 003† ≥45 yr5377Tumor diameter <20 mm45660 . 012† ≥20 mm10381ER status Negative46230 . 002* Positive102124No . of positive nodes 082690 . 117 1–34461 ≥42117Differentiation High31440 . 05† Mediate4854 Poor6949Metastasis No831110 . 001* Yes6536*p≤0 . 01 . †p≤0 . 05 were determined by chi-square test . We next examined LARP7 protein levels by immunohistochemistry in a human breast cancer tissue array containing 150 duplicated samples of normal human breast tissue , benign tumors , ductal carcinoma in situ ( DCIS ) , and invasive ductal carcinoma of varying pathological grades . In normal mammary tissues , LARP7 was highly expressed in the epithelial cells of the mammary lobuli and terminal ducts ( Figure 1C ) . This high expression was also observed in DCIS samples , but significantly attenuated in the invasive ductal carcinoma samples . These data confirm that LARP7 protein levels are also reduced during breast cancer progression , and thus suggest that LARP7 may play a potential tumor suppressor role in breast cancer . The association between cancer progression and reduced LARP7 expression was also confirmed in a panel of breast cancer cell lines . LARP7 was expressed at relatively high levels in the untransformed mammary epithelial cell lines ( MCF10A and EpH4 ) and noninvasive breast cancer cell lines ( MCF7 , BT474 , T74D , and ZR75B ) , but was markedly reduced in all four invasive and metastatic cancer lines examined ( MDA-MB-468 , BT549 , MDA-MB-231 , and MDA-MB-435; Figure 1D ) . This reduction was partially due to a decrease in the LARP7 mRNA levels as shown by qRT-PCR ( Figure 1—figure supplement 1 ) . Consistent with the demonstration that LARP7 is necessary for the integrity of the 7SK snRNA ( He et al . , 2008 ) , the reduced LARP7 expression in invasive breast cancer cell lines was accompanied by a decrease in 7SK snRNA ( Figure 1D ) . Meanwhile , phosphorylation of the Pol II CTD at Ser2 positions ( pSer2 ) strongly increased in the invasive breast cancer cell lines , indicating an increase in the P-TEFb kinase activity . Since the variations in LARP7 expression did not affect the cellular levels of CDK9 and Cyclin T1 , this increase in CDK9 kinase activity is likely due to the release of P-TEFb from the inhibitory 7SK snRNP as a result of reduced LARP7 and 7SK snRNA levels . Interestingly , HEXIM1 expression did not correlate with the malignant features of the breast cancer cell lines; it was moderately reduced in three invasive cancer lines and ZR75B non-invasive cancer line , but remained high in the metastatic MDA-MB-231 cells ( Figure 1D ) . Taken together , these data suggest a model that LARP7 is downregulated in invasive human breast cancer cells , leading to a decrease in 7SK snRNA and subsequently 7SK snRNP . This results in release of P-TEFb from the inhibitory 7SK snRNP and an increase in the phosphorylation of the P-TEFb substrates . Thus , active P-TEFb appears to play a key role in promoting breast cancer progression . To determine the precise role of LARP7 in breast cancer development , we first knocked down the expression of LARP7 in the untransformed MCF10A cells using short hairpin RNAs ( shRNAs ) . Two independent LARP7-targeting sequences ( shLARP7-1 and shLARP7-2 ) , when introduced separately into MCF10A cells , markedly decreased the levels of LARP7 protein and 7SK snRNA ( Figure 2A , upper panel ) . We have previously reported that these KD cells are partially transformed as evidenced by the disruption of epithelial polarity and morphological differentiation in the 3D IrECM ( He et al . , 2008 ) . During the culture of these cells , we noticed a change in cell morphology from the cobble stone-like shape typical of epithelial cells to a more spindle-like and scattered appearance ( Figure 2A , lower panel ) , indicating these cells may be undergoing EMT . 10 . 7554/eLife . 02907 . 006Figure 2 . Silencing LARP7 induces EMT in MCF10A cells . ( A ) Upper panel , the levels of 7SK snRNA and LARP7 protein in MCF10A parental cells and cells stably expressing scramble control , shLARP7-1 , or shLARP7-2 were examined by Northern blotting ( NB ) and Western blotting ( WB ) , respectively . Tubulin was used as a loading control . Lower panel , phase-contrast images of control and two shLARP7 pools . ( B ) Wound healing assay . Confluent cell monolayers were wounded , and wound closure was monitored at 0 hr and 16 hr . ( C ) Migration assay . MCF10A control or shLARP7 cells were subjected to a Transwell migration assay . The migrated cells were stained and counted . Data were collected from five fields in three independent experiments . ( D ) Immunofluorescence staining of E-cadherin ( green ) and actin stress fibers ( red ) . ( E ) qRT-PCR analysis of N-cadherin expression in control and two shLARP7 pools . ( F and G ) The levels of LARP7 protein ( F ) and mRNA ( G ) as well as 7SK snRNA level ( G ) in control , shLARP7-2 cells and rescue cells ( expressing an shRNA-resistant WT LARP7 cDNA ) were examined by Western blotting and qRT-PCR , respectively . ( H ) Phase contrast pictures of control , shLARP7-2 , and rescue cells . ( I ) The wound healing assay . ( J ) The Transwell migration assay . For C , E , G , and J panels: data are presented as mean ± SD . *: the p values ( p<0 . 05 ) were determined by the Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 006 EMT is characterized by a number of functional and molecular changes , including marked increase in cell migration and invasion , actin stress fiber formation , upregulation of mesenchymal markers and downregulation of epithelial markers ( Thiery et al . , 2009 ) . We therefore measured these characteristics in the MCF10A LARP7 KD cells in order to determine whether the knockdown induces EMT . In the wound healing assay , the KD cells showed markedly faster wound closure than the control cells ( Figure 2B ) . Consistently , these cells also displayed significantly accelerated cell migration ( Figure 2C ) . In addition , the LARP7 KD cells displayed a marked increase in actin stress fiber formation and a loss of E-cadherin from the adherens junctions ( Figure 2D ) . Furthermore , the level of N-cadherin , a mesenchymal marker , was increased nearly twofold upon LARP7 KD ( Figure 2E ) . Re-expression of a shRNA-resistant WT LARP7 in the shLARP7-2 cells ( Figure 2F ) effectively restored the loss of 7SK snRNA ( Figure 2G ) , restored the epithelial morphology ( Figure 2H ) and rescued the other EMT phenotypes ( Figure 2I , J ) , confirming that the enhanced EMT of shLARP7 cells is due to the loss of LARP7 expression . Together , these data indicate that reducing LARP7 promotes EMT in MCF10A cells . In light of the above demonstrations that LARP7 inhibits EMT and that downregulation of LARP7 occurs in invasive , but not noninvasive breast cancer tissues and cell lines , we asked whether downregulation of LARP7 via shRNA could directly promote malignant progression of noninvasive breast cancer cells . To test this , we stably knocked down LARP7 in two noninvasive breast cancer cell lines T47D ( Figure 3A ) and BT474 ( Figure 3—figure supplement 1A ) . This resulted in a reduced 7SK snRNA level and caused a significant increase in cell proliferation and anchorage-independent growth in soft agar ( Figure 3B–D , Figure 3—figure supplement 1B–D ) . The LARP7 KD cells also exhibited dramatically increased EMT as evidenced by accelerated cell migration and invasion ( Figure 3E , F , Figure 3—figure supplement 1E ) , downregulation of epithelial markers such as E-cadherin , DSP , and KRT19 and upregulation of mesenchymal markers Vimentin , N-cadherin , and matrix metalloproteinases ( MMPs ) ( Figure 3H , Figure 3—figure supplement 1F ) . Since EMT has recently been linked to the expansion of cancer stem cells ( CSCs ) ( Mani et al . , 2008; Bessede et al . , 2013 ) , we also examined whether the LARP7 KD cells displayed an increased CSC feature in the two-round mammosphere formation assay . Indeed , the KD cells generated bigger and larger number of mammospheres ( Figure 3G ) and displayed elevated levels of various stem cell markers including Oct4 , Sox2 , and ALDH1 ( Figure 3H ) . 10 . 7554/eLife . 02907 . 007Figure 3 . Silencing LARP7 promotes malignant progression of T47D breast cancer cells . ( A ) Western blotting shows effective knockdown of LARP7 in T47D cells stably expressing shLARP7s . ( B ) qRT-PCR analysis of 7SK and LARP7 RNA levels in T47D cells stably expressing scramble control or shLARP7s . ( C–F ) Silencing LARP7 in T47D cells results in increased cell proliferation ( C ) , anchorage-independent growth ( D ) , cell migration ( E ) , and invasion ( F ) . ( G ) Representative images of mammospheres formed by control or shLARP7 cells . The numbers of primary and secondary mammospheres were quantified and shown in the graph below . Scale bars represent 100 μm . ( H ) qRT-PCR analysis of various epithelial and mesenchymal markers , MMPs , and stem cell markers in control and two shLARP7 pools . DSP , desmoplakin; KRT19 , keratin 19 . PCR values were normalized to that of β-actin . ( I ) Representative images of the H&E-stained lung sections from mice injected with control or shLARP7-2 cells . Scale bar , 100 μm . Quantification of the number of metastatic nodules is shown in the bar graph below . n = 6 in each group . Data in this figure are presented as the mean ± SD , and the p values ( *p<0 . 05 ) were determined by the Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 00710 . 7554/eLife . 02907 . 008Figure 3—figure supplement 1 . Silencing LARP7 accelerates malignant progression of BT474 breast cancer cells . ( A ) Western blotting analysis of LARP7 levels in BT474 control and two shLARP7 pools . ( B ) qRT-PCR analysis confirms the reduced RNA levels of 7SK and LARP7 in BT474 cells stably expressing shLARP7s . ( C–E ) Silencing LARP7 in BT474 cells results in increased cell proliferation ( C ) , anchorage-independent growth ( D ) , and cell migration ( E ) . ( F ) qRT-PCR analysis of various EMT markers and MMPs in control and two shLARP7 pools . PCR values were normalized to that of β-actin . Data are presented as the mean ± SD from three independent measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 008 Finally , to determine whether LARP7 KD also results in an increase in breast cancer metastasis in vivo , the T47D control or shLARP7-2 cells were injected intravenously into the nude mice , and lung metastasis was examined 12 weeks later . Histological analyses revealed a significant increase in the number of metastatic lesions produced by shLARP7-2 cells when compared with that produced by the control cells ( Figure 3I ) . Taken together , our results suggest that knockdown of LARP7 enhances breast cancer EMT and CSC expansion in vitro and metastasis in vivo . Because LARP7 is a critical component of the inhibitory 7SK snRNP , we hypothesize that its ability to suppress tumor progression is due to the sequestration of P-TEFb in 7SK snRNP . To test this hypothesis , we first examined the kinase activity of P-TEFb upon LARP7 KD in an in vitro kinase assay using recombinant GST-CTD as a substrate . As predicted , the ability of endogenous CDK9 to phosphorylate the CTD was dramatically higher in the KD cells than in the control cells ( Figure 4A ) . 10 . 7554/eLife . 02907 . 009Figure 4 . P-TEFb is required for the increased cell migration and invasion induced by LARP7 KD . ( A ) An in vitro kinase assay using GST-CTD of Pol II as an exogenous substrate shows that silencing LARP7 results in activation of P-TEFb kinase . ( B ) Flavopiridol ( Flvp ) inhibits Ser2 phosphorylation of Pol II CTD in MCF10A and T47D cells . Cells were treated with varying concentrations of flavopiridol for 7 hr before lysis , and pSer2 and total Pol II levels were analyzed by Western blotting . Tubulin was used as a loading control . ( C–D ) Treatment of cells with 0 . 3 μM flavopiridol reverses the accelerated cell motility of shLARP7 cells in the wound healing assay ( C ) and cell migration in the Transwell assay ( D ) . Data are presented as the mean ± SD . p values were determined by the Student's t test . *p<0 . 05; comparison was between shLARP7 and scramble groups under DMSO treatment; #p<0 . 05; comparison was between Flvp- and DMSO-treated cells within the same cell lines . ( E ) Silencing CDK9 by siRNA reverses shLARP7-induced increase in cell migration . Two siCDK9s ( siCDK9-1 and siCDK9-2 ) were transfected into T47D shLARP7-2 cells , and cell migration was assessed by a Transwell assay and quantified in the graph below . The efficiency of CDK9 knockdown was examined by Western blotting ( upper panel ) . p values were determined by the Student's t test . *p<0 . 05 , compared between shLARP7-2 siCtrl and scramble; #p<0 . 05 , compared between shCDK9s and siCtrl . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 009 If the increased P-TEFb activity upon LARP7 KD is responsible for the EMT and enhanced transformation of breast cancer cells , inhibition of P-TEFb is expected to reverse the process . To test this , we employed flavopiridol , a known P-TEFb inhibitor , which effectively inhibits phosphorylation of CTD Ser2 in MCF10A and T47D cells at 0 . 3 μM ( Figure 4B ) . Treatment of LARP7 KD cells with this concentration of flavopiridol fully blocked the increase in cell motility and migration ( Figure 4C , D ) . In addition to the pharmacological inhibition of P-TEFb , silencing CDK9 expression by two different siRNAs also abolished the LARP7 KD-induced increase in cell migration ( Figure 4E ) . Taken together , these data are consistent with the model that the elevated P-TEFb activity in the LARP7 KD cells is responsible for the observed increase in EMT and invasion . Because P-TEFb is a transcription elongation factor that most likely affects breast cancer progression at the level of transcription , we decided to examine the expression of a panel of EMT regulators in the two T47D LARP7 KD cell lines ( shLARP7-1 and shLARP7-2 ) . Consistent with the enhanced EMT phenotypes , loss of LARP7 markedly increased the expression of many key EMT and metastasis genes ( Zeisberg and Neilson , 2009 ) , including Slug , FOXC2 , ZEB2 , Twist1 , ZEB1 , Snail , Twist2 , and SOX10 ( Figure 5A ) . Notably , treatment of these LARP7 KD cells with flavopiridol or introduction of siCDK9s strongly inhibited the expression of four of those genes: Slug , FOXC2 , ZEB2 , and Twist1 ( Figure 5B , C ) , suggesting that P-TEFb may directly affect their expression . 10 . 7554/eLife . 02907 . 010Figure 5 . Loss of LARP7 enhances transcription of key EMT genes . ( A ) LARP7 knockdown significantly increases the expression of key EMT transcriptional factors in T47D breast cancer cells as measured by qRT-PCR . The p values ( *p<0 . 05 ) were determined by the Student's t test . ( B ) Cells were treated with flavopiridol for 7 hr at the indicated concentrations . The expression of EMT genes was assessed by qRT-PCR . ( C ) Cells were transfected with siCtrl , siCDK9-1 or siCDK9-2 . After 48 hr , the expression of EMT genes was assessed by qRT-PCR . ( D ) Control or two shLARP7 T47D cells were subjected to ChIP analysis to determine the levels of CDK9 and phospho-Ser2 ( pSer2 ) Pol II occupancy at various locations of EMT-related transcription factors and non-responsive genes . qRT-PCR was performed using primers specific to the transcription start site ( TSS ) , interior and 3′-UTR of each gene , and the signals were normalized to that of input . Data are presented as the mean ± SD from three independent measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 010 To determine whether these four genes are direct targets of P-TEFb , we performed the chromatin immunoprecipitation ( ChIP ) assay to measure the occupancy of CDK9 and pSer2 CTD at these gene loci at three different positions: the promoter-proximal/transcription start site ( TSS ) , the interior of the open reading frame , and the 3′-untranslated region ( UTR ) . As shown in Figure 5D , an enrichment of both CDK9 and pSer2 CTD was detected at all three positions in Slug , FOXC2 , ZEB2 , and Twist1 genes upon LARP7 depletion , but not in other non-responsive genes such as β-actin and ubiquitin C or EMT genes that are not direct targets of P-TEFb such as Snail and Twist2 . These data suggest that LARP7 KD induces EMT by increasing P-TEFb-dependent expression of key EMT transcription factors . It is worth noting that although some of the EMT genes such as Snail and Twist2 are not directly bound by CDK9 , their expression may still be subjected to regulation by P-TEFb due to the fact that their transcription can be activated by Slug and Twist1 , respectively ( Smit et al . , 2009; Wels et al . , 2011; Guo et al . , 2012 ) , and therefore indirectly activated by P-TEFb . Since LARP7 depletion reduces 7SK snRNA levels , leading to disruption of the 7SK snRNP and P-TEFb activation , we next investigated whether the increased P-TEFb activity was due to redistribution of P-TEFb from the inhibitory 7SK snRNP to the transcriptionally active P-TEFb complexes . Toward this goal , CKD9 was immunoprecipitated from the nuclear extracts of T47D control and LARP7 KD cells , and its associated factors were analyzed by Western blotting . LARP7 KD decreased the association of P-TEFb with the 7SK snRNP components HEXIM1 and MePCE , while increased P-TEFb binding to the SEC components AFF4 , ELL2 , and AF9 ( Figure 6A ) . The increased CDK9 levels in the SEC was most likely due to the enhanced transcription of the three SEC genes as revealed by qRT-PCR ( Figure 6B ) , which in turn resulted in elevated levels of these proteins in the nuclear extracts ( Figure 6A ) . Thus , LARP7 KD releases P-TEFb from the 7SK snRNP and redistributes it to the active SEC . 10 . 7554/eLife . 02907 . 011Figure 6 . Silencing LARP7 redistributes P-TEFb from the 7SK snRNP to SEC . ( A ) Nuclear extracts were prepared from the control or shLARP7 cells and subjected to immunoprecipitation ( IP ) with anti-CDK9 antibodies . The 7SK snRNP and SEC formation was examined by Western blotting . ( B ) qRT-PCR analysis of AF9 , ELL2 , and AFF4 expression in T47D cells with or without LARP7 KD . The p values ( *p<0 . 05 ) were determined by the Student's t test . ( C ) ChIP assay was performed to determine the binding of ELL2 to the EMT-related genes and non-responsive genes . Primers at the TSS of each gene were used . Data are presented as the mean ± SD from three independent measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 011 To determine whether the elevated SEC formation in LARP7 KD cells could increase SEC's association with key P-TEFb-targeted EMT genes , we performed the ChIP assay to examine the binding of the signature SEC subunit ELL2 to the genes encoding EMT transcription factors . As shown in Figure 6C , ELL2 was enriched on the four P-TEFb-responsive EMT genes Slug , FOXC2 , ZEB2 , and Twist1 upon LARP7 KD , but not on non-responsive β-actin and ubiquitin C genes or Snail and Twist2 that are not direct P-TEFb targets , strongly implicating the involvement of the SEC in supporting the LARP7 KD-induced EMT . Taken together , these data indicate that silencing LARP7 in noninvasive breast cancer cells shifts the P-TEFb equilibrium from the inhibitory 7SK snRNP to the active SEC , leading to increased P-TEFb activity and expression of EMT-related transcription factors . If LARP7 downregulation accelerates malignant progression by increasing expression of EMT genes through elevated P-TEFb activity , we should expect P-TEFb inhibition or re-introduction of LARP7 protein to block EMT and transformation . To test this hypothesis , we first treated the highly invasive and metastatic MDA-MB-231 breast cancer cells with flavopiridol and noted that the treatment significantly inhibited cell migration ( Figure 7A ) . Furthermore , mRNA levels of Slug , a major EMT regulator , were repressed in a dose-dependent manner upon P-TEFb inhibition ( Figure 7B ) . These data suggest that inhibition of P-TEFb can effectively block cell migration in metastatic breast cancer cells . 10 . 7554/eLife . 02907 . 012Figure 7 . Inhibition of P-TEFb impairs EMT and survival of metastatic MDA-MB-231 cells . ( A ) Treatment of MDA-MB-231 cells with 0 . 3 μM flavopiridol ( Flvp ) significantly impairs cell migration as shown by the Transwell assay . The p value ( *p<0 . 05 ) was determined by the Student's t test . ( B ) qRT-PCR analysis of Slug expression in MDA-MB-231 cells that have been treated with flavopiridol for 7 hr at the indicated concentrations . ( C and D ) Clonogenic growth assay . MDA-MB-231 cells were transfected with control vector , Flag-tagged WT LARP7 , or Δ2A mutant . The expression of LARP7 and Δ2A was confirmed by Western blotting with anti-Flag . ( C ) . ( D ) Representative culture areas are shown in the top panel . The number of colonies was quantified and shown in the graph below . The p value ( *p<0 . 05 ) was determined by the Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 02907 . 012 Next , to investigate whether re-introducing LARP7 in metastatic breast cancer cells could block or reverse the malignant phenotypes , we transfected MDA-MB-231 cells with cDNAs encoding either WT LARP7 or the LARP7 Δ2A mutant ( Figure 7C ) . The Δ2A mutant was originally found in microsatellite instable gastric cancer and contains a deletion of two adenosines from a microsatellite repeat of 8 A's ( nucleotides 1206–1213 ) in the LARP7 C-terminal region , leading to a frameshift deletion of the C-terminal region of LARP7 . As a result , the Δ2A mutant cannot bind to 7SK snRNP and fails to suppress P-TEFb activity ( He et al . , 2008 ) . The transfected cells were selected with puromycin and subjected to a colony formation assay . Interestingly , the LARP7-overexpressing cells formed significantly fewer colonies than those harboring the control vector ( Figure 7D ) , suggest that re-expression of LARP7 impairs survival of these metastatic breast cancer cells . In contrast , cells expressing the Δ2A mutant formed as many colonies as the vector control cells , suggesting that the ability of LARP7 to assemble 7SK snRNP and suppress P-TEFb is necessary for its tumor suppressor potential . Taken together , these data indicate that LARP7 likely suppresses the survival and progression of malignant breast cancer cells through 7SK snRNP-dependent inhibition of P-TEFb .
In human cells , the 7SK snRNP represents the principle cellular reservoir of uncommitted P-TEFb , and its integrity is maintained through LARP7's direct interaction with and stabilization of the 7SK snRNA ( He et al . , 2008 ) . Here , we show that in noninvasive human breast cancer cells , disruption of this complex by knocking down LARP7 releases P-TEFb , redistributing it to the transcriptionally active SEC complex . This activation of P-TEFb promotes breast cancer EMT , invasion , cancer stem cell expansion , and metastasis by directly activating the expression of genes involved in EMT and metastasis . Thus , the sequestration of P-TEFb by 7SK snRNP is an effective anti-cancer mechanism , and the key 7SK snRNP component LARP7 is a potential tumor suppressor that specifically blocks breast cancer progression and metastasis . Our data highlight the importance of the transcription elongation machinery in regulating breast cancer EMT and metastasis and suggest a new therapeutic option to combat metastatic breast cancer through blocking P-TEFb activation . Currently , the CDK9 inhibitor flavopiridol is being evaluated in several phase I and II clinical trials for its anti-cancer effects either as a single agent or in combination with other drugs in treatment of esophageal cancer , B-cell chronic lymphocytic leukemia , endometrial carcinoma , recurrent/metastatic squamous cell carcinoma and most relevantly , previously treated locally advanced or metastatic breast cancer ( http://www . cancernetwork . com/review-article/current-clinical-trials-flavopiridol/page/0/2 ) . Thus , our studies are particular relevant and provide the mechanistic basis for targeting P-TEFb in patients with metastatic breast cancer . LARP7 is a member of the LARP family that contains four La domain-containing RNA binding proteins ( LARP1 , 4 , 6 , and 7 ) with distinct RNA target preferences ( Bayfield et al . , 2010 ) . Among the four members , LARP7 is the only one that binds to 7SK snRNA and is involved in transcription elongation controlled by Pol II . The role of the LARP family proteins in human cancer is so far poorly known , and our study is the first demonstration that LARP7 functions as a potential suppressor of human breast cancer metastasis . This function of LARP7 is additionally supported by the observation that the Drosophila homolog of LARP7 , multisex-combs ( MXC ) , also acts as a tumor suppressor to inhibit cell growth ( Remillieux-Leschelle et al . , 2002 ) . Moreover , the C-terminal region of LARP7 , which is essential for the sequestration of P-TEFb in 7SK snRNP , is required for LARP7's suppression of breast cancer survival and is frequently deleted in human gastric cancer ( Mori et al . , 2002 ) , suggesting that LARP7 may be a potential tumor suppressor in a broad range of human carcinomas . Finally , an inactivating mutation of LARP7 has been linked to a novel form of familial Primordial Dwarfism characterized by facial dysmorphism and intellectual disability ( Alazami et al . , 2012 ) . In these patients , the 7SK snRNA is depleted due to the defective LARP7 , indicating that LARP7 and 7SK snRNA , through their ability to control P-TEFb activity , also regulate embryonic development . Although it is formally possible that the LARP7 KD may also affect other yet-to-be-identified pathways in a P-TEFb-independent manner , our observation that both flavopiridol and siCDK9s efficiently rescued the increased cell migration of LARP7 KD cells lends a strong support to the notion that the observed effects of LARP7 KD on EMT and invasion are indeed mediated by P-TEFb activation . A key observation of our study is that among the many EMT genes whose expression is altered by P-TEFb activation , the transcription factors Slug , FOXC2 , ZEB2 , and Twist1 are direct targets of P-TEFb and bound by the P-TEFb-containing SEC . These four proteins are well known EMT transcription factors that function as decisive early drivers of EMT . While FOXC2 orchestrates the mesenchymal component of the EMT program ( Mani et al . , 2007 ) , Slug , ZEB2 , and Twist1 repress E-cadherin expression , a fundamental event in EMT , as well as the expression of proteins of tight junctions , gap junctions , and desmosomes , contributing to the de-differentiated state of tumor cells and facilitating EMT ( Peinado et al . , 2007; Sanchez-Tillo et al . , 2012 ) . In addition , these proteins are also closely linked to breast cancer stem cell properties and implicated in resistance to drug and radio-therapies ( Voulgari and Pintzas , 2009; Singh and Settleman , 2010; Hollier et al . , 2013 ) . Furthermore , these EMT early drivers also activate the expression of other EMT transcription factors . For example , Twist1 can induce Snail ( Smit et al . , 2009 ) , and Slug can upregulate Twist2 and ZEB1 expression ( Wels et al . , 2011; Guo et al . , 2012 ) . Thus , by directly activating the expression of these EMT early drivers , P-TEFb is capable of orchestrating the EMT program to drive malignant progression . Since P-TEFb and its associated factors are expressed in most cell types , it is possible that the ability of P-TEFb to promote tumor progression may not be restricted to breast cancer but also true for other malignant tumors . Indeed , the active P-TEFb-Brd4 complex has been linked to AML and CML ( Dawson et al . , 2011; Mertz et al . , 2011; Zuber et al . , 2011; Herrmann et al . , 2012; Winter et al . , 2012 ) , while multiple SEC components are direct targets of MLL-fusion proteins that are linked to various forms of leukemia ( Lin et al . , 2010; Yokoyama et al . , 2010 ) . We have found that in the noninvasive breast cancer cell lines , P-TEFb released from the 7SK snRNP is mainly redistributed to the SEC . The expression levels of two SEC components AF9 and ELL2 are also markedly upregulated , allowing more SEC formation . Since ELL2 is another key catalytic component of SEC besides P-TEFb , its increase is likely to further enhance transcriptional elongation of target genes . We found that the Brd4-P-TEFb complex was also increased moderately upon LARP7 KD ( data not shown ) , suggesting the possibility that Brd4 may also play a role in facilitating the P-TEFb-dependent breast cancer progression . Given that P-TEFb is mostly known as a general transcription factor , our observations raise two interesting questions: Why are EMT transcription factors particularly sensitive to P-TEFb activation in these cells , and how is the SEC involved in this process ? Our data that P-TEFb and SEC are associated with the genes encoding decisive EMT regulators but not non-responsive genes suggest that the direct target EMT genes may contain sequences that display high affinity binding to the SEC-P-TEFb complex . For example , these genes may contain the recently discovered Super-Enhancers that render them highly sensitive to activation by the Mediator and the SEC ( Hnisz et al . , 2013; Loven et al . , 2013; Whyte et al . , 2013 ) . Interestingly , previous proteomic and biochemical analyses have identified the Mediator as a SEC binding partner through interaction between the Mediator subunit MED26 and the SEC components ELL-associated factor 1 ( EAF1 ) and EAF2 ( Takahashi et al . , 2011 ) , suggesting that the Mediator may recruit the SEC to target genes and the two complexes may act together to activate genes containing the Super-Enhancers . The SEC has also been linked to the elongating Pol II on chromatin template through interaction between the SEC components ENL/AF9 and the Pol II-associated factor 1 ( PAF1 ) ( He et al . , 2011 ) . Finally , SEC may be recruited to its target promoters through tissue-specific and sequence-specific DNA binding partners , such as the HIV Tat protein in HIV-infected cells , MLL-fusion proteins in leukemia cells ( He and Zhou , 2011; Smith et al . , 2011 ) or retinoic acid receptor in differentiating mouse embryonic stem cells ( Lin et al . , 2011 ) . Besides the positive control exerted by P-TEFb and SEC , another possibility is that the EMT genes are strongly suppressed by negative transcription elongation factors DSIF and NELF , and thus are very sensitive to P-TEFb activation . Future studies will try to determine which of these mechanisms may be responsible for the heightened sensitivity of EMT transcription factors to the elongation control . Taken together , our study has demonstrated that LARP7 functions as a potential tumor suppressor in human breast cancer by suppressing the activity of P-TEFb and inhibiting EMT , invasion , and metastasis . Our study also provides the first demonstration that the transcription elongation machinery and the network of P-TEFb complexes play critical roles in regulating tumor EMT , cancer stemness , and metastasis by directly controlling the expression of upstream EMT regulators . These findings will facilitate the development of new drugs that target CDK9 for anti-cancer therapy .
The MCF10A mammary epithelial cells were cultured in DMEM-F12 medium supplemented with 5% horse serum , 20 ng/ml EGF , 10 μg/ml insulin , 0 . 5 μg/ml hydrocortisone , 100 ng/ml of cholera toxin , and penicillin/streptomycin . The EpH4 murine mammary epithelial cells were cultured as previously described ( Xu et al . , 2009 ) . The breast cancer cells BT474 , T47D , ZR75B , and BT549 were cultured in RPMI1640 media containing 10% FBS . MCF7 , MDA-MB-468 , MDA-MB-231 , and MDA-MB-435 cells were cultured in DMEM plus 10% FBS . The anti-LARP7 and anti-MePCE antibodies were generated as previously described ( He et al . , 2008; Xue et al . , 2010 ) . Antibodies against CDK9 , CyclinT1 , HEXIM1 , and Brd4 have been described earlier ( Yik et al . , 2003; Yang et al . , 2005 ) . Antibodies against AFF4 and Pol II CTD repeat YSPTSPS ( phospho S2 ) were purchased from Abcam ( Cambridge , UK ) . Antisera against Pol II ( 8WG16 ) were purchased from Santa Cruz Biotechnology ( Dallas , TX ) . The anti-ELL2 ( A302-505A-1 ) antibodies were purchased from Bethyl Laboratories , Inc ( Montgomery , TX ) . Antibodies against tubulin and E-cadherin were from Calbiochem ( San Diego , CA ) and BD ( Franklin Lakes , NY ) , respectively . Anti-Flag antibody was purchased from Sigma ( St . Louis , MO ) . The Oncomine database ( www . oncomine . org ) was searched for the expression profiles of P-TEFb and its associated factors . Only the data sets examining mRNA expression in cancer tissue with matched normal tissue controls ( cancer vs normal ) were included in this study . The threshold search criteria were p value<0 . 05 , fold change >2 , and gene rank percentile <10% . p values presented in this study were calculated using a two-sided Student's t-test . NKI295 gene expression data were downloaded from the Stanford Microarray Database ( http://microarray-pubs . stanford . edu/wound_NKI/explore . html ) . Survival data , stratified by expression of LARP7 , were analyzed by using SPSS 13 . 0 and tested for significance using the log-rank test . The breast cancer tissue array ( BR1503b ) was purchased from US Biomax , Inc . Immunohistochemistry was carried out using the Tyramide Signal Amplification Biotin System Kit ( PerkinElmer ) with anti-LARP7 antibodies ( 10 μg/ml ) , following the manufacturer's instructions . Images were captured using the Zeiss AxioImager M2 microscope . The intensity of LARP7 stain ( the number of pixels/area ) in each sample was quantified by analyzing at least three stained areas using the ImageJ plus software . Statistical analysis was performed using the SPSS 13 . 0 program and determined by one-way ANOVA . The small hairpin RNA ( shRNA ) vector targeting human LARP7 was introduced into breast cancer cells by retroviral infection as described previously ( Zhu et al . , 2007 ) . Briefly , shLARP7 in pSUPERretro-puro vector was transfected together with retroviral packaging vectors into 293T cells , and viral supernatant was used to infect the breast cancer cells . Pools of infected cells were selected in the presence of puromycin and analyzed by a variety of assays . The shLARP7 sequences are: shLARP7-1 , 5′-AATCACAGCTGGATTGAAA-3′; shLARP7-2 , 5′-AAGTTAATCACCAAAGCTG-3′ . A scrambled sequence having similar base compositions to the shLARP7s was used as a negative control . Lentivirus production and infection were conducted as previously described ( Moffat et al . , 2006 ) . The MCF10A shLARP7 stable cell lines and the rescue pool expressing the shRNA-resistant WT LARP7 have been reported previously ( He et al . , 2008 ) . siRNAs targeting CDK9 were introduced into cell by transfection . Sequences of the two siCDK9 constructs used in the study are: siCDK9-1: 5′-GGGAGAUCAAGAUCCUUCATT-3′; siCDK9-2: 5′-GGUGAUGCAGAUGCUGCUUTT-3′ . qRT-PCR was performed with ABI 7300 ( Applied Biosystem ) and DyNAmo HS SYBR Green qPCR kit ( Fisher Scientific ) as per manufacturer's instruction . The gene-specific primers ( Supplementary file 1A ) were used at a final concentration of 0 . 2 μM . All PCR reactions were performed in triplicates . NEs were prepared from various cell lines using a standard protocol ( Dignam et al . , 1983 ) . Immunoprecipitation and Western blotting were performed as previously described ( He et al . , 2008 ) . Briefly , NEs were incubated with specific antibodies at 4°C overnight and then with protein A beads ( Invitrogen ) for 2 hr . After washing with buffer D ( 20 mM HEPES-KOH [pH7 . 9] , 15% glycerol , 0 . 2 mM EDTA , 0 . 2% NP-40 , 1 mM dithiothreitol , and 1 mM phenylmethylsulfonyl fluride ) containing 0 . 3 M KCl ( D0 . 3M ) , the isolated proteins were eluted with 0 . 1 M glycine ( pH 2 . 0 ) and analyzed by Western blotting with indicated antibodies . The ChIP assay was performed according to the manufacturer's instructions ( EZ-ChIP: Catalog # 17-371; Millipore ( Billerica , MA ) . Briefly , 2 μg antibodies per reaction were incubated with chromatin fractions in an immunoprecipitation assay , and the isolated chromatin fraction was purified and subjected to qRT-PCR using primers listed in Supplementary file 1B . The in vitro kinase assay was performed as described ( Yang et al . , 2005 ) with minor modifications . Briefly , CDK9 were isolated by immunoprecipitation from NEs of breast cancer cells , washed with buffer D0 . 15M for three times , and then subjected to the kinase assay using 4 μg GST-CTD as an exogenous substrate in the presence of 5uCi γ-[P32]-ATP . The phosphorylated GST-CTD was analyzed by SDS-PAGE and visualized by autoradiography . Cells were cultured in six-well plates to full confluence . A plastic tip was used to generate a wound across the cell monolayer . The wound closure was measured after 16 hr . Migration assays were performed in Transwell chambers ( Corning ) . 1 × 105 cells in medium containing 1% FBS were seeded onto membranes in top chambers and allowed to migrate towards the bottom chambers filled with medium containing 10% FBS . Cells that remained in the upper chambers were removed with a cotton ball . Migrated cells were stained with crystal violet and photographed . Invasion assays were conducted using the BD Matrigel Invasion Chambers as per the manufacturer's protocol . Cells were fixed in 4% paraformaldehyde for 20 min and permeabilized in 0 . 1% Triton X-100 for 5 min . Rhodamine-conjugated phalloidin ( Molecular Probes: Eugene , OR ) was employed to stain actin stress fiber ( 30 min at room temperature ) . E-cadherin was detected by staining with anti-E-cadherin antibodies ( 1 μg/ml ) overnight at 4°C followed by incubating with Alexa 488-conjugated anti-mouse antibody ( Molecular Probes ) for 1 hr at room temperature . Cell nuclei were stained with Hoechst for 3 min . Growth medium ( 0 . 4 ml ) containing 0 . 66% Bacto Agar ( BD ) was added to a 24-well plate in triplicates and allowed to harden . 2000 breast cancer cells were suspended in 0 . 2 ml medium containing 0 . 375% agar and overlaid on the hardened bottom layer . Fresh medium ( 0 . 2 ml ) containing 0 . 375% agar was added to each well once a week for 4 weeks . The colonies were visualized by staining with 0 . 5 mg/ml 3- ( 4 , 5-dimethylthiazol-2-yl ) -2 , 5-diphenyl tetrazolium bromide ( MTT ) ( Sigma ) for 4 hr at 37°C . Two-round mammosphere assay was performed as previously described ( Dontu et al . , 2003 ) . Briefly , cells were trypsinized , counted , and plated on ultra-low attachment plates ( Corning ) at a density of 1000 cells/ml . Mammospheres were counted after 5 days . Cells were then dissociated , diluted , and re-seeded for a second round of mammosphere formation . To evaluate the metastatic potential of shLARP7 cells , 2 × 106 cells in 150 μl serum-free medium were injected into the tail veins of 6-weeks-old female nude mice . After 12 weeks , mice were sacrificed , and quantitation of metastatic colonies was performed on representative hematoxylin and eosin ( H&E ) -stained sections of formalin-fixed and paraffin-embedded lungs . MDA-MB-231 cells were transfected with empty vectors , WT LARP7 or Δ2A cDNAs using Lipofectamine 2000 ( Invitrogen ) . 48 hr after transfection , cells were seeded into six-well plates in triplicates and selected with puromycin ( 1 . 5 μg/ml ) for 3 days . Surviving cells were continued to grow in medium without antibiotics . After 2 weeks , the colonies were stained with 0 . 1% crystal violet and quantified . | To express a gene to make a protein , the gene's DNA must first be transcribed to produce molecules of messenger RNA . The start of the transcription process features two milestones . First , an enzyme called RNA Polymerase II starts the process . Shortly afterwards , however , the process pauses and only starts again when other proteins are recruited . This second step , called transcriptional elongation , is essential for gene expression in cells that are growing and specializing into specific cell types . However , it is unclear how important this second step is for the progression of human cancers , such as breast cancer . In humans , two proteins join together to form a complex called ‘positive transcription elongation factor b’ ( or P-TEFb for short ) . This elongation factor encourages the transcriptional elongation step by adding phosphate groups onto RNA Polymerase II and by outcompeting other proteins that act to stop the process . However , some of the P-TEFb proteins in the cell's nucleus are unable to do this because they are held within a complex , which also contains an RNA molecule and some other proteins including one called LARP7 . This protein–RNA complex is thought to help to prevent a number of cancers , for example breast cancer or stomach cancer; however the effect of P-TEFb proteins on cancers in humans is not known . Less LARP7 protein is made in breast cancer cells compared to healthy cells . And when Ji et al . reduced the levels of the LARP7 protein ( or the RNA molecule involved in the complex ) , the P-TEFb proteins were released from the complex and were free to encourage transcriptional elongation . This led to the increased expression of other proteins that switch other genes on or off , including genes that allow breast cancer cells to spread around the body . On the other hand , Ji et al . revealed that freeing the P-TEFb proteins from the complex in the nucleus did not appear to cause new tumors to develop or existing tumors to grow . Ji et al . suggest that the LARP7 protein normally helps to prevent the spread of breast cancers by keeping the P-TEFb proteins inactive as a part of the protein–RNA complex . One of the next challenges will be to see if drugs that can inhibit the P-TEFb proteins might be useful as new treatments for late stage breast cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2014 | LARP7 suppresses P-TEFb activity to inhibit breast cancer progression and metastasis |
The Positive Transcription Elongation Factor b ( P-TEFb ) phosphorylates Ser2 residues of the C-terminal domain ( CTD ) of the largest subunit ( RPB1 ) of RNA polymerase II and is essential for the transition from transcription initiation to elongation in vivo . Surprisingly , P-TEFb exhibits Ser5 phosphorylation activity in vitro . The mechanism garnering Ser2 specificity to P-TEFb remains elusive and hinders understanding of the transition from transcription initiation to elongation . Through in vitro reconstruction of CTD phosphorylation , mass spectrometry analysis , and chromatin immunoprecipitation sequencing ( ChIP-seq ) analysis , we uncover a mechanism by which Tyr1 phosphorylation directs the kinase activity of P-TEFb and alters its specificity from Ser5 to Ser2 . The loss of Tyr1 phosphorylation causes an accumulation of RNA polymerase II in the promoter region as detected by ChIP-seq . We demonstrate the ability of Tyr1 phosphorylation to generate a heterogeneous CTD modification landscape that expands the CTD’s coding potential . These findings provide direct experimental evidence for a combinatorial CTD phosphorylation code wherein previously installed modifications direct the identity and abundance of subsequent coding events by influencing the behavior of downstream enzymes .
The C-terminal domain of the RPB1 subunit of RNA polymerase II ( CTD ) is composed of a species-specific number of repeats of the consensus amino acid heptad YSPTSPS ( arbitrarily numbered as Tyr1 , Ser2 , Pro3 , Thr4 , Ser5 , Pro6 , and Ser7 ) ( Jeronimo et al . , 2016 ) . The CTD undergoes extensive post-translational modification ( PTM ) that recruits RNA processing and transcription factors that regulate progression through the various stages of transcription . These modification events are dynamic , highly regulated , and maintained through the complex interplay of CTD modification enzymes . Collectively these PTMs and recruited protein factors constitute the ‘CTD Code’ for eukaryotic transcription ( Buratowski , 2003 ) . Chromatin immunoprecipitation and next-generation sequencing technologies have revealed how phosphorylation levels of CTD residues change temporally and spatially during each transcription cycle ( Eick and Geyer , 2013 ) . The major sites of phosphorylation are Ser5 and Ser2 , directed by Transcription Factor II H ( TFIIH ) ( Feaver et al . , 1994 ) and P-TEFb in mammals ( Marshall et al . , 1996 ) , respectively . The other three phosphate-accepting residues ( Tyr1 , Thr4 , and Ser7 ) are also subject to modification , although their functions are less well-understood ( Jeronimo et al . , 2013 ) . In mammalian cells , the phosphorylations of Tyr1 and Ser7 rise and peak near the promoter along with Ser5 and gradually decrease as transcription progresses towards termination . The phosphorylation of Thr4 and Ser2 , on the other hand , don’t appear until later in the transcription cycle during elongation ( Eick and Geyer , 2013 ) . The molecular underpinnings resulting in this orchestration are poorly defined . A particularly apparent gap in current knowledge is if sequence divergence from the consensus heptad or previously installed PTMs influence coding events . The CTD code is generated through the interplay of CTD modifying enzymes such as kinases , phosphatases , and prolyl isomerases ( Bataille et al . , 2012 ) . Disruption of this process is implicated in various disease states . P-TEFb is of particular interest due to its overexpression in multiple tumor types and role in HIV infection ( Franco et al . , 2018 ) . As a major CTD kinase , P-TEFb promotes transcription by contributing to the release of RNA polymerase II from the promoter-proximal pause through its phosphorylation of Negative Elongation Factor ( NELF ) , DRB Sensitivity Inducing Factor ( DSIF ) , and Ser2 of the CTD ( Wada et al . , 1998 ) . Interestingly , P-TEFb seems to phosphorylate Ser5 of the CTD in vitro and mutation of Ser5 to alanine prevents the phosphorylation of CTD substrates . However , mutation of Ser2 to alanine did not result in this abolishment ( Czudnochowski et al . , 2012 ) . These results are in contrast to in vivo studies of P-TEFb specificity , where compromised P-TEFb kinase activity results in a specific reduction in levels of Ser2 phosphorylation ( Marshall et al . , 1996 ) . The discrepancies between P-TEFb specificity in vitro and in vivo make it difficult to reconcile P-TEFb’s function as a CTD Ser2 kinases ( Bartkowiak et al . , 2010; Czudnochowski et al . , 2012 ) . To resolve these inconsistencies , we utilize a multi-disciplinary approach to investigate the specificity of P-TEFb . Identification of phosphorylation sites was carried out using ultraviolet photodissociation ( UVPD ) mass spectrometry establishing the specificity of P-TEFb in vitro in single residue resolution . We reveal the tyrosine kinase c-Abl phosphorylates consensus and full-length CTD substrates in a conservative fashion , with only half of the available sites phosphorylated . The unique phosphorylation pattern of Tyr1 by tyrosine kinases like c-Abl directs the specificity of P-TEFb to Ser2 . The priming effect of pTyr1 on P-TEFb extends to human cells , where small-molecule inhibition of c-Abl-like Tyr1 kinase activities leads to a reduction of Tyr1 phosphorylation . Further ChIP-seq analysis shows that the loss of tyrosine phosphorylation increases promoter-proximal pausing with an accumulation of RNA polymerase II at the promoter region of the gene . Overall , our results reconcile the discrepancy of P-TEFb kinase activity in vitro and in cells , showing that Tyr1 phosphorylation can prime P-TEFb and alter its specificity to Ser2 . Importantly , these findings provide direct experimental evidence for a combinatorial CTD phosphorylation code wherein previously installed modifications direct the identity and abundance of subsequent coding events , resulting in a varied PTM landscape along the CTD allowing for diversified co-transcriptional signaling .
To define P-TEFb’s specificity directly on full-length RPB1 CTD substrates , we applied matrix-assisted laser desorption/ionization-mass spectrometry ( MALDI-MS ) and liquid chromatography ultraviolet photodissociation tandem mass spectrometry ( LC-UVPD-MS/MS ) to identify the substrate residues of this kinase . Ultraviolet photodissociation ( UVPD ) using 193 nm photons is an alternative to existing collision- and electron-based activation methods in proteomic mass spectrometry . This method energizes peptide ions via a single absorption event of high-energy photons resulting in a greater number of diagnostic fragment ions and the conservation of lower energy bonds like those of some PTMs including phosphorylations ( Brodbelt , 2014 ) . This method is applicable in both positive and negative ionization modes , results in a greater degree of peptide fragmentation , better certainty in PTM localization , and conservation of PTMs to ultimately ensure the detection of even low abundance or particularly labile modifications . Because endogenous RNA polymerase II is heterogeneously modified , we used recombinant yeast CTD GST fusion proteins , which contain mostly consensus heptad repeats ( 20 of 26 ) , as an unmodified substrate for PTM analysis ( Figure 1—figure supplement 1 ) . The stability and consistency of GST yeast CTD ( yCTD ) make it ideal for studying CTD modification along consensus heptads . With high kinase and ATP concentration ( 2 mM ) and overnight incubation ( ~16 hr ) , P-TEFb generates two phospho-peptides as detected by LC-UVPD-MS/MS: a major species phosphorylated on Ser5 ( Y1S2P3T4pS5P6S7 ) and a minor species phosphorylated on Ser2 ( S5P6S7Y1pS2P3T4 ) ( Figure 1A and Figure 1—figure supplement 2A–B ) . This is highly similar to patterns observed previously for bona fide Ser5 CTD kinases Erk2 and TFIIH ( Mayfield et al . , 2017 ) . These experiments confirm P-TEFb’s inherent in vitro preference for Ser5 when phosphorylating unmodified CTD ( Czudnochowski et al . , 2012; Portz et al . , 2017 ) . We next measured the total number of phosphates added to the CTD by P-TEFb . MALDI-MS analysis of yCTD treated with P-TEFb reveals a cluster of peaks with mass shifts relative to no kinase control ranging from 1906 . 1 to 2318 . 4 Da , each interspaced by 80 Da ( Figure 1B and Figure 1—figure supplement 2C ) . This corresponds to the addition of 24 to 29 phosphates to yCTD’s 26 heptad repeats . This finding in combination with our LC-UVPD-MS/MS analysis of P-TEFb treated yCTD indicates that P-TEFb phosphorylates the CTD in an average one phosphorylation per heptad manner , and these heptads are primarily phosphoryl-Ser5 ( pSer5 ) in vitro . To phosphorylate Ser2 and Ser5 , CTD kinases must discriminate very similar SP motifs in the CTD , Y1S2P3 and T4S5P6 , to maintain accuracy during transcription . Among the flanking residues of these two motifs , the unique structure of the tyrosine side chain likely contributes to the recognition of the serine residues subject to phosphorylation . Several factors suggest the chemical properties of residues at the Tyr1 position are important for CTD modification . First , residues at this first position of the heptad are highly conserved across species and substitution to non-aromatic residues is rare , suggesting significance to function ( Chapman et al . , 2008 ) . As evidence of this , even conservative mutation of the Tyr1 position to phenylalanine in both Saccharomyces cerevisiae and human cells is lethal , highlighting the significance of residue identity at this position ( Hsin et al . , 2014; West and Corden , 1995 ) . Secondly , we have shown that mutating the Tyr1 position to alanine prevents phosphorylation at other CTD residues by CTD kinases ( Mayfield et al . , 2017 ) , indicating the side chain at this position is important for kinase activity . Third , phosphoryl-Tyr1 ( pTyr1 ) is detected at the initiation of transcription in human cells ( Descostes et al . , 2014 ) . This positions pTyr1 well to influence and interact with subsequent modifications of the CTD and , potentially , direct subsequent enzyme specificities . To determine the effect of the chemical characteristics of residues at the Tyr1 position on CTD modification , we searched for naturally occurring Tyr1 substitutions . Drosophila melanogaster CTD contains a majority of heptads that diverge from consensus sequence with only 2 of its approximately 45 heptads being of the consensus sequence . Despite this highly divergent character , the Tyr1 position of D . melanogaster CTD is rather conserved and contains mostly tyrosine residues . For the six heptads that do not contain tyrosine , half are modestly substituted with phenylalanine . We were curious to determine if , like alanine , phenylalanine replacement at the Tyr1 position would abolish CTD kinase activity . These initial experiments were designed on D . melanogaster CTD because its heptads have diverse sequences that allow for the observation of shifts in electrophoretic mobility shift assay ( EMSA ) banding patterns , which might not be easily seen for consensus sequence CTD substrates . We generated GST-CTD constructs containing a portion of D . melanogaster CTD ( residues 1671–1733 , containing nine heptad repeats ) of either tyrosine containing wild-type ( dmCTD ) or with phenylalanine substitution at the Tyr1 position in all nine heptads ( dmCTDYtoF ) ( Figure 1—figure supplement 1D ) . The CTD variants purified from these constructs were phosphorylated with one of three established CTD kinases: Erk2 , a recently identified Ser5 CTD kinase that phosphorylates primed RNA polymerase II in developmental contexts ( Tee et al . , 2014 ) ; the kinase module of TFIIH that install Ser5 and Ser7 marks in vivo ( Feaver et al . , 1994 ) ; or P-TEFb which phosphorylates Ser2 in vivo ( Marshall et al . , 1996 ) . Unlike alanine substitution , all three kinases are active against the phenylalanine-substituted CTD construct ( Figure 1C ) . Surprisingly , the substitution of phenylalanine at the Tyr1 position alters the behavior of phosphorylated substrates in EMSA ( Figure 1C ) . While the wild-type variant assumes only one or two apparent intermediate species in EMSA , the YtoF variant of dmCTD exhibits multiple intermediates , suggesting the generation of a greater diversity of phosphorylated species . Additional analysis of Erk2 phosphorylated dmCTDYtoF using electrospray ionization mass spectrometry ( ESI-MS ) of the intact phosphorylated construct confirms the existence of multiple species revealing complex spectra composed of multiple overlapping peaks relative to the dmCTD control ( Figure 1—figure supplement 2D ) . To quantify the effect of phenylalanine replacement at the Tyr1 position on CTD kinase function , we measured the kinase activity of Erk2 and TFIIH using GST-yCTD or yCTDYtoF ( in which all Tyr1 positions have been mutated to phenylalanine ) substrates ( Figure 1—figure supplement 1A–B ) . Steady-state kinetics demonstrate that the replacement of tyrosine by phenylalanine has a markedly different effect on these two kinases . Erk2 shows a 2 . 5-fold higher specificity constant against the YtoF variant , as indicated by kcat/Km , compared to the WT construct ( Figure 1D ) . Erk2 has nearly identical kcat values ( 0 . 44 ± 0 . 02 s−1 vs . 0 . 45 ± 0 . 02 s−1 ) for the two substrates , but a much lower Km for the YtoF substrate ( 35 . 8 ± 3 . 2 µM for WT vs . 14 . 8 ± 2 . 2 µM for YtoF substrates ) . This difference in Km values suggests Erk2 has a binding preference for the phenylalanine substituted substrate . However , TFIIH activity is greatly compromised when Tyr1 is replaced by phenylalanine with a nearly 10-fold reduction in kcat/Km ( Figure 1E ) . Overall , our data demonstrate the chemical properties of the residues located at the first position of the heptad repeat have a significant impact on the phosphorylation of the CTD by CTD kinases . Even slight modification of this residue ( e . g . , loss of the hydroxyl group ) can have dramatic consequences for modification of the CTD . Although substitution of non-tyrosine residues at the Tyr1 position is relatively rare in nature and does not occur in human cells , Tyr1 phosphorylation is conserved from yeast to humans and plays a key role in transcriptional events ( Chapman et al . , 2008; Yurko and Manley , 2018 ) . Since the molecular mechanism explaining its diverse biological functions remains elusive , the sensitivity of CTD kinases to the chemical properties of Tyr1 side-chain motivated us to investigate if Tyr1 phosphorylation impacts subsequent phosphorylation events by reconstructing sequential CTD phosphorylation in vitro . In humans , Tyr1 phosphorylation rises along with Ser5 phosphorylation at the beginning of transcription ( Heidemann et al . , 2013 ) . However , experiments using synthetic CTD peptides with every Tyr1 residue phosphorylated have shown that Tyr1 phosphorylation inhibits subsequent phosphorylation by CTD kinases ( Czudnochowski et al . , 2012 ) . We suspect that the heavily phosphorylated synthetic peptide doesn’t mimic the physiological RNA polymerase II during transcription . Instead , we reconstructed the phosphorylation of the CTD using physiologically relevant Tyr1 kinases in vitro . Existing literature points to Abl-like non-receptor tyrosine kinases as mammalian Tyr1 CTD kinases , with c-Abl as a major candidate ( Baskaran et al . , 1997; Burger et al . , 2019 ) . Three lines of evidence support this notion: c-Abl phosphorylates CTD in vitro ( Baskaran et al . , 1993 ) and in cells since transient over-expression of c-Abl in primate COS cells results in increased Tyr1 phosphorylation ( Baskaran et al . , 1997 ) , and c-Abl immunoprecipitates with RNA polymerase II ( Baskaran et al . , 1999 ) . To elucidate the biophysical consequences of Tyr1 phosphorylation of the CTD , we reconstructed c-Abl phosphorylation of consensus sequence CTD in vitro using purified human c-Abl and the yCTD constructs . C-Abl readily phosphorylates yCTD in vitro as evidenced by EMSA and detection of Tyr1 phosphorylation using pTyr1 specific antibody 3D12 ( Figure 2A ) . We directly interrogate the sites of phosphorylation using LC-UVPD-MS/MS to identify phosphorylation sites in single residue resolution ( Mayfield et al . , 2017 ) . Using this method to analyze a peptide containing three heptad repeats treated by c-Abl ( 3CTD , Figure 1—figure supplement 1E ) , two single phospho-forms were detected , each containing a single phosphorylated tyrosine on either the first or second heptad of the variant ( Figure 2B and Figure 1—figure supplement 2E ) . These mass shifts confirm c-Abl phosphorylates consensus CTD sequences on Tyr1 in vitro . To test if modulating c-Abl activity can alter Tyr1 phosphorylation levels in human cells , we treated HEK293T cells with the c-Abl specific inhibitor imatinib ( Knight and McLellan , 2004 ) and monitored endogenous Tyr1 phosphorylation levels using phospho-Tyr1 specific antibody 3D12 ( Figure 2C ) . Imatinib has potent and specific inhibition against c-Abl and ABL2 ( which shares 93% sequence identity in the kinase domain as c-Abl ) ( Salah et al . , 2011 ) . Tyr1 phosphorylation decreases in a dose-dependent manner from 10–50% at imatinib concentrations of 10–30 μM after 24 hr of treatment ( Figure 2C ) . Overall , our result indicates controlling the kinase activity of c-Abl , or highly similar kinases , is sufficient to significantly modulate the level of Tyr1 phosphorylation of CTD in mammalian cells . We next quantified the maximal number of phosphates added to yCTD constructs using MALDI-MS . High-resolution MALDI-MS spectra of samples treated by c-Abl revealed peaks accounting for yCTD containing 5 to 13 phosphates ( with mass shifts ranging from 398 . 9 to 1049 . 4 Da ) ( Figure 2D ) , approximately half of yCTD’s available tyrosine residues within 26 heptad repeats . Further incubation with more kinase/ATP does not appear to add more than 13 phosphate groups to substrate CTD in these assays . Additional GST-CTD constructs containing 3–5 consensus heptad repeats ( Figure 1—figure supplement 1E–G ) treated with c-Abl were analyzed using MALDI-TOF to evaluate if c-Abl truly only phosphorylates half of the available Tyr1 sites even in the presence of high kinase/ATP concentrations and after prolonged incubation times . Three phosphorylation peaks were detected in the 5CTD variant with mass differences of 79 . 8 , 160 . 3 , and 239 . 1 Da relative to the unphosphorylated control , accounting for the addition of 1–3 phosphates ( Figure 2E ) . Phosphorylation of the 4CTD construct resulted in two peaks of phosphorylation with mass differences of 79 . 1 and 160 . 7 Da relative to unphosphorylated control , accounting for the addition of 1 or 2 phosphates ( Figure 2F ) . Similarly , two phosphates are added to the 3CTD variant that displayed mass shifts of 78 . 2 or 158 . 7 Da ( Figure 2G ) . These mass shifts suggest c-Abl does not phosphorylate consensus CTD in every heptad; instead , it favors phosphorylation of approximately half the available Tyr1 residues . With our knowledge of the previously undescribed Tyr1 phosphorylation pattern installed by c-Abl , we were curious if such a pattern could affect the phosphorylation of CTD by P-TEFb . We first determined if the pre-treatment of CTD by c-Abl alters the number of phosphates added by P-TEFb using MALDI-TOF . Since c-Abl phosphorylates tyrosine and P-TEFb phosphorylates serine residues as determined ( Figure 2A , B and 1A , respectively ) , if the two phosphorylation events are independent , the number of phosphates placed by the two kinases should be additive . C-Abl phosphorylation of yCTD alone adds up to 13 phosphates ( Figure 2D ) , and P-TEFb alone adds 24–29 phosphates ( Figure 1B ) . Interestingly , tandem treatment of yCTD with c-Abl followed by P-TEFb resulted in the addition of a total of 16 to 26 phosphates as detected by MALDI-MS , with a mass shifts of 1287 . 3 to 2073 . 2 Da ( Figure 3A ) . These data reveal c-Abl pre-treatment results in changes to P-TEFb’s phosphorylation along the CTD , evidenced by a reduction in the number of phosphate groups added by P-TEFb . To identify the position of phosphates added by P-TEFb when Tyr1 is phosphorylated , we quantified pSer2 and pSer5 by immunoblotting with antibodies recognizing Ser2 and Ser5 phosphorylations ( Figure 3B ) . Compared to a non-phosphorylated CTD , the pre-treatment of yCTD with c-Abl results in a significant increase in Ser2 phosphorylation of nearly 300% , as detected by pSer2 specific CTD antibody 3E10 , accompanied by a small and statistically non-significant decrease in pSer5 installed by P-TEFb ( Figure 3B and C ) . The increase of pSer2 is unique for P-TEFb-mediated phosphorylation of CTD since a similar tandem treatment of yCTD by c-Abl followed by either TFIIH or Erk2 showed no changes in pSer2 levels ( Figure 3C and Figure 3—figure supplement 1A ) . We propose two possible explanations for the apparent increase of pSer2 levels upon c-Abl/P-TEFb treatment: First , c-Abl interacts with and/or modifies P-TEFb and alters its specificity from Ser5 to Ser2 . Alternatively , c-Abl may phosphorylate substrate CTD and these phosphorylations prime P-TEFb specificity towards Ser2 residues of the CTD . To differentiate these two models , we inactivated c-Abl after its reaction with CTD but before the addition of P-TEFb . We used two independent methods to inactivate c-Abl prior to P-TEFb addition: introduction of the potent Abl inhibitor dasatinib to 10 µM or denaturation of c-Abl via heat-inactivation ( Figure 3—figure supplement 1B and C ) . In the first method , the introduction of dasatinib inhibits c-Abl activity towards the CTD but shows no effect in P-TEFb’s ability to phosphorylate the CTD substrate ( Figure 3—figure supplement 1C ) . In the second method , the heat-inactivation effectively abolishes the kinase activity of c-Abl ( Figure 3—figure supplement 1B ) . In both experiments , P-TEFb continues to install a greater amount of Ser2 phosphorylation relative to no c-Abl treatment controls ( Figure 3—figure supplement 1D ) . Therefore , the increase in the apparent Ser2 phosphorylation is not due to P-TEFb’s physical interaction with c-Abl but arises from c-Abl kinase activity against CTD substrates at Tyr1 . Since immunoblotting presents the issue of epitope masking in highly phosphorylated protein samples and lacks the ability to separate reaction products in high resolution , we sought a method to identify the phosphorylation sites on the CTD directly and precisely . To determine the phosphorylation pattern resulting from sequential kinase treatment , we used LC-UVPD-MS/MS to investigate the activity of P-TEFb in the context of Tyr1 phosphorylation . LC-UVPD-MS/MS provides single residue resolution and overcomes artifacts inherent to immunoblotting such as epitope masking . Unfortunately , full-length yeast CTD is resistant to proteolysis due to a lack of basic residues , hindering further analysis by tandem MS ( Schüller et al . , 2016; Suh et al . , 2016 ) . Novel proteases , such as chymotrypsin and proteinase K , that cleave at bulky hydrophobic residues like tyrosine have proven effective in the past for analyzing the native sequence of the CTD , but proteolysis becomes inhibited upon phosphorylation due to the modification on tyrosine ( Mayfield et al . , 2017 ) . Short synthetic peptides circumvent the need for proteases but poorly mimic the physiological CTD and are unlikely to reveal bona fide CTD kinase specificities . To overcome these technical challenges , we generated a full-length yeast CTD with lysine replacing Ser7 in every other repeat ( yCTD-Lys ) ( Figure 1—figure supplement 1C ) . This allowed for trypsin digestion into di-heptads , which represent the functional unit of the CTD ( Corden , 2013; Eick and Geyer , 2013 ) , and are amenable to MS/MS analysis . To validate that the introduction of lysine residues does not bias kinase specificity , we first mapped the phosphorylation pattern of c-Abl or P-TEFb individually along yCTD-Lys using LC-UVPD-MS/MS . When treated with c-Abl , two single phosphorylation species are found at equivalent abundances with tyrosine at the same or neighboring heptad of Lys replacement ( Figure 3D and Figure 3—figure supplement 2A , peak 1 and 2 ) and a small peak in which both Tyr1 residues are phosphorylated ( Figure 3D and Figure 3—figure supplement 2A , peak 3 ) . When treated with P-TEFb alone , we observed four single phosphorylated peptides: two almost equally abundant peaks containing di-heptads with a single Ser5 phosphorylation ( Figure 3E and Figure 3—figure supplement 2B , peak 1 and 2 ) and two peaks about ~40 fold less in intensity with pSer2 or pSer7 ( Figure 3E and Figure 3—figure supplement 2B , peaks 3 and 4 ) . This result shows that P-TEFb strongly favors pSer5 in unmodified CTD substrates , consistent with our previously analysis ( Figure 1A ) . Double phosphorylated species are also detected for di-heptads with both Ser5 residues phosphorylated as the predominant product ( Figure 3E and F and Figure 3—figure supplement 2B , peak 5 ) . Several very small peaks ( less than 100-fold lower in intensity ) are identified as peptides containing both Ser5 and pSer2 ( Figure 3E and Figure 3—figure supplement 2B , peak 6 ) . These results indicate that the existence of Lys residue does not seem to bias kinase activity and is consistent with our previous results that P-TEFb strongly prefers to phosphorylate Ser5 . When treated in tandem with c-Abl followed by P-TEFb and digested with trypsin , di-heptads ( YSPTSPSYSPTSPK ) in a variety of phosphorylation states are generated . These species were separated in liquid chromatography ( LC ) and revealed nine di-heptad species of varying abundances . LC purification separates the different phosphorylation states of the di-peptide ( Figure 3G ) . Some of the di-heptads contain only single phosphorylations due to incomplete reactions in vitro . To understand the effect of c-Abl CTD phosphorylation on P-TEFb , we focused on phosphorylated species with more than one phosphate added , especially those containing both tyrosine and serine phosphorylation ( Figure 3G ) . Tandem phosphorylation generated species unique to those observed in c-Abl or P-TEFb individual treatment ( elution at 27–31 min in LC , Figure 3G , peak 6 , 7 , 8 ) . The most abundant of these unique species ( Figure 3G peak 7 ) contains both Tyr1 and Ser2 phosphorylation ( Figure 3H and Figure 3—figure supplement 2C ) . Similarly , a close-by but less abundant peak also contains Tyr1 and Ser2 double phosphorylation although in a different location ( Figure 3G peak eight and Figure 3—figure supplement 2C ) . Only a small peak contains both Tyr1 and Ser5 phosphorylation ( Figure 3G peak six and Figure 3—figure supplement 2C ) . Although we cannot exclude the possibility of the existence of other phosphorylated species containing a mixture of tyrosine and serine phosphorylation , their quantity is likely very low and not detected in LC-UVPD-MS/MS analysis . P-TEFb’s serine residue preference is dramatically different between reactions on unmodified CTD substrate , where pSer5 predominates ( Figure 3E peak 5 ) , and those pre-treated with c-Abl where pSer2 is the primary product species ( Figure 3G peak 7 ) . Our results show that in di-heptads with Tyr1 phosphorylated , Ser2 becomes the primary target of P-TEFb phosphorylation . The high performance of LC chromatography also allowed us to confirm our phosphomapping within di-heptads of the yeast CTD that diverge from the consensus sequence ( Figure 1—figure supplement 1C ) . Three di-heptads of the divergent sequence were generated following trypsin digestion of the yCTD-Lys construct ( Figure 3—figure supplement 3 , sequences of YSPTSPAYSPTSPK , YSPTSPNYSPTSPK , and YSPTSPGYSPGSPK ) . Although these di-heptads exist in a much smaller amount than the dominant product YSPTSPSYSPTSPK , they can be resolved and purified in high-performance liquid chromatography and analyzed for phosphorylation position ( Figure 3—figure supplement 3 ) . In these three phosphorylated di-heptads , the sole detected product of tandem treatment is a di-heptad with Tyr1 and Ser2 phosphorylated ( Figure 3—figure supplement 3 , right panels ) . In contrast , all peptides phosphorylated by P-TEFb alone gave a predominant di-heptad species containing only pSer5 ( Figure 3—figure supplement 3 , left panels ) . No Tyr1 and Ser5 double phosphorylation species were captured , possibly due to low abundance . The phosphoryl mapping of the various di-heptads generated provides independent evidence that Tyr1 phosphorylation promotes Ser2 phosphorylation by P-TEFb even in the context of divergent heptads . To further corroborate the mass spectrometry results that the specificity of P-TEFb is altered from Ser5 to Ser2 upon Tyr1 phosphorylation , we generated a new yeast CTD variant with 13 repeats ( half of the full-length yeast CTD ) with every single Ser5 mutated to alanine ( S5A construct , Figure 1—figure supplement 1H ) . Previously , it was shown that replacing Ser5 in CTD prevents its phosphorylation by P-TEFb ( Czudnochowski et al . , 2012 ) . Treatment of the S5A constructs with P-TEFb alone results in the addition of little to no phosphate shown by MALDI-MS ( Figure 3I ) . However , when the S5A construct is treated with c-Abl , it accepts up to five phosphate groups ( Figure 3J ) . Subsequent treatment with P-TEFb results in an obvious shift in the MALDI-MS spectra with up to seven phosphates added to the final product ( Figure 3K ) . The results corroborate the conclusion drawn from the MS/MS results , indicating that upon Tyr1 phosphorylation S5A becomes a viable substrate for P-TEFb and adds at least two additional phosphates likely to Ser2 residues . The observation of pSer2 as the major product in the context of pre-existing pTyr1 is interesting because P-TEFb has consistently shown a strong preference for Ser5 in vitro . Using a combination of immunoblotting , LC-UVPD-MS/MS , mutagenesis , and MALDI-MS we found Tyr1 phosphorylation primes the CTD for subsequent modification on Ser2 by P-TEFb via alteration of its specificity from Ser5 . The observation that Tyr1 phosphorylation by c-Abl alters the specificity of P-TEFb from Ser5 to Ser2 prompted us to ask if other kinases recruited to the CTD at the beginning of transcription can also alter P-TEFb specificity in vitro . TFIIH is a kinase that acts during transcription initiation and promotes P-TEFb function in vivo ( Ebmeier et al . , 2017 ) . To evaluate if a modification or combination of modifications installed by TFIIH can promote the Ser2 specificity of P-TEFb as we see with Tyr1 phosphorylation , we reconstructed CTD phosphorylation in vitro by treating yCTD substrates sequentially with TFIIH followed by P-TEFb and analyzed the resultant phosphorylation pattern using LC-UVPD-MS/MS and immunoblotting ( Figure 3—figure supplement 4 ) . When followed by P-TEFb , three phosphorylated species are generated , as revealed by LC-UVPD-MS/MS: two major species containing Ser5 phosphorylation and a minor species containing Ser2 phosphorylation ( Figure 3—figure supplement 4A–B ) . These peptides are reminiscent of those generated by P-TEFb alone where Ser5 phosphorylation dominates ( Figure 1A ) . These data indicate that TFIIH-mediated phosphorylations do not alter P-TEFb specificity in vitro . Our kinase assays have shown that Tyr1 phosphorylation by c-Abl alters the specificity of P-TEFb allowing for Ser2 phosphorylation of the CTD in vitro . To evaluate the importance of pTyr1 to Ser2 phosphorylation in human cells , we sought to selectively reduce pTyr1 levels and monitor pSer2 via western blot ( Figure 4A and B ) . Available literature suggests that c-Abl is important to Tyr1 phosphorylation in RNA polymerase II but not the sole kinase responsible ( Baskaran et al . , 1999 ) . Other Abl-like kinases may likely compensate for the function of c-Abl by phosphorylating Tyr1 in human cells ( Baskaran et al . , 1997 ) . Therefore , we initially utilized the potent inhibitor dasatinib , which inhibits c-Abl as well as other tyrosine kinases similar to c-Abl , to treat HEK293T cells ( Winter et al . , 2012 ) . Tyr1 phosphorylation has also been implicated in stabilizing RNA polymerase II in the cytosol ( Hsin et al . , 2014 ) , so marked reduction of Tyr1 phosphorylation may lead to a decrease in the global level of RNA polymerase II resulting in an apparent decrease in CTD phosphorylation levels . To address this potential artifact , we optimized inhibitor concentration to a level at which global RNA polymerase II levels are not significantly altered as determined by immunoblotting against RNA polymerase II subunits POLR2A and POLR2C ( Figure 4 and Figure 4—figure supplement 1A ) . At 10 µM dasatinib , pTyr1 levels are reduced by 30% in HEK293T cells , and this is accompanied by a 29% decrease in Ser2 phosphorylation ( Figure 4A ) . Importantly , pSer5 levels were not significantly altered ( Figure 4A and Figure 4—figure supplement 1A ) . To more specifically target Abl-mediated Tyr1 phosphorylation , we utilized the highly specific inhibitor imatinib that has a much smaller inhibitory repertoire with strong inhibition to c-Abl and Abl2 ( Winter et al . , 2012 ) . Treatment of HEK293T cells with 20 μM imatinib results in a reduction in pTyr1 of 35% . This is accompanied by a statistically significant decrease in pSer2 levels of 15% ( Figure 4B ) . One potential concern is antibody masking by flanking phosphorylation because Tyr1 phosphorylation can block recognition of pSer2 by the 3E10 antibody ( Chapman et al . , 2007 ) . Thus , loss of pTyr1 by inhibition with small molecules should produce an increase in the pSer2 signal if pSer2 levels remain constant . The fact that we observe a significant reduction in the pSer2 signal suggests it is indeed decreasing , but the 15–29% reduction quantified is likely an underestimate of the decrease given existing knowledge about these antibodies . Compounded with the mass spectrometry results , our data support that pTyr1 promotes Ser2 phosphorylation . In both the dasatinib and imatinib treatments pSer5 , POLR2A , and POLR2C levels remain unaffected ( Figure 4 and Figure 4—figure supplement 1A ) . Data from this inhibitor-based approach are in line with our in vitro observation that Ser2 phosphorylation is specifically coupled to Tyr1 phosphorylation and extends these conclusions to cellular contexts . To understand the biological implication of coupled Tyr1 and Ser2 phosphorylations at the level of individual genes , we conducted ChIP-seq analysis for the distribution of RNA Polymerase II upon the inhibition of Tyr1 phosphorylation . To carry out this experiment , we inhibited c-Abl with the potent small-molecule inhibitor dasatinib in HEK293T cells under conditions where pTyr1 is significantly reduced , but overall Pol II amount is unaffected ( Figure 4A and Figure 4—figure supplement 1A ) . The sample was prepared for ChIP-seq studies using RNA polymerase II antibody ( 8WG16 ) for immunoprecipitation to analyze the distribution of RNA polymerase II in a genome-wide fashion . In comparison of the dasatinib treated cells with the vehicle controls , the distribution of RNA polymerase II along the gene body is altered in multiple genes ( Figure 5A ) . Signal was normalized to 10M reads for both samples , and a significant increase of peak height for RNA polymerase II was found in the promoter region of many genes , as demonstrated in representative genes Myc and FANCL ( Figure 5A ) . To quantify the change of distribution of the polymerase , we calculated the pausing index ( Zeitlinger et al . , 2007 ) which is the ratio of Pol II read density in the region −50 to +300 bp of Transcription starting site ( TSS ) to the rest of gene body 3000 bp downstream of Transcription end site ( TES ) . The genes are clustered into four groups based on the pausing index: the G0 cluster has a pausing score close to 0; the remaining genes were ranked based on their pausing scores from high to low with the G1 cluster containing genes with pausing scores less than the lower quartile , G2 with genes contained between the lower and upper quartile , and G3 with pausing scores above the upper quartile . Genes in G0 and G1 have little occupancy of the polymerase and might not be active ( Figure 5B ) . A meta-analysis , as visualized in box plot for the pausing index of the genes in G0 and G1 , shows no statistical difference between control and treatment samples ( Figure 4—figure supplement 1B and D ) . But a statistically significant increase can be observed close to two-fold in G2 and G3 genes ( Figure 5C and Figure 4—figure supplement 1C–D ) . The same trend is observed across biological duplicates . Overall , these results suggest that RNA polymerase II is stalled in the promoter region upon the inhibition of Tyr1 phosphorylation .
Our discovery that Tyr1 phosphorylation of the CTD alters the preference of P-TEFb from Ser5 to Ser2 resolves the controversy surrounding P-TEFb’s specificity ( Bartkowiak et al . , 2010; Czudnochowski et al . , 2012 ) . P-TEFb was initially identified as a CTD kinase that controls the elongation potential of RNA polymerase II , is required for the majority of RNA polymerase II transcription , and is specific for Ser2 in vivo ( Chao and Price , 2001; Marshall et al . , 1996; Ni et al . , 2004 ) . However , these early conclusions are at odds with in vitro data demonstrating P-TEFb is incapable of phosphorylating Ser2 of CTD peptides in vitro ( Czudnochowski et al . , 2012 ) . Two other kinases , CDK12 and CDK13 , display Ser2 kinase activity in cells but do not seem to play a major role in Ser2 phosphorylation in early transcriptional events ( Bartkowiak et al . , 2010; Chen et al . , 2007 ) . Investigations on the effect of CTD phosphorylations on P-TEFb specificity have revealed that Ser7 ( Czudnochowski et al . , 2012 ) and Ser5 ( this manuscript ) do not alter its preference for Ser5 . Using direct methods , like mass spectrometry confirmed by immunoblotting and EMSA , we identified that Tyr1 phosphorylation could alter the specificity of P-TEFb from Ser5 to Ser2 in vitro . It should be stressed that continuous heptad repeats with phosphorylated pTyr1 inhibit subsequent CTD modification by P-TEFb ( Czudnochowski et al . , 2012 ) . However , when treated biochemically with c-Abl , the pTyr1 pattern not only allows for P-TEFb phosphorylation but also shift the substrate preference from Ser5 to Ser2 . Furthermore , inhibition of Tyr1 phosphorylation leads to the reduction of Ser2 phosphorylation in human cells and the accumulation of Pol II in the promoter-proximal pausing stage of transcription . Therefore , we show that Tyr1 phosphorylation potentiates Ser2 phosphorylation of the CTD by altering P-TEFb specificity . The PTM state of the CTD has been correlated to the progression of transcription ( Jeronimo et al . , 2016 ) . Traditionally , such analyses are interpreted through a paradigm considering heptads phosphorylated on a single isolated residue with Ser5 phosphorylation dominating the initiation stage of transcription and Ser2 phosphorylation dominating elongation and termination ( Corden , 2013 ) . However , this simplification of CTD modification cannot explain the well-coordinated recruitment of the myriad CTD binding factors currently implicated in eukaryotic transcription ( Ebmeier et al . , 2017; Eick and Geyer , 2013; Harlen and Churchman , 2017 ) . Data presented here point to a sophisticated model in which the phosphorylation of Tyr1 at the beginning of transcription sets the stage for future coding events . This interplay between c-Abl and P-TEFb results in a chemically distinct phospho-CTD landscape compared to CTD phosphorylated by a single kinase . The combination of these modification modes likely contributes to a heterogeneous collection of modified heptads , which recruits the diverse array of CTD binding partners in a coordinated manner . These results are in good agreement with the ‘CTD code’ hypothesis proposed decades ago where different combinations of post-translational events result in different transcriptional outcomes . Tyr1 phosphorylation has been implicated in stabilizing RNA polymerase II in cells ( Hsin et al . , 2014 ) , transcription termination ( Mayer et al . , 2012 ) and anti-sense transcription ( Descostes et al . , 2014 ) but a coherent molecular basis for these disparate functions remains elusive . Our analysis provides a molecular mechanism demonstrating how Tyr1 phosphorylation can affect subsequent phosphorylation events carried out by other CTD kinases . The ability of Tyr1 phosphorylation to redirect signaling and influence subsequent modifications along the CTD , as revealed for P-TEFb , suggests these various roles for pTyr1 may arise indirectly . It can function through its impact on downstream CTD modifiers , highlighting integrated , indirect , and context-specific mechanisms for pTyr1 during co-transcriptional signaling . The final accumulation of individual species is dependent on the dynamic interplay of CTD kinases and phosphatases throughout the transcription cycle . Tyr1 phosphorylation is relatively transient , appearing at the transition from initiation to elongation and decreasing rapidly through the action of phosphatase ( s ) ( Eick and Geyer , 2013 ) . Despite this transient nature , pTyr1 is positioned in a vital window to alter P-TEFb specificity and regulate its phosphorylation pattern along RNA polymerase II . The adjustability of P-TEFb specificity by nearby Tyr1 phosphorylation reveals a novel mechanism for the regulation of P-TEFb kinase activity . With many binding partners in cells for P-TEFb , there might be additional regulators promoting the pSer2 activity of P-TEFb independent of or cooperatively with Tyr1 phosphorylation . The data presented reconcile P-TEFb’s in vitro and in vivo specificity and inspires new queries fundamental to CTD biology . P-TEFb is ubiquitously important for transcription across eukaryotic cells and often co-opted in disease states like HIV infection and cancer ( Franco et al . , 2018 ) . The integrated CTD code revealed here represents a unique mechanism to manipulate P-TEFb and potentially other CTD modifiers . Future inquiries using similar multi-disciplinary approaches will hopefully reveal CTD modification patterns in greater detail at different stages of the transcription process in single amino acid resolution . Information such as this will define the temporal and spatial signaling allowing for the recruitment of transcriptional regulators during active transcription . Overall , our findings support a model in which cross-talk between CTD modification enzymes increases the diversity and coding potential of CTD heptads . This expands the lexicon of phosphorylation marks and can provide more specific recruitment of transcription regulators allowing for the precise control of eukaryotic transcription .
CTD coding sequences ( Figure 1—figure supplement 1 ) were subcloned into pET28a ( Novagene ) derivative vectors encoding an N-terminal His-tag a GST-tag and a 3C-protease site to generate GST-CTD constructs as described previously ( Mayfield et al . , 2017 ) . 3CTD-5CTD and YtoF variants of CTD coding portions ( Figure 1—figure supplement 1 ) were amplified from synthetic DNA templates generated by IDT . The S5A variant DNA and the S7K spaced DNA constructs were purchased from Biomatik as synthetic genes , amplified and subsequently cloned into the pET28a derivative vector described above . Homo sapiens Erk2 was expressed from pET-His6-ERK2-MEK1_R4F_coexpression vector as a gift from Melanie Cobb ( Addgene plasmid #39212 ) ( Khokhlatchev et al . , 1997 ) . E . coli BL21 ( DE3 ) cells grown 37°C in Luria-Bertani ( LB ) media were used to overexpress recombinant GST-CTDs variants and Erk2 . Briefly , proteins were overexpressed in E . coli BL21 ( DE3 ) cells by growing at 37°C in LB media containing 50 μg/mL kanamycin to an OD600 of 0 . 4–0 . 6 . Expression was induced by the addition of isopropyl-β-D-thiogalactopyranoside ( IPTG ) to a final concentration of 0 . 5 mM . After induction , the cultures were grown at 16°C for an additional 16 hr . The cells were pelleted and lysed via sonication in lysis buffer [50 mM Tris-HCl pH 8 . 0 , 500 mM NaCl , 15 mM Imidazole , 10% Glycerol , 0 . 1% Triton X- 100 , 10 mM β-mercaptoethanol ( BME ) ] . The lysate was cleared by centrifugation , and the supernatant was initially purified using Ni-NTA ( Qiagen ) beads and eluted with elution buffer ( 50 mM Tris-HCl pH 8 . 0 , 500 mM NaCl , 200 mM Imidazole , and 10 mM BME ) . The protein was dialyzed against gel filtration buffer ( 20 mM Tris-HCl pH 8 . 0 , 50 mM NaCl , 10 mM BME for GST-yCTD and 20 mM Tris-HCl pH 7 . 5 , 200 mM NaCl , 10 mM BME ) . Finally , proteins were concentrated and ran on a Superdex 200 gel filtration column ( GE ) . Erk2 was purified using a previously published protocol ( Khokhlatchev et al . , 1997 ) . Homogeneity of the eluted fractions was determined via Coomassie Brilliant Blue stained SDS-PAGE . Samples were concentrated in vivaspin columns ( Sartorius ) . Abl kinase treated CTD reactions were prepared in buffer conditions containing 1 μg/μL GST-yCTD substrate , 0 . 0035 μg/μL c-Abl kinase , 50 mM Tris-HCl at pH7 . 5 , 50 mM MgCl2 and 2 mM ATP . TFIIH treated CTD reaction were prepared in buffer conditions containing 1 μg/μL GST-CTD substrate , 0 . 025 μg/μL TFIIH , 50 mM Tris-HCl at pH7 . 5 , 50 mM MgCl2 and 2 mM ATP . P-TEFb treated CTD reaction , were prepared in buffer conditions containing 1 μg/μL GST-CTD substrate , 0 . 0075 μg/μL P-TEFb , 50 mM Tris-HCl at pH7 . 5 , 50 mM MgCl2 and 2 mM ATP . Erk2-treated CTD reaction , as well as the controls with no kinase treatment , were prepared in buffer conditions containing 1 μg/μL GST-CTD substrate , 0 . 025 μg/μL Erk2 , 50 mM Tris-HCl at pH7 . 5 , 50 mM MgCl2 and 2 mM ATP . These reactions were incubated for various amount of time at 30°C along with control experiments setup under identical conditions but without kinases and then stored at −80°C until analysis . Tandem kinase treatments were performed by mixing 10 μg GST-CTD substrate treated with the c-Abl as described above ( incubated overnight for 16 hr ) and an equal volume of a solution containing the second kinase ( 0 . 05 μg/μL TFIIH or 0 . 015 μg/μL P-TEFb or 0 . 05 μg/μL Erk2 ) in tandem reaction buffer ( 50 mM Tris-HCl at pH7 . 5 , 50 mM MgCl2 , 2 mM ATP ) . These were incubated at 30°C for 16 hr and stored at −80°C until analysis . The Erk2 kinetic activity assay was performed in a 25 µl reaction volume containing 0–100 µM substrate ( GST yCTD or GST YtoF CTD ) and a reaction buffer of 40 mM Tris-HCl at pH 8 . 0 and 20 mM MgCl2 . The reaction was initiated by adding 187 nM of Erk2 and incubated at 28°C for 15mins before being quenched with 25 µl H2O and 50 µl of room temperature Kinase-Glo Detection Reaction ( Promega ) . The mixtures were allowed to sit at room temperature for 10 min before reading the bioluminescence in a Tecan Plate reader 200 . The readings obtained were translated to ATP concentration with the help of an ATP standard curve determined with the Kinase Detection Reagent . The TFIIH kinetic reactions were set up with 0–100 µM substrate ( GST yCTD or GST YtoF CTD ) , 0 . 2 µM TFIIH , 0 . 1 mg/ml Bovine Serum Albumin ( BSA ) and reaction buffer of 50 mM Tris-HCl pH 8 . 0 , 10 mM MgCl2 , 1 mM DTT . 500 µM ATP Mix ( 10 nCi/µl radiolabeled ATP , PerkinElmer ) was added to each tube to start the reactions . The tubes were subsequently incubated in a 30°C water bath for 30mins and quenched with 500 µl of quench buffer ( 1 mM potassium phosphate pH 6 . 8 , 1 mM EDTA ) to a reaction volume of 10 µl . Each reaction was loaded onto 0 . 45 µm nitrocellulose filters and washed three times with 1 mM potassium phosphate buffer to remove any excess labeled ATP . Filters were added to glass vials with scintillation fluid , Econo-Safe Economical Biodegradable Counting Cocktail ( Research Products International ) and set in a scintillation counter for 5 min reads each . The amount of phosphate incorporation was determined for each reaction using a set of 147 pmol labeled ATP standards that were read alongside each reaction set . Kinetic data obtained from the two assays described above were analyzed in R ( Hamilton , 2015; R Development Core Team , 2017 ) and fitted to the Michaelis-Menten kinetic equation to obtain respective kinetic parameters kcat ( s−1 ) and Km ( µM ) . SDS-PAGE analysis was performed using 10–15% acrylamide gels containing 1% SDS . GST-CTD samples were prepared by boiling with SDS-PAGE loading dye at 95°C for 5 min . This was also used to quench time-course reactions . A volume containing approximately 1 μg of phosphorylated GST-CTD substrate or no kinase control was loaded into wells and resolved at ~150V for 1 hr at room temperature . Gels were stained with Coomassie Brilliant Blue and visualized on G: BOX imaging systems ( Syngene ) . Approximately 5 μg of GST-CTD protein from the kinase reactions described were prepared for MALDI-MS . If necessary , the protein was digested with 3C protease by mixing sample in a 1:10 ratio of 3C-protease to GST-CTD variants . Proteins were equilibrated with dilute trifluoracetic acid ( TFA ) to a final concentration of 0 . 1% TFA and a pH of <4 . These samples were desalted using ZipTip ( Millipore ) tips according to manufacturer's instructions . These samples were mixed 1:1 with a 2 , 5-dihydroxybenzoic acid matrix solution ( DHB ) and spotted on a stainless steel sample plate . The spots were allowed to crystallize at ambient temperature and pressure . MALDI-MS spectra were obtained on an AB Voyager-DE PRO MALDI-TOF instrument with manual adjustment of instrument parameters to ensure the greatest signal to noise . Sample masses were determined by an internal calibration against the untreated GST-CTD variants . Data analysis , noise reduction , and Gaussian smoothing , if necessary , were performed in DataExplorer ( AB ) , R , and the R package smoother ( Hamilton , 2015; R Development Core Team , 2017 ) . Masses were determined as the highest local intensity peak of the post-processed data . Data were visualized in R-Studio using ggplot2 ( Wickhan , 2009 ) . All MALDI experiments were carried out three times independently with biological triplicates . HEK293T cells were purchased from ATCC and no mycoplasma contamination was detected . The cells were maintained in DMEM ( ISC BioExpress cat#T-2989–6 ) with splitting every other day and seeding at a concentration of 9 . 6 × 104 cells per 10 cm culture dish and incubated at 37°C at 5% CO2 . Cells to be treated with imatinib , dasatinib , or vehicle control were plated at 5 × 105 cells per well in 6-well tissue culture plates in fresh DMEM . Cells were incubated for 24 hr , and the media was replaced with fresh DMEM containing the indicated amount of inhibitor or equivalent portion of DMSO vehicle control for an additional 24–40 hr . Protein preparations were generated by direct in-well lysis . Media was then removed , and the cells were washed with ice-cold Phosphate-buffered saline ( PBS ) , and 200 μL RIPA buffer ( 150 mM NaCl , 10 mM Tris-HCl pH 7 . 5 , 0 . 1% SDS , 1% Triton X-100 , 1% deoxycholate , and 5 mM EDTA ) supplemented to 1X with HALT protease and phosphatase inhibitor cocktail ( Thermo Scientific ) was added directly to cells . Plates were incubated on ice for 15 min with gentle shaking , and the lysate was transferred to microcentrifuge tubes . Samples were briefly sonicated to reduce viscosity and spun to remove cell debris . Protein concentration was determined to utilize Pierce BCA Protein Assay Kit ( Thermo Scientific ) against a BSA standard curve . Samples were diluted with SDS-PAGE loading buffer and boiled at 95°C for 5 min . The sample was aliquoted and frozen at −80°C . Total protein from cell lysate ( 20–40 μg ) or GST-yCTD samples ( 50 to 500 ng , dependent on epitope ) was loaded onto a 4–20% gradient SDS-PAGE gel ( Biorad , Cat#:456–1096 ) and ran at 150V for 50 min at room temperature in a Mini-PROTEAN Tetra Cell ( Biorad ) . The proteins were transferred to PVDF membrane at 100 V for 1 hr on ice in a Mini-PROTEAN Tetra Cell ( Biorad ) . Membranes were blocked in 1X TBST ( 20 mM Tris-HCl pH 7 . 5 , 150 mM NaCl , 0 . 1% Tween-20 ) and 5% ( BSA ) or non-fat dry milk for 1 hr at room temperature with shaking . Blocked membranes were incubated in primary antibodies in either 1X TBST or 1X TBST+5% BSA at 4°C overnight or 1 hr at room temperature . The membranes were then washed six times with 1X TBST for 5 min each at room temperature and incubated with secondary antibodies in 1X TBST for 1 hr at room temperature . The membrane was washed once again and incubated with SuperSignal West Pico Chemiluminescent Substrate ( Pierce ) according to factory directions . Blots were imaged using a G:BOX gel doc system ( Syngene ) and quantified in ImageJ ( Schneider et al . , 2012 ) . Statistical analysis was performed in R ( R Development Core Team , 2017 ) . Blots normalized against Coomassie-stained bands were stained by incubating the membrane post-immunoblotting with stain solution ( 0 . 1% Coomassie Brilliant Blue R-250 , 40% ethanol , 10% acetic acid ) for 1 min . The stain was discarded , and the blot was briefly rinsed with distilled water . Blot was de-stained in de-stain solution ( 10% ethanol , 7 . 5% acetic acid ) until bands were visible . Blots were equilibrated with distilled water and imaged wet in a plastic blot protector using a G:BOX gel doc system ( Syngene ) and quantified as above . For dot blot , samples of GST-yCTD were treated with P-TEFb alone or c-Abl followed by P-TEFb . The heat-inactivated samples were prepared by heating the c-Abl treated GST-yCTD at 60°C for five minutes , while the Dasatinib inactivated samples were prepared by adding 10 µM Dasatinib to the c-Abl treated sample 15 min before incubation with P-TEFb . The dot blots were performed by adding 2X SDS page loading dye and briefly heating each sample after which 1 µg of the sample was loaded in three replicates onto a 0 . 45 µm Nitrocellulose membrane . The membrane was subsequently allowed to dry and blocked in 1X TBST ( 20 mM Tris-HCl pH 7 . 5 , 150 mM NaCl , 0 . 1% Tween-20 ) and 5% BSA for 1 hr at room temperature with shaking . The Anti-RNA Pol II phosphoSer2 antibody , 3E10 , ( Millipore ) was diluted 1:5000 times and was incubated overnight at 6°C to probe for Ser2 phosphorylation . The membranes were then washed five times with 1X TBST for 5 min each at room temperature and incubated with secondary antibody in 1X TBST for 1 hr at room temperature . The membrane was washed once again and incubated with SuperSignal West Pico Chemiluminescent Substrate ( Pierce ) according to manufacturer's instruction . Blots were imaged using a G:BOX gel doc system ( Syngene ) and quantified in ImageJ ( Schneider et al . , 2012 ) . Statistical analysis was performed by using the Data Analysis function in Microsoft Excel . Primary and secondary antibodies were stripped for re-probing by incubating membranes with a mild stripping buffer ( 200mM glycine , 0 . 1% SDS , 1% Tween 20 , pH 2 . 2 ) for 10 minutes . Stripping buffer was discarded and this step was repeated once more . The blot was washed twice with 1X PBS for 10 minutes . The blot was washed twice with 1X TBST for 10 minutes . The membrane was blocked and re-probed with secondary antibody and chemiluminsecent reagent , as described above , to insure complete removal of both primary and secondary antibodies . The membrane was then probed for desired epitope as described above . GST-3CTD samples ( approximately 1 µg/µL ) were digested on ice for 4 hr in 50 mM Tris-HCl at pH 8 . 0 with 150 mM NaCl using 3C-protease at a molar ratio of 100:1 protein: protease in a reaction volume of 20 µL . Digests were desalted on C18 spin columns and resuspended to 1 µM with 0 . 1% formic acid for LC-MS analysis . Separations were carried out on a Dionex Ultimate 3000 nano liquid chromatograph plumbed for direct injection . Picofrit 75 µm id analytical columns ( New Objective , Woburn , MA ) were packed to 20 cm using 1 . 8 µm Waters Xbridge BEH C18 ( Milford , MA ) . Mobile phase A was water , and B was acetonitrile , each containing 0 . 1% formic acid . Separations occurred over a 30 min linear gradient from 2–35% B . The flow rate was maintained at 0 . 3 µL/min during the separation . An Orbitrap Fusion Lumos Tribrid mass spectrometer ( Thermo Fischer Scientific , San Jose , CA ) equipped with a Coherent ExciStar XS excimer laser operated at 193 nm was used for positive mode LC-MS/MS analysis of the 3CTD peptides . The Lumos mass spectrometer was modified for ultraviolet photodissociation ( UVPD ) as described earlier ( Klein et al . , 2016 ) . Photoactivation in the low-pressure linear ion trap was achieved using 2 pulses at 2 mJ in a targeted m/z mode . The 3+ charge states of the singly and doubly phosphorylated peptide GPGSGMYSPTSPSYSPTSPSYSPTSPS were targeted for photoactivation . All data were acquired in the Orbitrap analyzer where MS1 and MS/MS spectra were collected at resolving powers of 60K and 15K ( at m/z 200 ) , respectively . MS1 spectra were acquired from m/z 400–2000 with an AGC setting of 5E5 . Each MS/MS spectrum consisted of two microscans collected from m/z 220–2000 with an AGC setting of 2E5 . Data analysis was performed using the XCalibur Qual Browser and ProSight Lite ( Fellers et al . , 2015 ) . For both targeted m/z values , the MS/MS spectrum for each phosphoform present was deconvoluted to neutral forms using Xtract with a signal-to-noise threshold of 3 . Sequence coverage was determined by matching the nine ion types observed with UVPD ( a , a• , b , c , x , x• , y , y-1 , z ) . Localization of the phosphorylation ( s ) was performed by adding a phosphate group ( +79 . 966 Da ) at each of the possible serine , threonine , and tyrosine residues to identify fragment ions containing the moiety and optimize characterization scores in ProSight Lite . Analysis of yCTD treated with P-TEFb was performed identically to previous analysis of yCTD treated with TFIIH and Erk2 ( Mayfield et al . , 2017 ) . GST-yCTD samples were prepared for bottom-up analysis using a two-step proteolysis method . First , overnight digestion with trypsin at 37°C was carried out using a 1:50 enzyme to substrate ratio , which cleaved the GST-portion of the protein while leaving the abasic 26mer CTD portion intact . The resulting digest was filtered through a 10 kDa molecular weight cutoff ( MWCO ) filter to remove tryptic GST peptides and buffer exchange the CTD portion into 50 mM Tris-HCl pH 8 . 0 and 10 mM CaCl2 for subsequent proteinase K digestion . Proteinase K was added in a 1:100 ration and digested overnight at 37°C . Samples were diluted to 1 μM in 0 . 2% formic acid for LC-MS . Analysis of yCTD-Lys treated by c-Abl , P-TEFb or c-Abl followed by P-TEFb is using a similar method as described above except the first digestion was done by 3C-protease and second by trypsin . A bottom-up analysis of yCTD was performed on a Velos Pro dual linear ion trap mass spectrometer ( Thermo Fisher ) equipped with a Coherent ExciStar XS excimer laser ( Santa Clara ) at 193 nm and 500 Hz as previously described for UVPD ( Gardner et al . , 2008; Madsen et al . , 2010 ) . Two pulses of 2mJ were used for photodissociation . Separations were carried out on a Dionex Ultimate 3000 nano liquid chromatography ( Thermo Fischer ) configured for preconcentration . Integrafrit trap columns were packed to 3 . 5 cm using 5 μm Michrom Magic C18 . Picrofrit analytical columns were packed to 20 cm using 3 . 5 μm Waters Xbridge BEH C18 ( Waters ) . Mobile phase A was water , and mobile phase B was acetonitrile; each contained 0 . 1% formic acid . Peptides were loaded onto the trap column for 5 min in an aqueous solvent containing 2% acetonitrile and 0 . 1% formic acid at a 5 μL/min flow rate . Separations occurred over a 20 min linear gradient in which percent phase B was increased from 2–15% during the first 15 min and further increased to 35% over the last 5 min . The flow rate was constant at 0 . 3 μL/min . A top seven data-dependent acquisition method was first used to identify the main phosphorylated species . A targeted analysis followed in which the singly phosphorylated heptad peptides were continually selected for UVPD activation ( between MS1 acquisitions occurring after every five MS/MS events ) to resolve partially co-eluting phospho-isomers . All MS experiments were carried out three times independently with biological triplicates . Resulting UVPD spectra were manually interpreted . For ChIP-seq experiments , HEK293T cells were seeded at 3 . 5 million cells in a 15 cm dish . After 24 hr , when the cells achieved a confluence of 40–50% , the media was replaced by fresh media containing 10 µM Dasatinib inhibitor or the DMSO control and allowed to grow for another 24 hr until the confluence of 80% was achieved . The cells were fixed with 1% formaldehyde in 15 ml of media , for 8 min at room temperature with intermittent swirling . The reaction was quenched by the addition of glycine to a final concentration of 0 . 125M and incubation for five minutes at room temperature . The cells were washed twice with 15 ml of ice-cold Dulbecco's phosphate-buffered saline and scraped off the surface . The cells were pelleted at a speed of 8000 g for 5 min , resuspended and aliquoted such that the number of control cells ( with only DMSO ) were normalized to the number of dasatinib treated cells . The cell pellet was frozen in a freezing mixture comprised of dry ice and ethanol . The cells were lysed by adding buffer LB1 [50 mM HEPES at pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , 10% glycerol , 0 . 5% NP-40 , 0 . 25% Triton-X 100 , 1x Protease inhibitor cocktail ( Thermoscientific ) ] and placing the tubes on a rotating wheel at 4°C for 10 min , following which they were spun at 2000 g for 5 min to isolate the nuclei as a pellet . These were washed with buffer LB2 [10 mM Tris-HCL at pH 8 . 0 , 200 mM NaCl , 1 mM EDTA , 0 . 5 mM EGTA + 1X Protease inhibitor cocktail ( thermoscientific ) ] , and subsequently the nuclei where resuspended in 300 µl of nuclear lysis buffer LB3 [10 mM Tris-HCL , pH 8 , 100mMNaCl , 1 mM EDTA , 0 . 5 mM EGTA , 0 . 1% Na-Deoxycholate , 0 . 5% N-lauroylsarcosine and 1x Protease inhibitor cocktail ( Thermoscientific ) ] . The nuclear lysate of 300 µl was sonicated using a Biorupter UCD 200 ( Diagenode ) for 25 cycles at maximum intensity ( 15 s ON 45 s OFF in a water bath at 4°C ) . After each of the 10 cycles , the samples were incubated on ice for 10 min . Following sonication 30 µl of buffer LB3 supplemented with 10% Triton X-100 was added into the sample and spun at full speed for 10 min to remove cell debris . 30 µl of the supernatant was taken as the input control for ChIP-seq , and the rest is used to prepare the samples . Magnetic Protein-G beads ( Thermo Fischer ) were incubated with respective antibody ( 1 µg per 10 µl of beads ) overnight on the rotating shaker at 4°C . The beads were then washed thrice with 5% BSA in PBS to remove any excess antibody , and the 300 µl of the sonicated lysate prepared above is added to it , with 800 µl of buffer LB3 and 100 µl of buffer LB3 supplemented with10% Triton X-100 . The samples were placed on a rotating wheel overnight at 4°C for the immunoprecipitation to occur . The beads were washed twice by a low salt buffer ( 0 . 1% sodium deoxycholate , 1% Triton X-100 , 1 mM EDTA , 50 mM HEPES at pH 7 . 5 , 150 mM NaCl ) followed by wash with high salt buffer ( 0 . 1% Na Deoxycholate , 1% Triton X-100 , 1 mM EDTA , 50 mM HEPES at pH 7 . 5 , 500 mM NaCl ) , lithium chloride buffer ( 250 mM LiCl , 0 . 5% NP-40 , 0 . 5% Na Deoxycholate , 1 mM EDTA , 10 mM Tris-HCl at pH 8 . 1 ) and finally washed twice with TE buffer ( 10 mM Tris-HCl at pH 8 . 1 and 1 mM EDTA ) . The beads were ultimately resuspended in 200 µl of elution buffer ( 1% SDS and 0 . 1M sodium bicarbonate ) and placed in the thermomixer at 65°C for 16 hr to enable reverse crosslinking . Both the input and treatment samples were with 70 µl of elution buffer ( 1% SDS and 0 . 1M sodium bicarbonate ) and underwent reverse crosslinking at 65°C for 16 hr . After the reverse crosslinking , phenol-chloroform extraction was used to extract the immunoprecipitated DNA , Library prep was done using a starting amount of 3 ng of DNA measured by Qubit HS ( Thermo Fischer ) using the NEBNext Ultra II DNA Library Prep Kit for Illumina ( NEB ) following the vendor manual . The libraries with multiplex index primers prepared above were pooled together and sequenced using the NextSeq single end 75 base pair sequencing platform . Reads were aligned to the human genome ( hg19 ) using bowtie with ‘--best --strata –m 1’ parameters ( Langmead et al . , 2009 ) . Only uniquely mapped reads were selected for downstream analysis . MACS2 was employed to call peaks by comparing immunoprecipitated chromatin with input chromatin using standard parameters and a q-value cutoff of 1e-5 ( Zhang et al . , 2008 ) . The peaks overlapped with the blacklist regions downloaded from UCSC were removed . Each sample was normalized to 10 million mapped reads and visualized in Integrative Genomics Viewer ( IGV ) ( Robinson et al . , 2011 ) . The pausing index was defined as the ratio of Pol II density in the promoter-proximal region and the Pol II density in the transcribed region ( Zeitlinger et al . , 2007 ) . The proximal promoter region is defined as −50 bp to +300 bp around the transcription start site ( TSS ) ; while the transcribed region ( gene body ) is from +300 bp to the 3000 bp downstream of transcription end site ( TES ) ( Rahl et al . , 2010 ) . MALDI data analysis , noise reduction , and Gaussian smoothing ( if necessary ) were performed in DataExplorer ( AB ) , R , and the R package smoother to provide interpretable data ( Hamilton , 2015; R Development Core Team , 2017 ) . Data were visualized in R-Studio using ggplot2 ( Wickhan , 2009 ) . LC masses were determined as the highest local intensity peak of the post-processed data . Tandem mass spectrometry data analysis was performed using the XCalibur Qual Browser and ProSight Lite . For both targeted m/z values , the MS/MS spectrum for each phosphoform present was deconvoluted to neutral forms using Xtract with a signal-to-noise threshold of 3 . Sequence coverage was determined by matching the nine ion types observed with UVPD ( a , a• , b , c , x , x• , y , y-1 , z ) . Localization of the phosphorylation ( s ) was performed by adding a phosphate group ( +79 . 966 Da ) at each of the possible serine , threonine , and tyrosine residues to identify fragment ions containing the moiety and optimize characterization scores in ProSight Lite ( Fellers et al . , 2015 ) . Western blots were quantified using ImageJ ( Schneider et al . , 2012 ) and statistical significance was determined by two-tailed unpaired Student’s t-test assuming unequal variances in Microsoft Excel . Statistical significance is reported in the figure legends . Results have been shown with ± standard deviation or SEM , as mentioned in the figure legends . Access Code: The ChIP-seq data for RNA polymerase II have been deposited into GEO with access codes GSE131838 . | DNA contains the instructions for making proteins , which build and maintain our cells . So that the information encoded in DNA can be used , a molecular machine called RNA polymerase II makes copies of specific genes . These copies , in the form of a molecule called RNA , convey the instructions for making proteins to the rest of the cell . To ensure that RNA polymerase II copies the correct genes at the correct time , a group of regulatory proteins are needed to control its activity . Many of these proteins interact with RNA polymerase II at a region known as the C-terminal domain , or CTD for short . For example , before RNA polymerase can make a full copy of a gene , a small molecule called a phosphate group must first be added to CTD at specific units known as Ser2 . The regulatory protein P-TEFb was thought to be responsible for phosphorylating Ser2 . However , it was previously not known how P-TEFb added this phosphate group , and why it did not also add phosphate groups to other positions in the CTD domain that are structurally similar to Ser2 . To investigate this , Mayfield , Irani et al . mixed the CTD domain with different regulatory proteins , and used various biochemical approaches to examine which specific positions of the domain had phosphate groups attached . These experiments revealed a previously unknown aspect of P-TEFb activity: its specificity for Ser2 increased dramatically if a different regulatory protein first added a phosphate group to a nearby location in CTD . This additional phosphate group directed P-TEFb to then add its phosphate specifically at Ser2 . To confirm the activity of this mechanism in living human cells , Mayfield , Irani et al . used a drug that prevented the first phosphate from being added . In the drug treated cells , RNA polymerase II was found more frequently ‘stalled’ at positions on the DNA just before a gene starts . This suggests that living cells needs this two-phosphate code system in order for RNA polymerase II to progress and make copies of specific genes . These results are a step forward in understanding the complex control mechanisms cells use to make proteins from their DNA . Moreover , the model presented here – one phosphate addition priming a second specific phosphate addition – provides a template that may underlie similar regulatory processes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2019 | Tyr1 phosphorylation promotes phosphorylation of Ser2 on the C-terminal domain of eukaryotic RNA polymerase II by P-TEFb |
Variation in floral displays , both between and within species , has been long known to be shaped by the mutualistic interactions that plants establish with their pollinators . However , increasing evidence suggests that abiotic selection pressures influence floral diversity as well . Here , we analyse the genetic and environmental factors that underlie patterns of floral pigmentation in wild sunflowers . While sunflower inflorescences appear invariably yellow to the human eye , they display extreme diversity for patterns of ultraviolet pigmentation , which are visible to most pollinators . We show that this diversity is largely controlled by cis-regulatory variation affecting a single MYB transcription factor , HaMYB111 , through accumulation of ultraviolet ( UV ) -absorbing flavonol glycosides in ligules ( the ‘petals’ of sunflower inflorescences ) . Different patterns of ultraviolet pigments in flowers are strongly correlated with pollinator preferences . Furthermore , variation for floral ultraviolet patterns is associated with environmental variables , especially relative humidity , across populations of wild sunflowers . Ligules with larger ultraviolet patterns , which are found in drier environments , show increased resistance to desiccation , suggesting a role in reducing water loss . The dual role of floral UV patterns in pollinator attraction and abiotic response reveals the complex adaptive balance underlying the evolution of floral traits .
The diversity in colour and colour patterns found in flowers is one of the most extraordinary examples of adaptive variation in the plant world . As remarkable as the variation that we can observe is , even more of it lays just outside our perception . Many species accumulate pigments that absorb ultraviolet ( UV ) radiation in their flowers; while these patterns are invisible to the human eye , they can be perceived by pollinators , most of which can be seen in the near UV ( Chittka et al . , 1994; Tovée , 1995 ) . UV patterns have been shown to increase floral visibility and to have a major influence on pollinator visitation and preference ( Brock et al . , 2016; Horth et al . , 2014; Rae and Vamosi , 2013; Sheehan et al . , 2016 ) . Besides their importance for pollinator attraction , patterns of UV-absorbing pigments in flowers have increasingly been recognized to have a role in responses to other biotic and abiotic factors , including defence against insect herbivory ( Gronquist et al . , 2001 ) , protection against UV radiation ( Koski and Ashman , 2015; Koski et al . , 2020 ) , and adaptation to different temperatures ( Koski and Ashman , 2016; Koski et al . , 2020 ) . Sunflowers are one of the most recognizable members of the Asteraceae family , which comprises circa 10% of all flowering plants ( Mandel et al . , 2019 ) . Besides cultivated sunflower , about 50 species of wild sunflowers are found across North America . Wild sunflowers are adapted to a variety of different habitats and display a remarkable amount of phenotypic and genetic diversity , which makes them a model system for studies of adaptation , speciation , and domestication ( Bock et al . , 2020; Heiser et al . , 1969; Todesco et al . , 2020 ) . In addition to being a major crop , sunflowers are also ubiquitous in popular culture , largely due to their iconic yellow inflorescences . Indeed , like many Asteraceae species , wild sunflowers have ligules ( the enlarged modified petals of the outermost whorl of florets in the sunflower inflorescence ) that appear of the same bright yellow colour to the human eye . However , ligules also accumulate UV-absorbing pigments at their base , while their tip reflects UV radiation ( Harborne and Smith , 1978; Wojtaszek and Maier , 2014 ) . Across the whole inflorescence , this results in a bullseye pattern , with an external UV-reflecting ring and an internal UV-absorbing ring . Considerable variation in the size of UV bullseye patterns has been observed between and within plant species ( Koski and Ashman , 2013; Koski and Ashman , 2016 ) ; however , few studies have investigated the ecological factors that drive this variation or the genetic determinants that control it ( Brock et al . , 2016; Koski and Ashman , 2015; Moyers et al . , 2017; Sheehan et al . , 2016 ) . Here , we explore the diversity of floral UV pigmentation in wild sunflowers and the genetic mechanisms and environmental factors that shape this variation .
A preliminary screening of 19 species of wild sunflowers , as well as cultivated sunflower , suggested that UV bullseye patterns are common across sunflower species ( Figure 1—figure supplement 1 ) . In several cases , we also observed substantial within-species variation for the size of UV floral patterns . Patterns of floral UV pigmentation have been previously investigated in the silverleaf sunflower Helianthus argophyllus , which is endemic to Southern Texas ( Figure 1—figure supplement 1 ) . Limited diversity was found between individuals , but transgressive segregation was observed in mapping populations; while several QTL affecting this trait were detected , genetic mapping resolution was insufficient to identify individual causal genes ( Moyers et al . , 2017 ) . To better understand the function and genetic regulation of variation for floral UV pigmentation , we focused on two widespread species of annual sunflowers , Helianthus annuus and Helianthus petiolaris . H . annuus , the common sunflower , grows across most of North America; it is probably the most diverse of the sunflower species and is the progenitor of domesticated sunflower ( H . annuus var . macrocarpus ) . H . petiolaris also has a broad distribution across North America , but prefers sandier soils . It includes two subspecies: subsp . petiolaris , which is common in the central plains of the United States , and subsp . fallax , which is found in the Southwestern USA and has repeatedly adapted to growing on sand dunes ( Heiser et al . , 1969; Todesco et al . , 2020 ) . Over two growing seasons , we measured floral UV patterns ( as the proportion of the ligule that absorbs UV radiation , henceforth ‘ligule ultraviolet proportion’ [LUVp] ) in 1589 H . annuus individuals derived from 110 distinct natural populations and 351 H . petiolaris individuals from 40 populations , grown in common garden experiments in Vancouver , Canada ( Todesco et al . , 2020 ) . The populations of origin of these plants were selected to represent the whole range of H . annuus , and most of the range of H . petiolaris ( Figure 1a and b , Figure 1—source data 1 ) . While extensive variation was observed within both species , it was particularly striking for H . annuus , which displayed a phenotypic continuum from ligules with almost no UV pigmentation to ligules that were entirely UV-absorbing ( Figure 1c–e , Figure 1—figure supplement 2 , Figure 1—source data 2 ) . Floral UV patterns have been proposed to act as nectar guides , helping pollinators orient towards nectar rewards once they land on the petal ( Daumer , 1958 ) , although recent experiments have challenged this hypothesis ( Koski et al . , 2014 ) . A relatively high proportion of H . annuus individuals in our survey ( ~13% ) had completely UV-absorbing ligules and therefore lacked UV nectar guides , suggesting that pollinator orientation is not a necessary function of floral UV pigmentation in sunflower . To identify the loci controlling variation for floral UV patterning , we performed a genome-wide association study ( GWAS ) . We used a subset of the phenotyped plants ( 563 of the H . annuus and all 351 H . petiolaris individuals ) for which we previously generated genotypic data at >4 . 6M high-quality single-nucleotide polymorphisms ( SNPs ) ( Todesco et al . , 2020 ) . Given their relatively high level of genetic differentiation , analyses were performed separately for the petiolaris and fallax subspecies of H . petiolaris ( Todesco et al . , 2020 ) . While no significant association was identified for H . petiolaris fallax ( Figure 2—figure supplement 1 ) , we detected several genomic regions significantly associated with floral UV patterning in H . petiolaris petiolaris , and a particularly strong association ( p=5 . 81e–25 ) on chromosome 15 in H . annuus ( Figure 2a and b ) . The chromosome 15 SNP with the strongest association with ligule UV pigmentation patterns in H . annuus ( henceforth ‘Chr15_LUVp SNP’ ) explained 62% of the observed phenotypic and additive variation ( narrow-sense heritability for LUVp in the H . annuus dataset is ~1 ) . Additionally , allelic distributions at this SNP closely matched that of floral UV patterns ( Figure 2c , compare to Figure 1a; Figure 1—source data 2 ) . Genotype at the Chr15_LUVp SNP had a remarkably strong effect on the size of UV bullseyes in inflorescences . Individuals homozygous for the ‘large’ ( L ) allele had a mean LUVp of 0 . 78 ( SD ±0 . 16 ) , meaning that ~3/4 of the ligule was UV-absorbing , while individuals homozygous for the ‘small’ ( S ) allele had a mean LUVp of 0 . 33 ( SD ±0 . 15 ) , meaning that only the basal ~1/3 of the ligule absorbed UV radiation . Consistent with the trimodal LUVp distribution observed for H . annuus ( Figure 1d ) , alleles at this locus showed additive effects , with heterozygous individuals having intermediate phenotypes ( LUVp = 0 . 59 ± 0 . 18; Figure 2d ) . The association between floral UV patterns and the Chr15_LUVp SNP was confirmed in the F2 progeny of crosses between plants homozygous for the L allele ( with completely UV-absorbing ligules; LUVp = 1 ) and for the S allele ( with a small UV-absorbing patch at the ligule base; LUVp < 0 . 18; Figure 2e , Figure 2—figure supplement 2 ) . Average LUVp values were lower , and their range narrower , when these populations were grown in a greenhouse rather than in a field . Plants in the greenhouse experienced relatively uniform temperatures and humidity , and were shielded from most UV radiation . These results suggest that although floral UV patterns have a strong genetic basis ( consistent with previous observations; Koski and Ashman , 2013 ) , their expression is also affected by the environment . While no obvious candidate genes were found for the GWAS peaks for floral UV pigmentation in H . petiolaris petiolaris , the H . annuus chromosome 15 peak is ~5 kbp upstream of HaMYB111 , a sunflower homolog of the Arabidopsis thaliana AtMYB111 gene ( Figure 2b ) . Together with AtMYB11 and AtMYB12 , AtMYB111 is part of a small family of transcription factors ( also called PRODUCTION OF FLAVONOL GLYCOSIDES [PFG] ) that controls the expression of genes involved in the production of flavonol glycosides in Arabidopsis ( Stracke et al . , 2007 ) . Flavonol glycosides are a subgroup of flavonoids known to fulfil a variety of functions in plants , including protection against abiotic and biotic stresses ( e . g . , UV radiation , cold , drought , herbivory ) ( Pollastri and Tattini , 2011 ) . Crucially , they absorb strongly in the near UV range ( 300–400 nm ) and are the pigments responsible for floral UV patterns in several plant species ( Rieseberg and Schilling , 1985; Sheehan et al . , 2016; Thompson et al . , 1972 ) . For instance , alleles of a homolog of AtMYB111 are responsible for the evolutionary gain and subsequent loss of flavonol accumulation and UV absorption in flowers of Petunia species , associated with two successive switches in pollinator preferences ( from bees , to hawkmoths , to hummingbirds; Sheehan et al . , 2016 ) . A homolog of AtMYB12 has also been associated with variation in floral UV patterns in Brassica rapa ( Brock et al . , 2016 ) . Analysis of sunflower ligules found two main groups of UV-absorbing compounds: glycoside conjugates of quercetin ( a flavonol ) and di-O-caffeoyl quinic acid ( CQA , a member of a family of antioxidant compounds that includes chlorogenic acid and that accumulates at high levels in many sunflower tissues; Koeppe et al . , 1970 ) . Both quercetin glycosides and CQA were more abundant at the base of sunflower ligules , and in ligules of plants with larger LUVp . However , this pattern was much more dramatic for flavonols , and they represented a much larger fraction of the total UV absorbance in UV-absorbing ( parts of ) ligules , suggesting that flavonols are the main pigments responsible for UV patterning in sunflower ligules ( Figure 3a and b ) . In Arabidopsis , AtMYB12 and AtMYB111 are known to have the strongest effect on flavonol glycoside accumulation ( Stracke et al . , 2007; Stracke et al . , 2010 ) . We noticed , from existing RNAseq data , that AtMYB111 expression levels are particularly high in petals ( Klepikova et al . , 2016; Figure 3c ) and found that Arabidopsis petals , while uniformly white in the visible spectrum , absorb strongly in the UV ( Figure 3d ) . To our knowledge , this is the first report of floral UV pigmentation in Arabidopsis , a highly selfing species that is seldom insect-pollinated ( Hoffmann et al . , 2003 ) . Accumulation of flavonol glycosides is strongly reduced , and UV pigmentation is almost completely absent , in petals of mutants for AtMYB111 ( myb111 ) . UV absorbance is further reduced in petals of double mutants for AtMYB12 and AtMYB111 ( myb12/111 ) . However , petals of the single mutant for AtMYB12 ( myb12 ) , which is expressed at low levels throughout the plant ( Klepikova et al . , 2016 ) , are indistinguishable from wild-type plants ( Figure 3d and e ) . This shows that flavonol glycosides are responsible for floral UV pigmentation also in Arabidopsis , and that AtMYB111 plays a fundamental role in controlling their accumulation in petals . To confirm that sunflower HaMYB111 is functionally equivalent to its Arabidopsis homolog , we introduced it into myb111 plants . Expression of HaMYB111 , either under the control of a constitutive promoter or of the endogenous AtMYB111 promoter , restored petal UV pigmentation and induced accumulation of flavonol glycosides ( Figure 3d and e ) . HaMYB111 coding sequences obtained from wild sunflowers with large or small LUVp were equally effective at complementing the myb111 mutant . Together with the observation that the strongest GWAS association with LUVp fell in the promoter region of HaMYB111 , these results suggest that differences in the effect of the ‘small’ and ‘large’ alleles of this gene on floral UV pigmentation are not due to differences in protein function , but rather to differences in gene expression . Analysis of HaMYB111 expression patterns in cultivated sunflower revealed that , consistent with a role in floral UV pigmentation and similar to its Arabidopsis counterpart , it is expressed specifically in ligules , and it is almost undetectable in other tissues ( Badouin et al . , 2017; Figure 3f ) . Similar to observations in Rudbeckia hirta , another member of the Heliantheae tribe ( Schlangen et al . , 2009 ) , UV pigmentation is established early in ligule development in both H . annuus and H . petiolaris as their visible colour turns from green to yellow before the inflorescence opens ( R4 developmental stage; Schneiter and Miller , 1981; Figure 3g , Figure 3—figure supplement 1 ) . HaMYB111 is highly expressed in the part of the ligule that accumulates UV-absorbing pigments , and especially in developing ligules , consistent with a role in establishing pigmentation patterns ( Figure 3h ) . We also observed a matching expression pattern for HaFLS1 , the sunflower homolog of a gene encoding one of the main enzymes controlling flavonol biosynthesis in Arabidopsis ( FLAVONOL SYNTHASE 1 , AtFLS1 ) , whose expression is regulated directly by AtMYB111 ( Stracke et al . , 2007; Figure 3i ) . Finally , we compared HaMYB111 expression levels in a set of 46 field-grown individuals with contrasting LUVp values , representing 21 different wild populations . HaMYB111 expression levels differed significantly between the two groups ( p=0 . 009; Figure 3j ) . Variation in expression levels within phenotypic classes was quite large; this is likely due at least in part to the strong dependence of HaMYB111 expression on developmental stage ( Figure 3g ) and the difficulty of accurately establishing matching ligule developmental stages across diverse wild sunflowers . These expression analyses further point to cis-regulatory rather than coding sequence differences between HaMYB111 alleles being responsible for LUVp variation . Accordingly , direct sequencing of the HaMYB111 locus from multiple wild H . annuus individuals , using a combination of Sanger sequencing and long PacBio HiFi reads , identified no coding sequence variants associated with differences in floral UV patterns , or with alleles at the Chr15_LUVp SNP ( Figure 3—figure supplement 2 , Supplementary files 1 and 2 ) . However , we observed extensive variation in the promoter region of HaMYB111 , differentiating wild H . annuus alleles from each other and from the reference assembly for cultivated sunflower ( Supplementary files 3 and 4 ) . Relaxing quality filters to include less well-supported SNPs in our LUVp GWAS did not identify additional variants with stronger associations than Chr15_LUVp SNP ( Figure 2—figure supplement 2 ) . However , many of the polymorphisms we identified by direct sequencing were either larger insertions/deletions ( indels ) or fell in regions that were too repetitive to allow accurate mapping of short reads , and would not be included even in this expanded SNP dataset . While several of these variants in the promoter region of HaMYB111 appeared to be associated with the Chr15_LUVp SNP , further studies will be required to confirm this , and to identify their eventual effects on HaMYB111 activity ( see discussion in the legend of Figure 3—figure supplement 2 ) . Interestingly , when we sequenced the promoter region of HaMYB111 in several H . argophyllus and H . petiolaris individuals , we found that they all carried the S allele at the Chr15_LUVp SNP , and that their promoter regions were generally more similar in sequence to those of H . annuus individuals carrying the S allele at the Chr15_LUVp SNP ( Supplementary files 3 and 4 ) . Similarly , in a set of previously re-sequenced wild sunflowers , we found the S allele to be fixed in several perennial ( Helianthus decapetalus , Helianthus divaricatus , and Helianthus grosseserratus ) and annual sunflower species ( H . argophyllus , Helianthus niveus , Helianthus debilis ) , and to be at >0 . 98 frequency in H . petiolaris ( Figure 1—source data 2 ) . Conversely , the L allele at Chr15_LUVp SNP was almost fixed ( >0 . 98 frequency ) in a set of 285 cultivated sunflower lines ( Mandel et al . , 2013 ) . Consistent with these patterns , UV bullseyes are considerably smaller in H . argophyllus ( mean LUVp ± SD = 0 . 27 ± 0 . 09 ) , H . niveus ( 0 . 15 ± 0 . 09 ) , and H . petiolaris ( 0 . 27 ± 0 . 12; Figure 1e ) than in cultivated sunflower lines ( 0 . 62 ± 0 . 23 ) . Additionally , while 50 of the cultivated sunflower lines had completely or almost completely UV-absorbing ligules ( LUVp > 0 . 8 ) , no such case was observed in the other three species ( Figure 1—figure supplement 3 ) . Although our results show that HaMYB111 explains most of the variation in floral UV pigmentation patterns in wild H . annuus , why such variation exists in the first place is less clear . Several hypotheses have been advanced to explain the presence of floral UV patterns and their variability . Like their visible counterparts , UV pigments play a fundamental role in pollinator attraction ( Horth et al . , 2014; Koski et al . , 2014; Rae and Vamosi , 2013; Sheehan et al . , 2016 ) . For example , in Rudbeckia species , artificially increasing the size of bullseye patterns to up to 90% of the petal surface resulted in rates of pollinator visitation equal to or higher than wild-type flowers ( which have on average 40–60% of the petal being UV-absorbing ) . Conversely , reducing the size of the UV bullseye had a strong negative effect on pollinator visitation ( Horth et al . , 2014 ) . To test whether the relative size of UV bullseye patterns affected pollination , we assessed insect visitation rates for wild H . annuus lines with contrasting UV bullseye patterns . An initial experiment compared inflorescences from pairs of plants from two populations ( ANN_03 from California and ANN_55 from Texas ) , which were selected to have large or small floral UV patterns . In this setup , inflorescences with large UV patterns received significantly more visits ( Figure 4a ) . While this experiment revealed a clear pattern of pollinator preferences , it involved plants from only two different populations , and effects of other unmeasured factors unrelated to UV pigmentation on visitation patterns cannot be excluded . Therefore , we monitored pollinator visitation in plants grown in a common garden experiment including 1484 individuals from 106 H . annuus populations , spanning the entire range of the species . Assaying a much more diverse population of H . annuus individuals should reduce effects on pollinator preferences of traits unrelated to floral UV pigmentation . Within this field , we selected 82 plants , from 49 populations , which flowered at roughly the same time and had comparable numbers of flowers . We selected plants falling into three categories of LUVp values , representatives of the more abundant phenotypic classes across the range of wild H . annuus ( Figure 1d ) : small ( LUVp = 0–0 . 3 ) , intermediate ( LUVp = 0 . 5–0 . 8 ) , and large ( LUVp >0 . 95 ) . Plants with intermediate UV patterns had the highest visitation rates ( Figure 4b , Figure 4—figure supplement 1 ) . Visitation to plants with small or large UV patterns was less frequent , and particularly low for plants with very small LUVp values ( <0 . 15 ) . Pollination rates are known to be yield-limiting in sunflower ( Greenleaf and Kremen , 2006 ) , and a strong reduction in pollination could therefore have a negative effect on fitness; this would be consistent with the observation that plants with very small LUVp values were rare ( ~1 . 5% of individuals ) in our common garden experiment , which was designed to provide a balanced representation of the natural range of H . annuus . Although pollinator preferences in this experiment could still be affected by other unmeasured factors ( nectar content , floral volatiles ) , these results are consistent with previous results showing that floral UV patterns play a major role in pollinator attraction ( Horth et al . , 2014; Koski et al . , 2014; Rae and Vamosi , 2013; Sheehan et al . , 2016 ) . They also agree with earlier findings in other plant species , suggesting that intermediate-to-large UV bullseyes are preferred by pollinators ( Horth et al . , 2014; Koski et al . , 2014 ) . While we cannot exclude that smaller UV bullseyes would be preferred by pollinators in some parts of the H . annuus range , this does not seem likely; the most common pollinators of sunflower are ubiquitous across the range of H . annuus , and many bee species known to pollinate sunflower are found in both regions where H . annuus populations have large LUVp and regions where they have small LUVp ( Hurd et al . , 1980 ) . Therefore , while acting as visual cues for pollinators is clearly a major function of floral UV bullseyes , it is unlikely to ( fully ) explain the patterns of variation that we observe for this trait . In recent years , the importance of non-pollinator factors in driving selection for floral traits has been increasingly recognized ( Strauss and Whittall , 2006 ) . Additionally , flavonol glycosides , the pigments responsible for floral UV patterns in sunflower , are known to be involved in responses to several abiotic stressors ( Korn et al . , 2008; Nakabayashi et al . , 2014b; Pollastri and Tattini , 2011; Schulz et al . , 2015 ) . Therefore , we explored whether some of these stressors could drive diversification in floral UV pigmentation . An intuitively strong candidate is UV radiation , which can be harmful to plant cells ( Stapleton , 1992 ) . Variation in the size of UV bullseye patterns across the range of Argentina anserina ( a member of the Rosaceae family ) has been shown to correlate positively with intensity of UV radiation . Flowers of this species are bowl-shaped , and larger UV-absorbing regions have been proposed to protect pollen from UV damage by absorbing UV radiation that would otherwise be reflected toward the anthers ( Koski and Ashman , 2015 ) . However , sunflower inflorescences are much flatter than A . anserina flowers , making it unlikely that any significant amount of UV radiation would be reflected from the ligules towards the disc flowers . Studies in another plant with non-bowl-shaped flowers ( Clarkia unguiculata ) have found no evidence of an effect of floral UV patterns in protecting pollen from UV damage ( Peach et al . , 2020 ) . Consistent with this , the associations between the intensity of UV radiation at our collection sites and floral UV patterns in H . annuus was weak ( H . annuus: R2 = 0 . 01 , p=0 . 12; Figure 4c , Figure 4—figure supplement 2 ) . Across the Potentillae tribe ( Rosaceae ) , floral UV bullseye size is also weakly associated with UV radiation , but is more strongly correlated with temperature , with lower temperatures being associated with larger UV bullseyes ( Koski and Ashman , 2016 ) . We found a similar , strong correlation with temperature in our dataset , with lower average summer temperatures being associated with larger LUVp values in H . annuus ( R2 = 0 . 44 , p=2 . 4 × 10–15; Figure 4d , Figure 4—figure supplement 2 ) . It has been suggested that the radiation absorbed by floral UV pigments could contribute to increasing the temperature of the flower , similar to what has been observed for visible pigments ( Koski et al . , 2020 ) . This possibility is particularly intriguing for sunflower , in which flower temperature plays an important role in pollinator attraction; inflorescences of cultivated sunflowers consistently face east so that they warm up faster in the morning , making them more attractive to pollinators ( Atamian et al . , 2016; Creux et al . , 2021 ) . By absorbing more radiation , larger UV bullseyes could therefore contribute to increasing temperature of the sunflower inflorescences , and their attractiveness to pollinators , in cold climates . However , UV wavelengths represents only a small fraction ( 3–7% ) of the solar radiation reaching the Earth’s surface ( compared to >50% for visible wavelengths ) , and might therefore not provide sufficient energy to significantly warm up the ligules ( Nunez et al . , 1994 ) . In line with this observation , different levels of UV pigmentation had no effect on the temperature of inflorescences or individual ligules exposed to sunlight ( Figure 4e–g , Figure 4—figure supplement 3 ) . While several geoclimatic variables are correlated across the range of wild H . annuus , the single variable explaining the largest proportion of the variation in floral UV patterns in this species was summer relative humidity ( RH; R2 = 0 . 51 , p=1 . 4 × 10–18; Figure 4h , Figure 4—figure supplement 2 ) , with lower humidity being associated with larger LUVp values ( i . e . , higher concentrations of flavonol glycosides in ligules ) . Lower RH is generally associated with higher transpiration rates in plants , leading to increased water loss , and flavonol glycosides are known to play an important role in responses to drought stress ( Nakabayashi et al . , 2014a ) ; in particular , Arabidopsis lines that accumulate higher concentrations of flavonol glycosides due to overexpression of AtMYB12 lose water and desiccate at slower rates than wild-type plants ( Nakabayashi et al . , 2014b ) . Similarly , in a set of plants representing seven independent natural populations of H . annuus , we found that completely UV-absorbing ligules desiccate at a significantly slower rate than largely UV-reflecting ligules ( Figure 4i ) . This is not due to general differences in transpiration rates between genotypes since we observed no comparable trend for rates of leaf desiccation in the same set of sunflower lines ( Figure 4j ) . Transpiration from flowers can be a major source of water loss for plants , and this is known to drive , within species , the evolution of smaller flowers in populations living in dry locations ( Galen , 2000; Herrera , 2005; Lambrecht , 2013; Lambrecht and Dawson , 2007; see Figure 4—figure supplement 4 ) . While desiccation rates are only a proxy for transpiration in field conditions ( Duursma et al . , 2019; Hygen , 1951 ) , and other factors might affect ligule transpiration in this set of lines , this evidence ( strong correlation between LUVp and summer RH; known role of flavonol glycosides in regulating transpiration; and correlation between extent of ligule UV pigmentation and desiccation rates ) suggests that variation in floral UV pigmentation in sunflowers is driven by the role of flavonol glycosides in reducing water loss from ligules , with larger floral UV patterns helping prevent drought stress in drier environments . One of the main roles of transpiration in plants is facilitating heat dispersion at higher temperatures through evaporative cooling ( Burke and Upchurch , 1989; Drake et al . , 2018 ) , which could explain the strong correlation between LUVp and temperature across the range of H . annuus ( Figure 4d ) . Consistent with this , summer RH and summer temperatures together explain a considerably larger fraction of the variation for LUVp in H . annuus than either variable alone ( R2 = 0 . 63 , p=0 . 0017; Figure 1—source data 1 ) , with smaller floral UV patterns being associated with higher RH and higher temperatures ( Figure 4—figure supplement 2 ) . Consistent with a role of floral UV pigmentation in the plant’s response to variation in both humidity and temperature , we found strong associations ( dB > 10 ) between SNPs in the HaMYB111 region and these variables in genotype-environment association ( GEA ) analyses ( Figure 4k , Figure 4—source data 4 ) . Despite a more limited range of variation for LUVp , a similar trend ( larger UV patterns in drier , colder environments ) is present also in H . petiolaris ( Figure 4—figure supplement 5 ) . Interestingly , while the L allele at Chr_15 LUVp SNP is present in H . petiolaris ( Figure 1—figure supplement 3 ) , it is found only at a very low frequency and does not seem to significantly affect floral UV patterns in this species ( Figure 2a ) . This could represent a recent introgression since H . annuus and H . petiolaris are known to hybridize in nature ( Heiser , 1947; Yatabe et al . , 2007 ) . Alternatively , the Chr_15 LUVp SNP might not be associated with functional differences in HaMYB111 in H . petiolaris , or differences in genetic networks or physiology between H . annuus and H . petiolaris could mask the effect of this allele , or limit its adaptive advantage , in the latter species . Connecting adaptive variation to its genetic basis is one of the main goals of evolutionary biology . Here , we show that regulatory variation at a single major gene , the transcription factor HaMYB111 , underlies most of the diversity for floral UV patterns in the common sunflower , wild H . annuus . Variation for these floral UV patterns correlates strongly with pollinator preferences , but also with geoclimatic variables ( especially RH and temperature ) and desiccation rates in sunflower ligules . While the effects of floral UV patterns on pollinator attraction are well-known , these associations suggest a role of environmental factors in shaping diversity for this trait . Larger floral UV patterns , due to accumulation of flavonol glycoside pigments in ligules , could help reduce the amount of transpiration in environments with lower RH , preventing excessive water loss and maintaining ligule turgidity . In humid , hot environments ( e . g . , Southern Texas ) , lower accumulation of flavonol glycosides would instead promote transpiration from ligules , keeping them cool and avoiding overheating . The presence of UV pigmentation in the petals of Arabidopsis ( also controlled by the Arabidopsis homolog of MYB111 ) further points to a more general protective role of these pigments in flowers since pollinator attraction is likely not critical for fertilization in this largely selfing species . It should be noted that , while we have examined some of the most likely factors explaining the distribution of variation for floral UV patterns in wild H . annuus across North America , other abiotic factors could play a role , as well as biotic ones ( e . g . , the aforementioned differences in pollinator assemblages , or a role of UV pigments in protection from herbivory; Gronquist et al . , 2001 ) . However , a role of floral UV patterns in reducing water loss from petals is consistent with the overall trend in increased size of floral UV patterns over the past 80 years that has been observed in herbarium specimens ( Koski et al . , 2020 ) ; due to changing climates , RH over land has been decreasing in recent decades , which could result in higher transpiration rates ( Byrne and O’Gorman , 2018 ) . Further studies will be required to confirm the existence of this trend and assess its strength . More generally , our study highlights the complex nature of adaptive variation , with selection pressures from both biotic and abiotic factors shaping the patterns of diversity that we observe across natural populations . Floral diversity in particular has long been attributed to the actions of animal pollinators . Our work adds to a growing literature demonstrating the contributions of abiotic factors to this diversity .
Sunflower lines used in this paper were grown from seeds collected from wild populations ( Todesco et al . , 2020 ) or obtained from the North Central Regional Plant Introduction Station in Ames , IA . For all experiments except the plants used for Figure 1—figure supplement 1b , sunflower seeds were surface sterilized by immersion for 10 min in a 1 . 5% sodium hypochlorite solution . Seeds were then rinsed twice in distilled water and treated for at least 1 hr in a solution of 1% PPM ( Plant Preservative Mixture; Plant Cell Technologies , Washington , DC ) , a broad-spectrum biocide/fungicide , to minimize contamination , and 0 . 05 mM gibberellic acid ( Sigma-Aldrich , St . Louis , MO ) . They were then scarified , dehulled , and kept for 2 weeks at 4°C in the dark on filter paper moistened with a 1% PPM solution . Following this , seeds were kept in the dark at room temperature until they germinated . For common garden experiments , the seedlings were then transplanted in peat pots , grown in a greenhouse for 2 weeks , then moved to an open-sided greenhouse for a week for acclimation , and finally transplanted in the field at the Totem Plant Science Field Station of the University of British Columbia ( Vancouver , Canada ) . For all other experiments , seedlings were transplanted in 2-gallon pots filled with Sunshine #1 growing mix ( Sun Gro Horticulture Canada , Abbotsford , BC , Canada ) . Plants grown in greenhouses at the Vancouver campus of the University of British Columbia were kept at 26°C during the day and 20°C during the night , supplemented with LED light on a cycle of 16 hr days and 8 hr nights . For the wild sunflower species shown in Figure 1—figure supplement 1b , following sterilization , seeds were scarified and then dipped in fusicoccin solution ( 1 . 45 µM ) for 15 min , dehulled , germinated in the dark for at least 8–10 days , and then grown in pots for 3 weeks before transplantation . One group of species was transplanted into 2-gallon pots filled with a blend of sandy loam , organic compost and mulch , and grown at the UC Davis Plant Sciences Field Station ( Davis , CA ) from July to October 2017 . Several additional species were grown in single rows covered with mulch and spaced 0 . 75 m apart at the Oxford Tract Facility field ( Berkeley , CA ) from June to October 2021 , or in a greenhouse facility at Berkeley , CA . A complete list of sunflower accessions and their populations of origin is reported in Figure 1—source data 1 and Figure 1—source data 2 . Seeds from the following Arabidopsis lines were obtained from the Arabidopsis Biological Resource Center: Col-0 ( CS28167 ) , myb111 ( CS9813 ) , myb12 ( CS9602 ) , and myb12/myb111 ( CS9980 ) . Seeds were stratified in 0 . 1% agar at 4°C in the dark for 4 days , and then sown in pots containing Sunshine #1 growing mix . Plants were grown in growth chambers at 23°C in long-day conditions ( 16 hr light , 8 hr dark ) . Two common garden experiments were performed , in 2016 and 2019 . After germination and acclimation , plants were transplanted at the Totem Plant Science Field Station of the University of British Columbia ( Vancouver , Canada ) . In the 2016 common garden experiment , each sunflower species was grown in a separate field . Pairs of plants from the same population were randomly distributed within each field . In the 2019 common garden experiment , plants were sown using a completely randomized design . In the summer of 2016 , 10 plants from each of the 151 selected populations of wild H . annuus , H . petiolaris , H . argophyllus , and H . niveus were grown . Plants were transplanted in the field on 25 May ( H . argophyllus ) , 2 June ( H . petiolaris and H . niveus ) , and 7 June 2016 ( H . annuus ) . Up to four inflorescences from each plant were collected for visible and UV photography . In the summer of 2019 , 14 plants from each of the 106 populations of wild H . annuus were transplanted in the field on 6 June . These included 65 of the populations grown in the previous common garden experiment , and 41 additional populations that were selected to complement their geographical distribution . At least three ligules from at least two different inflorescences for each plant were collected for visible and UV photography . Ligules were selected to be as far apart from each other as possible across the inflorescence , taking care to avoid damaged or otherwise unrepresentative ligules . Sample size for the common garden experiments was determined by the available growing space and resources . 10–14 individuals were grown for each population because this would provide a good representation of the variation present in each population , while maximizing the number of populations that could be surveyed . Researchers were not blinded as to the identity of individual samples . However , information about their populations of origin and/or LUVp phenotypes was not attached to the samples during data acquisition . Ultraviolet patterns were imaged in whole inflorescences or detached ligules ( see ‘Common garden’ section ) using a Nikon D70s digital camera , fitted with a Noflexar 35 mm lens and a reverse-mounted 2-inch Baader U-Filter ( Baader Planetarium , Mammendorf , Germany ) , which only allows the transmission of light between 320 and 380 nm . Wild sunflower species shown in Figure 1—figure supplement 1b were imaged using a Canon DSLR camera in which the internal hot mirror filter had been replaced with a UV bandpass filter ( LifePixel , Mukilteo , WA ) . Floral UV patterns were scored as LUVp , rather than total area or diameter of the UV bullseye , because LUVp is less influenced by genetic or environmental factors affecting inflorescence size ( Moyers et al . , 2017 ) . The length of the whole ligule ( LL ) and the length of the UV-absorbing part at the base of the ligule ( LUV-abs ) were measured using ImageJ ( Schindelin et al . , 2012; Schneider et al . , 2012 ) . LUVp was measured as the ratio between the two ( LUVp = LUV-abs/LL ) . In some H . annuus individuals , the upper , ‘UV-reflecting’ portion of the ligules ( LUV-ref ) also displayed a lower level of UV absorption; in those cases , these regions were weighted at 50% of fully UV-absorbing regions using the formula LUVp = ( LUV-abs/LL ) + ½ ( LUV-ref/LL ) . Partial UV absorbance in the tip of ligules was more common in plants with larger floral UV patterns ( Figure 1—figure supplement 2 ) . To avoid possible confounding effects , for all experiments plants in the ‘small’ and ‘intermediate’ LUVp classes were selected to have no noticeable UV absorbance in the tips of ligules . For UV pictures of whole inflorescences , LUVp values were measured for three representative ligules chosen to be as far apart from each other as possible , and the average of those three values was used as the LUVp for the inflorescence . LUVp values for all the inflorescences or detached ligules available for each plant were averaged to obtain the LUVp value for that individual . Infrared pictures for the experiments shown in Figure 4e–g and Figure 4—figure supplement 3 were taken using a Fluke TiX560 thermal imager ( Fluke Corporation , Everett , WA ) and analysed using the Fluke Connect software ( v1 . 1 . 536 . 0 ) . For time-series experiments on whole inflorescences , plants from populations ANN_03 ( from CA , USA , with large LUVp ) and ANN_55 ( from TX , USA , with small LUVp ) were germinated as above ( see ‘Common garden’ ) , grown in 2-gallon pots in a greenhouse until they produced four true leaves , and then moved to the field . On three separate days in August 2017 , pairs of inflorescences with opposite floral UV patterns at similar developmental stages were selected and made to face east . Infrared images were taken just before sunrise , ~5 min after sunrise , and then at 0 . 5 , 1 , 2 , 3 , and 4 hr after sunrise . For infrared pictures of detached ligules , plants were grown in a greenhouse . Plants with large LUVp came from populations ANN_03 ( CA , USA ) , ANN_16 ( NM , USA ) , and ANN_19 ( NM , USA ) ; plants with small LUVp came from populations ANN_55 and ANN_58 ( both from TX , USA ) . Flowerheads were collected and kept overnight in a room with constant temperature of 21°C , with their stems immersed in a beaker containing distilled water . The following day , pairs of inflorescences were randomly selected from the two LUVp categories , and representative , undamaged ligules were removed and arranged on a sheet of white paper . Infrared pictures were taken immediately before exposing the ligules to the sun , and again 5 , 10 , and 15 min after that , at around 1 pm on 5 October 2020 ( Figure 4—source data 2 ) . Whole-genome shotgun ( WGS ) sequencing library preparation and sequencing , as well as SNP calling and variant filtering , for the H . annuus and H . petiolaris individuals used for GWAS analyses in this paper were previously described ( Todesco et al . , 2020 ) . Briefly , DNA was extracted from leaf tissue using a modified CTAB protocol ( Murray and Thompson , 1980; Zeng et al . , 2002 ) , the DNeasy Plant Mini Kit , or a DNeasy 96 Plant Kit ( QIAGEN , Hilden , Germany ) . Genomic DNA was sheared to an average fragment size of 400 bp using a Covaris M220 ultrasonicator ( Covaris , Woburn , MA ) . Libraries were prepared using a protocol largely based on Rowan et al . , 2015 , the TruSeq DNA Sample Preparation Guide from Illumina ( Illumina , San Diego , CA ) , and Rohland and Reich , 2012 , with the addition of an enzymatic repeats depletion step using a Duplex-Specific Nuclease ( DSN; Evrogen , Moscow , Russia ) ( Matvienko et al . , 2013; Shagina et al . , 2010; Todesco et al . , 2020 ) . All libraries were sequenced at the McGill University and Génome Québec Innovation Center on HiSeq2500 , HiSeq4000 , and HiSeqX instruments ( Illumina ) to produce paired end , 150 bp reads . Sequences were trimmed for low quality using Trimmomatic ( v0 . 36 ) ( Bolger et al . , 2014 ) and aligned to the H . annuus XRQv1 genome ( Badouin et al . , 2017 ) using NextGenMap ( v0 . 5 . 3 ) ( Sedlazeck et al . , 2013 ) . We followed the best practice recommendations of the Genome Analysis ToolKit ( GATK ) ( Poplin et al . , 2017 ) and executed steps documented in GATK’s germline short variant discovery pipeline ( for GATK 4 . 0 . 1 . 2 ) . During genotyping , to reduce computational time and improve variant quality , genomic regions containing transposable elements were excluded ( Badouin et al . , 2017 ) . We then used GATK’s VariantRecalibrator ( v4 . 0 . 1 . 2 ) to select high-quality variants . SNP data were then filtered for minor allele frequency ( MAF ) ≥ 0 . 01 , genotype rate ≥ 90% , and to keep only biallelic SNPs . Filtered SNPs were then remapped to the improved reference assembly HA412-HOv2 ( Staton and Lázaro-Guevara , 2020 ) using BWA ( v0 . 7 . 17 ) ( Li , 2013 ) . These remapped SNPs were used for all analyses , excluding the GWAS for the region surrounding the HaMYB111 locus that used unfiltered variants based on the XRQv1 assembly ( Figure 2—figure supplement 3 ) . The SNP dataset used to determine the genotype at the Chr15_LUVp SNP in other species ( H . argophyllus , H . niveus , H . debilis , H . decapetalus , H . divaricatus , and H . grosseserratus ) was based on WGS data generated for Todesco et al . , 2020 and is described in Owens et al . , 2021 . Sequence data for the Sunflower Association Mapping population are reported in Hübner et al . , 2019 . Genome-wide association analyses for LUVp were performed for H . annuus , H . petiolaris petiolaris , and H . petiolaris fallax using two-sided mixed models implemented in EMMAX ( v07Mar2010 ) ( Kang et al . , 2010 ) or in the EMMAX module in EasyGWAS ( Grimm et al . , 2017 ) . For all runs , the first three principal components ( PCs ) were included as covariates , as well as a kinship matrix . Only SNPs with MAF > 5% were included in the analyses , and variants were imputed and phased using Beagle ( version 10Jun18 . 811 ) ( Browning et al . , 2018 #497 ) . A GWAS with MAF > 1% in H . petiolaris petiolaris failed to find any additional association between LUVp and variation at the Chr15_LUVp SNP ( the L allele is found at a frequency of ~2% in H . petiolaris petiolaris ) . Sample size was estimated to be sufficient to provide an 85% probability of detecting loci explaining 5% or more of the phenotypic variance in H . annuus . An 85% probability of detecting loci explaining 8% of variance in H . petiolaris was estimated for the whole species set ( 488 individuals ) ; upon analysis of resequencing data for this species , three distinct clusters of individuals were detected ( H . petiolaris petiolaris , H . petiolaris fallax , H . niveus canescens ) , and GWAS were performed independently on H . petiolaris petiolaris and H . petiolaris fallax ( the H . niveus canescens cluster included only 86 individuals ) . Subspecies dataset were found to provide sufficient power to detect strong associations with adaptive traits ( Todesco et al . , 2020 ) . Narrow-sense heritability ( h2 ) in the H . annuus dataset was estimated using EMMAX ( Kang et al . , 2010 ) , GEMMA ( Zhou and Stephens , 2012 ) , GCTA-GREML ( Yang et al . , 2011 ) , and BOLT_REML ( Loh et al . , 2015 ) . All software produced h2 values of ~1: while it is possible that the presence of a single locus of very large effect would lead to inflation of these estimates , all individuals in the GWAS populations were grown at the same time under uniform conditions , and limited environmental effects are therefore expected . Individuals from population ANN_03 from CA , USA ( large LUVp ) , and ANN_55 from TX , USA ( small LUVp ) , were grown in 2-gallon pots in a field . When the plants reached maturity , they were moved to a greenhouse , where several inflorescences were bagged and crossed . The resulting F1 seeds were germinated and grown in a greenhouse , and pairs of siblings were crossed ( wild sunflowers are self-incompatible ) . The resulting F2 populations were grown both in a greenhouse in the winter of 2019 ( n = 42 individuals for population 1 , 38 individuals for population 2 ) and in a field as part of the 2019 common garden experiments ( n = 54 individuals for population 1 , 50 individuals for population 2 ) . DNA was extracted from young leaf tissue as described above . All F2 plants were genotyped for the Chr15_LUVp SNPs using a custom TaqMan SNP genotyping assay ( Thermo Fisher Scientific , Waltham , MA ) on a Viia 7 Real-Time PCR system ( Thermo Fisher Scientific ) . Methanolic extractions were performed following Stracke et al . , 2007 . Sunflower ligules ( or portions of them ) and Arabidopsis petals were collected and flash-frozen in liquid nitrogen . For sunflower , all ligules , or part of ligules , were collected from the selected inflorescence ( avoiding damaged ligules ) . At least two ligules ( or parts of ligules ) were then randomly chosen , pooled , and weighed for methanolic extraction from each inflorescence . For Arabidopsis , hundreds of petals from several plants for each genotype were collected , pooled , and weighed to obtain a sufficient amount of tissue . The frozen tissue was ground to a fine powder by adding 10–15 zirconia beads ( 1 mm diameter ) and using a TissueLyser ( QIAGEN ) for sunflower ligules , or using a plastic pestle in a 1 . 5 ml tube for Arabidopsis petals . 0 . 5 ml of 80% methanol were added , and the samples were further homogenized and incubated at 70°C for 15 min . They were then centrifuged at 15 , 000 × g for 10 min , and the supernatant was dried in a SpeedVac ( Thermo Fisher Scientifics ) at 60°C . Samples were then resuspended in 1 µl ( sunflower ) or 2 . 5 µl ( Arabidopsis ) of 80% methanol for every milligram of starting tissue . The extracts were analysed by LC/MS/MS using an Agilent 1290 UHPLC system ( Agilent Technologies , Santa Clara , CA ) coupled with an Agilent 6530 Quadrupole Time of Flight mass spectrometer . The chromatographic separation was performed on Atlantis T3- C18 reversed-phase ( 50 mm × 2 . 1 mm , 3 µm ) analytical columns ( Waters Corp , Milford , MA ) . The column temperature was set at 40°C . The elution gradient consisted of mobile phase A ( water and 0 . 2% formic acid ) and mobile phase B ( acetonitrile and 0 . 2% formic acid ) . The gradient program was started with 3% B , increased to 25% B in 10 min , then increased to 40% B in 13 min , increased to 90% B in 17 min , held for 1 min , and equilibrated back to 3% B in 20 min . The flow rate was set at 0 . 4 ml/min and injection volume was 1 µl . A photo diode array ( PDA ) detector was used for detection of UV absorption in the range of 190–600 nm . MS and MS/MS detection were performed using an Agilent 6530 accurate mass Quadrupole Time of Flight mass spectrometer equipped with an ESI ( electrospray ) source operating in both positive and negative ionization modes . Accurate positive ESI LC/MS and LC/MS/MS data were processed using the Agilent MassHunter software to identify the analytes . The ESI conditions were as follows: nebulizing gas ( nitrogen ) pressure and temperature were 30 psi and 325°C; sheath gas ( nitrogen ) flow and temperature were 12 l/min , 325°C; dry gas ( nitrogen ) was 7 l/min . Full scan mass range was 50–1700 m/z . Stepwise fragmentation analysis ( MS/MS ) was carried out with different collision energies depending on the compound class . Total RNA was isolated from mature and developing ligules , or part of ligules , using TRIzol ( Thermo Fisher Scientific ) , and cDNA was synthesized using the RevertAid First Strand cDNA Synthesis kit ( Thermo Fisher Scientific ) . All ligules , or part of ligules , were collected from the selected inflorescence in a single tube ( avoiding damaged ligules ) and flash-frozen in liquid nitrogen . At least two full ligules ( or parts of ligules ) were then randomly chosen and pooled for RNA extraction from each inflorescence . Genomic DNA was extracted from leaves of Arabidopsis using CTAB ( Murray and Thompson , 1980 ) . A 1959-bp-long fragment ( pAtMYB111 ) from the promoter region of AtMYB111 ( At5g49330 ) , including the 5′-UTR of the gene , was amplified using Phusion High-Fidelity DNA polymerase ( New England Biolabs , Ipswich , MA ) and introduced in pFK206 derived from pGREEN ( Hellens et al . , 2000 ) . Alleles of HaMYB111 ( HanXRQChr15g0465131 ) were amplified from cDNA from ligules of individuals from populations ANN_03 ( large LUVp , from CA ) and ANN_55 ( small LUVp , from TX ) . These are the same populations from which the parental plants of the F2 populations shown in Figure 2e were derived . A comparison of the patterns of polymorphisms between these two alleles ( HaMYB111_large and HaMYB111_small ) , other HaMYB111 CDS alleles from wild H . annuus , and the cultivated reference XRQ sequence is shown in Figure 3—figure supplement 2 . These alleles were placed under the control of pAtMYB111 ( in the plasmid described above ) or of the constitutive CaMV 35S promoter ( in pFK210 , derived as well from pGREEN; Hellens et al . , 2000 ) . Constructs were introduced into Arabidopsis plants by Agrobacterium tumefaciens -mediated transformation ( strain GV3101 ) ( Weigel and Glazebrook , 2002 ) . At least five independent transgenic lines with levels of UV pigmentation comparable to the ones shown in Figure 3d were recovered for each construct . For expression analyses , qRT-PCRs were performed on cDNA from ligules using the SsoFast EvaGreen Supermix ( Bio-Rad , Hercules , CA ) on a CFX96 Real-Time PCR Detection System ( Bio-Rad ) . Expression levels were normalized against HaEF1α . HaEF1α ( HanXRQChr11g0334971 ) was selected as a reference gene because , out of a set of genes that showed constitutively elevated expression across different tissues and treatments in cultivated sunflower ( Badouin et al . , 2017 ) , it displayed the most robust expression patterns across ligules of different H . annuus and H . petiolaris individuals , and across ligule tips and bases in the two species . For the expression analyses shown in Figure 3h and i , portions of ligules were collected at different developmental stages from three separate inflorescences from one individual for each species ( biological replicates ) . Three qRT-PCRs were run for each sample ( technical replicates ) . For the expression analysis shown in Figure 3j , samples were collected from wild H . annuus individuals grown as part of the 2019 common garden experiment . Ligules were collected on the same day from developing inflorescences of 24 individuals with large LUVp ( from 10 populations ) and 22 individuals with small LUVp ( from 11 populations ) . qPCRs for three technical replicates were performed for each individual . These plants were genotyped for the Chr15_LUVp SNP using a custom TaqMan assay ( see ‘F2 populations and genotyping’ ) on a CFX96 Real-Time PCR Detection System ( Bio-Rad ) . Sample size for this experiment was determined by the number of available plants with opposite LUVp phenotypes and at the appropriate developmental stage on the day in which samples were collected . Primers used for cloning and qRT-PCR are given in the Key resources table . Fragments ranging in size from 1 . 5 to 5 . 5 kbp were amplified using Phusion High-Fidelity DNA polymerase ( New England Biolabs ) from genomic DNA of 20 individuals that had been previously resequenced ( Todesco et al . , 2020 ) and whose genotype at the Chr15_LUVp SNP was therefore known . Fragments were then cloned in either pBluescript or pJET ( Thermo Fisher Scientific ) and sequenced on a 3730S DNA analyzer using BigDye Terminator v3 . 1 sequencing chemistry ( Applied Biosystems , Foster City , CA ) . For long read sequencing , seed from wild H . annuus populations known to be homozygous for different alleles at the Chr15_LUVp SNP were germinated and grown in a greenhouse . After confirming that they had the expected LUVp phenotype , branches from each plant were covered with dark cloth for several days , and young , etiolated leaves were collected and immediately frozen in liquid nitrogen . High molecular weight ( HMW ) DNA was extracted from six plants using a modified CTAB protocol ( Stoffel et al . , 2012 ) . All individuals were genotyped for the Chr15_LUVp SNP using a custom TaqMan SNP genotyping assay ( Thermo Fisher Scientific , see above ) on a CFX96 Real-Time PCR Detection System ( Bio-Rad ) . Two individuals , one with large and one with small LUVp , were selected . HiFi library preparation and sequencing on a Sequel II instrument ( PacBio , Menlo Park , CA ) were performed at the McGill University and Génome Québec Innovation Center . Each individual was sequenced on an individual SMRT cell 8M , resulting in average genome-wide sequencing coverage of 6–8× . In September 2017 , pollinator visits were recorded in individual inflorescences of pairs of plants with large ( from population ANN_03 , from CA ) and small LUVp ( from population ANN_55 , from TX ) grown in pots in a field adjacent the Nursery South Campus greenhouses of the University of British Columbia . Populations ANN_55 and ANN_03 were chosen because they flowered at about the same time in our 2016 common garden experiment and had inflorescences of similar size and appearance . Pairs of size-matched inflorescences , made to face towards the same direction , were filmed using a Bushnell Trophy Cam HD ( Bushnell , Overland Park , KS ) in 12 min intervals . Visitation rates were averaged over 14 such movies ( Figure 4—source data 1 ) . The only other sunflowers present in the field were H . anomalus individuals , grown in a separate field about 15 m away . H . anomalus has uniformly small floral UV patterns ( Figure 1—figure supplement 1 ) , and is therefore unlikely to have affected pollinator preferences . In summer 2019 , pollinator visits were scored in a common garden experiment consisting of 1484 H . annuus plants at the Totem Plant Science Field Station of the University of British Columbia ( see the ‘Common gardens’ section for details on field design ) . Over 5 days , between 29 July and 7 August , pollinator visits on individual plants were directly observed and counted over 5 min intervals for a total of 435 series of measurements on 111 plants from 51 different populations ( Figure 4—source data 1 ) . Observers were careful to be at least 2 m away from the plant , and not to overshadow it . Visits to all inflorescences for each plant were recorded; pollinators visiting more than one inflorescence per plant were recorded only once . To generate a more homogenous and comparable dataset , measurements for plants with too few ( 1 ) or too many ( >10 ) inflorescences were excluded from the final analysis ( Figure 4—source data 1 ) . Twenty topo-climatic factors were extracted from climate data collected over a 30-year period ( 1961–1990 ) for the geographical coordinates of the population collection sites using the software package Climate NA ( Wang et al . , 2016; Figure 1—source data 1 ) . Additionally , UV radiation data were extracted from the glUV dataset ( Beckmann et al . , 2014 ) using the R package ‘raster’ ( Hijmans , 2020; R Development Core Team , 2020 ) . Correlations between individual environmental variables and LUVp was calculated using the ‘lm’ function implemented in R . A correlation matrix between all environmental variables , and LUVp was calculated using the ‘cor’ function in R and plotted using the ‘heatmap . 2’ function in the ‘gplots’ package ( Warnes et al . , 2009 ) . Plots of the interactions between relative humidity and average temperature in relation to LUVp were generated using the ‘interact_plot’ function implemented in the ‘interactions’ R package ( Long , 2020 ) . It should be noted that the values for climate variables used in these analyses are extrapolated from weather stations across North America , and not measured in situ , meaning that they might not account for microclimatic variation . For example , two populations in Southern Arizona do not fit the pattern we proposed – they have small floral UV patterns and high frequency of S alleles at the Chr15_LUVp SNP , despite being associated with relatively low RH values in our datasets . However , one of them ( ANN_13 ) was collected along the Verde River , near Deadhorse lake , and the description of the collection site is ‘riparian forest and wetland , ’ suggesting that humidity might be locally higher than in the surrounding region . Similarly , from satellite pictures , the collection site for the other population ( ANN_47 ) appears considerably more verdant than other collection sites in Arizona . GEAs were analysed using BayPass ( Gautier , 2015 ) version 2 . 1 . Population structure was estimated by choosing 10 , 000 putatively neutral random SNPs under the BayPass core model . The Bayes factor ( denoted BFis as in Gautier , 2015 ) was then calculated under the standard covariate mode . For each SNP , BFis was expressed in deciban units [dB , 10 log10 ( BFis ) ] . Significance was determined following Gautier , 2015 and employing Jeffreys’ rule ( Jeffreys , 1961 ) , quantifying the strength of associations between SNPs and variables as ‘strong’ ( 10 dB ≤ BFis < 15 dB ) , ‘very strong’ ( 15 dB ≤ BFis < 20 dB ) , and ‘decisive’ ( BFis ≥ 20 dB; Figure 4—source data 4 ) . Water loss was determined by measuring changes in the weight of detached ligules and leaves over time ( Duursma et al . , 2019; Hygen , 1951 ) . In the summer of 2020 , fully developed inflorescences and the one or two youngest fully developed leaves from each individual were collected from well-watered , greenhouse-grown plants that had large ( LUVp = 1 ) or small ( LUVp ≤ 0 . 4 ) floral UV patterns . They were brought immediately to an environment kept at 21°C and were left overnight with their stems or petioles immersed in a beaker containing distilled water . The following morning leaves from each plant , and three ligules removed from each inflorescence ( selected to be as far apart from each other as possible across the inflorescence and taking care to avoid damaged or otherwise unrepresentative ligules ) , were individually weighed and hanged to air dry at room temperature ( 21°C ) . Their weight was measured at 1 hr intervals for 5 hr , and then again the following morning . Leaves and ligules were then incubated for 48 hr at 65°C in an oven to determine their dry weight . Total water content was measured as the difference between the initial fresh weight ( W0 ) and dry weight ( Wd ) . Water loss was expressed as a fraction of the total water content of each organ using the formula [ ( Wi-Wd ) / ( W0-Wd ) ] × 100 , where Wi is the weight of a sample at a time i . The assay was performed on ligules from 16 inflorescences from 12 individuals belonging to seven different populations of H . annuus , and on leaves from 15 individuals from eight different populations . Of the individuals used for assays on leaves , 10 were also used for assays on ligules , 4 were half-siblings of individuals used for ligule assays , and 1 belonged to a different population ( Figure 4—source data 3 ) . | Flowers are an important part of how many plants reproduce . Their distinctive colours , shapes and patterns attract specific pollinators , but they can also help to protect the plant from predators and environmental stresses . Many flowers contain pigments that absorb ultraviolet ( UV ) light to display distinct UV patterns – although invisible to the human eye , most pollinators are able to see them . For example , when seen in UV , sunflowers feature a ‘bullseye’ with a dark centre surrounded by a reflective outer ring . The sizes and thicknesses of these rings vary a lot within and between flower species , and so far , it has been unclear what causes this variation and how it affects the plants . To find out more , Todesco et al . studied the UV patterns in various wild sunflowers across North America by considering the ecology and molecular biology of different plants . This revealed great variation between the UV patterns of the different sunflower populations . Moreover , Todesco et al . found that a gene called HaMYB111 is responsible for the diverse UV patterns in the sunflowers . This gene controls how plants make chemicals called flavonols that absorb UV light . Flavonols also help to protect plants from damage caused by droughts and extreme temperatures . Todesco et al . showed that plants with larger bullseyes had more flavonols , attracted more pollinators , and were better at conserving water . Accordingly , these plants were found in drier locations . This study suggests that , at least in sunflowers , UV patterns help both to attract pollinators and to control water loss . These insights could help to improve pollination – and consequently yield – in cultivated plants , and to develop plants with better resistance to extreme weather . This work also highlights the importance of combining biology on small and large scales to understand complex processes , such as adaptation and evolution . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"genetics",
"and",
"genomics"
] | 2022 | Genetic basis and dual adaptive role of floral pigmentation in sunflowers |
Plaque rupture occurs if stress within coronary lesions exceeds the protection exerted by the fibrous cap overlying the necrotic lipid core . However , very little is known about the biomechanical stress exerting this disrupting force . Employing optical coherence tomography ( OCT ) , we generated plaque models and performed finite-element analysis to simulate stress distributions within the vessel wall in 10 ruptured and 10 non-ruptured lesions . In ruptured lesions , maximal stress within fibrous cap ( peak cap stress [PCS]: 174 ± 67 vs . 52 ± 42 kPa , p<0 . 001 ) and vessel wall ( maximal plaque stress [MPS]: 399 ± 233 vs . 90 ± 95 kPa , p=0 . 001 ) were significantly higher compared to non-ruptured plaques . Ruptures arose in the immediate proximity of maximal stress concentrations ( angular distances: 21 . 8 ± 30 . 3° for PCS vs . 20 . 7 ± 23 . 7° for MPS ) ; stress concentrations excellently predicted plaque rupture ( area under the curve: 0 . 940 for PCS , 0 . 950 for MPS ) . This prediction of plaque rupture was superior to established vulnerability features such as fibrous cap thickness or macrophage infiltration . In conclusion , OCT-based finite-element analysis effectively assesses plaque biomechanics , which in turn predicts plaque rupture in patients . This highlights the importance of morpho-mechanic analysis assessing the disrupting effects of plaque stress .
Coronary artery disease ( CAD ) is one of the major causes of morbidity and mortality in the Western World ( Center for Disease Control , 2013 ) . One of the main challenges in modern therapy of patients with CAD is the detection of atherosclerotic plaques that do not yet limit flow in the coronary artery , but may potentially cause an acute coronary syndrome by rupturing with subsequent acute occlusion of the artery . Decades ago , these lesion entities have been defined as ‘vulnerable plaques’ ( Virmani et al . , 2003 ) . Pathology-based studies ( Virmani et al . , 2003; Falk , 1989; Burke et al . , 1997; Virmani et al . , 2000 ) as well as intravascular imaging ( Kato et al . , 2012; Kubo et al . , 2013; Reith et al . , 2014; Uemura et al . , 2012; Ehara et al . , 2004 ) have been used over time to explore potential features of plaque vulnerability , that is morphological characteristics which may predispose a plaque to rupture . The presently accepted features assessed using intravascular imaging include the thickness of the fibrous cap ( FCT ) ( Kato et al . , 2012; Kubo et al . , 2013; Reith et al . , 2014; Uemura et al . , 2012 ) , the extent of the necrotic lipid core ( Kato et al . , 2012 ) , the presence of macrophages ( Kato et al . , 2012; Reith et al . , 2014; Uemura et al . , 2012 ) , microvessels ( Kato et al . , 2012; Uemura et al . , 2012 ) , small calcifications ( Ehara et al . , 2004; Reith et al . , 2018 ) , or a positive remodeling ( Varnava et al . , 2002 ) . Features of coronary plaque vulnerability may also be assessed with different imaging modalities , such as coronary computed tomography angiography , and include also low-attenuation plaques and a higher plaque burden ( Conte et al . , 2017 ) . These features may predict the incidence of future major cardiac events ( Conte et al . , 2017; Prati et al . , 2020 ) , if such high-risk plaques are not timely recognized and sealed ( Dettori et al . , 2020 ) . From a mechanistic perspective , plaque rupture originates from an excessive stress concentration on the fibrous cap , which trespasses the tensile strength of the material and ultimately causes material failure and consequently plaque rupture . The fibrous cap yields , in this view , a protective effect , by avoiding the exposure of prothrombotic material present in the necrotic lipid core . Thus , it is not surprising that minimal FCT has been considered the most established parameter of plaque vulnerability ( Kato et al . , 2012; Kubo et al . , 2013; Reith et al . , 2014; Uemura et al . , 2012 ) : in fact , the thickness of the fibrous cap seems a good surrogate parameter of its ability to resist stresses . Still , evaluating solely the fibrous cap addresses only the main stabilizing factor in plaque biomechanics , leaving the destabilizing factors almost completely ignored . However , plaque rupture occurs if the stress within coronary lesions exceeds the protection exerted by the fibrous cap . Therefore , it is of utmost importance to also calculate the stress within the plaque as the disrupting force within the lesion and not only focus on the protective effects of the fibrous cap . However , very little is known about the determinants of the disrupting force pushing coronary plaques to rupture . In addition , the spatial relationship between rupture and the balance of stabilizing and destabilizing forces in the plaque remains largely unexplored: in other words , whether plaque rupture occurs at the site with the lowest FCT , as point of least resistance , or rather in sites of highest stress concentration , as point of maximal disruption is unclear . It is self-evident that such research questions cannot be addressed by ‘standard’ plaque imaging analysis . Therefore , we chose to develop a morpho-metric finite-element analysis to determine the influence of plaque morphology on the biomechanics of coronary lesions in patients and investigate the relationship between this calculated morpho-mechanic stress and plaque rupture . As a source of patient-specific , high-quality images of coronary plaques for this proof-of-concept study , we selected optical coherence tomography ( OCT ) , an established intravascular imaging technique which , due to a supreme resolution ( up to 10–20 µm ) , is able to depict peri-luminal structures with high accuracy ( Burgmaier et al . , 2014 ) . Based on the images obtained , we generated patient-specific reconstructions of coronary plaques , which we used as a basis to analyze stress concentration as a possible predictor of plaque rupture .
Of the 20 patients with type 2 diabetes enrolled in this study , 10 underwent coronary angiography due to acute coronary syndrome and 10 due to chronic coronary syndrome . All selected patients with acute coronary syndromes showed plaque rupture as the morphological correlate . Patients with ( n = 10 ) and without ( n = 10 ) plaque rupture did not differ with respect to their clinical characteristics , apart from a worse glycemic control in patients with plaque rupture ( HbA1c: 7 . 4 ± 1 . 4 vs . 6 . 1 ± 0 . 5 , p=0 . 026 ) . As expected , ruptured lesions presented a lower FCT ( minimal FCT: 49 ± 10 µm vs . 97 ± 15 µm , p<0 . 001; mean FCT: 94 ± 17 µm vs . 133 ± 12 µm , p=0 . 006 ) , a more extensive necrotic lipid core ( lipid arc: 178 ± 39° vs . 110 ± 8° , p=0 . 001; lipid volume index: 9876 ± 3088 vs . 3853 ± 1294 , p=0 . 011 ) , and a higher incidence of thin-capped fibroatheromas ( 70% vs . 0% , p=0 . 003 ) compared to non-ruptured lesions . Patients and lesions characteristics are reported in Table 1 . In ruptured plaques , the maximal stress within the fibrous cap ( peak cap stress ) was significantly higher than in non-ruptured ones ( 174 ± 67 vs . 52 ± 42 kPa , p<0 . 001 ) . Furthermore , also the maximal stress within the whole plaque ( maximal plaque stress ) was more than fourfold higher in ruptured plaques compared to stable ones ( 399 ± 233 vs . 90 ± 95 kPa , p=0 . 001 ) . Exemplary images of stress concentrations on the fibrous cap and overall in plaques , as well as a box-plot depicting stress concentrations in lesions with and without rupture are shown in Figure 1 . Variation of stress distribution in dependence of various model assumptions are reported in the Supplementary Results in Appendix 1 . After documenting much greater stress in ruptured compared with stable plaques , we assessed the effects of these morphology-driven peaks of intra-plaque stress on the lesion and if points of highest stress co-localize with plaque rupture within coronary lesions . Thus , we aimed to analyze the spatial correlation between the rupture point , as visualized in OCT images , and the point of maximal stress concentration , in order to further support the mechanistic role of stress concentrations in the fibrous cap in the genesis of plaque rupture . Only 50% of the plaque ruptures were in the shoulder region . The angle between the peak cap stress and the detectable rupture site was very low at 21 . 8 ± 30 . 3°; the angle between the maximal plaque stress and the detectable rupture site was even lower ( 20 . 7 ± 23 . 7° ) . Overall , 50% of plaque ruptures occurred within 10° of the peak cap stress and of the maximal plaque stress , suggesting that plaque rupture occurs in the very proximity of highest stress concentrations . A graphical presentation of spatial correlation between the sites presenting the peak cap stress and the maximal plaque stress in finite-element analysis and the rupture site in OCT pullback is shown in Figure 2 . In order to assess the diagnostic efficacy of finite-element analyses in predicting plaque rupture , we performed receiver operating characteristic ( ROC ) analysis for both peak cap stress and maximal plaque stress . In our cohort , we could demonstrate that both peak cap stress ( area under the curve [AUC] 0 . 940 ) and maximal plaque stress ( AUC 0 . 950 ) predict plaque rupture with excellent accuracy . Optimal cut-off for prediction of plaque rupture were 150 . 5 kPa ( Sensitivity 90%; Specificity 80% ) for peak cap stress and 169 . 5 kPa ( Sensitivity 90%; Specificity 90% ) for maximal plaque stress . ROC curves for the prediction of plaque rupture are shown in Figure 3 . To assess the possible additional value of finite-elements analysis compared to known features of plaque vulnerability , we performed ROC analysis for prediction of plaque rupture . Minimal FCT ( AUC 0 . 630 ) and mean FCT ( AUC 0 . 630 ) presented sufficient diagnostic accuracy; lipid volume index ( AUC 0 . 870 ) showed a very good diagnostic accuracy in predicting plaque rupture . Both peak cap stress and maximal plaque stress presented a significantly superior diagnostic efficiency in predicting plaque rupture compared to mean FCT ( p=0 . 027 vs . peak cap stress; p=0 . 036 vs . maximal plaque stress ) , minimal FCT ( p vs . peak cap stress: 0 . 027; p vs . maximal plaque stress = 0 . 036 ) and extent of macrophage infiltration ( p=0 . 003 vs . peak cap stress; p=0 . 001 vs . maximal plaque stress ) . Furthermore , diagnostic efficiency of both peak cap stress and maximal plaque stress was numerically superior ( albeit non-significant ) compared to that of lipid volume index ( p=0 . 409 vs . peak cap stress and p=0 . 386 vs . mean plaque stress ) . We combined OCT-derived morphologic features of plaque vulnerability by calculating two previously validated scores , shown in the CLIMA study ( Prati et al . , 2020 ) and in a publication by Burgmaier et al . , 2014 . Both scores could predict plaque rupture with very good to excellent diagnostic efficiency ( AUC for CLIMA score: 0 . 870; AUC for Burgmaier score: 0 . 900 ) , which however remained numerically lower than results of our morpho-mechanic analysis . ROC curves , including comparison with results of finite-element analyses , are shown in Figure 4 .
First , we could in vivo detect higher stress concentrations both within the fibrous cap and within the whole plaque in ruptured plaques compared to non-ruptured plaques . This is in line with previous simulation studies based on pathology specimens ( Loree et al . , 1992; Cheng et al . , 1993 ) and confirms our hypothesis of stress concentration as a central factor in the genesis of plaque rupture . As a further step in confirming this theory , we extended current knowledge by showing that plaque ruptures arise in the immediate proximity of maximal stress concentrations . This spatial coincidence between maximal stress concentrations and sites of plaque rupture confirms the causal link between plaque biomechanics on the one hand and plaque rupture and acute coronary syndromes on the other . In fact , on the basis of our data , it is tempting to speculate about the mechanistic process eventually leading to plaque rupture . Stress within the fibrous cap and overall within the plaque is concentrated in a pattern which is closely dependent on the morphology of the plaque ( for instance , in dependence of the extent and morphology of the necrotic lipid core and/or of calcification ) . A specific force is exerted on each point of the fibrous cap; should this stress concentration trespass the ultimate tensile strength of the vessel wall – which is highly likely to happen in points of maximal stress concentration – material failure and , eventually , plaque rupture occurs . An interesting question is , of course , the magnitude of the threshold that needs to be trespassed to cause material failure and plaque rupture . A previous pathology-based study by Cheng et al . , 1993 sets the stress threshold for plaque rupture to 300 kPa . In our study , we found 150 . 5 kPa in the fibrous cap and 169 . 5 kPa in the vessel wall to be the optimal thresholds for predicting plaque rupture . The difference between these values and the initially estimated threshold of 300 kPa may be explained through the different imaging modalities used to reconstruct the coronary plaque . In fact , in pathology specimens ( on which the 300 kPa value is based on ) , the fixation process may cause a shrinkage of the fibrous cap with a shortening of ca . 10–20% compared to the thickness measured in vivo with OCT . This effect may be sufficient to explain the numerical difference of the ‘critical’ peak cap stress needed for plaque rupture , especially considering – as pointed out by Finet et al . – that the relationship between FCT and peak cap stress is exponential ( Finet et al . , 2004 ) . Further clinical validation of these thresholds is , however , needed . Another possible explanation might be the hyperelastic behavior of certain plaque components , which has been described by some authors previously ( Yang et al . , 2009; Cardoso et al . , 2014 ) ; such an effect is , however , only present at large displacements ( >20–30% of the initial dimensions ) ( Yang et al . , 2009; Kobielarz et al . , 2020 ) , which are way over the average displacements reported in our study . In this range , a linear equation adequately depicts the behavior of the material . Also , stress on the lesion in patients is not only derived from the static factors included in our analysis , but also dynamic flow shear stress caused by blood flow ( Bourantas et al . , 2020 ) , which was not included in our study . Although dynamic flow shear stress was not included in our analysis , we could also show an excellent diagnostic efficiency of our finite-element analysis in predicting plaque rupture . Furthermore , the model we developed presented a clearly superior diagnostic efficiency compared with accepted parameters of plaque vulnerability as FCT ( Kato et al . , 2012; Kubo et al . , 2013; Reith et al . , 2014; Uemura et al . , 2012 ) or plaque macrophage infiltration ( Kato et al . , 2012; Reith et al . , 2014; Uemura et al . , 2012 ) and a numerically superior efficiency when compared to lipid volume index ( Kato et al . , 2012 ) in our cohort . The diagnostic efficiency of our model even presented a comparable efficiency when compared with the combination of lesion morphologies including FCT , plaque macrophage infiltration , and lipid volume index . In the light of these findings , it may be tempting to speculate about the strong link existing between stress concentrations and plaque rupture , which may explain the excellent predictive value of our model and may pave the way for widespread use of morpho-mechanical analysis in the clinical routine to detect vulnerable plaques . To the best of our knowledge , although we are first to develop an OCT-based finite-element model allowing patient-specific analysis of plaque biomechanics , several limitations have to be acknowledged . First of all , we faithfully reproduced plaque morphology in a 2D-based reconstruction; on the other hand , we are not taking into account several phenomena that may influence stress concentration , as for instance the longitudinal structure of the plaque or the longitudinal profile of the stenosis and flow shear stress – the relevance of these factors needs to be addressed in future studies . Furthermore , in spite of the excellent resolution of OCT in the near field , we cannot exclude imprecisions in the segmentation due to limited light penetration to the deeper vessel wall . Limited tissue penetration , in fact , does not allow to assess sites with positive remodeling , which may yield higher stresses due to accelerated plaque growth and therefore cause a higher rupture risk . This needs to be assessed through different imaging modalities , such as IVUS or coronary computed tomography . Moreover , dedicated computational techniques in order to reconstruct the deep structure of the lipid core ( Kok et al . , 2016 ) have not been employed , in order to simplify plaque reconstruction; moderate imprecisions in the deep contour of the lipid core cannot therefore be excluded . For our study , we employed a linear elastic model; other authors , though , suggest an hyperelastic behavior of the vessel wall , particularly for displacements > 20% of the initial dimensions ( Yang et al . , 2009; Kobielarz et al . , 2020 ) . In spite of an average displacement ‘small enough’ ( 10% ) to justify linearity , we cannot exclude underestimation of stress concentrations for very few plaques with larger displacements . To the best of our knowledge , although being the first to employ a morpho-metric approach in assessing plaque vulnerability on in vivo intravascular imaging , this pilot study still includes a low number of patients chosen among patients with and without plaque rupture , which may reduce the reliability of exact AUC and cut-off values in our predictive models; further analyses are needed to confirm the results of our proof-of-concept analysis .
We retrospectively selected 20 patients with type 2 diabetes mellitus and CAD , who underwent OCT prior to percutaneous coronary intervention at the Department of Cardiology of the University Hospital of the RWTH Aachen . Clinical presentation was stable CAD without evidence of plaque rupture in the OCT pullback ( n = 10 ) or acute coronary syndrome with plaque rupture ( n = 10 ) . Sample size calculation was performed based on previous results of calculation of mechanical stresses in histopathological samples , resulting in a minimal sample size of seven lesions per group in order to achieve α = 0 . 001 and power of 0 . 95; this was then arbitrarily rounded to 10 lesions per group . Informed consent of all patients was obtained prior to inclusion in the study . The study was approved by the Ethics Committee of the University Hospital of the RWTH Aachen ( EK 071/11 and EK 277/12 ) and is in accordance with the declaration of Helsinki on ethical principles for medical research involving human subjects . The acquisition of OCT pullbacks was performed as previously described in the literature ( Tearney et al . , 2012; Milzi et al . , 2017 ) . In brief , OCT images were acquired prior to coronary intervention using a frequency domain OCT C7XR system and the DragonFly catheter ( St . Jude Medical Systems; Lightlab Imaging Inc , Westford , MA ) . Blood removal was obtained by the injection of 14 ml contrast dye ( iodixanol ) at a flow rate of 4 ml/s through the guiding catheter . Image acquisition was obtained with automated pull-back rate of 20 mm/s . Analysis of plaque morphology was performed as previously described ( Tearney et al . , 2012; Milzi et al . , 2017 ) . In particular , as widely employed in clinical practice and in previous intravascular imaging studies , FCT is measured at different sites ( conventionally 3 ) per frame , with an analysis performed on different frames in 0 . 1–0 . 2 mm intervals . Usually , the rupture is localized in a single point or in a localized area of the cap , not impeding measurement even in the frame ( s ) with evident rupture , though in slightly different sites . In order to combine different features of plaque vulnerability for comparison with results of morpho-mechanical analysis , we calculated two different , established scores . The score validated in a publication from Burgmaier et al . ( following: Burgmaier Score ) is calculated as −2 . 401 + 1 . 568 * ( insert one if macrophages present; else 0 ) + 2 . 639 * ( medium lipid arc in multiples of 90° ) + 0 . 255 * ( lipid plaque length in mm ) − 0 . 738 * ( minimal FCT in multiples of 10 μm ) , as previously described ( Burgmaier et al . , 2014 ) . The score based on the CLIMA study ( following: CLIMA score ) attributed one point each to the presence of MLA < 3 . 5 mm2 , FCT < 75 µm , lipid arc circumferential extension > 180° , and presence of OCT-defined macrophages ( Prati et al . , 2020 ) . In order to obtain a patient-specific model of the coronary plaque , we selected a single frame from every OCT pullback . In ruptured plaques , the frame immediately preceding the site of the rupture was chosen; this was based on the consideration that , due to rupture of the fibrous cap and to the present artifacts ( caused for instance by thrombotic material ) , exact reconstruction of the rupture site may be inaccurate . For stable lesions , the frame with the minimal lumen area was selected; this was based on the need to select in a uniform way the site with the most advanced plaque development , assuming only a negative remodeling . After image selection , the operator ( AM ) was blinded to clinical presentation , as each frame was marked with a random ID . The selected image was then scaled 10:1 previous to segmentation of the different components of the atherosclerotic plaque , which was manually performed using commercial software ( AutoCAD 2017 , AutoDesk INC , San Rafael , CA ) . Plaque composition was analyzed according to the Consensus standards ( Tearney et al . , 2012 ) . Specifically for segmentation , we first delineated the vessel lumen , defined as the signal-poor region centrally located in the OCT image . We then proceeded to tracing the boundaries of the fibrous cap , which was defined as the signal-rich region surrounding the lumen; for clarity’s sake , this area will be denominated fibrous cap also when overlying a calcific or fibrocalcific plaque . When present , we segmented the necrotic lipid core , which was identified as a signal-poor region with poorly delineated borders , a fast signal drop-off , and little or no signal backscattering . When present , calcifications have also been segmented; as calcifications we considered signal-poor or heterogeneous regions with a sharply delineated border . In case of non-detectable borders of each of the segmented components due to artifacts or to the limited penetration of light , we delineated contours with the automatic interpolation function of the software . We conventionally shaped the coronary vessel as a cylinder and set its external diameter to 4 mm . A sample reconstruction is shown in Figure 5 . In order to obtain a three-dimensional model of the vessel , we performed a graphical extrusion of the segmented contours over a length of 10 mm . The modeled vessel was then imported in commercial software to perform finite-element analysis ( COMSOL Multiphysics 5 . 0 , Stockholm , Sweden ) . A solid mechanics physics was chosen , and the used mechanical properties for the different components of the plaque were extrapolated from previous literature ( Loree et al . , 1992; Cheng et al . , 1993; Finet et al . , 2004; Reith et al . , 2019 ) and are reported in Table 2 . Nevertheless , in the literature , there is no consensus regarding mechanical properties of the atherosclerotic plaque . This is a consequence of the very limited number of samples used for mechanical testing , of the need for pre-treatment of pathology samples ( which may potentially alter their mechanical properties ) , and of the intrinsic difficulty of the measurement of some properties ( specifically , the Poisson’s ratio , which is only indirectly measurable ) . In particular , some authors hypothesized an incompressible behavior of the vessel wall , which would lead to a Poisson’s ratio of 0 . 48 for this component ( instead of 0 . 27 as used in our simulations ) ( Yang et al . , 2009 ) . Moreover , very different values of stiffness of calcifications have been detected and employed in previous studies , ranging from about 10 MPa for mildly calcified tissues to even 17–25 GPa emulating the properties of bone tissue ( Loree et al . , 1992; Cheng et al . , 1993; Finet et al . , 2004; Kobielarz et al . , 2020; Holzapfel et al . , 2005; Barrett et al . , 2019; Wong et al . , 2012; Cahalane et al . , 2018 ) . Specifically , the use of smaller Young’s moduli is justified by the inhomogeneity of macrocalcifications , which may be only partly constituted by crystalline calcium Wong et al . , 2012; this could be associated , in previous studies , to the density of calcifications in computed tomography ( Cahalane et al . , 2018 ) . Though , as calcium density is not defined in OCT , we preferred to use 10 GPa , as the Young’s modulus of crystalline calcium , as calcification’s stiffness . Nevertheless , in order to exclude a relevant impact of these assumptions on our stress analysis , we assessed stress distribution in the presence of different assumptions . If not differently specified , the data presented in the rest of the manuscript derive from the model using constants shown in Table 2 . A pressure of 130 mmHg ( = 17 kPa ) on the luminal side was applied as external load and a simulation of the structural stress distribution in response to the load was performed . The round outer surface of the vessel was kept as fixed constraint . A suitable mesh was chosen per each simulated vessel , ranging from ‘fine’ to ‘very fine’ , in order to keep computational times reasonable . Average mesh properties are reported in Supplementary file 1 . Then , we analyzed the stress distribution as von Mises stress in a bi-dimensional cross-section of the vessel normal to the vessel axis; in order to avoid edge effects , we used cross-sections at a distance of 5 mm from each end . A graphical representation of the model is included in Figure 5—figure supplement 1 . The stress intensity on the fibrous cap was graphically shown on a blue-red color scale . The highest von Mises stress in the fibrous cap was defined as ‘Peak Cap Stress’; the highest von Mises stress in the vessel wall was defined as ‘Maximal Plaque Stress’ . The finite-element analysis was performed in a dedicated core lab from an operator ( EDL ) blinded to the clinical presentation of the patients . To exclude excess interobserver variability in manual segmentation , a different experienced OCT operator ( RD ) redraw the analyzed structures in a randomly selected 20% of the plaques in each group . Based on these segmentations , we run finite-element analysis . Rupture site was identified in OCT images as a clear , visible continuity interruption in the fibrous cap . The rupture site was marked from an operator ( AM ) blinded to the results of stress distribution . In case of non-punctual ruptures of the fibrous cap , an arbitrary middle point was used for calculation purposes . Then , results of stress analysis for both the fibrous cap and the overall vessel wall were reported on the OCT-image of the rupture site , correlating them to anatomic landmarks ( morphology of lumen; morphology of the fibrous cap; presence of calcifications , lipid deposits , cholesterol crystals , or macrophage accumulations ) . The angular distance between the rupture site and the sites where the simulation highlighted maximal stress concentrations was noted . Graphical representation was obtained with polar angle histograms generated with Origin ( OriginLab Corp , Northampton , MA ) . Continuous variables were reported as mean ± standard deviation , categorical as count ( percentage ) . Distributions of continuous variables were compared using t-test; for comparing categorical values , Fisher’s exact test was used for comparing distribution of categorical variables . To compare results of simulations with different assumptions regarding material properties , paired-samples t-test was used . Furthermore , in order to validate the finite-element simulation , we correlated results obtained in different models and after segmentation through different operators using one-way random effects model; results were expressed as intraclass correlation coefficient . We performed ROC analysis to validate diagnostic value of the results of simulation models as well as ‘classical’ features of plaque vulnerability ( FCT , lipid volume index , extent of plaque macrophages infiltration ) in predicting plaque rupture . Values with the highest Youden index were identified as optimal cut-off values; in case of equal Youden index between two or more data points , we selected one based on clinical judgment . In order to evaluate the value of a combination of OCT-derived morphologic parameters to predict plaque rupture , we performed multivariable logistic regression including minimal or mean FCT , lipid volume index , and macrophage volume index . Lipid volume index was calculated as the product of mean lipid angle and lipid length , as defined in previous works ( Kato et al . , 2012; Kubo et al . , 2013; Reith et al . , 2014; Uemura et al . , 2012; Ehara et al . , 2004; Reith et al . , 2018; Tearney et al . , 2012; Milzi et al . , 2017 ) . For calculation purposes , non-defined parameters such as FCT in case of calcified plaques or lipid volume index in non-lipidic plaques were set to zero . Then , ROC analysis was performed based on the predictive values of this multiple regression model . A classification of the diagnostic efficiency according to the values of the area under the curve ( AUC ) was used as described elsewhere ( Šimundić , 2009 ) . In order to compare diagnostic efficiency of results of morpho-mechanic analysis with combined features of coronary plaque vulnerability , we calculated the Burgmaier and CLIMA scores as previously described . Comparison of the diagnostic efficiency among different ROC curves was performed with the DeLong test , as previously described ( DeLong et al . , 1988 ) . All statistical analyses were performed with SPSS software ( v . 26 . 0 , IBM Corp . , Armonk , NY ) . Statistical significance was awarded for p<0 . 05 . In our proof-of-concept study , we demonstrate that OCT-based finite-element analysis is a feasible tool to determine plaque biomechanics , which in turn may predict plaque rupture in patients . Whereas the minimal fibrous cap thickness protects the plaque from its rupture , our data highlight the importance of morpho-mechanic analysis assessing the disrupting effects of plaque stress . These data need , however , to be verified in larger populations . This new method may offer valuable insights on the interplay between various plaque components in the determination of the net vulnerability of a plaque , bringing stress concentrations – the disrupting force of plaque rupture – back into clinical practice . | Heart attacks are caused by a blockage in arteries that supply oxygen to the heart . This often happens when fatty deposits ( or ‘plaques’ ) that line blood vessels break off and create a clot . To identify individuals most at risk of this occurring , physicians currently use symptoms , family history , blood tests , imaging and surgical procedures . But better methods are needed . Imaging blockages in the arteries of individuals who died from heart attacks highlighted certain plaque characteristics that increase the risk of a rupture . Further understanding the forces that lead to these fatty deposits breaking off may help scientists to develop improved heart attack prediction methods . Using patient-specific computer simulations , Milzi et al . show it is possible to predict where plaques are most likely to rupture in an individual , based on biomechanical stresses on the deposits in the artery . The models also showed how forces on the external layers of the plaque played a pivotal role in breakages . More research is needed to confirm the results of this study and to develop automated ways for measuring the stress exerted on plaques in the arteries . If that research is successful , biomechanical analyses of artery plaques in routine patient assessments may one day allow physicians to predict heart attacks and provide life-saving preventive care . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine"
] | 2021 | Coronary plaque composition influences biomechanical stress and predicts plaque rupture in a morpho-mechanic OCT analysis |
Early adversity is a risk factor for the development of adult psychopathology . Common across multiple rodent models of early adversity is increased signaling via forebrain Gq-coupled neurotransmitter receptors . We addressed whether enhanced Gq-mediated signaling in forebrain excitatory neurons during postnatal life can evoke persistent mood-related behavioral changes . Excitatory hM3Dq DREADD-mediated chemogenetic activation of forebrain excitatory neurons during postnatal life ( P2–14 ) , but not in juvenile or adult windows , increased anxiety- , despair- , and schizophrenia-like behavior in adulthood . This was accompanied by an enhanced metabolic rate of cortical and hippocampal glutamatergic and GABAergic neurons . Furthermore , we observed reduced activity and plasticity-associated marker expression , and perturbed excitatory/inhibitory currents in the hippocampus . These results indicate that Gq-signaling-mediated activation of forebrain excitatory neurons during the critical postnatal window is sufficient to program altered mood-related behavior , as well as functional changes in forebrain glutamate and GABA systems , recapitulating aspects of the consequences of early adversity .
Early-life experience plays a crucial role in the maturation and fine-tuning of neurocircuitry that drives emotional behavior in adulthood ( Hensch , 2004; Hensch , 2005; Bale et al . , 2010; Berardi et al . , 2000; Carr et al . , 2013; Kessler et al . , 2010 ) . Both clinical and preclinical evidence indicates that early-life adversity serves as a key risk factor for the development of adult psychopathology , increasing susceptibility to psychiatric disorders like anxiety , major depression and schizophrenia ( Carr et al . , 2013; Anda et al . , 2006; Knuesel et al . , 2014; Wright et al . , 1995; Glover , 2011 ) . Stressful experiences in adulthood can produce behavioral alterations that are often transient in nature , however perturbations in the vulnerable perinatal ‘critical window’ can program lasting changes in emotional behavior ( Heim and Nemeroff , 2001; McEwen , 2003; Ogle et al . , 2015 ) . Several animal models have been used to study the persistent behavioral changes caused by early- life perturbations , and have been instrumental in understanding specific underlying neural mechanisms involved in the programming of adult emotional behavior ( Glover , 2011; Heim and Nemeroff , 2001; Francis et al . , 1999; Walker and McCormick , 2009; Weinstock , 2008; Welberg and Seckl , 2001; Ansorge et al . , 2007; Ansorge et al . , 2004 ) . The prenatal period and the first few weeks after birth are marked by the establishment and functional maturation of several neurocircuits in rodent models , representing a critical period in which these circuits are particularly amenable to modification by environmental stimuli ( Hensch , 2005; Carr et al . , 2013 ) . Rodent models of early-life perturbations encompass those based on gestational stress ( Glover , 2011 ) , maternal immune activation ( Knuesel et al . , 2014; Wright et al . , 1995 ) , disruption of dam-pup interaction ( Liu et al . , 1997; Levine and Lewis , 1959 ) or pharmacological treatments ( Oberlander et al . , 2009; Cutler et al . , 1996; Sarkar et al . , 2014a ) , and exhibit both distinct and overlapping behavioral and physiological effects in adulthood ( Carr et al . , 2013; Heim and Nemeroff , 2001 ) . Strikingly , a commonality noted across these animal models is the fact that multiple molecular , cellular , functional and behavioral changes often persist throughout the animal’s lifespan ( Carr et al . , 2013; Heim and Nemeroff , 2001; Suri and Vaidya , 2015 ) . Amongst the underlying mechanisms implicated in the establishment of such long-lasting changes in response to early-life perturbations are a dysregulation of the hormonal stress response pathway ( Oberlander et al . , 2009; Kalinichev et al . , 2002; Leussis et al . , 2012; Wilber et al . , 2009; Fish et al . , 2006; Gillespie et al . , 2009 ) , serotonergic system ( Altieri et al . , 2015; Gross and Hen , 2004; Shah et al . , 2018 ) , and emergence of excitation-inhibition balance within key cortical neurocircuits ( Sohal and Rubenstein , 2019; Gatto and Broadie , 2010 ) . Common across several rodent models of early-life perturbations are alterations in G protein-coupled receptor ( GPCR ) signaling , including via the serotonin1A ( 5-HT1A ) receptor ( Gross et al . , 2002; Richardson-Jones et al . , 2010; Goodfellow et al . , 2009 ) , serotonin2A ( 5-HT2A ) receptor ( Benekareddy et al . , 2010; Benekareddy et al . , 2011; Weisstaub et al . , 2006; Malkova et al . , 2014; Wischhof et al . , 2015 ) , metabotropic glutamate receptors 1 and 5 ( mGluR1/5 ) ( Genty et al . , 2018; Lin et al . , 2018 ) , muscarinic acetylcholine receptor 1 ( M1 ) ( Proulx et al . , 2014 ) and α1 adrenergic receptors ( Loria and Osborn , 2017 ) . The emergence of aberrant emotional behavior in these animal models has been suggested to involve a key role for both excitatory Gq-coupled and inhibitory Gi-coupled GPCRs , in particular an appropriate balance of signaling between the Gq-coupled 5-HT2A receptor and the Gi-coupled 5-HT1A receptor in the forebrain has been hypothesized to be a critical determinant of the establishment of emotional behavior ( Sarkar et al . , 2014b; Sargin et al . , 2019; Lambe et al . , 2011; Vinkers et al . , 2010 ) . Enhanced signaling via the cortical 5-HT2A receptor is thought to be one of the common features noted across distinct models of early-life perturbations , including maternal separation ( Benekareddy et al . , 2010; Benekareddy et al . , 2011 ) , postnatal fluoxetine ( Sarkar et al . , 2014b ) and maternal immune activation ( Malkova et al . , 2014; Moreno et al . , 2011 ) . Interestingly , a systemic blockade of the Gq-coupled 5-HT2A receptor overlapping with early stress or postnatal fluoxetine treatment can prevent the emergence of adult anxiety and depressive behavior , and associated molecular and cellular correlates ( Benekareddy et al . , 2011; Sarkar et al . , 2014b ) . Furthermore , pharmacological stimulation of the 5-HT2A receptor during the postnatal critical window is sufficient to evoke a persistent increase in anxiety in adulthood ( Sarkar et al . , 2014b ) . Collectively , these observations motivate the key question of whether perturbed Gq-coupled signaling within the forebrain in the critical postnatal window plays an important role in the establishment of persistent changes in mood-related behaviors . Here , we have tested the hypothesis that enhanced Gq-mediated signaling in forebrain excitatory neurons during the postnatal critical window may be sufficient to program persistent alterations in mood-related behavior in adulthood . To address this central question we expressed the excitatory Designer Receptors Exclusively Activated by Designer Drugs ( DREADD ) in CamKIIα-positive forebrain excitatory neurons using a bigenic mouse line ( CamKIIα-tTA::TetO hM3Dq ) ( Alexander et al . , 2009 ) , and chemogenetically activated Gq signaling through oral administration of the DREADD agonist clozapine-N-oxide ( CNO; 1 mg/kg ) from postnatal Day 2 to 14 prior to behavioral analysis in adulthood . Our findings demonstrate that chemogenetic activation of Gq signaling in CamKIIα-positive forebrain excitatory neurons by chronic postnatal CNO ( PNCNO ) treatment enhances anxiety- and despair-like behavior , accompanied by impaired sensorimotor gating in adulthood . These long-lasting behavioral changes evoked by PNCNO treatment are associated with a persistent dysregulation of cortical and hippocampal glutamate/GABA metabolism , and perturbed hippocampal excitatory and inhibitory neurotransmission . The criticality of the postnatal time window is highlighted by our observation that the same perturbation performed in the juvenile window or in adulthood has no effect on mood-related behavior . Our findings provide evidence in support of the hypothesis that enhanced Gq signaling within forebrain excitatory neurons during the critical postnatal window is sufficient to evoke perturbed mood-related behavior in adulthood , recapitulating the enhanced vulnerability to psychopathology associated with early adversity .
To examine the persistent behavioral , metabolic , molecular and electrophysiological consequences of postnatal chemogenetic hM3Dq DREADD activation of forebrain excitatory neurons , CamKIIα-tTA::TetO-hM3Dq bigenic mice were generated ( Figure 1A ) . This bigenic mouse line is reported to exhibit selective expression of the hM3Dq DREADD in Ca2+/calmodulin-dependent protein kinase α ( CamKIIα ) -positive excitatory neurons in the forebrain ( Alexander et al . , 2009; Figure 1B ) . Western blotting and immunofluorescence analysis confirmed the presence of the HA-tagged hM3Dq DREADD in both the hippocampus and cortex of bigenic mouse pups ( P7 ) ( Figure 1C , D ) . Expression of the HA-tagged hM3Dq DREADD was not observed in either the hippocampal subfields or cortex of single-positive , genotype-control mouse pups ( P7 ) ( Figure 1D ) . Further , in order to delineate cell type specificity for the expression of the HA-tagged hM3Dq DREADD , we performed double immunofluorescence staining for the HA-tag with the excitatory neuron marker , CamKIIα , the inhibitory neuron marker , GABA and the astrocyte marker , glial fibrillary acidic protein ( GFAP ) ( Figure 1—figure supplement 1 ) . We noted that the HA-tagged hM3Dq DREADD exhibited robust co-localization with CamKIIα-positive neurons in the hippocampus and neocortex ( Figure 1—figure supplement 1A , B ) . We did not observe any co-localization of the HA-tagged hM3Dq DREADD with either inhibitory neuron marker , GABA or the astrocyte marker GFAP in any of the brain regions examined ( Figure 1—figure supplement 1C ) . Further , we noted that the HA-tagged hM3Dq DREADD expression was restricted to the forebrain regions , and we did not observe any HA immunofluorescence in subcortical brain regions including the hypothalamus , pallidum , and periaqueductal gray ( Figure 1—figure supplement 1D ) . Collectively , these results indicate the restricted cell type specific expression of the HA-tagged hM3Dq DREADD in forebrain CamKIIα-positive neurons of CamKIIα-tTA::TetO-hM3Dq bigenic mice . We next assessed whether , acute stimulation of the hM3Dq DREADD by the exogenous ligand , CNO at postnatal Day 7 ( P7 ) , resulted in enhanced neuronal activation within the forebrain , using two distinct strategies . First , we performed western blotting analysis to determine the expression levels of the neuronal activity markers , c-Fos and phospho-ERK , following a single dose of CNO ( 1 mg/kg ) administered via feeding to bigenic mouse pups at P7 ( Figure 1E , F ) . Western blotting analysis revealed a significant increase in both p-ERK/ERK ( Figure 1G , F1 , 6 = 5 . 872 , p=0 . 02 ) and c-Fos ( Figure 1H , F1 , 6 = 7 . 48 , p=0 . 002 ) levels in the hippocampus of CNO-treated bigenic mouse pups . We also observed a trend toward an increase in p-ERK/ERK levels ( Figure 1I , F1 , 6 = 8 . 462 , p=0 . 07 ) and significant increase in c-Fos levels ( Figure 1J , F1 , 6 = 5 . 608 , p<0 . 0001 ) in the cortex of CNO-treated bigenic mouse pups . Second , we carried out whole-cell patch clamp recordings in current clamp mode from the somata of CA1 pyramidal neurons in acute hippocampal slices derived from bigenic mouse pups ( P7 ) following CNO bath application ( Figure 1K , L ) . Bath application of 1 µM CNO evoked robust spiking activity in CA1 pyramidal neurons ( Figure 1M ) . These two approaches confirmed that as anticipated , acute postnatal hM3Dq DREADD activation of forebrain excitatory neurons resulted in enhanced neuronal activation . Given our treatment paradigm involved chronic administration of CNO to mouse pups during the early postnatal window ( P2–P14 ) , we further sought to understand the effects of chronic CNO-mediated hM3Dq DREADD activation in CamKIIα-positive forebrain excitatory neurons on neuronal activity , at an interim time point in the midst of chronic CNO administration ( P7 ) . Bigenic mouse pups were orally administered CNO ( 1 mg/kg; PNCNO ) from P2 to P7 , and then assessed via western blotting analysis for cortical and hippocampal levels of the neuronal activity marker , p-ERK ( Figure 1—figure supplement 2A ) , or through electrophysiological measurements of spontaneous excitatory/inhibitory currents in the hippocampus ( Figure 1N; Figure 1—figure supplements 3A and 4A ) . Electrophysiological recordings following this treatment paradigm were carried out in aCSF in the absence of CNO in the bath . We detected a significant increase in p-ERK/ERK levels in the hippocampus ( Figure 1—figure supplement 2B , C , F1 , 6 = 1 . 557 , p=0 . 04 ) and the cortex ( Figure 1—figure supplement 2B , D , F1 , 6 = 1 . 069 , p=0 . 0001 ) of PNCNO-treated mouse pups . Whole-cell patch clamp recording to measure sPSCs and intrinsic membrane properties in CA1 pyramidal neurons from acute hippocampal slices revealed a significant difference in sPSC amplitude in the PNCNO-treatment group , with a small but significant decrease in low amplitude events ( <100 pA; Figure 1—figure supplement 3B , C , p<0 . 0001 ) , accompanied by a significant increase in large-amplitude events characterized by the presence of a long-tail in sPSC amplitude event distribution ( Figure 1—figure supplements 3C , 4B and C ) . We also observed a significant reduction in the cumulative probability of sPSC interevent intervals in CA1 pyramidal neurons from the PNCNO- treatment group ( Figure 1—figure supplement 3D , p<0 . 0001 ) . CA1 pyramidal neurons in PNCNO-treated hippocampal slices displayed large network activity , characterized by compound negative peaks ( Figure 1—figure supplement 4B ) . Out of the total number of events analyzed , the frequency of events greater than 100 pA were almost double in PNCNO-treated CA1 neurons ( 4 . 85% ) as compared to controls ( 2 . 57% ) . We also noted a small fraction of events ( 0 . 8% ) with amplitudes greater than 250 pA in CA1 pyramidal neurons from PNCNO-treated mouse pups , which were not detected in vehicle-treated controls ( Figure 1—figure supplement 4D ) . In order to understand the influence of CNO-mediated hM3Dq DREADD activation of CamKIIα-positive excitatory neurons during the postnatal window on intrinsic excitability , we plotted an input-output curve by injecting increasing step currents and measured the number of action potentials ( Figure 1—figure supplement 3E ) . We observed no change in the number of action potentials generated in CA1 pyramidal neurons of PNCNO-treated mouse pups ( Figure 1—figure supplement 3F ) . Measurements of key intrinsic membrane properties revealed no change in input resistance ( RN ) , membrane time constant ( τ ) , sag voltage , and accommodation index in CA1 pyramidal neurons of PNCNO-treated mouse pups ( Table 1 ) . We did note a trend toward a depolarizing shift in the resting membrane potential ( RMP ) in CA1 pyramidal neurons of the PNCNO-treatment group as compared to their vehicle-treated controls ( Table 1 , F1 , 13 = 1 . 862 , p=0 . 06 ) . We next sought to parcellate the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the postnatal window on excitatory and inhibitory neurotransmission . Whole-cell patch clamp analysis was carried out to measure sEPSCs and sIPSCs in CA1 pyramidal neurons in acute hippocampal slices derived from bigenic mouse pups treated with CNO ( 1 mg/kg ) or vehicle ( Figure 1N ) . We observed a significant difference in sEPSC amplitude in CA1 pyramidal neurons of PNCNO-treated mouse pups as compared to vehicle-treated controls , as revealed by a small but significant decrease in low amplitude events ( <100 pA ) , and a significant increase in large-amplitude events characterized by the presence of a long-tail in sEPSC amplitude event cumulative distribution ( Figure 1O , Q , p<0 . 0001 ) . CA1 pyramidal neurons in hippocampal slices from PNCNO-treated mouse pups displayed large sEPSC events characterized by compound negative peaks as compared to vehicle-treated controls ( Figure 1O; bottom traces ) . We also noted a significant decline in the cumulative probability of sEPSC interevent intervals in CA1 pyramidal neurons from the PNCNO-treatment group ( Figure 1R , p<0 . 0001 ) . Further , we observed a significant reduction in the cumulative probability of sIPSC amplitude ( Figure 1P , S , p<0 . 0001 ) and an increase in the cumulative probability of sIPSC interevent intervals ( Figure 1T , p<0 . 0001 ) in CA1 pyramidal neurons from PNCNO-treated mouse pups . Our findings demonstrate the selective expression of the hM3Dq DREADD in the forebrain of CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups , and indicate that acute CNO treatment during postnatal life increases neuronal activity in the hippocampus and cortex . Further , as the main treatment paradigm used in our study is based on chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the postnatal window , our experiments at an interim juncture during postnatal treatment reveal significant changes in both neuronal activity marker expression and electrophysiological measures . Collectively , our results demonstrate that chronic CNO-mediated hM3Dq DREADD activation during the postnatal window results in enhanced expression levels of neuronal activity markers , elevated spontaneous network activity , an increase in spontaneous excitatory currents , and a concomitant decrease in spontaneous inhibitory currents in CA1 pyramidal neurons of the PNCNO-treatment group . We next sought to assess the persistent behavioral consequences of perturbing neuronal activity of CamKIIα-positive forebrain excitatory neurons during the early postnatal window using chronic CNO-mediated hM3Dq DREADD activation . CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups were orally administered the DREADD ligand , CNO ( 1 mg/kg ) , or vehicle , once daily from P2 to to P14 ( Figure 2A ) . Postnatal treatment with CNO did not alter the body weight , measured across the period of postnatal treatment or in adulthood ( Figure 2—figure supplement 1A–C ) . Chronic CNO-mediated hM3Dq DREADD activation in the early postnatal window did not alter the normal trajectory of sensorimotor development , as indicated by no change in the ontogeny of reflex behaviors , namely surface righting and negative geotaxis , in PNCNO-treated mouse pups as compared to their vehicle-treated controls ( Figure 2—figure supplement 2A–C ) . We examined the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on long-lasting changes in anxiety- and despair-like behavior . We subjected bigenic adult mice with a history of PNCNO or vehicle treatment to a battery of behavioral tasks , commencing 3-months post cessation of PNCNO treatment . We performed the open field test ( OFT ) , elevated plus maze ( EPM ) test , and the light-dark ( LD ) avoidance test to assess anxiety-like behavior , followed by the forced swim test ( FST ) to assess despair-like behavior in PNCNO-treated adult bigenic male and female mice ( Figure 2A , Figure 2—figure supplement 3A ) . We noted a significant increase in anxiety-like behavior in adult male mice with a history of PNCNO treatment on the OFT ( Figure 2B ) . The PNCNO-treatment group showed a significant decrease in percent distance traveled in the center ( Figure 2C , F1 , 28 = 1 . 097 , p=0 . 03 ) , number of entries to the center ( Figure 2E , F1 , 28 = 1 . 272 , p=0 . 02 ) , and total distance traveled in the OFT arena ( Figure 2F , F1 , 28 = 1 . 23 , p=0 . 003 ) . The percent time spent in the center of the OFT arena was unchanged across treatment groups ( Figure 2D ) . We also noted an increase in anxiety-like behavior in the EPM ( Figure 2G ) in the PNCNO-treated adult male mice , with a significant decline in the number of entries to the open arms ( Figure 2J , F1 , 28 = 2 . 829 , p=0 . 02 ) and a trend toward a decrease in the percent time spent in the open arms ( Figure 2I , F1 , 28 = 1 . 977 , p=0 . 08 ) . The percent distance traveled in the open arms ( Figure 2H ) and the total distance traversed in the EPM ( Figure 2K ) were unchanged . Behavioral analysis on the LD avoidance test ( Figure 2L ) , revealed an anxiogenic effect of PNCNO treatment in bigenic adult male mice , with a significant decrease in the number of entries to the light box ( Figure 2M , F1 , 28 = 1 . 229 , p=0 . 04 ) and a trend toward a decrease in the time spent in the light box ( Figure 2N , F1 , 28 = 1 . 378 , p=0 . 07 ) . We then evaluated the influence of chronic CNO-mediated hM3Dq DREADD activation of forebrain excitatory neurons in the early postnatal window on despair-like behavior in adulthood using the FST ( Figure 2O ) . We observed increased despair-like behavior in CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice with a history of PNCNO treatment , as noted by a significant increase in the time spent immobile in the FST ( Figure 2P , F1 , 15 = 7 . 862 , p=0 . 03 , Welch’s correction ) . Taken together , these results indicate that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window results in long-lasting increases in anxiety- and despair-like behavior in adult male mice . Following this , we sought to ascertain whether chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window , evokes a similar anxiogenic and despair-like behavioral phenotype in adult female mice ( Figure 2—figure supplement 3A ) . Bigenic adult female mice with a history of PNCNO treatment exhibited enhanced anxiety-like behavior on the OFT and EPM tests . In the OFT we noted a significant decrease in percent distance traveled in the center ( Figure 2—figure supplement 3B , C , F1 , 20 = 1 . 438 , p=0 . 01 ) , and number of entries to the center ( Figure 2—figure supplement 3E , F1 , 20 = 1 . 158 , p=0 . 008 ) , with no change observed in other measures ( Figure 2—figure supplement 3D , F ) . In the EPM , bigenic adult female mice with a history of PNCNO treatment , showed a significant decrease in the percent distance traveled ( Figure 2—figure supplement 3G , H , F1 , 20 = 3 . 139 , p=0 . 04 ) and the percent time spent in the open arms ( Figure 2—figure supplement 3I , F1 , 20 = 2 . 31 , p=0 . 004 ) as compared to their vehicle-treated controls , with no difference observed on other measures ( Figure 2—figure supplement 3J , K ) . PNCNO-treated bigenic adult female mice did not show any change in anxiety-like behavior on the LD avoidance test ( Figure 2—figure supplement 3L–N ) . PNCNO-treated bigenic adult female mice did not show any change in despair-like behavior assessed on the FST ( Figure 2—figure supplement 3O , P ) . Taken together , these results indicate that chronic CNO-mediated hM3Dq DREADD activation of forebrain excitatory neurons during the early postnatal window results in long-lasting increases in both anxiety- and despair-like behavior in adult male mice , whereas it evokes a persistent increase in anxiety- , but not despair-like behavior , in adult female mice . A caveat to note is that our experiments with adult CamKIIα-tTA::TetO hM3Dq males and females with a history of PNCNO treatment were performed on distinct cohorts at different times . This prevented us from performing a two-way ANOVA analysis to examine sexually dimorphic behavioral effects of excitatory DREADD-mediated chemogenetic activation of forebrain excitatory neurons during postnatal life . Henceforth , all our studies to assess the behavioral , metabolic , molecular and electrophysiological influence of chronic CNO-mediated hM3Dq DREADD activation of forebrain excitatory neurons during the early postnatal window have been restricted to male mice . We next sought to determine the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on stereotypic behavior . We subjected CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice , with a history of PNCNO treatment to the marble burial test ( Figure 2—figure supplement 4A , B ) . We observed no change in stereotypic behavior on the marble burial test , with no difference in the number of marbles buried by the PNCNO or vehicle-treated bigenic adult male mice ( Figure 2—figure supplement 4A–C ) . Our observations indicate that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window does not influence repetitive behavior in CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice . Considering the evidence that CNO metabolites can produce off-target behavioral effects ( Gomez et al . , 2017; MacLaren et al . , 2016 ) , we designed two sets of control experiments which assessed the influence of postnatal CNO administration in genotype-control or background strain mouse pups , and the resultant effects on the programming of adult anxiety- and despair-like behavior . First , we administered CNO ( 1 mg/kg ) or vehicle to genotype-control mouse pups , single- positive for either CamKIIα-tTA or TetO-hM3Dq once daily from P2 to P14 ( Figure 2—figure supplement 5A ) . Following a three-month washout period post cessation of CNO treatment , we assayed these mice for anxiety and depressive-like behavior . We did not observe any difference in anxiety-like behavior in the OFT ( Figure 2—figure supplement 5B–E ) in the PNCNO-treated genotype-control cohort as compared to vehicle-treated controls . We did note a small , but significant decrease in total distance traveled in the OFT arena ( Figure 2—figure supplement 5F , F1 , 22 = 1 . 372 , p=0 . 03 ) in the PNCNO-treated genotype-control group . Behavioral analysis on the EPM indicated no change in anxiety-like behavior in adult genotype-control mice with a history of PNCNO treatment ( Figure 2—figure supplement 5G–K ) . In addition , we did not observe any change in anxiety-like behavior in the LD avoidance test ( Figure 2—figure supplement 5L–N ) as a consequence of PNCNO treatment in genotype-control mice . Despair-like behavior was also unchanged across treatment groups , indicating that CNO treatment in genotype-control mice during the postnatal window does not alter behavior on the FST ( Figure 2—figure supplement 5O , P ) . The second control experiment to rule out potential off-target effects of chronic postnatal CNO treatment was performed in the background strain ( C57BL/6J ) . C57BL/6J mouse pups received oral administration of CNO ( 1 mg/kg ) or vehicle once daily from P2 to P14 , followed by behavioral testing commencing 3 months post cessation of the CNO treatment regime ( Figure 2—figure supplement 6A ) . To assess anxiety-like behavior C57BL/6J adult male mice with a history of PNCNO treatment were tested on the OFT , EPM , and LD avoidance test . We did not observe any change in anxiety-like behavior in the OFT ( Figure 2—figure supplement 6B–F ) , EPM ( Figure 2—figure supplement 6G–K ) and the LD avoidance test ( Figure 2—figure supplement 6L–N ) in the PNCNO-treated C57BL/6J adult male mice as compared to their vehicle-treated controls . Despair-like behavior , as assessed by immobility time on the FST was also unchanged across treatment groups , indicating no effect of postnatal CNO treatment in the C57BL/6J background strain ( Figure 2—figure supplement 6O , P ) . Collectively , these control experiments indicate postnatal CNO administration does not evoke off-target effects that influence anxiety- and despair-like behavior . Chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons using the hM3Dq DREADD agonist compound 21 ( C21 ) during the early postnatal window results in a long-lasting increase in anxiety-like behavior in adult mice . We further addressed whether an alternate hM3Dq DREADD agonist compound 21 ( C21 ) when utilized to evoke chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during postnatal life also programs persistent changes in adult anxiety-like behavior . We orally administered C21 to CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups once daily from P2 to 14 . We then subjected bigenic adult male mice with a history of PNC21 or vehicle treatment to the open field test ( OFT ) , and elevated plus maze ( EPM ) test to assess effects on anxiety-like behavior ( Figure 2—figure supplement 7A ) . We noted a significant increase in anxiety-like behavior in adult PNC21 male mice on the OFT ( Figure 2B ) . The PNC21 treatment group showed a significant decrease in percent time spent in the center of the OFT arena ( Figure 2—figure supplement 7D , F1 , 28 = 1 . 252 , p=0 . 025 ) . The percent distance traveled in the center ( Figure 2—figure supplement 7C ) , number of entries to the center ( Figure 2—figure supplement 7E ) , and total distance traveled in the OFT arena ( Figure 2—figure supplement 7F ) were unaltered across treatment groups . We also noted an increase in anxiety-like behavior in the EPM ( Figure 2—figure supplement 7G ) in the PNC21-treated adult male mice , with a significant decrease in percent distance traveled in the open arms ( Figure 2—figure supplement 7H , F1 , 28 = 2 . 295 , p=0 . 003 ) and the percent time spent in the open arms ( Figure 2—figure supplement 7I , F1 , 28 = 1 . 991 , p=0 . 002 ) . The number of entries to the open arms ( Figure 2—figure supplement 7J ) and the total distance traversed in the EPM ( Figure 2—figure supplement 7K ) were not changed . These results indicate that chronic C21-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window results in long-lasting increases in anxiety-like behavior in adult male mice . Chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during the juvenile window or in adulthood does not evoke any long-lasting changes in anxiety- and despair-like behavior . Given we observed that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the early postnatal window can program persistent changes in anxiety- and despair-like behavior , we next sought to ascertain whether the temporal window in which this perturbation is performed is critical to the establishment of these long-lasting behavioral changes . To address this question , we used chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in two distinct temporal windows , namely juvenile life ( P28–40 ) and adulthood . The time duration and dose of CNO treatment was maintained constant across the postnatal , juvenile , and adult treatment paradigms . CamKIIα-tTA::TetO-hM3Dq bigenic juvenile male mice received CNO ( 1 mg/kg; JCNO ) via oral administration once daily from P28-P40 ( Figure 3A; Figure 3—figure supplement 1A ) . Following a washout period , we subjected bigenic adult male mice with a history of JCNO treatment to behavioral tests for anxiety- and despair-like behavior . We observed no change in anxiety-like behavior in JCNO-treated mice in the OFT ( Figure 3B ) , with no difference noted for the percent distance traveled in the center ( Figure 3C ) , percent time spent in the center ( Figure 3D ) , number of entries to the center ( Figure 3—figure supplement 1B ) and the total distance traversed in the OFT arena ( Figure 3—figure supplement 1C ) . Behavioral testing on the EPM ( Figure 3E ) revealed no influence of JCNO treatment on anxiety-like behavior , with no difference noted for the percent distance traveled in the open arms ( Figure 3F ) , percent time spent in the open arms ( Figure 3G ) , number of entries to the open arms ( Figure 3—figure supplement 1D ) , and total distance traveled in the EPM arena ( Figure 3—figure supplement 1E ) . Further , we did not observe any difference in the number entries to the light box ( Figure 3H ) and the time spent in the light box ( Figure 3I ) in JCNO-treated mice on the LD avoidance test . JCNO and vehicle-treated bigenic male mice did not differ on despair-like behavioral measures on the FST , with no significant change in immobility time ( Figure 3J ) . These results indicate that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the juvenile window does not program any persistent changes in anxiety- and despair-like behavior . In order to test the effects of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in adulthood , we treated adult CamKIIα-tTA::TetO-hM3Dq bigenic male mice ( 3–4 months of age ) with CNO ( 1 mg/kg; i . p . ; ACNO ) or vehicle once daily for thirteen days ( Figure 3K; Figure 3—figure supplement 1F ) , following which we performed behavioral assays . Behavioral testing was carried out at two time windows of the treatment regime . The first round of behavioral testing was conducted during and soon after the cessation of CNO treatment to assess any immediate consequences on anxiety-like behavior ( Figure 3—figure supplement 2A ) . The second phase of behavioral testing commenced after a three-month washout period to assess long-lasting consequences of chronic CNO-mediated hM3Dq DREADD activation of forebrain excitatory neurons in adulthood on anxiety- and despair-like behavior ( Figure 3K; Figure 3—figure supplement 1F ) . The first phase of behavioral testing involved assays for anxiety-like behavior on the OFT , EPM and LD avoidance test during and soon after the cessation of CNO treatment . OFT was performed on Day 8 while the chronic CNO treatment was ongoing , and the EPM and LD avoidance test were carried out on Day 15 and Day 22 , respectively , soon after cessation of CNO treatment ( Figure 3—figure supplement 2A ) . No change in anxiety-like behavior was observed on the OFT ( Figure 3—figure supplement 2B–F ) , EPM ( Figure 3—figure supplement 2G–K ) , and LD avoidance test ( Figure 3—figure supplement 2L–N ) in the ACNO treatment group , during and soon after the cessation of CNO treatment . In the adult CNO treatment regime we did not subject mice to behavioral testing for despair-like behavior on the FST immediately after treatment , as swimming can serve as a strong stressor ( Can , 2011; Yankelevitch-Yahav et al . , 2015 ) , and we intended to assess for anxiety- and despair-like behavior following a three-month washout period in the same cohort . These findings indicate that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in adulthood does not evoke any change in anxiety-like behavior , during or in the short duration after the cessation of CNO treatment . The second phase of behavioral testing involved assessing for the long-lasting consequences of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in adulthood , with behavioral tests for anxiety- and despair-like behavior commencing 3 months post cessation of CNO treatment ( Figure 3K; Figure 3—figure supplement 1F ) . We did not observe any change in anxiety-like behavior on the OFT in the ACNO-treated bigenic adult male mice ( Figure 3L ) , with no change in the percent distance traveled in center ( Figure 3M ) , percent time spent in the center ( Figure 3N ) , number of entries to the center ( Figure 3—figure supplement 1G ) , and the total distance traveled in the OFT arena ( Figure 3—figure supplement 1H ) . Behavioral analysis of the EPM ( Figure 3O ) , indicated that the ACNO-treated group did not differ in the percent distance traveled in the open arms ( Figure 3P ) , percent time spent in the open arms ( Figure 3Q ) , number of entries to the open arms ( Figure 3—figure supplement 1I ) , and the total distance traveled in the EPM ( Figure 3—figure supplement 1J ) . Similarly , we did not observe any change in anxiety-like behavior in the LD avoidance test with no difference noted in the number of entries to the light box ( Figure 3R ) and the time spent in the light box ( Figure 3S ) across treatment groups . Further , we subjected ACNO and vehicle-treated bigenic adult male mice to the FST to assess for despair-like behavior , and noted no difference in the immobility time ( Figure 3T ) . These observations reveal that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in adulthood does not result in any long-lasting consequences in anxiety- and despair-like behavior . These observations collectively underscore the critical importance of the postnatal window in the long-term programming of anxiety- and despair-like behavior , as chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons is sufficient to establish persistent changes in these behaviors only when administered during the postnatal window , with no such effect noted when the same chemogenetic activation is performed either in juvenile life or in adulthood . Given prior evidence that a dysregulation of cortical excitation/inhibition balance during postnatal life contributes to the establishment of endophenotypes linked to schizophrenia ( Rosen et al . , 2015 ) , we next sought to examine whether chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the early postnatal window influenced sensorimotor gating behavior in adulthood ( Figure 4A ) . In order to assess for sensorimotor gating , we subjected CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice with a history of PNCNO treatment to the prepulse inhibition ( PPI ) test ( Figure 4B ) . We did not observe any significant alterations in the basal startle response across treatment groups ( Figure 4C ) . Strikingly , we noticed a significant PPI deficit at all prepulse tones , with a decline in percent PPI to tone ( 120 dB ) following a prepulse of + 4 dB ( 69 dB; Figure 4D , F1 , 19 = 2 . 063 , p=0 . 024 ) , + 8 dB ( 73 dB; Figure 4D , F1 , 21 = 1 . 136 , p=0 . 017 ) , and + 16 dB ( 81 dB; Figure 4D , F1 , 22 = 2 . 924 , p=0 . 041 , n = 12/ group ) above the background noise in PNCNO-treated bigenic adult male mice . These findings indicate that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window results in long-lasting deficits in sensorimotor gating . Next , we attempted to understand whether chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the juvenile time window or in adulthood can exert similar long-term effects on sensorimotor gating behavior . CamKIIα-tTA::TetO-hM3Dq bigenic male mice ( Juvenile group: P28-40; Adult group: 3–4 months of age ) were administered CNO ( 1 mg/kg ) or vehicle treatment once daily for thirteen days ( Figure 4—figure supplements 1A and 2A ) . Behavioral testing for sensorimotor gating on the PPI test ( Figure 4—figure supplements 1B and 2B ) commenced post a three-month washout period for both the JCNO and ACNO experiments . We did not observe any significant change in the basal startle response ( Figure 4—figure supplements 1C and 2C ) in either the JCNO or ACNO treatment groups as compared to their respective vehicle-treated controls . Further , we noted no significant difference in percent PPI in the JCNO or ACNO bigenic adult male mice to a 120 dB tone at all prepulse tones above the background noise ( Figure 4—figure supplements 1D and 2D ) . These behavioral experiments reveal that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons programs long-lasting changes in sensorimotor gating when the hM3Dq DREADD activation is performed in the postnatal window , but not in either the juvenile time window or in adulthood . Chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window results in long-lasting alterations in neuronal metabolic rate in the hippocampus and cortex . Dysregulation of glutamatergic and GABAergic neurotransmission within forebrain neurocircuitry , including the hippocampus and several cortical regions , is thought to causally contribute to the pathogenesis of several mood-related disorders including anxiety , major depression , and schizophrenia ( Sanacora et al . , 2012; Kendell et al . , 2005; Choudary et al . , 2005; Duman et al . , 2019 ) . In particular , metabolic dysfunction of glutamate and GABA systems are considered to be important endophenotypes of mood-related disorders ( Veeraiah et al . , 2014; Sekar et al . , 2019; Godfrey et al . , 2018; Hasler and Northoff , 2011 ) . Hence , we next sought to investigate the effects of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on the metabolic activity in glutamatergic and GABAergic neurons in the hippocampus and cortex in adulthood . We orally administered CNO ( PNCNO; 1 mg/kg ) or vehicle to CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups once daily from P2 to P14 , and performed metabolic analysis in adulthood using a trace approach by infusing [1 , 6-13C2]glucose ( Figure 5A; Figure 5—figure supplements 1 , 2A and 3A ) . [1 , 6-13C2]Glucose is transported and metabolized in the brain to PyruvateC3 via glycolysis . The pyruvateC3 is subsequently oxidized through the TCA cycles of glutamatergic and GABAergic neurons , and astrocytes to produce 13C labeled metabolites ( Figure 5—figure supplement 1 ) . The 13C labeling of brain metabolites was measured in 1H-[13C]-NMR spectra of brain tissue extracts . The metabolic rate of glucose oxidation in excitatory and inhibitory neurons was determined by using the three-compartment metabolic rate model ( Patel et al . , 2005; Tiwari et al . , 2013; Saba et al . , 2017 ) . First , we measured the concentration of different metabolites in the hippocampus and cortex from the non-edited 1H-[13C]-NMR spectrum using [2-13C]glycine as the reference ( Figure 5—figure supplement 2B ) . We did not observe any significant difference in the levels of glutamate , GABA , glutamine , aspartate , N-acetylaspartate , lactate , inositol , taurine , choline and creatine in the hippocampus and cerebral cortex of PNCNO-treated bigenic adult male mice as compared to their vehicle-treated controls ( Table 2 ) . We observed a significant decline in the levels of alanine in the hippocampus ( F1 , 12 = 2 . 012 , p=0 . 047 ) , but not in the cortex , of bigenic adult male mice with a history of PNCNO treatment ( Table 2 ) . Further , we measured the13C labeling of amino acids from [1 , 6– ( Ogle et al . , 2015 ) C2]glucose TCA from the 13C edited spectrum ( 13C only ) ( Figure 5—figure supplement 2C ) . The metabolic rate glucose oxidation in excitatory and inhibitory neurons from the hippocampus and cortex was determined from the 13C label trapped into different amino acids ( Patel et al . , 2005; Mishra et al . , 2018 ) . Bigenic adult male mice with a history of PNCNO treatment exhibited an elevated rate of hippocampal glutamate and GABA synthesis from [1 , 6 - 13C2]glucose as revealed by a significant increase in the concentration of 13C labeled GluC4 ( Figure 5B , F1 , 12 = 1 . 335 , p=0 . 05 ) , GABAC2 ( Figure 5E , F1 , 12 = 2 . 4 , p=0 . 045 ) and the metabolic rate of glucose oxidation in GABAergic neurons ( Figure 5G , F1 , 12 = 1 . 105 , p=0 . 001 ) in the hippocampus of PNCNO-treated adult mice . We did not note any difference in the concentration of 13C labeled GluC3 ( Figure 5C ) , GABAC4 ( Figure 5F ) , and the metabolic rate of glucose oxidation in glutamatergic neurons ( Figure 5D ) in the hippocampus . We also observed an overall increase in total neuronal metabolic rate of glucose oxidation in the hippocampus of PNCNO-treated mice as compared to vehicle-treated controls ( Figure 5H , F1 , 12 = 1 . 277 , p=0 . 05 ) . PNCNO-treated mice also showed a trend toward an increase in the concentration of 13C labeled AspC3 ( Figure 5—figure supplement 3C , F1 , 12 = 1 . 233 , p=0 . 07 ) with no change noted in the concentration of 13C labeled GlnC4 ( Figure 5—figure supplement 3B ) . In the cortex , PNCNO-treated adult mice significantly higher levels of 13C-labeled metabolites GluC4 ( Figure 5I , F1 , 12 = 2 . 609 , p=0 . 009 ) , GluC3 ( Figure 5J , F1 , 12 = 1 . 833 , p=0 . 026 ) from [1 , 6-13C2]glucose , and the metabolic rate of glucose oxidation in glutamatergic neurons of the cortex ( Figure 5K , F1 , 12 = 2 . 167 , p=0 . 007 ) . We observed a trend toward an increase in levels of 13C-labeled metabolite GABAC2 ( Figure 5L , F1 , 12 = 2 . 521 , p=0 . 09 ) and the metabolic rate of glucose oxidation in GABAergic neurons of the cortex ( Figure 5N , F1 , 12 = 1 . 523 , p=0 . 067 ) , with no change noted in 13C-labeled GABAC4 ( Figure 5M ) in the PNCNO-treatment group . There was a significant increase in the overall neuronal metabolic rate of glucose oxidation in the cortex ( Figure 5O , F1 , 12 = 1 . 884 , p=0 . 012 ) of the PNCNO-treatment group . Taken together , our data suggest a long-lasting increase in metabolic rate of neuronal glucose oxidation within the hippocampus and cortex following chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window . This suggests a persistent alteration in glutamatergic and GABAergic neurotransmission in the forebrain in adulthood as a consequence of postnatal chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons . To examine the influence of hM3Dq DREADD activation of forebrain excitatory neurons on neuronal activity , we focused on the hippocampus for the subsequent experiments . Chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window results in a long-lasting reduction in neuronal activity-related gene expression , and in c-Fos immunopositive cell numbers , in the adult hippocampus . In order to investigate the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on hippocampal neuronal activity , we adopted two complementary approaches . First , we performed qPCR analysis for neuronal activity and plasticity-related gene expression in hippocampi derived from PNCNO and vehicle-treated CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice ( Figure 6A; Gatto and Broadie , 2010; Loebrich and Nedivi , 2009 ) . We observed a significant decline in the expression of several neuronal activity-regulated genes namely Fos , Erk1 , Npas4 , Staufen1 , Staufen2 , Nrxn1 , Gphn , Shank , Psd95 , Fmrp , and Synapsin1b in the hippocampi derived from bigenic adult male mice with a history of PNCNO treatment ( Figure 6B ) . We did not observe any alteration in expression levels of Nr4a1 , Junb , Nlgn1 , Nlgn2 , Mecp2 , and Mef2c across treatment groups ( Figure 6B ) . The second approach we took was to perform cell counting analysis of c-Fos immunoreactive cell numbers within the hippocampal subfields namely , CA1 , CA3 , dentate gyrus ( DG ) , and the hilus of PNCNO and vehicle administered bigenic adult male mice ( Figure 6A ) . We observed a significant decline in c-Fos immunopositive cell number within the CA1 ( Figure 6C , F1 , 20 = 3 . 154 , p=0 . 004 ) and CA3 ( Figure 6D , F1 , 20 = 1 . 67 , p=0 . 012 ) subfields of the hippocampus in the PNCNO-treatment group . We did not note any change in c-Fos immunopositive cell numbers in the DG subfield ( Figure 6E ) , and in the hilus ( Figure 6F ) in the PNCNO group . Collectively , our findings provide evidence that chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window can program a persistent decline in the expression of several neuronal activity and plasticity-associated genes within the hippocampus , also accompanied by a reduction in the number of c-Fos immunopositive cells suggestive of an alteration in hippocampal neuronal activity . Chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window alters excitatory and inhibitory spontaneous currents in the hippocampi of adult male mice . Given that our gene expression profiling and c-Fos counting analyses pointed toward a possible change in hippocampal neuronal activity in adulthood as a consequence of postnatal hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons , we next performed electrophysiological studies to assess effects on hippocampal neurotransmission . Whole-cell patch clamp analysis was carried out in the somata of CA1 pyramidal neurons in acute hippocampal slices derived from PNCNO or vehicle-treated CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice ( Figure 7A; Figure 7—figure supplements 1A and 2A ) . In order to determine the long-lasting influence of chronic hM3Dq DREADD activation of forebrain excitatory neurons during the postnatal window on intrinsic excitability in adulthood , we plotted an input-output curve by injecting increasing step currents and measured the number of action potentials ( Figure 7—figure supplement 1B ) . No change was noted in the input-output curves obtained from CA1 pyramidal neurons in acute hippocampal slices derived from bigenic adult male mice with as history of PNCNO treatment ( Figure 7—figure supplement 1C ) . We then measured key intrinsic membrane properties using a hyperpolarizing current step of −100 pA for 500 ms . We did not observe any change in the resting membrane potential ( RMP ) , input resistance ( RN ) , membrane time constant ( τ ) , sag voltage , and accommodation index in CA1 pyramidal neurons of the PNCNO-treatment group ( Table 3 ) . Measurement of sPSCs in CA1 pyramidal neurons in acute hippocampal slices derived from PNCNO-treated bigenic adult male mice revealed a significant increase in the cumulative probability of sPSC amplitude characterized by a long-tail in sPSC amplitude event distribution ( Figure 7—figure supplement 1D , E; Figure 7—figure supplement 2B , C; p<0 . 0001 ) , accompanied by a significant reduction in the cumulative probability of sPSC interevent intervals ( Figure 7—figure supplement 1F; p=0 . 0009 ) . Out of the total number of sPSCs analyzed , we noted events with an amplitude greater than 100 pA occurred with a significantly greater frequency in CA1 neurons from the PNCNO-treatment group ( 2 . 2% ) , as compared to controls ( 0 . 32% ) . Further , we also observed a small fraction of events ( 0 . 66% ) with amplitudes greater than 250 pA in CA1 pyramidal neurons from the PNCNO-treated cohort , that were not detected in vehicle-treated controls ( Figure 7—figure supplement 2D ) . We next sought to distinguish the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on hippocampal excitatory and inhibitory neurotransmission in adulthood . We performed whole-cell patch clamp analysis to measure sEPSCs and mEPSCs in CA1 pyramidal neurons in acute hippocampal slices derived from bigenic adult male mice with a history of PNCNO treatment ( Figure 7A ) . We noted a significantly altered sEPSC amplitude distribution in CA1 pyramidal neurons of PNCNO-treated adult male mice ( Figure 7B , C; p<0 . 0001 ) , with a small but significant increase in low amplitude events ( <30 pA ) , and a significant decline in large-amplitude events . Further , we observed a significant decrease in cumulative probability of sEPSC interevent intervals in CA1 pyramidal neurons from the PNCNO-treatment group ( Figure 7D; p<0 . 0001 ) . We observed a small , but significantly enhanced cumulative probability of mEPSC amplitude ( Figure 7E , F; p<0 . 0001 ) , with no change observed in the cumulative probability of mEPSC interevent intervals ( Figure 7G ) in bigenic adult male mice with a history of PNCNO treatment . To assess effects on hippocampal inhibitory neurotransmission , we measured sIPSCs and mIPSCs in CA1 pyramidal neurons in acute hippocampal slices . We noted a significant increase in the cumulative probability of sIPSC amplitude ( Figure 7H , I; p<0 . 0001 ) , concomitant with a significant reduction in the cumulative probability of sIPSC interevent intervals ( Figure 7J; p<0 . 0001 ) in bigenic adult male mice with a history of PNCNO treatment . Further , we noted a significant reduction in the cumulative probability of mIPSC amplitude ( Figure 7K , L; p<0 . 0001 ) , with no change noted in the cumulative probability of mIPSC interevent intervals ( Figure 7M ) in CA1 neurons of PNCNO-treated adult male mice . Our electrophysiological studies performed on CA1 neurons in acute hippocampal slices of adult male mice with a history of chemogenetic activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window demonstrates the programming of persistent increases in spontaneous network activity , accompanied by significantly altered hippocampal excitatory and inhibitory neurotransmission .
The major finding of our study is that chronic chemogenetic hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons in the first two weeks of postnatal life is sufficient to program the emergence of enhanced anxiety- , despair- and schizophrenia-like behavior in adult male mice . In contrast , chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons in either the juvenile or adult temporal window did not result in any persistent changes in mood-related behavior . Chronic chemogenetic activation of forebrain excitatory neurons in postnatal life also resulted in persistent changes in glutamate/GABA metabolism in the hippocampus and cortex , accompanied by a long-lasting decline in hippocampal activity and plasticity-associated gene expression , and altered hippocampal spontaneous excitatory and inhibitory currents . Given prior reports that several models of early adversity exhibit enhanced signaling via Gq-coupled neurotransmitter receptors in the forebrain ( Benekareddy et al . , 2010; Benekareddy et al . , 2011; Malkova et al . , 2014; Proulx et al . , 2014; Sarkar et al . , 2014b; Moreno et al . , 2011 ) , our findings posit that enhanced Gq-signaling-mediated activation of forebrain excitatory neurons in the critical temporal window of postnatal life may serve as a putative mechanism to program enhanced risk for adult psychopathology , a hallmark feature of models of early adversity . Results from multiple rodent models including maternal separation ( MS ) , maternal neglect , and postnatal fluoxetine ( PNFlx ) indicate that the first two weeks of postnatal life are critical to the long-lasting programming of anxiety- and despair-like behavior ( Rebello et al . , 2014; Suri and Vaidya , 2015; Roque et al . , 2014; Freund et al . , 2013 ) . Evidence from several of these rodent models suggests enhanced functionality of Gq-coupled neurotransmitter receptors in the forebrain ( Benekareddy et al . , 2010; Benekareddy et al . , 2011; Malkova et al . , 2014; Proulx et al . , 2014; Sarkar et al . , 2014b; Moreno et al . , 2011 ) . Furthermore , pharmacological studies indicate that stimulation of the Gq-coupled 5-HT2A receptor during the early postnatal window can program persistent mood-related behavioral changes ( Sarkar et al . , 2014b ) , and that 5-HT2A receptor blockade overlapping with MS or PNFLx can prevent the emergence of adult anxiety- and despair-like behavior ( Benekareddy et al . , 2011; Sarkar et al . , 2014b ) . Our study directly tests the role of enhanced Gq-signaling-mediated activation of forebrain excitatory neurons in the early postnatal window in programming mood-related behavioral changes , and demonstrates that this perturbation when performed in the critical postnatal window , but not in juvenile or adult life , is sufficient to program the emergence of anxiety- , despair- and schizophrenia-like behaviors in adulthood . Adult male mice with a history PNCNO treatment showed both enhanced anxiety- and despair-like behavior , whereas adult female mice exhibited enhanced anxiety- , but not despair-like behavioral changes . Sexually dimorphic effects of early adversity have been previously reported , with females suggested to be resistant to some of the behavioral consequences of early adversity , in particular the programming of despair-like behavioral changes ( Leussis et al . , 2012; Roman et al . , 2004; Dimatelis et al . , 2016; Lundberg et al . , 2017; Desgent et al . , 2012; de Melo et al . , 2018 ) . Our results raise the possibility of sexually dimorphic behavioral consequences of early postnatal chemogenetic activation of forebrain excitatory neurons . A caveat of our study is that the neurometabolic , electrophysiological and molecular experiments were performed only in adult male mice , with a limited battery of behavioral analysis carried out in adult females , in part due to the large numbers of bigenic mice required to be maintained for these experiments . This precluded the possibility of careful analysis of sexual dimorphism in the consequences of enhanced Gq- signaling-driven within forebrain excitatory neurons in postnatal life , which will require detailed further experimentation . Given that thus far very few studies have used chemogenetic strategies during developmental time windows ( Teissier et al . , 2020; Wong et al . , 2018 ) , we characterized the consequences of hM3Dq DREADD activation of forebrain excitatory neurons in the postnatal window using both electrophysiological and biochemical approaches . Our observation of hM3Dq DREADD-mediated induction of robust spiking activity in postnatal slices parallels observations made in adulthood ( Alexander et al . , 2009; Pati et al . , 2019 ) . Chronic chemogenetic hM3Dq DREADD activation during postnatal life enhanced neuronal activity marker expression in the hippocampus and cortex , as well as increased network activity , enhanced spontaneous excitatory events , and reduced spontaneous inhibitory events in CA1 pyramidal neurons in PNCNO-treated mouse pups . Our studies also indicated that chronic chemogenetic hM3Dq DREADD activation of forebrain excitatory neurons during postnatal life does not impact normal physical growth , developmental milestones such as eye opening or ontogeny of reflex development . We also addressed the possibility that off-target effects of CNO ( Gomez et al . , 2017; MacLaren et al . , 2016 ) may impact our interpretations by extensively addressing effects of postnatal CNO treatment to genotype-control or background strain mouse pups , and noted no change in the emergence of anxiety- or despair-like behaviors in adulthood . These controls are particularly relevant given that very few reports have used chemogenetic perturbations during these early postnatal windows ( Teissier et al . , 2020; Wong et al . , 2018 ) . Our results do not allow us to completely rule out potential off-target effects of PNCNO treatment on locomotion in the OFT in adulthood . In this regard , our studies using an alternate DREADD agonist C21 to chemogenetically activate forebrain excitatory neurons during the postnatal window resulted in robust increases in anxiety-like behavior in adulthood , with no effects noted on total locomotion . This highlights the importance of using multiple DREADD ligands , especially when considering potential off-target effects of specific DREADD agonists on behavioral tasks . We next addressed whether chronic chemogenetic activation of CamKIIα-positive forebrain excitatory neurons in postnatal life recapitulates the effects of early adversity in programming changes in schizophrenia-like behavior ( Girardi et al . , 2014; Ellenbroek et al . , 1998 ) , and repetitive behavior ( Malkova et al . , 2012 ) . We noted a significant impairment in sensorimotor gating indicated by PPI deficits , but no change in stereotypic behavior , in adult mice with a history of PNCNO treatment . Deficient PPI is considered to be a behavioral deficit associated with schizophrenia-like behavior in both genetic or environmental perturbation based animal models ( Ellenbroek et al . , 1998; Khan and Powell , 2018; Mena et al . , 2016; Geyer et al . , 2001; Belforte et al . , 2010 ) . Preclinical genetic models targeting signaling pathways downstream to Gq ( PLC-β1-/- mice ) exhibit enhanced schizophrenia-like behavior ( McOmish et al . , 2008 ) . Further , loss of function of the Gq-coupled mGluR5 receptor in parvalbumin-positive interneurons increased both compulsive behavior and aberrant sensorimotor gating ( Barnes et al . , 2015 ) . It is important to note that PPI deficits are also common across various other neuropsychiatric conditions , in addition to schizophrenia ( Powell et al . , 2012; Nestler and Hyman , 2010 ) . Several reports indicate that early adversity during the perinatal window results in PPI impairments in adulthood ( Ellenbroek et al . , 1998; Ko et al . , 2014; Smith et al . , 2007; Fabricius et al . , 2008 ) . However , both the intensity and timing of the early stressor could program differing outcomes on PPI ( Ellenbroek et al . , 1998; Fabricius et al . , 2008; Ellenbroek and Cools , 2002 ) . For example , severe maternal deprivation evokes robust PPI deficits , whereas short duration maternal separation has no effect on PPI ( Ellenbroek et al . , 1998; Ellenbroek and Cools , 2002 ) . In this regard , our findings that chemogenetic activation of forebrain excitatory neurons produces an entire spectrum of mood-related behavioral changes , namely enhanced anxiety- , despair- and schizophrenia-like behaviors is suggestive of behavioral endophenotypes noted with the more severe of early stress models ( Nestler and Hyman , 2010; Bolton et al . , 2017; Walker et al . , 2017 ) . Our results that the timing of the chronic chemogenetic activation of forebrain excitatory neurons is central to determining consequent changes in mood-related behavior underscores the key importance of ‘critical’ periods for programming emotionality ( Bock et al . , 2014; Leonardo and Hen , 2008 ) . We observed no change in anxiety- , despair- and schizophrenia-like behaviors in either the juvenile or adult chronic CNO paradigms . Our treatment involved administration of the DREADD agonist CNO orally to pups/juveniles , and intraperitoneally to adult mice . Although we cannot assume that the effects observed across lifespan involved equivalent pharmacodynamics of CNO , it is noteworthy that there is no effect on anxiety-like behavior following chronic administration of CNO in both juvenile and adult CamKIIα-hM3Dq bigenic mice . This suggests that the chemogenetic activation of Gq signaling in forebrain excitatory neurons needs to be performed in the postnatal window ( P2–14 ) to program persistent mood-related behavioral changes . While our studies do not allow us to parcellate out the exact duration of this critical window , it likely encompasses the first two weeks of life . In this regard , both the MS and PNFLx models have critical periods spanning from P2 to 14 and P2–11 , respectively ( Rebello et al . , 2014; Roque et al . , 2014; Freund et al . , 2013 ) . These temporal windows overlap with distinct critical periods including the stress hyporesponsive period ( Schmidt et al . , 2003; Levine , 2001; Suchecki , 2018 ) , a neurodevelopmental window for the refinement of multiple cortical circuits ( Hensch , 2004; Hensch , 2005 ) , including the maturation of serotonergic afferents to the cortex ( Teissier et al . , 2017; Vitalis and Verney , 2017 ) , and the tuning of excitation-inhibition balance across cortical microcircuits ( Sohal and Rubenstein , 2019; Tatti et al . , 2017; Xue et al . , 2014 ) . The first two weeks of postnatal life also constitutes a window in which apoptotic cell death plays a key role in the shaping of neocortical microcircuitry ( Wong and Marín , 2019 ) . Activity within cortical pyramidal neurons can directly shape interneuron survival ( Wong et al . , 2018 ) , thus influencing the manner in which the optimal balance between excitatory and inhibitory neurons in cortical microcircuits is established . A recent report indicates that hM3Dq DREADD activation of the medial prefrontal cortex during postnatal life can abrogate the influence of maternal separation on oligodendrogenesis and despair-like behavior ( Teissier et al . , 2020 ) . However , the use of a pan-neuronal human synapsin promoter to drive the hM3Dq DREADD ( Teissier et al . , 2020 ) , and the absence of a non-maternal separation cohort makes it difficult to directly compare with our results . Our observations support the view that chemogenetic activation of forebrain excitatory neurons from P2 to 14 could impinge on several key neurodevelopmental processes , thus establishing a substrate for the emergence of perturbed mood-related behaviors in adulthood . Associated with the long-lasting behavioral changes programmed by chronic DREADD activation of CamKIIα-positive forebrain excitatory neurons , we noted persistent dysregulation of glutamate and GABA neurotransmitter metabolism , a decline in the expression of neuronal activity- and plasticity-related markers , as well as alterations in hippocampal spontaneous excitatory and inhibitory currents . The dysregulation of both glutamate and GABA systems is amongst the key factors in the pathophysiology of several psychiatric disorders including anxiety , depression , and schizophrenia ( Sanacora et al . , 2012; Kendell et al . , 2005; Duman et al . , 2019; Brambilla et al . , 2003; Bergink et al . , 2004 ) . Neuroimaging studies on human subjects with mood disorders and schizophrenia demonstrate altered volume and resting-state functional activity in several forebrain regions , including hippocampus , sensory and frontal cortices ( Koike et al . , 2013; Wolf et al . , 2011; Kühn and Gallinat , 2013 ) . A major endophenotype that reflects persistent alterations in neuronal activity in mood-related disorders is the levels and neurometabolic activity of glutamate and GABA , the major excitatory and inhibitory neurotransmitters , respectively ( Sanacora et al . , 2012; Kendell et al . , 2005; Duman et al . , 2019; Veeraiah et al . , 2014; Sekar et al . , 2019; Godfrey et al . , 2018; Hasler and Northoff , 2011 ) . Although 1H-MRS has been widely used to examine the levels of these neurotransmitters in both human patients and rodents ( Zieminska et al . , 2018; Dyke et al . , 2017 ) , there has been a scarcity of studies to investigate neurometabolic activity , which represent a functional readout of metabolic dynamics in these neurocircuits ( Patel et al . , 2004; de Graaf et al . , 2003 ) . We employed 1H-[13C]-NMR spectroscopy to measure the metabolic rate of excitatory and inhibitory neurons in conjunction with infusion of [1 , 6-13C2]glucose ( Patel et al . , 2005; Tiwari et al . , 2013; Patel et al . , 2004; de Graaf et al . , 2003 ) . The glutamate hypothesis of mood disorders is based on observations of elevated glutamate levels , associated with changes in glutamate receptors , biosynthetic and regulatory pathways both in human patients and rodent models of anxiety/despair-like behavior and schizophrenia ( Sanacora et al . , 2012; Duman et al . , 2019 ) . Consistent with this hypothesis , we observed an increase in glucose oxidation in the TCA cycle of glutamatergic neurons in the hippocampus and cortex of adult mice with a history of PNCNO treatment . The rate of neuronal glucose oxidation and neurotransmitter cycle are stoichiometrically coupled during the entire range of brain activity ( Patel et al . , 2005; Hyder et al . , 2006; Sibson et al . , 1998 ) . Hence , increased neuronal glucose oxidation in PNCNO-treated mice suggests enhanced excitatory and inhibitory neurotransmission . Furthermore , we observed an increase in metabolic rate of hippocampal GABAergic neurons , and a trend toward an increase in this measure in the cortex . It is important to note that though we see an increase in metabolic rate of both glutamatergic and GABAergic neurons , a history of PNCNO treatment does not influence the neurotransmitter pool of glutamate or GABA either in the hippocampus or cortex . While our observations for enhanced glutamatergic metabolic rate in PNCNO-treated animals are consistent with clinical and preclinical reports of enhanced glutamate function in mood disorders ( Sanacora et al . , 2012; Duman et al . , 2019 ) , our observations with GABA differ from the reports of reduced GABA levels observed in human subjects and several adult-onset stress based rodent models of mood-related behavioral changes ( Duman et al . , 2019; Pilc and Nowak , 2005; Kalueff and Nutt , 2007 ) . Thus far , neurometabolic studies on preclinical models of early-life stress , or in patients with a life history of early adversity have not been carried out , making it difficult to directly compare our observations . Collectively , our results suggest that driving enhanced Gq-signaling-based activation of forebrain excitatory neurons in the postnatal window can evoke persistent dysregulation of amino acid neurotransmitter system metabolism , which may contribute to the long-lasting behavioral changes . We focused our gene expression and electrophysiological studies on the hippocampus , which has been strongly implicated in mood-related disorders ( Santos et al . , 2018; Campbell and Macqueen , 2004; Femenía et al . , 2012 ) . Early stress influences both hippocampal neuronal morphology and plasticity , features that contribute to the behavioral sequelae of early trauma ( Kim et al . , 2006; Fenoglio et al . , 2006; Maccari et al . , 2014; McEwen et al . , 2016 ) . We noted a decreased expression of several activity- and plasticity-related markers within the hippocampus , observed months post the cessation of PNCNO treatment , indicative of persistent molecular changes that ensue from the transient postnatal perturbation . These markers included transcription/translation factors , scaffolding proteins , cell adhesion molecules , and ion channels previously implicated in the regulation of excitation-inhibition balance ( Gatto and Broadie , 2010; Loebrich and Nedivi , 2009 ) . These results are suggestive of the programming of an altered excitation/inhibition within the hippocampus , which is supported by our electrophysiological observations . Several studies in the past have investigated neurophysiological consequences of early adversity ( Ali et al . , 2011 ) . Previous results indicate persistent dysregulation of signaling via Gq-coupled neurotransmitter receptors ( M1 and 5-HT2A ) in the neocortex of maternally separated animals ( Benekareddy et al . , 2010; Proulx et al . , 2014 ) . Rodent models of rearing in an impoverished environment ( Cui et al . , 2006; Brunson et al . , 2005 ) , poor maternal care ( Bagot et al . , 2009; Weaver et al . , 2004; Meaney and Szyf , 2005 ) , neonatal novelty exposure ( Zou et al . , 2001 ) , and maternal separation ( Kehoe et al . , 1995; Gruss et al . , 2008; Salzberg et al . , 2007 ) are all associated with impairment of hippocampal long-term potentiation ( LTP ) . Further , in vivo electrophysiological recordings in a model of neonatal isolation indicate a decline in hippocampal outputs ( Bartesaghi , 2004; Bartesaghi et al . , 2006 ) . Our observations of reduced expression of the activity marker c-Fos in all hippocampal subfields , concomitant with an increase in cumulative probabilities of sIPSC amplitude and a reduction of sIPSC interevent intervals , is indicative of decreased activity in hippocampal networks in keeping with observations in early stress models ( Ali et al . , 2011 ) . Our results support an overall increase in inhibitory neurotransmission within the hippocampi of adult mice with a history of PNCNO treatment . The increase in GABA flux , enhanced metabolic rate of GABAergic neurons , and the shift toward increased inhibition noted in hippocampi of PNCNO-treated mice could arise as an adaptive compensation to increased DREADD-mediated excitation during the postnatal window . The effect on hippocampal excitatory neurotransmission on the other hand appears more complex , with an increase in low amplitude and a decline in larger amplitude spontaneous events , along with an increase in the frequency of sEPSC events . This was concomitant with an increase in the cumulative probability of mEPSC amplitude , and a decline in high amplitude mIPSC events . This suggests that the overall decline in inhibition is unlikely to be cell-autonomous , and probably emerges as a consequence of an alteration in the excitatory-inhibitory recurrent network of the hippocampus . Our results do not allow us to distinguish whether the dysfunctional glutamate/GABA metabolism and neurotransmission observed in PNCNO-treated mice serve as instructive/permissive to the development of psychopathology , or simply arise as compensatory adaptations due to increased neuronal activation of forebrain excitatory neurons in the postnatal window . This motivates future experiments to address the influence of driving Gq-signaling-mediated activation of forebrain excitatory neurons in the postnatal window on excitatory and inhibitory neurotransmission in both cortical and hippocampal networks , as well as the emergence of excitation-inhibition balance within these neurocircuits . A complementary set of studies involving the perturbation of inhibitory Gi signaling within forebrain excitatory neurons in early postnatal life , both baseline and within the background of early adversity , would also provide valuable insights into the cortical microcircuitry and signaling pathways that contribute to the programming of long-lasting behavioral changes in emotionality . In conclusion , we show that chemogenetic activation of forebrain excitatory neurons during postnatal life evokes a long-lasting increase in anxiety- , despair- , and schizophrenia-like behavior . These behavioral changes are accompanied by a dysregulation of glutamate/GABA metabolism in the cortex and hippocampus , as well as perturbed inhibitory and excitatory neurotransmission within the hippocampus in adulthood . Our perturbation evokes several of pathophysiological features associated with preclinical and clinical studies of early adversity . These findings suggest the intriguing possibility that early adversity could program specific aspects of long-lasting behavioral , molecular , metabolic and functional changes via a modulation of Gq-signaling-mediated neuronal activation of forebrain excitatory neurons within the critical temporal window of postnatal life .
The CamKIIα-tTA transgenic mice ( Mayford et al . , 1996 ) were gifted by Dr . Christopher Pittenger , Department of Psychiatry , Yale School of Medicine . The TetO-hM3Dq mice ( Cat . No . 014093; Tg ( TetO-CHRM3* ) 1Blr/J ) and C57BL/6J mice were purchased from Jackson Laboratories , USA . The genotypes of CamKIIα-tTA::TetO-hM3Dq animals were confirmed by PCR-based genotyping analysis . All experiments using bigenic mice utilized mice which were homozygous for both CamKIIα-tTA and TetO-hM3Dq . Single-positive animals , positive for either CamKIIα-tTA or TetO-hM3Dq , as well as the background strain C57BL/6J were used for control experiments . The animals were bred in the Tata Institute of Fundamental Research ( TIFR ) animal house facility . All animals were maintained on a 12 hr light-dark cycle ( 7 am to 7 pm ) , with ad libitum access to food and water . Slice electrophysiology experiments were carried out at the Jawaharlal Nehru Centre for Advanced Scientific Research ( JNCASR ) , Bengaluru and 1H-[13C]-NMR experiments were carried out at the Centre for Cellular and Molecular Biology ( CCMB ) , Hyderabad . Experimental procedures were carried out as per the guidelines of the Committee for the Purpose of Control and Supervision of Experiments on Animals ( CPCSEA ) , Government of India and were approved by the TIFR , JNCASR , and CCMB animal ethics committees . Care was taken across all experiments to minimize animal suffering and restrict the number of animals used . DREADD agonist , CNO ( Cat . No . 4936 , Tocris , UK ) was used to selectively activate the excitatory DREADD , hM3Dq . CNO was dissolved in 5% aqueous sucrose solution for oral delivery in postnatal and juvenile treatment experiments , and in physiological saline for intraperitoneal delivery in adult-onset treatments . The alternative DREADD agonist compound 21 ( C21; Cat . No . 5548 , Tocris , UK ) was dissolved in 10 µl DMSO , and then diluted in 5% aqueous sucrose solution to a 1 ml stock solution . Following this , the solution was aliquoted and stored at −80°C before using for oral delivery in postnatal experiments . For vehicle treatment , the base solution without the drugs was used . For slice electrophysiology experiments , stock solutions of CNO , 6-Cyano-7-nitroquinoxaline-2 , 3-dione disodium ( CNQX disodium salt; Cat . No . 1045 , Tocris , UK ) , DL-2-Amino-5-phosphonopentanoic acid sodium salt ( AP5 , Cat . No . 3693 , Tocris , UK ) , ( - ) -Bicuculline methochloride ( Cat . No . 0131 , Tocris , UK ) , and Tetrodotoxin citrate ( TTX; Cat . No . ab120055 , Abcam , UK ) were prepared and aliquots were stored at −20°C . For all slice electrophysiology experiments , acute slice preparations had bath application of drugs in artificial CSF ( aCSF ) using a perfusion system . To assess HA-tagged hM3Dq DREADD expression in the hippocampus and cortex of CamKIIα-tTA::TetO-hM3Dq bigenic mice on postnatal Day 7 ( P7 ) , we performed western blotting analysis for the HA antigen . To examine the influence of CNO-mediated hM3Dq DREADD activation on expression levels of neuronal activity markers ( c-Fos , phospho-ERK/ERK ) , we fed a single dose of 1 mg/kg CNO or vehicle to CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups ( P7 ) and sacrificed them 15 min post-feeding . In order to examine the effect of chronic CNO-mediated hM3Dq DREADD activation on expression levels of the neuronal activity marker p-ERK/ERK , we fed 1 mg/kg CNO or vehicle to CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups once daily from P2 to P7 and sacrificed them 15 min post-feeding on P7 . Tissue samples were dissected and stored at −80°C , and then homogenized in Radioimmunoprecipitation assay ( RIPA ) buffer ( 10 mM Tris-Cl ( pH 8 . 0 ) , 1 mM EDTA , 0 . 5 mM EGTA , 1% Triton X-100 , 0 . 1% sodium deoxycholate , 0 . 1% SDS , 140 mM NaCl ) using a Dounce homogenizer . The lysis buffer contained protease and phosphatase inhibitors ( Sigma- Aldrich , United States ) . Following the estimation of protein concentration using the Quantipro BCA assay kit ( Sigma-Alrich , United States ) , equal amounts of lysate were resolved on a 10% sodium dodecyl sulfate polyacrylamide gel and then transferred onto polyvinylidene fluoride membranes . Blots were blocked in 5% milk dissolved in TBST for 1 hr , and subsequently incubated overnight with respective primary antibodies that is rabbit anti-HA ( 1:1500 in 5% milk , Cat . No . H6908 , Sigma-Aldrich , United States ) , rabbit anti-c-Fos ( 1:1000 in 5% milk , Cat . No . 2250 , Cell Signalling Technology , United States ) , rabbit anti-actin ( 1: 10 , 000 in 5% BSA , Cat . No . AC026 , Abclonal Technology , United States ) , rabbit anti-p-ERK1/2 ( Thr202/Tyr204 ) ( 1:1000 in 5% BSA , Cat . No . 9101 , Cell Signalling Technology , United States ) , or rabbit anti-ERK1/2 ( 1:1000 in 5% BSA , Cat . No . 9102 , Cell Signalling Technology , United States ) . Following subsequent washes , blots were exposed to HRP conjugated goat anti-rabbit secondary antibody ( 1:6000 , Cat . No . AS014 , Abclonal Technology , United States ) for 1 hr . Signal was visualized using a GE Amersham Imager 600 ( GE life sciences , United States ) with a western blotting detection kit ( WesternBright ECL , Advansta , United States ) . Densitometric quantitative analysis was performed using ImageJ software . HA-tagged hM3Dq DREADD expression in the hippocampus and cortex of CamKIIα-tTA::TetO-hM3Dq bigenic mice ( P7 ) was visualized using immunofluorescent staining for the HA epitope . Pups single-positive for either CamKIIα-tTA or TetO-hM3Dq were used as the genotype-controls . Double immunofluorescence stainings were performed on brain sections derived from adult CamKIIα-tTA::TetO-hM3Dq bigenic mice . Mice were sacrificed by transcardial perfusion with 4% paraformaldehyde , and 40 µm thick serial coronal sections were obtained using a vibratome ( Leica , Germany ) . Following a permeabilization step at room temperature in phosphate-buffered saline with 0 . 4% Triton X-100 ( PBSTx ) for 1 hr , the sections were then incubated in the blocking solution [1% Bovine Serum Albumin ( Roche , Cat . No . 9048-49-1 ) , 5% Normal Goat Serum ( Thermoscientific , Cat . No . PI-31873 ) in 0 . 4% PBSTx] at room temperature for 1 hr . The sections were incubated with primary antibody , rabbit anti-HA ( 1:250; Rockland , Cat . No . 600-401-384 , USA ) . For double-label immunofluorescence experiments , to examine the co-localization of HA-tagged hM3Dq DREADD with markers of excitatory neurons , inhibitory neurons , and glial cells , tissue sections were exposed to the following antibody cocktails: rat anti-HA ( 1:200 , Roche diagnostics , Cat . No . 10145700 ) with rabbit anti-CamKIIα ( 1:200 , Santa Cruz , Cat . No . sc-12886-R ) , or rabbit anti-GABA ( 1:200 , Sigma , Cat . No . A2052 ) , or rabbit anti-GFAP ( 1:500 , Chemicon , Cat . No . AB5804 ) for 4 days at 4°C . Following sequential washes with 0 . 4% PBSTx , the sections were incubated with the secondary antibody , goat anti-rabbit IgG conjugated to Alexa Fluor 568 ( 1:500; Invitrogen , Cat . No . A-11011 , USA ) and goat anti-rat IgG conjugated to Alexa Fluor 488 ( 1:500; Invitrogen , Cat . No . A-21212 , USA ) , for 2 hrs at room temperature , followed by washes with 0 . 4% PBSTx . Sections were mounted on to slides using Vectashield Antifade Mounting Medium with DAPI ( Vector , H-1200 , USA ) and images were visualized on a LSM5 exciter confocal microscope ( Zeiss , Germany ) . To determine the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on neurometabolism in the hippocampus and cerebral cortex of vehicle and PNCNO-treated CamKIIα-tTA::TetO-hM3Dq bigenic male mice , the 13C labeling of brain metabolites were measured in tissue extracts using 1H-[13C]-NMR spectroscopy following infusion of [1 , 6-13C2]glucose . To determine the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on persistent changes in gene expression within the hippocampus , hippocampi derived from vehicle and PNCNO-treated CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice were subjected to qPCR analysis . Vehicle and PNCNO-treated adult CamKIIα-tTA::TetO-hM3Dq bigenic mice were anesthetized by CO2 inhalation and sacrificed by rapid decapitation . The hippocampi were then dissected in ice-cold PBS , snap-frozen in liquid N2 , and stored at −80°C . RNA extraction was performed using Trizol ( TRI reagent , Sigma-Aldrich , USA ) . The RNA was quantified using a Nanodrop ( Thermo Scientific , USA ) spectrophotometer followed by reverse transcription reaction to produce cDNA using PrimeScript RT Reagent Kit ( Takara , Clonetech , Japan ) . Specific primers against the genes of interest ( Figure 6—source data 1 ) were designed and qPCR was performed to amplify the genes of interest using the CF96X Real Time System ( BioRad , USA ) . The qPCR data were analyzed using the ΔΔCt method as described previously ( Bookout and Mangelsdorf , 2003 ) . Ct value for a particular gene was normalized to the endogenous housekeeping gene GAPDH ( Glyceraldehyde 3-phosphate dehydrogenase ) , which was unchanged across treatment groups . To determine the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window on persistent changes in neuronal activity within the hippocampus , brain sections were subjected to c-Fos immunohistochemistry and cell counting analysis . Vehicle and PNCNO-treated CamKIIα-tTA::TetO-hM3Dq bigenic adult male mice that were naïve for behavioral testing , were sacrificed by transcardial perfusion with 4% paraformaldehyde . Coronal sections of 40 µm thickness were obtained using the vibratome ( Leica , Germany ) . Sections were then blocked at room temperature for 2 hr in 10% horse serum with 0 . 3% TritonX-100 ( made in 0 . 1M Phosphate buffer ) following which they were incubated with rabbit anti-c-Fos antibody ( 1:1000 , Cat no . 2250 , Cell Signalling Technology , United States ) for 2 days at 4°C . Subsequently , they were subjected to incubation with the secondary antibody ( biotinylated goat anti-rabbit , 1:500 , Cat no . BA9400 , Vector Labs , United States ) for 2 hr at room temperature . Signal was amplified using an Avidin-biotin complex based system ( Vector lab , Vectastain ABC kit Elite PK1600 , United States ) and then visualized using Diaminobenzidine tetrahydrochloride substrate ( Cat no . D5905 , Sigma-Aldrich , United States ) . An experimenter blind to the treatment groups carried out cell counting of c-Fos immunopositive cells in the hippocampal subfields namely the CA1 , CA3 , and dentate gyrus ( DG ) using a brightfield microscope ( Zeiss Axioskop two plus , Germany ) at a magnification of 200X . Eight sections , separated by a periodicity of 200 µm , spanning the rostrocaudal extent of the hippocampus ( four dorsal and four ventral ) were selected from each mouse . Results are expressed as the number of c-Fos-positive cells per section for each hippocampal subfield . In order to determine the influence of CNO-mediated hM3Dq DREADD activation on spiking activity , drug-naïve CamKIIα-tTA::TetO-hM3Dq bigenic mouse pups were sacrificed on postnatal Day 7 following which current clamp recordings were performed with CNO bath application . To observe the effects of CNO-mediated chronic postnatal hM3Dq DREADD activation of CamKIIα-positive excitatory neurons on hippocampal neurotransmission , CamKIIα-tTA::TetO-hM3Dq bigenic mice pups were treated once daily with CNO or vehicle from postnatal Day 2–7 ( P2–7 ) and sacrificed on P7 for whole-cell patch clamp recording in aCSF . To determine the influence of chronic CNO-mediated hM3Dq DREADD activation of CamKIIα-positive forebrain excitatory neurons during the early postnatal window ( P2–14 ) on hippocampal neurotransmission that persists into adulthood ( 3–4 months ) , whole-cell patch clamp recording was performed on acute hippocampal slices derived from PNCNO and vehicle-treated adult CamKIIα-tTA::TetO-hM3Dq bigenic male mice . All experiments had two treatment groups and were subjected to a two-tailed , unpaired Student’s t-test using GraphPad Prism ( Graphpad Software Inc , USA ) . One-sample Kolmogorov-Smirnov test was performed to confirm normality . All graphs were plotted using GraphPad Prism ( Graphpad Software Inc , USA ) . Data are expressed as mean ± standard error of the mean ( S . E . M ) and statistical significance was set at p<0 . 05 . To account for type I errors , the qPCR data were further subjected to the two-stage linear step-up procedure of Benjamini , Krieger and Yekutieli method to calculate false discovery rate ( FDR ) at 5% . Vehicle and PNCNO-treatment groups were subjected to linear regression followed by ANCOVA in order to compare input-output curves and statistical significance was set at p<0 . 05 . For the analysis of spontaneous current data , amplitudes and interevent intervals of events recorded from vehicle or PNCNO-treatment groups were converted to corresponding cumulative probability distributions and then subjected to Kolmogorov-Smirnov two-sample comparison . Statistical significance was set at p<0 . 001 . | Stress and adversity in early childhood can have long-lasting effects , predisposing people to mental illness and mood disorders in adult life . The weeks immediately before and after birth are critical for establishing key networks of neurons in the brain . Therefore , any disruption to these neural circuits during this time can be detrimental to emotional development . However , it is still unclear which cellular mechanisms cause these lasting changes in behavior . Studies in animals suggest that these long-term effects could result from abnormalities in a few signaling pathways in the brain . For example , it has been proposed that overstimulating the cells that activate circuits in the forebrain – also known as excitatory neurons – may contribute to the behavioral changes that persist into adulthood . To test this theory , Pati et al . used genetic engineering to modulate a signaling pathway in male mice , which is known to stimulate excitatory neurons in the forebrain . The experiments showed that prolonged activation of excitatory neurons in the first two weeks after birth resulted in anxious and despair-like behaviors as the animals aged . The mice also displayed discrepancies in how they responded to certain external sensory information , which is a hallmark of schizophrenia-like behavior . However , engineering the same changes in adolescent and adult mice had no effect on their mood-related behaviors . This animal study reinforces just how critical the first few weeks of life are for optimal brain development . It provides an insight into a possible mechanism of how disruption during this time could alter emotional behavior . The findings are also relevant to psychiatrists interested in the underlying causes of mental illness after early childhood adversity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Chronic postnatal chemogenetic activation of forebrain excitatory neurons evokes persistent changes in mood behavior |
During hibernation , animals cycle between torpor and arousal . These cycles involve dramatic but poorly understood mechanisms of dynamic physiological regulation at the level of gene expression . Each cycle , Brown Adipose Tissue ( BAT ) drives periodic arousal from torpor by generating essential heat . We applied digital transcriptome analysis to precisely timed samples to identify molecular pathways that underlie the intense activity cycles of hibernator BAT . A cohort of transcripts increased during torpor , paradoxical because transcription effectively ceases at these low temperatures . We show that this increase occurs not by elevated transcription but rather by enhanced stabilization associated with maintenance and/or extension of long poly ( A ) tails . Mathematical modeling further supports a temperature-sensitive mechanism to protect a subset of transcripts from ongoing bulk degradation instead of increased transcription . This subset was enriched in a C-rich motif and genes required for BAT activation , suggesting a model and mechanism to prioritize translation of key proteins for thermogenesis .
Many mammals hibernate to conserve energy during extended periods of limited resource availability and harsh environmental conditions . As winter approaches in temperate climates , hibernators enter into a state of torpor . Torpor in ground squirrels involves active suppression of physiological processes to 2–5% of basal rates , which allows body temperature to lower to just above ambient , even as ambient temperatures fall to near freezing . This depressed state is not continuous throughout winter , however , instead it lasts for 1–3 weeks until it is punctuated by a spontaneous , rapid re-warming to 37°C; physiological rates during re-warming match or even exceed basal rates . The interbout arousal period is then sustained for 12–24 hr before torpor resumes . Cycles between torpor and arousal result in winter heterothermy or hibernation ( Figure 1 ) . Hibernation persists for 5–8 months before emergence in spring and maintenance of more typical mammalian homeostatic physiology throughout the summer period of growth and reproduction ( Figure 1 , see Carey et al . , 2003; for review ) . Although of broad medical interest for their ability to tolerate these extraordinary physiological extremes ( Carey et al . , 2003; Andrews , 2007; Carey et al . , 2012; Dave et al . , 2012 ) , many aspects of the hibernation phenotype remain poorly understood . Some of hibernation's most defining mysteries are the mechanisms that underlie the highly dynamic oscillations of the torpor–arousal cycle . 10 . 7554/eLife . 04517 . 003Figure 1 . The hibernating phenotype as a model for studying BAT metabolic regulation . ( A ) Schematic depicting the metabolic suppression and activation cycle of BAT during the highly recruited , winter hibernation phase ( blue shading ) of the annual cycle . Cartoon squirrels represent general phenotypic changes among annual and torpor–arousal cycles ( Hindle and Martin , 2014 ) . ( B ) Relationship of sample groups to body temperature over time . Blue highlighting on months indicates hibernation . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 003 Transcription and translation effectively cease at low body temperature during hibernation ( van Breukelen and Martin , 2001; van Breukelen and Martin , 2002 ) , yet organs maintain integrity and in some cases are quickly reactivated after 2 weeks of near inactivity in torpor . The need for immediate intense metabolic activation at low temperature is most pronounced in brown adipose tissue ( BAT ) ; early re-warming depends exclusively on non-shivering thermogenesis in this organ ( Cannon and Nedergaard , 2004 ) . Because of the constraints on gene expression during torpor , the rapid burst of metabolic activity that characterizes early re-warming may be particularly challenging for BAT . To balance the decreased transcription , mRNA degradation during torpor also must be reduced to maintain cellular integrity and permit function in early arousal . Just as with transcription , low body temperature ( i . e . , Q10 effects ) will slow rates of RNA degradation ( Burka , 1969; Bremer and Moyes , 2014 ) , but it is unclear how these two opposing activities will converge after two weeks to determine the steady-state abundance of specific RNAs at the end of a torpor bout . While the general consensus is that the transcriptome is largely stable during torpor ( reviewed by Tessier and Storey ( 2014 ) ) , this view is based upon results ( Frerichs et al . , 1998; O'Hara et al . , 1999; Knight et al . , 2000; Williams et al . , 2005 ) where ongoing degradation is not readily distinguishable from a stable transcriptome because of the sampling and normalization strategies employed . There are few clear examples of transcripts that diminish across a torpor bout ( Epperson and Martin , 2002 ) and others that appear to increase ( O'Hara et al . , 1999 ) , largely because few studies provide the necessary temporal resolution to quantify changes across a torpor bout . In this study , we interrogate BAT mRNA dynamics in 13-lined ground squirrels across the torpor–arousal cycle and the circannual rhythm of hibernation ( Figure 1B ) . We chose BAT because of its unique requirement to function quickly and maximally in the earliest moments of arousal , after spending two weeks at the transcriptionally prohibitive body temperatures of torpor . We used a transcriptional profiling approach developed for non-model organisms , EDGE ( Hong et al . , 2011 ) , on five precisely timed sample groups to capture multiple phases of the torpor–arousal cycle ( Figure 1B ) and , for comparison , three groups from the non-hibernating , homeothermic portion of the year ( Figure 1B ) .
A total of 38 EDGE-tag libraries , representing 8 distinct sampling groups ( Figure 1B ) , were sequenced , processed ( Figure 2—figure supplement 1 ) , and analyzed for changes associated with hibernation physiology . For each of the libraries , 90 . 1 ± 2 . 6% of the sequence reads aligned to ground squirrel genomic ( Supplementary file 3A in Grabek et al . , 2014 ) or mitochondrial DNA ( Figure 2A ) . After normalization , filtering , and annotation ( Figure 2B , Figure 2—figure supplement 1 ) , 14 , 798 EDGE-tags representing 8 , 089 unique genes remained ( Supplementary file 3B in Grabek et al . , 2014 ) . We first clustered the individual sample libraries by tag abundance using Random Forests ( Breiman , 2001 ) . Three main groups were evident ( Figure 2C ) : ( 1 ) ‘spring’ , independent of ambient temperature , spring cold ( SpC ) , and spring warm ( SpW ) ; ( 2 ) ‘winter warm’: interbout aroused ( IBA ) , entrance ( Ent ) , and summer active ( SA ) ; ( 3 ) ‘winter cold’: early torpor ( ET ) , late torpor ( LT ) , and early arousal ( EAr ) . Notably , BAT samples harvested from winter animals at warm body temperature clustered separately from those at low body temperature . This separation indicates the transcriptome is dynamic across a torpor bout . 10 . 7554/eLife . 04517 . 004Figure 2 . EDGE-tag library properties . Pie charts of EDGE tags mapped: ( A ) uniquely to 13-lined ground squirrel nuclear ( Genome Unique ) or mitochondrial ( Mito . Unique ) DNA; multiple locations ( Genome or Mito . Multiple ) ; or unmapped ( Unaligned ) ; or ( B ) the indicated distances from the nearest annotated Ensembl feature ( either overlapping or 3′ to the feature in kilobases ) . ( C ) Two-dimensional scaling plot showing Random Forests ( RF ) clustering of individual samples labeled by group symbol: spring warm ( SpW ) , spring cold ( SpC ) , summer active ( SA ) , interbout aroused ( IBA ) , entrance ( Ent ) , early torpor ( ET ) , late torpor ( LT ) , and early arousal ( EAr ) , as depicted in Figure 1 . ( D ) Line plots of EDGE-tag expression patterns for all 2 , 159 significant differentially expressed tags; mean scaled counts , solid line , ±SEM , dotted line . Total tags in each cluster are indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 00410 . 7554/eLife . 04517 . 005Figure 2—figure supplement 1 . Schematic illustrates library sequencing , read processing , tag annotation , and filtering after the creation of the EDGE-tag transcriptome libraries ( see ‘Materials and methods’ ) . Sequential actions are listed in each box , while the number of resulting reads/EDGE-tags are labeled between boxes . tpm = ‘Tags per million’ . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 005 The 14 , 798 tags were next tested for significant differential expression among the three main groups; changes were detected in 2 , 159 tags ( 14 . 6%; q < 0 . 05 ) representing 1 , 638 unique genes ( Supplementary file 3C in Grabek et al . , 2014 ) . These correlated well with quantitative changes in the BAT transcriptome of this species reported previously ( Hampton et al . , 2013 ) ; 91% of overlapping differentially expressed transcripts exhibited changes in the same direction among comparable states ( see ‘Materials and methods’ ) . DIANA hierarchical clustering identified six expression patterns among the differentially expressed tags ( Figure 2D and Supplementary file 3C in Grabek et al . , 2014 ) ; those in Clusters 1 and 2 were generally increased in spring compared to winter , while those in Clusters 3–6 were increased in winter , particularly in late torpor and early arousal . Distinct from the spring-enriched tags , those increased in winter were overwhelmingly enriched ( Huang da et al . , 2009 ) for functions related to BAT activation , such as lipid metabolism , lipid droplet formation , lipid transport , mitochondria and the TCA cycle ( Table 1 and Supplementary file 3D in Grabek et al . , 2014 ) . 10 . 7554/eLife . 04517 . 006Table 1 . DAVID functional annotations for each DIANA clusterDOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 006DIANA clusterFunctional annotation clusterEnrichment scoreAnnotations , nGenes , n1Cytosolic ribosome4 . 091112Zinc finger , C2H2-type2 . 19830Heme1 . 838Ribosome biogenesis1 . 6849Transcription1 . 674572Transcription6 . 03480Nuclear lumen5 . 91457RNA recognition motif , RNP-15 . 18318mRNA processing3 . 39417Transcription from RNA polymerase II promoter3 . 213173Mitochondrion outer membrane3 . 0746Triglyceride biosynthetic process1 . 78123Glutathione S-transferase , C-terminal-like1 . 4143Long-chain fatty acid transport1 . 37534Mitochondrial membrane5 . 843030Endoplasmic reticulum membrane3 . 942020Lipid particle2 . 7355Glucose metabolic process2 . 161111Peroxisome2 . 15995Mitochondrion5 . 612424Generation of precursor metabolites and energy2 . 661414Lipid droplet2 . 3833Lipid metabolism2 . 3177Oxidative phosphorylation2 . 25446Mitochondrion4 . 041727Neutral lipid biosynthetic process2 . 74144Glucose metabolic process2 . 54310Lipid catabolic process2 . 331311Adipocytokine signaling pathway2 . 17119The top five Functional Annotation Clusters , ordered by enrichment score ( >1 . 3 ) , are listed for each DIANA cluster . See Figure 2D for DIANA clusters and Supplementary file 3D in Grabek et al . , 2014 for all Functional Annotation Clusters . Surprisingly , a preponderance of winter-increased tags ( i . e . , transcripts ) reached their highest relative abundance during early torpor , late torpor , and/or early arousal ( Figure 2D , Clusters 4–6 ) despite near cessation of transcription in hibernators at low body temperature ( van Breukelen and Martin , 2002 ) . One clue to resolve this apparent paradox was provided by RPPH1 , the RNA subunit of RNaseP , whose relative abundance increased several 100-fold by early arousal ( q < 10^−14 , Figure 3A ) . Because RPPH1 is transcribed by Pol III , it is not typically polyadenylated ( Baer et al . , 1990 ) and should not have been recovered in these sequencing libraries . Nevertheless , RPPH1 acquired a long poly ( A ) tail at low body temperature ( Figure 3B , C ) , explaining its presence in the libraries and increase in the cold . 10 . 7554/eLife . 04517 . 007Figure 3 . Increased RPPH1 abundance is explained by the addition of a poly ( A ) tail . ( A ) Box plot of normalized tag counts for RPPH1 by state , triangle marks the mean . ( B ) Gel showing RPPH1 RT-PCR products from 3′ RACE ( 3′ R , lanes 1 and 3 ) and random hexamer ( RH , lanes 2 and 4 ) primed cDNA from early arousal ( EAr; lanes 1–2 ) and spring warm ( SpW; lanes 3–4 ) total RNA . Marker sizes are indicated on the left ( 50-bp ladder , lane M ) . ( C ) Multiple alignment of the RPPH1 genomic DNA and four 3′ end cDNA sequences from cloned 3′ RACE ( EAr ) products in B: uppercase , annotated RPPH1 RNA; lowercase , genomic DNA; underline , non-templated nucleotides . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 007 We considered three potential mechanisms that might explain increased transcript abundance at low body temperature: ( 1 ) elevated transcription; ( 2 ) relative stabilization; and ( 3 ) acquisition of a poly ( A ) tail . To probe these mechanisms , we quantified abundance and the effect of poly ( A ) tail length on the dynamics of RPPH1 and thirteen other transcripts , including three additional ncRNAs and ten mRNAs ( Supplementary file 1A; note that there are two isoforms of LIPE ) , during the torpor–arousal cycle . The absolute abundance of these transcripts was measured by RT-qPCR in total RNA , and short and long poly ( A ) RNA fractions ( Figure 4—figure supplement 1; Supplementary file 1A ) from interbout aroused , late torpor , early arousal , and spring warm animals ( n = 3 ) . Two classes of RNA dynamics were apparent; transcripts were either decreased ( labeled as Class I ) or stabilized ( labeled as Class II ) during torpor but not newly transcribed . Five Class I transcripts decreased during torpor , with poly ( A ) and total RNA mirroring the abundance of their EDGE tags ( compare IBA to LT in Figure 4A; see also Figure 4—figure supplement 2A and Supplementary file 1B ) . Interestingly , during early arousal , when core body temperature was still low , some of these transcripts increased , likely because heat generated early in the arousal process has returned BAT to a temperature permissive for transcription ( Osborne and Hashimoto , 2003 ) . These transcripts were largely bearing long poly ( A ) tails , which also appeared to shorten during torpor ( Figure 4—figure supplement 2B ) . Class I dynamics explain the DIANA Clusters 1–3 , where RNA decreased during torpor but then increased at the elevated body temperature of interbout arousal ( compare IBA to LT in Figure 2D ) , and likely the even larger collection of transcripts that were not differentially expressed ( e . g . , GAPDH , Figure 4—figure supplement 2A ) . Thus , it appears that most BAT transcripts slowly degrade over two weeks in torpor and are not replenished until body temperature recovers during the short euthermic period . 10 . 7554/eLife . 04517 . 008Figure 4 . Bulk RNA degradation with stabilization of selected transcripts and cycles of re-adenylation at low body temperature . ( A ) Class I , represented by RET proto-oncogene . Relative expression levels ( solid line; ±SEM , dotted line; y-axis ) of EDGE-tag counts ( far left ) , total RNA ( middle-left ) , poly ( A ) RNA ( green = long poly ( A ) RNA; orange = short poly ( A ) RNA; middle-right ) , and percent recovery ( y-axis ) of poly ( A ) RNA relative to total RNA ( far right ) among physiological states: interbout aroused ( IBA ) , late torpor ( LT ) , early arousal ( EAr ) , and spring warm ( SpW ) . Spearman correlations ( ρ ) to EDGE-tag expression labeled in three right boxes; *p ≤ 0 . 05 , **p ≤ 0 . 01 , ***p ≤ 0 . 005 . ( B–C ) Labeling is as in panel A . ( B ) Class IIA , represented by RPPH1 . ( C ) Class IIB , represented by STAP2 . ( D ) Western blot reveals three isoforms of the PNPLA2 protein ( left arrows ) among indicated ( top ) sample states; marker sizes are denoted on right , β-tubulin , below , served as a loading control . ( E ) Relative abundance ( solid lines; ±SEM , bars ) of the 55 and 47 kD PNPLA2 proteins among samples states . ( F ) Relative abundance pattern of the PNPLA2 48 kD protein , PNPLA2 long poly ( A ) and total RNA; hibernation states are double-plotted to reveal cyclical pattern of torpor and arousal . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 00810 . 7554/eLife . 04517 . 009Figure 4—figure supplement 1 . ePAT confirmation of RNA fractionation by poly ( A ) tail length . ( A ) A gel showing RT-PCR amplified CKB from TVN ( lane 2 ) and ePAT primed cDNA of one IBA , LT , SpW , and EAr short and long poly ( A ) RNA samples ( lanes 4–11; labeled along the top ) . A 50-bp ladder is shown in lanes 1 and 3 ( with M marked on top and several sizes denoted to the left of lane 1 ) . The TVN band marks the first 12 adenosines of the poly ( A ) tail , while all other bands from ePAT cDNA represent the total length of the poly ( A ) tail . ( B ) The mean ( +SEM ) short and long poly ( A ) tail lengths calculated from the ePAT ( -TVN ) band sizes of the samples within each RNA fraction . The short poly ( A ) tail is approximately 26 bp , while the long poly ( A ) tail is approximately 48 bp . *p < 0 . 05 by two-tailed Student's t-test . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 00910 . 7554/eLife . 04517 . 010Figure 4—figure supplement 2 . Class I RNA and poly ( A ) tail dynamics . ( A ) Class I RNA dynamics . In addition to RET ( Figure 4A ) , the measurements for the other transcripts ( labeled on left x-axis ) that belong to Class I are shown along horizontal panels . Relative expression levels ( n = 3; solid line; ±SEM , dotted line; y-axis ) of EDGE-tag counts ( far left ) , total RNA ( middle-left ) , poly ( A ) RNA ( green = long poly ( A ) RNA; orange = short poly ( A ) RNA; middle-right ) , and percent recovery ( y-axis ) of poly ( A ) RNA relative to total RNA ( far right ) among physiological states . See Supplementary file 1B , C for additional details of transcript classification and specific measurements . ( B ) The mean long:short poly ( A ) RNA ratios ( mean-scaled , solid line; ±SEM , dotted line; y-axis ) for all transcripts in Class I among the four sample states . Increased long:short poly ( A ) RNA ratio = lengthened poly ( A ) tail , while decreased ratio = shortened poly ( A ) tail . ( C and D ) Same as in B , except plots show mean poly ( A ) tail length changes for transcripts in Class IIA ( C ) and Class IIB ( D ) . All classes exhibited significant poly ( A ) tail length changes among sample states ( One-Way ANOVA; p = 0 . 003 for Class I; p < 10–9 for Class IIA; p = 0 . 006 for Class IIB ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 01010 . 7554/eLife . 04517 . 011Figure 4—figure supplement 3 . Class IIA RNA dynamics . In addition to RPPH1 ( Figure 4B ) , the measurements for the other transcripts ( labeled on left x-axis ) that belong to Class IIA are shown along the horizontal panels . Labeling is the same as in Figure 4—figure supplement 2A . See Supplementary file 1B , C for additional details of transcript classification and specific measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 01110 . 7554/eLife . 04517 . 012Figure 4—figure supplement 4 . Class IIB RNA dynamics . In addition to STAP2 ( Figure 4C ) , the measurements for the other transcripts ( labeled on left x-axis ) that belong to Class IIB are shown along the horizontal panels . Labeling is the same as in Figure 4—figure supplement 2A . See Supplementary file 1B , C for additional details of transcript classification and specific measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 012 Nine Class II transcripts were stabilized during torpor . While their EDGE-tags appeared to increase during torpor ( Figure 4B , C , left ) , this increase was not mirrored in total RNA ( Figure 4B , C , middle-left ) . We further sub-divided this class by differences in polyadenylation . Total RNA for five Class IIA transcripts remained stable among states ( Figure 4B , middle-left; Figure 4—figure supplement 3 ) . However , these transcripts increased in the short and long poly ( A ) fractions during late torpor and particularly early arousal ( Figure 4B , right; Figure 4—figure supplement 3 , Supplementary file 1C ) with concurrent poly ( A ) tail lengthening ( Figure 4—figure supplement 2C ) , correlating with their EDGE-tags ( Figure 4B , left; Figure 4—figure supplement 3; Supplementary file 1B ) . Total RNA decreased in torpor and early arousal for four transcripts in Class IIB ( Figure 4C , Figure 4—figure supplement 4 ) , whereas their polyadenylated fraction remained stable ( Figure 4C , middle-right; Figure 4—figure supplement 4 ) , resulting in an apparent increase ( Figure 4C , right; Figure 4—figure supplement 2D and 4; Supplementary file 1C ) and consistent with the EDGE-tag pattern ( Figure 4C , left; Figure 4—figure supplement 4; Supplementary file 1B ) . Thus , in contrast to Class I , Class II transcripts are stabilized throughout torpor with maintenance or acquisition of a poly ( A ) tail . The enhanced stability of this subset relative to all other transcripts apparently leads to their relative increases at low body temperature ( Figure 2D , DIANA Clusters 4–6 ) . We next tested whether the observed transcript dynamics in torpor–arousal cycles could impact the corresponding protein by measuring PNPLA2 . Three PNPLA2 protein isoforms , whose sizes were consistent with those predicted for mouse in UniProt ( UniProt Consortium , 2014 ) , were detected by Western blot ( Figure 4D ) . All appeared to cycle , but only the 48-kD band changed significantly ( Figure 4E , F ) , following the dynamics of the transcript with the long poly ( A ) tail despite no change in overall transcript abundance ( Figure 4F ) . Hence , the dynamics of this PNPLA2 protein isoform appears to be explained by polyadenylation changes in its transcript . To investigate how changes in rates of transcription and degradation could affect differential gene expression in torpor , we developed a mathematical model of transcript dynamics across the torpor–arousal cycle . We simulated a population of 50 ‘protected’ transcripts and a bulk population of 1 , 400 transcripts; these numbers are proportional to the 531 tags that were either increased or stabilized across a bout of torpor ( DIANA Clusters 5 and 6 , Figure 2D ) compared to the 14 , 267 tags in the full dataset . For this simulated population , the abundance of each RNA transcript was governed by a differential equation describing temperature-dependent rates of RNA synthesis and degradation ( Schwanhausser et al . , 2011 ) . To model RNA transcript dynamics across the torpor–arousal cycle ( Figure 5—figure supplement 1 ) , we introduced a representative 12-day body temperature profile , incorporating temperature-dependence into the rates of RNA synthesis and degradation based on Q10 effects ( Burka , 1969; van Breukelen and Martin , 2002 ) , as described in detail in Supplementary file 2 . In the 50-transcript subset , we implemented either fixed or temperature-dependent alterations to degradation and synthesis rates to determine the resulting protective effects on normalized transcript abundance following 10 days of torpor . We found that a temperature-dependent mechanism that protected a subset of transcripts relative to bulk RNA degradation ( Figure 5A , B ) was most consistent with the increased abundances observed experimentally . For a body temperature threshold of 10°C and degradation set to 3% of its rate in the warm animal , the relative abundance of the protected transcripts increased over twofold ( Figure 5C , D ) , best reflecting the experimental data . This effect was dose-dependent with the level of protection and was relatively insensitive to thresholds above 10°C ( Figure 5E ) . Although temperature-independent decreases in degradation rates also led to increases in the relative abundance of protected transcripts , this mechanism required implausible compensatory changes to either steady state RNA abundance or transcription rates in the warm animal ( Figure 5—figure supplement 2 ) . Due to the differential Q10 effects on transcription and degradation , increasing transcription rate did not produce relative abundance increases ( Figure 5—figure supplement 3 ) . Thus , in agreement with RT-qPCR data , mathematical modeling supports enhanced stabilization of a subset of transcripts via a temperature-dependent protective mechanism; this , rather than increased transcription , leads to the observed increase in their relative abundances at the low body temperature of torpor . 10 . 7554/eLife . 04517 . 013Figure 5 . Mathematical modeling dynamics for 1 , 400 bulk and 50 protected transcripts simulated over the 12-day torpor–arousal cycle . A temperature-dependent protective mechanism against degradation is implemented for protected transcripts: for body temperature below 10°C , degradation is set to 3% of its rate in the warm animal . Transcription rates for both population and degradation rates for bulk transcripts are adjusted for Q10 effects . Low body temperature during torpor causes the raw abundance of both ( A ) protected and ( B ) bulk transcripts to decrease . When abundances are normalized across the population , ( C ) protected transcripts appear to increase approximately twofold while the majority of ( D ) bulk transcripts still appear to degrade . ( E ) Systematically varying the temperature threshold and the percentage of the warm degradation rate associated with the protective mechanism reveals a dose-dependent relationship in which higher temperature thresholds and lower percentages are associated with larger fold increases over baseline in the protected subset . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 01310 . 7554/eLife . 04517 . 014Figure 5—figure supplement 1 . Mathematical modeling of transcript degradation . ( A ) Body temperature ( Tb ) over a 12 day torpor–arousal cycle drives transcript dynamics for 1 , 400 simulated transcripts . ( B ) The distributions of transcription and degradation rates result in an overall degradation of transcript abundance in all transcripts . ( C ) When transcript abundance is normalized , differences in rates cause some transcripts to appear to increase under baseline conditions . ( D ) Representative transcription and degradation rates are adjusted for Q10 effects as Tb varies . ( E ) Differential Q10 effects cause the ratio of transcription rate ( vsr ) to degradation rate ( kdr ) to vary with Tb and favor degradation at low Tb . This ratio specifies the temperature-dependent steady state abundance for each transcript . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 01410 . 7554/eLife . 04517 . 015Figure 5—figure supplement 2 . Results for Mechanism 2 . When lower rates of degradation were implemented for the subset of protected transcripts , the relative abundances of this subset were increased over the torpor–arousal cycle compared to baseline levels . This increase is illustrated by representative transcript time traces for raw and normalized abundance of 50 protected transcripts ( A , B; G , H ) and 1 , 400 bulk transcripts ( D , E; J , K ) where log mean mu for the distribution of half-lives is 5 . 5 for the protected transcripts compared to 2 . 5 for the bulk transcripts . This increase was more pronounced when the lower degradation rate was compensated by a lower transcription rate ( A–F ) compared to compensations in steady state abundance ( G–L ) . To quantify the dependence on degradation rate , we varied the log mean mu for the distribution of half-lives from 3 . 5 hr to 7 . 5 hr ( baseline mu value for bulk population was 2 . 5 hr ) . This corresponded to a change in average degradation rate from 3 . 63e-04 mRNAs/min to 6 . 69e-06 mRNAs/min . For lower degradation rates compensated by lower transcription rates , fold increase over baseline for protected transcripts showed a saturating dose dependent relationship with mu ( C ) . This mechanism had a minimal effect on bulk transcripts ( F ) . For lower degradation rates compensated by high steady state , fold increase over baseline for protected transcripts showed an inverted U-dependence on mu ( I ) : although this mechanism could produce large increases in the subset of protected transcripts , this effect was attenuated as small degradation rates caused large steady state abundances since degradation , but not transcription , is proportional to steady state values . The fold change in bulk transcripts decreased dose-dependently for this mechanism ( L ) . For mechanism 2 , the differential Q10 effect enhanced the effect of the decreased degradation rate at low body temperature . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 01510 . 7554/eLife . 04517 . 016Figure 5—figure supplement 3 . Results for mechanism 1 . When higher rates of transcription were implemented for the subset of protected transcripts , the relative abundances of this subset were decreased over the torpor–arousal cycle compared to baseline levels . This decrease is illustrated by representative transcript time traces for raw and normalized abundance of 50 protected transcripts ( A , B; G , H ) and 1 , 400 bulk transcripts ( D , E; J , K ) where log mean mu for the distribution of transcription rates is 1 . 33 for the protected transcripts compared to 0 . 033 for the bulk transcripts . The decrease in relative abundance was more pronounced when the increased transcription rate was compensated by an increased degradation rate ( A–F ) compared to compensations in steady state abundance ( G–L ) . To quantify the dependence on transcription rate , we varied the log mean mu for the distribution of transcription rates from 0 . 33 mRNAs/min to 3 . 33 mRNAs/min ( baseline mu value for bulk population was 0 . 033 mRNAs/min ) . This corresponded to a change in average transcription rate from 1 . 3967 mRNAs/min to 27 . 93 mRNAs/min . For increased transcription rates compensated by increased degradation rates , fold change over baseline for protected transcripts showed a saturating dose dependent relationship with mu ( C ) . Increased transcription rates compensated by steady state had a minimal effect on relative abundance of protected transcripts across mu values ( I ) . For both compensation mechanisms , increased transcription rates had a minimal effect on the abundance of bulk transcripts ( F and L ) . For Mechanism 1 , the differential Q10 effect prevented temperature-independent increases in transcription rates from increasing the relative abundance of protected transcripts . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 016 Finally , to address the possible mechanism underlying protection of selected transcripts from degradation , we examined transcript 3′ untranslated region ( UTR ) sequences for shared motifs ( Figure 6A ) ; because ground squirrel 3′ UTRs are largely unannotated , we defined 3′ UTRs as the 500 nt region immediately downstream of the stop codon . This choice was validated by taking a random sampling of 3′ UTR sequences in all clusters , which returned enrichment for a motif resembling the polyadenylation signal ( Colgan and Manley , 1997 ) when compared to a scrambled background set ( Figure 6B; [Bailey et al . , 2009] ) . To identify motifs unique to the protected RNA subset , the significantly changed transcripts were divided into two groups for comparison ( Figure 6A ) : ( 1 ) the positive set of transcripts that appeared to be stabilized ( e . g . , Figure 2D , DIANA Cluster 6 ) or increased in torpor ( e . g . , Figure 2D , DIANA Clusters 4–5 ) ; and ( 2 ) the negative set of transcripts that appeared to decrease in torpor ( e . g . , Figure 2D , DIANA Clusters 1–3 ) . When the positive set was compared to the negative set , EXTREME ( Quang and Xie , 2014 ) identified two significantly enriched C-rich motifs ( Motif 1 and 2 , Figure 6C , D ) . Significantly , DIANA Cluster 5 , comprised of the transcripts that most clearly increased in relative abundance from early to late in torpor , contained the greatest percentage of mRNAs with the two motifs ( Motif 1: 59 . 1%; Motif 2: 66 . 7% of transcripts , Figure 6C , D ) . DIANA Cluster 6 , comprised of transcripts that remained elevated and stable across a torpor bout , contained a higher proportion of transcripts with Motif 2 ( 50 . 8% , Figure 6D ) relative to the other DIANA clusters ( 8 . 5–39 . 8% , Figure 6D ) and the control , non-significant transcripts ( 21 . 1% , NS in Figure 6D ) . Additionally , the other winter-increased DIANA Clusters , 3 and 4 , were relatively enriched for these motifs as compared to the spring-increased DIANA Clusters 1 and 2 ( 38 . 4–43% vs 8 . 5–23 . 3% , Figure 6C , D ) , suggesting that these motifs play a broader role in enhanced transcript stability and/or translation during winter heterothermy . Although the transcripts in DIANA Cluster 4 appeared elevated in early torpor ( Figure 2D ) , their motif enrichment was similar to that of Cluster 3 . In contrast to those in Clusters 5 and 6 , these transcripts also appeared to largely degrade by late torpor ( compare ET to LT , Figure 2D ) ; hence their reduced motif enrichment is consistent with reduced transcript stability across a bout of torpor . 10 . 7554/eLife . 04517 . 017Figure 6 . Motif enrichment in the 3′ ends of protected transcripts . ( A ) Schematic shows methodology for identifying motifs enriched in the 3′ UTR regions ( 500-nt ) of transcripts increased or stabilized in torpor ( positive set; transcripts in DIANA Clusters 4–6 ) compared to transcripts decreased in torpor ( negative set; transcripts in DIANA Clusters 1–3 ) . The table below lists the number of unique transcripts within each DIANA Cluster used in the analysis , the sum of those in the negative or positive set and the number of non-significantly changed transcripts ( NS ) used in later comparisons . ( B ) The motif closely resembling the AAUAAA polyadenylation signal ( Colgan and Manley , 1997 ) identified by a random sampling of 3′ UTR sequences in all clusters compared to a scrambled background set . ( C ) Motif 1: the most significant motif identified in the positive set of transcripts when compared to the negative set . Bar plot below shows the percentage of transcripts in each DIANA Cluster or in the non-significant group ( NS ) that contains Motif 1 . Actual numbers of transcripts containing Motif 1 are labeled above bars ( see table in A for comparison ) . ( D ) Motif 2: the second and the only other significant motif identified in the positive set of transcripts when compared to the negative set . Labeling is the same as in C . ( E ) The top RNA-binding protein motif matches for Motif 1 and 2 using TOMTOM ( Gupta et al . , 2007 ) . These correspond to the poly ( C ) binding protein motifs reported by Ji et al . ( 2013 ) . ( F ) Box plot of normalized tag counts for PCBP3 by state , triangle marks the mean . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 017 To identify putative binding protein ( s ) for these enriched sequence motifs , we used TOMTOM ( Gupta et al . , 2007 ) , searching against a database of RNA binding motifs ( Ray et al . , 2013; Ji et al . , 2013 ) . Our motifs significantly matched those reported by Ji et al . ( 2013 ) ( Figure 6E ) , implying binding by a poly ( C ) binding protein . We detected expression of two poly ( C ) binding protein paralogs in our dataset: PCBP3 and PCBP4 . While PCBP4 did not vary with hibernation physiology , PCBP3 expression increased significantly in the winter groups ( q = 0 . 0067 , Figure 6F ) . The enrichment of PCBP binding motifs in the subset of transcripts that increased at low body temperature , together with the increased PCBP3 abundance in winter , suggest a role for PCBP3 in protecting a subset of BAT transcripts via its binding to the 3′ UTR C-rich motifs during torpor .
Our results show that enhanced stabilization and polyadenylation of a crucial group of transcripts , with evidence of ongoing bulk RNA degradation , occur during torpor and are likely tied to rapid activation of BAT . Dynamic polyadenylation has been demonstrated to control temporal and spatial regulation of translation in many systems ( Wu et al . , 1998; Kojima et al . , 2012 ) , including maternal Xenopus oocyte maturation ( reviewed in Vasudevan et al . [2006] ) . Generally , poly ( A ) tail elongation causes translational activation , whereas shortening leads to silencing and/or RNA degradation ( Weill et al . , 2012 ) . Similar to oocyte maturation , in hibernation , the transcriptional machinery is silenced during the two week period of torpor; therefore , post-transcriptional mechanisms affecting mRNA stability and translation serve in the rapid switch between a hypo- and hyper-metabolic state in BAT . For instance , PNPLA2 catalyzes the first committed step in triacylglycerol hydrolysis , resulting in diacylglycerol and free fatty acid ( Zimmermann et al . , 2004 ) . Its increased translation via poly ( A ) tail lengthening in early arousal would ensure immediate generation of free fatty acids for thermogenesis , while poly ( A ) shortening during interbout arousal offers a parsimonious means to silence protein translation and lower free fatty acids as metabolic activity declines . Additionally , dynamic polyadenylation likely controls translation of the other Class IIA transcripts . Although there is currently no evidence for this type of mechanism operating in human BAT , our results suggest a potential therapeutic strategy by which translation of key proteins can be prioritized for recovery from metabolic repression , and more specifically , for rapid activation of thermogenesis in BAT . A role for post-transcriptional regulation in hibernation was posited previously , based on poor correlations between mRNA and protein levels ( Shao et al . , 2010 ) . In this study , we provide evidence for stabilization of a specific , functionally relevant subset of mRNAs during torpor . Moreover , our findings are consistent with reports of mRNA degradation during torpor ( Epperson and Martin , 2002 ) and with global maintenance of poly ( A ) tails and their increased length ( Knight et al . , 2000 ) . These results also provide further explanation to histological observations , as RNP granules containing both ncRNAs and mRNAs are formed during torpor in BAT nucleoli ( Malatesta et al . , 2008; Malatesta et al . , 2011 ) . The Class II ncRNAs identified in our study are located in nucleoli , suggesting that sequestration into these RNP granules protects a subset of transcripts from the degradation affecting bulk BAT RNA , which is also consistent with the results of our mathematical modeling . Our results propose that a mechanism underlying enhanced stabilization of a subset of RNAs involves their binding by one of the poly ( C ) binding proteins ( PCBPs ) , as we detected their corresponding C-rich motifs in the 3′ regions of the torpor-stabilized mRNAs . Intriguingly , PCBPs are involved in many aspects of post-transcriptional control that are consistent with our observations , including enhanced stability of long-lived mRNAs ( Makeyev and Liebhaber , 2002 ) . Further roles include 3′ end-processing and alternative polyadenylation ( Ji et al . , 2011; Ji et al . , 2013 ) , and the addition , maintenance ( Wang et al . , 1999 ) , and elongation of poly ( A ) tails ( Vishnu et al . , 2011 ) . Finally , these proteins are involved in both translational silencing and enhancement ( Makeyev and Liebhaber , 2002 ) . Although these roles are established for the predominantly studied PCBP1 and PCBP2 , these paralogs were not detected in our dataset . Rather , we detected increased expression of PCBP3 during winter heterothermy , a pattern that would be expected for a role involving enhanced mRNA stabilization and polyadenylation during torpor . Furthermore , PCBP3's pattern runs in contrast to most of the other RNA binding proteins detected in this dataset , which , if changed , were largely decreased during winter heterothermy ( see Table 1 , functional annotations for DIANA Clusters 1 and 2 ) . In addition to a PCBP , other 3′ UTR binding proteins may be involved in enhancing mRNA stability during torpor . The poly ( A ) binding protein PABP1 and the TIA-1/R RNA binding proteins were recently shown to localize to discrete sub-nuclear foci during torpor in the livers of 13-lined ground squirrels ( Tessier et al . , 2014 ) . PABP1 specifically binds to the poly ( A ) tails of transcripts , influencing their length as well as overall transcript translation and stability ( Burgess and Gray , 2010 ) . Significantly , the PCBPs 1 and 2 functionally interact with PABP1 in order to prevent deadenylation and to maintain mRNA stability ( Wang et al . , 1999 ) . While further research is needed to determine whether the protection identified here extends to other tissues during torpor and to thoroughly examine the role of PCBP3 or its homologs in this protection , our results of enhanced stabilization and polyadenylation for a subset of crucial transcripts suggest a mechanistic link to PABP1 observations in other organs ( Knight et al . , 2000; Tessier et al . , 2014 ) . Although there are reports of elevated RBM3 , a cold-induced RNA binding protein , in several organs including BAT from ground squirrels and bears during hibernation ( Williams et al . , 2005; Yan et al . , 2008; Fedorov et al . , 2011 ) , we found no evidence for enrichment of RBM3 recognition motifs ( Liu et al . , 2013; Ray et al . , 2013 ) in our subset of stabilized transcripts . More broadly , our study highlights the importance of how transcriptome data is interpreted . While it is generally assumed that changes in steady-state mRNA levels stem from changes in the rate of transcription , varying the rate of degradation also changes steady-state levels; this phenomenon has been observed at both mRNA and protein levels in the cold acclimation of fish ( Sidell , 1977; Bremer and Moyes , 2014 ) . Recently , the balance between mRNA synthesis and degradation rates was examined on a global scale in cultured mammalian cells , demonstrating complex gene-specific effects of transcription , processing , decay , and translation ( Rabani et al . , 2011; Schwanhausser et al . , 2011 ) . In our study , the modeling results shown in Figure 5 demonstrate that transcript-specific variability in the rate of degradation may cause a subset of transcripts to increase in relative abundance . These transcripts are degrading , but more slowly than most of the bulk transcripts and therefore exhibit a small increase in relative abundance across the torpor bout . Thus , many of the smaller fold changes observed in gene expression datasets from hibernators ( Williams et al . , 2005; Yan et al . , 2008; Hampton et al . , 2013; Schwartz et al . , 2013 ) likely reflect intrinsic differences in the stability of specific mRNAs rather than specific mechanisms to regulate their transcription or decay . However , for changes greater than approximately twofold in a substantial number of transcripts , here ∼3 . 5% of the total , a specific regulatory mechanism appears to be required . Furthermore , particularly large fold changes , as observed here with RPPH1 , likely reflect the addition or lengthening of poly ( A ) tails . Our data suggest a model ( Figure 7 ) of RNA dynamics in hibernator BAT wherein key RNAs for BAT function are selectively stabilized during torpor while bulk transcripts decline through degradation in the absence of new transcription . Stabilization likely occurs by a temperature-dependent protective mechanism that is in place before body temperature reaches 5°C , such as PCPB3 binding to the 3′ UTRs of protected transcripts , which then leads to their relative increase as torpor progresses . At the end of torpor and onset of arousal , the stabilized mRNA subset with the longest poly ( A ) tails is translated immediately as BAT temperature becomes permissive . As BAT temperature rises , further polyadenylation of the remaining stabilized RNAs facilitates their translation , transcription resumes ( Osborne et al . , 2004 ) , and , during interbout arousal , transcripts that were previously degraded during torpor are replenished to their baseline levels . This dynamic cycle of transcription , degradation , stabilization , and polyadenylation in BAT leads to translation of the correct transcripts at the correct time with minimal energy expenditure . Specifically: ( 1 ) energy intensive translation during early arousal is directed to proteins needed for BAT activation; ( 2 ) the cell is not dependent on de novo transcription at the onset of the short bursts of metabolic activity , which could delay thermogenesis and induce stress; ( 3 ) inhibition of translation via shortening of poly ( A ) tails while body temperature is high or begins to decline conserves energy compared to mRNA degradation and subsequent re-synthesis . Thus , given a general suppression of transcription by low body temperature during two-week torpor periods , stabilization and dynamic polyadenylation provide an alternative mechanism to prioritize transcripts for immediate translation when BAT metabolic activity rapidly resumes . 10 . 7554/eLife . 04517 . 018Figure 7 . Model of BAT RNA dynamics in hibernation . Physiological stages of the torpor–arousal cycle are listed inside of the arrows and underneath representative animals . Key RNA changes are noted . See text for detailed explanation . DOI: http://dx . doi . org/10 . 7554/eLife . 04517 . 018
13-lined ground squirrels were procured and housed as described previously ( Hindle and Martin , 2014 ) . All animals except those in the summer active group ( SA; n = 5; Figure 1B ) were surgically implanted in late August or early September with both an intra-peritoneal datalogger ( iButton , Embedded Data Systems ) and a radiotelemeter ( VM-FH disks , Mini Mitter , Sunriver , OR ) for remote body temperature monitoring until tissue collection . All animal protocols were approved by the University of Colorado Institutional Animal Care and Use Committee . BAT samples were collected from the axillary pads in animals representing eight different seasonal and physiological groups . All groups with their approximate body temperature and time of year are depicted in Figure 1B . Five groups represent animals in the winter hibernating portion of the year , including: early torpor ( ET; Tb = 4°C for 5–10% of previous torpor bout; n = 4 ) and late torpor ( LT; Tb = 4°C for 80–95% of prior torpor bout , n = 4 ) , early arousal ( EAr; Tb = 5–12°C , n = 5 ) , interbout-arousal ( IBA; Tb = 37°C for 2–3 hr after Tb stabilization , n = 4 ) , and entrance into torpor ( Ent; Tb = 23–27°C , n = 5 ) . Three groups consisted of animals in the non-hibernating portion of the year: summer active ( SA; n = 5 ) , collected in late July or early August , and two spring groups: ( 1 ) spring cold ( SpC; n = 5 ) animals had spontaneously aroused from hibernation terminally , exhibiting no torpor for 10–20 days despite remaining in constant darkness at 4°C; and ( 2 ) a spring warm group ( SpW; n = 6 ) , at least 7 days after ambient temperature was raised first to 9°C for 5 days and then to 14°C . Animals were euthanized by exsanguination under isoflurane anesthesia , perfused with ice cold isotonic saline , and decapitated before dissection . Upon collection , BAT tissue was immediately snap-frozen in liquid nitrogen and then stored at −80°C . Total RNA was extracted from each BAT sample using the RNeasy Lipid Mini extraction kit ( Qiagen , Venlo , the Netherlands ) and assessed for quantity via NanoDrop and quality via the Bioanalyzer ( RIN ≥ 8; Agilent Technologies , Santa Clara , CA ) . EDGE-tag libraries were created according to the protocol described by Hong et al . ( 2011 ) and submitted for massively parallel high-throughput sequencing at the genomic services lab at the HudsonAlpha Institute of Biotechnology , Huntsville , AL . The EDGE-tag libraries for all groups except ET and SA were sequenced on individual lanes of an Illumina GAIIx ( Illumina , San Diego , CA ) . The ET and SA sample libraries were prepared subsequently; these were barcoded and all 10 samples sequenced on a single lane of an Illumina HiSeq , with two technical replicates added from the first round of sequencing . We examined the transcripts that appeared to increase in torpor for shared motif enrichment in their 3′ UTRs . Due to uncertainty in 3′ UTR structure and length , we included only transcripts that contained significantly differentially expressed ( D . E . ) tags <1 kb 3′ to their nearest protein coding feature . First , the transcripts were divided by whether they fell into the ‘torpor-increased’ DIANA clusters 4–6 or ‘torpor-decreased’ DIANA clusters 1–3 . Each transcript was counted only once; in cases where multiple tags mapped to the same transcript , the most 3′ D . E . tag was used for assignment of the transcript to a particular cluster . Annotation of the 3′ UTRs in the ground-squirrel genome is currently sparse and likely imprecise; thus , for each gene , we conservatively defined the ‘3′ UTR’ to be the 500-nt region immediately 3′ to the stop codon . To identify enriched motifs , we used the motif discovery algorithm EXTREME ( Quang and Xie , 2014 ) with the default settings , except allowing 0 gaps , and with the 500-nt ‘3′ UTR’ sequences from ‘torpor-increased’ transcripts input as the positive set and the 500-nt ‘3′ UTR’ sequences from the ‘torpor-decreased’ transcripts input as the negative set . We considered resulting motifs with E values <1 as significantly enriched in the positive set . The enriched motifs were used as input in the MAST tool ( Bailey and Gribskov , 1998 ) of MEME Suite ( Bailey et al . , 2009 ) , which counted the number of transcripts containing the motif ( E-value <10 ) within each DIANA cluster . As a control , we also counted the number of motif occurrences in the 500-nt ‘3′UTRs’ of 484 non-significant D . E . transcripts ( q > 0 . 97; 500 transcripts were originally chosen but 16 lacked 3′ sequence data , hence 484 transcripts ) . Finally , to detect motifs enriched in all clusters , 3′ UTR sequences from 20 randomly chosen transcripts in each cluster ( 120 total ) were compared against a scrambled background set in MEME ( Bailey and Elkan , 1994 ) , with the settings set at a maximum width of 8-nt and a search of the given strand only . To identify the putative RNA binding proteins that might recognize these motifs , the significantly enriched motifs were uploaded into the TOMTOM motif comparison tool ( Gupta et al . , 2007 ) of the MEME Suite . The database against which these motifs were first searched consisted of the RNA binding protein motifs described by Ray et al . ( 2013 ) ; however , no significant matches were found . We next added the nine C-rich motifs reported and provided by Ji et al . ( 2013 ) to the RNA binding protein motif database and repeated the search for significantly enriched motifs . The significance values for motif matches were calculated via Pearson correlation coefficient in TOMTOM . All motif logos were generated in TOMTOM . | Many mammals hibernate to avoid food scarcity and harsh conditions during winter . Hibernation involves entering a state called torpor , which drastically reduces the amount of energy used by the body . During torpor , body temperature also decreases . This is particularly exemplified in ground squirrels , whose body temperature can hover at just above or even below the point of freezing . However , hibernating mammals cannot remain in this state continuously over the months of hibernation but instead cycle between bouts of torpor lasting for 1–3 weeks and brief periods of ‘arousal’ lasting between 12–24 hr , during which their body rapidly warms up . The heat required to start warming up the hibernator is generated from a specialized form of fat called brown adipose tissue . Normally , the bursts of metabolic activity that are required to create this heat depend on certain proteins being produced . Making a protein involves ‘translating’ its sequence from template molecules called messenger RNA ( mRNA ) , which are ‘transcribed’ from the gene that encodes the protein . During the low body temperatures experienced during torpor , both of these processes stop . So how is the hibernator able to quickly and efficiently heat itself up during the arousal periods of hibernation ? Grabek et al . investigated this by analyzing the relative levels of mRNA in the brown adipose tissue of hibernating 13-lined ground squirrels . Using a special technique to sample and sequence small fragments of mRNA taken from brown adipose tissue , Grabek et al . compiled a profile of the mRNA molecules present at different points in the torpor–arousal cycle and compared this with a similar profile taken from squirrels that were not hibernating . From this analysis , Grabek et al . detected that a particular group of mRNA molecules that are required for producing heat increase in abundance during torpor , even though body temperature is low enough to stop gene transcription . This increased abundance does not occur because more of the mRNA molecules are made; instead , the mRNA molecules are modified to become more stable and long lasting . Once the animal warms up during arousal , gene transcription is reactivated and more new mRNA molecules are made . Grabek et al . suggest that the key mRNAs required for brown adipose tissue function are selectively stabilized during torpor through a temperature-dependent protective mechanism . These mRNAs are then preferentially translated into proteins during arousal to rapidly and efficiently heat the hibernator . Most other mRNA molecules degrade throughout torpor , and so their numbers decline as replacements are not transcribed until body temperature briefly recovers during arousal . Whether this protective mechanism is also used in other tissues during torpor remains a question for future work . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2015 | Enhanced stability and polyadenylation of select mRNAs support rapid thermogenesis in the brown fat of a hibernator |
Pericentromeric satellite repeats are enriched in 5-methylcytosine ( 5mC ) . Loss of 5mC at these sequences is common in cancer and is a hallmark of Immunodeficiency , Centromere and Facial abnormalities ( ICF ) syndrome . While the general importance of 5mC is well-established , the specific functions of 5mC at pericentromeres are less clear . To address this deficiency , we generated a viable animal model of pericentromeric hypomethylation through mutation of the ICF-gene ZBTB24 . Deletion of zebrafish zbtb24 caused a progressive loss of 5mC at pericentromeres and ICF-like phenotypes . Hypomethylation of these repeats triggered derepression of pericentromeric transcripts and activation of an interferon-based innate immune response . Injection of pericentromeric RNA is sufficient to elicit this response in wild-type embryos , and mutation of the MDA5-MAVS dsRNA-sensing machinery blocks the response in mutants . These findings identify activation of the innate immune system as an early consequence of pericentromeric hypomethylation , implicating derepression of pericentromeric transcripts as a trigger of autoimmunity . Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review . The Reviewing Editor's assessment is that all the issues have been addressed ( see decision letter ) .
In vertebrate genomes , the majority of cytosine residues within CpG dinucleotides are methylated at the 5 position of the cytosine ring ( 5-methylcytosine , 5mC ) ( Suzuki and Bird , 2008 ) . 5mC is established by the de novo DNA methyltransferases of the Dnmt3 family , and is propagated by the maintenance DNA methyltransferase , Dnmt1 ( Goll and Bestor , 2005 ) . In mice , frogs and zebrafish , mutation or morpholino-mediated depletion of Dnmt1 results in extensive genome-wide methylation loss and embryonic lethality ( Anderson et al . , 2009; Lei et al . , 1996; Rai et al . , 2006; Stancheva and Meehan , 2000 ) . In these species , global methylation deficiencies are linked to a variety of adverse outcomes including deregulation of gene expression , derepression of transposons , elevated levels of DNA damage and increased genome instability during mitosis ( Smith and Meissner , 2013 ) . Recent studies have further linked global hypomethylation to activation of antiviral signaling pathways in zebrafish mutated for dnmt1 and in cancer cells treated with the DNA methyltransferase inhibitor 5-azacytidine ( Chernyavskaya et al . , 2017; Chiappinelli et al . , 2015; Roulois et al . , 2015 ) . While these studies reinforce the general importance of DNA methylation in vertebrate development and tissue homeostasis , the extensive genome-wide loss of methylation in these models makes it difficult to assign significance to methylation deficiencies at any particular subclass of sequence . The pericentromeric satellite sequences that juxtapose chromosome centromeres represent an essential structural component of chromosomes and a significant source of 5mC in vertebrate genomes . These highly repetitive sequences appear particularly susceptible to methylation loss in cancer and senescent cells , although the consequences of this hypomethylation are not well understood ( Enukashvily et al . , 2007; Fanelli et al . , 2008; Nakagawa et al . , 2005; Narayan et al . , 1998; Qu et al . , 1999; Suzuki et al . , 2002; Tsuda et al . , 2002 ) . In contrast to global hypomethylation , loss of 5mC at pericentromeric repeats is compatible with human development . Individuals with the rare , autosomal recessive disorder Immunodeficiency , Centromere and Facial anomalies ( ICF ) syndrome show extensive hypomethylation of pericentromeric repeats , while methylation across the rest of the genome is relatively intact ( Tuck-Muller et al . , 2000; Velasco et al . , 2018; Weisenberger et al . , 2005 ) . Affected individuals usually die in late childhood or early adulthood , and exhibit variable symptoms including immunoglobulin deficiency , facial dysmorphism , growth retardation and a generalized failure to thrive ( Ehrlich et al . , 2008 ) . Chromosome anomalies including whole-arm deletions and multiradial chromosomes have also been reported in mitogen-stimulated lymphocytes from ICF-patients . However , similar chromosome anomalies are not observed in primary tissues from affected individuals ( Ehrlich , 2003 ) . Homozygosity mapping and whole-exome sequencing have separately implicated four genes in ICF syndrome: DNA Methyltransferase 3B ( DNMT3B , ICF type-1 ) , Zinc-finger and BTB domain containing 24 ( ZBTB24 , ICF type-2 ) , Cell division cycle associated 7 ( CDCA7 , ICF type-3 ) and Helicase , lymphoid-specific ( HELLS , ICF type-4 ) ( de Greef et al . , 2011; Thijssen et al . , 2015; Xu et al . , 1999 ) . Most of the described mutations in DNMT3B cause amino acid substitutions within the C-terminal catalytic domain , suggesting they may be hypomorphic . In contrast , the majority of mutations in ZBTB24 , CDCA7 and HELLS are predicted to cause loss of function . Mechanistically , ZBTB24 , CDCA7 and HELLS are thought to converge in a singular pathway that facilitates DNMT3B access to pericentromeric DNA ( Jenness et al . , 2018; Wu et al . , 2016 ) . To date , most studies of pericentromeric 5mC loss have been performed using transformed B-cell lines derived from ICF patients carrying mutations in DNMT3B ( Ehrlich et al . , 2008 ) . Attempts to generate viable mouse models of pericentromeric hypomethylation through mutation of ICF genes have had limited success . Mice harboring ICF-like mutations in Dnmt3b exhibit some characteristics of ICF syndrome including small size and facial anomalies . However , most mice die within 24 hr of birth ( Ueda et al . , 2006 ) . Global methylation profiles were not assessed in these mutants; but significant hypomethylation was reported at both pericentromeric repeats and retroviral sequences . Similar perinatal lethality was observed following deletion of the mouse HELLS orthologue . In this case , mutations were accompanied by roughly 50% reductions in 5mC , and methylation loss was noted at pericentromeres , retroviruses and some single copy sequences ( Tao et al . , 2011 ) . Deletion of the mouse Zbtb24 gene was reported to cause embryonic lethality; but methylation changes in these mutants have not been investigated ( Wu et al . , 2016 ) . Here , we describe a viable model of pericentromeric methylation loss , generated through mutation of the zebrafish zbtb24 gene . Homozygous mutant adults exhibited key phenotypic hallmarks of ICF syndrome including hypomethylation of pericentromeric satellite repeats . Hypomethylation of these repeats was first detected in mutants at 2 weeks post fertilization ( wpf ) and became more severe as animals matured . This progressive methylation loss allowed us to investigate the primary consequences of pericentromeric hypomethylation in the context of a vertebrate animal . Using this model , we link derepression of transcripts from hypomethylated pericentromeres to activation of an interferon-based innate immune response , and we demonstrate that this response is mediated through the MDA5-MAVS RNA sensing machinery . Our findings provide insights into the earliest consequences of pericentromeric hypomethylation , demonstrating an unappreciated function for methylation of pericentromeric repeats in protecting against autoimmunity .
The zebrafish genome encodes a single , well-conserved orthologue of ZBTB24 , which we mutated using TAL effector nucleases ( TALENs ) ( Figure 1A and Figure 1—figure supplements 1 and 2 ) . The recovered 7 . 9 kb deletion allele ( zbtb24mk22; here after referred to as zbtb24Δ ) , eliminates coding sequence between exons 2 and 5 ( Figure 1B ) . Animals that were homozygous for this deletion lacked detectable zbtb24 transcripts , suggesting zbtb24Δ is a null allele ( Figure 1—figure supplement 2D ) . Zbtb24Δ/Δ embryos were born to heterozygous parents at the expected Mendelian ratios and had no obvious morphological abnormalities during the first two weeks of development ( Figure 1C ) . Phenotypes that were reminiscent of ICF syndrome emerged as animals matured . Consistent with the small stature observed in ICF syndrome , by 3–4 weeks post fertilization ( wpf ) , zbtb24Δ/Δ mutant zebrafish were smaller than wild-type siblings raised under identical conditions , and this size reduction persisted into adulthood ( Figure 1D–F ) . As adults , zbtb24Δ/Δ mutants exhibited facial anomalies that were characterized by a quantifiable elongation of the snout ( Figure 1G–H ) . We also noted evidence of hypogammagloblulinemia in the presence of normal lymphoid cell numbers , which is an immunological hallmark of ICF syndrome ( Figure 1I–J ) . Significant death was noted among homozygous mutants at 4 months of age and fewer than 10% of zbtb24Δ/Δ animals survived beyond 8 months ( Figure 1K ) . Attempts to recover fertilized embryos by intercrossing or outcrossing zbtb24Δ/Δ adults were unsuccessful , suggesting that animals were sterile ( Figure 1—figure supplement 3A ) . Gonadal morphology in zbtb24Δ/Δ mutants appeared overtly similar to wild-type siblings in histological sections ( Figure 1—figure supplement 3B–C ) . However , testes size and sperm count were severely reduced in zbtb24 mutants , providing one potential explanation for impaired male fertility ( Figure 1—figure supplement 3D–G ) . Similar ICF-like phenotypes were observed in zebrafish that were homozygous for a second independently-isolated mutant allele of zbtb24 ( zbtb24mk19 ) ( Figure 1—figure supplement 4 ) . Taken together , these findings identify zbtb24 homozygous mutant zebrafish as a faithful animal model of ICF syndrome phenotypes . Pericentromeric satellite type-1 ( Sat1 ) repeats are found on all zebrafish chromosomes and comprise 5–8% of the zebrafish genome ( Phillips and Reed , 2000 ) . As expected , we found that Sat1 sequences from wild-type adults were resistant to digestion with the methylation sensitive restriction enzyme HpyCH4IV , indicating that these pericentromeric repeats were heavily methylated . In contrast to wildtype , Sat1 sequences from zbtb24Δ/Δ and zbtb24mk19/mk19 mutant adults were readily digested with HpyCH4IV , indicating extensive loss of methylation at these repeats ( Figure 2A–B and Figure 2—figure supplement 1A ) . Comparable Sat1 methylation deficiencies were observed when DNA was isolated from dissected adult brain , skin , muscle and fin , suggesting that these sequences were similarly hypomethylated in most adult somatic tissues ( Figure 2—figure supplement 1B and Figure 2—figure supplement 3A ) . Methylation levels at Sat1 repeats appeared normal in remaining sperm extracted from zbtb24Δ/Δ mutant adults , suggesting methylation loss may be restricted to somatic tissues ( Figure 2—figure supplement 1C–D ) . Somewhat unexpectedly , we found that pericentromeric methylation loss in zbtb24Δ/Δ mutants was progressive . While extensive hypomethylation of Sat1 sequences was detected in adults lacking zbtb24 , similar hypomethylation was not observed in mutants at 1 wpf ( Figure 2C–D ) . At 2 wpf , zbtb24Δ/Δ mutants exhibited roughly 3-fold increases in HpyCH4IV digestion , and sensitivity to digestion became increasingly pronounced in older animals ( Figure 2C–D ) . By 32 weeks , Sat1 sequences from zbtb24 mutants exhibited a 23-fold increase in HpyCH4IV digestion compared to wildtype , suggesting a greater than 95% reduction in methylation of these repetitive sequence blocks . Histone H3 lysine nine trimethylation levels were unaffected at Sat1 sequences in zbtb24Δ/Δ mutant adults ( Figure 2—figure supplement 2 ) . To clarify whether other sequences were also hypomethylated in zbtb24 mutants , we performed Enhanced Reduced Representation Bisulfite Sequencing ( ERRBS ) using genomic DNA isolated from the fins of three 6-month-old zbtb24Δ/Δ mutant adults and three wild-type siblings ( Garrett-Bakelman et al . , 2015 ) . At this stage , Sat1 sequences from isolated fins were 20-fold more sensitive to HypCH4IV in zbtb24 mutants compared to controls , indicating extensive loss of DNA methylation at pericentromeric repeats ( Figure 2—figure supplement 3A–B ) . We then used ERRBS data to interrogate the methylation status of 979 , 971 non-pericentromeric CpG sites across the genome in the same tissue samples . Our analysis revealed a strong correlation between genome wide 5mC levels in wild-type and zbtb24Δ/Δ mutant adults ( Pearson’s correlation value of 0 . 928 ) , although overall methylation levels appeared reduced by ~10% at all methylated sequence features in mutants ( Figure 2E–F and Figure 2—figure supplement 4 ) . Reductions consisted primarily of small-magnitude changes in 5mC across the genome , with only 1 . 3% ( 13 , 205 ) of examined CpG dinucleotides exhibiting methylation differences of greater than 20% . Consistent with these findings , at a threshold of 20% change ( p-value<0 . 01 ) , only 55 differentially methylated regions ( DMRs ) were identified between wild-type and zbtb24Δ/Δ adults ( Supplementary file 4 ) . Methylation levels at endogenous retroviruses and other transposable elements were also examined by methylation sensitive restriction digest . All tested elements were similarly resistant to digestion in zbtb24Δ/Δ mutant adults and wild-type siblings ( Figure 2—figure supplement 5 ) . Collectively , these data indicate that pericentromeres are a predominant site of methylation loss in zbtb24Δ/Δ mutants . To gain insights into the early consequences of methylation loss in zbtb24 mutants , we performed transcriptome analysis on RNA isolated from wild-type and zbtb24Δ/Δ zebrafish at 2 wpf . At this stage , zbtb24Δ/Δ mutants remain morphologically indistinguishable from wildtype , but show clear hypomethylation of pericentromeric sequences . RNA-seq identified 58 genes that were downregulated by more than 2-fold in zbtb24Δ/Δ larvae at 2 wpf , while 119 were upregulated by 2-fold or more ( Figure 3A ) . No gene enrichment signature was observed among downregulated genes . However , roughly 30% of upregulated genes were associated with activation of the innate immune system . In particular , we noted that upregulated transcripts included those associated with interferon stimulated genes ( ISGs ) and inflammatory cytokines ( Figure 3B ) . Consistent with these observations , Gene Set Enrichment Analysis ( GSEA ) identified significant enrichment of genes involved in viral response , a key function of innate immune pathways ( Figure 3C ) . Upregulation of ISGs was also observed in zbtb24Δ/Δ and zbtb24mk19/mk19 mutants by qRT-PCR at 3 wpf , whereas the same genes were expressed at wild-type levels at 1 wpf ( Figure 3D–E and Figure 3—figure supplement 1 ) . No immune-related genes ( and only one gene differentially upregulated in the RNA-Seq ) were found within 100 kb of identified DMRs , suggesting that direct loss of methylation at these sequences was unlikely to cause the response ( Figure 2—figure supplement 3D and Supplementary file 4 ) . Consistent with previous studies , we found that global methylation depletion using the DNA methyltransferase inhibitor 5-azacytidine also resulted in upregulation of immune response genes ( Figure 3—figure supplement 2 ) . The innate immune system represents an ancient defense system in which pathogen-associated molecular patterns ( PAMPs ) are recognized by pattern recognition receptors ( PRRs ) . These PRRs induce signaling cascades that drive the production of interferons and other inflammatory cytokines with antiviral and immune modulating functions ( Schneider et al . , 2014 ) . In addition to extracellular pathogens , PRRs also recognize PAMPs associated with cell-intrinsic stimuli including DNA damage , endogenous retroviral RNA and RNA-DNA hybrids ( Chiappinelli et al . , 2015; Härtlova et al . , 2015; Mankan et al . , 2014; Roulois et al . , 2015 ) . To clarify the origin of the response in zbtb24 mutants , we examined the major families of PRRs involved in innate immunity . These include the Toll-like receptors ( TLRs ) , which have broad functions in detecting PAMPs , the RIG-I like receptors ( RLRs ) , which are involved in the detection of cytosolic RNA and cGAMP synthase ( cGAS ) , which functions as a cytosolic sensor of DNA and RNA/DNA hybrids ( Crowl et al . , 2017 ) . Mutations in key mediator proteins required to propagate interferon signaling from each PRR family were introduced onto the zbtb24 mutant background and we tested the effect on ISG expression . Mutations in the zebrafish orthologs of mitochondrial antiviral-signaling protein ( mavs ) , which is an intermediate in RLR signaling and stimulator of interferon genes ( sting ) , which is involved in cGAS signaling were generated using CRISPR/Cas9 technology ( Figure 4—figure supplement 1A–B ) . The mutant allele of Myeloid differentiation primary response 88 ( myd88 ) , which is required for signaling through most TLRs , was previously described ( van der Vaart et al . , 2013 ) . As in prior experiments , significant increases of the ISGs , signal transducer and activator of transcription 1b ( stat1b ) and interferon regulatory factor ( irf7 ) were observed in zbtb24Δ/Δ larvae at 3 wpf by qRT-PCR ( Figure 4A–C ) . Introduction of myd88 or sting mutations had little impact on expression of these ISGs , as similar transcript levels were detected in zbtb24Δ/Δ single mutant animals compared to myd88hu3568/hu3568; zbtb24Δ/Δ or stingmk30/mk30; zbtb24Δ/Δ double mutants ( Figure 4A–B ) . Sustained ISG expression in these double mutants suggests limited roles for TLR and cGAS PRRs in mediating the interferon response in zbtb24 mutants . In contrast to myd88 and sting , mutation of mavs suppressed stat1b and irf7 upregulation in zbtb24Δ/Δ mutant animals . Expression levels of irf7 and stat1b were reduced 2- and 4-fold respectively in mavsmk28/mk28; zbtb24Δ/Δ double mutants when compared to zbtb24Δ/Δ single mutant zebrafish , indicating a requirement for mavs in the upregulation of these ISGs ( Figure 4C ) . This finding implicates RLR signaling in the activation of the innate immune system in zbtb24 mutants and suggests a cytosolic RNA trigger for the response . Given known roles for DNA methylation in transcriptional repression , we next tested whether loss of methylation at pericentromeric sequence resulted in increased levels of Sat1 transcripts that could trigger the RNA mediated interferon response . Consistent with this model , strong derepression of Sat1 RNA from hypomethylated pericentromeres was noted in zbtb24 mutant adults ( Figure 5A and Figure 5—figure supplement 1A ) , whereas transcripts for other dispersed repetitive elements remained unchanged between mutants and wildtype ( Figure 5—figure supplement 1B ) . Increases in Sat1 transcripts correlated with levels of irf7 expression in adult zebrafish ( r = 0 . 77 ) , and upregulation of Sat1 transcripts coincided with the window of ISG induction during development ( Figure 5B–C ) . Both sense and antisense transcripts were detected in mutants using TAG-aided sense/antisense transcript detection ( TASA-TD ) strand-specific PCR ( Henke et al . , 2015 ) , suggesting the potential for derepressed Sat1 transcripts to form double stranded RNAs ( Figure 5D–E ) . To determine whether Sat1 transcripts were sufficient to activate an innate immune response , in vitro synthesized RNA corresponding to Sat1 sense and antisense transcripts were injected into wild-type embryos at the 1 cell stage . Expression of the ISGs stat1b , irf7 , irf1b and mxa was then assessed at 8 hr post fertilization . Co-injection of sense and antisense Sat1 RNA was sufficient to reproducibly cause a 2 to 4-fold upregulation in expression of these ISGs , whereas combined injection of sense and antisense control transcripts encoding a fragment of zebrafish β-actin or GFP had no effect on expression of these genes ( Figure 5F and Figure 5—figure supplement 2 ) . Lower level upregulation of some , but not all ISGs was noted when sense or antisense Sat1 transcripts were individually injected into the embryo , suggesting that the response was primarily triggered by formation of Sat1 dsRNA ( Figure 5—figure supplement 2 ) . Collectively , these results functionally link the derepression of Sat1 transcripts to the activation of the innate immune response in zbtb24 mutants . Finally , we sought to identify the specific PRR required for the interferon response in zbtb24 mutants . The RLR family of PRRs consists of two RNA helicases that signal through Mavs: Melanoma Differentiation-Associated protein 5 ( Mda5 ) and Retinoic acid-inducible gene I ( Rig-I ) . Rig-I binds 5’ triphosphorylated RNA molecules , whereas Mda5 has been implicated in the recognition of long double-stranded RNAs in the cytosol ( Crowl et al . , 2017 ) . Given that 5’ triphosphorylation of RNAs is a typical viral signature that is unlikely to be present on endogenous RNA transcripts , we reasoned that Mda5 was a more likely candidate for the receptor . To test the requirement for mda5 , we generated a seven base-pair deletion in this gene that disrupted the DEAD box helicase domain ( Figure 4—figure supplement 1C ) . This mda5mk29 allele was then introduced onto the zbtb24 mutant background , and expression of the ISGs stat1b and irf7 was examined at 3 wpf and 6 wpf . Homozygous mutation of mda5 was sufficient to restore stat1b and irf7 expression to wild-type levels in zbtb24Δ/Δ mutant larvae , suggesting that Mda5 is the primary PPR required for the response ( Figure 6A and B ) . This requirement was further validated by RNA-seq , which revealed that a broad panel of ISGs that showed elevated expression in zbtb24 single mutants were no longer upregulated in mda5mk29/mk29; zbtb24Δ/Δ double mutants ( Figure 6C ) . Taken together , these results support a model in which derepression of transcripts from hypomethylated pericentromeres triggers activation of the innate immune system through the Mda5/Mavs viral RNA recognition pathway ( Figure 6D ) . These findings identify roles for pericentromeric RNA as a trigger of autoimmunity and reveal important functions for pericentromeric methylation in suppressing the generation of these immunostimulatory transcripts . Based on these results , we propose that induction of the innate immune system is one of the earliest in vivo consequences of pericentromeric methylation loss .
In this study , we describe a viable animal model of ICF syndrome which recapitulates key phenotypic hallmarks of the disease including slow growth , facial anomalies , immunoglobulin deficiencies and reduced lifespan . Given that previous attempts to model ICF syndrome have resulted in perinatal or embryonic lethality ( Geiman et al . , 2001; Ueda et al . , 2006; Wu et al . , 2016 ) , this zebrafish model provides an important new resource for understanding ICF disease etiology during juvenile and adult life stages . In particular , zbtb24 mutant zebrafish will be useful for understanding phenotypes such as immunoglobulin deficiency , which have not been observed in mouse models and are difficult to study in cell culture systems . As in ICF syndrome , zbtb24 mutant adult zebrafish exhibited extensive loss of methylation at pericentromeric sequences . For highly repetitive sequences , methylation sensitive restriction digest followed by Southern blot remains the most effective way to assess methylation levels . By this approach , we observed increases in HpyCH4IV digestion that are consistent with up to 95% reductions in methylation at Sat1 pericentromeric repeats in zbtb24 mutants . While similar hypomethylation was observed in all adult somatic tissues that we examined , we unexpectedly observed that methylation levels in sperm from zbtb24 mutants and wildtype animals appeared comparable . This finding raises the possibility that different pathways act to control pericentromeric methylation in germ and somatic cells . Methylation levels at pericentromeric Sat1 sequences could not be quantified by ERRBS , as this technique relies on MspI restriction digest to enrich for CpG containing sequences , and zebrafish Sat1 repeats are lacking in this restriction site . Nonetheless , ERRBS analysis suggested that the general methylation landscape in human ICF syndrome and in zbtb24 mutant zebrafish is similar . Methylome analysis of primary blood from ICF patients identified methylation changes of greater than 20% at roughly 3% of examined CpG dinucleotides . Significant changes in methylation of retroviruses and other dispersed repeats were not observed in these patients ( Velasco et al . , 2018 ) . Consistent with these findings , our ERRBS analysis revealed methylation changes of greater than 20% at roughly 1 . 3% of assayed CpG dinucleotides and found methylation of dispersed repeats to be similar between wildtype and in zbtb24 mutant zebrafish . The low-level methylation changes outside of the pericentromeres observed in ICF syndrome and our mutants raise the possibility that zbtb24 may have additional modest roles in maintaining methylation at non pericentromeric sequences . One important caveat of ERRBS analysis is that CpG poor sequences can be under represented , leaving open the possibility that additional DMRs in CpG poor regions of the genome were overlooked by our approach . The progressive loss of 5mC we observe in somatic tissues between larval and adult stages implicates Zbtb24 in regulating the long-term maintenance of methylation at pericentromeric repeats . We are unaware of any developmental or methylation milestones that can account for the onset of hypomethylation around 2 wpf . Rather , we speculate that the onset of methylation loss at this stage partly reflects the need to deplete maternally loaded zbtb24 prior to unmasking of the zbtb24 mutant phenotype and partly reflects the culmination of minor methylation losses due to lower fidelity maintenance over many rounds of cell division . We note that the onset of ICF-like growth defects in zbtb24 mutant zebrafish emerged in the weeks following Sat1 methylation loss . In at least one case of ICF syndrome type 2 , growth reductions and immunodeficiency were also reported to develop with age , raising the possibility that similar progressive methylation loss may impact ICF etiology in humans ( von Bernuth et al . , 2014 ) . It is also possible that Zbtb24 functions in both maintenance and establishment of pericentromeric methylation , but that requirements for establishment are masked by maternally deposited RNA in zbtb24 mutant zebrafish lines . Unfortunately , zbtb24 homozygous mutant zebrafish are sterile , preventing the generation of the maternal-zygotic mutants required to address this question . Previous studies have suggested that ZBTB24 is a transcription factor that may act to regulate DNA methylation through transcriptional control of the ICF gene CDCA7 ( Wu et al . , 2016 ) . Consistent with this model , we observe near complete loss of cdca7 expression in zbtb24 mutants in our RNA-seq data set and by qRT-PCR ( Figure 6—figure supplement 1 ) . A more recent study in cultured human cells proposed that ZBTB24 binding might be directly involved in recruiting DNMT3B to promote gene body methylation through recognition of AGGTCCTGGCAG motifs in human cells ( Thompson et al . , 2018 ) . Analysis using Find Individual Motif Occurrences ( FIMO ) ( Grant et al . , 2011 ) , did not reveal this motif in the promoter or gene body of zebrafish cdca7 or at Sat1 sequences . In the current study , we take advantage of the progressive Sat1 methylation loss in zbtb24 mutants to identify activation of interferon signaling as one of the earliest in vivo consequences of pericentromeric hypomethylation . This phenotype cannot be attributed to defects in adaptive immunity , as the zebrafish adaptive immune system is not functional until roughly 4 wpf ( Trede et al . , 2004 ) . Induction of an interferon response has been reported in the context of global hypomethylation in cancer cell lines treated with the DNA methyltransferase inhibitor 5-azacytidine and in zebrafish mutated for the maintenance DNA methyltransferase machinery ( Chernyavskaya et al . , 2017; Chiappinelli et al . , 2015; Roulois et al . , 2015 ) . In each of these cases induction of the interferon response was attributed to massive derepression of endogenous retroviral elements . Our results are distinguished from these earlier studies in that we identify hypomethylation of pericentromeric sequences and subsequent derepression of associated satellite transcripts as a previously unappreciated trigger of innate immunity . Immunostimulatory motifs have been noted in pericentromeric RNAs derived from mouse and humans , and transcripts derived from these repeats have been observed in p53 null mouse fibroblasts following global methylation loss ( Leonova et al . , 2013; Tanne et al . , 2015 ) . However , while these studies suggest the potential for pericentromeric hypomethylation to drive an interferon response in diverse vertebrate species , experimental evidence in support of this model has been lacking . Here we demonstrate a causative link between derepression of pericentromeric RNAs and the interferon response , and identify a requirement for Mda5/Mavs in mediating the response . Our findings suggest that aberrant Sat1 transcripts derived from pericentromeric repeats trigger this response , and that these transcripts may mimic features of double stranded RNA viruses in the cytosol . This finding raises the possibility that this pathway may also recognize additional endogenous RNAs that lack viral origin . While mutation of mda5/mavs rescued the interferon response in zbtb24 mutants , mda5/mavs mutation had little impact on other ICF phenotypes observed in zbtb24 mutants . Therefore , we find it unlikely that the interferon response drives ICF etiology . Rather this response represents an additional consequence of pericentromeric hypomethylation . Hypomethylation of pericentromeric sequences is compatible with human viability and is observed in abnormal cell contexts including cancer and senescence . Massive increases in pericentromeric transcripts and upregulation of interferon genes have both been noted in cancer ( Cheon et al . , 2014; Ting et al . , 2011 ) . Our data raise the possibility that pericentromeric hypomethylation and subsequent derepression of associated RNAs represents an important but underappreciated trigger of autoimmunity in a variety of disease states .
Zebrafish husbandry and care were conducted in full accordance with animal care and use guidelines with approval by the Institutional Animal Care and Use Committees at Memorial Sloan Kettering Cancer Center and the University of Georgia . Zebrafish were raised under standard conditions at 28° C . Wild-type lines were of the AB background . All mutant alleles are summarized in Supplementary file 1 . TALEN sequences were selected using Targeter 2 . 0 software ( Doyle et al . , 2012 ) . TAL repeat assembly was achieved using the Golden Gate assembly method , and assembled repeats were integrated into the GoldyTALEN scaffold ( Bedell et al . , 2012; Cermak et al . , 2011 ) . Assembled vectors served as templates for in vitro mRNA transcription using the T3 mMessage mMachine kit ( Ambion ) according to manufacturer’s instructions . 50–100 pg mRNA was injected into wild-type embryos at the one-cell stage . Injected embryos were raised to adulthood and F1 progeny were screened for germline transmission of mutations as previously described ( Li et al . , 2015 ) . Primers used for detection of mutations and subsequent genotyping are included in Supplementary file 1 . Target selection for CRISPR/Cas9 mediated mutagenesis was performed using CHOPCHOP ( Labun et al . , 2016 ) . sgRNA templates were generated either by cloning into pT7-gRNA as described by Jao et al . ( 2013 ) or using the oligo-based approach described in Gagnon et al . , 2014 and Burger et al . ( 2016 ) . All template oligos are listed in Supplementary file 3 . sgRNAs were in vitro transcribed from their respective templates using T7 RNA polymerase ( Promega ) as per manufacturer protocol . Cas9 RNA was in vitro transcribed from the pT3TS-nls-zCas9-nls plasmid ( Jao et al . , 2013 ) using the T3 mMessage mMachine Kit ( Ambion ) . For mutagenesis , 200–400 ng of sgRNA and ~500 ng of Cas9 mRNA were co-injected into wild-type embryos at the one-cell stage . Injected embryos were raised to adulthood , and F1 progeny were screened for germline transmission of mutations as previously described ( Li et al . , 2015 ) . Primers used for detection of mutations and subsequent genotyping are included in Supplementary file 1 . All bright field imaging of zebrafish larvae and adult was performed using Olympus MVX10 with CellSens Standard software . Standard-length was documented using ImageJ as defined in Parichy et al . ( 2009 ) . Photoshop ( Adobe ) adjustments to brightness and contrast were equally applied to all images of whole zebrafish in order to improve visualization . Adult zebrafish at 6 months were sacrificed with a combination of tricaine ( Sigma-Aldrich , CAS number 886-86-2 ) and rapid chilling . Whole kidneys were dissected using forceps and placed in 0 . 9 × PBS/5% FCS . Manual disaggregation using a P1000 pipette resulted in single cell suspensions . Cells were filtered over a 40 μm nylon mesh filter , and resuspended in PBS/FCS to give a final concentration of 100 , 000 cells/μl . FACS sorting of single cells were analyzed for forward/side scatter profiles . FACS data were analyzed using FloJo software . For Hematoxylin and Eosin ( H and E ) staining , adult zebrafish were fixed in 10% Neutral Buffered Formalin for 48 hr . Zebrafish were then decalcified in 0 . 5 M EDTA for 24 hr . After decalcification , fish were incubated overnight in 70% Ethanol before embedding in paraffin blocks . Sections were stained with H and E according to standard procedures . Adult zebrafish at 8 months were sacrificed with a combination of tricaine and rapid chilling . Whole testis was dissected using forceps and crushed in 100 ul of PBS . For determining sperm-count , sperm samples were diluted 1:20 for each fish . 10 ul of the diluted sample was then loaded onto a hemocytometer ( Bright-Line , Hauser Scientific ) for counting . The volume over the central counting area is 0 . 1 mm3 or 0 . 1 microliter . Average number of sperm counted over the central counting area was multiplied by 10000 to obtain the number of sperm/ml of the diluted sample . The obtained value was multiplied by the dilution factor to obtain the final sperm count . For Southern blot analysis , 1 μg of purified total genomic DNA was digested with the indicated methylation sensitive restriction enzyme , fractionated by electrophoresis through a 0 . 9% agarose gel and transferred to nylon membrane . Sperm DNA was isolated from sperm samples collected by crushing dissected testes in PBS . Probes were PCR amplified using primers in Supplementary file 2 and radiolabeled with 32P-dCTP using RediprimeTM II Random Prime Labelling System ( Amersham ) according to manufacturer protocol . Hybridization signals were imaged and analyzed using a Typhoon phosphorimager ( GE Life Sciences ) . Signal intensities were measured using ImageJ . Methylation changes at Sat1 was quantified as a ratio of the intensity of the unmethylated/methylated blot regions as indicated in the respective blot . HypCH4IV was selected for Sat1 methylation analysis over the more traditional MspI/HpaII isoschizomer pair because Sat1 sequences lack the CCGG sites that are recognized by these enzymes . ChIP was performed as described in Lindeman et al . 2009 with modifications . Briefly , zebrafish juveniles at 1 month were euthanized using tricaine . Chromatin was prepared from euthanized fish by lysing flash frozen samples using an automated pulverizer ( Covaris ) and crosslinking using 1% Formaldehyde for 5 mins . Chromatin shearing was performed using a Covaris S220 sonicator using the following conditions: 1 ml tubes with total chromatin from each fish in buffer containing 1% SDS were sonicated using peak intensity power of 140 , duty factor of 5 . 0 and 200 cycles per burst , for 14 min for zbtb24+/+ and 6 min for zbtb24Δ/Δ . Shearing was monitored using 1% agarose gel . To provide standardized input for each ChIP experiment , chromatin was diluted to A260 = 0 . 2 . For each ChIP , 2 μg antibody per 10 μl Dynabeads and 100 μl chromatin was incubated overnight at 4°C . Following antibodies were used in this study: anti-H3K9me3 antibody ( abcam ab8898 ) , anti-H3K27me3 ( Millipore 07–449 ) , anti-H3 ( abcam 1791 ) and IgG control ( abcam ab15008 ) . After elution , ChIP DNA and input controls were purified using QIAquick PCR purification kit ( Qiagen ) . Eluted DNA was analyzed by qPCR using primers targeting Sat1 ( Supplementary file 2 ) . 50 ng of high quality genomic DNA was prepared from fin tissue from 6-month-old adult zebrafish DNA was digested with MspI and bisulfite converted using the EZ DNA methylation kit ( zymo ) as in Garrett-Bakelman et al . ( 2015 ) . Bisulphite conversion rates ( calculated using non-CpG methylation conversion rates ) ranged from 99 . 6% to 99 . 7% for all samples ( Figure 2—figure supplement 2C ) . Amplified libraries were sequenced on the Hiseq2500 platform using a minimum of single-read 51 cycles . ERRBS data were filtered for sequence adapters , limited to the first 29 bp of the read ( Boyle et al . , 2012 ) , and mapped to the zebrafish genome ( danRer7 ) using BSmap ( v 2 . 90 ) ( Xi and Li , 2009 ) . Other than limiting to the first 29 bp all other BSmap parameters were the defaults . Methylation scores were calculated as the number of unconverted reads divided by the number of total reads at each CpG site . DMRs were called as described in Park and Wu ( 2016 ) . DMRs with at least a 0 . 2 change in methylation were determined using DSS ( delta = 0 . 2 , p . threshold = 0 . 01 ) . CallDMR function in DSS was used with default parameters except for p . threshold and delta as specified . Sat1 sequences are deficient in MspI sites , and are therefore not included in ERRBS data . For qRT-PCR , total RNA was isolated using Trizol ( Invitrogen ) and precipitated with isopropanol . RNA used for assaying expression of repeat sequences subsequently was treated with DNase using TURBO DNA-free Kit ( Ambion ) prior to analyses . RNA was converted to cDNA using GoScript Reverse Transcriptase Kit ( Promega ) and Real Time PCR was performed using an Applied Biosystems 7500 PCR Machine . Analysis was performed using the 2–ΔΔCt method , with relative mRNA levels of all transcripts normalized to β-actin1 or 18S . All primer sequences are listed in Supplementary file 2 . For Northern blot analysis , total RNA was extracted with using Trizol ( Invitrogen ) . 2 μg of RNA was subjected to electrophoresis on 1% agarose gel and transferred to Amersham Hybond-N+ membrane ( GE Healthcare ) . The membrane was probed with 32P-dCTP radiolabeled Sat1 DNA probe at 42°C . Hybridization signals were imaged and analyzed using a Typhoon phosphorimager ( GE Life Sciences ) . TAG-aided sense/antisense transcript detection ( TASA-TD ) strand-specific PCR was performed as described by ( Henke et al . , 2015 ) . Oligos used are listed in Supplemental file 3 . After RiboGreen quantification and quality control by Agilent BioAnalyzer , 500 ng of total RNA underwent polyA selection and TruSeq library preparation according to instructions provided by Illumina ( TruSeq Stranded mRNA LT Kit ) , with 8 cycles of PCR . Samples were barcoded and run on a HiSeq 2500 High Output in a 50 bp/50 bp paired end run , using the TruSeq SBS v4 Kit ( Illumina ) . An average of 45 . 3 million paired reads was generated per sample . The percent of mRNA bases averaged 62 . 8% . For single-mutant RNA-seq analysis presented in Figure 3 , reads were mapped to the Zebrafish genome ( danRer7 ) using the rnaStar aligner v2 . 5 . 0a ( Dobin et al . , 2013 ) . We used the two-pass mapping method outlined in Engström et al . ( 2013 ) . The first mapping pass used a list of known annotated junctions from Ensemble . Novel junctions found in the first pass were then added to the known junctions and a second mapping pass was done ( on the second pass the RemoveNoncanoncial flag was used ) . Expression counts ( counts per million , cpm ) were computed from the mapped reads using HTSeq v0 . 5 . 3 ( Anders et al . , 2015 ) and Ensemble D . rerio v79 gene annotations . Normalization and differential expression was performed using DESeq ( Anders and Huber , 2010 ) . For RNA-seq analysis presented in Figure 6 , raw RNA-seq FASTQ reads were trimmed for adapters and preprocessed to remove low-quality reads using Trimmomatic v0 . 33 ( arguments: LEADING:3 TRAILING:3 MINLEN:36 ) ( Bolger et al . , 2014 ) prior to mapping to the Danio rerio GRCz10 reference genome assembly . Reads were mapped using TopHat v2 . 1 . 1 ( Kim et al . , 2013 ) supplied with a reference General Features File ( GFF ) to the Danio rerioGRCz10 reference genome assembly , and with the following arguments: -i 10 -I 5000 --library-type fr-firststrand . Gene expression was estimated using Cuffquant ( a tool from Cufflinks v2 . 2 . 1 ) , with following arguments --library-type fr-firststrand . Expression level were normalized in FPKM units by Cuffnorm ( a tool from Cufflinks v2 . 2 . 1 ) , with following arguments --library-type fr-firststrand . Zebrafish embryos were treated with 5-aza-dC ( Sigma-Aldrich ) to the final concentration of 25 uM or 50 uM within the first 2 hr post fertilization , when zebrafish are sensitive to 5-aza-dC treatments as described in Martin et al . ( 1999 ) . At 24hpf , total RNA was collected for expression analysis . At 24hpf , genomic DNA was also collected and digested with methylation sensitive enzyme , HpaII , to test for global DNA hypomethylation . Sat1 RNA and control RNAs were in vitro transcribed using Riboprobe in vitro transcription systems ( Promega ) . Oligos to amplify the DNA template for in vitro transcription are included in Supplementary file 3 . Sense and anti-sense transcripts were transcribed in vitro using the T3 and T7 RNA polymerases respectively . RNA was purified illustra MicroSpin G-50 Columns ( GE Healthcare ) and 50 ng of sense and antisense RNA was co-injected into zebrafish embryos at the 1 cell stage . The Student unpaired 2-tailed t-test was used for statistical analysis unless specified otherwise . Statistical analysis was performed using GraphPad PRISM software . All ERRBS and RNA-Seq data reported in this paper have been deposited in GEO under the accession GSE116360 . | Cells package DNA into structures called chromosomes . When cells divide , each chromosome duplicates , and a structure called a centromere initially holds the copies together . The sequences of DNA on either side of the centromeres are often highly repetitive . In backboned animals , this DNA normally also has extra chemical modifications called methyl groups attached to it . The role that these methyl groups play in this region is not known , although in other DNA regions they often stop the DNA being ‘transcribed’ into molecules of RNA . The cells of people who have a rare human genetic disorder called ICF syndrome , lack the methyl groups near the centromere . The methyl groups may also be lost in old and cancerous cells . Researchers often use ‘model’ animals to investigate the effects of DNA modifications . But , until now , there were no animal models that lose methyl groups from the DNA around centromeres in the same way as seen in ICF syndrome . Rajshekar et al . have developed a new zebrafish model for ICF syndrome that loses the methyl groups around its centromeres over time . Studying the cells of these zebrafish showed that when the methyl groups are missing , the cell starts to transcribe the DNA sequences around the centromeres . The resulting RNA molecules appear to be mistaken by the cell for viral RNA . They activate immune sensors that normally detect RNA viruses , which triggers an immune response . The new zebrafish model can now be used in further studies to help researchers to understand the key features of ICF syndrome . Future work could also investigate whether the loss of methyl groups around the centromeres plays a role in other diseases where the immune system attacks healthy tissues . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"research",
"communication"
] | 2018 | Pericentromeric hypomethylation elicits an interferon response in an animal model of ICF syndrome |
Many stem cells divide asymmetrically in order to balance self-renewal with differentiation . The essence of asymmetric cell division ( ACD ) is the polarization of cells and subsequent division , leading to unequal compartmentalization of cellular/extracellular components that confer distinct cell fates to daughter cells . Because precocious cell division before establishing cell polarity would lead to failure in ACD , these two processes must be tightly coupled; however , the underlying mechanism is poorly understood . In Drosophila male germline stem cells , ACD is prepared by stereotypical centrosome positioning . The centrosome orientation checkpoint ( COC ) further serves to ensure ACD by preventing mitosis upon centrosome misorientation . In this study , we show that Bazooka ( Baz ) provides a platform for the correct centrosome orientation and that Baz-centrosome association is the key event that is monitored by the COC . Our work provides a foundation for understanding how the correct cell polarity may be recognized by the cell to ensure productive ACD .
Asymmetric division of adult stem cells that produces a self-renewing stem cell and a differentiating daughter cell is crucial for tissue homeostasis in diverse systems ( Morrison and Kimble , 2006 ) . Disruption of this balance is postulated to underlie many pathological conditions , including tumorigenesis/tissue hyperplasia ( due to excess self-renewal ) and tissue degeneration/aging ( due to excess differentiation ) . Intensive investigation has revealed the mechanisms that polarize cells and orient the division plane; however , less is known about how cells might respond to perturbation of cell polarity and whether/how cells might ensure that cell division occurs only after the establishment of correct polarity . A mechanism to coordinate the timing of two potentially independent events during the cell cycle is defined as a checkpoint . The spindle position checkpoint ( SPOC ) in budding yeast is a prominent example of a checkpoint that coordinates cell division with cell polarity; mitotic exit is delayed by the activity of SPOC if the spindles are not correctly oriented in a manner to ensure equal segregation of chromosomes into the mother and daughter cells ( Pereira and Yamashita , 2011 ) . A similar spindle orientation checkpoint mechanism is also reported in fission yeast ( Gachet et al . , 2006 ) . Despite the importance of asymmetric divisions in the development of multicellular organisms , the potential checkpoint mechanisms that ensure asymmetric cell divisions , similar to the SPOC in the budding yeast , are poorly defined . We have established Drosophila male germline stem cells ( GSCs ) as a model to study the checkpoint that coordinates polarization of cells and cell division ( Cheng et al . , 2008; Inaba et al . , 2010; Yuan et al . , 2012 ) . Drosophila male GSCs divide asymmetrically , producing one stem cell and one differentiating cell , the gonialblast ( GB ) . Asymmetric stem cell division is achieved by stereotypical positioning of the mother and daughter centrosomes in order to orient the spindle perpendicularly to the hub cells , the major component of the stem cell niche ( Yamashita et al . , 2003 , 2007 ) . Stereotypical centrosome behavior that occurs in preparation for asymmetric cell division has been described in other stem cell systems ( Rebollo et al . , 2007; Rusan and Peifer , 2007; Wang et al . , 2009; Conduit and Raff , 2010; Januschke et al . , 2011; Lu et al . , 2012; Salzmann et al . , 2014 ) , suggesting the evolutionarily conserved nature of the process . Asymmetric GSC division is further ensured by the centrosome orientation checkpoint ( COC ) that prevents mitotic entry in the presence of incorrectly oriented centrosomes ( Figure 1A ) ( Cheng et al . , 2008; Inaba et al . , 2010; Yuan et al . , 2012 ) . Upon sensing the centrosome misorientation , COC is activated to prevent mitotic entry ( Figure 1A ) . Thus , the defective COC can be suggested by the presence of misoriented spindles . We have shown that the centrosomal protein Cnn and a polarity kinase Par-1 are critical component of the COC , defects of which leading to high frequency of spindle misorientation ( Inaba et al . , 2010; Yuan et al . , 2012 ) . 10 . 7554/eLife . 04960 . 003Figure 1 . The apical centrosome associates with the Baz Patch . ( A ) The centrosome orientation in GSCs and the function of COC . ( B ) An example of an apical testis tip showing the Baz patch and centrosomes . The apical centrosome often associates with the Baz patch ( open arrow ) . The Baz patch ( solid arrow ) remains in GSCs with misoriented centrosomes . Centrosomes are indicated with arrowheads . The insets show Baz patches with or without the centrosome . ( B′ ) Baz-GFP only . Bar: 10 µm . The colored text indicates the fluorescence pseudocolor in the images in this and subsequent figures . The γ-tubulin staining indicates the centrosome . The Vasa staining indicates the germ cells . The hub is denoted with an asterisk . ( C ) The Baz patch is a small structure that is located on the GSC-hub interface . The arrowhead in ( C , C′ ) indicates the Baz patch stained with anti-Baz ( red ) . The yellow dotted line in ( C'' ) indicates the GSC-hub interface illuminated by GFP-E-cadherin ( DEFL , green ) expressed in the germline ( nos-gal4>UAS-DEFL ) . ( D ) Schematic describing the definition of centrosome orientation and Baz-centrosome docking . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 003 The physical basis of correct centrosome orientation monitored by the COC remains a mystery . In the case of the spindle assembly checkpoint ( SAC ) , the lack of microtubule attachment to the kinetochore ( or tension at the kinetochore ) is sensed as defective spindle assembly , triggering SAC activation to halt mitotic progression ( Musacchio and Salmon , 2007 ) . In the operation of the COC , what is sensed as correct or incorrect centrosome orientation to inactivate or activate the COC remains unknown . Here , we show that Bazooka ( Baz ) /Par-3 , a well-established polarity protein and a known substrate of Par-1 kinase , forms a small subcellular structure that anchors the centrosome right before mitotic entry . We provide evidence that the association between Baz and the centrosome is the key event that is interpreted to indicate ‘correct centrosome orientation’ by GSCs . We further show that Par-1-dependent phosphorylation of Baz is critical for GSC spindle orientation . Our study provides a framework of the mechanism by which GSC sense correct cell polarity .
Baz/Par-3 , which is a known physiological substrate of Par-1 , contributes to cell polarity and spindle orientation in diverse systems ( Watts et al . , 1996; Benton and St Johnston , 2003; Siller and Doe , 2009 ) . Because we previously found that Par-1 is a critical component of the COC ( Yuan et al . , 2012 ) , we examined the role of Baz in the centrosome orientation and/or COC . Baz has been reported to localize at the hub-GSC interface along with E-cadherin following overexpression in the germline ( nos > Baz-GFP ) ( Leatherman and Dinardo , 2010 ) , which was confirmed by using independent UAS-Baz-YFP construct ( see below ) . However , closer inspection using antibody staining and Flytrap Baz-GFP that expresses endogenous levels of Baz [CC01941 ( Kelso et al . , 2004; Buszczak et al . , 2007 ) ] revealed that Baz forms foci at the hub-GSC interface ( referred to as the ‘Baz patch’ hereafter ) , instead of entirely colocalizing with E-cadherin ( Figure 1B , C ) . The Baz patch is a small structure , with a size of approximately 1 . 5 µm , and this patch is considerably smaller than the GSC-hub interface that is marked by E-cadherin ( 4–6 µm ) ( Figure 1C ) . We noticed that the Baz patch was often closely associated with the apical centrosome ( 68 . 8 ± 2 . 2% of total GSCs; Figure 1C , arrowheads ) . We termed this close association of the Baz patch and the centrosome ‘Baz-centrosome docking’ . Baz-centrosome docking is a more specific criteria compared to centrosome orientation: ∼90% of total GSCs had ‘oriented’ centrosomes , a category that can be further subdivided into GSCs with ‘oriented , but not docked’ centrosomes ( ∼20% ) and those with ‘oriented and docked’ centrosomes ( ∼70% ) ( Figure 1D ) . The remaining ∼10% of total GSCs had misoriented centrosomes as reported previously ( Figure 1D ) ( Cheng et al . , 2008; Roth et al . , 2012; Yuan et al . , 2012 ) . Using a combination of multiple cell cycle markers , we found that Baz-centrosome docking is cell cycle dependent , reaching a peak of ∼80% during late G2 phase . The GSC cell cycle was judged by the following criteria ( Figure 2 ) . ( 1 ) A short pulse of ex vivo BrdU incorporation was used to detect cells in S phase . GSCs that were positive for BrdU were always connected with their differentiating daughters ( GBs ) , which were also positive for BrdU in synchrony . This finding suggests that GSCs ( and GBs ) enter S phase prior to cytokinesis and that the G1 phase is extremely short . ( 2 ) As cells progress into G2 phase , GSCs started to accumulate Dap , a Cip/Kip family CDK inhibitor , in the nucleus , as reported previously ( Meyer et al . , 2002 ) . Thus , late G2 GSCs were detected as nuclear Dap-positive cells . ( 3 ) Prior to nuclear Dap accumulation , GSCs that completed cytokinesis , negative for BrdU and had not accumulated nuclear Dap were judged as early G2 phase . 10 . 7554/eLife . 04960 . 004Figure 2 . Baz-centrosome docking is cell cycle-dependent . ( A ) A representative image of an undocked centrosome in S phase . The arrow indicates the centrosome , and the arrowhead indicates the Baz patch . The inset shows a magnified view . ( B ) A representative image of a late G2 GSC with nuclear Dap ( white ) . The arrowhead indicates the centrosome docked to the Baz patch . Spd-2 staining indicates the centrosome ( red ) . ( B′ ) Dap-myc only . ( C ) A representative image of a mitotic GSC . The yellow dotted line indicates a Baz crescent along the hub-GSC interface . At this point , Baz-centrosome docking cannot be assessed because Baz does not localize as foci ( ‘N/A’ in panel D ) . ( D ) The frequency of Baz-centrosome docking during the cell cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 004 Using these criteria , we correlated the cell cycle stages with the Baz-centrosome docking status . During S phase , the frequency of Baz-centrosome docking was low ( ∼20% , Figure 2A , D ) . GSCs in early G2 phase maintained low Baz-centrosome docking ( Figure 2D ) ; however , once GSCs reached late G2 phase , Baz-centrosome docking was elevated dramatically , reaching approximately 80% ( Figure 2B , D ) . As GSCs entered mitosis , Baz was broadly distributed to the hub-GSC interface instead of being confined as a small patch ( Figure 2C ) . This distribution resembled the Baz crescent that was observed in Drosophila neuroblasts ( Schober et al . , 1999; Wodarz et al . , 1999 ) . At this point in the cell cycle , Baz-centrosome docking could no longer be defined due to diffuse Baz localization ( Figure 2D , N/A ) . Together , these data demonstrate that Baz-centrosome docking is a cell cycle-dependent event that occurs just before mitotic entry . The tight association ( docking ) of the Baz patch and the centrosome just before mitotic entry led us to hypothesize that such association may be the physical basis for ‘correct centrosome orientation’ that allows GSCs to enter mitosis . We speculated that the Baz patch may provide a docking site for the centrosome to correctly orient and that once such docking occurs , GSCs may interpret this docking as ‘correct centrosome orientation’ . To address these possibilities , we first examined the potential role of Baz in centrosome orientation and COC function . We used centrosome and spindle misorientation as the major criteria for assessment of the function of centrosome orientation mechanism and the COC . If the centrosome orientation mechanism is defective , a high frequency of centrosome misorientation would result , although this scenario does not necessarily lead to spindle misorientation . If the COC is intact , GSCs with misoriented centrosomes would halt cell cycle progression prior to mitotic entry , resulting in a low frequency of spindle misorientation , even if the centrosomes are highly misoriented ( Figure 1A ) . If the COC is also defective , then GSCs with misoriented centrosomes would enter mitosis unchecked , resulting in a high frequency of misoriented spindles . We first attempted to knock down Baz using two independent RNAi lines ( validated in Figure 3—figure supplement 1 ) . In control GSCs , centrosomes were oriented in most of the GSCs ( up to ∼10% centrosome misorientation , Figure 3A , G ) as reported previously ( Yamashita et al . , 2003 , 2007 ) . On the contrary , RNAi-mediated knockdown of Baz caused a high frequency of centrosome misorientation in interphase GSCs ( Figure 3B , G ) . This result suggests that Baz is required for normal centrosome orientation in GSCs . Combined with the observation that the centrosome docks to the Baz patch , these results indicate that the Baz patch provides a physical platform for GSC centrosome association to achieve correct centrosome orientation . Baz RNAi caused a minor but statistically significant increase in the spindle misorientation ( Figure 3C , D , G ) , suggesting that Baz is also required for COC at least partially . 10 . 7554/eLife . 04960 . 005Figure 3 . Baz is required for centrosome orientation . ( A , B ) Control ( A ) and Baz RNAi ( B ) testes showing GSCs in interphase . Arrows indicate centrosomes . GSCs are indicated by broken lines . The hub is denoted by an asterisk . Bar: 10 µm . ( C , D ) Control ( C ) and Baz RNAi ( D ) GSCs in mitosis . Arrows indicate spindle poles . ( E ) Overexpressed Baz-YFP ectopically localizes to the lateral cortex of GSCs . Arrows indicate the ectopic patch docking to the centrosome . ( F ) Mitotic GSCs with misoriented spindles upon overexpression of Baz-YFP . ( G ) Frequencies of centrosome ( % of total GSC ) and spindle ( % of mitotic GSC ) misorientation upon Baz RNAi or Baz-YFP overexpression . N > 300 GSCs were scored for centrosome orientation , and N > 30 mitotic GSCs were scored for spindle orientation . ( H ) A model for Baz-centrosome docking and mitotic entry in control , Baz RNAi , and Baz overexpression . ( I ) Mitotic index of GSCs after incubation with or without colcemid for 4 . 5 hr in indicated genotypes . Increased mitotic index in the presence of colcemid indicates defective COC . p value indicates the statistical significance in an increase in mitotic index in the presence of colcemid . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 00510 . 7554/eLife . 04960 . 006Figure 3—figure supplement 1 . GFP fluorescent quantification of Baz patch ( Baz-GFP Flytrap ) upon knockdown of Baz ( GD1384 and JF01079 ) . The background ( cytoplasm signal in the same cell for same field ) was subtracted from each patch . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 006 To further explore the potential function of Baz , we overexpressed Baz in the germline ( nos > Baz-YFP ) . Upon overexpression , Baz often formed ectopic patches outside the hub-GSC interface ( Figure 3E , arrow ) and often broadly localized to the hub-GSC interface ( Figure 3F ) as reported previously ( Leatherman and Dinardo , 2010 ) . Ectopic Baz patches were often associated with misoriented centrosomes ( Figure 3E , inset ) , suggesting that the Baz patch has the ability to dock to centrosomes ectopically , even outside the hub-GSC interface . Strikingly , Baz overexpression led to a high frequency of spindle misorientation as well as centrosome misorientation ( Figure 3F , G ) . These data indicate that the Baz patch can ectopically dock the centrosome , and this docking is sufficient to satisfy the requirement of the COC , allowing GSCs to enter mitosis with misoriented spindles ( Figure 3H ) . The status of COC activity was further assessed by a recently developed assay , in addition to scoring the centrosome and spindle misorientation . Recently , we showed that the treatment of testes with colcemid , a microtubule depolymerizing agent , can serve as a sensitive assay to monitor the COC activity ( Venkei and Yamashita , 2015 ) . When testes were incubated with colcemid , spermatogonia , the differentiating progeny of GSCs , arrest in mitosis due to the activation of the spindle assembly checkpoint ( SAC ) . On the contrary , GSCs arrest in G2 phase of the cell cycle instead of mitosis due to centrosome misorientation and activation of COC , which operates prior to SAC . In the absence of functional COC ( such as in par-1 mutant ) , however , colcemid treatment leads to SAC-mediated mitotic arrest by bypassing COC-mediated G2 arrest ( Figure 3I ) . Thus , accumulation of mitotic GSCs in the presence of colcemid serves as a sensitive readout of defective COC . By using this method , we assessed the activity of COC in control , Baz RNAi and Baz-overexpressing GSCs ( Figure 3I ) . Indeed , the results confirmed the model obtained by the scoring of centrosome/spindle orientation: ( 1 ) Baz RNAi GSCs maintain relatively intact COC activity , and ( 2 ) Baz-YFP overexpression allows GSCs to enter mitosis , presumably through ectopic Baz-centrosome docking , which inactivates COC activity . In summary , the data presented here point to a model in which the Baz patch is a platform for centrosome anchoring , and the Baz-centrosome association ( docking ) is the cellular process that passes the COC to permit mitotic entry ( Figure 3H ) . Considering the ( partial ) requirement of Baz in COC , we speculate that Baz may play dual roles in COC , depending on the centrosome-docking status: it might contribute to activation of COC when undocked with centrosomes , whereas it might inactivate COC when docked with centrosomes . Similar to Baz overexpression , overexpression of a dominant-negative E-cadherin ( dCR4h ) leads to a high frequency of misoriented spindles ( Inaba et al . , 2010 ) . dCR4h lacks the extracellular domain and thus cannot engage in the homotypic interaction ( Oda and Tsukita , 1999 ) . As a result , overexpressed dCR4h localizes to the entire cortex of GSCs instead of being limited to the hub-GSC interface ( Inaba et al . , 2010 ) . Therefore , we speculated that the cytoplasmic domain of E-cadherin may participate in anchoring of the centrosomes , and ectopic anchoring between dCR4h and the centrosome may satisfy the conditions required to inactivate the COC to allow mitotic progression . Because Baz/Par-3 is recruited to adherens junctions ( Le Borgne et al . , 2002 ) , we speculated that Baz might function downstream of E-cadherin to anchor the centrosomes . To test this possibility , we first investigated the effect of dCR4h expression on Baz localization . GSC clones that express dCR4h were induced by heat-shock treatment ( hs-FLP , nos > stop > gal4 , UAS-dCR4h-GFP , UAS-GFP ) . In control GFP-negative GSCs , the Baz patch was observed as described above ( Figure 4A , arrow ) . On the contrary , in GSCs that expressed dCR4h , Baz broadly localized along the hub-GSC interface , as opposed to concentrated localization as a patch ( Figure 4A , broken lines ) , and the frequency of Baz patch-positive GSCs was significantly reduced ( Figure 4B ) . Furthermore , we observed that overexpression of dCR4h often formed ectopic Baz patches away from the hub-GSC interface , and such a Baz patch was associated with the centrosomes ( approximately 10% of GSCs expressing dCR4h , Figure 4C ) . In addition , Baz patch was undetectable in most of GSC clones that are homozygous for E-cadherin loss of function mutation ( shg10469 ) confirming that Baz localization depends on E-cadherin ( Figure 4D , E ) . These data indicate that the cytoplasmic tail of E-cadherin indeed recruits Baz , which in turn anchors the centrosome . 10 . 7554/eLife . 04960 . 007Figure 4 . Baz functions downstream of E-cadherin during centrosome orientation . ( A ) An apical tip of the testis containing clones ( GFP+ ) that express a dominant-negative form of E-cadherin ( dCR4h ) . The arrow indicates normal Baz patch in a control GSC that does not express dCR4h . Broken lines indicate diffused Baz localization upon expression of dCR4h . ( A′ ) anti-Baz only . The hub is denoted by an asterisk . Bar: 10 µm . ( B ) Frequency of GSCs with the Baz patch in control vs dCR4h-expressing GSCs . N > 100 GSCs were scored . ( C ) An example of ectopic Baz patch away from the hub-GSC interface ( arrow ) that docks the centrosome upon expression of dCR4h . ( D ) Frequency of Baz patch in control vs shg10469 ( E-cadherin loss of function allele ) GSC clones . ( E ) An example of GFP- , shg10469 clone without Baz patch ( yellow line ) . Control ( GFP+ ) GSCs with Baz patch ( arrowheads ) are juxtaposed ( cyan lines ) . ( F ) An example of a misoriented spindle upon expression of dCR4h . Arrows indicate spindle poles . ( G ) An example of oriented spindles in GSCs that express both dCR4h and Baz RNAi . ( H ) Frequencies of centrosome ( % of total GSC ) and spindle ( % of mitotic GSC ) misorientation upon expression of dCR4h in the presence or absence of Baz RNAi . N > 300 GSCs were scored for centrosome orientation , and N > 30 mitotic GSCs were scored for spindle orientation . ( I ) Mitotic index of GSCs after incubation with or without colcemid for 4 . 5 hr indicate genotypes . Increased mitotic index in the presence of colcemid indicates defective COC . p value indicates the statistical significance in an increase in mitotic index in the presence of colcemid . ( J ) A model for Baz-centrosome docking and mitotic entry in dCR4h-expressing GSCs with or without Baz RNAi . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 007 If this model is correct , the ability of dCR4h to misorient the spindles may rely on the presence of Baz . To test this possibility , we examined the effect of Baz knockdown on dCR4h-mediated centrosome/spindle misorientation . Overexpression of dCR4h alone led to a high frequency of spindle misorientation , as reported previously ( Figure 4F , H ) ( Inaba et al . , 2010 ) ; however , co-expression of Baz RNAi with dCR4h significantly suppressed spindle misorientation ( Figure 4G , H ) . These results clearly demonstrate that the spindle misorientation caused by dCR4h overexpression is due to ectopic recruitment of Baz , which in turn anchors the centrosomes . By using colcemid treatment , we further assessed COC status under these conditions . Expression of dCR4h resulted in mild abrogation of COC ( Figure 4I ) , as predicted by high frequency of spindle misoreintation upon expression of dCR4h ( Figure 4H ) . Interestingly , despite rescue of spindle misorientation by Baz RNAi , COC defect was not rescued under these conditions ( Figure 4I ) . We speculate that Baz RNAi reduces the level of ectopic Baz due to dCR4h overexpression , leading to reduced level of spindle misorientation , without rescuing COC defect . Taken together , these results show that Baz functions downstream of E-cadherin to anchor the centrosome , and that E-cadherin-Baz-centrosome interaction is the critical aspect in satisfying the COC ( Figure 4J ) . The above data suggest that Baz is a critical component of the GSC centrosome orientation and that Baz-centrosome docking is the physical basis that is monitored by the COC . Our previous study demonstrated that Par-1 , a physiological kinase of Baz , is a critical component of the COC ( Yuan et al . , 2012 ) . The mechanism by which Par-1 mediates sensing of the correct centrosome position is unknown . Therefore , we set out to examine whether phosphorylation of Baz by Par-1 is important for COC function . Two conserved serine residues of Baz protein , serine 151 ( S151 ) and serine 1085 ( S1085 ) , are known to be phosphorylated by Par-1 ( Krahn et al . , 2009 ) . To begin to address the relationship between Baz and Par-1 in the COC , we first examined the phosphorylation status of Baz using phosphorylation-specific antibodies against phospho-S151 ( pS151 ) and phospho-S1085 ( pS1085 ) ( Krahn et al . , 2009 ) . We detected pS151 ( Figure 5A ) , but not pS1085 , at the Baz patch; thus , we focused only on pS151 function in subsequent experiments . Interestingly , phosphorylation of the Baz patch is dependent on the cell cycle and/or centrosome orientation status . Baz phosphorylation was considerably weaker in GSCs in early G2 phase with undocked but oriented centrosomes ( Figure 5B , D ) compared to that in GSCs in late G2 phase when the centrosome is docked to the Baz patch ( Figure 5A , D ) . Once cells entered mitosis , Baz phosphorylation became undetectable ( Figure 5C ) , as the Baz patch diffused ( Figure 2C ) . Using Par-1 RNAi , which was previously validated in GSCs ( Yuan et al . , 2012 ) , we found that the pS151 signal intensity was significantly reduced in Par-1 RNAi GSCs ( Figure 5E ) , suggesting that Par-1 is required for phosphorylation of Baz-S151 , as has been shown in other cell types ( Benton and St Johnston , 2003; Krahn et al . , 2009 ) . 10 . 7554/eLife . 04960 . 008Figure 5 . Par-1-dependent Baz-S151 phosphorylation is required for the centrosome orientation checkpoint . ( A–C ) Phosphorylation of Baz-S151 was monitored by a phospho-S151-specific antibody during cell cycle . The Baz-patch is indicated by the arrow . Red indicates Baz-pS151 , green indicates Baz-GFP , and blue indicates Vasa . Bar: 10 μm . The hub is denoted with an asterisk . ( D ) The quantification of Baz-S151 phosphorylation levels . The signal was normalized by Baz-GFP ( pixel intensity of pS151 staining was divided by the pixel intensity of Baz-GFP ) . The background ( cytoplasm signal in the same cell ) was subtracted from both pS151 and Baz-GFP prior to calculation . N > 10 GSCs were scored . ( E ) The quantification of Baz-S151 phosphorylation level in control vs Par-1 RNAi GSCs . N > 10 GSCs was scored . ( F ) Frequencies of centrosome ( % of total GSC ) and spindle ( % of mitotic GSC ) misorientation in control vs Par-1 RNAi GSCs with or without expression of wild type Baz , Baz-SA or Baz-SE . N > 300 GSCs were scored for centrosome orientation , and N > 30 mitotic GSCs were scored for spindle orientation . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 008 To address whether this Par-1-dependent phosphorylation of Baz is important in the COC function , we examined the effect of overexpression of a non-phosphorylatable form or a phosphomimetic form of Baz . Upon expression of the nonphosphorylatable form of Baz ( Baz-SA , Baz-S151A S1085A ) , GSCs showed a high frequency of centrosome and spindle misorientation ( Figure 5F ) . These results clearly demonstrate that the phosphorylation of Baz by Par-1 is critical for COC function . Furthermore , we found that overexpression of wild-type Baz as well as the phosphomimetic form of Baz ( Baz-SE , Baz-S151E ) suppressed spindle misorientation caused by Par-1 RNAi ( Figure 5F ) . In contrast , overexpression of the non-phosphorylatable form of Baz ( Baz-SA ) did not rescue spindle misorientation due to Par-1 RNAi ( Figure 5F ) . These results clearly demonstrate that Par-1 executes its function in the COC mainly through phosphorylation of Baz . Notably , Baz-SE did not rescue the centrosome misorientation phenotype , and overexpression of Baz-SE in a wild-type background resulted in a high frequency of centrosome misorientation , despite near complete suppression of spindle misorientation caused by Par-1 RNAi . These observations suggest that although phosphorylation of Baz is sufficient for COC function , dephosphorylation is also important for anchoring the centrosomes . The cell cycle-dependent phosphorylation cycle of Baz ( Figure 5A–D ) is also consistent with the idea that the phosphorylation and dephosphorylation cycle of Baz may be important for anchoring the centrosomes and monitoring centrosome orientation . Taken together , these results reveal the critical function of Par-1-mediated Baz phosphorylation in the function of the COC . To further assess the outcome of COC defect in these mutant conditions described above , we scored the frequency of symmetric GSC divisions . After inducing GSC clones at a low frequency ( i . e . , dominantly single GSC clone/testis ) , symmetric outcome can be assessed by scoring the frequency of ‘doublet’ clones ( i . e . two GSC clones are juxtaposed each other ) ( Figure 6A–C ) ( Salzmann et al . , 2013 ) . Wild-type GSC clones showed a basal level of doublet frequency ( ∼10% ) ( Figure 6D ) . This is due to expected low frequency events: ( 1 ) two juxtaposing GSCs become clones independently , and ( 2 ) ‘crawling back’ of GBs to the niche causes symmetric outcome of the division as described previously ( Sheng and Matunis , 2011 ) . dCR4h clones showed increased doublet frequency due to symmetric divisions as described previously ( Salzmann et al . , 2013 ) . GSC clones of Baz overexpression , Baz RNAi or Par-1 RNAi showed significant increase in the frequency of doublet clones ( Figure 6D ) . In contrast , GSC clones of Baz-SA as well as dCR4h with Baz RNAi did not increase the frequency of doublet clones . We did not examine Baz-SE , since Baz-SE yielded extremely low frequency of GSC clone induction for unknown reasons . These results demonstrate a correlation between COC defect/spindle misorientation and symmetric GSC division . However , the extent of spindle misorientation and the doublet clone frequency did not perfectly correlate , suggesting that there may be additional mechanisms to contribute to symmetric outcome: for example , spindle misorientation may not necessarily lead to symmetric GSC divisions , if spindle orientation is corrected after entering mitosis or the GSC clones are defective in niche adhesion in addition to spindle orientation . 10 . 7554/eLife . 04960 . 009Figure 6 . COC is required to prevent symmetric GSC divisions . ( A ) An assay system to examine symmetric GSC divisions . GFP+ clone is induced at a low frequency using hs-FLP , nos > stop > gal4 , UAS-GFP by a 20-min heatshock . GFP clones were examined 24 hr post heatshock . When such GSCs undergo symmetric stem cell division , it will generate doublet clones ( two GFP+ GSCs are juxtaposed each other ) . ( B , C ) Representative images of singlet ( B ) and doublet ( C ) GSC clones . Hub is indicated by the asterisk . Clones are indicated by dotted lines . Bar: 10 µm . ( D ) Frequency of doublets after 24 hr post heatshock . JF01079 line was used for Baz RNAi . ( E ) Model of Baz function in COC ( see text for detail ) . DOI: http://dx . doi . org/10 . 7554/eLife . 04960 . 009
Although intensive investigations have revealed the mechanisms of cell polarity and asymmetrical cell division along the polarity axis , much less is known about how cells ensure the correct temporary order of cell polarization and cell division . Precocious cell division before establishment of correct polarity would lead to a deleterious outcome , such as a failure in cell fate determination; however , the presence of checkpoint mechanisms to ensure asymmetric division has not been thoroughly investigated . In Drosophila neuroblasts , which divide asymmetrically by stereotypically oriented spindles , a phenomenon called ‘telophase rescue’ has been reported: many mutants that compromise correct spindle orientation in neuroblasts eventually divide asymmetrically ( Lu et al . , 1998; Schober et al . , 1999; Wodarz et al . , 1999; Peng et al . , 2000 ) . In ‘telophase rescue’ , asymmetric outcome of the division is restored by correcting the localization of basal polarity proteins ( Schober et al . , 1999; Peng et al . , 2000 ) . Such correction might indicate the presence of an orientation/polarity checkpoint , although the mechanistic basis remains unknown . Thus , the COC may serve as a model system to study a new class of checkpoints that specialize in monitoring division orientation in multicellular organisms . In the present study , we showed that Baz is a critical player in centrosome orientation and its checkpoint in Drosophila male GSCs . Baz forms a subcellular structure ( Baz patch ) at the hub-GSC interface , which anchors the apical centrosome prior to mitotic entry . Our data indicate that Baz-centrosome docking is the cellular event that is recognized by the COC as correct centrosome orientation . The data presented in this study point to the following working model ( Figure 6E ) : ( 1 ) Baz patch is formed at the hub-GSC interphase in an E-cadherin-dependent manner . ( 2 ) Baz patch functions to recruit the centrosome . In the absence of Baz , centrosome is highly misoriented . ( 3 ) Baz is ( partially ) required for the COC activity . Baz patch that is not docked to the centrosome might contribute to the activation of COC to prevent mitotic entry . ( 4 ) Once Baz patch is docked to the centrosome , this docking is interpreted as 'correct centrosome orientation’ , leading to inactivation of COC and thus mitotic entry . This model indicates that Baz plays dual roles in COC depending on its centrosome-docking status: in the absence of docking , Baz patch activates COC , whereas centrosome-docked Baz functions to inactivate COC . Our results show that Par-1-mediated phosphorylation of Baz is critical for spindle orientation , although the mechanistic details of phosphorylated Baz function are yet to be determined . The fact that both phosphomimetic form ( Baz-SE ) as well as wild-type Baz can rescue spindle misorientation in Par-1 RNAi GSCs suggests that timing of phosphorylation might not be so critical , in spite of observed phosphorylation–dephosphorylation cycle of Baz during normal cell cycle . With the currently available data , it is unclear how temporal regulation of Baz phosphorylation relates to steps of Baz-centrosome docking , mitotic entry , and spindle orientation . Furthermore , it is puzzling that overexpression of Baz causes high frequency of spindle misorientation in wild type , whereas the overexpression of the same construct in Par-1 RNAi background lowers spindle misorientation . Future investigation is required to understand how distinct isoforms of Baz ( phosphorylated vs non-phosphorylated ) participate in distinct aspects of centrosome/spindle orientation . In summary , our study reveals a cellular mechanism by which stem cells integrate information about cell polarity to regulate their cell cycle progression . Such a mechanism ultimately functions to ensure the asymmetric outcome of stem cell division . We speculate that the orientation checkpoint may be present in many other multicellular organisms , and the understanding of the COC in Drosophila may provide a conceptual framework for understanding orientation checkpoint mechanisms in general .
All fly stocks were raised on standard Bloomington medium at 25°C . The following fly stocks were used: UAS-Baz RNAi ( TRiP . JF01079 , obtained from the Bloomington Stock Center ) ; UAS-Baz RNAi ( GD1384 , obtained from the Vienna Drosophila Center ) ; Baz-GFP ( Flytrap project [Morin et al . , 2001; Kelso et al . , 2004; Buszczak et al . , 2007] ) ; nos-gal4 ( Van Doren et al . , 1998 ) , UAS-Baz-YFP ( obtained from Cheng-Yu Lee ) ; UAS-Baz-S151A S1085A ( Benton et al . , 2002 ) , UAS-Par-1 RNAi ( a kind gift from Bingwei Lu [Zhang et al . , 2007] ) ; UAS-dCR4h , UAS-DEFL ( a kind gift from Hiroki Oda [Oda and Tsukita , 1999] ) ; dap1gm- ( myc ) ( a kind gift from Christian F Lehner [Meyer et al . , 2002] ) , FRT42D shg10469 ( Uemura et al . , 1996 ) and nos > stop > gal4 ( Salzmann et al . , 2013 ) . For construction of Baz-S151E , a point mutation was introduced at the S151 residue by site-directed mutagenesis using the polymerase chain reaction ( PCR ) , and the mutant was subcloned into pUAST-EGFP-attB . All transgenic flies were generated using PhiC31 integrase-mediated transgenesis systems ( Groth et al . , 2004 ) by BestGene , Inc . Immunofluorescence staining was performed as described previously ( Cheng et al . , 2008 ) . The following primary antibodies were used: mouse anti-γ-tubulin ( 1:100; GTU-88 , Sigma-Aldrich , St . Louis , MO ) , mouse anti-Fasciclin III ( FasIII; 1:20; developed by C Goodman and obtained from the Developmental Studies Hybridoma Bank [DSHB] , Iowa City , IA ) , rabbit anti-Thr3-phosphorylated histone H3 ( 1:200; Cell Signaling Technology , Danvers , MA ) , rabbit anti-Vasa ( 1:100; Santa Cruz Biotechnology , Santa Cruz , CA ) , rat anti-Vasa ( 1:20; developed by AC Spradling and D Williams and obtained from DSHB ) , mouse anti-c-myc ( 1:100; clone 9E10 , DSHB ) , rabbit anti-c-myc ( 1:30; c3956; Sigma-Aldrich , St . Louis , MO ) , rabbit anti-Spd-2 ( Giansanti et al . , 2008 ) ( a kind gift from Maurizio Gatti , Dipartimento di Biologia e Biotecnologie Università di Roma ) , guinea pig anti-Baz ( 1:500; from Cheng-Yu Lee [University of Michigan] and Chris Doe [University of Oregon] ) , and rabbit anti-Baz-pS151 and Baz-pS1085 ( Krahn et al . , 2009 ) ( a kind gift from Andreas Wodarz [Georg-August-Universitat Gottingen] ) . Guinea pig anti-Baz ( 1:10 , 000 ) was also generated using the synthetic peptide Ac-VSEPDASKPRKTWLLEDGDHEGGFASQRC-amide ( Covance , Denver , PA ) , which showed the same staining pattern with other anti-Baz antibodies , thus used interchangeably in the experiments reported here . AlexaFluor-conjugated secondary antibodies were used at a dilution of 1:200 ( Life Technologies , Carlsbad , CA ) . Images were taken using a Leica TCS SP5 confocal microscope with a 63× oil immersion objective ( NA = 1 . 4 ) and processed using Adobe Photoshop software . 45-min ex vivo BrdU pulse labeling was performed as previously described ( Roth et al . , 2012 ) . BrdU was detected by immunofluorescence staining using rat anti-BrdU antibody ( 1:50; Abcam , ab6326 , Cambridge , MA ) . Statistical analysis was performed using Microsoft Excel 2010 or GraphPad Prism 6 software . Pixel intensity analyses for staining of Baz phospho-specific S151 , nuclear Dap , and Baz-GFP were performed using ImageJ software . For centrosome and spindle orientation scoring , >300 GSCs were scored for centrosome misorientation , and >30 mitotic GSCs were scored for spindle misorientation . Centrosome misorientation was indicated when neither of the two centrosomes were closely associated with the hub-GSC interface during interphase . Spindle misorientation was indicated when neither of the two spindle poles were closely associated with the hub-GSC interface during mitosis . Data are shown as means ±standard deviation . The p-value ( two-tailed student's t-test ) is provided for comparison with the control . | The tissues of an animal's body are built from cells that are originally derived from stem cells . Each stem cell can divide and give rise to another stem cell and a cell that will become a more specific type of cell—such as a nerve cell , muscle cell , or sperm cell . If this asymmetric cell division is disrupted , it can result in developmental disorders and diseases such as cancer . When a cell divides , a structure known as the spindle separates the copies of the chromosomes into the two newly formed cells . The spindle consists of long protein filaments that extend from two smaller structures known as centrosomes , which are found at opposite sides of the cell . The position of these centrosomes governs the orientation of the spindle , which in turn determines the plane in which cell division takes place . Thus , cells that need to divide with a certain orientation must have a mechanism that ensures that their centrosomes are correctly positioned . However , the existence of such a mechanism has been underexplored , and it remains unclear how the alignment of the centrosomes is controlled . Inaba et al . analyzed how stem cells in the male fruit fly divide asymmetrically to form one stem cell and second cell that develops into sperm . The experiments revealed that a protein called Bazooka ( or Baz for short ) closely associates with the centrosomes just before the cells start to divide . Many other animals—such as humans and worms—have proteins that are closely related to Bazooka , which are needed for asymmetric cell divisions . When Inaba et al . reduced the levels of the Bazooka protein in the fruit fly cells , a large number of these cells ended up with centrosomes that were incorrectly aligned . As a result , these cells' spindles were also oriented incorrectly . These findings suggest that the interactions between Bazooka and the centrosomes inform a cell when the centrosomes are correctly orientated . However , further work will be required to determine the details of how Bazooka controls asymmetric cell divisions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cell",
"biology"
] | 2015 | The polarity protein Baz forms a platform for the centrosome orientation during asymmetric stem cell division in the Drosophila male germline |
Human gut Bacteroides use surface-exposed lipoproteins to bind and metabolize complex polysaccharides . Although vitamins and other nutrients are also essential for commensal fitness , much less is known about how commensal bacteria compete with each other or the host for these critical resources . Unlike in Escherichia coli , transport loci for vitamin B12 ( cobalamin ) and other corrinoids in human gut Bacteroides are replete with conserved genes encoding proteins whose functions are unknown . Here we report that one of these proteins , BtuG , is a surface-exposed lipoprotein that is essential for efficient B12 transport in B . thetaiotaomicron . BtuG binds B12 with femtomolar affinity and can remove B12 from intrinsic factor , a critical B12 transport protein in humans . Our studies suggest that Bacteroides use surface-exposed lipoproteins not only for capturing polysaccharides , but also to acquire key vitamins in the gut .
Our understanding of the factors that shape gut microbial community composition is largely based on the primary economy of this ecosystem: the flow of carbon from the diet to bacterial biomass and fermentation products . However , an accompanying secondary economy of essential vitamins and other cofactors , which are much less abundant , also plays a critical role in determining bacterial growth rates and resulting microbiome dynamics ( Sonnenburg and Sonnenburg , 2014 ) . Small molecule cofactor biosynthesis is an energetically costly process and alternate cofactor-independent enzymes can be less efficient ( Roth et al . , 1993 ) , favoring microbes that can best acquire these nutrients from their environment . The complex organometallic cofactor vitamin B12 ( cobalamin ) is representative of this challenge: de novo biosynthesis requires the coordinated activity of nearly 30 dedicated enzymes ( Roth et al . , 1993 ) . Although bacteria have evolved cofactor-independent alternatives to many B12-dependent enzymes ( e . g . B12-independent methionine synthase MetE and ribonucleotide reductase NrdEF ) , even species that lack vitamin biosynthetic machinery maintain their B12-dependent enzymes ( e . g . , methionine synthase MetH and ribonucleotide reductase NrdZ ) for use when vitamin B12 is available in the environment ( Degnan et al . , 2014a; Degnan et al . , 2014b; Young et al . , 2015 ) . In the human gut , most species encode B12-dependent enzymes . The great majority of these species also encode transport systems for capturing B12 from the environment , either instead or in addition to vitamin biosynthetic pathways ( Degnan et al . , 2014a ) . Although bacteria utilize many vitamin B12-like molecules ( corrinoids ) , transport and utilization proteins and regulatory elements are typically referred to as B12-dependent , in keeping with their initial characterization . The machinery used by Gram-negative bacteria to transport vitamin B12 and other corrinoids has been studied extensively in Escherichia coli and has been used as a model for TonB-dependent transport . In E . coli , extracellular B12 is transported into the periplasm through the outer membrane β-barrel protein BtuB in a TonB-dependent manner ( Bassford et al . , 1976; Bassford and Kadner , 1977 ) . The periplasmic protein BtuF subsequently binds to and delivers B12 to the ABC-type transporter BtuCD in the inner membrane , which brings the vitamin into the cytoplasm ( Cadieux et al . , 2002 ) . In previous studies , we established that the most abundant Gram-negative bacteria in the human gut ( Bacteroidetes ) encode a diverse array of B12 transport systems in B12-riboswitch regulated loci , often with multiple locus architectures per genome ( Degnan et al . , 2014a ) . Bacteroides thetaioatomicron serves as a model for defining the role of these transporters , as it does not encode B12 biosynthetic machinery , and its repertoire of B12 transporters determine in vivo fitness in a diet- and community context-dependent manner ( Degnan et al . , 2014a; Goodman et al . , 2009 ) . In the course of these studies , we noticed that B . thetaiotaomicron and other human gut Bacteroides also maintain a heterogeneous repertoire of additional genes , nearly all of unknown function , in these B12 transport loci . E . coli also encodes additional genes within its B12 transport operons ( e . g . , btuE is positioned between btuC and btuD ) , but these do not play a role in B12 transport ( de Veaux et al . , 1986; Arenas et al . , 2010 ) . Here we report that unlike BtuE in E . coli , Bacteroides accessory proteins can play a critical role in B12 transport . Using the B . thetaiotaomicron homolog of a highly conserved accessory protein ( BT1954; hereafter called BtuG2 ) as a representative , we establish that BtuG2 localizes to the cell surface , interacts with BtuB , and determines the ability of B . thetaiotaomicron to transport B12 and persist in the mammalian gut . Furthermore , BtuG2 directly binds vitamin B12 with femtomolar affinity , thereby enabling it to acquire this vitamin from intrinsic factor , a critical B12 transport protein in humans .
Nearly all of the Bacteroides B12 transport loci include homologs of a hypothetical gene that is exclusively found in the Bacteroidetes phylum ( Figure 1A ) . We refer to these homologs as BtuG; one of which ( BtuG2 from B . thetaiotaomicron , marked with a star in Figure 1A ) has been crystallized and adopts a seven-bladed β-propeller fold ( PDB 3DSM; Figure 1B ) . B . thetaiotaomicron encodes three genetic loci with vitamin B12 transport genes ( locus1 , locus2 and locus3; Figure 1A ) . Each locus encodes a btuG homolog ( btuG1 , btuG2 and btuG3 , respectively ) adjacent to a homolog of btuB ( Figure 1A ) . Using BtuG1 , BtuG2 and BtuG3 from B . thetaiotaomicron as representatives , we identified 112 putative BtuG homologs in 313 genome-sequenced human gut bacterial strains ( Degnan et al . , 2014a ) by an initial BlastP search ( Supplementary file 1 ) . 106 are encoded in 106 of 109 btuB-containing operons identified previously ( Degnan et al . , 2014a ) . The six remaining btuG homologs identified by BlastP have operon annotations affected by incomplete genome assemblies , however , five are associated with a corrinoid riboswitch and one or more btu transport genes . Two of the three remaining btuB-containing operons encode a divergent btuG gene ( e-value >1e-10 , but Phyre2 match to the BtuG2 crystal structure ) . No homologs were detected outside of the Bacteroidetes using the defined BlastP parameters ( Supplementary file 1 ) . We sought to test whether BtuG is needed for B12-dependent growth , given the near universal genetic linkage of btuG homologs to known B12 transport genes . Because B . thetaiotaomicron encodes three homologous B12 transport loci ( and corresponding btuG genes ) that complicate the ability to assign functions to specific genes , we established a simplified genetic background that lacks locus1 and locus3 ( hereafter referred to as the B . thetaiotaomicron ‘parent’ strain ) . We created an in-frame , unmarked deletion of btuG2 in this parent strain and tested its ability to grow in minimal media supplemented with cyanocobalamin ( B12-dependent growth; B . thetaiotaomicron encodes MetH but not MetE ) or methionine ( B12-independent growth ) . We chose cyanocobalamin concentrations of 37 nM and 0 . 37 nM because these concentrations repress ( 37 nM ) or activate ( 0 . 37 nM ) B12 riboswitches in this species ( Sonnenburg et al . , 2005; Martens et al . , 2008; Degnan et al . , 2014a ) . As compared with its growth in methionine medium , the btuG2 deletion strain had a ~4 hr longer lag phase in 37 nM cyanocobalamin medium and did not display any growth during 72 hr in 0 . 37 nM cyanocobalamin medium ( Figure 1C ) . By contrast , the parent and complemented strains grew indistinguishably in all three media . Vitamin B12 riboswitches are RNA aptamers that bind B12 directly and repress downstream gene expression , providing a biosensor for intracellular levels of this cofactor ( Fowler et al . , 2010 ) . To determine the contribution of BtuG2 to intracellular corrinoid accumulation , we used the B12 riboswitch of locus2 as a biosensor . Quantification of gene expression by qRT-PCR revealed that the parent strain represses riboswitch-dependent gene expression ~70 fold in culture medium containing 37 nM cyanocobalamin as compared to medium with 0 . 37 nM and 0 nM cyanocobalamin . By contrast , the btuG2 deletion strain fails to repress B12 riboswitch-regulated gene expression in any concentration of extracellular cyanocobalamin ( Figure 1D ) . These data suggest that BtuG2 contributes significantly to B12 accumulation within B . thetaiotaomicron cells . B12 transport machinery encoded in locus2 are critical for B . thetaiotaomicron fitness in gnotobiotic mice ( Goodman et al . , 2009 ) . To compare the relative contribution of BtuG2 and BtuB2 to fitness in the gut , we colonized germfree mice with a 1:1 mixture of the parent strain and ∆btuG2 ( Figure 1E and Figure 1—figure supplement 1 ) or ∆btuB2 ( Figure 1F ) , and monitored the relative abundance of each strain in fecal samples collected over time . In both groups of mice , the parent strain dominated while the abundance of the mutant strain dropped continuously until it was no longer detected on day 24 ( Figure 1E–F ) . Surprisingly , these data indicate that the absence of BtuG2 has a similarly deleterious impact on in vivo fitness as the absence of the outer membrane transporter BtuB2 . Together , these studies suggest that BtuG plays a critical role in mediating B . thetaiotaomicron B12 transport in the gut . We next sought to determine the subcellular localization of BtuG2 to better understand its role in B12 transport . Aligning the first 90 amino acids of 114 homologs of BtuG ( Supplementary file 1 ) using ClustalW ( Larkin et al . , 2007 ) and displaying their conservation as a sequence logo ( Crooks et al . , 2004; Schneider and Stephens , 1990 ) revealed a number of important clues: firstly , nearly all homologs have a conserved cysteine residue within their first ~18–40 amino acids ( Cys-32 of BtuG2 ) ( Inouye et al . , 1983 ) ; secondly , this cysteine is preceded by a conserved lipobox-like sequence typical of lipidated proteins ( VFGS in BtuG2 ) ( von Heijne , 1989 ) ; thirdly , this cysteine is followed by a conserved lipoprotein export signal or LES ( MKWD in BtuG2 ) , a feature exclusive to Bacteroidetes that allows lipoproteins to be flipped from the inner to the outer leaflet of the outer membrane ( Figure 2A ) ( Lauber et al . , 2016 ) . Thus , from their primary sequences alone , we predicted BtuG homologs to be surface-exposed lipoproteins . To test whether BtuG2 is indeed surface-exposed , we treated intact cells grown in minimal medium with methionine ( but without cyanocobalamin ) with varying concentrations of proteinase K ( 0–100 µg/mL ) , ran whole cell lysates on an SDS-PAGE gel , and probed by Western blot for BtuG2 or a periplasmically localized control protein SusA ( BT3704 ) appended with a C-terminal HA-tag ( Shipman et al . , 1999 ) . While SusA was protected from protease treatment even at the highest concentration of protease , BtuG2 was progressively degraded at increasing concentrations of protease , consistent with it being surface-exposed ( Figure 2B ) . Moreover , BtuG2 associated most strongly with the membrane of fractionated B . thetaiotaomicron cells ( parent strain ) , and was even found in the supernatant fraction , consistent with its localization to the interface of the outer membrane of the cell and the extracellular milieu ( Figure 2C ) . Alteration of these surface-localizing sequence features diminished BtuG2 stability , complicating efforts to directly assess the contribution of these sequences to protein localization ( Figure 2—figure supplement 1A ) . Deleting the putative signal sequence ( aa 2 – 31 ) , mutating Cys-32 to alanine , or changing three residues of the LES ( Lys-34 , Trp-35 and Asp-36 ) to alanine all resulted in a lack of BtuG2 detection in whole cell lysates by Western blot , despite normal levels of transcription ( Figure 2—figure supplement 1A–B ) . By contrast , replacing the putative signal sequence with aa 1 – 18 of the known surface-exposed lipoprotein SusD ( BT1762; aa 1 – 18 ) ( Shipman et al . , 2000; Glenwright et al . , 2017 ) , or replacing the signal sequence and LES ( BtuG2 aa 1 – 37 ) with the signal sequence and LES of SusD ( aa 1 – 24 ) complemented a btuG2 deletion strain both in terms of protein production and B12-dependent growth ( Figure 2—figure supplement 1C–D ) . These results indicate that the N-terminal residues of BtuG2 are critical for protein stability , and that these residues can be functionally replaced with the corresponding sequences from the surface-exposed lipoprotein SusD . Given that BtuG2 exhibits sequence signatures and protease sensitivity indicative of a surface-exposed lipoprotein , is localized to the cell membrane , and contributes to B12 acquisition , we hypothesized that it might interact with the outer membrane B12 transporter BtuB2 . Consistent with this hypothesis , we found that deletion of btuB2 changes the predominant localization of BtuG2 from the membrane fraction to the culture supernatant ( Figure 2C ) . Furthermore , tandem affinity purification ( TAP ) of BtuB2-associated proteins in growing B . thetaiotaomicron cells readily pulls down BtuG2 ( Figure 2D ) . By contrast , TAP of an untagged strain , or of a strain with the TAP epitopes appended to an unrelated TonB-dependent outer membrane β-barrel-type transporter ( SusC; BT1763 ) fails to pull down BtuG2 . These data suggest that BtuG2 associates with known B12 transport machinery . All components of the canonical vitamin B12 transport pathway—BtuB , BtuF and BtuCD—bind cyanocobalamin directly during the process of transport from outside the cell into the periplasm and ultimately the cytoplasm . To test whether BtuG2 also binds cyanocobalamin , we expressed and purified BtuG2-10xHis in E . coli . We then took advantage of the ability of cyanocobalamin to absorb light at a wavelength of 362 nm , and aromatic amino acids within BtuG2 to absorb light at 280 nm , by performing size-exclusion chromatography with multi-angle light scattering ( SEC-MALS ) on BtuG2 after co-incubation with equimolar cyanocobalamin . At both wavelengths , we observed nearly identical traces corresponding to the elution volume for BtuG2 , indicating that BtuG2 can directly bind cyanocobalamin in vitro ( Figure 3A ) . Purified , recombinant BtuG homologs from B . vulgatus , B . uniformis and B . coprophilus ( BVU2056 , BACUNI04578 and BACCOPRO02032 , respectively ) also exhibit this function ( Figure 3—figure supplement 1 ) . We then sought to determine the kinetics and affinity of cyanocobalamin binding by BtuG2 . Because BtuG2-cyanocobalamin saturation occurs too rapidly to measure the dissociation constant accurately by isothermal titration calorimetry ( data not shown ) , we used surface plasmon resonance ( SPR ) to determine a KD of 1 . 87 ± 0 . 76 × 10−13 M for BtuG2-cyanocobalamin binding and 1 . 93 ± 0 . 63 × 10−13 M for BtuG2-dicyanocobinamide binding ( Figure 3B ) . Both ligands bind BtuG2 at a 1:1 ratio . BtuG2 binds to cyanocobalamin with a measured kon = 1 . 40 ± 0 . 05 × 109 M−1s−1 and a koff = 2 . 59 ± 0 . 96 × 10−4 s−1; similarly , BtuG2 binds to dicyanocobinamide with a measured kon = 2 . 61 ± 1 . 56 × 109 M−1s−1 and a koff = 4 . 54 ± 1 . 38 × 10−4 s−1 . Together , these measurements establish that BtuG2 binds cyanocobalamin and a corrinoid precursor with femtomolar affinity , at a rate generally observed for diffusion-limited enzymes and proteins . This ligand interaction is maintained for over 1 hr on average before spontaneous dissociation ( Corzo , 2006 ) . Notably , surface electrostatic analysis of the crystal structure of BtuG2 reveals a predominantly positive electrostatic potential on the face of BtuG2 displaying the C-terminal 6x-His tag , and a predominantly negative electrostatic potential on the opposing face ( Figure 3—figure supplement 2A ) . The coordinated cobalt ion of a corrinoid carries a positive charge ranging from +1 to +3 , depending in part on the upper ligand ( –CN , –Me , –Ado , –OH ) ( Obeid et al . , 2015 ) . Therefore , if surface electrostatic charges on BtuG2 are involved in orienting corrinoids to facilitate protein-ligand interactions , these forces should draw the ligand into the negative electrostatic face of BtuG2 . Because BtuG2 modulates intracellular corrinoid levels , is surface-exposed , associates with BtuB2 and binds cyanocobalamin directly , we reasoned that it might act as a critical extracellular step in the process of capturing and transporting corrinoids . To test whether BtuG2 can function extracellularly , we measured growth of btuG2 mutant cells upon supplementation with supernatants from btuB2 mutant cultures ( this strain disproportionally partitions BtuG2 to the supernatant; Figure 2C ) . To this end , we collected , filter-sterilized , ultra-centrifuged , and concentrated culture supernatants from B . thetaiotaomicron ∆locus1 ∆locus3 ∆btuB2 ( or ∆locus1 ∆locus3 ∆btuG2 as a control ) strains grown to exponential phase in minimal medium lacking cyanocobalamin and supplemented with methionine . Concentrated supernatants were incubated in the presence or absence of 0 . 37 µM cyanocobalamin , diluted and concentrated repeatedly to remove unbound ligand and residual methionine , and introduced to recipient cells in medium lacking both cyanocobalamin and methionine . Under these conditions , ∆btuG2 recipient cells grow robustly when provided with BtuG2-containing culture supernatant that had been incubated with cyanocobalamin ( Figure 4A ) . By contrast , the same recipient strain failed to grow when provided culture supernatant from a ∆btuG2 strain incubated with cyanocobalamin , or when provided BtuG2-containing culture supernatant incubated with PBS instead of cyanocobalamin ( Figure 4A ) . Further , BtuG2-containing culture supernatant incubated with cyanocobalamin failed to rescue growth of recipient cells lacking corrinoid transport machinery . These data suggest that BtuG2 can function from the exterior of cells in trans to promote cyanocobalamin-dependent growth . Notably , BtuG2-6xHis purified from E . coli failed to restore cyanocoblamin-dependent growth to a ∆btuG2 mutant , suggesting that lipidation or some other Bacteriodes-specific feature may be important for protein function ( data not shown ) . The observation that BtuG2 can function in trans raises the possibility that this protein could act as a public good , secreted into the environment and shared across cells in the population . However , BtuG2-producing parent cells fail to rescue the cyanocobalamin-dependent growth defect of a ∆btuG2 strain when the two strains are co-cultured ( Figure 4B ) , suggesting that this protein primarily functions in cis . This is consistent with its localization as a membrane-bound lipoprotein ( Figure 2 ) and the competitive defect of a ∆btuG2 strain in the presence of parent B . thetaiotaomicron in the mouse gut ( Figure 1E and Figure 1—figure supplement 1 ) . Because BtuG2 is surface-exposed and binds cobalamin with femtomolar affinity , we wondered if it might affect the function of human B12-binding proteins that transport this vitamin through the length of the gastrointestinal tract . Humans absorb cobalamin from their diet with the help of two carrier proteins . The first , haptocorrin , is secreted from salivary glands and binds the vitamin as it is released from food broken down in the stomach; the second , intrinsic factor ( IF ) , is released from parietal cells in the stomach and binds cobalamin in the duodenum following the degradation of haptocorrin by host enzymes ( Nielsen et al . , 2012 ) . IF then carries the vitamin through several meters of intestinal tract to the distal ileum , where receptors on intestinal epithelial cells allow for the uptake of IF and its vitamin cargo ( Nielsen et al . , 2012 ) . As IF traverses the small intestinal lumen , it encounters increasing densities of gut microbes ( from ~103 to ~108 cells/g ) ( Scheithauer et al . , 2016 ) . To test whether B . thetaiotaomicron can use BtuG2 to acquire vitamin B12 from IF , we first incubated recombinant human IF with cyanocobalamin , diluted and concentrated repeatedly to remove unbound ligand , provided the IF-cyanocobalamin complexes to B . thetaiotaomicron cells in minimal media without exogenous cyanocobalamin or methionine , and measured culture growth over time . The parent B . thetaiotaomicron strain grew readily when provided IF-cyanocobalamin ( Figure 5A ) . By contrast , IF alone was not sufficient to allow bacterial growth , and recipient cells lacking btuG2 do not grow when provided IF-cyanocobalamin . Further , the parent strain is unable to grow when provided the filtrate from the last IF-cobalamin wash step , indicating that an insignificant amount of cobalamin dissociates from IF during dilution and concentration ( Figure 5A ) . These results indicate that btuG2-encoding B . thetaiotaomicron cells are capable of acquiring cyanocobalamin from IF . To test whether BtuG2 acquires cobalamin from IF directly , we incubated cyanocobalamin with IF and/or recombinant BtuG2 and determined the relative amounts of the vitamin associated with each protein by SEC-MALS . As expected , incubation of BtuG2 with cyanocobalamin produces a distinct absorbance peak at 362 nm corresponding to the elution volume for BtuG2 ( Figure 5B ) . Similarly , incubation of IF with cyanocobalamin produces a 362 nm absorbance peak at the elution volume corresponding to IF . Notably , addition of BtuG2 to IF-cobalamin shifts the majority of the 362 nm absorbance to the elution volume for BtuG2 ( Figure 5B ) . This suggests that BtuG2 can directly acquire cobalamin from an IF-cobalamin complex . We next sought to determine whether this direct transfer of cobalamin from IF to BtuG2 allows B . thetaiotaomicron to grow on cobalamin acquired from IF-cobalamin complexes . Indeed , ∆btuG2 recipient cells grow readily when provided ∆btuB2 culture supernatants supplemented with IF-cobalamin complexes ( Figure 5C ) . The same recipient cells do not grow when provided ∆btuG2 culture supernatants supplemented with IF-cobalamin complexes , when provided ∆btuB2 culture supernatants supplemented with IF alone , or when provided ∆btuB2 culture supernatants supplemented with the filtrate from the last wash of IF-cobalamin complexes ( Figure 5C ) . Collectively , these results suggest that B . thetaiotaomicron can use BtuG2 to acquire cobalamin that is already bound to the host protein responsible for transporting this vitamin through the gastrointestinal tract .
Early studies documenting increased vitamin requirements for germfree animals suggested that the microbiota plays a critical role in contributing these essential cofactors to the host ( Barnes , 1967 ) . In the case of vitamin B12 , however , it is unlikely that the microbiota makes a significant contribution to host vitamin supply . Indeed , B12 constitutes less than 2% of fecal corrinoid pools in humans , and supplementation studies suggest that gut microbes efficiently convert dietary B12 into alternate corrinoids that cannot be used by humans ( Allen and Stabler , 2008 ) . In this report , we describe a novel microbial factor , BtuG , that could pose a more direct obstacle to the host’s absorption of this essential vitamin . What evolutionary forces drove the Bacteroidetes , unlike other Gram-negative phyla , to incorporate an additional component into their B12 transport pathway that binds B12 with such high affinity ? One possible answer lies in the gut environment , where bacteria co-exist at densities of 1011 cells per gram or higher ( Whitman et al . , 1998 ) . Under these conditions , adaptations that increase corrinoid capture could allow cells to minimize their requirement for energetically costly vitamin biosynthetic pathways . Indeed , many human gut Bacteroidetes encode incomplete vitamin B12 biosynthesis pathways; B . thetaiotaomicron is missing this pathway entirely . Selection for increased corrinoid binding affinity in BtuG could conversely permit mutations that decrease the ability of BtuB to directly capture these molecules from the environment: E . coli , which transports B12 via BtuB and lacks any BtuG homolog , grows readily on 0 . 4 nM B12 ( Di Girolamo et al . , 1971 ) , while B . thetaiotaomicron requires BtuG under these conditions ( Figure 1C ) . An additional consequence of selection for increased corrinoid binding affinity in BtuG is that such adaptations allow this protein to compete with the host proteins for dietary cobalamin . In this way , the ability of BtuG to acquire B12 from IF may have emerged as a byproduct of inter-microbial competition for gut corrinoids . Notably , BtuG has evolved this femtomolar affinity for B12 while still maintaining the capacity to release the vitamin for transport into the cell . Gut bacteria can interfere with host cobalamin absorption in patients with small intestinal overgrowth of bacteria , leading to cobalamin deficiency ( Giannella et al . , 1971 ) . Early efforts to find the responsible bacteria using radiolabeled cyanocobalamin reported that Bacteroides were particularly adept at removing cyanocobalamin from IF in vitro ( Giannella et al . , 1972; Schjönsby et al . , 1973 ) . Other studies describe patients with small intestinal bacterial overgrowth whose cobalamin deficiency was corrected by antibiotics that target Bacteroides but not Proteobacteria ( Schjönsby et al . , 1977 ) ; however , the precise factors responsible for these phenomena were never defined . Our data suggest that BtuG , which is universally present among the Bacteroides , could be one extracellular factor responsible for these observations . Although BtuG-mediated vitamin piracy may be well tolerated by a healthy host , conditions of small intestinal bacterial overgrowth or diminished IF production could alter this effect . Notably , even minor B12 deficiencies ( historically assigned to the ‘subclinical’ range ) could have health consequences ( McCaddon , 2013; Moore et al . , 2012 ) . A comparison of measured binding kinetics of BtuG2 and IF may explain how this bacterial protein extracts cobalamin from IF , which also binds the vitamin very strongly . Reported equilibrium dissociation constants for IF and cobalamin span a broad range ( KD ~10−9–10−15 M ) ; the most extreme of these describes a KD ~5 × 10−15 M , a kon ~7 × 107 M−1s−1 and a koff ~4 × 10−7 s−1 ( Fedosov et al . , 2005; Brada et al . , 2001; Fedosov et al . , 2006 ) . By comparison , BtuG2 has a KD ~2 × 10−13 M , a kon ~1 × 109 M−1s−1 and a koff ~3 × 10−4 s−1 ( Figure 3B ) . This kon rate , which is similar to that previously measured in diffusion-limited enzymes and proteins ( Corzo , 2006 ) , is orders of magnitude greater than the kon for IF . This suggests that BtuG2 has a superior ability to bind up free cobalamin compared with IF . The different environments in which these proteins encounter free cobalamin could explain these differences in kon rates . In humans , IF first encounters free cobalamin in the duodenum , where the vitamin is released from haptocorrin by host enzymes . At this juncture , there are few gut microbes present ( ~103 per gram ) to compete with IF for free cobalamin ( Scheithauer et al . , 2016 ) . By contrast , microbial densities in the large intestine exceed 1011 per gram , thus introducing a stronger element of competition for free corrinoids among gut microbes ( Whitman et al . , 1998 ) . Therefore , BtuG2 may face a stronger selective pressure to augment its kon rate in order to remain competitive against other BtuG homologs in the microbiota . The koff rate of IF is orders of magnitude lower than the koff rate of BtuG2 , corresponding to an average protein–ligand association time of ~700 hr for IF–cobalamin versus ~1 hr for BtuG2–cobalamin . Although both of these proteins bind B12 for delivery to their respective receptor , their different tasks may explain the observed koff rates . IF binds cobalamin in the proximal small intestine ( duodenum ) , but its uptake receptors on epithelial cells are located exclusively in the distal small intestine ( ileum ) . Therefore , IF must maintain its association with cobalamin while traversing several meters of intestinal tract before the host can absorb its nutrient cargo . By contrast , BtuG2 localizes to the bacterial cell surface and likely acts in cis to capture extracellular corrinoids for subsequent delivery to the outer membrane receptor BtuB on the same cell . This difference in protein function may impose differences in selective pressure for koff rates . Although BtuG2 and IF thus achieve these strong binding affinities by different kinetics , mixing the proteins results in a transfer of cobalamin from IF to BtuG2 ( Figure 5 ) . It is unclear whether this occurs through a direct interaction between BtuG2 and IF . Diffusion-limited enzymes and proteins ( e . g . , acetylcholinesterase and superoxide dismutase ) can employ surface electrostatic charges to affect the fluid environment in their immediate vicinity in ways that can enhance the likelihood of protein–ligand contact beyond the frequency determined through diffusion alone ( Tan et al . , 1993; Getzoff et al . , 1983 ) . Acetylcholinesterase , for example , exhibits contrasting electrostatic charge distributions on opposing faces of the enzyme , creating an electrostatic dipole that is reported to drive the interaction between the enzyme and its positively charged ligand ( Figure 3—figure supplement 2A ) ( Ripoll et al . , 1993; Tan et al . , 1993 ) . BtuG2 also presents one face with a predominantly positive electrostatic potential and the other with a strikingly negative electrostatic potential ( Figure 3—figure supplement 2A-B ) . While the β-propeller structure of BtuG2 , which resembles a disk with a central hole , may contribute to the formation of a dipolar electrostatic field through the middle of the protein , these properties are not intrinsic to seven-bladed β-propeller proteins ( Figure 3—figure supplement 2A-B ) . Because the coordinated cobalt ion of corrinoids carries a positive charge , surface electrostatic charges could be involved in orienting free corrinoids to facilitate protein–ligand interactions by repelling corrinoids away from the positive electrostatic face while drawing them into the negative electrostatic face of BtuG2 . This uncommon surface electrostatic charge distribution in BtuG2 could potentially alter IF–cobalamin stability without direct interaction between the two proteins . Although our evidence of BtuG lipidation is indirect , the use of cell surface-level machinery to enhance the uptake of key nutrients is not unprecedented for gut microbes . For example , the Bacteroides each encode dozens of Sus-like systems , defined by the presence of an outer membrane β-barrel protein ( e . g . SusC ) , a cell surface-exposed lipoprotein ( e . g . , SusD ) , and other components that directly interact to digest and import various polysaccharides into the cell ( Koropatkin et al . , 2012; Glenwright et al . , 2017 ) . Our studies suggest that surface-exposed nutrient binding proteins may determine the ability of these bacteria to not only capture carbon , but also to drive the ‘secondary economy’ of critical vitamins that power microbial growth in the gut .
Culturing of Bacteroides thetaiotaomicron VPI-5482 was carried out in an anaerobic chamber ( Coy Laboratory Products , Grass Lake , MI , USA ) , filled with 70% N2 , 20% CO2 , and 10% H2 by volume , using minimal media with vitamin B12 omitted ( Martens et al . , 2008 ) supplemented where specified with 500 µM DL-methionine and/or vitamin B12 ( 0 , 0 . 37 or 37 nM ) . Escherichia coli S17-1 lambda pir or BL21 Rosetta ( DE3 ) strains were grown in LB medium and incubated aerobically at 37°C . Culture media were supplemented with antibiotics as needed at the following concentrations: ampicillin 100 μg/mL , chloramphenicol 30 µg/mL , erythromycin 25 μg/mL , gentamicin 200 μg/mL , tetracycline 2 μg/mL , and 5-fluoro-2′-deoxyuridine ( FUdR ) 200 μg/mL . Plasmid constructs ( Supplementary file 1 ) were created , maintained and transformed using standard molecular cloning procedures . Primers ( Supplementary file 1 ) were obtained from the Keck Biotechnology Resource Laboratory ( Yale University , New Haven , CT , USA ) and DNA amplification was performed using KAPA HiFi ReadyMix ( Kapa Biosystems , Wilmington , MA , USA ) . Gene deletions in B . thetaiotaomicron were carried out as previously reported using B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 as a parent strain ( Degnan et al . , 2014a ) by amplifying flanking regions ( ~1000 bp ) of genes of interest and joining them by splicing by overlap extension ( SOE ) PCR or Gibson assembly . The concatenated fragments were inserted into the suicide vector pExchange-tdk ( Martens et al . , 2008 ) via ligation or Gibson assembly . Clones were sequence-verified and introduced into B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 by conjugation . Following counter selection , gene deletions were confirmed by PCR . Gene complementation constructs were created using pNBU2 vectors ( with or without oligonucleotide barcodes ) introduced in single copy into B . thetaiotaomicron as previously described ( Martens et al . , 2008 ) . Complementation constructs contained 425 bp upstream of BT1957 , the first gene in locus2 , to capture the locus2 promoter and vitamin B12 riboswitch ( us1957 , Supplementary file 1 ) . Overnight cultures of B . thetaiotaomicron strains grown in minimal medium supplemented with methionine ( no B12 ) were pelleted , washed three times in minimal medium without methionine or vitamin B12 , and used to inoculate wells of a 96-well plate in triplicate containing minimal media with methionine or vitamin B12 ( 0 , 0 . 37 or 37 nM ) . The plate was incubated anaerobically under constant agitation for 72 hr at 37˚C and OD600 measurements were taken at regular intervals using a BioTek Eon microplate spectrophotometer . B . thetaiotaomicron strains carrying unique oligonucleotide barcodes were co-cultured in minimal media in triplicate as previously described ( Degnan et al . , 2014a; Martens et al . , 2008 ) . Briefly , B . thetaiotaomicron strains were grown overnight in minimal media supplemented with methionine , washed and resuspended in minimal media without methionine or vitamin B12 . OD600 was measured and used to create a 1:10 mixture of competing strains ( 1 part parent strain to 10 parts ∆btuG2 strain ) , which was then used to inoculate , at 1:1000 , minimal media supplemented with methionine or vitamin B12 ( 0 , 0 . 37 , or 37 nM ) . These inoculations were then incubated anaerobically at 37 ˚C under static conditions or shaking ( 250 rpm ) . Cultures were passaged at 1:1000 into fresh media every 24 hr and an aliquot was stored at −20 ˚C for gDNA extraction ( Truett et al . , 2000 ) . Relative strain abundances were determined by quantitative PCR ( qPCR ) using a CFX96 thermocycler ( Bio-Rad , Hercules , CA , USA ) and SYBR FAST Universal Mastermix ( KAPA Biosystems , Wilmington , MA , USA ) ( Degnan et al . , 2014a ) . Strain abundances were analyzed using a standard curve and efficiency-corrected ∆Cq method was used to determine relative fold changes ( Bookout et al . , 2006 ) . All animal experiments were performed using protocols approved by the Yale University Institutional Animal Care and Use Committee . Male and female germfree 8- to 12-week-old Swiss Webster mice were individually caged and maintained in flexible plastic gnotobiotic isolators with a 12 hr light/dark cycle . Mice were provided with standard autoclaved mouse chow ( 5K67 LabDiet; Purina , St . Louis , MO , USA ) ad libitum . Germfree mice were colonized with 200 µL bacterial glycerol stocks by oral gavage . Mice were divided into groups ( n = 4 – 5/group ) . Each mouse in the first group was gavaged with 108 CFU each of B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 att1::pNBU2_tetQ_BC01 ( ‘parent’ strain ) and B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 ∆btuG2 att1::pNBU2_tetQ_BC14 ( ‘∆btuG2’ strain ) . Each mouse in the second group was gavaged with 108 CFU each of the parent strain and B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 ∆btuB2 att1::pNBU2_tetQ_BC14 ( ‘∆btuB2’ strain ) . Each mouse in the third group was gavaged with ~107 CFU each of the parent strain , the ∆btuG2 strain , and B . thetaiotaomicron ∆tdk ∆locus1 ∆locus3 ∆btuG2 att1::pNBU2_tetQ_BC16_us1957_btuG2 ( ‘∆btuG2 + btuG2’ strain ) . Fecal samples were collected over time and stored at −80°C before genomic DNA extraction . DNA was extracted as described previously ( Cullen et al . , 2015 ) . The relative abundance of each strain was determined using oligonucleotide barcode-specific primers ( Supplementaryfile 1 ) in a qPCR assay as described above . B . thetaiotaomicron strains were grown in triplicate anaerobically at 37 ˚C to mid-log phase ( OD600 ~0 . 3 ) in minimal media supplemented with methionine and vitamin B12 ( 0 , 0 . 37 , or 37 nM ) . RNA was extracted using a cell lysis buffer ( 10 mM Tris pH 8 . 0 , 1 mM EDTA , 0 . 2 mg lysozyme , 0 . 5 mg proteinase K ) and an RNeasy kit ( Qiagen , Hilden , Germany ) . DNA was removed using DNA-free DNA Removal Kit ( Invitrogen , Carlsbad , CA , USA ) , and RNA was again cleaned using an RNeasy kit ( Qiagen , Hilden , Germany ) . cDNA was then made with SuperScript II Reverse Transcriptase ( Invitrogen , Carlsbad , CA , USA ) using the manufacturer’s instructions , and RNA was removed with 1 N NaOH at 65 ˚C for 30 min and neutralized with 1 N HCl . Samples were then cleaned using a PCR purification kit ( Qiagen ) and cDNA was quantified using a Qubit ( Invitrogen , Carlsbad , CA , USA ) . Quantitative PCR was performed using SYBR FAST Universal Mastermix ( KAPA Biosystems , Wilmington , MA , USA ) and gene-specific primers ( Supplementary file 1 ) . Samples were normalized first to 16S rRNA expression for each individual sample and replicate , and then normalized to the expression level of each strain in 0 nM B12 ( Figure 1D ) or the btuG2 complement strain in 0 . 37 nM B12 ( Figure 2—figure supplement 1B ) . A standard curve and efficiency-corrected ∆Cq method was used to determine relative fold changes ( Bookout et al . , 2006 ) . Detection of BtuG2 from B . thetaiotaomicron lysates was performed by Western blot analysis using a custom-made rabbit anti-BtuG2 polyclonal antibody ( Cocalico Biologicals , Reamstown , PA , USA ) . A B . thetaiotaomicron strain expressing an HA-epitope tagged allele of the periplasmic protein SusA ( Shipman et al . , 1999 ) was grown to OD600 ~0 . 8 in minimal media with methionine , pelleted and washed in 1x cOmplete EDTA-free protease-inhibitor cocktail ( Roche , Basel , Switzerland ) before being pelleted and stored at −80 ˚C . Pellets were thawed and resuspended in PBS with proteinase K ( 0 , 10 , 50 or 100 µg/mL; AmericanBio , Natick , MA , USA ) , and incubated at 37 ˚C aerobically under continuous agitation ( 250 rpm ) for 8 hr . Cells were then pelleted and washed 3 times in 1x cOmplete EDTA-free protease-inhibitor cocktail , pelleted and stored at −80 ˚C . Thawed cells were lysed using BugBuster reagent ( Millipore Sigma , Burlington , MA , USA ) , 20 µg of clarified protein lysate was loaded onto an SDS-PAGE gel , transferred to a PVDF membrane and probed with rabbit anti-BtuG2 and rabbit anti-HA ( Santa Cruz Biotechnology , Dallas , TX , USA ) . B . thetaiotaomicron cultures were grown to OD600 ~0 . 6 in minimal media with methionine . Cells were pelleted ( ~3000 x g for 15 min at 4˚C ) and supernatant was filtered through 0 . 2 µm filter and stored temporarily on ice . Pellets were resuspended in breakage buffer ( 50 mM Tris pH 8 , 5 mM EDTA , 2 mM PMSF , 10% glycerol ) , lysed at 4˚C by sonication ( 40 Amps; 15 s ‘on’ and 30 s ‘off’; 3 min total ) , and clarified lysates were ultracentrifuged at 100 , 000 x g for 1 hr at 4˚C to separate membranes ( insoluble ) from cytoplasm/periplasm ( soluble ) fractions . Membrane fractions were resuspended in 250 µl of breakage buffer , while the cytoplasm/periplasm fraction was concentrated by centrifugal filtration ( 30K; Millipore Sigma , Burlington , MA , USA ) to 250 µl . Membrane and cytoplasm/periplasm fractions were temporarily stored on ice . Filtered supernatant was utracentrifuged at ~100 , 000 x g for 1 hr at 4˚C to remove outer membrane vesicles . The soluble fraction was then concentrated by centrifugal filtration to 250 µl . 20 µl each of membrane , cytoplasm/periplasm , and supernatant fractions were loaded onto SDS-PAGE gels and analyzed by Western blot . PVDF membranes were probed with rabbit anti-BtuG2 and mouse anti-RpoB ( Santa Cruz Biotechnology , Dallas , TX , USA ) as a cytoplasmic control . B . thetaiotaomicron strains were grown to OD600 ~0 . 6 in minimal media with methionine . Cells were pelleted , supernatant was removed , and cells were lysed with BugBuster reagent . Co-immunoprecipitation was carried out using FLAG HA Tandem Affinity Purification kit ( Millipore Sigma , Burlington , MA , USA ) according to the manufacturer’s instructions . Eluates were probed with rabbit anti-BtuG2 by Western blot . Overnight cultures of B . thetaiotaomicron strains were grown in minimal media with methionine . Supernatant donor strains were subcultured ( 1:100 ) into 60 mL fresh minimal media with methionine and allowed to grow to OD600 ~0 . 6 . Supernatant recipient strains were washed 3 times in minimal media without methionine or vitamin B12 , subcultured to a final OD600 ~0 . 001 in 1 mL minimal media without methionine or vitamin B12 , and incubated at 37 ˚C anaerobically for 8 – 12 hr . Supernatant donor cultures were pelleted and supernatant filtered ( 0 . 2 µm ) . Filtered supernatants were then ultracentrifuged at 100 , 000 x g for 1 hr at 4 ˚C to remove outer membrane vesicles and concentrated to <2 mL by centrifugal filtration ( 30K; Millipore Sigma , Burlington , MA , USA ) . Where indicated , concentrated supernatants were supplemented with 0 . 37 µM vitamin B12 or PBS and incubated for 20 min at room temperature aerobically on a nutator . Supernatants were then washed 4 times in 70 mL minimal media without methionine or vitamin B12 by centrifugal filtration ( 30K ) . Washed , concentrated supernatants were then applied to recipient cell cultures and incubated at 37 ˚C anaerobically . CFU measurements were taken at regular intervals . For assays involving IF , recombinant human IF ( Xeragenx LLC ) was incubated in 0 . 37 µM vitamin B12 or PBS for 20 min at room temperature on a nutator , and washed 4 times in 70 mL minimal media without methionine or vitamin B12 via centrifugal filtration ( 30K ) . IF ( ± vitamin B12 ) was provided either alone to recipient cell cultures ( e . g . , Figure 5A ) or with B12-free donor supernatants ( e . g . , Figure 5C ) at a final concentration of 10 nM IF per replicate . Cultures were incubated at 37˚C anaerobically and CFU measurements were taken at regular intervals . Recombinant BtuG2-6xHis or BtuG2-10xHis was expressed and purified from E . coli BL21 Rosetta ( DE3 ) carrying a modified pET21 vector . Cells were grown to OD600 ~0 . 6 before being induced for 3 hr in 0 . 5 mM IPTG at 37 ˚C under constant agitation . Cell pellets were lysed using BugBuster reagent ( Millipore Sigma , Burlington , MA , USA ) . Lysates were incubated for 1 hr at 4 ˚C with Ni-NTA agarose beads ( Qiagen , Hilden , Germany ) and washed with 12 – 18 mL of wash buffer ( 50 mM NaH2PO4 , 300 mM NaCl , 20 mM imidazole pH 7 . 4 ) , and eluted with 6 mL elution buffer ( 50 mM NaH2PO4 , 300 mM NaCl , 250 mM imidazole , pH 7 . 4 ) . BtuG2-6xHis or BtuG2-10xHis fractions were dialyzed overnight in 20 mM Tris pH eight before being spun through a PierceTM strong anion exchange column ( ThermoFisher Scientific , Waltham , MA , USA ) according to the manufacturer’s instructions . Eluted proteins were dialyzed twice for 4–8 hr in 4 L of PBS , pH 8 . Proteins were quantified by Bradford assay ( Bio-Rad , Hercules , CA , USA ) according to the manufacturer’s instructions . BtuG2-6xHis was incubated at a 1:1 molar ratio of cyanocobalamin to protein for 30 min at 25 ˚C and analyzed by SEC-MALS . BtuG2-10xHis was enriched for protein monomers by size exclusion chromatography and quantified using a NanoDrop spectrophotometer ( ThermoFisher Scientific , Waltham , MA , USA ) for use in SPR experiments ( described below ) . pET21_NESG was used to express and purify C-terminal 6xHis-tagged versions of the BtuG homologs BVU2056 , BACUNI04578 and BACCOPRO02032 as described above for BtuG2 . Each BtuG homolog , or BSA or PBS as controls , was incubated at a 1:1 molar ratio of cyanocobalamin to protein for 30 min at 25 ˚C before being spun through a centrifugal filter ( 30K; Millipore Sigma , Burlington , MA , USA ) to elute unbound cyanocobalamin . Proteins were washed 3 times with 400 µl PBS . Retained protein-cyanocobalamin complexes were resuspended in 100 µl PBS and analyzed by spectrophotometry for absorbance at 360 nm , corresponding to cyanocobalamin . Binding studies were performed at 25°C using a Biacore T100 optical biosensor ( GE HealthCare , Biacore , Piscataway , NJ , USA ) . Recombinant BtuG2-10xHis was purified from E . coli BL21 Rosetta ( DE3 ) as described above and immobilized on a NTA chip . Cyanocobalamin was injected at 0 . 8 , 0 . 4 , 0 . 2 , 0 . 1 , and 0 . 05 nM and the binding was monitored in single cycle kinetics in PBS ( Karlsson et al . , 2006 ) . Binding responses were double-referenced against non-specific binding to dextran and the NTA surface alone , and against injections of buffer alone . Binding affinity was determined by fitting the kinetics of the binding reaction to a 1:1 binding model using BioEvaluation software ( GE HealthCare , Biacore , Piscataway , NJ , USA ) . | Eating is the first step in an hours-long process that extracts the nutrients we need to live . It not only nourishes us , but also a vast community of bacteria in our gut called the microbiota . The gut microbiota acts like an extension of our immune system and helps us stay healthy in many ways . For example , it blocks pathogens from making us sick . But too many gut bacteria in the wrong parts of our intestines can be harmful . Some people are prone to developing a dangerous overgrowth of bacteria in their small intestine where most of our dietary nutrients get absorbed . This overgrowth can lead to many problems including vitamin B12 deficiency even when they eat plenty of it . To understand why , scientists must learn how microbes affect our ability to absorb nutrients from food and how the microbes themselves capture nutrients like vitamin B12 as they pass through our digestive tract . Now , Wexler et al . show that some gut microbes may be able to pirate vitamin B12 from us as it passes through the digestive tract . Wexler et al . showed that a protein called BtuG on the surface of a type of gut bacteria called Bacteriodes grabs onto vitamin B12 with extraordinary strength . In fact , these bacterial proteins bind to vitamin B12 so strongly that they can even pry it away from our own vitamin B12 collecting protein . When Bacteriodes with and without BtuG were placed in mice with no gut bacteria of their own , bacteria with BtuG rapidly outcompeted those lacking the protein . The experiments suggest that competition for vitamin B12 among microbes has favored bacteria that are better at capturing the nutrient . More studies are needed to learn whether BtuG contributes to vitamin B12 deficiencies in humans with gut bacteria overgrowth and determine the best ways to combat such deficiencies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Human gut Bacteroides capture vitamin B12 via cell surface-exposed lipoproteins |
Clarifying gene expression in narrowly defined neuronal populations can provide insight into cellular identity , computation , and functionality . Here , we used next-generation RNA sequencing ( RNA-seq ) to produce a quantitative , whole genome characterization of gene expression for the major excitatory neuronal classes of the hippocampus; namely , granule cells and mossy cells of the dentate gyrus , and pyramidal cells of areas CA3 , CA2 , and CA1 . Moreover , for the canonical cell classes of the trisynaptic loop , we profiled transcriptomes at both dorsal and ventral poles , producing a cell-class- and region-specific transcriptional description for these populations . This dataset clarifies the transcriptional properties and identities of lesser-known cell classes , and moreover reveals unexpected variation in the trisynaptic loop across the dorsal-ventral axis . We have created a public resource , Hipposeq ( http://hipposeq . janelia . org ) , which provides analysis and visualization of these data and will act as a roadmap relating molecules to cells , circuits , and computation in the hippocampus .
Gene expression profiling can be a powerful tool to understand the functionality and organization of cells and networks . For example , by relating the specific enriched or depleted genes to their corresponding ontologies in a given population , functional hypotheses can be generated at the intrinsic level ( e . g . , voltage-gated channel subunits ) and network level ( e . g . , ligand-receptor interactions ) . A different approach , agnostic to functional correlates of genes , can be taken by using gene expression profiles as a means to genetically delineate different populations of cells , either across classes or within a given class . In this way , gene expression profiling simultaneously clarifies complementary aspects of molecular , cellular , and circuit properties of cells . Transcriptional profiling in the mouse brain is becoming a powerful tool in neuroscience , owing to a host of complementary innovations and technologies . A variety of transgenic mice have emerged over the last decade ( Gong et al . , 2003; 2007; Taniguchi et al . , 2011 ) , which enables access to genetically defined populations of neurons , and a variety of techniques now exist for purifying labeled cells from surrounding tissue ( Okaty et al . , 2011a ) to obtain cell-class-specific transcriptomes ( Okaty et al . , 2011b ) . Although large-scale , quantitative gene expression profiling across neurons has typically been performed by microarray ( Belgard et al . , 2011; Siegert et al . , 2012; Sugino et al . , 2006 ) , more recently the technically superior RNA-seq ( Shin et al . , 2014 ) is finding application in the neurosciences ( Cembrowski et al . , 2016; Zeisel et al . , 2015; Zhang et al . , 2014 ) . Complementing these quantitative profiling methods is the mouse Allen Brain Atlas ( ABA ) ( Lein et al . , 2007 ) , providing histological information from in situ hybridization ( ISH ) assays . A combination of these techniques has been previously applied to study principal cells in the hippocampus . Microarray work has been used to study CA1 pyramidal cells ( Kamme et al . , 2003; Sugino et al . , 2006 ) as well as cells of the trisynaptic loop ( Deguchi et al . , 2011; Greene et al . , 2009; Lein et al . , 2004; Nakamura et al . , 2011; Zhao et al . , 2001 ) . Mining of the ABA has revealed molecularly defined subregions in multiple principal cell classes ( Dong et al . , 2009; Fanselow and Dong , 2010; Thompson et al . , 2008 ) . Most recently , RNA-seq work has been used to study CA1 pyramidal cells ( Cembrowski et al . , 2016; Zeisel et al . , 2015 ) . This work has helped to identify genetic differences within and across regions and as well revealed differences within canonical neuronal populations . Despite this extensive investigation , many aspects of the hippocampal transcriptome remain unresolved or warrant further investigation . Previous transcriptional profiling has predominantly focused on CA1 and CA3 pyramidal cells; markedly less work has examined DG granule cells and CA2 pyramidal cells ( but see Fanselow and Dong , 2010; Lein et al . , 2004; 2005 ) and no profiling has been performed for DG mossy cells . Additionally , recent work has suggested that DG granule cells may be a heterogeneous population along the dorsal-ventral axis ( Fanselow and Dong , 2010 ) , but this has not been systematically investigated . Perhaps most importantly , the technical superiority of RNA-seq may reveal governing organizational rules that may have not been resolved with ISH or microarray ( Cembrowski et al . , 2016 ) , suggesting that a systematic RNA-seq based approach may provide unparalleled insight into the transcriptional organization of the hippocampus . Here , we manually purified labeled excitatory cell populations from microdissected hippocampal regions , and used RNA-seq in combination with histological information from ABA to characterize gene expression quantitatively and histologically . This approach enabled analysis of hippocampal gene expression in a cell-class- and region-specific manner . We use this approach to examine both previously characterized and novel transcriptomes , and to understand the organizational schemes of gene expression within and across neuronal populations . These data and analysis tools are publicly available ( http://hipposeq . janelia . org ) , enabling users to examine gene expression at multiple levels of granularity in the hippocampus and providing a molecular blueprint to predict and investigate phenotypes at cellular , systems , and behavioral levels .
The hippocampus is grossly comprised of five excitatory cell populations; namely , granule and mossy cells of the dentate gyrus ( DG ) , and pyramidal cells of CA3 , CA2 , and CA1 . We sought to obtain and analyze transcriptomes for each of these five excitatory populations , which we operationally refer to as five distinct cell 'classes' for the remainder of the manuscript . Additionally , following recent work illustrating prominent dorsal-ventral differences within multiple canonical cell classes in the hippocampus ( Cembrowski et al . , 2016; Dong et al . , 2009; Fanselow and Dong , 2010; Thompson et al . , 2008 ) , we endeavored to profile the excitatory cells comprising the trisynaptic loop at the dorsal and ventral poles of the hippocampus; operationally , hereafter we will refer to cell classes at opposite poles to be from distinct 'regions' . Thus , in total , we aimed to profile eight distinct excitatory neuronal populations based upon cell-class and region specificity ( Figure 1a ) . 10 . 7554/eLife . 14997 . 003Figure 1 . Generation of hippocampal RNA-seq database . ( a ) Datasets included in the hippocampal RNA-seq characterization . Note , operationally , cell 'class' refers to gross cell type and 'region' refers to dorsal vs . ventral location . ( b ) Protocol underlying the generation of raw RNA-seq data . In a transgenic line in which cells of interest were fluorescently labeled ( left ) , the region of interest was microdissected ( dashed box ) . The isolated region was then dissociated , and labeled neurons were manually purified ( middle ) . RNA-seq data was generated from the purified cells . ( c ) Protocol underlying the processing of RNA-seq data . Raw reads were aligned , and then expression was quantified and statistically analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 00310 . 7554/eLife . 14997 . 004Figure 1—figure supplement 1 . Transgenic lines used to create cell-class- and region-specific transcriptomes . For each RNA-seq dataset , the corresponding transgenic expression pattern and approximate microdissected region is shown . Scale bar: overview: 500 μm; inset: 200 μm . Images of trisynaptic loop and CA2 expression patterns reprinted from Neuron , 89 ( 2 ) , Cembrowski et al . , Spatial Gene-Expression Gradients Underlie Prominent Heterogeneity of CA1 Pyramidal Neurons , 351–368 , 2016 , with permission from Elsevier . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 00410 . 7554/eLife . 14997 . 005Figure 1—figure supplement 2 . Reproducibility and purity of RNA-seq data . ( a ) Representative scatterplot of FPKM values for all genes for two replicates of dorsal CA3 , with a Pearson correlation coefficient r = 0 . 98 . ( b ) Correlation coefficients across replicates for each cell population . ( c ) Representative FPKM values corresponding to ERCC spike-in controls . Red points indicate undetected spike-in control; i . e . , FPKM=0 . Here , the Pearson correlation coefficient r = 0 . 94; for all replicates , r = 0 . 94 ± 0 . 01 ( n = 24 replicates ) . ( d ) FPKM values for genes corresponding to interneurons and non-neuronal cells . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 00510 . 7554/eLife . 14997 . 006Figure 1—figure supplement 3 . Reproducibility of RNA-seq quantification and differential expression . ( a ) Comparison of FPKM- vs . CPM-based enrichment for CA2 marker genes in Figure 3b . ( b , c ) As in Figure 3c , d , but for CPM-based analysis . ( d ) As in a , but for mossy cell marker genes of Figure 4b . ( e , f ) as in Figure 5b , c , but for CPM-based analysis . Insert: comparison of the number of differentially expressed genes for FPKM- vs . CPM-based approaches . ( g ) As in Figure 6c , but for CPM-based analysis . ( h ) Representative example of FPKM values for datasets obtained with TopHat and STAR alignment ( dorsal CA1 correlation r = 0 . 98; all datasets r = 0 . 98 ± 0 . 00 , Pearson correlation , mean ± SD , n = 8 datasets ) . ( i ) Representative example of differential expression results obtained from Tophat and STAR alignment ( dorsal vs . ventral CA1: 1015 genes identified using Tophat alignment , 1072 genes identified using STAR alignment ) . Colored points denote differentially expressed genes , with green color used here to better visualize data points . ( j ) Overlap in differentially expressed genes from the representative example in i . Here , 955/1015 = 94 . 1% of genes found using TopHat alignment were also identified with STAR . Across entire dataset , 95 . 0 ± 1 . 3% of differentially expressed genes found by TopHat approach were shared with STAR , with STAR identifying 6 . 8 ± 1 . 3% more genes than TopHat on average ( mean ± SD , n = 28 pairwise comparisons for each ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 006 To transcriptionally profile each of the eight populations ( Figure 1b , c ) , we first identified transgenic mouse lines that would allow for class and region specificity when combining local microdissections with fluorescence-based purification ( see Materials and methods; Figure 1—figure supplement 1 ) . We then microdissected the region of interest from the corresponding transgenic animal; this tissue was subsequently dissociated and the fluorescently labeled cells were purified by manual selection ( 112 ± 6 cells per biological replicate , mean ± SEM , n = 24 replicates ) ( Hempel et al . , 2007 ) . The sorted sample underwent library preparation and sequencing , the resulting raw RNA-seq reads were aligned , and expression was quantified and analyzed across samples ( see Materials and methods ) . To assess reproducibility , three biological replicates were ascertained for each dataset . Replicate datasets , corresponding to the same class-region pair , were well correlated with each other ( r = 0 . 98 ± 0 . 02 , mean ± SD , Pearson’s correlation coefficient; Figure 1—figure supplement 2a , b ) , and each replicate was devoid of marker gene cohorts associated with interneurons and non-neuronal cells ( Figure 1—figure supplement 2d ) . Thus , our obtained transcriptomes were internally consistent and cell-class specific , ensuring the integrity of our dataset . We began by exploring the gross relationships of hippocampal transcriptomes . Using hierarchical clustering ( see Materials and methods; Figure 2a ) we found the initial bifurcation corresponded to a divide between granule cells and non-granule cells , consistent with previous microarray ( Greene et al . , 2009 ) and ISH work ( Thompson et al . , 2008 ) . The second broad division of the dendrogram partitioned mossy cells from pyramidal cells and the final bifurcation in each limb corresponded to dorsal-ventral differences in each cell class , although the degree of within-class similarity was frequently comparable to across-class similarity ( Cembrowski et al . , 2016 ) . 10 . 7554/eLife . 14997 . 007Figure 2 . Gene expression in the hippocampus exhibits a variety of cell population- and region-specific expression . ( a ) Left: the hierarchical structure of gene expression in the hippocampus calculated by agglomerative clustering . Middle and right: Expression across replicates for marker genes associated with broad hippocampal populations ( middle ) or specific cell classes and regions ( right ) . Marker genes were selected based upon two-fold enrichment in all replicates in the target population ( s ) relative to all other replicates ( see Materials and methods ) . FPKM values displayed in the heat map were normalized on a gene-by-gene ( i . e . , column-by-column ) basis by the highest expressing sample for each gene . ( b ) Confirmation of gene expression profiles by ISH . In corresponding bar plots , RNA-seq FPKM values for each class/region dataset are displayed , with coloring adhering to the conventions of Figure 1a and fill vs . crosshatch indicating dorsal vs . ventral datasets . Scale bar: 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 007 We next cross-validated our RNA-seq hits with ABA ISH data ( see Materials and methods ) . From RNA-seq , many marker genes could be identified that corresponded to specific dendrogram bifurcations , both across broad hippocampal populations ( Figure 2a , left ) as well as within cell classes across regions ( Figure 2a , right ) . Importantly , these RNA-seq hits gave good agreement with ABA histological data , correctly predicting the enriched populations in ~81% of cases ( 124/153 genes where coronal ISH images were available , Figure 2b , Supplementary file 1; see Materials and methods ) . The consistency of RNA-seq with existing ISH data indicates that the two datasets can be used in conjunction to study spatial patterns of gene expression and delineate genetic boundaries across excitatory cell classes in the hippocampus . Complementing this , the quantitative whole-genome nature of RNA-seq enables well-principled numerical insight into the extent and properties of gene enrichment . For the remainder of the manuscript , we leverage these advantages to first examine individual cell classes , and then subsequently elucidate transcriptomes across cell classes and regions of the hippocampus . To examine the extent to which our RNA-seq dataset both recapitulated and expanded upon previous work , we searched for CA2-specific marker genes in our RNA-seq dataset . This search identified 41 genes with >3-fold enrichment in CA2 relative to all other populations , using relatively conservative search parameters ( Figure 3a , b; see Materials and methods ) . We compared these genes against previously known CA2-enriched genes from both literature ( Dudek et al . , 2016 ) and ABA mining ( Lein et al . , 2007 ) . Notably , although some of our retrieved genes were previously identified as enriched in CA2 ( Figure 3a ) , the majority of discovered genes were novel hits ( 66% , n=27/41; Figure 3b ) . Thus , our dataset recapitulated previous findings , but moreover revealed a host of previously unidentified genes with greater CA2 specificity ( Figure 3—figure supplement 1 ) , directly demonstrating the utility of RNA-seq relative to previous methodologies . Many novel genes were associated with functionally relevant neuronal ontologies , including cell adhesion and axon guidance ( Srgap2 , Vcan ) , neuropeptide signaling ( Ntsr2 ) , and calcium binding ( Scgn ) ( genes highlighted in orange , Figure 3b ) . 10 . 7554/eLife . 14997 . 008Figure 3 . Gene expression properties of CA2 pyramidal cells . ( a ) Heat map of replicate FPKM values for previously identified CA2 marker genes . ( b ) Heat map of replicate FPKM values for novel CA2 marker genes identified by RNA-seq . Orange indicates genes with previously characterized neuronal relevance . ( c ) The number of CA3- , CA2- , and CA1-specific genes , when restricting comparisons to solely these three cell populations in the dorsal hippocampus . A gene is denoted as X-fold enriched in a given CA region if it the average FPKM value is at least X-fold greater than the other CA regions . ( d ) Multidimensional scaling demonstrating the distance between CA3 , CA2 , and CA1 pyramidal cells . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 00810 . 7554/eLife . 14997 . 009Figure 3—figure supplement 1 . Recapitulation and extension of previous CA2 marker gene results . ( a ) Normalized FPKM values for CA2 marker genes identified in previous literature . Note that many previous marker genes have relatively high expression in non-CA2 populations . ( b ) As in ( a ) , but for all CA2 marker genes identified through ABA ISH CA2 Fine Structure Search . ( c ) As in ( a ) and ( b ) , but for RNA-seq-identified marker genes , to compare specificity relative to previous marker genes . ( d ) Quantitative comparison of the number of marker genes as a function of fold change identified from previous literature ( green ) , ABA ISH fine structure search ( blue ) , the union of previous literature and Fine Structure Search results ( teal ) , and RNA-seq ( black ) . Circular data point illustrates 3-fold enrichment criterion used to obtain RNA-seq marker genes shown in ( c ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 009 Although area CA2 shares some similarities with neighboring CA3 and CA1 regions , CA2 also exhibits features unique among these principal cells . Consequently , the extent to which CA2 pyramidal cells embody their own unique characteristics versus sharing properties with CA1 and/or CA3 is a subject of ongoing research ( Dudek et al . , 2016 ) , which can be directly and comprehensively addressed by transcriptome comparisons . Analyzing gene expression in dorsal CA3 , CA2 , and CA1 , we found each cell population had a similar number of enriched genes ( Figure 3c ) . Complementing this , applying multidimensional scaling to visualize the distances between CA3 , CA2 , and CA1 ( see Materials and methods ) , we found that the three regions were approximately equidistant ( Figure 3d ) . These results illustrated that CA2 is largely its own distinct region , rather than being a weighted combination of CA3 and CA1 features; i . e . , the physical intermediacy of CA2 did not correlate with transcriptional intermediacy . We also investigated mossy cells , a relatively uncharacterized excitatory cell population found within the hilus of the dentate gyrus . As with CA2 , we first investigated the extent to which mossy cells exhibited enriched genes relative to all other hippocampal excitatory neurons . Previous work has found one gene enriched in mossy cells ( Calb2 ) ( Fujise et al . , 1998 ) , which was recapitulated by our analysis ( Figure 4a ) ; in addition we identified 59 mossy cell-enriched genes in the hippocampus ( Figure 4b ) . Many of these genes play roles in neuronally relevant ontologies ( genes highlighted in orange , Figure 4b ) , including cellular adhesion and axon guidance ( Cntn6 , Ephb6 ) , calcium signaling ( Hpcal1 ) , ligand-receptor signaling ( Drd2 , Gal , Glp1r , Grm8 , Nmb ) , and regulation of transcription ( Prrx1 ) . 10 . 7554/eLife . 14997 . 010Figure 4 . Gene expression properties of hilar mossy cells . ( a ) Heat map of replicate FPKM values for the previously identified mossy cell marker gene Calb2 . ( b ) Heat map of replicate FPKM values for novel mossy cell marker genes identified by RNA-seq . Orange indicates genes with previously characterized neuronal relevance . ( c ) ISH profiles ( bottom ) for marker genes identified by RNA-seq ( top ) . Scale bar , overview: 500 μm; expanded: 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 010 Cross-validating these genes in the ABA , we found excellent agreement between mossy cell-enriched genes from RNA-seq and expression in the hilar cells ( 95%; n=36/38 agreement where coronal images available , Supplementary file 2; see Materials and methods ) . Interestingly , although some genes seemed to be expressed relatively uniformly across the long axis ( e . g . , Csf2rb2 , Figure 4c ) , many genes seemed to be enriched at specific locations along the long axis . For example , Nmb and Thbs2 exhibited expression near the dorsal pole of the hippocampus but lacked expression at the ventral horn of the hippocampus . Conversely , Calb2 and Tm4sf1 exhibited expression concentrated near the ventral pole of the hippocampus . In addition to this differential labeling across the hippocampus , differences were also seen in the labeling density at corresponding enriched regions ( e . g . , cf . Nmb with Thbs2 dorsally and Calb2 with Csf2rb2 ventrally ) , suggesting that mossy cells are a transcriptionally heterogeneous population of cells . Although significant work has been done examining dorsal-ventral differences in CA3 and CA1 ( Cembrowski et al . , 2016; Thompson et al . , 2008 ) , differences in dentate gyrus granule cells have received relatively little attention . Previous work ( Fanselow and Dong , 2010 ) has suggested that domains specified by the dorsal marker gene Lct and the ventral marker Trhr may correspond to tripartite molecular divisions of the dentate gyrus ( Figure 5a ) : here , Lct and Trhr expression respectively represent the dorsal and ventral divisions , whereas the intermediate domain is characterized by weak expression of both genes . Importantly , our RNA-seq work recapitulated both of these marker genes , suggesting that our data could be used to quantitatively explore the degree and patterns of granule cell heterogeneity . 10 . 7554/eLife . 14997 . 011Figure 5 . Dorsal-ventral differences in dentate gyrus granule cells . ( a ) RNA-seq ( top ) and ISH ( bottom ) profiles of Lct and Trhr , two previously identified marker genes respectively enriched in dorsal and ventral granule cells . Scale bar: 500 μm . ( b ) FPKM scatterplot of average dorsal and ventral GC transcriptomes . Data points represent individual genes , with genes highlighted in red indicating differential expression . ( c ) Number of genes enriched at the poles of DG as a function of fold change . ( d ) Example ISH profiles ( bottom ) of dorsal GC marker genes obtained by RNA-seq ( top ) . ( e ) As in ( d ) but for ventral GC marker genes . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 01110 . 7554/eLife . 14997 . 012Figure 5—figure supplement 1 . Dorsal-ventral differences in dentate gyrus granule cells . ( a ) Examples of regionally enriched marker genes with neuronally relevant functionality . ( b ) RNA-seq and ISH profiles of representative novel dorsal marker genes . Scale bar , overview: 500 μm; expanded: 100 μm . ( c ) As in b , but for novel ventral marker genes . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 012 We first used our RNA-seq dataset to examine gene expression for granule cells at the two poles of the dentate gyrus . Notably , hundreds of genes were differentially expressed between these two poles ( Figure 5b ) , and corresponded to large fold changes ( Figure 5c ) . Many of these genes were involved in neuronally relevant functions and were cross-validated by ISH ( 58% , n=33/57 agreement where coronal images available , Figure 5—figure supplement 1 , Supplementary file 3; see Materials and methods ) . Given the agreement between RNA-seq and ISH , we used the ABA to investigate genetic domains in granule cells . The genetic domain specified by Lct was recapitulated by multiple dorsal marker genes ( Figure 5—figure supplement 1b ) , including Gsg1l , Spata13 , and Stra6 ( Figure 5d ) . Conversely , little agreement was seen between ventral marker genes: the domain specified by Trhr was found to differ from the patterns observed for other markers ( e . g . , Cpne7 , Grp , and Nr2f2; Figure 5e , Figure 5—figure supplement 1c ) . These findings indicate that the genetic boundary defined by Lct may correspond to a transcriptionally well-defined subpopulation , but an equally well-defined ventral subpopulation does not appear to be present . Despite this , all novel marker genes validated by ISH appeared to be expressed in gradients across the long axis ( Figure 5—figure supplement 1b , c; we did not find genes that were selectively expressed in either the upper or lower blades of DG granule cells ) , suggesting that granule cell transcriptional identity exists in a continuous spectrum in this axis . The granule cell marker genes Lct and Trhr were previously shown to be enriched in other class-region populations; namely , Lct is expressed in dorsal CA1 and Trhr is expressed in ventral CA3 ( Cembrowski et al . , 2016; Dong et al . , 2009; Thompson et al . , 2008 ) ( see also Figure 5a ) . This raises the intriguing possibility that there exist genes that are enriched in a region- but not class-specific manner; i . e . , genes enriched dorsally or ventrally across multiple cell classes . We next analyzed this the context of the dorsal versus ventral cell classes of the trisynaptic loop . We first identified the number of expressed genes that were >2 fold enriched between poles on a class-by-class basis ( top values , Figure 6a ) . From here , we searched for genes that were associated with enrichment at the same pole across multiple class comparisons; e . g . , the genes Cadm2 and Mgll were found to be >2 fold dorsally enriched in every dorsal-ventral comparison , whereas Resp18 and Efnb2 were found to be ventrally enriched ( Figure 6b; corroborated by ABA , Figure 6d ) . In general , many genes were found that obeyed region-specific enrichment across multiple cell classes ( Figure 6c ) . To compare this empirically determined number of region-enriched genes relative to the number expected by chance , we calculated the total number of genes that were expressed in each cell class/region combination ( defined as the number of genes with FPKMavg>10; Figure 6a ) , and compared this experimental data to a null model where gene names were drawn at random from the list of expressed genes ( see Materials and methods ) ( Figure 6c ) . Interestingly , in almost every possible comparison ( n=5/6 pairwise enrichment combinations and 2/2 triplicate enrichment combinations ) , the number of enriched genes that were shared across cell classes in a region-specific manner were significantly greater than that expected by chance . 10 . 7554/eLife . 14997 . 013Figure 6 . Regionally enriched genes invariant to principal cell class . ( a ) For each trisynaptic loop dataset , the number of enriched genes when comparing to the same cell class at the opposite pole ( >2-fold difference ) , as well as the total number of expressed genes for each dataset ( FPKMMIN>10 ) , are shown ( top and bottom values respectively ) . ( b ) Example genes enriched in a region- , but not cell-class- , specific manner . ( c ) The number of dorsally and ventrally enriched genes shared across cell classes . Both the observed RNA-seq data ( horizontal lines ) and null distribution ( mean ± 2SD ) are shown . ( d ) Sagittal ISH profiles of the example region-enriched genes . Scale bar: 500 μm . ( e ) Heat map of all genes found enriched in a region-specific manner across the trisynaptic loop ( n=37; null distribution predicts 6 . 0 ± 7 . 1 ( mean ± 2SD ) , p<1e-6 ) . Orange text: genes also identified as differentially expressed in medial entorhinal cortex ( MEC ) ( Ramsden et al . , 2015 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 01310 . 7554/eLife . 14997 . 014Figure 6—figure supplement 1 . Dorsal and ventral genes enriched across hippocampus and MEC . Left: atlas showing dorsal-ventral extent of MEC in sagittal section . Middle , right: example ISH profiles of dorsally- and ventrally-enriched genes identified by RNA-seq and found to be predictive of MEC enrichment ( Ramsden et al . , 2015 ) . Scale bar: 500 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 014 We examined the ontologies associated with the 37 genes that were enriched across all dorsal-ventral comparisons ( i . e . , the n=12 dorsally and 25 ventrally enriched genes from the triplicate comparisons of Figure 6c ) . Although the enriched genes spanned a variety of ontologies , many genes that emerged as being region- but not class-enriched corresponded to cellular adhesion and axon guidance; for example , the cellular adhesion molecules Cadm1 , Cadm2 , the ephrins/receptors Epha5 , Epha7 , Efnb2 , as well as Dagla , Odz4 , Timp2 and Negr1 ( Figure 6e ) . Given the abundance of genes expressed in a region-specific manner in the hippocampus , we next examined whether these region-specific genes would predict similar patterning outside of the hippocampus . Recently , RNA-seq has been conducted on the dorsal and ventral poles of the medial entorhinal cortex ( MEC ) ( Ramsden et al . , 2015 ) , providing a direct comparison with our data . Strikingly , of the 37 genes identified as regionally enriched in all principal cells of the trisynaptic loop , 43% ( n=16/37 ) were identified as differentially expressed in the dorsal-ventral axis of MEC with an identical directionality ( Figure 6e , orange text ) . Examining the spatial patterns of gene expression in the MEC in sagittal sections , we found that these regional-specific genes exhibited a broad range of expression profiles across the dorsal-ventral axis , attenuating in labeling density and/or intensity on a gene-by-gene basis ( Figure 6—figure supplement 1 ) . Attenuation was also generally not constrained to a fixed cell class: although some genes exhibited expression restricted to single lamina ( e . g . , Inf2 , Etv1 ) , other genes were more broadly expressed ( e . g . , Efnb2 across two laminae; Cadm2 , Crym , Hap1 , and Odz4 across 3+ laminae ) . The preceding work considered cell classes and regions determined a priori to analysis . To complement this , we next used a wholly data-driven approach to analyze the hippocampal transcriptome through Weighted Gene Co-expression Network Analysis ( WGCNA; see Materials and methods ) ( Zhang and Horvath , 2005 ) . This method identifies highly correlated expression of gene modules across subsets of samples . We used the top 1000 most variable genes from the full hippocampal dataset , and using WGCNA , identified eight gene modules that were enriched in various subpopulations of hippocampal excitatory neurons ( Figure 7a , b; see Materials and methods ) . 10 . 7554/eLife . 14997 . 015Figure 7 . Weighted gene co-expression network analysis ( WGCNA ) of hippocampal excitatory neuron transcriptomes . ( a ) Top and middle: hierarchical clustering and normalized expression of the 1000 most variable genes , respectively . Bottom: colors denoting the modules obtained from WGCNA . ( b ) Six modules obtained from a , with the average expression shown and significantly enriched terms highlighted . Each module is named according to the gross overall expression profile across datasets . ( c ) Genes associated with long-term potentiation significantly enriched in modules . ( d ) As in c , but with genes associated with Parkinson’s disease . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 015 The functional implications of these modules were then examined by using DAVID ( Huang et al . , 2009a; 2009b ) ( see Materials and methods ) to identify statistically significant Gene Ontology and KEGG Pathway terms ( Kanehisa and Goto , 2000; Kanehisa et al . , 2014 ) . Interestingly , although by definition the genes present within a given module were not shared across modules , many associated ontologies and pathways were common between modules . For example , the KEGG annotation 'Long Term Potentiation' was enriched for distinct modules associated with DG granule cells , CA1 pyramidal cells , and CA2/3 pyramidal cells ( Figure 7c ) , illustrating that specific genes that underlie LTP vary between cell classes despite all cell classes expressing genes related to LTP . Similarly , this approach also enabled us to identify specific modules with disease annotations; for example , both mossy cells and dorsal dentate gyrus granule cells were enriched for genes associated with disease terms ( e . g . , Parkinson’s; Figure 7d ) .
A major result of our RNA-seq analysis was the identification of region- and/or class-specific marker genes . Our profiling recapitulated many marker genes known a priori , but in addition , demonstrated an abundance of previously unappreciated marker genes . Consequently , our work greatly expands the total number of marker genes and enables several complementary lines of inquiry based upon these findings . First , the marker genes uncovered here serve as candidates for obtaining genetic access to well-defined populations of neurons . Due to the inherently quantitative nature of RNA-seq , the strength and specificity of candidate genes can be evaluated by examining expression in both on-target and off-target populations; promoters associated with genes that are sufficiently restricted in expression can be employed for designing transgenic animals or viruses to enable genetic access . We emphasize that the Cre lines used here for labeling cells for transcriptional profiling ( Figure 1—figure supplement 1 ) may already provide sufficient genetic access to many excitatory subpopulations of the hippocampus; thus , this work offers both existing and new ways to selectively target and manipulate specific populations of neurons . Second , many of the marker genes identified here can be tied to specific functionality ( e . g . , Figure 3b , 4b , Figure 5—figure supplement 1a ) , allowing novel marker genes to be used for hypothesis generation . These hypotheses can be investigated through a variety of perturbations ( e . g . , siRNA , knockouts , or CRISPR-Cas gene editing ) combined with other experiments ( e . g . , physiology , behavior ) . Similarly , many marker genes identified here are annotated as specific disease-related genes ( e . g . , Figure 7b , d ) , which can help to reveal both the molecules and cell classes to examine in pathological conditions and disease models . Our work here identified two components of transcriptional variability in the cells of the trisynaptic loop; namely , differences across different neuronal populations ( across-class ) and differences across the dorsal-ventral axis ( across-region ) . On a gene-by-gene basis , these two organizational principles could be observed either individually or simultaneously ( Figure 8 ) . Transcriptional differences across classes of the trisynaptic loop have been identified and appreciated for some time ( e . g . , Greene et al . , 2009; Lein et al . , 2004 ) . Indeed , it is likely that these transcriptional differences are fundamental in producing across-class variability in morphology , physiology , and connectivity , ultimately underlying the diverse and distinct roles that the different cell classes are believed to play in hippocampal processing ( Mizuseki et al . , 2012; Neunuebel and Knierim , 2014 ) . Across-region differences , at a cell-class-specific level , have recently been identified to be present in CA3 and CA1 pyramidal cell populations ( Cembrowski et al . , 2016; Thompson et al . , 2008 ) . Here , we show that granule cells of the dentate gyrus also adhere to this rule , with marked differences present in the transcriptomes between dorsal and ventral poles ( Figure 5b , c ) . Indeed , the transcriptional distance between dorsal and ventral granule cells is similar to that between different populations of cells ( Figure 2a ) , a finding similar to that found between CA3 and CA1 pyramidal cells ( Figure 2a ) ( Cembrowski et al . , 2016 ) . 10 . 7554/eLife . 14997 . 016Figure 8 . Gene expression patterns of excitatory hippocampal neurons . Genes can be expressed in relatively similar abundances across the hippocampus ( Slc17a7; upper left ) , vary in a cell-class-specific ( Fibcd1; upper right ) or region-specific manner ( Cadm2; lower left ) , or vary in both cell-class- and region-specific manners simultaneously ( Wfs1; lower right ) . In each panel , light magenta denotes the spatial extent of the hippocampus , dark magenta illustrates the CA1 region in particular , and the dots indicate the location and intensity of labeling from ISH . RNA-seq profiling results are provided for each gene . Images from the Allen Brain Explorer v2 . DOI: http://dx . doi . org/10 . 7554/eLife . 14997 . 016 Finding genes that are regionally enriched across cell classes ( Figure 6 ) is surprising and warrants further investigation . The fact that many of these genes are involved in cell adhesion and axon guidance , in conjunction with the observation that they are enriched along the dorsal-ventral axis in multiple areas of the brain , suggests that they may generally be used for maintaining polarity in the mature brain . During development , gradients of gene expression are used for proper patterning of neural circuits ( Sansom and Livesey , 2009 ) ; in a similar fashion , these genes may reflect the mature counterpart that actively maintains spatial identity . A central goal of neuroscience is to disentangle and understand the vast complexity of neuronal populations , both within and between cell classes . Next-generation RNA-seq provides a comprehensive means of clarifying cellular identities in the hippocampus and complements other gene expression analyses ( Dong et al . , 2009; Fanselow and Dong , 2010; Lein et al . , 2007; Thompson et al . , 2008 ) . Our work here furthers understanding of across-class differences , but also emphasizes the high degree of transcriptional variability that can be present within a given population ( e . g . , across the dorsal-ventral axis of the hippocampus ) . Pyramidal cells in CA1 and CA3 ( Cembrowski et al . , 2016; Thompson et al . , 2008 ) , as well as both mossy cells ( Figure 4 ) and granule cells ( Figure 5 ) of the dentate gyrus , appear to exhibit a high degree of heterogeneity across the long axis . Notably , this within-class , dorsal-ventral heterogeneity can exhibit different organizational principles; for example , CA3 pyramidal cells have been shown to conform to discrete subpopulations ( Thompson et al . , 2008 ) , whereas CA1 pyramidal cells ( Cembrowski et al . , 2016 ) , DG mossy cells ( Figure 4 ) , and DG granule cells ( Figure 5 ) do not exhibit clear subdomain organization . It is important to emphasize that our results , although illustrating a high degree of transcriptional heterogeneity in principal cells of the hippocampus , likely underestimate the total amount of variability for several reasons . First , comparing the dorsal and ventral poles may miss differences present along other dimensions , as well as differences present at spatially intermediate locations . Second , our population-level approach may miss heterogeneity that is present at a subpopulation or single-cell level , including transcriptional signatures associated with specialized but sparse excitatory cell classes ( e . g . , radiatum giant cells ( Gulyas et al . , 1998 ) , semilunar granule cells ( Williams et al . , 2007 ) , and CA3 granule cells ( Szabadics et al . , 2010 ) ) . Finally , transcriptional properties may vary in ways other than geographical location; for example , differential gene expression for cells of the same region that target different downstream locations ( Cembrowski et al . , 2016; Sorensen et al . , 2015 ) . All data presented here are accessible on the Hipposeq website ( http://hipposeq . janelia . org ) , an interactive database that allows user-friendly analysis and visualization of gene expression data for individual genes , cohorts of genes , and entire transcriptomes . Through this site , data can be mined for a priori genes of interest , or alternatively investigated with discovery-based analysis tools . Raw and processed data are also available for download , enabling the user to export data into their own environment for more specialized analyses . This website and associated dataset expand upon and complement other existing publicly available gene expression databases , mostly notably the ABA ( Lein et al . , 2007 ) . Our RNA-seq approach offers advantages that circumvent traditional issues associated with ISH; namely , providing class-specific data with a large dynamic range , helping to circumvent confounds associated with image-based analyses of gene expression that can be limited by changes in cell density and/or labeling across sections . Of course , RNA-seq also has limitations relative to ISH , including an inherent lack of spatial information . In this way , the combination of Hipposeq and the ABA ISH atlas provides a powerful set of tools that enables quantitative and histological whole genome insight into gene expression in the hippocampus .
All transgenic mice used ( namely , granule cells from both blades of the DG: Rbp4-Cre KL100 , mossy cells of the hilus: Lypd1-Cre NR151 , CA3 and CA2 pyramidal cells: Mpp3-Cre KG118 , CA1 pyramidal cells: Vipr2-Cre KE2 ) were generated by the Gene Expression Nervous System Atlas ( GENSAT ) project ( Gong et al . , 2003; Gong et al . , 2007 ) . Transgenic lines were maintained on a C57bl/6J background , with each line backcrossed at least one generation prior to use ( note that strain-specific gene expression differences are likely minor [Sandberg et al . , 2000] ) . Cre expression was reported by an Ai9 ( tdTomato ) mouse cross ( Madisen et al . , 2010 ) , and double-positive mice of either sex were sacrificed ( age P26-P35 , either single- or group-housed ) within a 3 hr time window approximately midway through the light cycle , with microdissection locations shown in Figure 1—figure supplement 1 . In all cases , manual sorting to purify for fluorescent neurons from microdissected slices was performed according to previous methods ( Hempel et al . , 2007 ) . For each cell class/region combination , three biological replicates ( i . e . , cell class/region from a different animal of the same genotype ) with sufficient reproducibility ( within-class Pearson correlation coefficient >0 . 90 , a criterion determined after analysis ) were obtained; biological replicates were re-obtained for datasets for correlations <0 . 90 . This was the only exclusion criterion for datasets . No technical replicates were used in this study . Three biological replicates have been previously shown to be sufficient in detecting differences in gene expression known a priori ( Cembrowski et al . , 2016 ) . On average , 112 ± 6 cells ( mean ± SEM , n = 24 replicates ) were recovered in the final purified pool for library preparation and sequencing . Total RNA was isolated from each sample using PicoPure RNA Isolation kit ( Life Technologies , Frederick , MD ) including the on-column RNase-free DNase I treatment ( Qiagen , Hilden , Germany ) following the manufacturers’ recommendations . Eluted RNA ( 11 ul ) was dried in a speed vac to approximately 2–4 ul . ERCC control RNAs ( Life Technologies ) were added using 1 ul of 1:100 , 000 dilution for every 50 cells . cDNA was amplified from this input material using Ovation RNA-seq v2 kit ( NuGEN , San Carlos , CA ) . Approximately half of the resulting cDNA was used to make the sequencing libraries using the Ovation Rapid DR Multiplexing kit ( NuGEN ) . Four barcoded libraries were pooled per sequencing lane on a HiSeq 2500 ( Illumina , San Diego , CA ) and single-end 100 bp reads were generated . No randomization or blinding was used for sorting , library preparation , or sequencing . Reads for each library ( 37 . 9 ± 1 . 3 million per replicate , n = 24 replicates ) were mapped using TopHat v2 . 0 . 6 ( http://ccb . jhu . edu/software/tophat/index . shtml ) ( Trapnell et al . , 2009 ) against the mouse genome build NCBIM37 ( mm9 ) combined with sequences corresponding to ERCC spike-in controls . The following options were used: '--num-threads 8 --GTF mouseGtf . gtf' , where mouseGtf . gtf reflects the concatenated Ensembl NCBIM37 transcript annotation file and the annotated ERCC spike-in controls . With these settings , 77 . 5 ± 1 . 0% ( 29 . 6 ± 2 . 0 million per replicate , n = 24 ) of all reads aligned at least once to either the annotated transcriptome or genome . After mapping , quantification and differential expression of the annotated mouseGtf . gtf was performed using Cuffdiff v2 . 1 . 1 ( http://cole-trapnell-lab . github . io/cufflinks/ ) ( Trapnell et al . , 2010 ) using the accepted_hits . bam files for all replicates , with three biological replicates used for each dataset . The following options were used: “--frag-bias-correct mouseFa . fa --mask-file mouseMask . gtf --max-bundle-frags 10000000 --num-threads 8 --multi-read-correct --no-effective-length-correction” , where mouseFa . fa is the Ensembl NCBIM37 reference FASTA and mouseMask . gtf is a mask file that ignores all alignments corresponding to genes annotated in mouseGtf . gtf annotated as tRNA , rRNA , or snRNA . In addition , from inspecting gene tracks we also noticed that a few loci ( namely , Nat8l , Psd2 , Xist , Gm15459 , and Gm10335 ) would occasionally produce many identical reads to one specific sequence with few or no alignment reads found elsewhere; these loci were also masked . Finally , when not explicitly examining ERCC controls ( Figure 1—figure supplement 2c ) , ERCC 'loci' were also included in the mask . When considering spike-in controls , FPKM values were found to be highly reproducible across replicates ( r = 0 . 94 ± 0 . 01 , n = 24 replicates , Pearson correlation coefficient ) . The resulting data were analyzed in the R environment using a combination of cummeRbund v3 . 0 ( http://compbio . mit . edu/cummeRbund/ ) ( Goff et al . , 2013 ) and custom scripts . General analysis conventions were as follows: a conventional threshold of FDR <0 . 05 was used for differential expression , allowing both under- and overexpressed genes to be identified ( i . e . , two-sided ) ; a gene was considered X-fold enriched in a given region , relative to other regions , when the mean FPKM value was at least X-fold greater for all corresponding pairwise comparisons ( e . g . , for gene A to be X-fold enriched in dorsal CA1 relative to dorsal CA2 and dorsal CA3 , FPKMA , CA1dorsal > X∙FPKMA , CA2dorsal and FPKMA , CA1dorsal > X∙FPKMA , CA3dorsal ) ; Pearson correlation coefficients were used to compare across datasets; error bars for FPKM values were taken from Cuffdiff’s 95% CI model; and gene expression was required to obey FPKM>10 in at least one population to be included in differential expression or enriched population analyses . No randomization or blinding was used for computational handling of data . Raw and processed RNA-seq datasets were deposited in the National Center for Biotechnology Information ( NCBI ) Gene Expression Omnibus ( GEO ) , accession number GSE74985 , and analysis scripts can be downloaded from Github ( https://github . com/cembrowskim/hipposeq . git ) . For hierarchical clustering of datasets ( Figure 2a ) , whole-transcriptome averaged FPKM values were processed by adding a pseudocount of 1 to all values , log10 transformed , and normalized on a sample-by-sample basis by the sample sum of the transformed FPKM value . The pairwise Jensen-Shannon distance was calculated across samples , and agglomerative clustering was performed on the distance matrix using complete linkage . When plotting normalized FPKM heat maps ( Figures 2a , 3a , b , 4a , b , 6e , 7a , c , d; Figure Supplements 1-2d , 3-1a , b , c , 5-1a ) , each gene ( i . e . , column ) was normalized by dividing the expression values by the highest FPKM value across samples . When visualizing expression in single-gene bar plots ( Figures 2b , 4c , 5a , d , e , 6b; Figure Supplements 5-1b , c , 6–1 ) , x-axis values are individual class/region datasets , such that filled bars represent dorsal samples , cross-hatched bars represented ventral samples , and coloring adheres to the convention of Figure 1a; y-axis values are gene expression values in FPKM . For identifying marker genes corresponding to the clusters of Figure 2a , we searched for genes that were more than two fold enriched in every replicate of the desired sample ( s ) relative to the remaining sample ( s ) . For identifying cell class-specific genes ( Figure 3a , b; Figure 4a , b ) , we searched for genes that were more than 3-fold enriched on average in the desired population , relative to all other populations . For identifying regionally enriched granule cell marker genes with neuronal relevance ( Figure 5—figure supplement 1a ) , we searched for genes that were more than 3-fold enriched at either of the two poles relative to the opposite pole . For multidimensional scaling ( MDS ) of CA3 , CA2 , and CA1 dorsal datasets ( Figure 3d ) , replicate FPKM values were processed by adding a pseudocount of 1 to all values , log10 transformed , and normalized on a sample-by-sample basis by the sample sum of the transformed FPKM value . The pairwise Jensen-Shannon distance was calculated across samples , and agglomerative clustering was performed on the distance matrix using complete linkage . Two-dimensional classical MDS was performed by using cmdscale in R with default arguments . When examining the number of regionally enriched genes invariant to principal cell class ( Figure 6 ) , genes obeying FPKM>10 for each class/region combination were obtained ( 'expressed genes' ) , and from this list , for each dorsal-ventral comparison genes also being >2 fold enriched at either pole were identified ( 'enriched genes' , Figure 6a ) . The overlap in enriched genes across respective regions was then determined ( horizontal lines , Figure 6c ) . To derive chance levels for overlap , Monte Carlo simulations were performed . Here , genes were drawn at random without replacement from the expressed genes list from each dataset , with the total number of genes drawn equal to the number of enriched genes for each pairwise comparison . The ensuing results were analyzed analogously to empirical data . This process was repeated 1000 times to characterize the chance distributions for all comparisons . For comparing to MEC RNA-seq ( Figure 6—figure supplement 1 ) , we first identified all genes that were >2 enriched in each dorsal-ventral comparison for cell class in the trisynaptic loop ( n=37 ) . Next , we compared this list to those identified as differentially expressed ( FDR < 5% ) from RNA-seq on microdissected dorsal and ventral poles of the MEC ( S7 Dataset , Ramsden et al . , 2015 ) . Genes present in this list that obeyed the same directionality of enrichment are highlighted in Figure 6e ( dorsal: 6/12; ventral: 10/27 ) . WGCNA analysis ( Zhang and Horvath , 2005 ) was performed ( Figure 7 ) by first identifying the 1000 most variable genes across datasets ( FPKMMIN=5 ) . The correlation matrix C of these genes was subsequently obtained according to Cij=rij , where rij was the Pearson correlation coefficient of genes i and j across datasets . The dissimilarity matrix D was then calculated by taking Dij=1− ( Cij+12 ) . Divisive hierarchical clustering was then performed on D by the diana method in R with default parameters , and modules were obtained by choosing a cut height of 10 on the computed dendrogram . Genes associated with each cohort were analyzed by DAVID ( Database for Annotation , Visualization and Integrated Discovery ) ( Huang et al . , 2009a , 2009b ) . The default Gene Ontology terms ( GOTERM_BP_FAT , GOTERM_CC_FAT , GOTERM_MF_FAT ) ( Ashburner et al . , 2000 ) and KEGG ( Kyoto Encyclopedia of Genes and Genomes ) pathways ( Kanehisa and Goto , 2000; Kanehisa et al . , 2014 ) were analyzed by the Functional Annotation Chart , and terms that obeyed a Benjamini ( adjusted ) p-value <0 . 05 were considered for further analysis . First , to cross validate results of the TuxedoSuite pipeline with alternative quantification and differential expression software , we used HTSeq to quantify expression in a count-based fashion ( Figure 1—figure supplement 3a–g ) . For each sample , the mapped reads from TopHat were quantified by HTSeq using htseq-count with the previous mouseGtf . gtf and the following options: “--format=bam --stranded=no” . After quantification , the count data for individual samples were merged into one file , and analyzed by DESeq2 and custom scripts in the R environment . Values are reported in counts per million ( CPM ) , and a CPM cutoff of 20 was used for fold change analysis , which retained a similar number of genes to FPKM=10 threshold used elsewhere . To examine the choice of alignment software on quantification and differential expression ( Figure 1—figure supplement 3h–j ) , in a second set of analysis we first aligned reads with STAR v2 . 5 . 1a ( Dobin et al . , 2013 ) rather than TopHat . For each sample , reads were mapped according to “--runThreadN 8 --genomeDir starGenomeDir --outSAMtype BAM SortedByCoordinate” , where starGenomeDir was the directory containing the STAR genome index files . Output BAM files were then processed for quantification and differential expression according to the Cuffdiff approach described above . When cross-validating the results of RNA-seq , we examined coronal ISH images from the ABA ( Lein et al . , 2007 ) ( except for Figure 6d , where sagittal sections were used to visualize dorsal and ventral trisynaptic loops in the same section ) . To validate genes identified by RNA-seq as enriched in hierarchical clustering subgroups ( Figure 2a ) , ISH expression profiles at the corresponding dorsal or ventral section were examined , and the validation was counted as a success if there was obvious expression by eye in the enriched subgroup ( occurring in ~81% of cases; 124/153 genes , Supplementary file 1 ) . Similar approaches were used for mossy cell marker genes ( Figures 3b and 4b , Supplementary file 2 ) . Expression in all representative ISH images shown in the text was consistent with other sections near the same anterior-posterior location in the same animal , as well as with at least one additional animal in the ABA ( with the exception of Nr2f2 ( Figure 5e ) , which had ubiquitous labeling in sagittal sections , inconsistent with the coronal expression pattern employing a different animal and probe , as well as Inf2 ( Figure 6—figure supplement 1 ) , which had only one animal ) . To examine reproducibility of RNA-seq vs . ISH cross-validation , two additional observers were independently shown a randomly chosen subset of images corresponding to enriched genes of Figure 2a and asked to identify the cell class ( es ) that exhibited expression . This scoring was performed blind to the RNA-seq result and incorporated negative control images wherein RNA-seq expression was not cell class-specific . This blind , outside-observer assessment correctly identified the enriched populations at a similar success rate ( 77% and 88% of genes for two independent observers , with n = 20/26 and 23/26 randomly selected genes correctly identified respectively ) . Images of large regions of tissue ( i . e . , complete dorsal and ventral CA1 ) were acquired on a whole-slide digital scanner ( Pannoramic 250 Flash , Perkin Elmer , Waltham , PA ) using a 20x objective . Cellular resolution images were acquired with a confocal microscope ( LSM 710 Carl Zeiss Microscopy , Jena , Germany ) using a 20x objective . Some images were post-processed in Fiji , including pseudocoloring to adhere to the coloring conventions of different cell classes . | Both mouse and human brains are made up of many millions of cells called neurons that are interconnected to form circuits . These neurons are not all the same , because different classes of neurons express different complements of genes . Neurons that express similar genes tend to look and act alike , whereas neurons that express different genes tend to be dissimilar . Cembrowski et al . have used a technique called next-generation RNA sequencing ( RNA-seq ) to determine which genes are expressed in groups of neurons that represent the main cell types found in a part of the brain called the hippocampus . This brain region is important for memory , and was chosen because the location and appearance of the main cell types in the hippocampus were already well understood . The approach revealed that the main types of neurons in the mouse hippocampus are all very different from each other in terms of gene expression , and that even neurons of the same type can exhibit large differences across the hippocampus . Cembrowski et al . created a website that will allow other researchers to easily navigate , analyze , and visualize gene expression data in these populations of neurons . Future work could next make use of recent technological advances to analyze gene expression in individual neurons , rather than groups of cells , to provide an even more detailed picture . It is also hoped that understanding the differences in gene expression will guide examination of how the hippocampus contributes to memory and what goes wrong in diseases that affect this region of the brain . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2016 | Hipposeq: a comprehensive RNA-seq database of gene expression in hippocampal principal neurons |
Risk for Atrial Fibrillation ( AF ) , the most common human arrhythmia , has a major genetic component . The T-box transcription factor TBX5 influences human AF risk , and adult-specific Tbx5-mutant mice demonstrate spontaneous AF . We report that TBX5 is critical for cellular Ca2+ homeostasis , providing a molecular mechanism underlying the genetic implication of TBX5 in AF . We show that cardiomyocyte action potential ( AP ) abnormalities in Tbx5-deficient atrial cardiomyocytes are caused by a decreased sarcoplasmic reticulum ( SR ) Ca2+ ATPase ( SERCA2 ) -mediated SR calcium uptake which was balanced by enhanced trans-sarcolemmal calcium fluxes ( calcium current and sodium/calcium exchanger ) , providing mechanisms for triggered activity . The AP defects , cardiomyocyte ectopy , and AF caused by TBX5 deficiency were rescued by phospholamban removal , which normalized SERCA function . These results directly link transcriptional control of SERCA2 activity , depressed SR Ca2+ sequestration , enhanced trans-sarcolemmal calcium fluxes , and AF , establishing a mechanism underlying the genetic basis for a Ca2+-dependent pathway for AF risk .
Atrial fibrillation ( AF ) is the most common arrhythmia in humans , characterized by irregularly irregular atrial electrical activity , resulting in asynchronous atrial contraction . AF is a global problem , affecting more than 33 million people and approximately 25% of Americans over the age of forty ( Nishida and Nattel , 2014; Weng et al . , 2018 ) . AF is associated with significant morbidity and mortality due to thromboembolic events , heart failure , and sudden cardiac death . AF also significantly complicates overall health care management , with AF patients costing five times more to treat than patients without AF ( Andrade et al . , 2014 ) . The total annual cost to treat AF patients in the US is on the order of 26 billion dollars ( Nishida and Nattel , 2014 ) . AF is a highly significant and growing public health concern . A genetic basis for AF risk has been described in the last decade . Large community-based cohort studies indicate that heritability provides between 40% and 62% of AF risk ( Nishida and Nattel , 2014; Christophersen et al . , 2009 ) . An emerging paradigm describes AF as a multifactorial disease with genetic predisposition that will determine the propensity of secondary clinical insults to cause AF . This model highlights the importance of understanding the molecular mechanisms underlying the genetic predisposition to AF . Genome-wide association studies ( GWAS ) studies have identified common risk variants and familial mutations at the T-box transcription factor 5 ( TBX5 ) locus that result in increased risk for AF ( McDermott et al . , 2008; Sinner et al . , 2014 ) . Adult-specific Tbx5 knockout mice demonstrate primary spontaneous and sustained AF , providing evidence supporting the genetic implication at this locus . GWAS have also implicated multiple genes involved in cardiomyocyte calcium handling , including Atp2a2 , encoding the sarcolemmal calcium ATPase SERCA2 , and Sln and Pln , encoding direct binding SERCA2 inhibitors sarcolipin and phospholamban , respectively . We have previously demonstrated that these cardiomyocyte calcium control genes are direct TBX5 targets ( Nadadur et al . , 2016 ) . These observations suggested that tight transcriptional control of SERCA2 activity may be central to atrial rhythm robustness and that variation in SERCA2 expression and activity may contribute to AF risk . The cellular mechanisms causing the irregular electrical activity in AF are believed to include an abnormal myocardial substrate and formation of an ectopic trigger . Abnormal substrate refers to altered electrical conduction between cardiomyocytes . Ectopic trigger refers to cardiomyocyte ectopy , or initiation of electrical activity at regions outside of the sinoatrial node . Both of these cellular phenomena are observed in Tbx5 adult-specific mutant mice and have been associated with abnormal cellular calcium handling ( Dobrev , 2010; Voigt et al . , 2012; Voigt et al . , 2014; Vest et al . , 2005; Shanmugam et al . , 2011; Neef et al . , 2010; Macquaide et al . , 2015; Liang et al . , 2008; Lenaerts et al . , 2009; Hove-Madsen et al . , 2004; Greiser et al . , 2011; El-Armouche et al . , 2006; Brundel et al . , 1999 ) . We described the TBX5-dependent gene regulatory network essential for atrial rhythm control and identified downstream ion channels and transporters potentially important to rhythm control ( Nadadur et al . , 2016; Yang et al . , 2017 ) . Triggered activity in the form of early and delayed afterdepolarizations ( EADs and DADs ) observed in Tbx5-deficient atrial cardiomyocytes could be rescued by heavy buffering of cytoplasmic calcium ( Nadadur et al . , 2016 ) . Tbx5-dependent calcium handling has thereby emerged as a potential mediator of the myocardial physiologic abnormalities resulting in AF . We sought to define the Tbx5-dependent cellular mechanisms responsible for abnormal calcium-dependent electrical activity . We found that Tbx5-dependent AF is associated with abnormal sarcoplasmic reticulum ( SR ) calcium uptake due to depressed SERCA2 expression , depressed SERCA function , and increased phospholamban expression . Decreased SR calcium uptake is compensated by increased Ca2+ extrusion from cardiomyocytes via sodium-calcium exchanger ( NCX ) current ( INCX ) , which provides a mechanism for TBX5-dependent action potential ( AP ) prolongation and the propensity for triggered cellular ectopy . In the setting of enhanced NCX mediated Ca2+ efflux and depressed SR uptake , compensatory increases in L-type calcium current ( ICaL ) balance calcium extrusion to maintain steady state calcium homeostasis . Together these calcium handling alterations contribute to AP prolongation and triggered activity . We further demonstrated that calcium handling abnormalities , AP alterations , and triggered activity are all normalized by knockout of phospholamban , which prevents Tbx5-dependent AF . These results establish a direct link between depressed SR Ca2+ sequestration , enhanced NCX activity , and AF . This model suggests that targeting calcium handling pathways may be a treatment approach for a subpopulation of AF patients .
We previously reported that Tbx5-deficient atrial cardiomyocytes demonstrated AP prolongation and myocardial ectopy . We hypothesized that these defects were caused by cellular calcium handling abnormalities . We therefore surveyed the expression of known calcium handling genes in the adult-specific Tbx5 knockout model . We assessed gene expression in Tbx5fl/fl;R26CreERT2 and control R26CreERT2 mice at 10 weeks of age following tamoxifen ( TM ) treatment at 8 weeks of age . Consistent with previous observations , the adult Tbx5fl/fl;R26CreERT2 but not control mice developed spontaneous AF , showing an irregularly irregular heartbeat , by telemetric electrocardiogram ( ECG ) recordings ( Figure 1A , B ) . As previously shown , APs and [Ca]i transients were prolonged in Tbx5fl/fl;R26CreERT2 ( Figure 1C ) ( Nadadur et al . , 2016 ) . We assessed expression of genes important to cellular calcium handling in the left atrium by quantitative PCR ( Figure 1D ) . mRNA transcripts for RyR2 ( Ryr2 ) and SERCA2 ( Atp2a2 ) , two of the main determinants of sarcoplasmic reticulum ( SR ) calcium flux , were decreased by 61% and 71% respectively in Tbx5fl/fl;R26CreERT2 mice compared to R26CreERT2 controls ( p=0 . 026 and p=0 . 001 for Ryr2 and Atp2a2 respectively ) consistent with previous studies ( Nadadur et al . , 2016 ) . In addition , phospholamban ( Pln ) mRNA expression was increased by 69% in Tbx5fl/fl;R26CreERT2 compared to R26CreERT2 ( p=0 . 023 ) , which would be expected to further depress SERCA2 activity . There was no significant difference in mRNA expression of the alpha 1C subunit of the L-type calcium channel ( Cacna1c ) , the cardiac sodium calcium exchanger ( Ncx1 ) , or any of the calmodulins 1–3 ( Calm1 , Calm2 , Calm3 ) ( Figure 1C ) . These data are consistent with the hypothesis that the myocardial electrophysiology deficits in the Tbx5-deficient AF model may be due to abnormal calcium handling . We examined the relationship between myocardial electrophysiology deficits and calcium flux in Tbx5-mutant atria . In steady state , with each cardiomyocyte contraction cycle , calcium entering the cardiomyocyte ( L-type calcium channel , ICaL ) is extruded from the cell ( predominantly via inward INCX ) . Similarly , calcium leaving the SR via RyR2 release or SR leak pathways is taken back up into the SR via SERCA2 . We examined the effect of altered TBX5-dependent gene expression on these aspects of cardiomyocyte calcium flux . Given the observed changes in Ryr2 and Atp2a2 mRNA abundance , we hypothesized that AP prolongation in Tbx5 deficient cardiomyocytes was due to calcium handling defects downstream of initial Ca2+ entry through ICaL . To test this , we recorded APs in the presence and absence of the L-type Ca2+ channel blocker nifedipine . This approach blocks Ca2+ entry into the cell and indirectly removes the effect of Ca2+ entry on downstream Ca2+ handling pathways , including SR Ca2+ release/reuptake as well as the electrogenic effect of calcium transport out of the cell via inward INCX . 30 μM nifedipine completely inhibited L-type calcium current , preventing Ca2+ entry or release of SR calcium in control R26CreERT2 and Tbx5fl/fl;R26CreERT2 ( Figure 2—figure supplement 1 ) . In control R26CreERT2 atrial cardiomyocytes , the effect of nifedipine on AP duration ( APD ) was small , with 19 ± 4% shortening of APD at 90% repolarization ( APD90 ) ( p=0 . 008 ) ( Figure 2A ) . However , in Tbx5fl/fl;R26CreERT2 atrial cardiomyocytes , nifedipine had a profound effect: APD at 50% repolarization ( APD50 ) was shortened by 16 ± 6% and APD90 by 61 ± 6% ( p=0 . 02 and 0 . 007 respectively ) ( Figure 2B , C ) . Western blot with densitometry analysis for CaV1 . 2 showed no significant difference in protein expression ( Figure 2D ) , in line with the qPCR data ( Figure 1D ) , consistent with no TBX5-driven direct transcriptional regulation of L-type calcium channels . However , peak ICaL current was increased 92 ± 34% ( p=0 . 027 ) in Tbx5fl/fl;R26CreERT2 atrial cardiomyocytes compared to control R26CreERT2 ( Figure 2E and F ) . The inactivation kinetics at peak ICaL were accelerated Tbx5fl/fl;R26CreERT2 compared to control R26CreERT2 ( τ = 26 . 7 ± 3 . 4 ms vs . τ = 40 . 0 ± 3 . 0 ms; p=0 . 05 ) . Steady-state ICaL inactivation was unchanged ( Figure 2—figure supplement 2 ) . These data suggest that increased ICaL may contribute to TBX5-loss associated AP prolongation and EADs . However , nifedipine also blocks SR Ca2+ release as well as downstream Ca2+ extrusion pathways , which also affect AP duration . Further , since late AP repolarization is dramatically prolonged ( negative to −30 mV where ICaL is largely inactive ) we hypothesized that Tbx5-deficiency disrupts Ca2+ handling pathways downstream of ICaL . Because RyR2 is a critically important sarcolemmal calcium extrusion channel and Ryr2 mRNA was downregulated in Tbx5-mutant atria , we investigated the Tbx5 dependent regulation of RyR2 protein expression and function . RyR2 protein expression was significantly decreased in left atria of Tbx5fl/fl;R26CreERT2 mice compared to R26CreERT2 mice by western blot ( Figure 3A ) , consistent with the observed downregulation of Ryr2 mRNA ( Figure 1B ) . We hypothesized that decreased RyR2 contributed to abnormal Ca2+ release from the SR and tested this by measuring local spontaneous RyR2-mediated Ca2+ release events ( Ca2+ sparks ) using confocal linescans ( Figure 3B ) . The frequency of Ca2+ sparks in Tbx5fl/fl;R26CreERT2 atrial cardiomyocytes was decreased in comparison with R26CreERT2 atrial cardiomyocytes at different pacing frequencies from 0 to 2 Hz ( Figure 3C ) . A decrease in calcium sparks can be due to either decreased RyR2 open probability or a reduced SR calcium load . To differentiate these possibilities , we first examined RyR2 function in the setting of reduced RYR2 expression by performing a [3H]-ryanodine binding assay . [3H]-ryanodine binding to RyR2 correlates with RyR2 open probability ( Dobrev , 2010 ) . Despite reduced ryanodine receptor expression , overall ryanodine binding was unchanged over the majority of the physiological range of calcium values , with no shift in calcium sensitivity ( Figure 3D ) . This observation suggests that the alterations in spark frequency were not due to changes in total RyR2 open probability . Instead , it may be caused by diminished SR Ca2+ uptake , a SERCA-dependent property . We next focused on the balance of diastolic calcium efflux pathways as potential mediators of Ca2+ mishandling by measuring SR Ca2+ content and protein expression and function of SERCA2 and NCX1 . We observed that SERCA2 protein expression was decreased while NCX1 protein expression was increased in Tbx5fl/fl;R26CreERT2 in comparison with R26CreERT2 atria ( Figure 4A , B ) . To define steady state SR Ca2+ content , we loaded cardiomyocytes with Fluo-4 AM and paced with a train of field stimuli to achieve a steady state Ca2+ content , peak Ca2+ content , and rate of Ca2+ removal were measured ( Figure 4C , D ) . The [Ca2+]i transient peaks were unchanged , but [Ca2+]i transient decay rates , corresponding to SR Ca2+ uptake and cellular Ca2+ extrusion , were slowed in Tbx5fl/fl;R26CreERT2 compared to R26CreERT2 atrial cardiomyocytes ( Figure 4E , F ) ( Nadadur et al . , 2016 ) , consistent with defective Ca2+ removal from the cytosol . We also measured [Ca2+]i transients in voltage clamp mode using 40 ms square wave voltage clamp pulses from −80 to 0 mV ( Figure 4—figure supplement 1 ) . Similar to the field stimulation experiments , [Ca2+]i transient decay rates were slowed , but [Ca2+]i transient peaks were decreased by 23 ± 4% ( p=0 . 02 ) in Tbx5fl/fl;R26CreERT2 cardiomyocytes compared to R26CreERT2 , which suggests that AP prolongation is essential to maintaining peak twitch [Ca2+]i . The latter experiment is also consistent with depressed SR loads in Tbx5fl/fl;R26CreERT2 compared to R26CreERT2 atrial myocytes . We hypothesized that decreased SERCA2 expression caused decreased SR load . We examined SERCA activity by synchronizing the opening of RyR2 channels while preventing Ca2+ extrusion through NCX using caffeine containing , sodium-free , Tyrode solution . This provides a measurement of the maximum release of Ca2+ into the cytosol from the SR , a measure of the SR Ca2+ load ( Figure 4C , D ) . SR [Ca2+] was reduced by 24 ± 8% ( p=0 . 0005 ) in Tbx5fl/fl;R26CreERT2 compared with R26CreERT2 atrial cardiomyocytes ( Figure 4G ) . SERCA activity was assessed from [Ca2+]i decay rate after SR release in the absence of external sodium ( NCX inactive ) . Peak SERCA activity was reduced by 31 ± 9% ( p=0 . 006 ) in Tbx5fl/fl;R26CreERT2 compared with R26CreERT2 atrial cardiomyocytes ( Figure 4C , H ) . NCX activity was assessed as the rate of change in [Ca2+]i decay in Na+ containing caffeine solution , preventing net SR uptake . Since NCX activity depends on [Ca]i , we plotted NCX as a function of the [Ca]i signal . NCX activity was ~60% higher in Tbx5fl/fl;R26CreERT2 in comparison with R26CreERT2 atrial cardiomyocytes ( Figure 4D , I ) . Thus , removal of Tbx5 causes decreased SR Ca2+ load and decreased SERCA function , but increased NCX mediated Ca2+ extrusion . Increased inward NCX activity promotes cardiomyocyte depolarization , providing a mechanism for prolonged APs and increased ectopy in Tbx5-mutant atrial cardiomyocytes . We hypothesized that Tbx5 deficiency reduces SERCA activity by decreasing SERCA2 protein expression ( Figure 4A ) and increasing expression of phospholamban ( Pln ) , a negative regulator of SERCA2 ( Figure 1D ) . If these were the primary causes of decreased SERCA function in Tbx5-mutant atria , reduced PLN or PLN phosphorylation ( relieving inhibition of SERCA2 ) would be expected to normalize SERCA function . Western blot analysis showed that PLN expression was significantly increased in Tbx5fl/fl;R26CreERT2 compared with R26CreERT2 atria ( Figure 5A ) . In addition , PLN phosphorylation was also increased at serine 16 in Tbx5fl/fl;R26CreERT2 compared to R26CreERT2 . These data suggest PLN phosphorylation may be a compensatory mechanism in response to decreased SERCA expression and activity , but is insufficient to normalize SERCA function ( Figure 4 ) . Thus , we hypothesized that reduction of Pln gene expression would be more effective in restoring SERCA function . We assessed if Pln deficiency can affect SERCA function in a dose dependent manner by crossing the Tbx5fl/fl;R26CreERT2 with germline Pln knockout mice ( Pln-/-;R26CreERT2 ) ( Luo et al . , 1994 ) . We compared SR load and SERCA function in adult-specific Tbx5; Pln double mutant mice versus Tbx5 mutant mice . We measured SR load and SERCA function using caffeine-induced SR release in atrial cardiomyocytes from R26CreERT2 , Tbx5fl/fl;R26CreERT2 , Pln-/-;R26CreERT2 , Tbx5fl/fl;Pln-/+;R26CreERT2 mice , and Tbx5fl/fl;Pln-/-;R26CreERT2 mice . Control Pln deficient mice ( Pln-/-;R26CreERT2 ) had increased steady state SR load and SERCA activity relative to R26CreERT2 ( Figure 5C , D ) . The decreased SR load and SERCA function observed in Tbx5 mutant mice ( Tbx5fl/fl;R26CreERT2 ) was converted to elevated SR load and SERCA function after the removal of Pln ( Tbx5fl/fl;Pln-/-; R26CreERT2 ) ( Figure 5C , D ) . Pln loss alone increased peak twitch calcium . However , in the setting of combined Tbx5;Pln deficiency , peak twitch calcium and tau twitch were normalized to R26CreERT2 values ( Figure 5E , F ) . We next tested the possibility that decreased SERCA function was the mechanism of TBX5-deficiency-driven AP prolongation and triggered activity and that decreased Pln may rescue these defects . As we previously showed , Tbx5fl/fl;R26CreERT2 atrial cardiomyocytes exhibited significantly prolonged APs and frequent EADs and DADs compared to R26CreERT2 atrial cardiomyocytes ( Figure 6A , B ) ( Nadadur et al . , 2016 ) . APs of Pln-/-;R26CreERT2 atrial cardiomyocytes were similar to R26CreERT2 controls ( Figure 6C ) . The prolonged AP duration observed Tbx5fl/fl;R26CreERT2 was rescued in both Tbx5fl/fl;Pln+/-;R26CreERT2 and Tbx5fl/fl;Pln-/-;R26CreERT2 atrial cardiomyocytes ( 43 ± 10% and 38 ± 5% shorter than Tbx5fl/fl;R26CreERT2 respectively; p=0 . 01 , 0 . 0007 ) ( Figure 6D , E ) . Along with normalization of AP duration , we observed significantly fewer EADs and DADs in Tbx5fl/fl;Pln-/-;R26CreERT2 cardiomyocytes ( Figure 6F ) . The data demonstrate the importance of TBX5-driven SERCA activity on cellular electrophysiology and triggered activity in atrial cardiomyocytes and decreased Pln rescues both SERCA function and cardiomyocyte electrophysiological abnormalities in Tbx5-mutant mice . The data above show reducing Pln gene dosage rescues calcium handling defects , AP prolongation and triggered activity observed in Tbx5-mutant atrial cardiomyocytes . We hypothesized that normalizing these cardiomyocyte cellular defects would reduce AF susceptibility in Tbx5 knockout mice ( Figure 7 ) . We performed intracardiac burst pacing . All Tbx5fl/fl; R26CreERT2 mice ( 6/6 ) paced into AF , compared to none of the R26CreERT2 ( 0/5 ) or Pln-/-;R26CreERT2 littermate controls ( 0/7 ) . Consistent with our hypothesis , AF susceptibility was significantly decreased in Tbx5; Pln compound knockouts: only 1/11 of Tbx5fl/fl;Pln-/-;R26CreERT2 paced into AF ( Figure 7F ) . Thus , Tbx5-deficiency induced AF is due to calcium handling abnormalities , specifically decreased SR load and SERCA activity , and that modulation of the SERCA2 inhibitor , Pln , normalized SERCA activity and AF susceptibility caused by Tbx5 loss .
We and others have hypothesized that ectopic triggers of AF can be due to abnormal atrial calcium handling ( Greiser et al . , 2011; Bers and Grandi , 2011 ) . Here we define this relationship in a model of spontaneous AF . We analyzed the major calcium transport pathways in atrial myocytes and demonstrated that the critical calcium handling deficit associated with Tbx5-loss is depressed SERCA-mediated SR calcium uptake . We report significant reduction of SERCA2 protein expression and function , consistent with human paroxysmal or chronic AF ( Voigt et al . , 2012; Voigt et al . , 2014; El-Armouche et al . , 2006; Brundel et al . , 1999 ) . The mechanism causing cardiomyocyte depolarizations from depressed SERCA activity must be indirect , given that SERCA2 is localized to the intracellular SR membrane and therefore does not directly contribute to membrane potential itself . Instead , slowed SR calcium uptake from depressed SERCA activity provides higher cytosolic calcium driving force for calcium extrusion from the cell via electrogenic inward INCX . We demonstrate increased NCX1 protein expression with Tbx5 knockout , a finding also observed in human and other animal models of AF ( Neef et al . , 2010; Lenaerts et al . , 2009; Greiser et al . , 2011; El-Armouche et al . , 2006 ) . Since protein expression , electrochemical driving force , and allosteric calcium regulation can all affect amplitude of inward INCX ( Blaustein and Lederer , 1999 ) , we measured NCX activity following loss of Tbx5 . NCX activity was significantly increased at all levels of calcium ( Figure 4I ) . Thus , increased NCX function coupled with prolonged [Ca2+]i transients , drives increased inward INCX , providing additional depolarizing current during the AP , contributing to its prolongation . While increased NCX function may partially compensate for the depressed SERCA function to bring down calcium levels , it may also promote calcium-induced DADs in the setting of inappropriately timed SR calcium release events . Previous modeling in ventricular cardiomyocytes predicted countervailing functions of SERCA and NCX ( Li et al . , 2011 ) , which we observed in Tbx5 knockout mice . Our data further support that DADs , which are classically thought to relate to SR calcium overload can still occur with depressed SR loads in the appropriate context of depressed SERCA and elevated NCX function ( Voigt et al . , 2012 ) . Modeling suggests that compensatory increases in L-type calcium current in the setting of depressed SERCA function could be required to maintain systolic and diastolic calcium levels ( Li et al . , 2011 ) . In line with the modeling , we observed enhanced peak ICaL with loss of Tbx5 . This may account for early AP prolongation as well as EADs . It is interesting that peak [Ca]i is depressed using controlled square wave voltage clamp pulses ( Figure 4—figure supplement 1 ) , which suggests that 40 ms is insufficient to maintain calcium entry in Tbx5 knockout , even in the setting of enhanced ICaL . However , in the setting of AP prolongation ( Figure 2E–G ) peak twitch calcium levels are maintained . Additionally , the increase in calcium entering the cell through ICaL during the AP would be expected to balance a net increase in NCX mediated calcium extrusion ( Figure 4I ) , a requirement for steady state [Ca2+]i homeostasis . However , our observations that L-type calcium channel expression is TBX5-independent ( Figures 1D and 2D ) and that genetically targeting only the Ca2+ efflux pathways in our model is sufficient to restore normal electrical activity ( Figures 5–7 ) suggest that the ICaL change is not a primary TBX5-dependent effect . Furthermore , although ICaL is increased , it quickly inactivates in TBX5 knockout cardiomyocytes ( Figure 2F ) . Together with our observation that nifedipine normalizes the APD , this supports that enhanced calcium entry impacts APD via secondary [Ca]i dependent mechanisms . In addition to identifying the role of altered NCX and SERCA function , we assessed the importance of TBX5-driven RyR2 expression . RYR2 is a known susceptibility locus for AF and RYR2 mutations are correlated with AF ( Fatkin et al . , 2017; Di Pino et al . , 2014 ) . We observed that RyR2 protein expression was significantly depressed following Tbx5 loss . Defective RyR2 function has also been associated with AF ( Vest et al . , 2005; King et al . , 2013 ) . Despite TBX5-dependent RyR2 expression , the ryanodine binding assay ( Xu et al . , 1998 ) suggested that RyR2 function is generally preserved over the physiologic range of calcium in Tbx5-mutant atria ( Figure 3D ) . This suggests a compensatory mechanism must occur allowing for preserved RyR2 open probability in the setting of depressed protein expression . For example , CaMKII is a potential regulator of RyR2 function which could increase the open probability and thereby increase steady leakage or favor spontaneous local Ca2+ release events from the SR ( Vest et al . , 2005; Neef et al . , 2010; Fischer et al . , 2015 ) in Tbx5-deficient mice . In line with such compensation , we could not detect any differences in the calcium rise kinetics during controlled square wave voltage clamp pulses ( Figure 4—figure supplement 1 ) . Nevertheless , RyR2 compensation in the setting of reduced expression could contribute to abnormal triggered activity in the setting of Tbx5 loss , and is an important topic for further investigation . We found that depressed SERCA function in TBX5 knockout was completely normalized with heterozygous or homozygous phospholamban knockout , which normalized AP duration , decreased frequency of afterdepolarizations , and reduced AF inducibility . This finding demonstrates the importance of SERCA2 to the pathophysiology of AF in the Tbx5-loss model . While phospholamban has been associated with AF by GWAS ( Fatkin et al . , 2017; Federico et al . , 2017 ) , its functional role is less clear . Although PLN is predominantly found in the ventricle ( Bhupathy et al . , 2007 ) , we showed not only its expression is increased in the atria in the context of Tbx5 loss , but also PLN participates in rheostatic control of SERCA activity in the atria , which is sufficient to protect against AF inducibility ( Figures 5–7 ) . Our findings are further supported by patient studies . For example , in patients who experience post-operative AF , SERCA2 is significantly decreased in the atrial tissue ( Zaman et al . , 2016 ) , but those with PLN mutations have decreased AF susceptibility in the context of arrhythmogenic right ventricular cardiomyopathy ( Bourfiss et al . , 2016 ) . Thus , AF is a heterogeneous disease and there can be variability in how the calcium handling proteins are expressed ( Dai et al . , 2016 ) in different disease settings . The genetic background of an individual may be a critical determinant of how calcium handling moieties are disrupted to result in AF . In summary , the most important features of the Tbx5-dependent SERCA2 and PLN regulatory axis are reduced SR uptake and load ( Figure 8 ) . In this setting , enhanced inward INCX and ICaL contributes to AP prolongation , and , more importantly , to cardiomyocyte ectopy . Finally , we demonstrate PLN as a potential means to augment SERCA function , restoring normal atrial myocyte electrical activity and normal sinus rhythm in Tbx5 knockout mice . Thus , the Tbx5 knockout model represents an excellent system to study pharmacologic rescue of SERCA activity , prevention of cardiomyocyte ectopy , and AF . AF has become an increasingly common cause of morbidity and mortality , underlying over one-third of stroke cases and significantly increases the risk for heart failure ( Nishida and Nattel , 2014 ) . Consequently , AF poses a significant socioeconomic burden . AF does not always exist in isolation , but rather in conjunction with other predisposing factors such as obesity , thyroid hormone alterations , or heart failure . Interestingly , disruptions in calcium handling proteins such as the SERCA2-PLN regulatory axis are implicated as predisposing factors . In AF compounded by heart failure , decreased SERCA2 and phosphorylated PLN , and increased NCX1 expression were observed ( Lugenbiel et al . , 2015 ) . Decrease in phosphorylated PLN coupled with an increase in total PLN has been found in animal models of obesity , potentially increasing risk of AF ( Lima-Leopoldo et al . , 2014; Lima-Leopoldo et al . , 2008 ) . These findings suggest a need to evaluate an individual’s genetic background as well as changes in calcium handling proteins when considering predisposing factors for AF . Currently , there are few effective and targeted AF therapies , in part due to an incomplete understanding of the mechanisms underlying AF . Recent studies of specific genetic loci for AF susceptibility have opened new opportunities to identify specific mechanisms at play in subpopulations of AF patients . Understanding these specific mechanisms may facilitate more effective personalized therapies to target specific atrial Ca2+ handling abnormalities . Our data is consistent with the knowledge that pharmacologic regulators of NCX1 or SERCA2 may normalize defects in cellular calcium handling in the atrium ( Dobrev , 2010; Jost et al . , 2013; Nagy et al . , 2014; Ferrandi et al . , 2013; Parikh et al . , 2012 ) . For example , a selective NCX1 inhibitor , ORM-10103 , was shown to prevent cellular Ca2+ handling abnormalities in ischemic ventricular cardiomyocytes , possibly by limiting calcium entry through outward INCX ( Jost et al . , 2013; Kormos et al . , 2014 ) . The benefit of NCX inhibition might also be considered in human cases of AF with increased NCX activity . Further , resveratrol , which increases SERCA2 activity , has been shown to decrease AF , suggesting that targeting SERCA2 activity may be a viable therapeutic approach ( Bai et al . , 2016; Chong et al . , 2015 ) . In addition to providing specific insight into treating TBX5-loss associated AF , our findings may be more broadly applied . These data suggest that pharmacological treatment of AF may be improved by assessing for a defect in the TBX5-SERCA2-PLN axis followed by specifically targeting the defect to restore normal cardiomyocyte electrical activity . We expect this work and continued efforts to uncover mechanisms responsible for AF in subpopulations of patients will play a key role in advancing personalized therapeutics for AF .
The Tbx5fl/fl , Pln-/- and Rosa26CreERT2 lines have all been previously described and were kept in a mixed genetic background ( Luo et al . , 1994; Bruneau et al . , 2001; Ventura et al . , 2007 ) . Double knockout mice were generated by crossing Tbx5fl/fl;R26CreERT2 mice with germline Pln-/- mice . After two generations , we obtained Tbx5fl/fl;R26CreERT2 mice with either loss of one ( Pln+/- ) or both ( Pln-/- ) copies of Pln . All experiments were done using age- and genetic strain-matched littermate controls . Tamoxifen was administered for three consecutive days at a dose of 0 . 167 mg/kg body weight by intraperitoneal injection at 6–10 weeks of age , as previously described ( Nadadur et al . , 2016 ) . All experiments were performed in accordance to The University of Chicago Institutional Animal Care and Use Committee ( IACUC ) approved protocol . 8-to 10- week-old mice were anesthetized using isoflurane , and telemetry transmitters ( ETA-F10 , Data Science International ) were implanted in the back with leads tunneled to the right upper and left lower thorax , as previously described ( Wheeler MT et al . , JCI 2004 ) . Baseline recordings were obtained for 24 hr after a post-implant recovery period of one day . ECG data was analyzed using LabChart 8 ( AD Instruments ) . Detailed protocols for intracardiac electrograms have been previously described ( Nadadur et al . , 2016 ) . Briefly , 8- to 10- week-old mice were anesthetized with isoflurane and a vertical skin cut-down at the right jugular vein was performed . A 1 . 1 F octapolar catheter ( EPR-800 , Millar Instruments ) was advanced in the right jugular vein to perform electrical stimulation . The catheter was connected to ADI BioAmp and PowerLab apparatus and signals were recorded using LabChart Software ( ADInstruments ) . Atrial induction pacing was performed using burst pacing and the presence of at least three cycles of atrial tachycardia or fibrillation at least twice was considered positive . Langendorff perfusion with 2 mg/mL of Collagenase Type 2 ( Worthington Biochemical ) at 5 ml/min was used to isolate atrial cardiomyocytes . Cardiomyocytes were then plated on laminin coated glass bottom dishes for 30 min prior to incubation with 10 µM Fluo-4/AM ( Molecular Probes/Invitrogen ) in normal Tyrode’s solution containing ( in mM ) : 140 NaCl , 4 KCl , 10 glucose , 10 HEPES , and 1 MgCl2 , 1 CaCl2 pH 7 . 4 using NaOH for 20 min at room temperature . Cells were perfused with prewarmed Tyrode for 10 min prior to imaging . Imaging was performed on an Olympus microscope with a 20x objective lens , a LAMBDA DG-4 power source with 488 nm excitation and 515 nm emission filters and a PMT ( photomultiplier tube ) to record whole cell signal . Electrical field stimulation ( Grass stimulator; Astro-Med ) was performed at 1 Hz . SERCA and NCX measurements were performed by flowing sodium free Tyrode with 10 mM caffeine followed by sodium free Tyrode alone or Tyrode with caffeine respectively . Cells were returned to normal Tyrode in both cases at the end of the recording . [Ca2+]i transients are presented as total fluorescence intensity normalized to resting fluorescence ( F/F0 ) obtained from steady-state resting conditions before field stimulation . [Ca2+]i transients and sparks were acquired in line-scan mode ( 3 ms per scan; pixel size 0 . 12 µm ) using a Zeiss confocal microscope . APs and voltage clamp recordings were recorded using standard ruptured patch protocol ( Nadadur et al . , 2016 ) . We used current clamp mode with 0 . 5 nA ×2 ms current clamp pulses to measure APs . Voltage clamp mode was used to measure capacitance transients and to study [Ca2+]i transients with fixed duration depolarizations . Cardiomyocytes are kept at 37°C and perfused with Tyrode solution ( 140 NaCl , 4 KCl , 1 MgCl2 , 1 CaCl2 , 10 HEPES , 10 Glucose , and pH 7 . 4 with NaOH ) . Internal pipette solution composition was ( in mM ) : 20 KCl , 100 K-glutamate , 10 HEPES , 5 MgCl2 , 10 NaCl , 5 Mg-ATP , 0 . 3 Na-GTP . Patch pipettes ( World Precision Instruments ) were pulled to have a mean resistance of 3 . 5–5 MΩ . An Ag–AgCl pellet and 3M KCl agar bridge was used to ground the bath . Liquid junction potentials , were always corrected after cell rupture . External solution for ICaL contained ( in mM ) : 120 Tetraethylammonium-chloride , 10 CsCl , 10 Glucose , 10 HEPES , 1 . 5 MgCl2 , 1 CaCl2 , pH 7 . 4 with CsOH . Internal pipette solution contained ( in mM ) : 100 Cs-methanesulfonate , 30 CsCl , 10 HEPES 5 EGTA , 2 MgCl2 , 5 Mg-ATP , pH 7 . 2 with CsOH . ICaL was recorded during 200 ms voltage clamp pulses from a holding potential of −40 mV to test potentials ranging from −40 to 60 mV , with pulses applied every 2 s in 5 mV increments . Peak current amplitudes were normalized to the cell capacitance ( Cm ) and presented as current density ( A/F ) . Steady-state inactivation of ICaL was investigated using two-pulse protocol . Holding potential was −80 mV . The first pulse depolarized membrane from −60 to 20 mV with 10 mV increments during 500 ms , the second pulse depolarized the membrane to 10 mV for 50 ms . The inactivation curves were fit to a Boltzmann distribution . Acquisition was performed using an Axopatch-200B amplifier connected to a Digidata1550A acquisition system ( Axon Instruments , Foster City , CA , USA ) . In recording filtering at 2 kHz was performed using the amplifier Bessel and sampled at 10 kHz . Analysis was performed using pCLAMP10 ( Axon Instruments ) and a home written analysis code . Atrial tissue was collected and homogenized as described previously ( Alvarado et al . , 2017 ) , in a buffer containing 0 . 9% NaCl , 10 mM Tris-HCl pH 6 . 8 , 20 mM NaF and protease inhibitors . Equal amounts of protein , as determined by Bradford assay , were loaded . 50 µg of tissue homogenate , in Laemmli buffer , was separated by SDS-PAGE in 4–20% TGX or AnyKD precast gels ( Bio-Rad ) . Proteins were transferred to PVDF membrane using the iblot2 transfer system ( ThermoFisher ) or wet transfer . Primary antibodies were as follows: anti-RyR2 ( 1:2000; MA3-925 , ThermoFisher ) , SERCA2 ( 1:1000; MA3-919 , ThermoFisher ) , NCX ( 1:1000; MA3-926 , ThermoFisher ) , PLN ( 1:5000; A010-14 , Badrilla ) , pT17-PLN ( 1:5000; A010-13 , Badrilla ) , pS16-PLN ( 1:5000; A010-12 , Badrilla ) , Cav1 . 2 ( 1:200; ACC-003 , Alomone ) , GAPDH ( 1:10000; MAB374 , Millipore ) . Secondary antibodies were: goat anti-mouse-HRP ( 1:5000; 31437 , ThermoFisher ) or goat anti-rabbit-HRP ( 1:5000; 31463 , ThermoFisher ) . Secondary antibody concentrations were 5x higher when using the ibind Flex system . SuperSignal ECL reagent ( ThermoFisher ) was used to develop membranes followed by imaging with a ChemiDoc MP apparatus ( Bio-Rad ) . Band intensities were quantified with the ImageLab software ( Bio-Rad ) or using ImageJ ( NIH ) . Binding assays were carried out following a protocol previously described ( Federico et al . , 2017 ) . Binding mixtures contained 100 µg of protein from homogenates prepared from pooled atria ( 5–7 mice ) , 0 . 2 M KCl , 20 mM Hepes ( pH 7 . 4 ) , 6 . 5 nM [3H]ryanodine ( PerkinElmer ) , 1 mM EGTA and enough CaCl2 to set free [Ca2+] between 10 nM ( pCa2+ 8 ) and 100 µM ( pCa2+ 4 ) . The ratio between Ca2+ and EGTA was determined using MaxChelator ( WEBMAXCLITE v1 . 15 http://maxchelator . stanford . edu/webmaxc/webmaxclite115 . htm ) . Following a 2 hr incubation at 36°C , reactions were filtered through Whatman GF/B Filters using a Brandel M24-R Harvester . [3H]ryanodine binding was determined using a Beckman LS6500 scintillation counter and BioSafe II scintillation cocktail ( RPI Corp ) . Non-specific binding was quantified in the presence of 2 µM unlabeled ryanodine ( MP Biomedicals ) and subtracted . Left atrial tissue of Tbx5fl/fl;R26CreERT2 and R26CreERT2 mice was removed two weeks after receiving tamoxifen and RNA was isolated using a Trizol ( Invitrogen ) based method . Reverse transcription reaction was carried out using the qScript cDNA synthesis kit ( Quanta ) according to the manufacturer’s protocol . Quantitative RT-PCR was performed using the Power SYBR Green PCR Master Mix ( Applied Biosystems ) and run on an Applied Biosystems AB7500 machine . Relative fold changes were calculated using the comparative threshold cycle method ( 2-ΔΔCt ) , using glyceraldehyde-3-phosphate dehydrogenase ( Gapdh ) gene expression level as internal control . Values are represented as mean ±standard error of the mean ( ±SEM ) . Statistical significance for quantitative metrics of APs , SERCA , NCX , SR load , ICaL , spark frequency , and [Ca2+]i transients were determined using hierarchical statistical methods ( Sikkel et al . , 2017 ) . Statistical significance for mRNA , and protein expression studies was determined using Student’s t-test . Statistical significance of the nifedipine effect on AP duration was determined using two-tailed paired t-test . A two-tailed Fisher’s exact test was used for statistical significance of count-based analysis of AF inducibility and EAD and DAD count . Statistical significance is designated as *p<0 . 05 , **p<0 . 01 , and ***p<0 . 001 . | The human heart contains four distinct chambers that work together to pump blood around the body . In individuals with a condition called atrial fibrillation , two of the chambers ( known as the atria ) beat irregularly and are unable to push all the blood they hold into the other two chambers of the heart . This can cause heart failure and increases the likelihood of blood clots , which may lead to stroke and heart attacks . Small molecules called calcium ions play a crucial role in regulating how and when the atria contract by driving electrical activity in heart cells . To contract the atria , a storage compartment within heart cells known as the sarcoplasmic reticulum releases calcium ions into the main compartment of the cells . Calcium ions also enter the cell from the surrounding tissue . As the atria relax , calcium ions are pumped back into the sarcoplasmic reticulum or out of the cell by specific transport proteins . Individuals with mutations in a gene called Tbx5 are more likely to develop atrial fibrillation than other people , but it was not clear how such gene mutations contribute to the disease . Here , Dai , Laforest et al . used mice with a mutation in the Tbx5 gene to study how defects in Tbx5 affect electrical activity in heart cells . The experiments found that the Tbx5 gene was critical for calcium ions to drive normal electrical activity in mouse heart cells . Compared with heart cells from normal mice , the heart cells from the mutant mice had decreased flow of calcium ions into the sarcoplasmic reticulum and increased flow of calcium ions out of the cell . These findings provide a direct link between atrial fibrillation and the flow of calcium ions in heart cells . Together with previous work , these findings indicate that multiple different mechanisms could lead to atrial fibrillation , but that many of these involve changes in the flow of calcium ions . Therefore , personalized medicine , where clinicians uncover the specific mechanisms responsible for atrial fibrillation in individual patients , may play an important role in treating this condition in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2019 | A calcium transport mechanism for atrial fibrillation in Tbx5-mutant mice |
Cells are fundamental units of life , but little is known about evolution of cell states . Induced pluripotent stem cells ( iPSCs ) are once differentiated cells that have been re-programmed to an embryonic stem cell-like state , providing a powerful platform for biology and medicine . However , they have been limited to a few mammalian species . Here we found that a set of four mammalian transcription factor genes used to generate iPSCs in mouse and humans can induce a partially reprogrammed pluripotent stem cell ( PRPSCs ) state in vertebrate and invertebrate model organisms , in mammals , birds , fish , and fly , which span 550 million years from a common ancestor . These findings are one of the first to show cross-lineage stem cell-like induction , and to generate pluripotent-like cells for several of these species with in vivo chimeras . We suggest that the stem-cell state may be highly conserved across a wide phylogenetic range .
Stem cells are in an early undifferentiated state and have the potential to differentiate into a variety of cell types and tissues , both in vitro and in vivo , including in developing embryos and grafted adult tissues ( Badylak et al . , 2012 ) . Accordingly , stem cells provide a powerful platform to study development ( Arendt , 2008 ) , tissue regeneration ( Langer and Vacanti , 1999; Rosselló et al . , 2009 ) , disease mechanisms ( Colman and Dreesen , 2009 ) , and gene therapeutic approaches to the brain and other organs ( Hwang et al . , 2011 ) . Embryonic stem cells ( ESCs ) have the potential to be differentiated to most if not all cell types ( pluripotent ) , whereas more differentiated stem cells , such as those in the skin , have a more restricted differentiation potential ( multipotent or unipotent ) ( Collas et al . , 2007 ) . Induced pluripotent stem cells ( iPSCs ) are once mature cells that have been de-differentiated to become like the embryonic state ( Thomson et al . , 1998; Takahashi and Yamanaka , 2006; Takahashi et al . , 2007; Yu et al . , 2007; Maherali and Hochedlinger , 2008; Stadtfeld et al . , 2008 ) . One major advantage of iPSCs is that they can be made from differentiated cells , such as skin or fibroblasts , of an individual and do not require isolating cells from 2–6 day old embryos , which is controversial for human studies ( Lo and Parham , 2009 ) . The finding that simple over-expression of four genes ( Oct4 , Sox2 , Klf4 and c-myc ) was sufficient to generate iPSCs from adult cells of mice ( Takahashi and Yamanaka , 2006 ) and humans ( Takahashi et al . , 2007; Yu et al . , 2007; Sommer et al . , 2009 ) made the process of generating and studying stem cells much more tractable in certain other mammalian species , where it was once difficult to generate stem cells , such as in rats ( Li and Ding , 2010 ) and pigs ( Wu et al . , 2009 ) . However , important issues in biology are addressed in experimental systems other than mammals , specifically in birds ( Jarvis , 2004 , 2007; Jarvis et al . , 2005 ) , fish ( Fetcho et al . , 2008 ) , and flies ( Kuo et al . , 2006; Yu et al . , 2006 ) . Some of these animals have traits similar to humans that are not found in closely related non-human primates or commonly used laboratory animals . These include vocal learning in parrots and songbirds ( Jarvis , 2004 ) , widespread adult neurogenesis in non-mammalian vertebrates ( Nottebohm , 2002; Kaslin et al . , 2008 ) , and vascularization and organ regeneration in zebrafish ( Poss et al . , 2002; Stoletov and Klemke , 2008; Yaqoob and Schwerte , 2010 ) . Another important reason is that some traits are more easily studied in simpler organisms before they are applied to humans . The arthropod Drosophila melanogaster is an attractive genetic model due to the short life span , large number of offspring , and applicability of many genetic techniques ( van Ham et al . , 2009 ) . Drosophila have been used to model Parkinson’s , Huntington’s , and Prion disease . Unfortunately , production of non-mammalian stem cells has been limited , due to problematic or unknown isolation procedures , and insufficient maintenance methods ( Lavial and Pain , 2010 ) . For these reasons , there has been a desire to generate stem cells for these species , allowing disease and mechanistic models to be explored , and possibly transgenic animals to be generated . Induced stem cells could provide such a model . Here we attempted to generate an iPSC state for non-mammalian vertebrate and invertebrate model species spanning over 550 million years from a common ancestor ( Figure 1A ) ( Sullivan et al . , 2006 ) : in birds ( galliformes and songbirds ) , fish ( zebrafish ) , and insect ( Drosophila ) . We found that the four transcription factor genes used to induce mammalian stem cells can produce a partial iPSC state that varies with degree of relationship to mammals . Moreover , the mammalian ( mouse ) homolog of these genes induced this partial iPSC reprogrammed state in the non-mammalian cells of all species tested , including inducing the ability of the vertebrate cells to incorporate into embryonic chimeras . We use the term partial iPSC or iPSC-like cells to denote cells that are transformed and show some iPSC characteristics . These findings are the first that we are aware of to generate iPSC-like cells across multiple non-mammalian species , using mammalian genes , in animal models where stem cells have been difficult or impossible to isolate ( Zwaka , 2008; Lavial and Pain , 2010 ) . 10 . 7554/eLife . 00036 . 003Figure 1 . Phylogeny of species used and stem cell gene homologies . ( A ) Phylogenetic relationships of the species studied relative to mouse: birds ( galliforms and songbirds ) , fish ( zebrafish ) , and an insect ( Drosophila ) . The phylogenetic tree is based on ( Sullivan et al . , 2006 ) . ( B ) General structure and sequence comparisons of the putative homologs of the four stem cell inducing transcription factors included in the cassette ( Figure 1—figure supplement 1; Oct-4 , Sox-2 , C-myc , Klf-4 ) across species . Although overall homologies vary significantly , DNA binding sites are highly conserved . Gene sequences were either from published studies ( Lavial et al . , 2007; Camp et al . , 2009 ) or from those predicted in sequence databases ( Ensembl ) . Conserved domains ( color coated boxed regions with accession numbers ) were found using the Ensembl orthologue function and NCBI’s HomoloGene . Detailed sequence homologies can be seen in Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 00310 . 7554/eLife . 00036 . 004Figure 1—figure supplement 1 . Schematic representation of the pHAGE-STEMCCA vector map . The critical induction genes depicted are transcribed on a single multicistronic mRNA transcribed under the control of the human EF1 ? promoter . The mRNA contains an IRES element separating two fusion cistrons ( Oct4 and Klf4; and Sox2 and cMyc ) . The LoxP site can be used to excise the four genes after the cells have been induced to become stem cells , and thus prevent c-myc from causing aberrant growth in transgenics or re-activation of the exogenous stem cell genes . For a full map of the vector see ( Sommer et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 00410 . 7554/eLife . 00036 . 005Figure 1—figure supplement 2 . Alignments of the coding sequence of the putative Oct4 , Sox2 , Klf4 , and c-myc homologs across vertebrate species . The DNA binding domain are highlighted by a red box . Alignments were generated using T-coffee ( www . tcoffee . org ) . Color-coding indicates degree of amino acid similarity ( red , very similar or identical; blue , completely different ) . Accession numbers for the specific sequences used are in Supplementary file 1C ( same sequences used to generate primers ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 00510 . 7554/eLife . 00036 . 006Figure 1—figure supplement 3 . Alignments of the coding sequence of the putative Oct4 , Sox2 , Klf4 , and c-myc homologs across invertebrate species . Accession numbers for the specific sequences used are in Supplementary file 1C ( same sequences used to generate primers ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 006
In an ongoing effort to generate stem cells for transgenic songbirds with targeted gene manipulations , as opposed to non-targeted ( Agate et al . , 2009 ) as a means to study the molecular basis of vocal learning ( Jarvis , 2004 ) , we attempted the iPSC approach . We decided to transduce embryonic fibroblast cells of zebra finch and galliforms ( quail and chicken ) with a lentivirus retroviral vector ( called STEMCCA [Sommer et al . , 2009] ) containing the four genes from the mouse driven by the human EF1α promoter ( Map in Figure 1—figure supplement 1 ) . We surmised that the mouse genes might work in birds despite the separation of ∼300 million years ago ( MYA ) from a common ancestor with mammals ( Figure 1A ) , because although there were stretches of low homology and divergent sequences in three of the four genes ( Oct4 , Klf4 , and c-myc ) , the overall conservation between birds and mammals was good ( 80–98% overall amino acid identity; Figure 1B–E , Figure 1—figure supplement 2 ) . Furthermore , all four genes had highly conserved DNA binding domains ( Figure 1—figure supplement 2; red boxes ) . For an iPSC positive control , we isolated mouse embryonic fibroblasts and transfected them with the same lentiviral cassette ( Supplementary file 1A ) . For non-iPSC positive controls , we used established ESC lines of mouse ( Nagy et al . , 1993 ) and chicken ( Pain et al . , 1996 ) . For two negative controls , we transduced fibroblasts of each species with the same lentivirus vector , but containing GFP in place of the four mouse transcription factors , and grew the cells either in our stem cell media or complete media ( see media composition in Supplementary file 1B ) . For a third negative control , we cultured non-transfected fibroblasts in stem cell media for each species to make sure media alone could not induce the cells ( Supplementary file 1A , B ) . The two negative control groups grown in stem cell media exhibited similar qualitative and quantitative characteristics , and therefore , to diminish redundancy , the data shown is from the GFP-transduced fibroblasts . We repeated our experiments at least seven independent times , with 12–18 wells per species in 48 well plates ( ‘Materials and methods’ ) , and used established guidelines to evaluate iPSCs ( Maherali and Hochedlinger , 2008; Kim and Daley , 2009 ) . The transformed avian cells showed a number of stem cell features absent from control fibroblasts and present in our mouse ESC and iPSC controls , and chicken ESC controls . This included , within 5 days , transformation from fibroblast morphology ( Figure 2A ) to colonies with characteristic clustered stem cell-like morphology ( Figure 2B ) . These colonies had strong alkaline phosphatase ( ALP ) enzyme activity ( Figure 2D ) , a characteristic of early and mature stem and tumor cells ( O’Connor et al . , 2008 ) , whereas the starting fibroblasts did not ( Figure 2C ) . They expressed Stem Cell Specific Antigen-1 ( SSEA-1; Figure 2F ) , while none was detected in control fibroblasts ( Figure 2E ) . An average of 20% of the wells had iPSC-like cells , as measured by colony morphology and ALP activity ( measured from seven independent experiments for each avian species ) . Later iterations with different media conditions produced transformed cells in up to 90% of the wells ( Dai et al . , unpublished date ) . The higher the viral titer used , the more colonies were produced ( Figure 2—figure supplement 1 ) ; the highest titer , 109 U/ml , was used in the above experiments . We noticed some differences between the mouse and avian colonies , in that the mouse colonies as well as the individual cells within the colonies appeared on average slightly larger , while avian cells appeared more clustered . Similar differences have been observed when comparing human and mouse colonies ( Nichols and Smith , 2009 ) . The mouse and avian iPSC-like colonies were similar to those in established lines of mouse and chicken ESCs that we treated under the same growth conditions , including differences between the species ( Figure 2G–H ) . These features were absent from our control mouse and avian fibroblasts treated under the same conditions with and without the lentiviral GFP-vector lacking the four transcription factors ( Figure 2A , C , E; and not shown ) . 10 . 7554/eLife . 00036 . 007Figure 2 . Generation of iPSC-like cells from differentiated cells of mouse , birds , fish , and Drosophila using the mouse transcription factors . ( A ) Non-transduced mouse , avian and zebrafish embryonic fibroblasts , and Drosophila S2 cell line . ( B ) Transformed cells ( colonies ) after 20 days ( first passage ) , using optimal titers ( Figure 2—figure supplement 1 ) . ( C ) Non-transduced cells labeled for ALP activity . ( D ) Colonies formed by transformed cells labeled for ALP activity after the first passages ( 10th passage staining can be seen in Figure 2—figure supplement 2 ) . ( E ) Non-transduced cells and F , transduced cells after colony formation reacted with a Stage Specific Embryonic Antigen-1 ( SSEA-1; green fluorescence ) antibody . ( G ) Colonies of embryonic stem cells ( positive controls ) . ( H ) Embryonic stem cells labeled for ALP activity ( positive controls ) . Black scale bars , 100 μm; green and red bars , 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 00710 . 7554/eLife . 00036 . 008Figure 2—figure supplement 1 . Colony formation in vertebrate cells as a function of species and titer . After transduction with different viral titers , iPSC-like colonies were counted in 35-mm plates . Higher titers produced more colonies , although the highest titer did result in greater variability . The mouse cells gave the highest number of colonies . This could be due to the efficacy of transducing mouse cells with mouse genes or a species difference . Higher titers were used for transductions presented in this paper , as they provided the higher number of colonies . Error bars , S . E . M ( n = 11 independently transduced plates for each species and titer ) . Statistics in Supplementary file 1D . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 00810 . 7554/eLife . 00036 . 009Figure 2—figure supplement 2 . Alkaline phosphetase staining ( red color labling ) in chicken iPSC-like colonies after the 10th passage , and growth of fibroblast feeder layer cells that are not labeled . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 009 Like our mouse control iPSCs , the transformed avian cells ( chicken , quail , and finch ) expressed the four exogenous mammalian genes ( Figure 3A–D; as determined by quantitative RT-PCR with mouse specific probes; Supplementary file 1C ) . After the first and second passages ( 3–4 weeks ) , three of the endogenous avian homologs ( Oct4 , Sox2 , c-myc ) were significantly upregulated 10–100-fold in the presence of their mammalian counterparts ( except c-myc in quail; Figure 3A–D; green ) . The levels of induction of the endogenous and exogenous expression of these three genes in our chicken and mouse cells were similar to the control chicken and mouse ES cell . The level of induction in quail and zebra finch was lower ( 4–40-fold ) , but still statistically significant ( p<0 . 0001 , ANOVA ) with no overlap in the expression detected in five replication experiments relative to the embryonic fibroblast controls . The fourth gene , Klf4 , was upregulated in our mouse control iPSC and ESC , but not upregulated in any of the avian species ( Figure 3A–D ) . However , Klf4 was also not upregulated in the established control chicken ESC line ( Figure 3C–D ) , relative to the chicken embryonic fibroblast . All avian species also showed significant induced expression of two other endogenous stem cell markers , nanog and vasa , not present in the STEMMCA vector , with levels more similar among species but lower than the mouse ( Figure 3E–G ) . After about the fifth passage ( 2–3 months ) , the exogenous mouse genes were either completely ( mouse and chicken ) or partially ( quail and finch ) silenced , and this was associated with a concomitant further increase in some of the endogenous species-specific homologs ( Figure 3G–J; including c-myc in quail as well as vasa and nanog , Figure 3K–L ) . However , Klf4 was still very low relative to the starting fibroblast controls in the avian cells , except for a small increase in some of the finch cell lines ( Figure 3J ) . 10 . 7554/eLife . 00036 . 010Figure 3 . Upregulation of stem cell genes in mouse , birds , fish , and Drosophila by mouse transcription factors . ( A–D ) qRT-PCR of exogenous ( black ) mouse and endogenous ( green ) species-specific expression of Oct-4 ( A ) , Sox-2 ( B ) , c-myc ( C ) , and Klf-4 ( D ) in iPSC-like cells of each species after the second passage relative to ( normalized ) non-transduced fibroblast controls ( blue ) . Mouse and chicken ESCs were included as positive controls ( red ) . Primers used are shown in Supplementary file 1C . Several values overlap among cell types ( e . g . , mouse exogenous and endogenous Oct-4 and Klf-4 ) and are thus not distinguishable in the graph . ( E–F ) qRT-PCR of Nanog ( E ) and Vasa ( F ) homologs in the different cell types across species . ( G–L ) qRT-PCR after the fifth passage show that the exogenous mouse genes are significantly downregulated or silenced . These values were normalized to the same fibroblast values as in the second passage . Nanog and Vasa expression levels exhibit no significant difference from passage two levels , except in chicken cells . Expression levels were also measured for 12th passage iPSC-like cells ( Figure 3—figure supplement 1 ) and fifth passage iPSC-like cells were normalized against adult tissue ( Figure 3—figure supplement 2 ) . ( M ) qRT-PCR of exogenous and endogenous ( homologs ) Drosophila specific genes in the transformed S2 cells , and N , other genes known to be involved in early embryogenesis in Drosophila . Expression levels were also measured with iPSC-like cells generated from a primary drosophila cell line ( BG2; Figure 3—figure supplement 3 ) . Error bars , S . E . M within cell populations . p-values for all comparisons are shown in Supplementary file 1D , ANOVA , ( Tukey’s post hoc , p<0 . 001; n = 5 replicates of independent transformed lines ) . ( O ) Time course of self-renewal and proliferation of stem cells ( iPSC-like cells and ESCs ) relative to control fibroblast ( or S2 ) as measured by the MTT [ ( 3- ( 4 , 5-Dimethylthiazol-2-yl ) -2 , 5-diphenilytetrazolium bromide] assay ( read at 570 nm ) ( error bars not shown for clarity ) . ESCs and iPSC-like cells maintain high proliferation levels , while primary fibroblasts decay . ( P ) Telomerase activity was greatly increased ( lower mean Cycle Threshold , CT ) in iPSC-like cells and control ESCs over control fibroblast cells . Error bars , S . E . M ( n = 5 independent cell line replicates for both MTT and telomerase data ) . Statistics shown in Supplementary file 1D . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01010 . 7554/eLife . 00036 . 011Figure 3—figure supplement 1 . Comparison of iPSC-like expression patterns after the 5th passage and 12th passage . Cmyc exhibited a slight down regulation by the 12th passage relative to the fifth , while Oct-4 was slightly upregulated . Neither Klf-4 nor Sox2 exhibited significant changes . *p<0 . 001 , ANOVA , followed by Tukey's post hoc; n = 5 replicates of independent transformed lines . Error bars , S . E . M within cell populations . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01110 . 7554/eLife . 00036 . 012Figure 3—figure supplement 2 . Gene expression profiles under different normalization basis . Fifth passage iPSC-like cells for mice and aves were normalized relative to embryonic fibroblasts and adult tissue , and compared . Adult tissue RNA was purchased for mice and chicken ( Zyagen Cat MR-201 and CR-201 respectively ) , while for finch and quail they were isolated from brains of animals in the lab using a total RNA isolation kit . The comparison shows several significant , but small differences . Expression of Oct-4 was significantly higher when compared to adult tissue in all species . The finch showed significant differential expression in all genes , except Klf4 . *p<0 . 001 , ANOVA , followed by Tukey's post hoc; n = 5 replicates of independent transformed lines . Error bars , S . E . M within cell populations . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01210 . 7554/eLife . 00036 . 013Figure 3—figure supplement 3 . Drosophila BG2 cells also exhibited some transformation . ( A ) BG2 before and 7 days after transfection with the STEMCCA gene casset . Note the clustered colony morphology in the later . ( B ) RT-PCR analyses of mouse ( black ) and homologous drosophila homolog ( green ) of the four inducing transcription factors , relative to starting levels in non-transformed cells ( blue ) . ( C ) RT-PCR analyses of other Drosophila stem cell genes . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 013 Using modified media conditions containing differentiation inhibitors ( Dai et al . , unpublished date ) , we have been able to passage the iPSC-like chicken cells at the same rate as the mouse iPSC ( currently > 20 passages ) and these avian colonies still stain with ALP ( Figure 2—figure supplement 2 for the tenth passage ) and the endogenous avian versions of the re-programming genes , with only minor differences compared to the fifth passage ( Figure 3—figure supplement 1 for the 12th passage ) . When comparing expression of these genes in the iPSC cells with adult avian cells as opposed to the control embryonic fibroblasts , the relative levels of some factors ( such as Oct-4 ) were still significantly increased above the adult levels ( Figure 3—Figure Supplement 2 ) . All of these findings were consistent for each avian species , given the low variation ( S . E . M . ) across independent replicates ( Figure 3A–L , Supplementary file 1D ) . Based on this success , we mimicked transduction conditions for another non-mammalian vertebrate , the zebrafish ( ∼400 MYA removed from mammals; Figure 1 ) , by transducing an embryonic clonal fibroblast line ( ATCC , CRL-2147 ) with the STEMCCA lentivirus in fish-specific complete media supplemented with our stem cell media reagents ( Supplementary file 1B ) . Although the homologies between mouse and fish for two ( Oct-4 and Klf4 ) of the four genes are less than they are with birds ( Figure 1B and Figure 1—Figure Supplement 2 ) , our rational to pursue this route was strengthened by a study that found that the downstream target genes of Oct4 are relatively conserved between zebrafish and mouse , and the mouse Oct4 can rescue zebrafish mutants ( Onichtchouk et al . , 2010 ) . We found similar results for transformed zebrafish cells as for bird putative iPSC and ESC . This included cell colony formation ( Figure 2B ) , ALP activity ( Figure 2D ) , and expression of SSEA1 protein ( Figure 2E ) , initial high expression ( Figure 3A–D ) and then silencing of the exogenous mouse genes by the fifth passage ( Figure 3G–J ) , and absence of induction of endogenous Klf4 ( Figure 3D–J ) . There was also induction of the endogenous stem cell marker Vasa ( Figure 3E–F; K–L ) . The only significant difference between the zebrafish and birds was lack of Nanog induction in the fish cells ( Figure 3E ) . The average zebrafish colony size was also smaller ( Figure 2B ) . Our results with vertebrate cells prompted us to consider whether these same mammalian genes can induce iPSC-like features in a yet more distant relative , in Drosophila , an invertebrate ( 550 MYA removed; Figure 1A ) ( Sullivan et al . , 2006 ) . Although there are even greater divergences between mouse and Drosophila genes , we could still find putative homologs ( 62–64% identity ) with highly conserved DNA binding domains ( Figure 1—figure supplement 3 ) . Thus , we transduced the commonly used Drosophila S2 line with the STEMCCA lentivirus or transfected with a plasmid containing the four factors and a Metallothionein inducible promoter . We decided to try both vector approaches , because , to our knowledge , there had been no successful attempts in transduction of genes into fly cells using lentivirus . Surprisingly , the lentivirus and its recombinant promoters worked in the drosophila cells , generating GFP labeled cells ( Figure 6—figure supplement 1C ) . We found that the transformed Drosophila S2 cells with the STEMCCA lentivirus or plasmid containing the four factors showed colony formation , although the colonies were notably fewer , smaller in size , and even darker than the vertebrate colonies ( Figure 2B ) . The Drosophila colonies , like those of vertebrates , showed ALP activity ( Figure 2D ) . They also expressed the exogenous mouse genes ( Figure 3M ) and , similar to the avian and fish transformed cells , the Drosophila transformed cells had induced expression of two of four endogenous homologues to the mammalian cassette , VVL ( Oct4 homolog ) and dMyc ( c-myc ) , low induction of SoxN ( Sox2 ) , and no induction of Luna ( putative Klf4; Figure 3M ) . There was also a significant upregulation of four of six other known endogenous Drosophila adult stem cell markers , Dichaete , Escargot , Snail , and Vasa ( Figure 3N ) ( Wilson and Dearden , 2008; Palasz and Kaminski , 2009 ) . Because the starting S2 cells are polyploid and are known to be highly proliferative to begin with ( Moutinho-Pereira et al . , 2010 ) , we wondered if these properties could have contributed to the induction process . Thus we tried another Drosophilia cell line , BG2 , which is derived from the central nervous system and which is less proliferative ( Ui-Tei et al . , 2000 ) . After 7 days and passaging , the BG2 cells also exhibited transformation phenotypes . Of nine independent transfection replicates in 48 well plates , three of them were successful , exhibiting morphological changes and expression of endogenous mouse genes ( Figure 3—figure supplement 3 ) , including in the majority ( 90% ) of the wells . Thus the differences between experiments had more to do with other conditions than cell type differences between BG2 and S2 cells . However , there were differences between the two cell types ( S2 and BG2 ) in the expression of induced genes . Like the S2 transformed cells , the BG2 transformed cells showed overexpression of SoxN , Escargot , Snail and Vasa . Unlike the S2 cells , the BG2 cells did not show significant overexpression of Diachaete , asense and VVL ( Figure 3—figure supplement 3 ) . This suggests that the starting state of the cells could make a difference , as seen with mammalian cells ( Kim et al . , 2009 ) . Proliferation levels of the mouse , avian , and fish transformed cells ( measured by an MTT metabolic assay ) were above their respective fibroblast controls after the first to third passage , depending on species ( Figure 3O , green vs blue ) . The MTT levels at this time approached that of the mouse and chicken ESC lines ( in red ) . Vertebrate cells with the GFP-vector alone treated under identical conditions instead showed a continuous decrease in MTT levels ( in blue ) , which was clearly associated with senescence . The Drosophila result was somewhat different , since the control S2 cells are already highly proliferative . However , transformed S2 cells exhibited enhanced proliferation levels at all passages ( Figure 3O; we did not assess the BG2 cells with MTT ) . Despite the increased proliferation levels , we initially were not able to get the cells to grow well beyond the fourth to fifth passage . We thus attempted to determine optimal conditions for maintenance of the avian iPSC-like cells . We initially used chicken embryonic stem cell media ( chicken ES media; Supplementary file 1B; [Pain et al . , 1996] ) , which allowed transduction but resulted in few passages . We then discovered that with inclusion of 3i inhibitors of cell differentiation ( Li and Ding , 2010 ) and doubling of LIF , growth was still slow , but the modified media supported proliferation up to about seventh passages ( the condition used for most of our tests ) . Conversely , after decreasing LIF by half and doubling two of the 3i inhibitors , the transformed chicken iPSC-like cells became just as highly proliferative as the mouse iPSC and ESC cells ( reported on in more detail in Dai et al . , unpublished date ) . We are currently passaging these chicken iPSC-like cells 1–2 times per week at a 1:4 dilution , and are above passage 20; we could be further along in passages , but froze the cells at various times over one year to postpone growth in order to conduct other experiments . Thawing the frozen cells does not prevent them from continuing to proliferate at a high rate . Telomerase activity was also activated in all transformed vertebrate cells , and at levels comparable to those seen in the mouse and chicken ESC lines ( Figure 3P ) . Telomerase activity is a characteristic feature of immortal cell lines ( Thomson et al . , 1998 ) . Unlike the vertebrate cells , however , the transformed Drosophila cells did not have telomerase activity ( not shown ) , confirming known absence of telomerase in the Drosophila genome ( Gomes et al . , 2010 ) . We karyotyped some of the avian species to assess chromosomal normalcy . The chicken iPSC-like lines ( Figure 4C , male shown ) displayed a normal karyotype of macro chromosomes in the majority of the spreads analyzed ( 90%; 18 out of 20 ) compared to a standard ( Nanda et al . , 1999 ) , control fibroblasts ( Figure 4D ) . The majority of the zebra finch ( female ) iPSC-like lines ( 90% ) also displayed normal karyotype of macro chromosomes ( Figure 4E , female shown ) . For zebra finch cells , standards were not available , and thus , the control fibroblasts were used as a reference ( Figure 4F ) . The minority of cells that were not normal had tetraploid spreads , but in both iPSC-like and control cells: two out of 20 in the chicken iPSC-like cells and controls , two out of 20 in the finch iPSC-like cell , and one out of 20 in the finch control . This was a result of a doubling of the chromosome complement , which is common in cultured cells . These results with at least the avian cells suggest that major chromosomal arrangements did not occur as a result of the transformation . 10 . 7554/eLife . 00036 . 014Figure 4 . Karyotyping and in vitro pluripotency of iPSC-like cells . ( A ) Embryoid bodies ( EB ) from iPSC-like cells in differentiation media . ( B ) qRT-PCR gene expression analyses of Nestin ( ectoderm marker ) , Brachyury ( mesoderm ) , and Gata-4 ( endoderm ) homologs in undifferentiated iPSC-like cells ( green ) and in EBs ( yellow ) from mouse , bird and fish relative to their control fibroblasts ( normalized; blue ) . Error bars , S . E . M . ( n = 5 replicates of independently generated cell lines or EBs ) . Statistics in Supplementary file 1D . ( C–F ) Karyotypes of macro chromosomal arrangements of the chicken iPSC-like ( C ) , chicken control fibroblasts ( D ) , zebra finch iPSC-like cells ( E ) and zebra finch control fibroblasts ( F ) , exhibiting 18 normal chromosomes . ZZ is female and ZW is male in birds . Black scale bar , 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 014 To assess pluripotency in vitro , we attempted to generate embryoid bodies ( EB; ‘Materials and methods’ [Takahashi and Yamanaka , 2006] ) . Formation of EBs was achieved from the avian , fish , and Drosophila iPSC-like cells , and they appeared similar to those formed from our chicken and mouse ESC lines , and control mouse iPSCs ( Figure 4A ) . The Drosophila EBs were more irregularly shaped . No EB formation occurred with the control cells of any of the species ( fibroblast or S2 ) , indicating that EB formation was specific to the iPSC-like cells and established ESCs . Differentiation into the three germ cell lineages was supported by quantitative RT-PCR of lineage-enriched genes showing over-expression relative to the fibroblasts of Brachyury ( mesoderm ) , Nestin ( endoderm ) , and Gata-4 ( ectoderm ) in all vertebrate species ( Figure 4B ) ( Leahy et al . , 1999; Murakami et al . , 2004; Hailesellasse Sene et al . , 2007 ) . Conversely , the expression of these genes was much lower in our undifferentiated mouse , avian , or fish iPSC-like cells ( i . e . , the iPSC-like , green ) . The in vitro pluripotency results suggest that the iPSC-like cells have the potential to differentiate into multiple cell types , but EBs do not necessarily have advanced differentiated cell types , nor do they conclusively demonstrate the potential for incorporation in vivo . To assess pluripotency in vivo , we employed two strategies: generation of ( 1 ) teratomas and ( 2 ) chimeric embryos with the iPSC-like cells ( ‘Materials and methods’ ) . We did not attempt to do so with the Drosophila cells , as the early embryo is nearly one large cytoplasm partially divided up by membranes ( Mavrakis et al . , 2009 ) . Teratomas were attempted for avian species by injecting the iPSC-like cells into the testes of SCID nu/nu mice in 18 animals for each avian species ( nine with control fibroblasts and 9 with iPSC-like cells ) . After 35 days , two ( out of nine ) of the chicken iPSC-like and three ( out of nine ) quail iPSC-like cells injected mice developed teratomas . These teratomas exhibited organized formation of endoderm ( such as neuronal rossetts , Figure 5A , D ) , mesoderm ( such as bone , Figure 5B , E ) , and ectoderm ( such as G . I Tract , Figure 5C , F ) , demonstrating pluripotency in vivo . None of the controls generated teratomas ( Figure 5H , I ) . So far , none of the zebra finch iPSC-like cells formed teratomas , suggestive of possible species differences for in vivo pluripotency . 10 . 7554/eLife . 00036 . 015Figure 5 . Teratoma formation generated by chicken and quail iPSC-like cells . ( A–C ) Teratoma formation after injections of chicken iPSC-like cells in testes of SCID mice , showing aberrant growth of ( A ) neural like cells ( neuronal rosettes , endoderm , black arrows ) , ( B ) bone-like cells ( mesoderm , black arrows ) , and ( C ) gastrointestinal tract-like cells ( endoderm , black arrows ) . Similar features are seen in the quail generated teratomas ( D–F ) . ( G ) Control testes without cell injections showing normal tissue morphology . ( H ) Testis with control chicken fibroblasts injected showing no germline formation . ( I ) Testis injected with control quail fibroblasts that did not generate teratomas . Panels A–F are at 40 × magnification , whereas G–I are at 4 × in order to get a broader view . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 015 For the chimeric studies , we simultaneously transduced chicken and zebrafish fibroblast cells with the STEMMCA and the GFP lentiviruses , or transduced the cells with the GFP lentivirus after their second to fifth passage from frozen stocks . In both cases , we obtained GFP labeled colonies that still had the characteristic morphology of the iPSC-like cells ( Figure 6—figure supplement 1 ) . Cells were collected , washed , mechanically disassociated , counted , resuspended , and injected into embryonic 1-day ( ED1 ) old chickens or 1–2 hr post fertilized ( 1–2hpf ) zebrafish embryos , respectively . We then fixed the embryos 1–5 days later . We conducted control experiments in parallel with GFP-labeled fibroblasts ( early first to second passage ) injected into the embryos . We obtained animals up until ED4 for chicken and 72hpf for fish . We found that recombinant GFP-labeled chicken and fish iPSC-like cells successfully incorporated into the developing animals ( Figure 5 ) . This required about 5000 cells for chicken and 100–200 for fish . The rate of chimera formation was about 16% for the chicken ( four out of 25 attempts ) and 10% for fish embryos ( 10 out of 103 attempts ) . Embryos injected with iPSC-like cells were subject to higher mortality than those injected with control fibroblasts . In chicken , about 50% of the embryos did not reach the 3rd day of incubation , compared to only 20% for control cells . Similarly with zebrafish , about 60% of embryos injected with the iPSC-like cells did not survive , while the rate was negligible in control fibroblast injected embryos . These results are consistent with lower survival rates observed in iPSC-injected mice ( Maherali and Hochedlinger , 2008 ) , and could be due to multiple factors , such as the iPSC causing tumors and some other type of aberrant growth . Interestingly , after 1 day , some of the iPSC-like-GFP injected fish embryos produced a secondary axis ( Figure 6—figure supplement 2C ) , suggesting a disruption in the developmental program . There were some zebrafish embryos , which , after 1 day of incorporation , exhibited a cluster of GFP labeled fibroblast derived cells ( Figure 6—figure supplement 2 ) , although to a lower intensity than the iPSC GFP homologs ( Figure 6—figure supplement 2B–C ) . However , the starting fibroblast cells did not survive in the 72hpf fish embryos or the ED4 chicken , and thus did not generate fluorescently labeled older chimeras ( Figure 6A , E ) . 10 . 7554/eLife . 00036 . 016Figure 6 . In-vivo pluripotency of iPSC-like cells from chicken and fish . ( A and C ) 4 day old chicken embryos that had been injected with GFP-labeled chicken fibroblasts ( A ) or GFP-labeled iPSC-like cells ( C ) 3 days earlier ( GFP labeled cells in Figure 6—figure supplement 1 ) . Incorporated GFP-labeled cells ( green ) are spread throughout the body for the iPSC cells but , not fibrobloast . ( B and D ) Histological sections stained with antibodies to GFP ( brown ) confirming absence of label in chicken fibroblast injected animals ( B ) , and presence of label in multiple tissue types in the iPSC-like injected animals ( D ) . ( E and G ) 3-day old ( 72hpf ) zebrafish embryos injected with GFP-labeled zebrafish fibroblasts ( E ) or GFP-labelled iPSC-like cells ( G ) , respectively . ( F and H ) Histological sections stained with antibodies to GFP ( brown ) confirming absence in controls ( F ) and presence of labeled cells in iPSC-like injected animals ( H ) . Arrows in all images point to GFP-labeled cells; P . duct = pronephric duct ( P . duct ) . 1 day old fish embryo is shown in Figure 5—figure supplement 2 . Black bars , 30um; white scale bar , 3000 μm for the chicken and 350 μm for the fish . 1 day old post fertilization zebrafish embryos ( Figure 6—figure supplement 2 ) , and chicken embryos with partial incorporation ( Figure 6—figure supplement 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01610 . 7554/eLife . 00036 . 017Figure 6—figure supplement 1 . iPSC-like cells for ( A ) chicken , ( B ) zebrafish , and ( C ) Drosophila , transfected with a GFP expressing lentivirus . Post induction to iPSC-like state . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01710 . 7554/eLife . 00036 . 018Figure 6—figure supplement 2 . 1 day old post fertilization zebrafish embryos . ( A ) Embryos generated with control fibroblast cells exhibiting some localized flourescent cells . ( B ) Generated with iPSC-like GFP cells distributed in several parts of the embryo . ( C ) A double axis embryo , generated with iPSC-like GFP cells , showning one axis with high gfp flourescence , and none on the other . Explanation of histology sections is the same as in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 01810 . 7554/eLife . 00036 . 019Figure 6—figure supplement 3 . Partial incorporation of chicken iPSC-like cells in chicken embryos as demonstrated by fluorescence and immunohistochemistry . ( A ) Chicken embryo ( day 2 ) expressing with incorporated cells expressing GFP in the neuraltube , ( B ) Chicken embryo ( day 3 ) expressing GFP in the mouth and GI Tract C . Chicken embryo expressing GFP in the developing brain . DOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 019 Immunolabeling of GFP in tissue sections confirmed cell incorporation and allowed localization of the incorporated cells . In some animals ( both chicken and fish ) , the cells incorporated in nearly all organs of the body ( Figure 6C , G ) , but for the most part they incorporated sporadically ( chicken iPSC-like cells incorporation , Figure 6—figure supplement 3 ) . In the chicken , the iPSC-like cells differentiated into many cell types , including into muscle , intestines , skin , and brain , while in the fish , most of the incorporation was observed in the stomach and the head ( Figure 6D ) . A separate study will be conducted to see how long the embryos can live with the injected cells and whether they incorporated into mature gonads for germline transmission . No GFP immunolabel was detected in the 72hpf zebrafish or ED4 chicken controls ( Figure 6B , F ) . These features are similar to those seen in the early mouse chimeras created with mouse iPSC cells ( Takahashi and Yamanaka , 2006 ) and chicken chimeras created with chicken ESC cells ( Lavial et al . , 2007 ) . Like those studies , most of our surviving embryos looked normal with no overt differences from animals treated under the same conditions without injected cells or injected with non-transduced fibroblast cells . These findings demonstrate that the non-mammalian vertebrate cells we generated with mammalian genes are pluripotent for at least non-germline cells in a developing animal in vivo , and functionally behave like mouse and chicken ESC and iPSC in vivo up until the ages analyzed .
Our results indicate that at least partially reprogrammed iPSC can be generated in non-mammalian species and that the mammalian genes are sufficient to do so in both non-mammalian vertebrates and invertebrates . While our study was under review , others ( Lu et al . , 2012 ) have recently shown that for one bird species , the quail , some of these properties can be induced using the human genes . We also had tested the four human transcription factors in all aves ( chicken , quail , zebra finch ) and zebrafish cells and found results similar to the ones obtained with the mouse factors ( ‘Materials and methods’ and data not shown ) . We cautiously call these cells partial iPSC or iPSC-like compared to authentic iPSCs , which refer to cells capable of giving rise to not only most cell types of an adult animal , but also to functional gametes for non-human species . Characteristics which the non-mammalian iPSC-like cells we generated have in common with mouse iPSCs and ESCs are colony morphology , marker expression of induced genes , reactivation of some endogenous pluripotency genes , transgene-independent self-renewal , embryoid body formation , teratoma generation ( for chicken and quail ) , and the ability to contribute to different cell lineages in chimeric embryos ( for chicken and zebrafish; Table 1 ) . These findings suggest that the induction process is relatively conserved . 10 . 7554/eLife . 00036 . 020Table 1 . Comparison of characteristics of the IPSCs or PRPSCs cells across speciesDOI: http://dx . doi . org/10 . 7554/eLife . 00036 . 020Stem Cell markersSelf-renewalPluripotencyiPSC or PRPSC cellsMorphologyAlkaline phosphataseInduced endogenous homologsExogenous silencingGene expressionProliferationTelomeraseEB formation/ TeratomaGene expressionChimera formationMouseESC-like clustersYESOct4 , Sox2 , c-myc , Klf4YESNanog Vasa SSEA-1YESYESlarge aggregatesthree germ linesYESChickenESC-like clustersYESOct4 , Sox2 , c-myc , −Klf4YESNanog Vasa SSEA-1YESYESlarge aggregates/ Teratomas formedthree germ linesYESQuailESC-like clustersYESOct4 , Sox2 , c-myc , −Klf4YESNanog Vasa SSEA-1YESYESlarge aggregates /Teratomas formedthree germ linesNDFinchESC-like clustersYESOct4 , Sox2 , c-myc , −Klf4YESNanog Vasa SSEA-1YESYESsmall aggregatesthree germ linesNDZebrafishESC-like clustersYESOct4 , Sox2 , c-myc , −Klf4YES- Nanog Vasa SSEA-1YESYESsmall aggregatesthree germ linesYESDrosophilaDarker , some clustersYESOct4 ( VVL ) , SoxN , d-myc , −Luna ( klf4 ) NDVasa Dichaete Escargot SnailYESNAsmall aggregatesNANDA large number of similarities are found . Species differences are highlighted in bold . NA , not applicable; ND , not done . Some differences to mouse iPSC cells include lack or little induction of Klf4 , overall initial slower growth of cells , lower overall fold-expression increase in endogenous stem cell genes , and presence of autologously derived fibroblast cells for some of the avian species in the middle passages ( Dai et al . , unpublished date ) ; the latter two traits are similar to some human ES cells ( Draper et al . , 2004 ) . For example , in the quail and zebra finch iPSC-like cells Oct4 is 20–80-fold higher than in control fibroblasts , but lower than the 100–400-fold increase seen in the chicken and mouse cells and yet not the 100–1000-fold increase typically seen in mouse cells ( Soldner et al . , 2009 ) . We further note that our fibroblast controls were from early embryos , which already had some Oct4 and Nanog expression ( detected in PCR reactions by the 28th cycle ) . Thus , we believe that not only protocol differences exist between studies that affect expression levels , but we clearly find species differences in induction levels for all four genes . Importantly , despite these species differences , the lower levels of one or more of these genes was still sufficient to generate pluripotent cells . Thus , we conclude that is it is not necessary for these genes to be induced 500–1000-fold in order for the cells to show some level of pluripotency in vitro or in vivo across species . Apparent differences between species include SSEA1 , Nanog , and Klf4 ( Table 1 ) . All species we studied showed induced SSEA-1 , but this gene is not induced in human iPSCs or ESCs ( Takahashi et al . , 2007 ) , indicating that human cells might be different from other vertebrates or at a different stem cell state . Nanog was induced in all species except zebrafish . Nanog is the third master transcription factor in the stem cell regulatory system ( Nichols et al . , 2009; Silva et al . , 2009 ) that promotes self renewal in the absence of LIF ( Chambers and Tomlinson , 2009 ) . However , recent studies suggest that Nanog may not be integrated in a pluripotency regulatory circuit in fish ( Camp et al . , 2009 ) . We note , though , that the so-called fish Nanog homolog ( NCBI Accession # NP_001091862 ) of these and many other studies show low sequence identity across much of the protein coding sequence with birds ( 32% ) and mammals ( 31% ) . In this regard , it is possible that the fish Nanog homolog has not been really identified or is not present in the zebrafish genome . The absence or very low induction of endogenous Klf4 in all of the non-mammalian species may either be due to an inability of the mammalian genes to re-program this gene , redundancy of the Klf family ( Jiang et al . , 2008 ) , or a lineage specific difference of mammals . The later hypothesis is supported by the low expression of Klf4 in chicken ESCs relative to control fibroblasts . Induction of endogenous Klf4 has so far been found in several mammalian species , such as pig , mouse , and rats ( Roberts et al . , 2009 ) . It has been shown that Klf4 preferentially regulates genes involved in cell adhesion , either activating or inhibiting adhesion , and that cell adhesion can inhibit proliferation ( Swamynathan et al . , 2008 ) . This function is consistent with our findings that relative to mice , the non-mammalian iPSC-like cells are more adhesive to each other , which is known to slow proliferation even in mouse iPSCs ( Chen et al . , 2011 ) . Future investigation could test these hypotheses by over-expressing mouse or species-specific Klf4 without silencing , and assessing long-term proliferation . The induced expression of endogenous Oct4 , Sox2 , ( and partially of c-myc ) homologs in all species and their continued expression when the mouse transgenes became silenced suggests that these genes may be playing more conserved inductive roles . Oct-4 and Sox-2 , along with Nanog , are known as master transcription factors for the pluripotent stem cell state in mammalian cells ( Chan et al . , 2011 ) . Oct4 and Sox2 can dimerize and when bound to DNA motifs of their target genes , they activate gene regulatory networks involved in both self-renewal and pluripotency ( Remenyi et al . , 2003; Shi and Jin , 2010 ) . Sox2 alone has been shown to play various roles in different tissue types or cell states ( Tomioka et al . , 2002 ) . Oct-4 has been deemed the most important of these master factors in the mammalian stem cell regulatory system ( Sterneckert et al . , 2011 ) . Although not a master factor , c-myc is known to induce proliferation , by repressing growth arresting genes ( Gartel and Shchors , 2003 ) . This makes it a key contributor in inducing the self-renewal state of the cell . Recently , other factors that are less oncogenic have been shown to be suitable substitutes for c-myc , such as Gliss1 ( Maekawa et al . , 2011 ) . These substitutes may be useful in future studies in non-mammalian species . A recent study also showed that an intermediate state of stem cell induction can exist , by using just a few of the transcription factors ( Lin et al . , 2011 ) . Different from our iPSC-like cells they did not successfully achieve pluripotency in vivo . Thus , our cells appear to , at least , be further along in the programming stage than these cells . All together , we show that there are at least 7–8 stem cell marker genes up-regulated and three lineage specific germ layer genes down-regulated in the induced cells across species , relative to the embryonic fibroblasts . While we have not performed full-scale transcriptome analysis or DNA methylation studies , the results suggest that enough of a gene regulatory network was induced for the vertebrate cells to maintain a stem cell-like state in the absence of continued exogenous mouse transgene intervention ( Maherali and Hochedlinger , 2008 ) , to become stably proliferative in optimal media conditions for the avian cells , and to become pluripotent in vivo . It is important to caution several factors about the Drosophila results . First , the induction of stem cell characteristics was less prominent than in the vertebrates . When embryoid bodies were generated , they did so at a much lower rate with Drosophila cells than with the vertebrate cells . Putative endogenous homolog expression did not occur in two of the genes for the S2 cells and in another three for the BG2 cells . Second , proliferation was enhanced , but the starting S2 cells were already relatively proliferative . This difference might be because S2 cells are partially differentiated , aneuploid , renewable cells ( Moutinho-Pereira et al . , 2010 ) . As such , these cells are a work-horse type of cell line , such as HeLa cells for humans , more so than true primary somatic cells . One interpretation of these findings is that the S2 cells underwent transdifferentiation as a result of the presence of mammalian stem cell genes , but for a stem cell state; that is switching from one highly proliferative state to another . Altogether , the findings suggest that the closer the relationship to mammals , the more reprogrammed characteristics the cells showed . Although we were hoping that the mammalian genes would induce stem cell-like cells in other species , we were quite surprised that they did so and at the efficiency discovered . Substituted gene function among vertebrates ( Enard et al . , 2009 ) and between vertebrates and invertebrates ( Lavial et al . , 2007 ) has been demonstrated previously , but we are not aware of a systematic set of genes doing so . It is possible that a stem cell gene regulatory network and the stem cells themselves share more conserved molecular similarities than differentiated divergent cells . This idea is supported by the fact that Oct4 has been shown to regulate some of the same genes , including Sox2 , in mammals and fish ( Onichtchouk et al . , 2010 ) , and embryonic cells and the three germ layers are more similar to each other in distantly related organisms than their adult cells , which are more divergent ( Parikh et al . , 2010 ) . This is further supported by the finding that the iPSC features were maintained in the vertebrate cells even after the mouse transgenes were suppressed . As such , our findings suggest that stem cells could represent a primordial animal cell state conserved across the animal kingdom . This can be tested by comparing the transcriptomes of iPSCs and more differentiated cell types across species . Future studies may also focus on transducing cells with their species-specific genes , testing non-integrating genome mechanisms of transduction , using promoters that have been shown to work at high efficiencies in a particular species , super-induction and inducing cells to different stem cell states ( Ye and Cheng , 2010 ) . Along these lines , it could even be possible for different stem cell types , such as germ cells , epi-stem cells , and other primed states to be present within the current conditions ( Dejosez and Zwaka , 2012 ) . In summary , the generation of in vivo incorporating stem cells for non-mammalian species should help advance studies of these experimental model systems . These cells can be used as tools to study differentiation , evolution , and disease across a wide range of species , including cancer ( Takashima et al . , 2013 ) . The induced cells might serve as a platform to study cellular evolution at the molecular level .
Embryonic fibroblasts were collected at embryonic day 12 . 5 for mouse and the comparable stages ( embryonic day 8 ) ( Butler and Juurlink , 1987 ) for chicken , quail , and zebra finch . Briefly , several embryos ( n = 4 ) were extracted from the womb or eggs , their head , limbs , and liver removed , and the remaining contents were minced manually using forceps . The minced contents were placed in a 15 ml tube and treated with 0 . 25% trypsin ( 0 . 25% Trypsin/EDTA , Gibco; 1–2 ml per embryo ) for 30 min at 37°C , pipetting briefly every 5 min to enhance dissociation . Trypsin was neutralized with complete media ( Supplementary file 1B ) , cells were spun down , counted ( hemocytometer ) , re-suspended in complete media and plated at a concentration of one embryo per 150 mm dish for mouse and per 100 mm or 60 mm dish for chicken/quail and zebra finch , respectively . When grown to confluent layers , all fibroblasts were passaged in complete media twice before cells were frozen in aliquots . Zebrafish fibroblast cells were purchased ( ATCC , CRL2 ) and maintained per supplier’s specifications at 26°C in zebrafish complete media ( Supplementary file 1B ) . For Drosophila , Schindler’s cell line ( S2 ) , an epithelial-like cell line , was purchased ( ATCC , CRL 1963 ) and passaged ( 1:10 ) and maintained per supplier’s specifications in Drosophila complete medium ( Supplementary file 1B ) . BG2 cells were purchased from the Drosophila Genome Research Center ( ML-dmBG2; number 51 ) , and maintained with growth culture conditions provided by the center . Mouse embryonic stem cells ( ESCs; line R1 [Nagy et al . , 1993] ) were cultured using standard conditions ( Joyner , 1999 ) . Chicken ESCs ( 25th passage ) were provided by Dr Bertrand Pain ( Clermont University , France ) and cultured according to their protocol ( Lavial et al . , 2007 ) . Adult cell lines for mouse , aves , and fish were either generated or purchased ( Supplementary file 1E ) . Lentiviral vectors were generated in human embryonic kidney ( HEK ) 293T cells ( Cell Biolabs , San Diego , CA , Cat # LTV-100 ) , using a third-generation lentiviral system , following a previously described protocol ( Cockrell and Kafri , 2007 ) . Prior to transfection , the cells were plated on 10 cm collagen coated plates at a density that resulted in 60–70% confluency at the time of transfection . A transfection mix was prepared with either 5 , 10 , or 15 μg of DNA of the STEMCCA vector or control GFP lentiviral vectors ( EF1α-GFP; both kindly provided by Dr Gustavo Mostoslavsky ) , packaging cassette ( REV and Gag/Pol , 10 μg ) and the VSV-G ( 5μg ) envelope expression cassette , respectively . The cells were then transfected with the mix , using 40 μl of Lipofectamine ( Invitrogen , Carlsbad , CA ) per plate . 8 hr after the addition of DNA , the transfected cells were washed with PBS and fresh complete media as used for mouse cells ( Supplementary file 1B ) . Media with viral particles were collected every 24 hr for the next 48 hr and stored at 4°C until complete . Viral particles were separated from cellular debris by centrifugation at 4000g for 5 min followed by filtration through a 0 . 45-micron filter . The titer was measured using Quick-Titer ( Cell Biolabs Inc , Cat # VPK-112 ) and promptly stored at −80°C . If necessary , titer concentrations were increased by ultracentrifugation ( SW-29 rotor ) at 50 , 000g for 2 hr , followed by re-suspension in PBS ( pH = 7 . 2 ) . We also used a commercially available human stem cell cassette with GFP ( Biosettia , cat# iPSC-p4F01 ) on the avian cells . We established DNA preps and lentiviral vectors as above . Maximum titer was significantly less than with the STEMCCA cassette ( 2 . 5 × 108 U/ml ) . For Drosophila transductions , we also generated a plasmid with the Metallothionein inducible promoter from the vector pMT/BiP/V5-His A ( Invitrogen ) . The four transcription factors in the STEMMCA cassette described above were cloned into pMT/BiP/V5-His A in two steps: first , the Oct-4 and Klf-4 segment , followed by the Sox-2 , c-myc segment . The cloning was confirmed by sequencing using plasmid and gene specific primers . Transduction was performed using the ViraDuctin system , as per supplier’s protocol ( Cell Biolabs , Cat # LTV-201 ) in complete media ( Supplementary file 1B ) . Before transduction , cells were thawed and cultured in complete media until 80% confluent . After transduction , cells were grown for 5 days ( 2 days for Drosophila ) , then passaged ( first passage ) , and let to grow for approximately 20 days ( 8 days for Drosophila ) in 3i Media ( Stem Cell Sciences , UK , SCS-SF-ES-01 ) or our mouse stem cell media for mouse cells ( Ying et al . , 2008 ) , our modified version of avian stem cell media ( Pain et al . , 1996 ) for avian cells , fish stem cell media and Drosophila stem cell media ( Supplementary file 1B ) . Drosophila cells grew faster than the vertebrate fibroblasts , and thus , markers were observable at earlier time points . Cells of all species were then subsequently passaged when cultures reached confluency , which was every 7–10 days for the vertebrate cells or every 3 days for Drosophila cells , and divided 1:2 ( Aves and Fish ) or 1:10 ( Drosophila , due to more rapid growth ) . Before we performed detailed analyses on multiple transfections , viral transduction efficiency values were assessed at three different STEMCCA concentrations in 48 well plates and cell colony forming units quantified in the vertebrate species ( Figure 2—figure supplement 1 ) . We measured 11 independently transduced plates , and analyzed differences based on titer and species . Based on these initial transduction experiments , most subsequent transductions were performed at 108 U/ml for mouse and 9 . 5 × 109 U/ml for all other species ( birds , fish , fly ) to achieve similar colony forming unit levels as starting points for our analyses . For subsequent analysis , in order to achieve statistical confidence , we transfected 12 to 30 wells seeded with primary cells , in seven different independent experiments . Each well was independently transfected . Samples of the cells were then extracted at various time points , to identify the presence of exogenous or endogenous genes and proteins , via RT-PCR and immunocytochemistry , respectively . For all species , negative control groups were conducted utilizing fibroblasts transduced with a GFP containing lentivirus and grown in the stem cell media ( Supplementary file 1A ) . For in vivo pluripotency experiments , both fibroblasts and iPSC-like cells were first transduced with the GFP lentiviral vector ( titer 108 ) , following the same transfection protocol . We also performed post induction GFP transfection on the Drosophila cells , although these were not used for in vivo studies . To transduce S2 cells with the Metallothionein inducible promoter plasmid , we used a previously described protocol ( Santos et al . , 2007 ) . To induce expression of the transcription factors , 1–2 days after transfection , copper sulfate was added to the medium to a final concentration of 500 μM ( 5 μl of a 100 mM CuSO4 stock ) . To transduce avian cells with the human STEMCCA lentivirus , we used the protocol described in the preceding paragaph . Colonies were observed after around the 7th day , but they numbered less than with cells transduced with the mouse genes . These colonies showed alkaline phosphatase staining and formed embryoid bodies ( not shown ) . Cells or embryoid bodies were spun down and RNA isolated using a standard kit ( Promega SV total RNA isolation system , Z3105 ) . RNA was quantified using a NanoDrop 2000c ( Thermo Scientific , Waltham , MA ) and then stored in −80°C . Complementary DNA ( cDNA ) was produced by reverse transcription ( RT ) in a 20 μl reaction using the supplier’s protocol ( 10 μl of 2X RT buffer and 1 μl of 20X Superscript II enzyme; Applied Biosystems ) . The cDNA was then used as a template to perform PCR gene expression assays in 20 μl reactions containing 1 μl template ( ∼2 μg/μl ) , 10 μl 2X Gene Expression Master Mix ( BioRad , Hercules , CA ) and forward and reverse TaqMan primer probes ( Generated by Applied Biosystems ) or in 20 μl reactions containing the same reagents , but in place of TaqMan primers , custom PCR primers and 1 μl SYBR green ( BioRad ) . To discriminate between endogenous and exogenous expression of the stem cell genes across species , different primers were generated for mouse and the non-mammalian species , using non-overlapping sequences . To discriminate between mouse exogenous and endogenous genes , primers to the WPRE region of the vector were used . Using this strategy , the estimated relative amount of endogenous expression was calculated as the expression level of the WPRE segment subtracted from the total RNA of the mouse specific transcription factors . Primer sequences are listed in Supplementary file 1C . The reactions were performed in a Cx96 real-time machine ( BioRad ) . Cycling conditions were 95°C for 10 min , followed by 35 cycles of denaturation at 95°C for 15 s and annealing/extension at 60°C for 1 min . No-template controls were run for each primer set and probe . 18S rRNA endogenous control was run for each sample using TaqMan primers that recognized the RNA in all species tested ( Cat# Eukaryotic 18S RNA HS99999901_S1; Applied Biosystems ) . The results were normalized to the endogenous 18S expression and to the gene expression level of the control fibroblast/primary cell groups using the ΔΔCT method common for RT-PCR analyses . All primers showed efficiency levels above 90% , using the protocol in the MIQE guidelines ( minimal information for publication of real-time PCR experiments ) ( Bustin et al . , 2009 ) . For statistical analysis , 2-way ANOVAs were performed on two factors ( genes and cell types [iPSC , fibroblast , ESC , EB] ) on n = 5 independently transduced lines ( replicates ) for each of the vertebrate species or n = 3 independent lines for the Drosophila cells . Alkaline phosphatase ( ALP ) activity was measured using the STEMTAG Immunohistochemical Kit ( Cat# CBA-300 , Cell Biolabs ) , following the manufacturer’s protocol . Control fibroblasts , ESCs , and iPSCs were washed with PBS , and fixed with either 4% paraformaldehyde or the kit’s fixing solution for 10 min at room temperature . The fixing solution was then aspirated , the staining solution was placed in each well for 30 min and stored in the dark at room temperature . The wells were washed with dH20 3 times and images were taken immediately under a stereomicroscope without coverslipping . A dark blue/purple color product indicates the presence of ALP enzymatic activity normally found in stem cells , whereas differentiated cells will not stain . The same protocol was also employed , in some instances , with Vector Red as an indicator ( Vector Laboratories , inc , Burlingame , CA ) . To assess proliferation , we used the MTT ( 3-[4 , 5-Dimethylthiazolyl-2]-2 , 5-diphenyltetrazolium bromide ) Quantitative Cell Proliferation Assay ( Cat# 30-1010K; ATCC ) . Tetrazolium salts are reduced metabolically by the cells , resulting in a colorimetric change . The resulting intracellular purple formazan is solubilized and quantified spectrophotometrically ( at 570 nm ) . Cells ( induced and controls ) for all species were plated at 10 , 000 cells/well ( in quintuplets , from independently transduced cells ) and incubated for 24 hr . 10 μl of the MTT reaction solution was added to each plate and incubated for 3 hr . 100 μl of detergent was added to each plate , stored for 2 hr in the dark ( room temperature ) , and the absorbance was measured at 570 nm using a Molecular Devices Emax Microplate Reader . ANOVA was performed to test for differences between cell types and species ( n = 5 independent lines , per species ) . Statistical significance was considered at p<0 . 05 . Telomerase expression is low or absent in most somatic tissues , but not in germ cells , stem cells , and tumors ( Meyne et al . , 1989 ) . The telomerase binds to a particular repeat seq , uence TTAGGG present at the ends of chromosomes of most eukaryotic species and extends them during cell replication . Telomerase enzymatic activity was determined using the Quantitative Telomerase Detection Kit ( BioMax , USA , MT3012 ) , following the manufacturer’s protocol . Cell extracts containing proteins and RNA were generated from the ESC , iPSC , and control fibroblast , and then telomerase activity was measured . If telomerase is present , it adds nucleotide repeats to the end of an oligonucleotide substrate of the kit , which is subsequently amplified by real time qPCR . Quantitation was carried out by the PCR software of the BioRad Cx96 system . Positive control ( template provided with kit ) and negative control ( heat inactivated samples ) reactions were performed . Cycling conditions for the BioRad Cx96 real-time machine were as follows: 48°C for 10 min and 95°C for 10 min , followed by 40 cycles of 95°C for 15 s ( denaturation ) and 60°C for 1 min ( annealing/extension ) . All reactions were performed in quintuplets . Paired t-tests were performed to test for differences of telomerase in the iPSC-like and control fibroblasts of each cell line . Statistical significance was considered at p<0 . 05 . Karyotyping was performed as previously described ( Bangs and Donlon , 2005 ) , by Karyologic , inc . Briefly , cells were seeded in t-25 tissue culture flasks , and allowed to grow . Colchicine ( Colcemid , Invitrogen 15210-040 ) was added to each flask ( 0 . 25 ml/5 ml media ) and incubated at 37°C , 5% CO2 for 12 hr . Cells were then trypsinized , transferred to 15 ml tubes and spun down at 1200 RPM , for 8 min . Cells were then resuspended in 0 . 0075 KCL and incubated at room temperature ( 6 min ) before being spun down again . Cells were then fixed with Methanol/Acetic acid fixative ( 3:1 ) and stored overnight . Cell suspensions were then dropped into cold slides , dried and baked for 20 hr at 65°C . In order to assess the banding of the chromosomes , slides were treated with 0 . 05% trypsin 0 . 02 EDTA at room temperature for 12 s , rinsed quickly in 100% ethanol and then in Gurr’s phosphate buffer ( pH 6 . 8 , Invitrogen #10582-013 ) . Slides were then stained with Karyomax Giemsa ( Invitrogen #10092-013 ) , per manufacturer instructions . To assess the chromosomes , Applied Imaging Genus Cytovision Software ( v2 . 8 ) was used . In order to form embryoid bodies ( EBs ) , the hanging drop method was used ( Keller , 1995 ) . After harvesting the iPSCs or control fibroblasts ( or S2 cells ) directly from culture on the stem cell media , they were washed with PBS ( pH7 . 4; Gibco ) to remove any LIF and resuspended in ‘differentiation media’ which is complete media for each species excluding LIF , cytokines , chemical inhibitors and mercaptoethanol . The cells were then micropipetted in 20 μl drops containing ∼500 cells each on the lids of bacteriological plates ( Sigma , 100 mm ) . The lids were inverted over a dish filled with 10 ml PBS and incubated for 2–3 days . After the embryoid bodies had formed from the iPSC-like cells , the drops were flushed from the lid with differentiation media and grown in suspension culture for another 3–5 days . Embryoid bodies were then collected via pipette , RNA extracted ( as above ) , and qRT-PCR analysis conducted ( as above ) . For SSEA-1 labeling , reactions were performed on cells cultured on coverslips in 24 well plates . The primary SSEA-1 antibody ( Cat# 480 , Santa Cruz Biotechnology , Dallas , TX ) was diluted ( 1:200 ) in PBS . A secondary anti-mouse IgM conjugated to a green fluorescent molecule ( Abcam , Cambridge , MA ) was diluted ( 1:500 ) and incubated at 4°C , overnight . The cells were then washed 3X in PBS and coverslipped with DAPI solution ( VectaShield; Vector Labs ) . Images were taken using a fluorescent microscope ( Olympus Bx61 ) . For GFP labeling ( performed by the Duke University Pathology Lab ) , chicken or zebrafish embryos , or positive control tissue slides ( Mouse GFP positive brain sections ) , were cut at 5 μm on a paraffin block and mounted into glass slides . These were dried for at least 30 min at 60°C in an oven . The slides were deparaffinized in three changes of xylene ( 5 min each ) , 2 changes of 100% EtOH ( 3 min each ) , and 2 changes of 95% EtOH ( 3 min each ) . Rehydration was performed in dH2O for 1 min . To block endogenous peroxidase activity , 3% hydrogen peroxide was used for 10 min , followed by a rinse in dH2O to remove antigens . For the primary antibody ( Anti-Rabbit GFP Abcam ab290 , diluted at 1:100 in PBS [pH = 7 . 1] ) , 200 mls of the citrate , pH 6 . 1 , antigen-retrieval buffer from Dako ( 10X concentrate ) were used . The buffer was preheated to 80°C in a Black and Decker vegetable steamer for 20 min . The slides were then cooled to room temperature in running tap water ( about 15 min ) . Slides were thoroughly rinsed in water and placed in TBST . After antigen retrieval , 10% normal rabbit serum was applied to the slides and incubated for 60 min at room temperature . Afterwards , they were washed with PBS and the excess was drained . After incubation , Vectastain Elite ABC was used , followed by DAB chromagen ( Dako ) , and incubated for 5 min , followed by washing . All slides were counterstained in hematoxylin for 30 s . Slides were rinsed in tap water until clear and coverslipped . Chicken and quail iPSC-like cells and control fibroblasts were grown in 6 well plates , detached , and spun down ( 200g , 5 min ) . The supernatant was removed , and cells were cleaned and re-spun with PBS ( 1X , pH: 7 . 2 ) . The concentration of cells was adjusted to 5 × 106 cells per ml . 5-week-old male SCID mice ( N:NIH-bg-nu-xid; Charles River Laboratories , Raleigh , NC ) were used for each experiment . Animals were anesthetized with intraperitoneal injections of ketamine–xylazine ( 50 and 5 µg/g , respectively ) in saline . 100 μl of the cell solution was injected into the mouse testes . Afterwards , the mice were let to recover from the anesthesia on a heating pad ( Kent Scientific ) . After 5 weeks , the mice were sacrificed , and the testes were removed to assess teratoma formation by histology ( H&E , above ) . Zebrafish were raised as described ( Akimenko et al . , 1995 ) using standard methods in the Poss Lab zebrafish facility ( Duke University ) . Briefly , GFP labeled control and iPSC-like cells were prepared as described above for chicken cells , and adjusted to a concentration of 106 cells/ml in PBS ( pH = 7 . 1 ) . The cell mixture was placed in a borosilicate glass needle fitted to an Eppendorf CellTram Microinjector . Approximately 100 cells were introduced to blastodisc region of a just fertilized ( 0 hpf ) zebrafish embryo . Embryos were then maintained at 31°C . After 1–3 days , the injected embryos were observed under a fluorescent microscope , to determine GFP labeled cell incorporation . Embryos were then fixed in 4% PFA for 20 min , and placed in 70% ethanol and stored at 4°C for immunohistochemical analysis . | Stem cells are ‘pluripotent’—in other words , they have the potential to become many other cell types . This ability makes them extremely valuable for research . They also hold substantial promise for medical applications , since they can be used to replace cells lost or damaged by disease or injury . Embryos represent a rich source of stem cells; however , obtaining these cells from human embryos raises obvious ethical and practical concerns , and they have also been difficult to isolate from many species . A recent discovery circumvented these issues for humans and several mammalian species commonly studied in the laboratory . This technique can turn cells from adult mammals into ‘induced pluripotent stem cells’ , or iPSCs , by switching on four genes . Nevertheless , no analogous method has yet been established to create similar cell populations in non-mammalian organisms , which are also important models for human development and disease . Now , Rosselló et al . have shown that cells from both invertebrate and non-mammalian vertebrate species—including birds , fish and insects—can be reprogrammed into cells that closely resemble iPSCs . Intriguingly , these cells were created by switching on the same four genes that generate iPSCs in mammals , even though vertebrates and invertebrates are separated by around 550 million years of evolution . Rosselló et al . used a viral vector that carries the four stem-cell genes ( from the mouse ) into target cells from the different species . The genetically altered cells developed into iPSC-like cells with many of the characteristics of natural mammalian and bird stem cells . To confirm that the cells were pluripotent , Rossello et al . first showed that the cells could develop into primitive early embryos called embryoid bodies . For the vertebrate species tested , the embryoid bodies contained cells from each of the three main vertebrate embryo cell types . Secondly , iPSC-like cells from two organisms—chicks and zebrafish—formed various mature cell types when injected into developing chick or zebrafish embryos . These results have two important implications . They suggest that the genetic mechanisms by which cells can be reprogrammed into a stem-like state have been conserved through 550 million years of evolution; additionally , they demonstrate that stem-like cells can be generated from important experimental organisms , and provide an important tool for both biological and biomedical research . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine"
] | 2013 | Mammalian genes induce partially reprogrammed pluripotent stem cells in non-mammalian vertebrate and invertebrate species |
Cardiomyocyte ( CM ) maturation in mammals is accompanied by a sharp decline in their proliferative and regenerative potential shortly after birth . In this study , we explored the role of the mechanical properties of the underlying matrix in the regulation of CM maturation . We show that rat and mouse neonatal CMs cultured on rigid surfaces exhibited increased myofibrillar organization , spread morphology , and reduced cell cycle activity . In contrast , compliant elastic matrices induced features of CM dedifferentiation , including a disorganized sarcomere network , rounding , and conspicuous cell-cycle re-entry . The rigid matrix facilitated nuclear division ( karyokinesis ) leading to binucleation , while compliant matrices promoted CM mitotic rounding and cell division ( cytokinesis ) , associated with loss of differentiation markers . Moreover , the compliant matrix potentiated clonal expansion of CMs that involves multiple cell divisions . Thus , the compliant microenvironment facilitates CM dedifferentiation and proliferation via its effect on the organization of the myoskeleton . Our findings may be exploited to design new cardiac regenerative approaches .
During the early postnatal period , cardiomyocytes ( CMs ) undergo a switch from a proliferative , hyperplastic mode to non-proliferative , hypertrophic growth that persists throughout life ( Li et al . , 1996; Soonpaa et al . , 1996; Soonpaa and Field , 1998 ) . This process is accompanied by multiple , highly synchronized cellular changes . The expression of cell-cycle and embryonic markers falls precipitously in CMs ( Walsh et al . , 2010 ) , while the expression of CM differentiation genes increases , along with the appearance of sarcomeres , contractile units that undergo cross-striation to form a functional myoskeletal system within CMs . Further , mechanical and electrical cell-to-cell communication between CMs is established via intercalated discs , containing gap junctions , adherens junctions and desmosomes ( Noorman et al . , 2009; Sheikh et al . , 2009 ) . In parallel , up-regulation of extracellular matrix ( ECM ) components , as well as increased cross-linking of matrix proteins , also occurs postnatally , leading to an overall stiffening of the heart tissue ( Janmey and Miller , 2011; Swift et al . , 2013; Majkut et al . , 2014 ) . The mechanisms whereby these complex developmental processes are regulated and coordinated , and their putative effects on CM maturation , are yet enigmatic . Over the first week of postnatal life in mice , most CMs exit the cell cycle and differentiate ( Soonpaa et al . , 1996 ) , although some CMs undergo an additional burst of proliferation in the pre-adolescent period , driven by the thyroid hormone ( Naqvi et al . , 2014 ) . Mammalian CMs display distinct cycling phases: a fetal/neonatal phase in which nuclear division ( karyokinesis ) is immediately followed by cell division ( cytokinesis ) , and a later neonatal phase , in which karyokinesis proceeds without cytokinesis , leading to binucleation ( Soonpaa et al . , 1996; Li et al . , 1997a , 1997b; Zebrowski and Engel , 2013 ) . Binucleation , hypertrophic growth , increased myofibrillar organization , and cell cycle withdrawal are all manifestations of the differentiated state of adult CMs ( Ahuja et al . , 2007; Naqvi et al . , 2009; Zebrowski and Engel , 2013 ) , while the ‘terminal differentiation’ state is yet uncertain . Naturally , there are physiological advantages in the postnatal maturation of CMs . Nonetheless , loss of the proliferative potential of differentiated CMs creates a major barrier to cardiac regeneration after injury . In humans , myocardial infarction ( MI ) is a life-threatening disease leading to permanent loss of hundreds of millions of CMs , followed by an inflammatory response and formation of scar tissue , that progressively lead to cardiac dysfunction and heart failure ( Virag and Murry , 2003; Ausoni and Sartore , 2009 ) . Cardiac regeneration does exist in lower vertebrates such as newts and fish . The zebrafish heart , for example , is able to fully regenerate after injury , without scarring ( Poss et al . , 2002 ) . The lack of regenerative potential of the mammalian heart was challenged by Porrello and colleagues ( Porrello et al . , 2011 ) , who showed that the neonatal murine heart displays a transient regenerative phase that diminishes within the first week after birth . Both mammalian and zebrafish heart regeneration is characterized by increased CM proliferation associated with sarcomere disassembly , attributed to a CM dedifferentiation process ( Poss , 2007; Jopling et al . , 2010 , 2011; Porrello et al . , 2011 ) . Recent studies with diverse cell types indicate that the rigidity of the underlying ( or surrounding ) matrix strongly influences cell structure , cytoskeletal organization , migration , polarization , and regulation of gene expression and cell fate ( Pelham and Wang , 1997; Discher et al . , 2005; Engler et al . , 2006; Vogel and Sheetz , 2006; Engler et al . , 2008; Geiger et al . , 2009; Prager-Khoutorsky et al . , 2011; Shin et al . , 2011 ) . In particular , it was demonstrated that interaction with rigid surfaces ( hundreds of kPa and stiffer ) promotes the formation of large , integrin-mediated adhesions , and consequently , development of a well-organized and polarized cytoskeleton ( Prager-Khoutorsky et al . , 2011 ) . Following this logic , we tested here the hypothesis that the decline in CM proliferative and regenerative capacities can be attributed to changes in the mechanical properties of the pericellular environment , which facilitate the assembly of a tightly organized myoskeletal structure . We hypothesize that matrix stiffening , which occurs in the heart after birth , and even more so following MI , is part of a mechanism that blocks CM cell cycle activity . In this study , we discovered that soft matrices promote CM dedifferentiation , as evidenced by CM rounding , myofibrillar disassembly , increased CM cell division , and clonal expansion . In contrast , a rigid matrix facilitates CM differentiation , characterized by tightly packed arrays of long and aligned sarcomeric bundles , cell cycle arrest , and binucleation . Furthermore , we demonstrated that matrix rigidity specifically affects CM cytokinesis , but not karyokinesis . We next demonstrated that disruption of the highly organized architecture of the CM myoskeleton , using the myosin-II inhibitor blebbistatin , induced CM cell cycle re-entry . We suggest that the mechanical properties of the postnatal heart play a key role in the acquisition of the fully differentiated phenotype , and inhibition of the specialized contractile system in CMs could be used to promote CM dedifferentiation .
To study the effect of matrix rigidity on CM cellular characteristics , we plated newborn ( P1 ) rat CMs on PDMS substrates with different rigidities , ranging from stiff ( 2 MPa ) to compliant ( 20 kPa and 5 kPa ) , and analyzed the sarcomeric and cellular organization . Myosin heavy chain ( MHC ) , Myomesin ( an M-band protein ) and cardiac Troponin T ( cTnT ) were expressed in CMs cultured on both rigid and compliant substrates; however , CMs on the rigid substrate displayed aligned sarcomeres with defined and registered striations , whereas CMs on the compliant substrate had disorganized sarcomeres , with misaligned myoskeletal structures ( Figure 1A ) . Furthermore , CMs cultured on the rigid substrate appeared larger , elongated , and mostly triangular , whereas CMs on the compliant substrate were round , smaller , and less polarized . CMs cultured on the 20 kPa substrate had intermediate parameters in terms of their polarization and spreading morphologies , compared to the 2 MPa and the 5 kPa substrates ( Figure 1B–D ) . 10 . 7554/eLife . 07455 . 003Figure 1 . Compliant substrates alter neonatal cardiomyocyte cell shape and sarcomeric organization , and promote their proliferation . ( A ) Representative images of the sarcomeric patterns of MHC ( top ) , myomesin ( middle ) , and cTnT ( bottom ) on rigid 2 MPa ( left ) and soft 5 kPa ( right ) substrates . Scale bar: 20 µm . ( B ) An overview of neonatal P1 rat CMs on rigid ( left ) and soft substrates ( right ) . Scale bar: 50 µm . ( C ) Axial ratio of P1 rat CMs ( D ) Cell area of P1 rat CMs ( n = 1 , 108 , for C , D ) . ( E , F ) Representative immunofluorescence images of proliferating P1 rat CMs on various PDMS substrates . Scale bar: 20 µm . CMs were stained for ( E ) sarcomeric cTnT and proliferation marker Ki67 , or ( F ) cardiac MHC and PH3 ( PH3+/MHC+ are denoted with arrows ) . ( G ) Quantification of Ki67+/cTnT+ P1 rat CMs on different rigidities ( n = 4 , 165 ) . ( H ) Quantification of PH3+/MHC+ P1 rat CMs on different rigidities ( n = 1 , 964 ) ( I ) Representative images of neonatal P1 mouse CMs on 2 MPa , 20 kPa , and 5 kPa substrates . ( J ) Quantification of Ki67+/cTnT+ P1 mouse CMs on different rigidities ( n = 11 , 876 ) . Statistical significance was determined using ANOVA followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 , two ( ** ) if p < 0 . 01 , and three ( *** ) if p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 00310 . 7554/eLife . 07455 . 004Figure 1—figure supplement 1 . Compliant substrates alter P1 mouse cardiomyocyte cell shape , polarity , and sarcomeric organization . ( A ) Cell area of P1 mouse CMs cultured on rigid and compliant substrates ( n = 1 , 118 ) . ( B ) Axial ratio of CMs on the different substrates ( n = 1 , 115 ) . ( C ) Cell perimeters of CMs ( n = 1 , 131 ) . Statistical significance was determined using ANOVA , followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) for p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 004 To determine the effects of substrate rigidity on CM proliferation , we stained CMs with the cell-cycle markers Ki67 and phospho-histone-3 ( PH3 ) , together with cTnT and MHC ( Figure 1E , F ) . Proliferating CMs were observed on all tested matrices ( Ki67+/cTnT+ , and PH3+/MHC+ ) ( Figure 1E , F , respectively ) . CMs cultured on either glass or on the stiff ( 2 MPa ) substrate presented similar proliferative capacities ( data not shown ) . However , CM proliferation was increased on the compliant substrates ( 5 kPa and 20 kPa ) by ∼50–65% , respectively , relative to the rigid substrate ( Figure 1G , visualized with Ki67 ) , and by ∼25–110% on the 5 kPa and 20 kPa substrates , respectively , marked by PH3+ MHC+ CMs ( Figure 1H ) . Taken together , these findings suggest that myoskeletal disassembly , a tendency toward cell rounding , and increased CM proliferation are more compatible with compliant matrices , in comparison with the rigid matrix . Inspired by the important finding of the transient regenerative potential in the neonatal mouse heart ( Porrello et al . , 2011 ) , we established a mouse culture system of P1 neonatal CMs cultured on 2 MPa , 20 kPa , and 5 kPa substrates . As in rat CMs , mouse neonatal CMs grew on all substrates , and developed normal beating within 48–72 hr after seeding . We quantified the percentage of proliferating Ki67+ cTnT+ CMs , and observed that both the 5 kPa and 20 kPa substrates facilitated CM proliferation by ∼30–50% , respectively , relative to the rigid 2 MPa substrate ( Figure 1I , J ) . There was no significant difference in the area of cells plated on the different substrates ( Figure 1—figure supplement 1A ) . Similar to the rat system , CMs cultured on the rigid substrate were elongated and more polarized , compared to those on the compliant substrates ( Figure 1—figure supplement 1B ) . A decrease in CM cell perimeter , indicative of roundness , was observed on the compliant substrates , compared to the rigid one ( Figure 1—figure supplement 1C ) . Hence , compliant substrates promote CM proliferation as well as alterations in CM cell shape , in both rat and mouse neonatal cultures . Various proliferation markers ( e . g . , Ki67 shown in Figure 2A , BrdU , PH3 , and Aurora B ) , are widely used for cell-cycle assessment and quantification of cell proliferation; however , they sometimes prove insufficient , especially in CMs in which polyploidization and binucleation are natural outcomes of cell proliferation in the postnatal period ( Bersell et al . , 2009; Zebrowski and Engel , 2013 ) . In order to explore the effects of matrix rigidity on CM cell division and the formation of new CMs , we established a live-cell imaging system of CMs derived from transgenic mice expressing the R26R-tdTomato reporter under the regulation of the Myh6 promoter ( Myh6-Cre;R26R-tdTomato-lox ) ( Figure 2B ) . Live-cell video microscopy revealed two distinct CM cell-cycle phenotypes: karyokinesis followed by binucleation ( Figure 2C , Video 1 ) , as opposed to karyokinesis followed by cytokinesis , resulting in the formation of two new CMs ( Figure 2D , Video 2 ) . In the first , CMs were relatively large , spread , and immotile ( Figure 2C: 50′ , 60′ ) . These CMs underwent karyokinesis , leading to the formation of binucleated CMs ( Figure 2C: 50′ , 60′ ) , while remaining well spread and attached to the substrate ( Figure 2C ) . 10 . 7554/eLife . 07455 . 005Figure 2 . Distinct rigidity-dependent cardiomyocyte division mechanisms . ( A ) Immunofluorescent images of dividing neonatal CMs in culture . The newly formed CMs are still connected , and both nuclei are Ki67+ ( red ) . ( B ) A schematic drawing of the experimental design . ( C ) Video frames of a P1 Myh6-Cre;R26R-dTomato-lox CM undergoing karyokinesis followed by binucleation , without changing its morphology . ( Inset ) A fibroblast undergoing typical cell division . ( D ) Video frames of a P1 Myh6-Cre;R26R-dTomato-lox CM undergoing karyokinesis , followed by complete cytokinesis . The original CM is highlighted in 0′ . The two new CMs are highlighted in 120′ . Yellow arrows mark karyokinesis ( 20′ , 30′ ) ; white arrows mark the new nuclei ( 40′ ) , or new CM ( 60′ , 110′ ) . ( E ) P1 mouse CM karyokinesis on different rigidities . ( F ) P1 mouse CM cytokinesis on different rigidities . ( G ) P1 mouse CM binucleation on different rigidities ( n = 2 , 167 for E , F , G ) . ( H ) Quantification of new CMs on different rigidities ( n = 2 , 878 ) . Statistical significance was determined using ANOVA followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 , and two ( ** ) if p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 00510 . 7554/eLife . 07455 . 006Figure 2—figure supplement 1 . Binucleated cardiomyocytes re-enter the cell cycle . ( A ) A binucleated P1 Myh6-Cre;R26R-dTomato-lox CM ( highlighted ) re-enters the cell cycle , but fails to divide , and remains binucleated . ( B ) A binucleated P1 Myh6-Cre;R26R-dTomato-lox CM ( highlighted ) re-enters the cell cycle , changes its morphology , and divides , forming new CMs . ( C ) A schematic drawing of binucleated CM cell cycle re-entry; however , the CM remains binucleated . ( D ) A schematic drawing of a binucleated CM re-entering the cell cycle , undergoing rounding , and completing cytokinesis , to form new CMs . ( E ) Binucleated CMs cell-cycle re-entry events occurring on matrices of varying rigidities . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 00610 . 7554/eLife . 07455 . 007Video 1 . Mouse cardiomyocyte binucleation . A 24 hr time-lapse video of a representative P1 Myh6-Cre;R26R-dTomato-lox CM undergoing karyokinesis followed by binucleation . Time-lapse-10 min . Scale bar: 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 00710 . 7554/eLife . 07455 . 008Video 2 . Mouse cardiomyocyte cytokinesis . A 48 hr time-lapse video of a representative P1 Myh6-Cre;R26R-dTomato-lox CM undergoing karyokinesis followed by cytokinesis . Time-lapse-10 min . Scale bar: 30 µm . One of the mononucleated daughter cells undergoes consecutive cell division , and karyokinesis followed by cytokinesis . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 008 In contrast , CMs that completed cell division ( cytokinesis ) underwent a step of mitotic rounding ( Figure 2D ) , which is common for most proliferating cells ( Lancaster and Baum , 2014 ) ; moreover , these CMs often underwent consecutive cell divisions . Strikingly , we found that compliant matrices did not affect nuclear cell division ( karyokinesis ) , yet promoted cytokinesis and inhibited CM binucleation prominence , as determined by quantification of the division frequency , observed by live-cell imaging ( Figure 2E–G ) . In order to demonstrate an actual increase in CM number , we quantified the amount of CMs at the beginning and after 48 hr . A significant increase in the number of newly formed CMs was observed on the 20 kPa substrate , relative to the rigid 2 MPa ( Figure 2H ) . Interestingly , we could observe rare events of cytokinesis even in binucleated CMs cultured on the 20 kPa substrate , resulting in two daughter CMs ( Figure 2—figure supplement 1E ) . These successful events were also accompanied by mitotic rounding ( Figure 2—figure supplement 1 ) . Taken together , our findings demonstrate that culturing CMs on compliant matrices facilitate CM cell rounding and division ( cytokinesis ) that lead to formation of new CMs . In contrast , our results demonstrated that the rigid matrix promotes karyokinesis without cytokinesis , leading to CM binucleation ( Figure 2G ) . To further investigate the molecular status of CMs undergoing cytokinesis , we designed an assay that enabled us to correlate between the live imaging videos , in which we could visualize cell division processes ( Figure 2 ) , with molecular and lineage analyses of the dividing cells ( Figure 3 ) . Accordingly , correlated live cell-immunofluorescence microscopy was performed on CMs derived from P1 transgenic Myh6-Cre;R26R-tdTomato-lox mice cultured on 2 MPa , 20 kPa and 5 kPa substrates in grid-containing plates . The ‘Tomato’ cells represent CMs that express , or previously expressed , the Myh6 gene , a signature of CM differentiation . Under regular conditions , we detected almost 100% of dTomato+/cTnT+ double positive CMs ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 07455 . 009Figure 3 . Compliant matrices induce cardiomyocyte dedifferentiation . ( A–C ) Correlative live-cell-immunofluorescence of CM dedifferentiation on ( A ) 2 MPa , ( B ) 20 kPa , and ( C ) 5 kPa matrices . Recently divided P1 Myh6-Cre;R26R-dTomato-lox CMs , following time-lapse imaging ( A , B , C , left panel , and A′ , B′ , C′ ) , correlated with the expression of cTnT ( A′′ , B′′ , C′′ ) , and Ki67 ( A′′′ , B′′′ , C′′′ ) . CMs following cell division are denoted with white arrows . ( D ) An overview of a field of P1 Myh6-Cre;R26R-dTomato-lox CMs ( red , left panel ) , immunostained for cTnT ( green , middle panel ) , and a merge of Tomato+/cTnT− images ( right panel ) . Scale bar: 50 µm . ( E ) Quantification of Tomato+/cTnT− CMs on 2 MPa , 20 kPa , and 5 kPa ( n = 19 , 869 ) . ( F ) An overview of a field of P1 Myh6-Cre;R26R-dTomato-lox CMs ( red , left panel ) , immunostained for MHC ( green , middle panel ) , and a merge of the Tomato+/MHC− images ( right panel ) . Scale bar: 20 µm . ( G ) Quantification of Tomato+/MHC− CMs on 2 MPa , 20 kPa , and 5 kPa substrates . Scale bar: 20 µm ( n = 13 , 474 ) . ( H ) P1 Myh6-Cre;R26R-dTomato-lox/cTnT− CMs expressing Nkx2 . 5 . ( I ) Schematic diagram showing CM dedifferentiation from a mature to an immature proliferative state , facilitated by compliant matrices . Statistical significance was determined using ANOVA followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 00910 . 7554/eLife . 07455 . 010Figure 3—figure supplement 1 . dTomato cardiomyocytes express cTnT . ( A , B ) Representative image of P1 Myh6-Cre;R26R-dTomato-lox-derived CMs co-stained with cTnT antibody ( A ) , and their quantification showing less than 1% of Tomato+/cTnT negative CMs ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 010 Time-lapse imaging was performed for 48 hr , and immediately thereafter , cultures were fixed and immunostained for cTnT and Ki67 . We first examined the time-lapse videos for CMs undergoing complete cell division ( karyokinesis plus cytokinesis ) , and identified the two daughter cells by using the grid coordinates ( Video 3–5 ) . By correlating the last frame of each time-lapse video ( Figure 3A–C ) with cTnT staining , we revealed that the dividing CMs ( tdTomato positive; Figure 3A′ , B′ , C′ ) lost cTnT expression , either completely ( Figure 3A′′ , B′′ ) , or partially ( Figure 3C′′ ) . This result is consistent with a CM dedifferentiation process , in which CMs , originating from the Myh6 lineage , lost cTnT expression ( Figure 3A′′ , B′′ , C′′ ) . Furthermore , the majority of these cells expressed Ki67 on all three matrices , indicating that these CMs maintain a proliferative potential ( Figure 3A′′′ , B′′′ , C′′′ ) . 10 . 7554/eLife . 07455 . 011Video 3 . Mononucleated cardiomyocyte cytokinesis on a 2 MPa substrate . 48 hr time-lapse imaging ( with 10 min intervals ) of P1 Myh6-Cre;R26R-dTomato-lox CMs on a 2 MPa substrate . Scale bar: 30 µm . The last frame of this video is shown in Figure 3A , and was correlated with immunofluorescence staining , as shown in Figure 3A′′ , A′′′ . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 01110 . 7554/eLife . 07455 . 012Video 4 . Mononucleated cardiomyocyte cytokinesis on a 20 kPa substrate . 48 hr time-lapse imaging ( with 10 min intervals ) of P1 Myh6-Cre;R26R-dTomato-lox CMs on a 20 kPa substrate . Scale bar: 30 µm . The last frame of this video is shown in Figure 3B , and was correlated with immunofluorescence staining , as shown in Figure 3B′′ , B′′′ . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 01210 . 7554/eLife . 07455 . 013Video 5 . Mononucleated cardiomyocyte cytokinesis on a 5 kPa substrate . 48 hr time-lapse imaging ( with 10 min intervals ) of P1 Myh6-Cre;R26R-dTomato-lox CMs on a 5 kPa . Scale bar: 30 µm . The last frame of this video is shown in Figure 3C , and was correlated with immunofluorescence staining , as shown in Figure 3C′′ , C′′′ . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 013 To quantify CM dedifferentiation on the different matrix rigidities , we performed immunofluorescence analysis of P1 CMs cultured on the different substrates . We counted the number of tdTomato+/cTnT− CMs ( Figure 3D , E ) or MHC− ( Figure 3F , G ) . A ∼twofold increase in CM dedifferentiation ( Tomato+/cTnT− and Tomato+/MHC− ) was observed on the 20 kPa , and an ∼1 . 5-fold increase , on the 5 kPa matrices , relative to the rigid matrix ( Figure 3E , G ) . Moreover , all Myh6 lineage-positive ( tdTomato+ ) CMs in the culture expressed Nkx2 . 5 , indicating that these cells derived from the cardiac lineage ( Figure 3H ) . Our findings , showing dividing CMs that downregulated cTnT and/or MHC sarcomeric proteins and express Nkx2 . 5 , strongly suggest that the compliant matrices promote the dedifferentiation of CMs , consistent with their role in promoting cytokinesis ( Figure 3I ) . Our findings thus far suggest that disruption of the CM myoskeleton promotes CM cell cycle re-entry . We therefore tested whether inhibition of the CM myoskeleton function could boost CM proliferation . For that purpose , we used blebbistatin , a small molecule myosin II inhibitor that effectively blocks the actin-myosin interaction in several striated muscle ( including CMs ) , in smooth muscle , and in non-muscle cells ( Dou et al . , 2007 ) . Neonatal mouse CMs were cultured on PDMS substrates with the different rigidities , left to adhere for 72 hr , and then treated with 20 µM blebbistatin for 24 hr . After 24 hr of treatment , we fixed the cells and immunostained them for cTnT and Ki67 . We found that blebbistatin severely disrupted CM myoskeleton morphology , leading to massive changes in CM cell shape , manifested by major flattening , dramatic fragmentation of the sarcomeric arrays , and complete loss of striations ( Figure 4A , B ) . 10 . 7554/eLife . 07455 . 014Figure 4 . Blebbistatin induces CM proliferation . ( A , B ) Representative images of P1 mouse CMs cultured in the absence of blebbistatin ( A ) , or presence of 20 µM blebbistatin ( B ) , on 2 MPa ( top ) , 20 kPa ( middle ) , and 5 kPa ( bottom ) substrates . Scale bar: 20 µm . ( C ) Quantification of Ki67+/cTnT+ P1 mouse CMs on different rigidities in the presence or absence of 20 µM blebbistatin ( n = 25 , 554 ) . Statistical significance was determined using ANOVA followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 , and two ( ** ) if p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 014 We next analyzed CM proliferation in the presence or absence of blebbistatin , by quantifying the percentage of Ki67+/cTnT+ CMs ( Figure 4C ) . This analysis indicated that blebbistatin-treated CMs display higher cell cycle activity compared to control ( untreated ) CMs on the 2 MPa and 20 kPa rigidities ( Figure 4C ) . Hence , we conclude that relaxation of the acto-myosin myoskelton in CMs by blebbistatin promotes cell cycle re-entry . The lack of blebbistatin effect on the 5 kPa substrate is consistent with the relaxed CM morphology induced by the 5 kPa substrate in the absence of blebbistatin . Taken together , these results suggest that complaint matrices promote CM proliferation primarily via their effect on myoskeletal organization . A closer examination of CMs derived from the Myh6 lineage that were grown on the 20 kPa matrix and lost cTnT and/or MHC expression , revealed clusters of cells ( more than 2 ) , suggesting that these cells were derived from a common CM that underwent multiple cell divisions ( Figure 5—figure supplement 1 ) . To explore this notion further , we took advantage of the R26R-Confetti reporter line ( Snippert et al . , 2010 ) . In this setting , the R26R-Confetti reporter generated stochastic , multicolor Myh6 lineage-derived CMs with four fluorescent proteins , each marking individual clones ( Figure 5A–F; high magnification areas showing heterogeneous multicolor CM clones in vivo D , E , or individual YFP+ CM clones in F ) . To asses the clonal potential of CMs grown on the different rigidities , we cultured P1 CMs derived from Myh6-Cre;R26R-Confetti-lox hearts , and imaged them at 4 , 5 , 7 , 8 9 , and 11 days in culture . At each time point , the exact same fields were imaged , in order to enable the tracing of CM cell division and clonal expansion . Unicolor clones , likely originating from a single CM , were identified on all three rigidities . 10 . 7554/eLife . 07455 . 015Figure 5 . The compliant 20 kPa matrix promotes cardiomyocyte clonal expansion . ( A ) A schematic drawing of the clonal tracing strategy of Myh6-Cre;R26R-Confetti-lox mice . ( B ) In vivo section of a neonatal Myh6-Cre;R26R-Confetti-lox heart . Scale bar: 100 µm . ( C ) In vitro culture of Myh6-Cre;R26R-Confetti-lox—derived CMs . Scale bar: 20 µm . ( D–F ) in vivo images of Myh6-Cre;R26R-Confetti-lox heart sections at high magnification . Scale bar: 20 µm . ( D–E ) Heterogeneous population of CMs . ( F ) A clone of YFP+ CMs , surrounding a single RFP+ CM . ( G ) 20 kPa substrate facilitates a 2 . 4-fold increase in CM number , relative to 2 MPa . ( H ) 20 kPa substrate facilitates an increase in the percentage of clones . ( I ) Total number of cells over time on 2 MPa , 20 kPa , and 5 kPa substrates , showing a greater initial number of cells , as well as a greater final number of cells , on 20 kPa substrates ( n = 6 , 401 for G , n = 5 , 110 for H , I , ) . ( J ) Formation of a YFP-positive clone on a 20 kPa substrate , throughout the experiment . Clonal expansion is indicated by an increase in the number of cells . Statistical significance was determined using ANOVA , followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 , and two ( ** ) if p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 01510 . 7554/eLife . 07455 . 016Figure 5—figure supplement 1 . Formation of dedifferentiated cardiomyocyte clones on a 20 kPa matrix . A cluster of Tomato+/cTnT−/Nkx2 . 5+ expressing cells ( more than 2 ) on a 20 kPa matrix , suggesting clonal behavior of cells derived from a common CM ( or a few CMs ) that underwent multiple cell divisions ( circled clone ) . Scale bar: 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 01610 . 7554/eLife . 07455 . 017Figure 5—figure supplement 2 . Clonal behavior of individual P1 Myh6-Cre;R26R-Confetti-lox—derived cardiomyocytes on different rigidities . Development of P1 Myh6-Cre;R26R-Confetti-lox -derived clones at different time points , on matrices with rigidities of 2 MPa ( left ) , 20 kPa ( middle ) , and 5 kPa ( right ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 017 We quantified the total amount of CMs that were formed between day 4 and day 11 , the percentage of single-color clones , and the number of cells in each clone , for each of the three rigidities ( Figure 5G–I , respectively . Figure 5—figure supplement 2 ) . We found ∼2 . 4-fold higher numbers of total CMs formed on the 20 kPa relative to the 2 MPa substrate ( Figure 5G ) . A ∼twofold increase in the percentage of unicolor clones was observed on the 20 kPa substrate , compared to both the rigid 2 MPa and softest 5 kPa substrates ( Figure 5H ) . Next , we quantified the number of cells in all the clones , at each time point , and found that the 20 kPa matrix induced an impressive increase in CM cell number ( Figure 5I ) . A typical clonal expansion of an individual YFP-positive clone , from only 2 cells at Day 4 , to 9 cells within a week ( Day 11 ) , is shown ( Figure 5J ) . This analysis revealed that neonatal CMs can undergo clonal expansion in vitro , a process which involves multiple cell divisions; moreover , this potential can be induced most readily on the 20 kPa matrix , compared to either rigid or softer matrices .
In this study , we explored the effect of the microenvironment , specifically its rigidity , on CM cell fate , with an emphasis on CM proliferation and dedifferentiation . Notably , most studies that address the impact of rigidity on cell fate decisions have focused on the differentiation processes ( Engler et al . , 2008; Jacot et al . , 2008; Bajaj et al . , 2010; Bhana et al . , 2010 ) . The terminology and characteristics of the reverse process are much less clear , but generally refer to the shift from a differentiated to a less-differentiated cellular stage within the same lineage ( Jopling et al . , 2011 ) . Here , we show that neonatal CMs cultured on compliant ( compared to rigid ) 2D matrices lose several key manifestations commonly associated with postnatal cardiac muscle differentiation in vivo , such as robust , well-aligned myoskeletel organization , expression of differentiation genes , and cell cycle arrest . Hence , we conclude that compliant substrates promote CM dedifferentiation and proliferation ( Figure 6 ) . In line with these findings , we demonstrated that dedifferentiated CMs were able to undergo clonal expansion , which involves multiple cell divisions , suggesting a shift to a progenitor cell state . 10 . 7554/eLife . 07455 . 018Figure 6 . The crosstalk between matrix rigidity and CM cell fate . A schematic model showing the effect of matrix rigidity on CM cell fate . Rigid substrates maintain the differentiated state of CMs , which is incompatible with cytokinesis . Blocked cytokinesis leads to binucleated CMs . A mitotic checkpoint determines the ability of CMs to re-enter the cell cycle . Compliant substrates control a second checkpoint , mechanical in nature , which enables cytokinesis . Compliant substrates induce CM dedifferentiation into a less organized , immature and proliferative state . This state involves cytoskeletal disarrangements leading to CM mitotic rounding , thus enabling cell division . DOI: http://dx . doi . org/10 . 7554/eLife . 07455 . 018 What is the mechanism underlying the matrix rigidity-dependent CM dedifferentiation process ? In particular , we wondered whether there is a clear hierarchy between the mechanical properties of the heart during embryonic and postnatal stages . We propose that the rigidification of the heart after birth from 10-20 kPa to 40–55 kPa ( Berry et al . , 2006; Engler et al . , 2008; Jacot et al . , 2010 ) promotes CM terminal differentiation by triggering the formation of a rigid myoskeleton that mechanically interferes with CM cytokinesis , without affecting karyokinesis . Ultimately , this physiological transition leads to an accumulation of binucleated CMs , correlating with the transition from embryonic hyperplastic growth of CMs , to postnatal hypertrophic growth ( Figure 6 ) . It was previously shown that CM shape is linked to sarcomeric alignment ( Bray et al . , 2008 ) ; circular CMs do not assemble actin networks or sarcomeric arrays , whereas rectangular CMs do . Consistent with this , we show high polarity and organization of triangular CMs on rigid substrates , as opposed to low polarity , roundness , and disorganized sarcomeres on compliant substrates . Eukaryotic cell division requires cellular and cytoskeletal remodeling: cells become round and spherical upon entering mitosis ( Heng and Koh , 2010; Lancaster et al . , 2013 ) . This mitotic rounding is , to some extent , considered a hallmark of cell division , and the interplay between the actin cytoskeleton , cell shape , and the developing mitotic spindle are all involved in this rounding ( Cadart et al . , 2014; Lancaster and Baum , 2014 ) . Thus , it appears that cytoskeletal organization and cell-cycle control are linked together . In line with this notion , we demonstrate that inhibition of the organized CM contractile apparatus following blebbistatin treatment resulted in increased CM cell cycle re-entry . We show an increase in Ki67-expressing CMs in the presence of blebbistatin on the 2 MPa and 20 kPa rigidities . Blebbistatin treatment strongly perturbed CM myoskeletal organization and consequently induced CM cell cycle activity , suggesting that the development of a robust myoskeleton in CMs is linked to their cell cycle arrest . We suggest that compliant matrices promote CM proliferation primarily via their effect on the organization of the myoskeleton . More broadly , we propose that the rigid matrix increases the forces imposed on the CMs' actin myoskeleton and this facilitates CM differentiation , while inhibition of sarcomeric organization , function and contractility by blebbistatin has the opposite effect , eventually , leading to CM dedifferentiation . That said , it was previously shown that long blebbistatin treatment inhibits cell division , leading to generation of polyploid megakaryocytes ( Shin et al . , 2011 ) . Polyploidity and binucleation are natural outcomes of postnatal CM proliferation events ( Zebrowski and Engel , 2013 ) . It is possible that the high CM cell-cycle activity observed in the presence of blebbistatin may ultimately result in inhibition of cytokinesis , leading to CM binucleation , or even multinucleation . One of the most important findings in this study is that cytokinesis , but not karyokinesis , is affected by matrix rigidity: it is facilitated by the compliant substrates , or inhibited by the rigid substrate . Taken together , the data presented here suggest that committed CMs cultured on the 20 kPa ( and , to some extent , the 5 kPa ) substrate lose the environmental mechanical barrier , leading to the disorganization of the myoskeleton , which facilitates the acquisition of a rounded morphology compatible with cytokinesis and cell division ( Figure 6 ) . The mechanisms underlying CM binucleation , due to failure to complete cytokinesis , are poorly understood . Recently , the view that binucleated CMs cannot divide was challenged , showing that a proliferative burst on P15 , induced by a thyroid hormone surge , led to a 1 . 4-fold increase in CM number ( Naqvi et al . , 2014 ) . We observed rare events of binucleated CMs attempting to divide , though most remain binucleated . However , the few successful division events were facilitated by growth on the 20 kPa matrix , further supporting the notion that CM can respond to matrix rigidity . Interestingly , transgenic expression of a constitutively active ( ca ) form of ErbB2 in mouse CMs during the neonatal and adult periods could stimulate CM division , expanding the postnatal proliferative and regenerative windows into adulthood . Stimulation of CM division by caErbB2 involved both CM dedifferentiation , and hypertrophy leading to cardiomegaly . In line with the observed dedifferentiation phenotype , caErbB2 signaling promoted proliferation of mono- and bi-nucleated CMs ( D'Uva et al . , 2015 ) . In other model systems , rigidity was shown to influence proliferation and stemness of cells . In line with our data , mouse embryonic fibroblasts ( MEFs ) , vascular smooth muscle cells ( VSMCs ) , and MCF10A cells proliferate best on ∼24 kPa ( Klein et al . , 2009 ) . Moreover , muscle stem cells ( MuSCs ) that were cultured on a soft substrate that mimics muscle tissue rigidity ( ∼12 kPa ) display delayed differentiation , self-renewal , and maintained stemness ( Gilbert et al . , 2010 ) . In agreement with our findings , these researchers propose that decreased rigidity preserves stemness by altering cell shape , resulting in cytoskeletal rearrangements . In both neonatal mice and adult zebrafish , cardiac regeneration processes were accompanied by loss of sarcomeric structures and re-acquisition of cell division , both attributed to CM dedifferentiation ( Poss , 2007; Jopling et al . , 2010; Porrello et al . , 2011 ) . Our correlative live-imaging and immunostaining results support the notion that dedifferentiated CMs disassemble their sarcomeres , in order to divide . The fact that CMs have lost MHC or cTnT expression , and the partial disassembly of the sarcomeres during cytokinesis , further support our model of CM dedifferentiation , rather than the presence of progenitor cells . The Myh6-Cre;R26R-confetti-lox system enabled us to document clonal behavior in neonatal CMs . We show that clonal expansion is rigidity-dependent , with an optimal clonal efficiency on the 20 kPa . Clone formation originated at various time points and progressed at different rates , suggesting a continuous proliferative capacity , rather than a final ‘burst’ of proliferation . Whether neonatal or adult CMs are capable of undergoing multiple cycles of cell division in vivo , has yet to be determined . Our findings bear obvious relevance for cardiac regeneration after injury . In this unfortunate event , CMs in the vicinity of the scar tissue sense a stiffer environment , compared to the healthy adult heart ( Jacot et al . , 2010 ) . Thus , if fully differentiated CMs present in the vicinity of the damaged area , can still maintain the capacity to dedifferentiate and resume cell division if confronted with a sufficiently compliant scaffold , a new avenue toward the regeneration of heart tissue may be opened .
Experiments were approved by the Animal Care and Use Committee of the Weizmann Institute of Science . Neonatal rat CMs were isolated and cells were cultured , as previously described ( Shneyvays et al . , 2002 ) . Briefly , 1-day-old newborn Wistar rats were decapitated; the hearts were harvested , cut into small pieces ( 1–2 mm ) , and washed several times in phosphate-buffered saline ( PBS ) to remove excess blood . To obtain a CM-rich culture , the hearts were digested in RDB , a proteolytic enzyme extracted from the fig tree ( Biological Institute , Nes Ziona , Israel ) in 6–8 cycles of 10 min each , at room temperature . The cells were centrifuged at 1500 rpm for 5 min . The pellet was resuspended and cultured in a 10 cm dish for 30 min-1 hr for pre-plating , to produce a CM-rich culture . CMs were then cultured on glass or PDMS surfaces . CMs began to beat spontaneously within 2–3 days , on all of the tested substrates . CM isolation was performed using the Neonatal Heart Dissociation Kit ( gentleMACS , Miltenyi Biotec , Auburn , CA , USA ) . CMs were cultured in DMEM/F12 containing Na-pyruvate , non-essential amino acids , penicillin ( 100 U/ml ) , streptomycin ( 100 μg/ml ) , 2 mM L-glutamine , 5% horse serum , and 20% fetal bovine serum ( FBS ) . Briefly , 1-day-old newborn wild-type ( WT ) or Myh6-Cre;R26R-tdTomato-lox CMs were decapitated; the hearts were harvested and transferred into a 10 cm dish containing PBS , and remaining blood was carefully pumped . Hearts were transferred into the gentleMACS C Tube , and enzyme mix was added . C Tubes containing hearts and enzyme mix were incubated for 15 min at 37 °C , transferred to the gentleMACS Dissociator , and the gentleMACS Program ‘mr_neoheart_01’ was run . These incubation and dissociation steps were repeated 2–3 times . C tubes containing hearts were centrifuged at 2000 rpm for 5 min . The pellet was resuspended and cultured in a 10 cm dish for 30 min-1 hr for pre-plating , to produce a CM-rich culture . CMs were then cultured on fibronectin-coated PDMS surfaces . CMs began to beat spontaneously within 2–3 days , on all of the tested substrates . For live-cell imaging and imuunofluorescence assays of Myh6-expressing cells , we crossed mice carrying the Cre coding sequence inserted after the alpha myosin heavy chain promoter ( Myh6-cre ) , which can drive high-efficiency gene recombination in CMs ( Agah et al . , 1997 ) , with ROSA26-flox-STOP-tdTomato indicator mice ( R26R-tdTomato ) ( Madisen et al . , 2010 ) , thus creating red fluorescent-labeled CMs that could be tracked . For clonal analysis of Myh6-expressing cells , Myh6-Cre mice were crossed with the ROSA26-flox-STOP-Confetti reporter mice ( Snippert et al . , 2010 ) . ROSA26-flox-STOP-Confetti reporter mice ( R26R-confetti reporter mice ) were kindly provided by Dr . Shalev Itzkovitz , Weizmann Institute of Science . Myh6-Cre;R26R-confetti-lox CMs were cultured on different PDMS matrices on grid-bottomed MatTek dishes . Using a DeltaVision Elite system ( Applied Precision , USA ) on an Olympus IX71 inverted microscope , running softWoRx 6 . 0 , fluorescence images were acquired at 10× or 20× magnifications , by a CoolSnap HQ2 CCD camera ( Roper Scientific , USA ) . For clonal analysis , we utilized the DeltaVision collect panels option . Panel stitching was performed using softWoRx . Polydimethylsiloxane ( PDMS ) substrates of varying rigidities were prepared , using a Sylgard 184 silicone elastomer kit ( Dow Corning , USA ) . The silicone elastomer component was mixed with the curing agent , degassed , and spin-coated at 2000 rpm for 2 min with a Spin Processor WS-650MZ-23NPP/LITE ( Laurell Technologies ) on MatTek #0 , #1 or MatTek #1 . 5 glass-bottomed dishes ( MatTek Corporation ) , or microscopy coverslips ( Electron Microscopy Science ) for live-cell and for immunofluorescence experiments , or MatTek gridded #2 glass-bottomed dishes for the correlation live-cell immunofluorescence experiments , to obtain a 35 ± 5 μm-thick PDMS layer . Subsequently , crosslinking of the elastomer was carried out at 70°C overnight . The compliance of the PDMS substrates was verified by using the methodology as previously described ( Prager-Khoutorsky et al . , 2011 ) . Briefly , polymerized slabs of PDMS were used for bulk measurements of Young's moduli , using an Instron universal testing machine ( Instron ) . Elastomer to curing agent ratios of 10:1 , 50:1 and 75:1 corresponded to Young's moduli of 2 MPa , 20 kPa and 5 kPa , respectively . Dishes with a layer of PDMS were functionalized with 20 μg/ml fibronectin at 4°C overnight . Before cell plating , plates were washed with PBS and growth medium . CMs were cultured for 72 hr , and then treated with the myosin II inhibitor blebbistatin ( 20 µM , Sigma ) for 24 hr . Following blebbistatin treatment , cells were permeabilized , fixed , and immunostained for sarcomere and proliferation markers . Cells were permeabilized with 3% paraformaldehyde ( PFA ) in PBS containing 0 . 25% Triton X-100 for 3 min , and then fixed with 3% PFA in PBS for 20–30 min , washed 3 times in PBS , blocked with 10% goat serum in PBS , and stained for the following sarcomeric markers: cardiac troponin T ( cTnT , abcam ) ; MF20 ( MHC , Myosin heavy chain Developmental Studies Hybridoma Bank [DSHB] ) ; cardiac MHC ( Abcam ) ; myomesin ( DSHB ) ; Nkx2 . 5 ( Abcam ) , and the following proliferation markers: Ki67 ( Abcam ) and phospho-histone-3 ( PH3 , Santa Cruz ) . Secondary antibodies used were Alexa-488 , Cy3 , Alexa- 647 , conjugated anti-mouse IgG1 , and Cy5-conjugated anti-mouse IgG2b . Cells were also stained for DAPI , to visualize the nuclei . Live-cell imaging and sample examination were performed using a DeltaVision Elite system ( Applied Precision , USA ) , on an Olympus IX71 inverted microscope , running softWoRx 6 . 0 . Fluorescent images were acquired at 10× , 20× , 40× and 60× magnifications , by a CoolSnap HQ2 CCD camera ( Roper Scientific , USA ) . Time-lapse imaging was carried out for 24 hr or 48 hr at 10 min intervals , and acquired at a 20× or 10× magnification ( 20×/0 . 5NA or 10×/0 . 3NA objectives ) . Correlated live-cell immunofluorescence microscopy was performed on gridded #2 glass-bottomed dishes ( MatTek Corporation ) , using either 10×/0 . 3NA or the 20×/0 . 5NA objectives for time-lapse imaging , and the 10×/0 . 3NA or 20×/0 . 85NA objectives for immunofluorescence correlation . Images of Myh6-Cre;R26R-confetti-lox mice in vivo sections were taken with a Nikon Ti-E inverted fluorescence microscope equipped with a 10× , or 100× oil-immersion objectives and a Photometrics Pixis 1024 CCD camera using MetaMorph software ( Molecular Devices , Downington , PA ) . The image-plane pixel dimension was 1 . 3 µm for 10× magnification , and 1 . 3 µm for 100× magnification . CM projected area and best-fit ellipse aspect ratio of cardiac troponin T-stained cells were calculated , using ImageJ software . Descriptions , titles and arrows have been added to all videos using ‘Final Cut Pro’ editing software . Generally , all experiments were carried out with n ≥ 3 . In all panels , numerical data are presented as mean +s . e . m . Statistical significance was determined using ANOVA , followed by post-hoc Tukey's ( HSD ) test . Results are marked with one asterisk ( * ) if p < 0 . 05 , two ( ** ) if p < 0 . 01 , and three ( *** ) if p < 0 . 001 . | Heart muscle contracts and relaxes in a regular rhythm to pump blood around the body . Soon after birth , the cells that form our heart muscle stop multiplying . As we grow , these cells increase in size and their internal skeleton—called myoskeleton—becomes more complex , to withstand the demands of pumping more blood . However , because the cells can no longer divide , the body is unable to replace heart muscle cells that are damaged by a heart attack or other illness . This lack of ability to regenerate heart muscle is a major challenge for medicine . While researchers have documented many of the changes that occur in heart muscle cells ( known as cardiomyocytes ) after birth , it is not known exactly what triggers these changes . A network of proteins and other molecules—also known as a matrix—surrounds the cardiomyocytes and affects their behavior . Here , Yahalom-Ronen et al . investigated the degree to which the mechanical properties of this matrix affect the ability of cardiomyocytes to divide . In the experiments , cardiomyocytes from newborn rodents were grown on matrices with different rigidities . The cells grown on rigid matrices stopped dividing and became larger with a more robust myoskeleton . These cells also contained two nuclei , which indicates that these cells have become mature cardiomyoctyes . In contrast , heart cells grown on a softer matrix continued to multiply . These cells also began to lose some of the features that distinguish mature cardiomyocytes from the cardiomyocytes found in embryos . Next , Yahalom-Ronen et al . treated the cardiomyoctes with a drug that stops them from contracting , which led to increases in cell multiplication . Yahalom-Ronen et al . 's findings suggest that the stiffness of the matrix that surrounds heart muscle cells regulates their ability to divide and mature . In the future , these findings may pave the way towards the development of soft scaffolds that can stimulate the regeneration of adult human heart . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2015 | Reduced matrix rigidity promotes neonatal cardiomyocyte dedifferentiation, proliferation and clonal expansion |
We take a functional genomics approach to congenital heart disease mechanism . We used DamID to establish a robust set of target genes for NKX2-5 wild type and disease associated NKX2-5 mutations to model loss-of-function in gene regulatory networks . NKX2-5 mutants , including those with a crippled homeodomain , bound hundreds of targets including NKX2-5 wild type targets and a unique set of "off-targets" , and retained partial functionality . NKXΔHD , which lacks the homeodomain completely , could heterodimerize with NKX2-5 wild type and its cofactors , including E26 transformation-specific ( ETS ) family members , through a tyrosine-rich homophilic interaction domain ( YRD ) . Off-targets of NKX2-5 mutants , but not those of an NKX2-5 YRD mutant , showed overrepresentation of ETS binding sites and were occupied by ETS proteins , as determined by DamID . Analysis of kernel transcription factor and ETS targets show that ETS proteins are highly embedded within the cardiac gene regulatory network . Our study reveals binding and activities of NKX2-5 mutations on WT target and off-targets , guided by interactions with their normal cardiac and general cofactors , and suggest a novel type of gain-of-function in congenital heart disease .
The mammalian heart is a highly modified muscular vessel whose lineage programmes are governed by conserved gene regulatory networks ( GRNs ) ( Davidson and Erwin , 2006 ) . The cardiomyocyte GRN is controlled by lineage-restricted transcription factors ( TFs ) , which interact to form a recursively wired sub-network termed the cardiac ‘kernel’ ( Davidson and Erwin , 2006 ) . Kernel TFs , such as NKX2-5 , GATA4 , TBX5/20 , and serum response factor ( SRF ) , show regionally restricted expression and act as selector proteins that help define developmental and organ-specific territories . A well-studied kernel TF is NKX2-5 , a member of the NK-2 homeodomain ( HD ) factor subclass that plays an early role in development of hearts and heart-like organs of diverse species ( Elliott et al . , 2010 ) . Loss of Nkx2-5 in mice leads to arrested heart morphogenesis due to blocked progenitor growth , defective chamber and conduction system development , and a deranged GRN ( Prall et al . , 2007 ) . In humans , NKX2-5 is one of the most commonly mutated single genes in congenital heart disease ( CHD ) , with dominant alleles causing atrial septal defect and atrioventricular conduction block most commonly , and a host of more severe defects at lower penetrance ( Elliott et al . , 2010 ) . Many CHD-causing NKX2-5 mutations are located within the conserved HD ( Figure 1A ) , which serves as both a sequence-specific DNA-binding domain and protein-binding interface for interactions with other kernel TFs ( Elliott et al . , 2010 ) . It is widely assumed that CHD is caused by haploinsufficiency and an inability of mutant proteins to recognise target genes . However , we know little about disease causation at the genome level , and in fact , most CHD mutations lie outside of the HD ( Figure 1A ) , often in conserved domains with largely unknown functions . 10 . 7554/eLife . 06942 . 003Figure 1 . DNA adenine methyltransferase identification ( DamID ) identifies a robust set of NKX2-5 targets in HL-1 cardiomyocytes . ( A ) Structure of the human NKX2-5 protein ( TN , tinman domain; NK2SD , NK-2 specific domain; YRD , tyrosine-rich domain ) . Bars and arrows indicate missense and termination mutations associated with congenital heart disease ( CHD ) , respectively . ( B ) Top over-represented motifs discovered de novo in NKX2-5 peaks using Trawler or Weeder . NKX2-5 , GATA , and Nuclear Factor 1 ( NF1 ) binding motifs deposited in TRANSFAC are shown . ( C ) Distribution of NKX2-5 , GATA , and NF1 binding sequences in NKX2-5 peaks . ( D ) Yeast-two-hybrid assay . NKX2-5 and NF1 proteins were fused to Gal4-activation and DNA-binding domains , respectively . Positive signs ( + ) show interaction as growth on selective medium from three independent experiments . ( E ) Normalized median expression of NKX2-5-target genes in 91 murine cell types ( data collected from BioGPS ) . Tissues with the highest median expressions are shown in colour , including heart ( red ) . ( F ) Expression of NKX2-5 target genes and random genes in HL-1 cells . Data collected from ( Mace et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00310 . 7554/eLife . 06942 . 004Figure 1—figure supplement 1 . DamID validation in HL-1 cardiomyocytes . ( A ) Expression of V5-tagged DNA adenine methyltransferase ( Dam ) and NKX2-5-Dam in HEK Ecr-293 cells in presence of 10 mM Ponasterone A detected by immunofluorescence microscopy using anti-V5 antibodies . ( B ) Expression of Dam alone ( ‘D’ ) and Dam/NKX2-5 fusion proteins ( WT , ∆HD , and Y191C ) in HEK Ecr-293 cells in absence ( − ) or presence ( + ) of 5 mM Ponasterone A detected by western blotting using anti-NKX2-5 antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00410 . 7554/eLife . 06942 . 005Figure 1—figure supplement 2 . Expression of PGK-GFP control 24 hr post-transduction of HL-1 cells transduced with lentivirus ( LV ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00510 . 7554/eLife . 06942 . 006Figure 1—figure supplement 3 . PCR-amplified methylated fragments of HL-1 genomic DNA 40 hr post-transduction with Dam alone and Dam-NKX2-5 . Controls are ‘not transduced’ cells , or cells transduced with empty lentivirus ( ‘LV’ ) , or with Dam alone and amplified with ‘no ligase’ or ‘no DpnI’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00610 . 7554/eLife . 06942 . 007Figure 1—figure supplement 4 . False discovery rate of Dam/TF fusion protein binding peaks as determined using CisGenome/TileMapv2 with moving average ≥ 3 . 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00710 . 7554/eLife . 06942 . 008Figure 1—figure supplement 5 . Chromatin immunoprecipitation ( ChIP ) -PCR validation of NKX2-5 WT target peakes determined by DamID . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00810 . 7554/eLife . 06942 . 009Figure 1—figure supplement 6 . NKX2-5 binds the NKE but not the NF1-like motif . ( A ) Cell-free expression of NKX2-5 WT , NKX2-5Y191C , and NKX2-5∆HD proteins revealed by western-blotting using anti-V5 antibodies . ( B ) Electrophoretic mobility shift assay using NKX2-5 and DIG-labelled NKE ( lanes 1–7 ) and NF-1 oligos ( lanes 8–9 ) . Controls include no protein ( lane 1 ) , supershifts after addition of anti-V5 ( lane 3 ) or anti-NKX2-5 ( lane 4 ) antibodies and competition with unlabelled oligos ( lane 5 ) . See ‘Material and methods’ for details . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 00910 . 7554/eLife . 06942 . 010Figure 1—figure supplement 7 . Identification of NKX2-5 , SRF , and ELK1/4 target genes in HL-1 cells . UCSC Genome Browser screen shots showing DamID TF association in HL-1 cells with known NKX2-5 target genes ( Myocd [Ueyama et al . , 2003] , Actc1 [Chen and Schwartz , 1997] , and Gata4 [Riazi et al . , 2009] ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01010 . 7554/eLife . 06942 . 011Figure 1—figure supplement 8 . Nuclear localisation of NKX2-5 and histone modifications in HL-1 cardiomyocyte nuclei . Scale bars represent 10 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01110 . 7554/eLife . 06942 . 012Figure 1—figure supplement 9 . Proportional Venn diagram showing the overlapping binding peaks between NKX2-5 determined by DamID ( this study ) or ChIP-seq ( He et al . , 2011 ) and ( van den Boogaard et al . , 2012 ) . Only peaks that fall in the regions covered by the Affymetrix mouse promoter microarrays are represented . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 012 Here , we take a functional genomics approach to understanding the mechanism of NKX2-5 CHD at the chromatin level . We applied the technique of DNA adenine methyltransferase identification ( DamID ) ( Vogel et al . , 2006 ) to identify target genes of NKX2-5 wild type and NKX2-5 mutant proteins mimicking those found in patients with CHD . While binding of severe NKX2-5 mutants to targets was compromised , they nonetheless associated with hundreds of targets , including many normally bound by NKX2-5 wild type , and could regulate a subset of these in cellular assays . We demonstrate that severe NKX2-5 mutant proteins retained an ability to interact with NKX2-5 wild type and other cardiac TFs via a novel tyrosine-rich protein:protein interface in NKX2-5 that lies outside of the HD . NKX2-5 mutant proteins also bound hundreds of ‘off-targets’ not bound by NKX2-5 wild type , via altered DNA-recognition or via interactions with previously unrecognised cofactors . These cofactors , which include members of the ETS family , are embedded in the cardiac GRN , indicating a broad role for ubiquitous TFs in cardiac network logic and mechanism of CHD .
DamID is a sensitive enzymatic method developed to identify protein:DNA interactions ( van Steensel and Henikoff , 2000; van Steensel et al . , 2001 ) . It is complementary to chromatin immunoprecipitation ( ChIP ) but avoids artifacts associated with chromatin crosslinking and poor-quality ChIP antibodies ( Waldminghaus and Skarstad , 2010; Teytelman et al . , 2013 ) . DamID involves creation of a fusion between a chromatin-binding protein of interest and Escherichia coli DNA adenine methyltransferase ( Dam ) . When expressed , Dam fusions bind to DNA , whereby Dam locally methylates adenine within its target sites ( 5′GATC3′ ) . Target DNA can then be released using DpnI , a nuclease specific for the methylated Dam site , and analysed on arrays ( Vogel et al . , 2006 ) . Here , we used DamID in HL-1 cells , which resemble atrial cardiomyocytes , to profile NKX2-5 WT and mutants that mimic those found in CHD ( Claycomb et al . , 1998 ) . Mouse Nkx2-5 cDNAs were cloned into a lentiviral Dam vector carrying heat shock and ponasterone A-inducible promoters ( Vogel et al . , 2006 ) . Only N-terminal Dam fusions ( Dam-NKX2-5 ) were analysed because C-terminal fusions were sterically compromised ( data not shown ) . Ponasterone A induction after viral transduction of HEK Ecr-293 cells confirmed nuclear localisation and the correct size of fusions ( Figure 1—figure supplement 1 ) . Transduction of HL-1 cells was efficient ( ∼97%; Figure 1—figure supplement 2 ) . DamID fusions were expressed from the uninduced heat shock protein 68 promoter at very low levels ( undetectable by western blotting ) , which was sufficient for Dam enzymatic activity on chromatin ( Figure 1—figure supplement 3 ) but avoided skewing of the network . DamID is dependent on the density of DpnI sites ( 5′GATC3′ ) , which occur on average every 260 bp in the mouse genome , as in the fly genome ( van Steensel and Henikoff , 2000 ) . After DpnI digestion of virus-transduced HL-1 cell DNA , PCR-amplification lead to fragments of ∼200-2000 bp ( Figure 1—figure supplement 3 ) . We determined that of all perfect NKX2-5 high-affinity binding sites ( NKE; 5′AAGTG3′ ) in the mouse genome , 90% had a DpnI site within 1 kb upstream and downstream ( median of ∼289 bp ) , suggesting that DamID captures the vast majority of NKX2-5 direct targets . Methylated genomic fragments were hybridised to Affymetrix Mouse promoter microarrays in biological triplicate . These microarrays represent 7 . 3% of the mouse genome , and we estimate from published NKX2-5 ChIPseq data that they would capture 20–30% of all NKX2-5 peaks ( He et al . , 2011; van den Boogaard et al . , 2012 ) . DamID peaks were determined using CisGenome after subtraction of signal obtained from cells expressing Dam alone ( Ji et al . , 2008 ) ( see Supplementary file 1 , and UCSC genome browser http://genome . ucsc . edu/cgi-bin/hgTracks ? db=mm9&type=bed&hgt . customText=ftp://ftp . ncbi . nlm . nih . gov/geo/series/GSE44nnn/GSE44902/suppl/GSE44902_DamID . bed . gz [Kent et al . , 2002] ) . For NKX2-5 WT , we identified 1524 peaks , which displayed low-false discovery rates ( <0 . 08; Figure 1—figure supplement 4 ) . ChIP-PCR confirmed NKX2-5 occupancy over NKX2-5 DamID peaks for 10 out of 11 targets tested , while DamID-negative regions were not occupied ( Figure 1—figure supplement 5 ) . De novo motif discovery using Weeder or Trawler ( Pavesi and Pesole , 2006; Haudry et al . , 2010 ) revealed that the top 3 over-represented motifs were 5′AAGTG3′ , identical to the NKE; 5′GATAA3′ , identical to the GATA-binding motif; and 5′TGCCAA3′ , similar to the binding motif of Nuclear Factor 1 ( NF1 ) ( Figure 1B ) . After counting of their exact sequences , the NKE was present in 79% of NKX2-5 peaks , with half of these bearing over 2 and up to 12 NKEs . The GATA- and NF1-binding motifs were present in 59% and 36% of peaks , respectively ( Figure 1C ) . These results suggest that the majority of NKX2-5 WT targets are directly bound via the NKE and also support the combinatorial action of NKX2-5 with GATA factors on many targets ( Durocher et al . , 1997; He et al . , 2011 ) . NKX2-5 did not bind directly to the NF1-like site in vitro ( Figure 1—figure supplement 6 ) , suggesting that NKX2-5 could act in combination with NF1 in HL-1 cells . Using a yeast two-hybrid ( Y2H ) assay , we showed that NKX2-5 interacted with NF1-B1 and NF1-B3 but not NF1-A or NF1-X ( Figure 1D ) , as for the related family members NKX2-1 and NF1 in lung ( Bachurski et al . , 2003 ) . This suggests that NF1-B1/3 regulate a subset of target genes in combination with NKX2-5 in cardiomyocytes . Using GREAT ( McLean et al . , 2010 ) , NKX2-5 peaks were assigned to 1490 unique genes , including previously identified NKX2-5 target genes ( Figure 1—figure supplement 7 ) . NKX2-5 also bound its own promoter , potentially reflecting auto-regulation ( Prall et al . , 2007 ) . Gene Ontology ( GO ) analysis of DamID NKX2-5 targets showed that the most enriched biological processes were heart development and muscle contraction ( Supplementary file 1 ) , attesting to the specificity of DamID . The most represented words amongst the top 50 GO terms , were muscle , development , regulation , cardiac , and cell . Other enriched processes included cytoskeleton organisation and related terms , and metabolic GO terms such as regulation of cellular carbohydrate catabolic process , regulation of glycolysis , and regulation of generation of precursor metabolites and energy , indicating that NKX2-5 exerts high-level network control over metabolic processes . In silico analysis using the BioGPS repository of 91 tissue transcriptomes revealed that the median expression level of NKX2-5 target genes was highest in adult hearts ( Figure 1E ) . Transcriptome data for HL-1 cells ( Mace et al . , 2009 ) showed that NKX2-5 targets were strongly skewed towards higher expression values compared to genes randomly selected from the array , although some genes were expressed at lower levels ( Figure 1F ) . We note from immunofluorescence studies that the most intense NKX2-5 staining co-localised in speckle-like nuclear foci with active histone marks H3K9Ac and H3K4me3 , but not with repressive mark H3K27me3 ( Figure 1—figure supplement 8 ) . Having established a robust NKX2-5 WT target list , we set out to probe the mechanisms of CHD by identifying the genome-wide targets of NKX2-5 mutant proteins . We first constructed N-terminal Dam fusions for the human CHD mutation NKX2-5Y191C ( Benson et al . , 1999 ) , which disrupts tyrosine 54 of the HD , necessary for the unique binding site specificity of the NK-2 homeoprotein sub-class to the NKE ( Figure 2A ) . In vitro , NKX2-5Y191C dimerises normally with co-factors but shows 80-fold reduced binding to the NKE ( Kasahara et al . , 2000; Kasahara and Benson , 2004 ) . We also created NKX2-5∆HD , in which the HD was deleted and replaced by a glycine-linker . We anticipated that NKX2-5∆HD would be functionally dead , since the HD has been shown to serve as both DNA-binding domain and interface for homo- and hetero-dimerisation ( Elliott et al . , 2010 ) . However , we were interested in testing whether the presence of other conserved domains in NKX2-5 and often mutated in NKX2-5-related CHD ( Figure 2A ) could confer any functionality to the NKX2-5 mutants . Both mutants were stable ( Figure 1—figure supplements 1 , 6 ) and , as anticipated , did not bind efficiently to the NKE of the known target Nppa ( natriuretic peptide precursor A ) ( Durocher et al . , 1996 ) in vitro ( Figure 1—figure supplement 6 ) . 10 . 7554/eLife . 06942 . 013Figure 2 . NKX2-5 mutants bind to hundreds of targets in HL-1 cells . ( A ) Structure of NKX2-5Y191C and NKX2-5∆HD mutant proteins ( ‘G’ indicates a Glycine linker ) . ( B ) Overlapping binding peaks between NKX2-5 WT and mutants . Proportional Venn diagrams show peaks unique to WT NKX2-5 ( A sets ) ; common to WT and mutant proteins ( B sets ) ; unique to mutant proteins ( C sets ) . ( C ) Representative autocorrelation function ( ACF ) curves and fits for NKX2-5 WT and mutant proteins measured by photoactivatable fluorescence correlation spectroscopy ( paFCS ) in HL-1 cells . ( D ) Top over-represented motifs discovered de novo using Weeder in A- , B- and C-target sets of NKX2-5 WT and mutant comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01310 . 7554/eLife . 06942 . 014Figure 2—figure supplement 1 . ( A ) Image showing paGFP-NKX2-5 and histone 2B-RFP fluorescence in a single HL-1 nucleus with fluorescence correlation spectroscopy ( FCS ) measurement performed at the crosshair point . ( B ) Typical ACF curves for each NKX2-5 WT and mutant proteins measured by paFCS in HL-1 cells . ( C ) Individual fit residuals for data shown in Figure 2C . ( D ) Summary of diffusion parameters for photoactivatable GFP ( paGFP ) -fusion proteins in HL-1 cells . Data are mean ± standard error of n measurements . τDslow , dwell time of the slow-diffusing component; αslow , anomalous parameter of the slow-diffusing component; Γ/4slow , transport coefficient for anomalous diffusion ( G = 4D for non-anomalous diffusion ) of the slow-diffusing component; τDfree , dwell time of the free-diffusing component; Dfree , diffusion coefficient of the free-diffusing component; %free , percentage of the free-diffusing component . Statistical significance indicated as *** , p < 0 . 0001; ** , p < 0 . 001; * , p < 0 . 01 by Student's t-test for differences compared with NKX2-5 WT . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01410 . 7554/eLife . 06942 . 015Figure 2—figure supplement 2 . Enrichment of known motifs in NKX2-5 peak subsets . Heat maps showing motifs significantly enriched in subsets of the NKX2-5 WT/Y191C ( A ) or NKX2-5 WT/∆HD ( B ) comparisons . TF-binding sites from the TRANSFAC ( * ) , JASPAR ( ** ) and UniPROBE ( Badis et al . , 2009 ) ( *** ) databases with a positive raw Clover score in ≥1 set are shown . The NKE is highlighted in a green box and ETS-like motifs are highlighted with black arrow . SRF† shows the secondary SRF motif , not the CArG box . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01510 . 7554/eLife . 06942 . 016Figure 2—figure supplement 3 . Enrichment of GO terms in NKX2-5 target subsets . Heat maps showing GO terms enriched in target gene subsets of the NKX2-5 WT/Y191C ( A ) or NKX2-5 WT/∆HD ( B ) comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 016 Surprisingly , in HL-1 cells , both NKX2-5Y191C and NKX2-5∆HD associated with a large number of loci ( 1149 and 792 peaks , respectively; Figure 2B; Supplementary file 1 ) with low false discovery rates ( Figure 1—figure supplement 4 ) . Even though NKX2-5Y191C cannot bind DNA , the most over-represented motifs were 5′AAGTGT3′ ( NKE ) , 5′GATAA3′ ( GATA ) , and 5′TGCCAA3′ ( NF1-like ) , exactly as for NKX2-5 WT fusion ( Supplementary file 1 ) . NKX2-5Y191C peaks also showed over-representation of the motif 5′TAATC3′ , which is similar to the binding sites of many non-NK-2 class HD proteins , including HOX proteins , as well as that of NK-2 proteins that lack HD tyrosine 54 , such as NKX1-2 ( Berger et al . , 2008 ) . NK-2 class proteins which do carry Y54 in their HDs , including NKX2-5 , also bind to this HOX-like site , albeit with a 10-fold-reduced affinity compared to that of the NKE ( Chen and Schwartz , 1995 ) . For NKX2-5∆HD peaks , there were many over-represented motifs , although none resembled known cardiac TF-binding sites ( Supplementary file 1 ) . The ability of NKX2-5Y191C and NKX2-5∆HD to bind a large number of targets was unexpected . Hence we sought to confirm using an independent approach that these mutants could interact with chromatin in vivo . We employed photoactivatable fluorescence correlation spectroscopy ( paFCS ) ( Kaur et al . , 2013 ) , measuring the diffusion dynamics of NKX2-5 WT and mutants within individual HL-1 nuclei ( Figure 2—figure supplement 1A ) . Expression vectors encoding NKX2-5 proteins fused to photoactivatable GFP ( paGFP ) were introduced by transfection . As for other TFs , NKX2-5 WT showed a biphasic behaviour , with a freely diffusing fraction similar to paGFP ( adjusted for molecular mass ) and a slower , chromatin-interacting fraction . NKX2-5 WT displayed slower diffusion ( dwell time τDslow = 121 ms ) than that mutants ( NKX2-5Y191C: τDslow = 64 ms; NKX2-5ΔHD: τDslow = 11 ms ) , which diffused more slowly than paGFP alone ( Figure 2C , Figure 2—figure supplement 1B-D ) . This confirmed that NKX2-5Y191C and NKX2-5ΔHD physically interact with chromatin ( Kaur et al . , 2013 ) and suggests that a gradation in chromatin tethering correlates with the apparent severity of the NKX2-5 mutation ( WT>Y191C>ΔHD>paGFP ) . We compared NKX2-5 WT and mutant peaks and adopted a simple nomenclature ( A , B , and C ) for peak subsets , where A is bound by NKX2-5 WT only , B is bound by both WT and a mutant , and C is bound only by a mutant ( Figure 2B ) . In both comparisons , the A sets predominated ( 81% and 53% of NKX2-5 WT peaks in the NKX2-5 WT/∆HD and NKX2-5 WT/Y191C comparisons , respectively ) , and their most enriched motif determined de novo was the NKE ( Figure 2D ) , demonstrating that , as anticipated , NKX2-5 mutants failed to bind to most WT targets , including Nppa . Therefore , the HD mutations result in a severe loss of function . However , a high proportion of mutant peaks overlapped with NKX2-5 WT peaks ( B sets; 36% of 792 NKX2-5∆HD peaks and 63% of 1149 NKX2-5Y191C peaks ) . To investigate the A- and B-set logic further , we used Clover ( Frith et al . , 2004 ) to calculate the enrichment of known TF motifs present in TRANSFAC ( BIOBASE ) and JASPAR databases ( Bryne et al . , 2008 ) in A and B sets separately ( Figure 2—figure supplement 2 ) . The NKE was significantly enriched in both A and B sets , showing that A and B sets contain direct NKX2-5 targets . A- and B-sets also exhibited distinct enrichment signatures for many known TF motifs , for example , those for cardiac kernel TFs GATA4 , HAND1 , and MEF2 were only enriched in the B-set of the NKX2-5 WT/∆HD comparison . Thus , both mutants can bind a subset of direct WT targets despite crippled DNA binding . This likely occurs via dimerisation with endogenous NKX2-5 WT and/or its cofactors ( see below and [Kasahara et al . , 2000] ) . Indeed , for NKX2-5∆HD , which completely lacks a DNA-binding domain , our data suggest that a protein:protein interface within NKX2-5 that is distinct from the HD can guide the mutant protein to NKX2-5 WT targets . At a GO term level , the A- and B-sets were also distinct ( Figure 2—figure supplement 3; Supplementary file 1 ) . For the Y191C/∆HD comparison , the A-set showed virtually no enrichment for GO terms and the B-set was primarily enriched in functions related to heart development . For the WT/∆HD comparison , the overlap in GO terms between the A-set and B-set was also very small , with the A-set being primarily enriched in terms related to cellular functions , such as actin filament organisation and transcription , and the B-set being enriched in functions related to heart and blood vessel development . This suggests that NKX2-5 mutants preferentially associate with genes involved in cardiac related processes . Based on the DamID and pFCA data above showing an association between NKX2-5 mutants and chromatin , we tested the functionality of NKX2-5 mutants in vivo , selecting Inhibitor of DNA binding 3 ( Id3 ) as a B set gene bound by NKX2-5 WT , Y191C , and ΔHD for detailed analysis ( Figure 3A ) . Id3 encodes a bHLH factor that represses cardiogenic differentiation in heart progenitor cells and is down-regulated as progenitors differentiate and levels of NKX2-5 increase ( Ding et al . , 2006 ) . Using qRT-PCR , we confirmed previous microarray data that Id3 is up-regulated in Nkx2-5 null mouse mutant hearts , similar to other cardiac progenitor genes ( Prall et al . , 2007 ) , whereas Nppa , a known directly activated NKX2-5 target , is down-regulated ( Figure 3—figure supplement 1 ) . The Id3 downstream region , when cloned into a minimal promoter/luciferase vector in antisense orientation and transfected into HEK 293 cells , stimulated transcription ∼18-fold . Activity was stimulated a further fourfold by SRF ( Figure 3—figure supplement 1 ) , which DamID indicated also bound to this region ( Figure 3A and below ) . NKX2-5 WT , Y191C , and ΔHD repressed both the basal and SRF-stimulated activity ( Figure 3B ) , suggesting that NKX2-5 mutants retain some WT functionality as repressors of Id3 . 10 . 7554/eLife . 06942 . 017Figure 3 . NKX2-5 mutant proteins retained partial functionality . ( A ) UCSC Genome Browser screen shot showing DamID transcription factor ( TF ) association in HL-1 cells with Inhibitor of DNA binding 3 ( Id3 ) . ( B ) Normalised Rluc activity in HEK 293 cells transfected with NKX2-5 and serum response factor ( SRF ) . Id3 DR was in the antisense orientation . p-values were calculated relative to controls ( ** p < 0 . 01; *** p < 0 . 001; **** p < 0 . 0001; ns = not significant ) or SRF alone ( †† p < 0 . 01; ††† p < 0 . 001; †††† p < 0 . 0001 ) . The quantity of vector is given in ng . ( C ) Schematic representation of ‘Materials and methods’ for the generation of embryonic stem ( ES ) cell lines for inducible expression of NKX2-5 WT and mutants . Cardiac differentiation was initiated in embryoid bodies ( EBs ) , followed by induction of cardiogenesis differentiation ( from day 4 ) on plates . ( D ) FACS quantification of PDGFRa/FLK1+ multipotent cardiovascular progenitors ( MCPs ) in NKX2-5 dox-inducible ES cell lines ( at day 5 ) . The mean of 6–7 independent experiments is shown . ( E , F ) FACS quantification of cTNT + cardiomyocytes ( D ) and CD31+ endothelial cells ( E ) at day 8 . ( G ) RT-PCR quantification of A- and B-set target genes in NKX2-5 WT or Y191C dox-inducible ES cell lines at day 5 in three independent experiments ( 24 hr post-induction ) . Results were normalised to expression in uninduced cells . p-values were calculated using a t-test ( * p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01710 . 7554/eLife . 06942 . 018Figure 3—figure supplement 1 . WT and mutant NKX2-5 targets identified by DamID in HL-1 cells . ( A ) Nppa and Id3 expression in Nkx2-5LacZ/+ heterozygous and Nkx2-5LacZ/LacZ null homozygous embryos relative to wild type ( +/+ ) at E8 . 5 . Expression was normalised to HPRT . ( B ) Normalised Renilla Luciferase ( Rluc ) activity in HEK 293 cells transfected with NKX2-5 and SRF . The 1270-bp Id3 DR region was cloned in sense or antisense orientations upstream of Rluc driven by a minimal TATA promoter . p-values were calculated relative to controls using an unpaired t-test ( *** p < 0 . 001; ns = not significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 01810 . 7554/eLife . 06942 . 019Figure 3—figure supplement 2 . ( A , B ) RT-PCR quantification of Nkx2-5 ORF and endogenous Nkx2-5 ( specifically detected using primers in the Nkx2-5 3′UTR , which is absent in the inducible construct ) in NKX2-5 WT/Y191C/∆HD dox-inducible ES cell lines at day 6 ( WT:∆HD comparison ( A ) ; 48 hr post-dox induction ) and day 5 ( WT:Y191C comparison ( B ) ; 24 hr post-induction ) . Results were normalised to expression in uninduced cells . p-values were calculated using a t-test ( * p < 0 . 05; ns = not significant ) . ( C ) Detection of induced NKX2-5 WT , NKX2-5Y191C , and NKX2-5∆HD by immunofluorescence microscopy in cytospinned cells at Day 6 ( 48 hr post-dox induction ) . Scale bars represent 50 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 019 To test functionality of NKX2-5 mutants in a developmental context , we generated mouse embryonic stem ( ES ) cells in which expression of exogenous NKX2-5 WT or mutants could be conditionally induced by doxycycline ( dox ) -dependent activation of a ubiquitous promoter ( Bondue et al . , 2008 , 2011 ) . Following induction of cardiogenesis , NKX2-5 WT , Y191C , and ∆HD proteins were activated in most or all embryoid body ( EB ) cells with dox at day 4 , 1 day before , endogenous NKX2-5 appears in a minority of cells within cardiogenic clusters ( Figure 3C ) . Comparable levels of NKX2-5 WT and mutant mRNAs and protein were detected 1–2 days post-induction ( Figure 3—figure supplement 2 ) . Induced NKX2-5 WT strongly repressed formation of FLK1+/PDGFRα+ multipotent cardiovascular progenitors at day 5 ( Figure 3D ) , and of cardiomyocytes ( cTNT+ ) and endothelial cells ( CD31+ ) at day 8 ( Figure 3E-F ) . Repression of cardiac lineages by NKX2-5 in ES cells is consistent with its demonstrated early role as a negative feedback regulator of cardiac induction and second heart field gene expression ( Prall et al . , 2007 ) , and as an inhibitor of reprogramming of cardiac fibroblasts to a cardiomyocyte fate ( Ieda et al . , 2010 ) . NKX25∆HD was nuclear in only 15% of cells at day 6 ( likely due to deletion of the nuclear localisation signal within the HD; Figure 3—figure supplement 2C ) and had no effect on cardiac or endothelial cell differentiation ( Figure 3E-F ) and so was not considered further . NKX2-5Y191C was nuclear and , while showing no effects on myocardial progenitors or endothelial cells in three independent experiments , showed a trend towards inhibition of cardiomyocyte numbers at day 8 ( ∼44%; p = 0 . 079 ) , suggesting that it also retained some repressive activity in this assay . To gain deeper insights into the underlying effects , we therefore measured alterations in gene expression in the ES cell lines at day 5–6 in three independent experiments , testing a selection of A and B-set genes derived from the NKX2-5WT:Y191C comparison . The majority of both A ( 8/12 ) and B ( 12/13 ) set genes tested were modulated >1 . 5 fold by NKX2-5 WT ( total 20/25 ) , providing strong evidence that DamID selects NKX2-5 WT developmental targets . Within B-set genes , most were repressed by NKX2-5 WT , consistent with the repression of cardiogenesis by NKX2-5 WT at early stages of cardiac lineage specification ( Prall et al . , 2007 ) . A subset of A and B set genes was also regulated by NKX2-5Y191C ( A-set: 3/12; B-set: 4/13; Figure 3G ) , despite no change in early myocardial progenitors in the NKX2-5Y191C line , with regulation by Y191C being in the same direction as for NKX2-5 WT . As examples , Actc1 , Nkx2-5 , Hand2 , and Gata4 were repressed by both NKX2-5 WT and NKX2-5Y191C , while Nppa and Tbx3 were activated by both . These data show that while NKX2-5Y191C loses its ability to bind and regulate most normal NKX2-5 targets , it nonetheless retains some of the regulatory capabilities of NKX2-5 WT . Our results showing that the NKE is enriched in both B-sets suggest that NKX2-5 mutants can bind targets by dimerising with NKX2-5 WT . While dimerisation to NKX2-5 WT has been demonstrated for NKX2-5Y191C ( Kasahara and Benson , 2004 ) , it was surprising for NKX2-5ΔHD because the HD has been previously described as being essential for NKX2-5 homodimerisation ( Elliott et al . , 2010 ) . Transfected NKX2-5∆HD was largely cytoplasmic in CV-1 cells ( Figure 4A ) , which lack endogenous NKX2-5 . This is likely a consequence of the deletion of the NKX2-5 nuclear localisation signal located in the amino terminus of the HD ( Kasahara and Izumo , 1999 ) . However , NKX2-5∆HD became nuclear when co-expressed with NKX2-5 WT , which itself was nuclear when transfected alone ( Figure 4A ) . Cofactors GATA4 , TBX5 , and TBX20 , which have been reported to bind NKX2-5 WT via the HD also induced nuclear translocation of NKX2-5∆HD ( Figure 4—figure supplement 1 ) . In HL-1 cells , which express endogenous NKX2-5 and its cardiac cofactors , transfected NKX2-5∆HD was predominantly nuclear although some protein was present in the cytoplasm ( Figure 4B ) . These data demonstrate that NKX2-5 WT and its cofactors are able to carry NKX2-5∆HD into the nucleus via an unknown protein:protein interaction not involving the HD . 10 . 7554/eLife . 06942 . 020Figure 4 . The YRD is essential for interaction between NKX2-5 and NKX2-5ΔHD . ( A ) Intracellular localisation of V5-tagged NKX2-5ΔHD and HA-tagged NKX2-5 WT . CV-1 cells were transfected with NKX2-5ΔHD only ( top row ) , NKX2-5 WT only ( middle row ) , or both ( bottom row ) . Solid and dashed lines highlight the cellular and nuclear boundaries , respectively . ( B ) Intracellular localisation of transfected V5-tagged NKX2-5 WT ( top row ) and NKX2-5ΔHD ( bottom row ) in HL-1 cells . ( C ) NKX2-5 homo- and hetero-dimerisation measured by Rluc-PCA in HEK 293T cells . F1 and F2 represent the N- and C-terminal Rluc fragments , respectively . Data are represented as mean of the normalised luciferase activity ± SEM . Significance was calculated using an unpaired t-test ( **** p < 0 . 0001 ) . ( D ) NKX2-5 homo- and hetero-dimerisation measured by processed spectral Förster Resonance Energy Transfer ( psFRET ) . Cerulean and Venus represent the donor and acceptor molecules . FRET efficiency is represented as mean ± SEM . p-values were calculated using a t-test between each pair and its appropriate controls . Significance is as follows: † for p-value < 0 . 05; ! for p-value < 0 . 001 ; # for p-value < 0 . 0001; values are given if p-value >0 . 05 . ( E ) Representative psFRET images ( false coloured using fire look up table ) used in ( D ) . ( F ) Yeast-two-hybrid assay . Proteins were fused to Gal4-activation and DNA-binding domains . Positive signs ( + ) show interaction as growth on selective medium from three independent experiments ( nd = not determined ) . FRET , Förster Resonance Energy Transfer . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 02010 . 7554/eLife . 06942 . 021Figure 4—figure supplement 1 . NKX2-5∆HD interacts with NKX2-5 cardiac cofactors and ETS-factors in vivo . Expression of HA-tagged TBX5 , TBX20 , and GATA4 ( red ) induces nuclear translocation of V5-tagged NKX2-5DHD ( green ) in CV-1 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 02110 . 7554/eLife . 06942 . 022Figure 4—figure supplement 2 . NKX2-5 dimerisation with TBX5 , TBX20 , and GATA4 measured by Rluc-PCA . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 02210 . 7554/eLife . 06942 . 023Figure 4—figure supplement 3 . Density of NKX2-5 WT , NKX2-5∆HD , and NKX2-5Y191C peak sets and of probes relative to the TSS . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 023 To demonstrate the presence of this interface independently , we used a Renilla luciferase protein fragment complementation assay ( Rluc-PCA ) ( Stefan et al . , 2007 ) with a weakened Cytomegalovirus ( CMV ) promoter to avoid protein aggregation . This assay confirmed NKX2-5 WT homo-dimerisation , hetero-dimerisation between NKX2-5 WT and NKX2-5∆HD , and homo-dimerisation of NKX2-5∆HD ( Figure 4C ) . It also confirmed interactions between NKX2-5∆HD and GATA4 , TBX5 , and TBX20 ( Figure 4—figure supplement 2 ) . We further employed processed spectral Fluorescence Resonance Energy Transfer ( psFRET ) ( Chen et al . , 2007 ) , detecting robust Fluorescence Resonance Energy Transfer ( FRET ) from NKX2-5 WT/WT homodimers or WT/∆HD heterodimers in the nucleus and from ∆HD/ΔHD homodimers in the cytoplasm ( Figure 4D–E ) , the latter result demonstrating that dimerisation was independent of chromatin binding . To define the dimerisation domain , we used the Y2H assay ( Figure 4F ) . Full-length NKX2-5 interacted with peptides carrying the NKX2-5 HD ( aa137-196 ) or C-terminal region ( aa198-318 ) but not with the N-terminal region ( aa1-135 ) . The C-terminal peptide could also interact with itself . The conserved tyrosine-rich domain ( YRD ) , present within the C-terminal region ( Figure 1A ) , was essential for the interaction between full-length NKX2-5 and the C-terminus , being blocked by the NKX2-5YRDY−A mutation in which all 9 tyrosines of the YRD are mutated to alanine . We have previously shown that NKX2-5YRDY−A does not affect DNA binding , although it has a strong dominant-negative activity leading to lethal CHD-like phenotypes in high-level heterozygous ES cell chimaeras , and when placed over the null allele creates a phenocopy of NKX2-5 loss-of-function ( Elliott et al . , 2006 ) . Mutation of other conserved domains within the C-terminus ( NK2-specific domain; Nkx2-5 box ) had no affect on the interaction ( Figure 4F ) . While the NKX2-5 HD is known to be a homophilic interaction domain ( Elliott et al . , 2010 ) , our Y2H data identify the YRD as a novel dimerisation interface within NKX2-5 . This YRD-dependent interaction , demonstrated in both yeast and mammalian cells , provides the mechanism for heterodimerisation of NKX2-5 WT with NKX2-5∆HD , as well as homodimerisation of NKX2-5∆HD . When comparing NKX2-5 WT and mutant peaks , the presence of C sets revealed that mutant proteins bound a set of unique targets ( Figure 2B ) that we call ‘off-targets’ potentially reflecting a gain-of-function . Proportionally , off-target sets were much larger for the more severe NKX2-5∆HD than for NKX2-5Y191C ( 64% of total NKX2-5∆HD targets [504 peaks]; 37% of NKX2-5Y191C targets [427 peaks] ) . Off-target peaks had specific characteristics . For example , for both NKX2-5∆HD and Y191C mutants , C-sets peaks were uni-modal centred at the transcription start site , while A- and B-set peaks displayed a bimodal distribution around the transcription start site , as for total NKX2-5 WT peaks ( Figure 4—figure supplement 3 ) . Testing NKX2-5 mutant off-target genes for over-representation of GO terms , we found that both off-target sets were enriched in GO terms chromatin organisation , macromolecular complex assembly , and cell cycle , but not cardiac terms ( Figure 2—figure supplement 3 ) . These results indicated that off-targets are selected via a specific logic , most likely via interaction with non-cardiac-restricted cofactors . Using Clover ( Frith et al . , 2004 ) to calculate the enrichment of all known TF motifs present in both C sets ( Figure 2—figure supplement 4 ) , we noted that off-target peak sets for NKX2-5∆HD and Y191C were not enriched in the NKE , nor in fact DNA-binding sites of virtually all other known cardiac TFs ( Figure 2—figure supplement 4 ) . The most over-represented motif for NKX2-5Y191C was 5′TAAT3′ ( Figure 2D ) , which is similar to the HOX-like site over-represented in NKX2-5 WT targets mentioned above . Our DamID findings on NKX2-5Y191C are consistent with in vitro data showing that the affinity of an analogous Drosophila NK2 class HD mutant ( vnd/NK2 Y54M ) for the NKE is reduced by 10-fold , while its affinity for the TAAT core was unchanged ( Weiler et al . , 1998 ) . For both mutants , off-target peak sets were enriched in other motifs specific to many broadly expressed TFs that are not known to be part of the cardiac GRN ( Figure 2—figure supplement 4 ) . We focused on the enrichment of TF-binding sites in the C-set of NKX2-5∆HD , based on the hypothesis that DNA-binding cofactors of NKX2-5 guide NKX2-5∆HD to its off-targets . The motifs for E−26 transformation-specific ( ETS ) TFs occurred most frequently in NKX2-5∆HD C-set peaks ( 42% ) and were significantly enriched compared to random peaks , suggesting that ETS factors could play a role in guiding NKX2-5∆HD to its off-targets . ETS factors are broadly expressed TFs that are activated by phosphorylation downstream of tyrosine kinase receptors . They are important for extracellular signal-gating in specification of cardiac progenitors in Ciona ( Davidson et al . , 2006 ) , eve-positive extra-cardiac progenitors in Drosophila ( Halfon et al . , 2000 ) , and for specification of vascular endothelial cells and Gata4 expression in endocardial cushions in mammals ( Schachterle et al . , 2011 ) . However , the broader roles of ETS factors in mammalian cardiogenesis are uncharted . Many ETS factors are expressed in embryonic hearts ( Schachterle et al . , 2011 ) , and we selected ternary complex factors ELK1 and ELK4/SAP-1/TCF-1 , whose motifs occurred in 26% of C-set peaks , for further analysis . We speculated that these ubiquitous TFs are cofactors of NKX2-5 and , in the absence of compelling cardiac specificity conferred by the HD , guide NKX2-5∆HD to a subset of off-targets . In the next set of experiments , we set out to establish this principle . Antibodies specific for total ELK1 or its phosphorylated form ( pELK1 ) showed that in HL-1 cells ELK1 was mostly cytoplasmic and unphosphorylated , although in a minority of cells ELK1 was phosphorylated and nuclear , where it co-localised with endogenous NKX2-5 ( Figure 5A , B ) . In CV-1 cells , transfected ELK1 was also variably partitioned between the nucleus and cytoplasm as in transfected neurons ( Lavaur et al . , 2007 ) . When phosphorylated and nuclear , pELK1 guided co-transfected NKX2-5∆HD into the nucleus of CV-1 cells . Conversely , when unphosphorylated , it retained co-expressed NKX2-5 WT within cytoplasmic inclusions ( Figure 5C; Figure 5—figure supplement 1 ) . These data suggest that ELKs and NKX2-5 are interacting cofactors . 10 . 7554/eLife . 06942 . 024Figure 5 . NKX2-5 WT and NKX2-5ΔHD interact with ELK1/4 . ( A ) Nuclear expression of endogenous NKX2-5 and pELK1 in HL-1 cells . Scale bar represents 10 µm . ( B ) Co-localisation of transfected V5-tagged NKX2-5ΔHD and endogenous total ELK1 in HL-1 cells . Scale bar represents 10 µm . ( C ) Co-localisation of transfected HA-tagged ELK1 and V5-tagged NKX2-5ΔHD or NKX2-5 WT in CV-1 cells . Arrowheads and arrows show cells with cytoplasmic and nuclear NKX2-5 staining , respectively . ( D ) Interactions between NKX2-5 , NKX2-5ΔHD , or SRF and ELK4 , ELK1 , or ELK1Y159A measured by the Rluc-PCA in HEK 293T cells . ( E ) Interactions between ELK1 and NKX2-5 or NKX2-5ΔHD measured by psFRET in CV-1 cells . p-values: † <0 . 05; ! <0 . 001; values are given if p-value>0 . 05 . ( F ) Representative psFRET images used in ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 02410 . 7554/eLife . 06942 . 025Figure 5—figure supplement 1 . HA-tagged ELK1 expression ( green ) induces nuclear translocation of V5-tagged NKX2-5∆HD ( red ) in CV-1 cells . Cells were co-transfected with NKX2-5∆HD and ELK1 ( top ) ; NKX2-5∆HD and ELK1Y158A ( middle ) ; or NKX2-5 WT and ELK1Y158A ( lower ) . Anti-phosphorylated Elk1 antibody was used in top and middle panels . The arrowheads and arrows show cells with cytoplasmic and nuclear NKX2-5 staining , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 02510 . 7554/eLife . 06942 . 026Figure 5—figure supplement 2 . ( A ) Interactions between TBX5 and ELK1/ELK4 measured by the Rluc-PCA in HEK 293T cells ( ** p < 0 . 01; *** p < 0 . 001; ****p < 0 . 0001 ) . ( B ) NKX2-5 dimerisation with ETS factors measured by Rluc-PCA in HEK 293T cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 026 Using Rluc-PCA in HEK 293 cells , we confirmed that both ELK1 and ELK4 interacted with NKX2-5 WT and NKX2-5∆HD , as well as SRF and TBX5 ( Figure 5D , Figure 5—figure supplement 2 ) . NKX2-5 WT also interacted with ETS-family members ELK3 , ETS1 , ERF , and GABPα ( Figure 5—figure supplement 2 ) . psFRET confirmed the NKX2-5/ELK1 interaction in both the nucleus and cytoplasm in CV-1 cells ( Figure 5E–F ) , demonstrating that the interaction was independent of ELK1 phosphorylation and chromatin binding . SRF is a well-described cofactor of both NKX2-5 and ELK1 , albeit in different contexts ( Treisman , 1994 ) . SRF is an effector of several signalling pathways and plays a specific role during early heart formation via its combinatorial action with NKX2-5 and GATA4 on target genes . To test whether the NKX2-5/ELK1 interaction was mediated by SRF , we utilised Rluc-PCA . The ELK1 B-box mutation Y159A , which disrupts the interaction between ELK1 and SRF ( Ling et al . , 1998 ) , still interacted with NKX2-5 ( Figure 5D ) and , as for ELK1 WT , could induce nuclear translocation of NKX2-5∆HD or retain NKX2-5 WT in cytoplasmic inclusions ( Figure 5—figure supplement 1 ) . Thus , the ELK/NKX2-5 interaction does not require SRF . In the Y2H assay , we confirmed an interaction between ELK1 and both NKX2-5 WT and NKX2-5∆HD ( Figure 4F ) . ELK1 did not interact with the HD , and its interaction with NKX2-5 was blocked by the YRDY−A mutation , demonstrating that the YRD , in addition to being a haemophilic interaction domain contributing to NKX2-5 homodimerisation , is a protein:protein interface essential for the interaction between NKX2-5 and ELK1 . Our results are consistent with previous data showing that a tyrosine-rich region within NK2 class homeo-protein NKX3-1 mediates interaction with prostate-derived ETS factor ( Chen and Bieberich , 2005 ) . The data suggest that ETS factors including ELK1/4 are functional cofactors of NKX2-5 WT during cardiogenesis . In order to confirm binding of ELK1/4 to NKX2-5∆HD off-targets , we identified ELK1 and ELK4 target peaks in HL-1 cells using DamID . For ELK1 and ELK4 , we generated robust data for both N- and C-terminal Dam fusions . All data sets showed low false discovery rates ( Figure 1—figure supplement 4 ) and significant overlap between N- and C-terminal Dam fusions ( p < 0 . 001 ) , resulting in 1217 and 875 overlapping peaks for ELK1 and ELK4 fusions , respectively ( Figure 6A; Supplementary file 1 ) . De novo motif discovery identified the known DNA-binding site for both ELK1 and ELK4 as the only over-represented motif ( Figure 6B; Supplementary file 1 ) . The overlap between ELK1 and ELK4 peaks ( p < 0 . 001 ) was high ( 48% of ELK1 and 67% of ELK4 peaks; Figure 6A ) . 10 . 7554/eLife . 06942 . 027Figure 6 . ELK1 and ELK4 co-occupy NKX2-5ΔHD off-targets in HL-1 cells . ( A ) Overlapping peaks between ELK1 , ELK4 , and SRF as shown by proportional Venn diagram . ( B ) Top binding motifs discovered de novo with Weeder or Trawler in ELK1 , ELK4 , and SRF peaks . The TRANSFAC SRF motif and the ELK1/4 motifs determined by ( Wei et al . , 2010 ) in vitro are shown on the right . ( C ) Normalised median expression of ELK1 , ELK4 , and SRF target genes in 91 murine cell types , including the heart in red ( data collected from BioGPS ) . ( D ) Density of ELK1 , ELK4 , and SRF peaks and probes relative to the TSS . ( E ) Normalised Rluc activity in HEK 293 cells . NKX2-5 WT and mutants were co-transfected with a pGL4 . 24 luciferase reporter under the control of the Rad50- or Snai2-promoters . p-values < 0 . 01 calculated relative to control are denoted by ** . # shows significant difference ( p < 0 . 01 ) . ( F ) RT-PCR quantification of NKX2-5Y191C off-target genes in NKX2-5 WT or Y191C dox-inducible ES cell lines at day 5 ( 24 hr post-induction ) . Results were normalised to expression in uninduced cells . p-values were calculated using a t-test ( * p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 027 ELK1/4 peaks could be assigned to 1423 and 1051 unique target genes , respectively . GO analysis showed highest over-representation of terms cytoskeletal organisation and RNA processing ( Supplementary file 1 ) , suggesting that , overall , ELK1/4 regulate many generic cellular functions . ELK1/4 target gene median expression was low in heart and highest in mast cells and macrophages ( Figure 6C ) . In contrast to NKX2-5 WT peaks , ELK1/4 peaks showed unimodal enrichment centred across the transcription start site ( TSS ) , identical to those of NKX2-5∆HD off-targets ( Figure 6D , S6C ) . These data highlight the architectural differences between the majority of NKX2-5 WT peaks and those bound by ELK1/4 and NKX2-5∆HD . Having established a robust target set for ELK1/4 , we determined how many peaks within the NKX2-5∆HD C-set were actually occupied by ELK1/4 in HL-1 cells . Of the 506 NKX2-5∆HD C-set peaks , 72 ( 14 . 2% ) were bound by ELK1/4 , representing more than half of the 26% of this C-set predicted by Clover to carry an ELK1/4 binding motif . By comparison , <1% of NKX2-5∆HD C-set peaks overlapped with randomly generated peak sets . In the case of NKX2-5Y191C C-set peaks , 6 . 5% were bound by ELK1/4 . These data provide strong support for our hypothesis that ELK1/4 directs NKX2-5 mutants to a subset of off-targets . To explore the significance of ETS co-occupancy for selection of off-targets , we used DamID to determine the targets of an NKX2-5 mutant carrying tyrosine–alanine ( Y-A ) substitutions in the YRD because this mutant does not interact with ELK1/4 . Comparisons between NKX2-5YRDY−A and NKX2-5 WT targets revealed robust B and C-set ( 650 and 380 peaks , respectively; Figure 2B and Supplementary file 1 ) . However , in the C-set , only 3 . 2% of peaks were occupied by ELKs . This lends further support for the notion that NKX2-5∆HD binds a subset of its off-targets via cofactors interacting with the YRD . The proportionally large size of the NKX2-5YRDY−A C-set also suggests that this mutation is comparable in severity to NKX2-5Y191C ( Figure 2B ) , supporting the genetic evidence ( Elliott et al . , 2006 ) . Dysregulation of C-set genes may contribute to CHD . To test whether NKX2-5 mutants can influence the expression of their off-targets , we cloned the promoter regions of Rad50 and Snai2 , which bind to NKX2-5∆HD , NKX2-5Y191C , ELK1 , and ELK4 , but not to NKX2-5 WT , into a luciferase reporter . In HEK 293 cells , both promoters were stimulated modestly by NKX2-5 WT but repressed by ELK1 or ELK4 ( Figure 6E ) . NKX2-5∆HD had no activity alone , but significantly enhanced repression by ELK1 or ELK4 , an activity not displayed by NKX2-5 WT . NKX2-5Y191C had a similar activity , albeit weaker . These data show that NKX2-5∆HD indirect binding to off-targets can modify gene expression , at least when over-expressed . We next tested whether NKX2-5 mutants could alter off-target gene expression in the inducible ES cell lines described above . For off-targets from the NKX2-5WT:Y191C comparison , NKX2-5 WT modulated the expression of 6/22 C-set genes tested by > 1 . 5-fold ( Figure 6F ) , although the effect on 2 of these ( Dock4 and Sparc ) could be accounted for by NKX2-5 WT binding to a distinct DamID peak , and consistently , NKX2-5 WT and Y191C had opposite effects on these two genes . Thus , NKX2-5 WT influenced a minority of NKX2-5Y191C off-targets . NKX2-5Y191C repressed Dock4 and activated 7/22 other C-set genes with 6 of these not regulated by NKX2-5 WT at the >1 . 5 fold significance threshold . Other genes showed a similar trend . Thus , NKX2-5Y191C could regulate ≥36% of its off-targets in this system . As shown above , ETS motifs were enriched in NKX2-5ΔHD off-targets . They were also present although not significantly enriched in total NKX2-5 WT targets . Enrichment was also low in previously reported genome-wide NKX2-5 targets determined using ChIP-seq , but higher in targets of GATA4 , SRF , and TBX5 ( He et al . , 2011 ) . These and our DamID data suggest that ELK1/4 play a role in the normal cardiac GRN . To explore this further , we examined the overlap between ELK1/4 and NKX2-5 WT target genes . Of 1490 NKX2-5 WT target genes , 21% ( p-value < 2 . 2e-16 , Fisher exact test ) were also bound by ELK1 or ELK4 ( Figure 7A ) , and in 52% of these ( including Nkx2-5 , Actc1 , Id2 , and Id3 ) , the NKX2-5 and ELK1/ELK4 peaks overlapped . Overlaps between SRF and ELK1 targets , and between SRF and NKX2-5 targets have been documented previously , although not in the same cell type ( Boros et al . , 2009a; He et al . , 2011; Schlesinger et al . , 2011 ) . To complete our comparisons , we therefore determined SRF targets in HL-1 cells using DamID . 10 . 7554/eLife . 06942 . 028Figure 7 . ELK1 and ELK4 are embedded in the cardiac gene regulatory network . ( A ) Overlapping target genes between NKX2-5 , ELK1/ELK4 , and SRF visualised using Cytoscape ( spring-embed layout ) . Cluster size is proportional to gene numbers . For each cluster , rectangles indicate the normalised median expression of target genes in 91 murine cell types ( BioGPS ) . Cell types are ordered by increasing expression values from left to right . Top 3 over-represented DAVID Gene Ontology ( GO ) annotations are indicated for each cluster . ( B ) Regulatory interactions between Elk1/3/4 and cardiac TFs from DamID experiments and published data sets ( network constructed with BioTapestry ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06942 . 028 For SRF , we generated data for both N- and C-terminal Dam fusions . SRF peaks showed low false discovery rates ( Figure 1—figure supplement 4 ) and significant overlaps between N- and C-terminal Dam fusions ( p < 0 . 001 ) , resulting in 1214 high-confidence peaks ( Figure 6A; Supplementary file 1 ) . These peaks could be assigned to 1314 unique target genes . De novo motif discovery identified the known SRF motif ( CArG box ) as the only over-represented motif ( Figure 6B; Supplementary file 1 ) and present in 26% of targets , supporting the high specificity of DamID . We note that a previous ChIP study in HL-1 cells reported SRF targets with only a very small fraction containing the CArG box ( Schlesinger et al . , 2011 ) . For SRF target genes , GO analysis showed high over-representation of cardiac ( e . g . , heart development ) and generic ( cytoskeletal organisation and RNA processing ) terms ( Supplementary file 1 ) . SRF target median expression was highest in heart and skeletal muscle ( Figure 6C ) , consistent with its known functions . Like NKX2-5 WT , SRF peaks displayed a bimodal distribution around the TSS ( Figure 6D ) . The overlap between ELK1/4 and SRF peaks was low ( 15 and 22% of ELK1 and ELK4 peaks , respectively; and 20% of SRF peaks; Figure 6A ) , consistent with previous ChIP studies in serum-starved HeLa cells suggesting that most ELK1 targets are not co-bound by SRF ( Boros et al . , 2009a ) . In HL-1 cells , targets co-occupied by ELK1/4 and SRF nonetheless included the Fos gene—the defining target of ternary SRF-ELK complexes ( Treisman , 1994 ) . Targets bound uniquely by ELK1/4 or SRF showed over-representation of GO terms related to catabolic and metabolic processes , and as for total ELK1/4 targets , were only modestly expressed in heart ( Figure 7A; Supplementary file 1 ) . In contrast , targets bound by NKX2-5 only , NKX2-5 , and SRF , or NKX2-5 , SRF , and ELK1/4 were highly expressed in heart . NKX2-5 unique targets were enriched in GO terms anion transport and regulation of cell motion , while targets co-bound by NKX2-5 and SRF were enriched in GO terms related to heart contraction . Co-targets of all 3 factors ( NKX2-5 , SRF , ELK1/4 ) or of NKX2-5 and ELK1/4 were most over-represented in GO terms heart development , blood vessel morphogenesis , and cytoskeleton organisation . These data show that , overall , ELK1/4 , SRF , and NKX2-5 regulate many generic cellular functions , while NKX2-5 and SRF also have more executive roles in regulating heart development and contraction . Furthermore , ELK1/4 appear to control very specific subsets of the cardiac GRN . By integrating the targets of all WT TFs profiled with those of other cardiac TFs previously published , we found that Elk1/4 were embedded in the developmental cardiac GRN with many cross and feedback connections ( Figure 7B ) , suggesting that ELK1/4 play a significant role in normal and disease cardiac GRN logic .
Here , we used DamID in HL-1 cells to probe the cardiac GRN by identifying target genes for kernel TFs NKX2-5 and SRF , and for the MAP kinase signalling-dependent cardiac transcriptional cofactors ELK1 and ELK4 . To begin to understand the mechanism of CHD , we adopted the DamID method and identified targets of NKX2-5 mutants that mimic those found in CHD . These are the first functional genomics studies to probe CHD , providing new insights into the structure and function of NKX2-5 , and network regulation in normal heart development and CHD . Previous studies comparing DamID and ChIP experiments performed in Drosophila reported ‘a high degree of overlap’ ( Moorman et al . , 2006; Negre et al . , 2006; Tolhuis et al . , 2006; van Bemmel et al . , 2010; Yin et al . , 2011 ) . We compared our NKX2-5 peak set with NKX2-5 ChIPseq peaks generated previously in HL-1 cells ( He et al . , 2011 ) and adult hearts ( van den Boogaard et al . , 2012 ) , restricting our analyses to genomic regions covered by the promoter microarrays ( Figure 1—figure supplement 9 ) . We found that the overlap between all three data sets was low—specifically 18% and 6% DamID peak overlap , respectively . This low overlap highlights the problems of comparing data from different platforms and different laboratories , and warrants deeper analysis . We note that only DamID was performed with 3–4 replicates per experiment . Furthermore , de novo motif discovery identified the known high-affinity NKE exclusively within NKX2-5 DamID peaks . Both of the published studies identified a similar but variant NKE using ChIPseq . Cardiac- and muscle-related GO terms were more highly enriched in NKX2-5 target genes determined by DamID compared to published ChIPseq peaks from adult hearts included in Figure 1—figure supplement 9 ( van den Boogaard et al . , 2012 ) , and no such GO term were detected in data from HL-1 cells ( He et al . , 2011 ) . These results provide strong validation for the DamID method . Our study using DamID confirms for the first time at a genome-wide level in vivo , the prevailing view that severe mutations in cardiac kernel TF ( such as NKX2-5∆HD , Y191C , and YRDY−A ) fail to bind the majority of WT targets ( A sets ) even in the presence of normal levels of NKX2-5 WT . Loss-of-binding of mutant proteins will be associated with alterations in the level of expression of many genes in the cardiac network through haploinsufficiency , which will be a major contributor to the structural and functional heart defects found in human CHD patients . Surprisingly , however , NKX2-5 mutants still recognised a large number of genomic sites and some of these were normally bound by NKX2-5 WT ( B-sets ) . B-sets included direct and indirect NKX2-5 WT target genes and were enriched in heart- and muscle-related GO terms . In the ES cell system , NKX2-5 mutant proteins retained WT-like functionality on some B-set genes . We propose that B-sets are recognised by mutants via dimerisation with cofactors or with NKX2-5 WT . The NKX2-5 HD is a recognised interface for NKX2-5 homo-dimerisation and interactions with cofactors , such as GATA4 , HAND1 , TBX5/20 , and MEF2 . NKX2-5∆HD lacks the HD and embedded nuclear localisation domain and represents a severe mutation that has lost the ability to bind to the majority ( 81% ) of normal NKX2-5 WT targets ( A-set ) . However , even this severe mutant retained binding to a subset of direct and indirect NKX2-5 WT targets ( B-set ) . NKX2-5∆HD binding to B-set peaks could only occur through heterodimerisation with NKX2-5 WT or cofactors . We have shown using a number of approaches that NKX2-5∆HD retained its ability to interact with NKX2-5 WT , GATA4 , and TBX5/20 , and we identified the YRD located within the C-terminus of NKX2-5 as a novel interface that collaborates with the HD to support these interactions . Our characterisation of the YRD as a novel homophilic and heterophilic interaction domain provides the molecular mechanism that explains dimerisation of NKX2-5∆HD to NKX2-5 WT or its cofactors in the absence of the HD . We predict that the HD and YRD act supportively and synergistically in the assembly of macromolecular TF complexes . The YRD is an ancient domain that has coevolved with the NK2-class HD and is essential for NKX2-5 function in early mouse embryogenesis ( Elliott et al . , 2006 ) . In humans , half of NKX2-5 mutations associated with CHD are outside of the HD , and three frame-shift mutations located within the YRD have been identified ( Benson et al . , 1999; Gutierrez-Roelens et al . , 2002; Ikeda et al . , 2002 ) . Both NKX2-5 mutants studied also recognised a large number of targets uniquely ( C-set ) , which we term off-targets . The relative size of the C-set within the total mutant target set ( C-set/B-set ratio ) correlates with the location and predicted severity of NKX2-5 mutations ( ∆HD>Y191C≥ YRDY−A ) , and thus B-set and C-set size and content may prove to be valuable signatures for understanding the relationship between genotype , phenotype , and clinical outcomes in CHD . NKX2-5∆HD off-targets were associated with GO terms chromatin organisation and cell cycle , and peaks were enriched in the binding sites for ETS family and many other broadly expressed TFs , attesting to an underlying logic in their selection . A high percentage of NKX2-5 mutant off-targets ( 14 . 2% for NKX2-5∆HD and 6 . 5% for NKX2-5Y191C ) were occupied by ELK1/4 in HL-1 cells , compared to <1% when randomly generated peaks were overlapped . Furthermore , only 3% of C-set peaks of the NKX2-5YRDY−A mutant were bound by ELK1/4 . Although ETS factors have previously been predicted to be involved in heart development based on motif enrichment in predicted cardiac enhancers in fly and humans ( Pham et al . , 2007; Narlikar et al . , 2010; Jin et al . , 2013 ) , this is the first evidence showing a direct interaction with cardiac TFs . Studies of the Ciona intestinalis cardiac GRN demonstrated that ETS family gene Ets1/2 is activated by FGF/MAP Kinase signalling , which specifies the identity of founder cells of the cardiac lineage ( Davidson et al . , 2006 ) . Interestingly , most overrepresented in candidate Ets1/2 target genesis was the motif 5′ATTA3′ , ( Davidson et al . , 2006; Woznica et al . , 2012 ) , which is similar to the HOX-like motif overrepresented in NKX2-5Y191C off-target peaks in HL-1 cells , potentially indicating a conserved FGF/MAP Kinase-activated synergy between ETS and HD factors at the earliest stages of cardiac lineage specification . ELK1 and ELK4 interacted with NKX2-5 WT and NKX2-5∆HD through the YRD , and this interaction was sufficiently strong for pELK1 to transport NKX2-5∆HD into the nucleus when co-expressed and for unphosphorylated ELK1 to tether NKX2-5 WT in cytoplasmic inclusions . This interaction provides a mechanism for how severe NKX2-5 mutants lacking a functional HD can be directed to a sub-set of off-targets via interaction with broadly expressed TFs . Severe NKX2-5 mutants may be drawn to off-targets due to the breakdown of compelling cardiac specificity on true targets . We have also shown that ETS factors ELK1 and ELK4 are normal cofactors of NKX2-5 , embedded in the cardiac GRN kernel . The cardiac GRN kernel is traditionally conceptualised as a set of cardiac-restricted TFs that are critical for defining organ territories and imposing organ specificity to GRN logic . However , the cardiac kernel is likely to have evolved in the context of a host of ancient TFs controlling ubiquitous cellular processes such as metabolism , cell cycle , chromatin dynamics , and cytoskeleton . Furthermore , it is well known that TFs are amongst the targets of the complex paracrine signalling pathways that underpin pattern formation , tissue specification , and differentiation in metazoans . Our genome-wide studies suggest that ELK1/4 interact with , regulate , and are regulated by , cardiac TFs , and are therefore recursively wired into the cardiac kernel at a high level . Other studies show that organ-restricted NK-2 class HD TFs collaborate with ETS factors on specific target genes ( Chen and Bieberich , 2005; Lin et al . , 2006 ) or provide signal gating for decisions affecting lineage fate within developmental fields ( Halfon et al . , 2000; Hollenhorst et al . , 2011 ) . ELK1/4 bind and potentially regulate a large number of genes in HL-1 cells involved in ubiquitous cellular processes , most prominently metabolism . However , smaller gene sets , most significantly those encoding heart developmental and cytoskeletal genes , were co-bound by NKX2-5 and ELKs , SRF and ELKs , or ELK1/4 , NKX2-5 , and SRF . One interpretation is that subsets of ELK1/4 targets , potentially those at critical hubs , have been drawn into the cardiac GRN to exert strategic ( signal-gated ) and fine network control over its outputs . Considering the implications of this and related studies more broadly , network links between cardiac and canonical signal-gated TFs are likely to be much greater than currently appreciated and generate massive potential for regulatory coding that needs to be tested in future studies . Our results may have implications for the mechanism of CHD . While we acknowledge that our findings are derived from studies in a cell culture system , they compellingly suggest that NKX2-5 mutants , even those lacking DNA-binding ability , can bind to a subset of normal targets and off-targets , where they could affect transcription either positively or through dominant-negative action . Importantly , both NKX2-5∆HD and NKX2-5Y191C retained some functionality on select B- and C-set targets in vitro , an activity driven by heterodimerisation with cofactors . Our results confirm the inability of severe mutants to bind most normal targets , and therefore that haploinsufficiency is likely the dominant component of CHD mechanism . However , we predict that allele-specific dominant-negative and gain-of-function effects arising from dysregulation of the hundreds of off-targets will further destabilise the cardiac GRN and could contribute to disease . We have not yet analysed the expression of off-targets in mouse models or human CHD samples . Defining the targets and off-targets of a range of human CHD mutations and analysing their expressions in appropriate animal and human models will be needed to rigorously test the hypothesis .
Animal experimentation was performed with approval of the Garvan Institute/St Vincent's Hospital Animal Ethics Committee ( Project numbers 10/19 and 10/01 ) . ChIP experiments were performed in HL-1 cells following the Abcam Cross-link Chromatin ( X-ChIP ) protocol . Proteins and DNA were cross-linked with 0 . 75% formaldehyde at room temperature for 10 min . Immunoprecipitation of chromatin complexes was done using Santa Cruz antibodies cat no sc365207 for NKX2-5 and sc2025 for IgG control . Target gene levels were quantitatively measured by PCR . Primers were designed to validate DamID peaks and are provided in Supplementary file 2B . DamID experiments were performed after modification of published protocols ( Vogel et al . , 2007 ) . In brief , HL-1 cells at confluency were transduced with lentiviral vectors allowing undetectable expression of TF/Dam fusion proteins in the absence of induction . After 40 hr , genomic DNA was extracted using a Gentra PureGene Cell kit ( QIAGEN , Venlo , Netherlands ) , digested by DpnI at 37°C for 6 hr , and amplified by ligation-mediated PCR . PCR products were further fragmented with DNaseI at 24°C for 1 min , and labelled and hybridised to Affymetrix mouse 1 . 0R promoter microarrays according to manufacturers’ instructions . Three independent DamID experiments were performed utilising biological triplicates in first two studies and quadruplicates in the third study ( Supplementary file 2C ) . Peaks can be directly visualised in the UCSC Genome Browser following this http://genome . ucsc . edu/cgi-bin/hgTracks ? db=mm9&type=bed&hgt . customText=ftp://ftp . ncbi . nlm . nih . gov/geo/series/GSE44nnn/GSE44902/suppl/GSE44902_DamID . bed . gz . Microarray data were deposited in NCBI's Gene Expression Omnibus ( GSE44902 ) . HL-1 cells were seeded onto Lab-Tek chambers ( Thermo Scientific , Waltham , USA ) with 5 mg/ml fibronectin and 20 mg/ml gelatine and transfected with indicated paGFP fusion proteins and histone 2B ( H2B ) -RFP ( to identify nuclei ) using Lipofectamine 2000 ( Invitrogen ) , as per manufacturer's instructions . Fluorescence correlation spectroscopy ( FCS ) measurements were performed 48 hr post-transfection , following photoactivation of the paGFP fusion proteins with 405 nm light , as previously described ( Kaur et al . , 2013 ) using the Zeiss LSM780 laser scanning confocal microscope , with the avalanche photodiodes of the Confocur 3 module ( Zeiss , Jena ) . The autocorrelation function ( ACF ) G ( t ) for the fluorescent intensity over the 20 s measurement was calculated using the ZEN software FCS module ( Zeiss , Jena , Germany ) . Mean ACFs ( of 7 repetitions ) were fitted using the ZEN Software ( Zeiss ) using a model that comprised two 3D diffusion terms , one free and one anomalous , and another term to account for triplet state photophysics ( see [Kaur et al . , 2013] for further details ) . | Many genes working within large gene networks influence the development of heart muscle cells in humans and other animals . The activity of these genes is controlled in part by proteins called transcription factors , which bind to DNA and act as molecular switches . One transcription factor that is particularly important for the development of heart muscle cells is called NKX2-5 . Mice lacking NKX2-5 have abnormal hearts and many humans who are born with congenital heart disease carry mutations in the gene that encodes this protein . Many of these mutations alter a section of the protein called the homeodomain , and therefore interfere with the ability of NKX2-5 to bind to DNA or associate with other important cardiac proteins called cofactors . However , it is not clear how such mutations alter the behaviour of NKX2-5 across all of its targets . Bouveret et al . have now used a technique called ‘DNA adenine methyltransferase identification’ to study how NKX2-5 interacts with other proteins and DNA . The experiments found that , as expected , the mutant NKX2-5 proteins were unable to associate with many of the usual gene and protein targets of normal NKX2-5 . However , the mutant proteins were still able to bind to some of their usual targets , plus many other targets that the normal NKX2-5 protein was not able to bind to . A particular NKX2-5 mutant protein that the experiments analysed was missing the entire homeodomain , yet it was still able to associate with the normal NKX2-5 protein and bind to cofactors that help NKX2-5 to find its usual targets . This finding led Bouveret et al . to discover the role of a section of the NKX2-5 protein called the tyrosine-rich domain , which in the absence of the homeodomain can direct interactions of NKX2-5 with itself and its cofactors . Bouveret et al . 's findings suggest that protein cofactors of NKX2-5 help mutant NKX2-5 proteins retain some of their normal activities , but also direct the mutant proteins to abnormal gene targets , which could contribute to congenital heart disease . The next steps are to carry out experiments in animals to confirm these findings , and to understand the activities of mutant NKX2-5 and other mutant transcription factors across the whole genome . This could lead to new therapeutic approaches to treat congenital heart disease and other conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | NKX2-5 mutations causative for congenital heart disease retain functionality and are directed to hundreds of targets |
Bacterial small RNAs ( sRNAs ) are key elements of regulatory networks that modulate gene expression . The sRNA RydC of Salmonella sp . and Escherichia coli is an example of this class of riboregulators . Like many other sRNAs , RydC bears a ‘seed’ region that recognises specific transcripts through base-pairing , and its activities are facilitated by the RNA chaperone Hfq . The crystal structure of RydC in complex with E . coli Hfq at a 3 . 48 Å resolution illuminates how the protein interacts with and presents the sRNA for target recognition . Consolidating the protein–RNA complex is a host of distributed interactions mediated by the natively unstructured termini of Hfq . Based on the structure and other data , we propose a model for a dynamic effector complex comprising Hfq , small RNA , and the cognate mRNA target .
The expression of genetic information is controlled and synchronised through intricate regulatory networks . In bacteria , the control of gene expression post-transcription is mediated in part by small RNAs ( sRNAs ) , and their contributions enrich the computational complexity and repertoire of regulatory circuits ( Beisel and Storz , 2011 ) . Bacterial sRNAs are typically 50 to 300 nucleotides in length ( Storz et al . , 2011 ) , and each control expression of a distinct set of target mRNAs , which they recognise with high specificity and apparent precision . One of the major facilitators of sRNA activity in bacteria is the protein Hfq , which promotes pairing of an sRNA with its target mRNA in solution ( Møller et al . , 2002; Vogel and Luisi , 2011; Panja and Woodson , 2012; De Lay et al . , 2013 ) . Indeed , the kinetics of target pairing seem an essential aspect of sRNA action in vivo , as many of these sRNAs affect rates of translation or decay , either positively or negatively depending on target and context ( Storz et al . , 2004; Fröhlich and Vogel , 2009; Desnoyers and Massé , 2012; Papenfort et al . , 2013 ) . It can be envisaged how Hfq acts as a catalyst for such recognition in vivo , but the question naturally arises how sRNAs generally achieve precision , accuracy , and speed in producing their effects and avoid undesired off-target consequences . The key to understanding these fundamental processes of sRNA-based regulation is to determine how RNAs are bound and presented by Hfq and other effector proteins . The previous crystal structures of truncated Hfq variants have provided clues as to how the protein recognises short stretches of single-stranded RNA . Hfq bears an α + β fold that is the signature of the highly conserved family of Sm/Lsm proteins that draws members from all domains of life ( Kambach et al . , 1999; Wilusz and Wilusz , 2013 ) . Like other proteins of this extensive group , the bacterial Hfq self-assembles into a ring-like architecture . Hfq of Escherichia coli and other eubacteria forms a compact hexameric toroid that presents two structurally non-equivalent concave surfaces for molecular recognition; these faces are referred to as the proximal and distal faces , with the former exposing an N-terminal α-helix ( Schumacher et al . , 2002; Link et al . , 2009 ) . Crystallographic studies have identified interactions of short RNA polymers with either of these two surfaces and have inferred sequence preferences that have been corroborated and refined by mutagenesis and solution binding studies ( Schumacher et al . , 2002; Link et al . , 2009; Sauer and Weichenrieder , 2011; Robinson et al . , 2013; Peng et al . , 2014 ) . According to these results , the proximal site interacts preferentially with uridine-rich sequences , while the distal site favours the binding of ARN or ARNN motifs ( with R being a purine and N any nucleotide ) ( Schumacher et al . , 2002; Link et al . , 2009; Sauer and Weichenrieder , 2011 ) . In addition to the distal and proximal faces , the torus-shaped Hfq hexamer bears a convex circumferential rim that has recently been identified as a surface contributing to RNA binding ( Sauer et al . , 2012; Zhang et al . , 2013 ) . Structural studies of the hetero-heptameric Sm assembly from the mammalian splicing machinery have identified RNA binding surfaces that share some similarities with the bacterial counterpart ( Leung et al . , 2011 ) . Bacterial sRNAs may recognize their cognate target mRNAs in different ways . In some sRNAs , a target-recognition region is presented at or near the 5′-end ( Papenfort et al . , 2010 ) . For another subset of sRNAs , several pairing regions are non-continuous and each can independently bind distinct targets ( Beisel and Storz , 2011; Shao et al . , 2013 ) . In both cases , the sRNA recognition site ( referred to as the ‘seed’ ) and the target can have imperfect complementarity of various lengths with interactions being as short as 6 base-pairs , as seen for example in the SgrS/ptsG mRNA pair from E . coli ( Kawamoto et al . , 2006 ) . To fully understand the mechanism of sRNA mediated regulation , it is important to address the questions of how the intricate RNA folds are recognised , how the seeds are presented , and how the pairing of sRNAs with mRNAs is facilitated . As the sRNAs , mRNAs , and their complexes can all be remodelled upon binding Hfq , the rules for recognition are likely to be highly dependent on context . In this study , we describe the structure of the sRNA RydC in complex with the full-length Hfq protein of E . coli . Previous studies demonstrated that RydC is involved in biofilm regulation ( Bordeau and Felden , 2014 ) as well as in the control of membrane stability through the positive regulation of an isoform of the cfa mRNA encoding cyclopropane fatty acid synthase ( Fröhlich et al . , 2013 ) . RydC is proposed to have a pseudoknot fold that exposes a 5′ seed sequence , and the in vivo stability of RydC requires Hfq ( Antal et al . , 2005; Fröhlich et al . , 2013; Bordeau and Felden , 2014 ) . The crystal structure of the RydC–Hfq complex , together with biochemical and in vivo data , defines key interactions and suggests a hypothetical model for how the sRNA and mRNA targets might be matched through Hfq binding .
Despite its low abundance in comparison with other sRNAs under standard growth conditions , RydC has been repeatedly recovered with the Hfq protein in pull-down assays ( Zhang et al . , 2003; Sittka et al . , 2008; Chao et al . , 2012 ) . A distinctive feature of RydC is its predicted pseudoknot fold which is highly conserved and whose disruption by point mutations renders the RNA unstable in vivo ( Fröhlich et al . , 2013 ) . A double mutant of RydC ( G37C and G39C within helix 1; hereafter , RydC-S1 ) was predicted computationally to form a distinct structure with one short stem-loop at the 5′-end and a strong terminator hairpin at the 3′-end ( Figure 1A ) . 10 . 7554/eLife . 05375 . 003Figure 1 . Association of RydC with the chaperone Hfq and its pseudoknot fold are required for sRNA stability . ( A ) Secondary structure of the RydC pseudoknot and the predicted alternative structure are formed by the double mutant RydC-SI . Substitutions G37C and G39C are indicated with asterisks . ( B , C ) Stabilities of RydC and RydC-SI were determined by Northern blot analyses . Total RNA samples were extracted prior to and at indicated time-points after inhibition of transcription by rifampicin in late exponential phase ( OD600 of 1 ) from Salmonella strains ΔrydC ( JVS-0291 ) or Δhfq ( JVS-0584 ) , carrying plasmids expressing RydC ( pKF42-1 ) or RydC-SI ( pKF60-1 ) from the constitutive PL promoter . Error bars represent standard deviation calculated from three biological replicates . ( D ) Predicted duplex formed between the RydC 5′-end ( nts 2–11 ) and the longer isoform cfa mRNA ( nts −99 to −109 relative to the translational start site ) originating at transcriptional start site 1 ( TSS1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00310 . 7554/eLife . 05375 . 004Figure 1—figure supplement 1 . Electrophoretic mobility shift assay ( EMSA ) of Hfq with RydC or RydC-S1 . ( A ) Electrophoretic mobility shift assay ( EMSA ) with in vitro synthesized 5′-end-labelled RydC and RydC-S1 RNAs ( RydC* and RydC-S1* , 4 nM ) in the presence of increasing concentrations of Hfq protein as indicated . ( B ) Preformed Hfq/RNA* ( RydC* and RydC-S1* , 4 nM ) complexes were incubated with increasing concentrations of cold competitor RNA ( RydC-S1 for RydC*; RydC for RydC-S1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 004 To assess the molecular mechanism underlying the intrinsic instability of the RydC-S1 mutant , we compared its turnover rate to the wild-type RNA in the presence and absence of the chaperone Hfq ( Figure 1B/C ) . To this end , both RydC and RydC-S1 were expressed in Salmonella under the control of the constitutive PL promoter from high-copy plasmids . At exponential growth ( OD600 of 1 ) , transcription was stopped by the addition of rifampicin , and RNA levels were monitored at various time-points after treatment . As previously shown ( Fröhlich et al . , 2013 ) , RydC is very stable in wild-type cells ( half-life ( t1/2 ) >32 min ) , while RydC-S1 decayed rapidly ( t1/2 ∼2 min ) . In contrast , the half-life of RydC is already markedly reduced in hfq mutant cells ( t1/2 ∼4 min ) and there is some further acceleration of the decay rate for RydC-S1 , matching that of RydC in the absence of Hfq . This observation led us to assume that RydC stability depends on both the integrity of the pseudoknot fold and the ability to associate with the Hfq protein . Both the wild-type RydC and RydC-S1 mutant bind Hfq in electrophoretic mobility shift assays with in vitro synthesized RNA although the affinity of the mutant is lower than for the wild-type RydC ( Figure 1—figure supplement 1A ) . Competition experiments suggest that the binding sites for the RydC and RydC-S1 overlap at least partially ( Figure 1—figure supplement 1B ) . This suggests that the destabilisation in vivo might result mostly from the integrity of the pseudoknot fold , but to some degree , the Hfq binding alone may inhibit the degradation of RydC from initiating in single-stranded regions as suggested for other sRNAs ( Moll et al . , 2003; Saramago et al . , 2014 ) . The strong association of RydC with Hfq both in vivo and in vitro as well as its highly compact structure made it an ideal candidate for co-crystallizations . Co-crystals of full length E . coli Hfq and Salmonella RydC were obtained that diffracted to 3 . 48 Å resolution , and the crystal structure of the RydC–Hfq complex was solved by molecular replacement using a model of the structured hexameric core of E . coli Hfq ( i . e . , lacking the C-terminal residues beyond amino acid 65 ) . Unbiased , interpretable density for the RNA was apparent in the early maps calculated from the positioned protein hexamer , and a model for most of RydC could be fitted into the electron density ( Figure 2—figure supplement 1 shows a map in which the RNA was omitted from the refinement and phase calculations ) . As the crystal structure is limited to a resolution of 3 . 48 Å , the map does not provide unequivocal alignment of RydC sequence with structure; however , the path of the majority of the RNA and the duplex regions could be modelled confidently . The asymmetric unit of the crystal comprises one full Hfq hexamer and one RydC ( Figure 2 ) . The fold of the RydC is consistent with the pseudoknot structure predicted from solution studies and sequence alignments ( Antal et al . , 2005; Fröhlich et al . , 2013 ) . The RNA bridges two adjacent Hfq hexamers in the crystal lattice , effectively forming a distorted sandwich with a wedge-shape ( Figure 3A ) . The 3′-end of the RNA interacts with the proximal face of the principle hexamer , while the 5′-end interacts with the upper portion of the lateral surface of another hexamer in the asymmetric unit ( see below for details , paragraph ‘Interactions of RydC with the lateral surface of Hfq’ ) . 10 . 7554/eLife . 05375 . 005Figure 2 . Crystal structure of the RydC/Hfq complex . The content of the asymmetric unit is shown . The left panel presents a view along the molecular sixfold axis of the Hfq ring core ( roughly encompassing residues 3–72 ) , while the right panel is viewed along the perpendicular direction . The 3′-end poly-U tail of RydC is inserted in a groove on the proximal face of the hexamer . The N- and C-terminal regions are indicated with letters ( cyan and blue for different Hfq protomers ) ; these are natively unstructured and in multiple conformations , but some portions could be modelled . The RydC phosphate backbone is shown as an orange cartoon , bases are shown as sticks . Nucleotides 20–21 are disordered , but the backbone was modelled to help view the RNA structure and is indicated in grey cartoon style , as is also the last modelled phosphate at the 5′-end of RydC . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00510 . 7554/eLife . 05375 . 006Figure 2—figure supplement 1 . Electron density calculated after refinement of a model in which a portion of a RNA duplex region was omitted ( nucleotides 26–33 and 50–57 ) . The model was refined with REFMAC for 10 cycles and a Fo–Fc map calculated . The difference map is shown in green and is contoured at 3σ . The refined position of the RNA is superimposed on the map . The omitted RydC is shown in orange stick representation , while the rest of RydC and Hfq are coloured grey . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00610 . 7554/eLife . 05375 . 007Figure 2—figure supplement 2 . Analytical Ultracentrifugation of Hfq , RydC , and Hfq–RydC mixtures at different RNA:protein ratios . A 1:1 complex of RydC:Hfq6 can be detected ( yellow trace ) when the two components are mixed in a 1:1 ratio , but at higher ratios of RydC:Hfq , a 1:2 RydC:Hfq6 can form as well . Protein–RNA mixtures containing different molar ratios ( 1:1 , 1:2 , 1:10 ) of RydC to Hfq hexamer were incubated at room temperature for 30 min in 10 mM Tris–HCl , 20 mM NaCl , 10 mM EDTA , 5 mM DTT . The concentration of RydC in each experiment was 1 . 5 μM . Hfq alone was also run at a concentration of 173 μM . The sedimentation of the different species was followed both at 260 nm and 275 nm , corresponding respectively to the maximum absorbance of RNA and Hfq . The program Sedfit was used to extrapolate the sedimentation coefficients ( S ) for RydC , Hfq , and the protein–RNA complexes and their observed mass , after assuming that vHfq = 0 . 7248 , vRydC = 0 . 5500 and after calculating the vcomplex with the following equation: vcomplex = ( MWprotein × vprotein + MWRNA × vRNA ) / ( MWprotein + MWRNA ) = 0 . 6835 . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00710 . 7554/eLife . 05375 . 008Figure 2—figure supplement 3 . Size exclusion chromatography dynamic light scattering ( SEC-MALS ) . ( A ) Elution profile for the RydC:Hfq complex from a Superdex 200 column superimposed on the estimated molecular masses . Fractions over the peak were analysed with a denaturing protein gel ( B ) and a denaturing RNA gel ( C ) , confirming that both RydC and Hfq are present in the peak . The table in the lower panel summarises the results of the SEC-MALS data evaluations . SEC-MALS measurements were made with a Superdex 200 10/300 column connected to a miniDAWN TREOS multi-angle static Light Scattering and an Optilab T-rEX ( refractometer with EXtended range ) detector ( Wyatt Technology Corporation , Santa Barbara CA , USA ) . Mixtures of protein and RNA in different molar ratios were run through an S200 10/300 size exclusion chromatography column equilibrated in buffer A ( 50 mM Tris , pH 7 . 5 , 50 mM NaCl , 50 mM KCl , 5 mM MgCl2 , 2 mM DTT ) . Both RydC and Hfq were diluted in buffer A . The RNA was first annealed ( 90°C for 2 min , 30 min at room temperature ) before mixing it with an equal molar amount of Hfq6 . The final concentrations of RydC and Hfq6 in the mixture were equal to 30 μM; Hfq alone was at a molar concentration of 50 μM . BSA ( 2 mg/ml ) was run as positive control . Predicted and observed molecular weights for Hfq and Hfq/RydC complex are listed below . SamplePredicted molecular weightObserved molecular weight*PolydispersityHfq66 , 996 Da64 , 070 Da ( ±0 . 101% ) 1 . 000 ( ±0 . 142% ) Hfq:RydC 1:188 , 895 Da91 , 650 Da ( ±0 . 218% ) 1 . 000 ( ±0 . 308% ) DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00810 . 7554/eLife . 05375 . 009Figure 3 . Interactions established by the 3′-end and 5′-end regions of RydC with two adjacent Hfq hexamers in the crystal lattice . ( A ) RydC is sandwiched between two Hfq hexamers in the crystal lattice . Two asymmetric units are shown with the RydC in the centre ( coloured in orange ) bridging two Hfq hexamers . The symmetry-related Hfq molecule is labelled with an asterisk ( * ) . The 3′-end region of RydC in the main asymmetric unit is pointed by an arrow and contacts the proximal face of the principal Hfq hexamer , another arrow indicates the position of the 5′-end of RydC , interacting with a portion of the rim of the Hfq hexamer in the neighbouring asymmetric unit . Two black rectangles define the 5′-end and 3′-end RydC–Hfq contact regions , which are enlarged in panels B and C respectively . ( B ) Contacts established between the 5′-end seed region of RydC and the rim of a symmetry related Hfq molecule . The main interactions at this interface involve the following amino acid–nucleotide pairs: R17-G8 , H71-G8 , R16-U9 , N13-U9 , F39-U9 , Q5-G11 , P10-A10 . Residues belonging to the symmetry related Hfq molecules are labelled with an asterisk ( * ) . ( C ) Interactions of the poly-U 3′-end of RydC with the recessed groove of the proximal face of Hfq . The uridines and cytosine stack on F42; this and other interactions are similar to those seen in the U6/Hfq crystal structure ( Sauer and Weichenrieder , 2011 ) . The side chains of critical residues that contribute to keep RydC in the channel are shown in stick representation . The adjacent protomers of Hfq are coloured in blue and cyan , RydC is shown in orange . Side chains are shown for all residues , main chains are shown for Q41 in order to highlight the position of atoms involved in interactions with neighbouring residues . The RNA used in the crystallisations has a guanine on the 3′-end originating from the template for in vitro transcription , but this base may be in multiple conformations . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 00910 . 7554/eLife . 05375 . 010Figure 3—figure supplement 1 . Distances of the main contacts between Hfq and the 5′- and 3′-ends of RydC . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 010 The organisation of the RNA in the crystal would suggest that a single RydC molecule could form a closed complex with two sandwiching Hfq hexamers . However , solution studies using analytical ultracentrifugation ( AUC ) and multi-angle laser light scattering ( SEC-MALS ) indicate that the RydC can form a stable 1:1 complex with the Hfq hexamer under various buffer conditions when the components are present in unitary stoichiometry ( Figure 2—figure supplements 2–3 ) . When Hfq is present at twice the concentration of RydC under low ionic strength conditions , some 2:1 Hfq:RydC species can be identified; moreover , a higher order complex can be observed by electrophoretic mobility shifts when Hfq is in excess over RNA ( results not shown ) . These observations suggest that the stoichiometry of Hfq/RNA complexes is very sensitive to solution conditions . In vivo , it seems unlikely that one RydC can find two Hfq hexamers when the protein is limiting to total RNA ( Wagner , 2013 ) , and we suggest that the 1:1 complex is likely to represent the physiologically relevant species . We propose that the 2:1 sandwiching complex we observe in the crystal is not the biologically relevant species , but that it encompasses the full set of interactions that would form in a closed 1:1 RydC:Hfq complex . We suggest that the contacts made by RydC with the neighbouring Hfq hexamers can form on the proximal face of the same , single hexamer to form an idealised 1:1 complex . This hypothetical model will be elaborated on further below . The 3′-end of RydC has a U-rich tail that - together with a strong hairpin structure - facilitates Rho-independent transcription termination and which is an important determinant for recognition of many sRNAs ( Otaka et al . , 2011; Sauer and Weichenrieder , 2011; Ishikawa et al . , 2012 ) . The RydC/Hfq structure reveals that the U-rich tail is bound in a recessed channel on the proximal face of Hfq . The electron density is not well defined in this region , and there may be disorder or multiple conformations of the polyU tail , but the features do show that uridines U61 , U62 , and U64 ( as well as C63 ) each forms an aromatic base stacking interaction with the ring of F42 ( Figure 3C; distances provided in Figure 3—figure supplement 1 ) . K56 forms hydrogen bonds with the O2 group of the uracils , consistent with the importance of this residue in binding RNA on the proximal face ( Mikulecky et al . , 2004 ) . The model indicates that there are likely to be hydrogen bonds between the uridine bases and Q41 and Q8 , as well as between the furanose 2′ OH group with the imidazole group of H57 . These interactions are consistent with the crystal structure of Salmonella Typhimurium Hfq/polyU ( Sauer and Weichenrieder , 2011; PDB code 2YLC ) and with fluorescence quenching experiments for binding U6 ( Robinson et al . , 2013 ) . Similar interactions are seen in the crystal structure of Staphylococcus aureus Hfq in complex with AUUUUUG ( Schumacher et al . , 2002; PDB code 1KQ2 ) and E . coli Hfq in complex with AUUUUUA ( Wang et al . , 2011 , 2013; PDB code 4HT9 ) . In principle , the RNA can visit each of the uridine pockets following a clockwise or counter-clockwise path , but the polarity seems to be consistently clockwise in the reference frame that views the proximal face . The RNA exits the groove to engage on the rim of the proximal face of Hfq through numerous interactions with protein side- and main-chain atoms , as we describe further below . On the exposed rim of the proximal face , U24 stacks onto F39 and U23 packs onto U24 ( Figure 4A ) . The U46/U47 bases form similar interactions with the rim of Hfq in a neighbouring asymmetric unit ( Figure 4B ) . U24 and U47 contact N13 and likely make a hydrogen bonding interaction . Both U23/U24 and U46/U47 pairs are highly conserved in an alignment of the available RydC sequences ( Fröhlich et al . , 2013 ) . The U24/U25 and U46/U47 interactions with the rim are consolidated by the N-terminal residues G4 and Q5 and likely include amide backbone and lysine side chain interactions ( Figure 4A , B; distances provided in Figure 4—figure supplement 1 ) . 10 . 7554/eLife . 05375 . 011Figure 4 . Contacts between the conserved U–U pairs of RydC and the Hfq rim . ( A ) The U23/U24 pair interacts with F39 , R16 , R17 , and N13 on the convex rim of Hfq . The U23/U24 interactions with the rim are consolidated by the N-terminal residues G4 and Q5 of a neighbouring Hfq protomer ( coloured in cyan ) . ( B ) The U46/U47 pair makes similar interactions with Hfq in a neighbouring asymmetric unit . Residues belonging to the symmetry related Hfq molecule are labelled with an asterisk ( * ) in panel B . Side chains are shown for all residues , main chain is shown for G4 in both panels to help visualize the residue position . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 01110 . 7554/eLife . 05375 . 012Figure 4—figure supplement 1 . Distances of the main contacts between Hfq and the U23/U24 and U46/U47 base steps of RydC . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 012 Although highly conserved , the U23/U24 and U46/U47 pairs are not required for the base-pairing to the RydC target , cfa mRNA , because a chimeric RNA assembled from the RydC 5′-end ( nt . 1 to 13 ) and the 3′-end of the sRNA MicA is as efficient in target gene regulation as the wild-type RydC ( Fröhlich et al . , 2013 ) . To address the function of the highly conserved U residues , three mutant variants of RydC were constructed ( Figure 5A ) . RydC-LI ( U23G U24C ) , RydC-LII ( U46C U47G ) , as well as RydC-LI/II ( U23G U24C U46C U47G ) were expressed from a constitutive promoter in wild-type Salmonella cells , and RNA stability was assessed by rifampicin treatment ( Figure 5B ) . Compared to wild-type RydC ( t1/2 ∼32 min ) , mutant RydC-LII had a reduced half-life ( t1/2 ∼11 min ) . Both variants RydC-LI and the double mutant RydC-LI/LII were even less stable ( t1/2 ∼2 min ) and especially the RydC-LI and RydC-LI/LII variants were less potent than the wild-type to regulate the reporter target Cfa–GFP ( Figure 5C ) . The discrepancy between steady-state levels and regulatory potential observed between the different RydC variants could be the result of alternative binding patterns via LI and LII leading to changes in RNA orientation on Hfq . Thus , while functional conclusions may be limited at this point , our in vivo data suggest a requirement of the conserved U pairs in the loops of the RydC pseudoknot structure for RNA stability and indicate a functional importance for the interactions observed in the crystal structure of the complex . 10 . 7554/eLife . 05375 . 013Figure 5 . Requirement for the conserved UU steps for in vivo stability of RydC . ( A ) Schematic of RydC pseudoknot fold with substitutions of U23G and U24G ( RydC-LI ) and U46G and U47G ( RydC-LII ) . ( B ) Stabilities of RydC and its variants RydC-LI , RydC-LII , and RydC-LI/LII . RNA levels were determined by Northern blot analyses . Total RNA samples were extracted prior to and at indicated time-points after inhibition of transcription by rifampicin in late exponential phase ( OD600 of 1 ) from Salmonella strain ΔrydC ( JVS-0291 ) carrying plasmids expressing RydC ( pKF42-1 ) , RydC-LI ( pKF224-9 ) , RydC-LII ( pKF223 ) , or RydC-LI/II ( pKF225 ) from the constitutive PL promoter . Error bars represent standard deviation calculated from three biological replicates . ( C ) Target regulation by RydC loop mutants . Salmonella ΔrydC ( JVS-0291 ) express a plasmid-borne translational fusion of the RydC target cfa to the green fluorescent protein ( Cfa–GFP; pKF31-1 ) in the presence of either a control ( pJV300 ) or plasmids expressing RydC ( pKF42-1 ) , RydC-LI ( pKF224-9 ) , RydC-LII ( pKF223 ) , or RydC-LI/II ( pKF225 ) from the constitutive PL promoter . Bacteria were grown to late exponential phase ( OD600 of 1 ) and Cfa–GFP and GroEL proteins were detected by Western blot ( upper two panels ) . RydC variants as well as 5S RNA were detected by Northern blot ( lower two panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 013 The observed U23/U24 and U46/U47 interactions with the rim region may account for the findings from solution studies that E . coli Hfq has greater affinity for 16-mer polyU compared to the 6-mer U6 , because the longer polymer could form the contacts in the recessed core with the U-rich 3′-end and simultaneously interact with the rim through F39 and N13 . The U23/U24 and U46/U47 interactions may also account for the partial fluorescence quenching observed for F39 in the presence of U6 ( Robinson et al . , 2013 ) , because excess U6 may bind to the rim and interact with F39 in a similar way to the contacts observed for U23/U24 and U46/47 pairs of RydC . Earlier studies identified residues R16 , R17 , R19 , and K47 as forming a lateral RNA binding surface in Salmonella Hfq ( Sauer et al . , 2012; highlighted in cyan in Figure 6C ) . These residues are part of an arginine patch that is proposed to interact with both the sRNA and the mRNA molecules ( Sauer et al . , 2012 ) . In our structure , R16 and R17 make phosphate backbone contacts near U23 and U24 ( Figure 4A ) , while R19 and K47 are not interacting with the RNA . Residues R16 and R17 belonging to Hfq in the neighbouring crystal unit cell are also involved in the binding of the 5′-end of RydC ( Figure 3B ) . The nucleotide stretch 8–12 , representing the last portion of the RydC seed region , is in fact kept in an extended conformation by the interaction with helical region 8–17 from a symmetry related Hfq molecule ( Figure 3B ) . This may account for observations from previous studies showing that mutations on this region of the rim do not impact on the ability of Hfq to bind sRNA sequences but rather on its capacity to stimulate annealing of sRNA with an mRNA target ( Panja et al . , 2013 ) . 10 . 7554/eLife . 05375 . 014Figure 6 . A potential RNA interaction surface on the circumferential rim of the Hfq hexamer . ( A ) Interaction between a duplex region of RydC and the rim surface of an Hfq molecule belonging to the neighbouring asymmetric unit in the crystal . Symmetry related RydC and Hfq are labelled with an asterisk ( * ) . ( B ) A detailed view into the interaction site , showing the main protein residues making contacts with the duplex RNA . The side chain of R66 is not shown as it is disordered in the structure . ( C ) A surface representation of the RydC/Hfq complex , highlighting in colour in one protomer , the residues of the proposed extended rim that recognises RNA . The residues coloured cyan are from earlier studies that identified the rim as an RNA binding region ( R16 , R17 , R19 , K46; Sauer et al . , 2012 ) , and the yellow residues include an extended surface proposed here to engage duplex RNA ( P21 , Q33 , Q35 , T49 , R66 ) . Also labelled are the residues of this extended surface that engage the U–U steps in RydC ( N13 , R16 , R17 , F39; F39 and N13 are not visible in this view ) . ( D ) Rates and equilibrium constants for binding of Hfq and Hfq/RydC to cfa for wild-type ( HfqWT ) and rim-mutant protein ( HfqRim; R19A , P21A , Q33A , Q35A , T49A , R66A ) . The measurements were performed by fitting the averaged responses from three independent experiments . It was not possible to fit the weaker binding response of HfqRim binding RydC . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 01410 . 7554/eLife . 05375 . 015Figure 6—figure supplement 1 . Interactions between RydC and the proximal face of a symmetric Hfq molecule . ( A ) Hfq belonging to a neighbouring asymmetric unit is shown at the top right with the RNA contacting its distal face . ( B ) An expanded view into the distal face showing nucleotides A6 and C42 bound to the two pockets formed by residues I30 and Y25 . Crystallographic and solution binding studies show that E . coli Hfq has preferences to bind an A-A-N motif on the distal face ( Robinson et al . , 2013 ) . The contacts observed for RydC A6 and C42 with the distal are fortuitous lattice interactions , but they mimic interactions of the A-A-N repeats on the distal face . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 01510 . 7554/eLife . 05375 . 016Figure 6—figure supplement 2 . Binding rates and equilibrium constants for Hfq66 and Hfq66/RydC to immobilized cfa . The measurements were performed by fitting the response from three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 016 The RNA mediates numerous other interactions with neighbouring Hfq hexamers . One of the contacts mimics interactions of the A-A-N repeats on the distal face ( Figure 6—figure supplement 1 ) . A more extensive and potentially more interesting interaction occurs between the duplex region of RydC and the rim surface of an Hfq hexamer in a neighbouring asymmetric unit ( Figure 6A ) . The residues mediating this interaction are R19 , R66 , P21 , T49 , Q33 , and Q35 ( Figure 6B ) . To explore the contribution of the potential rim contacts to mediate the interaction between RydC and cfa , an Hfq mutant in the lateral rim ( where residues R19 , P21 , Q33 , Q35 , T49 , R66 have been mutated to A ) was expressed and purified , and its binding rates and affinities were estimated from interferometry ( ForteBio Octet Red96 ) . For this experiment , the 5′ region of cfa ( cfa1 , nucleotides from TSS1 to −72 relative to the AUG , 139 nts , Fröhlich et al . , 2013 ) was prepared with biotin attached to the 5′-end , which allowed the mRNA fragment to be immobilised on a streptavidin sensor surface . As shown in Figure 6D , both the wild-type protein ( HfqWT ) and the Hfq rim mutant ( HfqRim ) have similar affinities and association rate constants for the immobilised cfa . In the presence of RydC , the association of Hfq with cfa changes . The binding affinity of HfqWT for cfa decreases by a factor of roughly 5 , suggesting that the formation of the Hfq–sRNA complex affects the Hfq affinity for the mRNA fragment . Moreover , the presence of RydC decreases by an order of magnitude the association rate constant of Hfq for cfa . This could result from cfa being bound in a slightly different way in the Hfq/cfa complex as compared to the Hfq/RydC/cfa ternary complex . Strikingly , in the presence of RydC , the HfqRim mutant no longer has detectable binding to cfa . This finding suggests that the RydC interacts differently with HfqWT and the HfqRim lateral surface mutant in a way that perturbs cfa binding . Thus , the rim interactions shown in Figure 6A , B , C are crucial for forming a productive sRNA:Hfq:mRNA assembly . The C-terminal regions of Hfq , beyond residue 70 , are generally disordered , but density could be observed for some of the protomers into which the model could be partially extended . Although poorly ordered , the C-terminal tails appear to be making distributive contacts over the surface of RydC . The N-terminal regions of three Hfq protomers , exposed on the proximal face of the protein , also interact with RydC , mainly in the vicinity of nucleotides 23–25 , 51–52 , and 59 . To explore if the observed interactions between the Hfq C-termini and RydC might be important in vivo , we compared the stabilities of the sRNA in the presence of wild-type Hfq and an Hfq mutant truncated at residue 70 ( Figure 7 ) . We observed that the half-life of RydC was greatly reduced in the absence of the C-terminal end ( t1/2 ∼5 min ) and was comparable with the half-life observed for an hfq null mutant ( t1/2 ∼4 min ) . However , upon initial decay of a subfraction of RydC , the remaining population appears to be more stable and is only degraded at a rate comparable to RNA in the wild-type background . In vitro binding experiments show that the affinity of purified Hfq truncated at residue 66 ( Hfq66 ) for binding cfa is similar to that of the wild-type protein; however , the presence of RydC decreases substantially the binding of Hfq66 ( Figure 6—figure supplement 2 ) . The curve for Hfq66/RydC binding to cfa could not be modelled , but the affinity is likely to be reduced by two orders of magnitude compared with the binding of Hfq66 alone to cfa . This reduction is much greater than the fivefold decrease in affinity when comparing the mRNA binding of the wild-type Hfq and its Hfq/RydC complex ( Figure 6D ) . These observations suggest that the C-terminal tails are important for facilitating the formation of the sRNA:mRNA:Hfq ternary complex . 10 . 7554/eLife . 05375 . 017Figure 7 . Role of the Hfq C-terminus in RydC stability in vivo . Stabilities of RydC were determined by Northern blot analyses in Salmonella strains ΔrydC ( WT; JVS-0291 ) , ΔrydC Δhfq ( Δhfq; JVS-10665 ) , and ΔrydC hfq70 ( hfq70; JVS-11150 ) expressing RydC from the constitutive PL promoter ( pKF42-1 ) . Total RNA samples were extracted prior to and at indicated time-points after inhibition of transcription by rifampicin in late exponential phase ( OD600 of 1 ) . Error bars represent standard deviations calculated from three biological replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 017 The C-terminal tails appear to influence the rate of Hfq–RNA complex formation , as indicated by the order-of-magnitude lower on-rate of Hfq66 compared to HfqWT for binding cfa ( Figure 6D and Figure 6—figure supplement 2 ) . Moreover , the interactions of the tails with sRNA may explain the puzzling observation that the HfqWT–RydC has a lower kon than HfqWT for association with cfa . In this case , we envisage that the tails , sequestered with the sRNA , are no longer available to ‘fish’ for and accelerate association with the mRNA partner .
The crystal structure of RydC/Hfq suggests key elements of molecular recognition of this effector complex . We observe interactions of the 3′ polyU tract with the recessed pore in the Hfq proximal face and interactions of U–U dinucleotide steps in single-stranded regions with the surface near the rim . Based on results from in vivo crosslinking and RNA sequencing from enterohemorrhagic E . coli , it has been proposed that the consensus Hfq binding site on many sRNAs includes a U–U dinucleotide associated with an unpaired region ( Tree et al . , 2014 ) . Our results are consistent with this prediction and provide a structural rationalisation: the unpaired region is required to engage the U–U step on the rim . Interactions with the rim may account for the capacity of longer U-tracts to bind Hfq more strongly ( Murina et al . , 2013; Panja et al . , 2013 ) , since these are anticipated to bridge between the rim and the 3′-polyU-binding pore . The rim might also be the binding site for the internal U-rich region in SgrS that is required for stable association with Hfq ( Ishikawa et al . , 2012 ) . The crystal structure suggests that one RydC can be sandwiched between two Hfq hexamers , and a similar model has been proposed for interaction of Hfq with the sRNA DsrA ( Wang et al . , 2011 ) . However , solution data indicate that the stoichiometry of the complex is sensitive to buffer conditions and that a 1:1 complex is preferred at higher salt concentrations , but a 2:1 Hfq hexamer:RydC complex can be formed in lower ionic strength buffer ( Figure 2—figure supplement 2 ) . We propose that the 1:1 complex is the physiologically relevant species in vivo and that the sandwiching complex we observe in the crystal is likely to mimic a folding intermediate en route to forming a closed 1:1 RydC:Hfq complex . Accordingly , the contacts formed by the RydC with the neighbouring Hfq hexamers can both occur on the proximal face of the same , single hexamer to form the 1:1 complex ( shown schematically in Figure 8 ) . 10 . 7554/eLife . 05375 . 018Figure 8 . Hypothetical model of the RydC/mRNA/Hfq effector complex . Portions of the protomers forming the Hfq core have been represented with six spheres , from which the disordered C-terminal tails extend radially . RydC ( orange ) sits on the proximal face of Hfq , with the 3′-end U-rich tail interacting with the central channel and the two conserved U–U pairs ( U23/U24 and U46/U47 ) making contacts with the lateral face of the hexamer . The two double strands conferring the pseudoknot structure to RydC are indicated as S I and S II . The target mRNA cfa , depicted in yellow , associates with the distal face of Hfq , and it is proposed to form a duplex with the 5′-end ‘seed’ region of RydC that is recognised by the circumferential rim of Hfq . The six long C-termini are depicted in the cartoon , and the shorter N-termini are not shown for clarity . The association of the two RNA partners is aided by the C-terminal tails of Hfq , which extend towards the RNA molecules , embracing them and stabilizing the seed/target pairing . The natively unstructured termini make distributed interactions with the RNA , visiting dynamically multiple points on the nucleic acid . DOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 018 Solution studies indicate that the C-termini of Vibrio cholerae and E . coli Hfq are natively unstructured ( Beich-Frandsen et al . , 2011; Vincent et al . , 2012 ) . For most of the Hfq subunits in the crystal structure studied here , the C-terminal regions beyond residue 77 are disordered , but broken density in the vicinity is likely to be due to multiple conformations of the C-terminal region . These regions make consolidating interactions with the RNA and appear to be distributed over the exposed surfaces of the nucleic acid . A structural condensation of the C-terminal domains with sRNA binding is also consistent with data from X-ray and neutron solution scattering ( Henderson et al . , 2013 ) . Both the structural data as well as in vivo experiments suggest that the C-terminal end of Hfq contributes to the association with RydC . The distributive interactions of the C-terminal domain are not expected to provide specificity to the recognition but may enhance the stability of equilibrium complexes as well as boosting the association rates by increasing the effective molecular encounter radius . The distributive interaction may account for the finding that deletion of the C-terminal extension decreases binding to mRNAs ( Večerek et al . , 2008 ) and that the C-terminal tail can bind to some RNA species ( Robinson et al . , 2013 ) . The same might be true for yeast Lsm1 mutants with C-terminal truncations that have reduced affinity for RNA ( Chowdhury et al . , 2012 ) . These results are in agreement with a study that compared the distribution of Salmonella RNAs associating with Hfq variants of different species ( Sittka et al , 2009 ) . Similar to several orthologues of various bacterial species , Hfq of the archaeon Methanocaldococcus jannaschii lacks the unstructured C-terminus . M . jannaschii Hfq is generally able to complement a Salmonella hfq null mutant; however , the distribution of associated sRNAs recovered from pull-down experiments varies greatly compared to the wild-type . While a number of Salmonella sRNAs ( including , for example , the Hfq-dependent SroC , InvR , and GcvB ) are not reduced in their cellular abundance , a second class of sRNAs is strongly affected by the absence of the C-terminal end . This latter group includes RydC and also several sRNAs highly abundant in the wild-type ( e . g . , RprA , SdsR , and ArcZ ) , and it is tempting to speculate that this subpopulation of sRNAs is associated with Hfq through supporting interactions with the C-terminal tail . The differing roles of the tails for different sRNAs may account for the apparently conflicting results reported in the literature on the effects of the C-terminus on sRNA binding and activity ( Olsen et al . , 2010 ) . Finally , it should be noted that the C-terminal tails have different lengths in different Hfq homologs , and it is likely that their role differs in different species ( Vincent et al . , 2012 ) . As a general principle , the interactions of natively unstructured regions with RNA may be important in other nucleic acid/protein complexes ( Tompa and Csermely , 2004; Kucera et al . , 2011; Jonas and Izaurralde , 2013 ) , and they can contribute to cooperative binding ( Hunter and Anderson , 2009; Motlagh et al . , 2014 ) . We envisage that these interactions would help to keep the sRNA and mRNA in proximity without constraining them , so that the mRNA might travel along the distal face while the sRNA was held in a more fixed conformation on the proximal face . In this way , the complex can glide along an mRNA until the seed and target region meet . Once this occurs , they will close like a zipper to form a short duplex that could be engaged on the convex rim of the Hfq . The mismatch of molecular symmetries and the high flexibility of both the N- and C-termini of Hfq ensure that the complex will be intrinsically heterogeneous in conformation , but highly dynamic . Based on the current crystal structure , including the lattice interactions observed , and on other available structural and biophysical data , we propose a speculative model for the ternary complex of Hfq/sRNA and mRNA target ( Figure 8 ) . The model shares some similarity to the schematic proposed by Panja et al . ( 2013 ) and predicts that the 5′ seed will be exposed and available for interaction with the target to form a duplex region that binds to the convex rim of the Hfq hexamer . We envisage that a different model will be required to explain the interactions of sRNA using recognition regions that are internal and not in the 5′ seed . One prediction is that the base-pairing sites in this group of sRNAs will be presented near the rim of Hfq . This model awaits testing .
Wild-type E . coli Hfq was expressed and purified as previously described by Bandyra et al . ( 2012 ) . Hfq mutants were expressed from pBAD vector in Top10 Δhfq strain ( JVS-02001 ) to avoid hetero-hexamer formation . The mutant proteins were purified with the same protocol used for the wild-type Hfq , with a HiTrap Heparin purification step added ( buffer A: 50 mM Tris , pH 8 . 0 , 100 mM NaCl , 100 mM KCl; buffer B: buffer A + 1 M NaCl ) . The presence of mutations or truncations was confirmed by mass spectrometry analyses . For Salmonella RydC RNA in vitro transcription ( IVT ) , a DNA template was amplified by PCR from pKF42-1 using primers RydC_T7_fwd and RydC_T7_rev ( Table 1 ) . 10 . 7554/eLife . 05375 . 019Table 1 . OligonucleotidesDOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 019NameSequence 5′-3′RydC_T7_fwdGTTTTTTTTTTAATACGACTCACTATAGGCTTCCGATGTAGACCCGTRydC_T7_revCAGAAAACGCCTGCGTCTAACCAGGACCCGJVO-0322CTACGGCGTTTCACTTCTGAGTTCJVO-4363AGAAAACGCCTGCGTCJVO-4558GTTTTTTTTAATACGACTCACTATAGGTTGTTTATATTACGATAATT JVO-4721GTTTTTTTTTTAATACGACTCACTATAGGTTCCGATGTAGACCCGTCCJVO-4722AGAAAACGCCTGCGTCTAACJVO-5165GTTTTTTTTTTAATACGACTCACTATAGGTTCCGATGTAGAGCGGTCCJVO-9044CCCACGGACAATTCCGTJVO-10909CAGCAACAATGCCGGTGGCGGCGCCAGCAATAACTACCATTAAGGTCCATATGAATATCCTCCTTAGJVO-10910ATTATCCGACGCCCCCGACATGGATAAACAGCGCGTGAACGTGTAGGCTGGAGCTGCTTCJVO-10913CGAGACGCAGGCGTTTTCJVO-10914P-CCAGGACCCGTGACGJVO-10915GCCGCCTGCGTCACGGJVO-10916P-GGAGGACGGGTCTACATCP: 5′ phosphorylation Transcription mixtures containing DNA-template , ribonucleotides ( rNTPs ) , T7 RNA polymerase , DTT ( dithiothreitol ) , and RNase Out ( Invitrogen , UK ) were incubated in transcription buffer ( 400 mM Tris ( pH 7 . 9/8 . 0 ) , 250 mM MgCl2 , 20 mM spermidine ) for 4 hr at 37°C , and then Turbo DNase ( Ambion ) was added to digest the template . The IVT product was separated on a 10% polyacrylamide gel containing 7 M urea . The band corresponding to the transcribed RydC was visualized by UV-shadowing , excised from the gel , and the RNA was electro-eluted overnight using an EluTrap system ( Whatman , UK ) . The RNA sample was concentrated by ultrafiltration through a Vivaspin 500 concentrator ( 5 kDa cut-off ) . The RNA–protein complex was prepared by mixing purified Hfq6 and RydC in a ratio of 2:1 , the Hfq hexamer concentration being 15 mg/ml . Crystals of the Hfq–RydC complex were grown by sitting drop vapour diffusion after adding an equal volume of crystallization buffer ( 0 . 2 M trisodium citrate , 0 . 1 M sodium cacodylate ( pH 6 . 5 ) , 15% vol/vol isopropanol ) to the protein–RNA mixture . Crystals were harvested using 25% PEG400 as cryoprotectant and flash frozen in liquid nitrogen . X-ray data were collected on station I24 at Diamond Light source . Data were collected at 100 K at a wavelength of 0 . 9778 Å . Images obtained from the two best crystals having the same space group and cell dimensions were merged and scaled together , before proceeding with the structure determination . The structure was solved by molecular replacement with PHASER ( McCoy et al . , 2007 ) using residues 5–66 of E . coli Hfq as the search model . Electron density for RNA was apparent in the early maps . Maps were improved with density modification with PARROT using histogram matching with tRNA/aminoacyl-tRNA synthase experimental structure factors as the reference distribution . This structure has a similar solvent content and RNA/protein mass ratio as the RydC/Hfq crystals , and density modification with this method improved the map and visualisation of the RNA . The structure was refined using BUSTER ( Bricogne et al . , 2011 ) and with jelly body restraints in refinement with REFMAC ( Murshudov et al . , 2011 ) . The model was built using COOT ( Emsley et al . , 2010 ) . The protein and RNA stereochemistry were validated by using both Coot validation tools and Procheck from the CCP4 suite ( Laskowski et al . , 1993; Emsley et al . , 2010 ) . The Ramachandran plot of the model places 92 . 9% of residues in most favoured regions , 5 . 4% in additional allowed regions , 1 . 6% in generously allowed regions , and none in disallowed regions . X-ray data collection and refinement statistics are summarized in Table 2 . Figures were prepared using PYMOL ( DeLano , 2006 ) . The coordinates and structure factors have been deposited in the PDB with accession code 4v2s . 10 . 7554/eLife . 05375 . 020Table 2 . X-ray data collection and refinement statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 020Data collection Space groupP212121 Cell dimensions a , b , c ( Å ) 71 . 94 , 73 . 36 , 137 . 95 α , β , γ ( ° ) 90 , 90 , 90 Resolution ( Å ) 24 . 47–3 . 48 ( 3 . 69–3 . 48 ) * Rmerge0 . 152 ( 0 . 886 ) I/σI9 . 6 ( 2 . 3 ) Completeness ( % ) 99 . 1 ( 99 . 4 ) Multiplicity5 . 8 ( 5 . 9 ) Refinement Resolution ( Å ) 24 . 32 ( 3 . 48 ) No . reflections9190 Rwork/Rfree0 . 22/0 . 28 No . atoms4470 Protein3238 RNA1187 Water45 B-factors Overall94 . 69 Protein93 . 36 RNA122 . 60 Water54 . 90 R . m . s . deviations Bond lengths ( Å ) 0 . 010 Bond angles ( ° ) 1 . 475*Values in parentheses are for the highest-resolution shell . Kinetic measurements with Bio-Layer Interferometry were performed using an Octet RED96 equipped with Streptavidin sensors ( ForteBio , UK ) on 96-well plates . The experiment was performed in the binding buffer ( 25 mM Tris , pH 7 . 5 , 50 mM NaCl , 50 mM KCl , 1 mM MgCl2 , 1 mM DTT ) , which was also used to prepare all dilutions , for dissociation and neutralisation . A fragment of cfa mRNA ( cfa1 , nucleotides from TSS1 to −72 relative to the AUG , 139 nts , Fröhlich et al . , 2013 ) was in vitro transcribed from a PCR template ( JVO-4558*JVO-9044 on pKF31-1 ) and labelled with biotin in an IVT reaction using fivefold excess of GMP-biotin ( TriLink Biotechnologies , San Diego CA , USA ) over GTP . 5′ biotin labelled cfa fragment was immobilised on the biosensor that was subsequently submerged into 10 μM solution of maltose binding protein ( MBP ) labelled with biotin . The binding of wild-type and mutant Hfq were assayed at 0 , 5 , 10 , 25 , 75 , 150 , 250 , and 500 nM protein over 400 s , in the absence or presence of 1 μM chemically synthesized RydC ( Dharmacon , GE Healthcare , UK ) . The dissociation was monitored over 300 s and was followed by regeneration of the sensors using 1 M MgCl2 . Another set of tips was saturated with MBP-biotin and the measurements were then repeated for all Hfq and Hfq–RNA concentration series . The data were fitted with Data Analysis software ( ForteBio ) with a 1:1 binding model . The plots were prepared with Profit ( Quantum Soft , Switzerland ) using the following equation for response fit: Y = Rm × Xn/ ( Kdn + Xn ) , where Y is the observed binding , X is the molar concentration of the ligand , Rm is the maximum specific binding , and n is the Hill’s coefficient . Sequences of all oligonucleotides employed in this study are listed in Table 1 . All plasmids used in this study are summarized in Table 3 . 10 . 7554/eLife . 05375 . 021Table 3 . PlasmidsDOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 021Plasmid trivial namePlasmid stock nameRelevant fragmentCommentOrigin , markerReferenceUsed in Figurectrl . pJV300pPL control plasmid , expresses an ∼50 nt nonsense transcript derived from rrnB terminatorColE1 , AmpR ( Sittka et al . , 2007 ) 5CpPL-RydCpKF42-1RydCExpresses Salmonella RydC from constitutive PLlacO promoterColE1 , AmpR ( Fröhlich et al . , 2013 ) 1B/C; 5B/C; 7pPL-RydC-SIpKF60-1RydC-SIExpresses Salmonella RydC-SI ( SNEs G37C; G39C ) from constitutive PLlacO promoterColE1 , AmpR ( Fröhlich et al . , 2013 ) 1B/CpPL-RydC-LIpKF224-9RydC-LIExpresses Salmonella RydC-LI ( SNEs U23G; U24C ) from constitutive PLlacO promoterColE1 , AmpRThis study5B/CpPL-RydC-LIIpKF223RydC-LIIExpresses Salmonella RydC-LII ( SNEs U46C; U47G ) from constitutive PLlacO promoterColE1 , AmpRThis study5B/CpPL-RydC-LI/LIIpKF225RydC-LI/LIIexpresses Salmonella RydC-LI/LII ( SNEs U23G; U24C and U46C; U47G ) from constitutive PLlacO promoterColE1 , AmpRThis study5B/Ccfa::gfppKF31-1cfa::gfpexpresses cfa::gfp translational fusion ( −210 rel . to AUG + 15 codons of cfa ) from constitutive PLtetO-1 promoterpSC101* , CmR ( Fröhlich et al . , 2013 ) 5CpKD4Template plasmid for KmR mutant constructionoriRγ , AmpR ( Datsenko and Wanner , 2000 ) pKD46γ-β-exotemperature-sensitive plasmid to express λRED-recombinase from arabinose-inducible ParaB promoteroriR101 , AmpR ( Datsenko and Wanner , 2000 ) pCP20FLP–ci857Temperature-sensitive Flp recombinase expression plasmidpSC101 , AmpR , CamR ( Cherepanov and Wackernagel , 1995 ) Hfq wild-typepEH–10– ( hfq ) HfqExpresses wild-type Hfq protein without tagsAmpRKind gift of I . Moll ( Max F . Perutz Laboratories , University of Vienna , Austria ) 2 , 3 , 4 , 6HfqRimpBAD-RimHfq R19A , P21A , Q33A , Q35A , T49A , R66AExpresses mutant of Hfq protein without tagspMB1 , AmpRThis study6Hfq66pBAD-66Hfq amino aids 1-66Expresses truncated Hfq protein without tagspMB1 , AmpRThis studyS5 For plasmids expressing loop mutants of RydC from the PL-promoter , plasmid pKF42-1 served as template for PCR amplification with primer pairs JVO-10915/JVO-10916 ( pPL-RydC-LI; pKF224-9 ) , JVO-10913/JVO-10914 ( pPL-RydC-LII; pKF223 ) , and the linear fragments were purified and self-ligated . Similarly , pPL-RydC-LI/LII ( pKF225 ) was constructed by self-ligation of a PCR product of JVO-10915/10 , 916 using pKF223 as template . Competent E . coli TOP10 were used for all cloning purposes . For the in vivo analyses , bacteria were grown aerobically in L-Broth ( LB ) medium at 37°C . Where appropriate , liquid and solid media were supplemented with antibiotics at the following concentrations: 100 µg/ml ampicillin , 50 µg/ml kanamycin , and 20 µg/ml chloramphenicol . A complete list of bacterial strains employed in this study is provided in Table 4 . Salmonella enterica serovar Typhimurium strain SL1344 ( JVS-0007 ) is referred to as wild-type strain and was used for mutant construction . Single mutant derivatives were constructed by the λRed recombinase one-step inactivation method . To obtain JVS-10701 ( SL1344 hfq70::STOP ) , Salmonella cells carrying the pKD46 helper plasmid were transformed with a DNA fragment amplified from pKD4 using JVO-10909/JVO-10910 . Subsequently , the KanR cassette of λRed-derived mutants was eliminated by transformation with the FLP recombinase expression plasmid pCP20 ( Datsenko and Wanner , 2000 ) . Phage P22 transduction was used to transfer chromosomal modifications to a fresh Salmonella wild-type background as well as to obtain strains carrying multiple mutations . 10 . 7554/eLife . 05375 . 022Table 4 . StrainsDOI: http://dx . doi . org/10 . 7554/eLife . 05375 . 022Trivial nameStockGenotype; relevant markersSource/referenceUsed in FigureSalmonellaJVS-0007SL1344; StrR hisG rpsL xylLaboratory stockΔrydCJVS-0291ΔrydC::KanR ( Papenfort et al . , 2008 ) 1B/C; 5B/C; 7ΔhfqJVS-0584Δhfq ( Sittka et al . , 2007 ) 1B/CΔrydC ΔhfqJVS-10665ΔrydC::KanR Δhfq ( Fröhlich et al . , 2013 ) 7JVS-10701hfq70::STOPThis studyΔrydC hfq70JVS-11150hfq70::STOP ΔrydC::KanRThis study7E . coliTOP10JVS-2000F− mcrA Δ ( mrr-hsdRMS-mcrBC ) Φ80lacZΔM15 ΔlacX74 recA1 araD139 Δ ( ara-leu ) 7697 galU galK rpsL endA1 nupG λ−Invitrogen To prepare whole-cell samples for Western blot analyses , bacteria were collected by centrifugation ( 16 , 000×g; 2 min; 4°C ) , and pellets were resuspended in 1X protein loading buffer ( Fermentas , UK ) to a final concentration of 0 . 01 OD/µl . Samples corresponding to 0 . 1 OD were loaded per lane and resolved by SDS-PAGE , after which proteins were transferred to PVDF membranes as described in Sittka et al . ( 2007 ) . GFP fusion proteins and GroEL were detected using commercially available antibodies ( GFP: 1:5000; mouse; Roche # 11814460001 and GroEL: 1:10 , 000; rabbit; Sigma-Aldrich G6532 ) . Anti-mouse or anti-rabbit secondary antibodies conjugated with horseradish peroxidase ( 1:10 , 000; GE Healthcare ) were used in all cases . Signals were visualized using the Western Lightning reagent ( PerkinElmer , Waltham MA , USA ) and an ImageQuant LAS 4000 CCD camera ( GE Healthcare ) . To monitor RNA half-life of RydC mutant variants , cells were grown to an OD600 of 1 and treated with rifampicin ( final concentration: 500 µg/ml ) to abrogate transcription . RNA samples were withdrawn at indicated time-points , and RNA decay was determined by Northern blot analysis as previously described ( Urban and Vogel , 2007; Fröhlich et al . , 2012 ) . RydC and its variants were detected with the universal oligo JVO-4364; 5S RNA served as loading control ( JVO-0322 ) . Formation of complexes between sRNAs and Hfq in vitro was analysed by gel shift assays . For RNA in vitro synthesis , ∼200 ng of template DNA carrying a T7 promoter sequence was amplified by PCR ( RydC: JVO-4721*JVO-4722 on pKF42-1; RydC-S1: JVO-5165*JVO-4722 on pKF60-1 ) and reverse transcribed and 5′-end-labelled as described previously ( Fröhlich et al . , 2013 ) . Labelled RNA RydC or RydC-S1 ( 4 pmol ) was denatured ( 95°C , 2 min ) , chilled on ice for 5 min , and supplemented with 1× structure buffer and 1 μg yeast RNA . Upon addition of purified Hfq ( concentration as indicated in the figure legends ) or Hfq dilution buffer ( control; 1× structure buffer , 1% ( vol/vol ) glycerol , 0 . 1% Triton X-100 ) , samples were incubated at 37°C for 10 min . Unlabelled competitor RNA was added at the indicated concentrations , and samples were incubated for additional 10 min . Prior to loading , reactions were mixed with native loading buffer ( 50% glycerol , 0 . 5× TBE , 0 . 02% ( wt/vol ) bromophenol blue ) and separated by native PAGE ( 6% PAA ) . Gels were dried and signals were determined on a Typhoon FLA 7000 phosphorimager . | A crucial step in the production of proteins is the translation of messenger RNA molecules . Other RNA molecules called small RNAs are also involved in this process: these small RNAs bind to the messenger RNA molecules to either increase or decrease the production of proteins . Bacteria and other microorganisms use small RNA molecules to help them respond to stress conditions and to changes in their environment , such as fluctuations in temperature or the availability of nutrients . The ability to rapidly adapt to these changes enables bacteria to withstand harmful conditions and to make efficient use of resources available to them . Many small RNA molecules use a protein called Hfq to help them interact with their target messenger RNAs . In some cases this protein protects the small RNA molecules when they are not bound to their targets . Hfq also helps the small RNA to bind to the messenger RNA , and then recruits other enzymes that eventually degrade the complex formed by the different RNA molecules . Previous research has shown that six Hfq subunits combine to form a ring-shaped structure and has also provided some clues about the way in which Hfq can recognise a short stretch of a small RNA molecule , but the precise details of the interaction between them are not fully understood . Now Dimastrogiovanni et al . have used a technique called X-ray crystallography to visualize the interaction between Hfq and a small RNA molecule called RydC . These experiments reveal that a particular region of RydC adopts a structure known as a pseudoknot and that this structure is critical for the interactions between the RydC molecules and the Hfq ring . Dimastrogiovanni et al . find that one RydC molecule interacts with one Hfq ring , and they identify the contact points between the RydC molecule and different regions of the Hfq ring . Based on this information , Dimastrogiovanni et al . propose a model for how the RydC:Hfq complex is likely to interact with a messenger RNA molecule . The next step will be to test this model in experiments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | Recognition of the small regulatory RNA RydC by the bacterial Hfq protein |
The hippocampus is critical for episodic memory and computational studies have predicted specific functions for each hippocampal subregion . Particularly , the dentate gyrus ( DG ) is hypothesized to perform pattern separation by forming distinct representations of similar inputs . How pattern separation is achieved by the DG remains largely unclear . By examining neuronal activities at a population level , we revealed that , unlike CA1 neuron populations , dentate granule cell ( DGC ) ensembles activated by learning were not preferentially reactivated by memory recall . Moreover , when mice encountered an environment to which they had not been previously exposed , a novel DGC population—rather than the previously activated DGC ensembles that responded to past events—was selected to represent the new environmental inputs . This selection of a novel responsive DGC population could be triggered by small changes in environmental inputs . Therefore , selecting distinct DGC populations to represent similar but not identical inputs is a mechanism for pattern separation .
During learning and memory , the hippocampus , a key structure for episodic memory , receives information from the cortex through multiple parallel pathways to each of its main subregions , including the dentate gyrus ( DG ) , CA3 and CA1 ( Squire , 1992; Rolls and Kesner , 2006; Rolls , 2010 ) . The DG receives excitatory inputs from entorhinal cortex ( EC ) layer II neurons via the perforant pathway and relays the information to CA3 through mossy fibers . The CA3 in turn projects to CA1 , which sends back-projections to deep layers of the EC , forming the classic tri-synaptic pathway ( EC→DG→CA3→CA1 ) . CA3 also receives direct inputs from EC through the perforant pathway and there are extensive interconnections among CA3 neurons via recurrent collateral fibers . In addition to inputs from CA3 , CA1 receives inputs directly from EC layer III neurons through the temporoammonic pathway , forming a monosynaptic pathway ( EC→CA1 ) . In this complex neural network , each pathway and each subregion is likely to carry out specific functions during learning and memory . Based on these network connections and the anatomical characteristics of each subregion , theories about specific functions of the individual hippocampal subregions in learning and memory have been proposed by computational modeling ( Rolls and Kesner , 2006; Rolls , 2010 ) . In particular , the DG is postulated to function as a pattern separator by de-correlating inputs from EC ( Marr , 1971 ) because of its sparse activity and its considerably larger population of neurons compared to the EC and CA3 . The pattern separation function of the DG is supported by accumulating evidence from behavioral studies , reporting that animals with lesions or blocked plasticity in the DG were impaired in discriminating similar spatial and contextual information ( Gilbert et al . , 2001; McHugh et al . , 2007; Goodrich-Hunsaker et al . , 2008; Nakashiba et al . , 2012 ) . Nevertheless , how the DG achieves the pattern separation function remains elusive . In vivo physiological recordings of dentate granule cells ( DGCs ) have shown that changes in environmental inputs only evoke the rate remapping of DGCs but not the global remapping predicted by computational models ( Leutgeb et al . , 2007 ) . Through the powerful mossy fiber synapses , outputs of the DG are passed to the downstream recurrent network in CA3 , which is hypothesized to be the site for memory storage ( Treves and Rolls , 1994 ) . Computational studies have suggested that it is advantageous to have two extrinsic afferent systems for the autoassociative network in CA3—one with strong synapses for memory formation and the other with associatively modifiable synapses for memory retrieval ( Treves and Rolls , 1992 ) . Therefore , it has been speculated that the mossy fiber inputs from the DG may be particularly suitable for memory formation , whereas the direct inputs from EC may be responsible for information recall . On the other hand , CA1 is considered to be a feed forward neural network and is the main output region for the hippocampus ( Rolls and Kesner , 2006; Rolls , 2010 ) . Experimental evidence from genetic and physiological studies has demonstrated the importance of CA1 for both memory formation and retrieval ( Riedel et al . , 1999; Dupret et al . , 2010; Goshen et al . , 2011 ) . Because the large size of the DGC population is a key factor for the computational hypothesis of pattern separation , we utilized TetTag transgenic mice to examine the population neuronal activity of the dorsal DG to test whether DGCs undergo global remapping at the population level . To examine the specificity of the responsiveness of the DGCs , population activity in CA1 was also analyzed . Our results revealed a novel mechanism for pattern separation in the DG: the selection of distinct DGC populations to represent different contextual information . In addition , we observed that memory recall preferentially reactivated the neuronal population involved in learning in CA1 but not in the DG , suggesting that , in a complex neural network , memory recall may not reinstate the activities in every pathway involved in memory formation .
We studied the population activity of neurons in the hippocampus by examining the transient expression of immediate early genes ( IEGs , such as Fos , Arc and Egr1 ) , which is commonly used as an indicator of recent neuronal activity ( Guzowski et al . , 2005 ) . To compare the activities in the same neuronal population in response to two events at sequential time points , we used TetTag bi-transgenic mice in which neuronal activities at a given time window can be persistently labeled ( Figure 1A , Figure 1—figure supplement 1; Reijmers et al . , 2007 ) . In these mice , neuronal activity can activate the Fos promoter and induce the expression of tetracycline-controlled transactivator ( tTA ) from the Fos-tTA transgene . In the absence of doxycycline ( dox ) , a drug that binds to tTA and prevents tTA from binding to the tetracycline responsive promoter ( tetO ) , the resulting tTA can activate the expression of the tau-LacZ marker from the transgene: tetO-tau-lacZ:tTA* . At the same time , a tetracycline-insensitive form of transactivator ( tTA*: tTA containing H100Y point mutation ) is also expressed , allowing the persistent tau-LacZ expression irrespective of dox treatment . Thus , if the mice are removed from dox treatment for an initial experience and euthanized shortly after a second experience , the activity of the same neuronal ensemble in response to these two sequential experiences can be assessed by examining the expressions of tau-LacZ and IEGs , which correspond to neuronal activities of the first and second experiences , respectively . 10 . 7554/eLife . 00312 . 003Figure 1 . Induction of tag ( tau-LacZ ) expression by removing dox treatment . ( A ) A brief cartoon illustrating the TetTag transgenic system . ( B ) and ( E ) Experimental designs . Dox treatment is illustrated by blue shading . ( C ) and ( D ) There are few neurons in either CA1 ( C ) or the DG ( D ) ( outlined by DAPI [blue] ) expressing LacZ marker ( green ) if mice are kept on a dox diet until enriched environment ( EE ) exposure . Samples are also stained with FOS ( red ) . Each channel in the inset ( outlined by the square ) is presented below the corresponding overall image , with arrows indicating the LacZ-positive neurons . ( F ) and ( G ) In mice that were removed from dox treatment 2 days before EE exposure , many LacZ-positive neurons can be observed in both CA1 ( F ) and the DG ( G ) . In both subregions , many tagged neurons are also co-stained with FOS . The scale bar in ( D ) represents 50 μm for ( C , D , F , and G ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 00310 . 7554/eLife . 00312 . 004Figure 1—figure supplement 1 . The TetTag system . The TetTag transgenic mice contain two transgenes: one transgene expresses tTA by Fos promoter and the other transgene expresses dox-insensitive tTA* and tau-LacZ downstream of the tetO promoter . Neuronal activity can trigger activation of Fos promoter and lead to the expression of tTA , the transactivator of the tetO promoter . In the presence of dox ( left panel ) , tTA cannot induce the expression of either tau-lacZ or tTA* ( mutated tTA i . e . insensitive to dox ) ( see Neuron A as an example ) . When mice were removed from the dox diet ( middle panel ) , tTA , whose expression is induced by neuronal activity , can bind to the tetO promoter to activate the expression of the tau-lacZ marker gene and tTA* . tTA* and tetO form a transcription feedback loop that can sustain the expression of tau-lacZ even after mice are put back on dox treatment ( see Neuron B ) . Putting mice on dox food will close the time window to mark the activated neurons with tau-lacZ ( Neuron C ) . ( This Figure is adapted from Figure 1A in Reijmers et al . , 2007 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 00410 . 7554/eLife . 00312 . 005Figure 1—figure supplement 2 . Induction of tau-LacZ expression in the hippocampus by removing mice from dox treatment . ( A ) Very few neurons in the hippocampus are labeled by tau-LacZ ( green ) in mice maintained on dox diet . DAPI ( blue ) staining is used to show the anatomy of the hippocampus . The suprapyramidal blade and the infrapyramidal blade of the DG are labeled . ( B ) Removing mice from dox treatment effectively induces tau-LacZ expression in the DG and CA1 but not CA3 of the hippocampus . ( C ) and ( D ) Confocal images showing that few CA3 neurons are labeled by LacZ marker under either on Dox or off Dox conditions . The yellow arrows in ( D ) point to the LacZ-labeled mossy fibers , the axons of DGCs projecting to CA3 . The scale bar in ( A ) denotes 200 μm for panels ( A , B ) . The scale bar in ( D ) denotes 100 μm for panels ( C , D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 00510 . 7554/eLife . 00312 . 006Figure 1—figure supplement 3 . Quantification of activities and FOS intensity in mice exposed to an enriched environment during the dox-off window . ( A ) Quantification the numbers of FOS-positive , LacZ-positive and FOS+LacZ double positive cells in CA1 and the DG . Asterisk indicates the numbers of DAPI-positive cells in the DG are estimated values ( see 'Materials and methods' ) . ( B ) and ( C ) The intensity of FOS staining between LacZ-positive and LacZ-negative cells is not significantly different in CA1 ( B: frequency distribution; C: t184 = 0 . 864 , p>0 . 38 ) . ( D ) and ( E ) The intensity of FOS staining between LacZ-positive and LacZ-negative cells is not significantly different in the DG ( D: frequency distribution , E: t158 = 0 . 873 , p>0 . 38 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 006 First , we tested whether expression of tau-LacZ markers in the hippocampus of TetTag mice could be regulated by dox . We exposed mice to an enriched environment under either a dox-on or dox-off condition ( Figure 1B , E , see ‘Materials and methods' ) and found that removing dox treatment effectively induced tau-LacZ expression in neurons of the DG and CA1 ( Figure 1B–G , Figure 1—figure supplement 2A–B ) , with most of the LacZ-positive neurons displaying typical morphologies of granule cells and pyramidal neurons in the DG and CA1 , respectively ( Figure 1F , G ) . Furthermore , many LacZ-positive cells also co-expressed FOS , with over 70% and 85% of LacZ-positive cells expressing FOS in the DG and CA1 , respectively , suggesting that the expression of LacZ did not affect the expression of IEGs in the same neuron ( Figure 1F , G , Figure 1—figure supplement 3A ) . It was also notable that the efficiency of tagging was low , compared to the endogenous FOS labeling , particularly in the CA1 region . A low efficiency of tagging was also observed in basolateral amygadala ( Reijmers et al . , 2007 ) . This low and variable induction efficiency across brain regions was possibly caused by low penetrance and variable expressivity of the transgenes , a common problem for transgenic mice . To test if the tagged population represents activities in the general population , we measured the intensity of FOS staining in the LacZ-positive and LacZ-negative neurons . In both DG and CA1 , the FOS intensity was similar between LacZ-positive and LacZ-negative populations ( Figure 1—figure supplement 3B–E ) . Therefore , it was likely that LacZ tagged neurons were representatives of the activated population , although we could not formally rule out the possibility that only a specific population of activated neurons ( e . g . the population with the highest activities ) could be tagged . The induction efficiency was even lower in CA3 with few neurons tagged ( Figure 1—figure supplement 2 ) , preventing further analysis in this region . We next tested the activity-dependent expression of tau-LacZ markers in TetTag mice . After removing them from dox treatment , we exposed some mice to a fear conditioning chamber ( ctxA , Figure 2—figure supplement 4 ) and kept others in their home cage ( HC ) ( ‘Materials and methods' ) . While LacZ-positive neurons could be readily detected in both CA1 and the DG in the HC mice ( Figure 2A , B ) , substantially more LacZ tagged neurons were observed in the ctxA mice in both CA1 and the DG ( Figure 2C–F; t-test , in CA1 , HC , 1 . 5 ± 0 . 5% , n = 4; ctxA , 5 . 1 ± 0 . 5% , n = 3; p<0 . 007; in the DG , HC , 1 . 9 ± 0 . 7%; ctxA , 6 . 9 ± 1 . 0%; p<0 . 016 ) . Therefore , the dox-regulated and activity-dependent expression of LacZ in both the DG and CA1 suggested the feasibility of studying neuronal activities at a population level in these hippocampal subregions using TetTag mice . 10 . 7554/eLife . 00312 . 007Figure 2 . Activity-dependent induction of tag ( tau-LacZ ) expression . ( A ) and ( B ) The expressions of LacZ marker ( green ) in CA1 ( A ) and DG ( B ) of the mice in the home cage ( HC ) group . The overall anatomies are highlighted by the DAPI staining ( blue ) . ( C ) and ( D ) The expressions of LacZ marker in CA1 ( C ) and the DG ( D ) of the mice in the context A ( ctxA ) group . ( E ) and ( F ) Quantification demonstrates that the numbers of LacZ-positive neurons are significantly higher in the ctxA group compared to the HC group in both CA1 ( E ) and the DG ( F ) . The scale bar in ( D ) represents 100 μm for panels ( A–D ) . Asterisk indicates statistically significant difference between groups . Data are shown as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 00710 . 7554/eLife . 00312 . 008Figure 2—figure supplement 1 . Contexts used for contextual fear conditioning . Context A' and B are modified from context A . Context C is completely different from context A and is located in another test room . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 008 To study the activity of neuronal populations in the DG and CA1 during event learning and subsequent memory recall , we used a contextual fear conditioning paradigm combining contextual pre-exposure and immediate foot shock ( Fanselow , 1990 , 2000 , 2010 ) . This is a task in which the hippocampus has been demonstrated to be critically involved in forming a conjunctive representation of the conditioning context during pre-exposure ( Barrientos et al . , 2002; Rudy et al . , 2002; Stote and Fanselow , 2004 ) . We chose this task because the formation of the contextual memory , which is dependent on the hippocampus , can be temporally separated from the subsequent context-shock association , which presumably relies mostly on the function of amygdala ( Rudy and O'Reilly , 2001; Rudy et al . , 2002; Reijmers et al . , 2007; Han et al . , 2009 ) . With dox treatment removed , we pre-exposed one group of mice ( preA , n = 12 ) to the fear conditioning chamber ( context A ) to tag the activated neurons ( LacZ+ ) in contextual learning ( Figure 3A , Figure 2—figure supplement 1 ) . After the last pre-exposure ( on day 5 ) , mice were put back on dox treatment to prevent further tagging . 2 days later , mice were subjected to immediate shock in the conditioning chamber and their conditioned fear memory was tested 1 day after immediate shock . Mice were perfused shortly after the memory test for neuronal activity analysis . For comparison , another group of mice ( preC , n = 11 ) was subjected to the identical protocol except that they were pre-exposed to an environment ( context C , Figure 2—figure supplement 1 ) that was completely different from the conditioning chamber . Because the mice associated contextual information during pre-exposure with the subsequent aversive stimulus ( i . e . foot shock ) in this protocol , it was not surprising that preA mice but not preC mice displayed a high level of freezing behavior when the mice were tested for their conditioned response in context A ( Figure 3B; t-test , t21 = 3 . 424 , p<0 . 0026 ) . 10 . 7554/eLife . 00312 . 009Figure 3 . Memory recall preferentially reactivates the neuron population in response to learning in CA1 but not in the DG . ( A ) The pre-exposure-immediate shock paradigm for contextual fear conditioning . Dox treatment is illustrated by blue shading . Contextual learning mainly takes place during pre-exposure in the absence of dox treatment . LacZ and IEGs ( FOS or EGR1 ) are regarded as indicators of learning-induced activity and retrieval-induced activity , respectively . Dox treatment is illustrated by blue shading . ( B ) preA mice display significantly more freezing behavior than preC mice . ( C ) and ( D ) During pre-exposure , the proportions of LacZ-positive neurons in either CA1 ( C ) or the DG ( D ) are not significantly different between preA and preC mice . ( E ) During the retrieval test , preferential reactivation of the LacZ-positive population in CA1 is revealed by quantifying the percentage of FOS-positive neurons in the total population ( activation rate ) and the percentage of LacZ-FOS double-positive cells in the LacZ-positive population ( reactivation rate ) . ( F ) There is no preferential reactivation of LacZ-positive DGCs in preA mice , whereas LacZ-positive DGCs are significantly less likely to be reactivated in preC mice compared to preA mice . The reactivation rate is not significantly different from the activation rate in preA mice but is significantly lower than the activation rate in preC mice . ( G ) Reactivation indexes suggesting the differential reactivations of learning-induced neuronal ensembles by recall in CA1 and the DG ( ANOVA: region x group interaction , F1 , 1 = 5 . 016 , p<0 . 037; main region effect , F1 , 21 = 24 . 49 , p<0 . 0001; main group effect , F1 , 21 = 50 . 10 , p<0 . 0001 ) . Asterisk indicates statistically significant difference between groups . Hash indicates statistically significant difference from chance . Data are shown as mean ± SEM ( ns: no significant difference; HC: home cage; sac: sacrifice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 00910 . 7554/eLife . 00312 . 010Figure 3—figure supplement 1 . Representative confocal images illustrating the expression of IEGs and LacZ in CA1 ( tau-LacZ in green , FOS in red , RBFOX3 in blue ) and the DG ( tau-LacZ in green , EGR1 in red , DAPI in blue ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 01010 . 7554/eLife . 00312 . 011Figure 3—figure supplement 2 . Quantification of the entire z-series of confocal images in the DG for the contextual fear conditioning experiment 1 . ( A ) The level of LacZ induction by pre-exposure is not significantly different between preA and preC mice ( t21 = 0 . 4071 , p>0 . 64 ) . ( B ) During the recall test , there is no preferential reactivation of LacZ-positive DGCs in preA mice , whereas LacZ-positive DGCs are significantly less likely to be reactivated in preC mice ( ANOVA: group x activity rates interaction , F1 , 1 = 22 . 40 , p<0 . 0001; almost significant main effect of activity rates , F1 , 21 = 4 . 201 , p=0 . 053; almost significant main effect of group , F1 , 21 = 4 . 183 , p=0 . 053 ) . The reactivation rate is not significantly different from the activation rate in preA mice but is significantly lower than the activation rate in preC mice ( Bonferroni post hoc test , activation rate vs reactivation rate , p>0 . 05 for preA and p<0 . 001 for preC ) . In addition , the reactivation rate ( EGR1+LacZ/LacZ% ) , but not the activation rate ( EGR1% ) , in preC mice is significantly lower than that in preA mice ( Bonferroni post hoc test , preA vs preC , p<0 . 01 for reactivation rates and p>0 . 05 for activation rates ) . ( C ) The reactivation index is significantly higher in preA mice compared to the preC group ( t-test , t21 = 4 . 277 , p<0 . 0001 ) . Moreover , the reactivation index is not significantly different from chance in preA mice but is significantly below chance in preC mice ( one sample t-test , chance = 0: preA , t11 = 1 . 582 , p>0 . 14; preC , t10 = 6 . 316 , p<0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 011 To investigate the activities in neuron populations of the DG and CA1 , we concurrently examined the expression of tau-LacZ and the expression of IEGs to evaluate the neuronal activities during contextual pre-exposure and during memory recall test , respectively ( Figure 3—figure supplement 1 , see ‘Materials and methods' ) . We focused our analysis on the dorsal hippocampus , because this region has been shown to be tightly associated with learning and memory . For technical convenience , FOS and EGR1 were used as markers to assess the recall-activated neurons in CA1 and the DG , respectively , and we designated the percentage of IEG positive neurons in the total numbers of neurons quantified as the activation rate ( ‘Materials and methods' ) . To measure the proportion of the neurons that were activated by the recall test in the neuronal population that was previously activated during pre-exposure , we quantified the percentage of LacZ+IEG double positive neurons in the LacZ tagged population ( designated as the reactivation rate ) . We were not able to detect a significant difference in either CA1 or the DG in the percentage of LacZ positive neurons in the total numbers of neurons quantified between preA and preC mice ( Figure 3C , D; t-test , CA1: t21 = 0 . 5005 , p>0 . 62; DG: t21 = 0 . 8504 , p>0 . 40 ) , suggesting that contexts A and C had equivalent simulating effects . To investigate how the neurons involved in memory formation responded to subsequent memory recall , we compared the reactivation rates to the corresponding activation rates . In CA1 , whether or not the neurons that were activated during pre-exposure were preferentially activated again by the recall test in context A depended on the identity of the pre-exposure context ( Figure 3E; ANOVA: group x activity rates interaction , F1 , 1 = 11 . 60 , p<0 . 0027; main effect of activity rates , F1 , 21 = 44 . 04 , p<0 . 0001; main effect of group , F1 , 21 = 8 . 238 , p<0 . 0092 ) . Because both groups of mice were tested in context A , there was no significant difference in the activation rate between preA and preC mice , as expected ( Figure 3E , Bonferroni post hoc test , p>0 . 05 ) . In contrast , the reactivation rate of preA mice was significantly higher than that of preC mice ( Figure 3E , Bonferroni post hoc test , p<0 . 001 ) , indicating that a previous learning experience affected neuronal responses at the time of memory recall . In preA mice , which underwent pre-exposure and retested in context A , the reactivation rate was significantly higher than the activation rate ( Figure 3E , Bonferroni post hoc test , p<0 . 0001 ) , suggesting that CA1 neurons that were activated during learning were preferentially reactivated by subsequent memory recall . By contrast , in preC mice , neurons responding to context C during pre-exposure were not preferentially activated by the subsequent test in context A ( Figure 3E , Bonferroni post hoc test , activation rate vs reactivation rate in preC mice , p>0 . 05 ) , suggesting that recall-induced preferential reactivation of the CA1 neuron population involved in memory formation depended on retrieval of the same memory trace . We further quantified the degree of this reactivation preference by a reactivation index , which normalized the reactivation rate by the corresponding activation rate ( ‘Materials and methods' ) . In CA1 , the reactivation index in preA but not preC mice was significantly above chance ( Figure 3G; one sample t-test , chance = 0: preA , t11 = 12 . 12 , p<0 . 0001; preC , t10 = 1 . 337 , p>0 . 20 ) , with the index of preA mice being significantly higher than that of preC mice ( Figure 3G; t-test , t21 = 3 . 115 , p<0 . 0048 ) . These data allow us to propose that the CA1 neuronal ensemble that is responsible for contextual learning is likely reinstated for the recall of the same memory trace . The consistency of this finding with previous reports that showed that CA1 is involved in both memory formation and retrieval ( Riedel et al . , 1999; Goshen et al . , 2011 ) further validates our methodology of using TetTag mice for population neuronal activity study in the hippocampus . In contrast to CA1 , memory recall did not induce the preferential reactivation of the population of DGCs that was activated during learning , as indicated by the similar activation rate and reactivation rate in preA mice ( Figure 3F; ANOVA: group x activity rates interaction , F1 , 1 = 18 . 51 , p<0 . 0003; no main effect of activity rates , F1 , 21 = 2 . 910 , p=0 . 1028; main effect of group , F1 , 21 = 6 . 213 , p<0 . 022; Bonferroni post hoc test , activation rate vs reactivation rate in preA , p>0 . 05 ) . To our surprise , in preC mice , the reactivation rate of DGCs was significantly lower than the activation rate ( Figure 3F; Bonferroni post hoc test , p<0 . 001 ) , indicating that the DGC population responding to context C was significantly less likely to be activated by context A compared to the general DGC population . Compared to activation rates that were not significantly different between preA and preC mice ( Figure 3F; Bonferroni post hoc test , p>0 . 05 ) , the reactivation rate of preC mice was significantly lower than that of preA mice ( Figure 3F; Bonferroni post hoc test , p<0 . 001 ) . Thus , rather than the DGC population responding to context C , preC mice activated a different population of DGCs in response to context A . These results were further confirmed by the analysis of reactivation indexes . The reactivation index was significantly higher in preA mice compared to that of preC mice ( Figure 3G; t-test , t21 = 5 . 032 , p<0 . 0001 ) , with the index in preC but not preA mice significantly below chance ( Figure 3G; one sample t-test , chance = 0: preA , t11 = 1 . 550 , p>0 . 14; preC , t10 = 5 . 314 , p<0 . 0003 ) . To substantiate these results , we re-analyzed the data in the DG by quantifying the activities in the entire z-series of confocal images and obtained similar results ( see ‘Materials and methods' , Figure 3—figure supplement 2 ) . Given that the quantification of the entire z-series increased the sampling size , all subsequent analyses were carried out using this approach . Furthermore , we re-measured the activation and reactivation rates in a subset of preA and preC mice , using the expression of FOS as the indicator for DGC activities in the recall test . Similar results were obtained using either FOS or EGR1 as activity indicators in the same cohort of mice ( Figure 4 ) . In summary , these analyses of the population activities of DGCs demonstrated that neurons in the DG and CA1 responded differently during memory processing . Unlike CA1 pyramidal neurons , DGCs activated by learning an event were not preferentially reactivated by retrieving the same memory . Instead , distinct ensembles of DGCs were selected in response to different events . 10 . 7554/eLife . 00312 . 012Figure 4 . Similar results are obtained in the activity analysis of DGCs using either FOS or EGR1 as IEG markers in the same cohort of mice . ( A ) Activity analysis using FOS as IEG marker and RBFOX3 as neuronal marker demonstrates the selection of different populations of DGCs to represent different environmental inputs ( ANOVA: group x activity rate interaction , F1 , 1 = 9 . 038 , p<0 . 017; Bonferroni post hoc test , reactivation rate vs activation rate , p>0 . 05 for preA mice , p<0 . 05 for preC mice; preA , n = 6; preC , n = 4 ) . ( B ) Reactivation index calculated from the analysis using FOS as IEG marker . The index in preC is significantly smaller than preA ( t-test , t8 = 3 . 911 , p<0 . 0045 ) and the chance level ( one sample t-test , chance = 0 , t3 = 3 . 558 , p<0 . 038 ) , whereas the index in preA is not different from chance ( one sample t-test , chance = 0 , t5 = 0 . 6153 , p>0 . 56 ) . ( C ) Activity analysis using EGR1 as IEG marker in the same cohort of mice has similar activity pattern as those analyzed by FOS ( ANOVA: group x activity rate interaction , F1 , 1 = 7 . 405 , p<0 . 026; Bonferroni post hoc test , reactivation rate vs activation rate , p>0 . 05 for preA mice , p<0 . 05 for preC mice ) . The numbers of DGCs in the granule cell layers were quantified from DAPI images . ( D ) Reactivation index calculated from the analysis using EGR1 as IEG marker . The index in preC is significantly smaller than preA ( t-test , t8 = 3 . 519 , p<0 . 0079 ) and the chance level ( one sample t-test , chance = 0 , t3 = 4 . 403 , p<0 . 022 ) , whereas the index in preA is not different from chance ( one sample t-test , chance = 0 , t5 = 0 . 6815 , p>0 . 52 ) . Asterisk indicates statistically significant difference between groups . Hash indicates statistically significant difference from chance . Data are shown in mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 012 To determine how these results could be affected by HC activity , an inevitable part of both pre-exposing and re-exposing experiences , and whether the emotional value of the learned context was critical for population reactivation , we performed a new experiment with two modifications of the previous procedures . First , one group of mice ( HC mice , n = 4 ) were kept in their HC without exposure to any context during the dox-off window while the other group was exposed to context A ( ctxA , n = 7 ) ; second , the immediate-shock procedure was omitted so that the pre-exposed context remained emotionally neutral for animals at the re-exposure ( Figure 5A ) . After all mice were put back on dox treatment , the HC mice were subsequently re-exposed to context A , whereas ctxA mice were further divided into two groups and re-exposed to either context A ( ctxA/A , n = 3 ) or C ( ctxA/C , n = 4 ) . 10 . 7554/eLife . 00312 . 013Figure 5 . Neither home cage activity nor emotional value of context has a significant impact on reactivation patterns in CA1 and DG . ( A ) The experimental design . Dox treatment is illustrated by blue shading . ( B ) HC mice have a significantly lower number of LacZ-positive cells in CA1 compared to ctxA mice . ( C ) Preferential reactivation of CA1 neurons responding to pre-exposure by re-exposure occurs only in ctxA/A mice but not HC or ctxA/C mice . ( D ) HC mice have a significantly lower number of LacZ-positive cells in the DG compared to ctxA mice . ( E ) In HC and ctxA/C mice but not ctxA/A , the reactivation rate is significantly lower than the corresponding activation rate in the DG , suggesting different populations of DGCs are selected in response to distinct experiences ( F ) reactivation indexes analysis . Asterisk indicates statistically significant difference between groups or rates . Hash indicates statistically significant difference from chance . Data are shown as mean ± SEM ( HC: home cage; sac: sacrifice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 013 Consistent with the findings described in Figure 2 , exposure to context A resulted in higher levels of LacZ induction in both CA1 and DG ( Figure 5B , D; t-test , HC vs ctxA , in CA1 , t9 = 3 . 578 , p<0 . 006; in DG , t9 = 3 . 131 , p<0 . 013 ) . In ctxA/A mice , which were pre-exposed and re-exposed to the same context , the reactivation rates in CA1 were significantly higher than the activation rates , suggesting preferential activation of neurons that responded during pre-exposure by re-exposure; however , this preferential reactivation was not found in either HC or ctxA/C mice , whose experiences at pre-exposure and re-exposure were different ( Figure 5C; ANOVA: group x activity rates interaction , F2 , 1 = 13 . 99 , p<0 . 0024; main effect of group , F2 , 8 = 15 . 68 , p<0 . 0017; main effect of activity rates , F1 , 8 = 18 . 59 , p<0 . 0026; Bonferroni post hoc test , for reactivation rate vs activation rate , p<0 . 001 in ctxA/A and p>0 . 05 in HC and ctxA/C; for reactivation rate , HC vs ctxA/A , p<0 . 001; ctxA/A vs ctxA/C , p<0 . 0001 ) . In the DG , there was no preferential reactivation of neurons which were activated by pre-exposure in ctxA/A mice , whereas the reactivation rates were significantly lower than the corresponding activation rates in HC and ctxA/C mice ( Figure 5E; ANOVA: group x activity rates interaction , F2 , 1 = 12 . 98 , p<0 . 0031; no main effect of group; main effect of activity rates , F1 , 8 = 73 . 79 , p<0 . 0001; Bonferroni post hoc test , for reactivation rate vs activation rate , p<0 . 001 in HC and ctxA/C and p>0 . 05 in ctxA/A; planned comparisons for reactivation rate , HC vs ctxA/A , p<0 . 05 , ctxA/A vs ctxA/C , p=0 . 055 ) . These results were further confirmed by the analysis of the reactivation index ( Figure 5F; two-way ANOVA: group x region interaction , F2 , 1 = 7 . 740 , p<0 . 012; main effect of group , F2 , 8 = 18 . 72 , p<0 . 001; main effect of region , F1 , 8 = 167 . 8 , p<0 . 0001; compared to chance by one sample t-test: ctxA/A in CA1 , t2 = 7 . 241 , p<0 . 0185; HC in DG , t3 = 8 . 477 , p<0 . 0034; ctxA/C in DG , t3 = 6 . 909 , p<0 . 0062 ) . These data indicated that the pattern of neuronal activation and reactivation of HC mice was drastically different from that of ctxA/A mice , with reactivation rates in HC mice significantly lower than those of ctxA/A mice in both CA1 and the DG . Thus , HC activity does not seem to have a dramatic impact on the population reactivation pattern in CA1 and the DG . In addition , a similar activity pattern was found in ctxA/A and ctxA/C mice compared to that of preA and preC mice in the fear conditioning experiment ( Figure 3 ) , suggesting that the emotional value of contexts did not drastically influence the neuronal activity in CA1 or the DG of the hippocampus . Because the DG was postulated to function as a pattern separator to form distinct representations of similar inputs ( Marr , 1971; O'Reilly and McClelland , 1994; Rolls and Kesner , 2006; Rolls , 2010 ) , we asked whether small changes in contextual inputs might affect the selection of responding neuron populations in the DG . We trained a new cohort of mice for contextual fear conditioning in context A and subsequently tested them in either context A ( testA , n = 10 ) or context B ( testB , n = 11 ) ; the latter was modified from but still shared many common components with context A ( similar but not the same ) ( Figure 6A , Figure 2—figure supplement 1 , see ‘Materials and methods' ) . testA mice displayed a higher level of freezing than testB mice ( Figure 6B; t-test , t19 = 2 . 123 , p<0 . 047 ) , suggesting that mice were able to detect the small changes in context . 10 . 7554/eLife . 00312 . 014Figure 6 . Population activities in the DG but not CA1 are sensitive to small changes in environmental inputs . ( A ) Mice subjected to the pre-exposure-immediate shock paradigm in context A were tested for memory retrieval in either context A ( testA ) or context B ( testB ) , which was modified from context A . Dox treatment is illustrated by blue shading . ( B ) testA mice display significantly more freezing behavior than testB mice . ( C ) and ( D ) During pre-exposure , the percentage of LacZ-positive neurons in total population is not significantly different between testA and testB mice in either CA1 ( C ) or the DG ( D ) . ( E ) Activity of CA1 neurons during retrieval test . While neither activation rates nor reactivation rates are significantly different between groups , reactivation rates are significantly higher than the activation rates in both testA and testB mice . ( F ) During the retrieval test , there is no preferential reactivation of LacZ-positive DGCs in testA mice , whereas LacZ-positive DGCs are significantly less likely to be reactivated in testB mice compared to testA mice . The reactivation rate is significantly lower than the corresponding activation rate in testB mice but not in testA mice . ( G ) Reactivation indexes suggesting the differential reactivations of learning-induced neuronal ensembles by recall in CA1 and the DG ( ANOVA: region x group interaction , F1 , 1 = 62 . 98 , p<0 . 0001; main region effect , F1 , 19 = 215 . 4 , p<0 . 0001; main group effect , F1 , 19 = 25 . 45 , p<0 . 0001 ) . Asterisk indicates statistically significant difference between groups . Hash indicates statistically significant different from chance . Data are shown as mean ± SEM ( ns: no significant difference; HC: home cage; sac: sacrifice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 014 We then examined neuronal activities and found that equivalent numbers of LacZ positive neurons were tagged in testA and testB mice in both CA1 and the DG ( Figure 6C , D; t-test , CA1: t19 = 0 . 2054 , p>0 . 83; DG: t19 = 1 . 319 , p>0 . 20 ) . ANOVA analysis of the activity rates in CA1 revealed that reactivation rates in both testA and testB mice were significantly higher than the corresponding activation rates , and there was no significant difference in either activation rates or reactivation rates between testA and testB mice ( Figure 6E; ANOVA: main effect on activity rates , F1 , 19 = 176 . 2 , p<0 . 0001; no group effect , F1 , 19 = 0 . 01493 , p>0 . 90; no group x activity rate interaction , F1 , 1 = 0 . 1676 , p>0 . 68; Bonferroni post hoc test , activation rates vs reactivation rates , testA , p<0 . 0001 , testB , p<0 . 0001; Bonferroni post hoc test , testA vs testB , activation rate , p>0 . 05 , reactivation rate , p>0 . 05 ) . Moreover , the reactivation indexes for CA1 were not significantly different between testA and testB mice ( Figure 6G; t-test , t19 = 0 . 4206 , p>0 . 67 ) and were above chance in both groups of mice ( Figure 6G; one sample t-test , chance = 0: testA , t9 = 16 . 58 , p<0 . 0001; testB , t10 = 6 . 629 , p<0 . 0001 ) . These results extended our previous finding and indicated that recall-evoked preferential reactivation of CA1 neurons that were responsive during memory formation was resistant to perturbation by small alterations in environmental inputs . In contrast to CA1 , there was a significant interaction between group and activity rates in the DG ( Figure 6F; ANOVA: group x activity rate interaction , F1 , 1 = 36 . 94 , p<0 . 0001 , main effect on activity rates , F1 , 19 = 42 . 19 , p<0 . 0001; no group effect , F1 , 19 = 0 . 8841 , p>0 . 35 ) . Similar to the results of the previous fear conditioning experiment ( Figure 3F ) , the reactivation rate in the DG was significantly lower than the corresponding activation rates in testB but not testA mice ( Figure 6F , Bonferroni post hoc test , activation rates vs reactivation rates , testA , p>0 . 05 , testB , p<0 . 0001 ) and the reactivation rate , but not the activation rate , of the testB mice was significantly lower than that of testA mice ( Figure 6F , Bonferroni post hoc test , testA vs testB , activation rate , p>0 . 05 , reactivation rate , p<0 . 01 ) . Moreover , the reactivation index in testB mice was significantly below chance ( Figure 6G; one sample t-test , chance=0: t10 = 8 . 321 , p<0 . 0001 ) and was significantly lower than that in testA mice ( Figure 6G; t-test , t19 = 6 . 810 , p<0 . 0001 ) , which was not significantly different from chance ( Figure 6G; one sample t-test , chance = 0: t9 = 0 . 4784 , p>0 . 64 ) . These results demonstrate that small environmental changes were enough to evoke responses in distinct ensembles of DGCs but not CA1 neurons ( Figure 6G ) , indicating that this selection of a unique population of DGCs to represent a particular event serves as a mechanism for the function of pattern separation .
By examining neuronal activity at the population level , we discovered that the DG and CA1 of the hippocampus displayed differential neuronal responses at a population level during learning and memory ( see Figure 7 for a model ) . In particular , our data revealed that the selection of separated populations of DGCs in the dorsal DG to represent similar but non-identical environmental inputs was a mechanism for pattern separation . 10 . 7554/eLife . 00312 . 015Figure 7 . A model for population codes in CA1 and the DG during learning and memory . Experience and learning of an event ( event 1 , green ) evoke activities in ensembles of neurons in CA1 and the DG ( green cells ) . When mice subsequently encounter the same event , which will most likely induce memory recall ( event 2 = event 1 ) , the population of CA1 neurons responding to event 1 is preferentially reactivated ( red cells ) , whereas DGCs responding to event 1 are reactivated at chance level ( event 1-responsive DGCs have neither an advantage nor a disadvantage to be reactivated compared to the total DGC population ) . Neurons that are responsive to both events are in yellow . When mice encounter a second event that is similar but not identical to event 1 ( event 2 ≈ event 1 ) , there is still a preference to activate the CA1 neurons that are activated by event 1 . However , in the DG , another population of DGCs that does not respond to event 1 will likely be selected to respond to event 2 . Hence , small changes in inputs can evoke a population code change in the DG but not CA1 , providing a neural basis for the pattern separation function of the DG . When mice encounter a second event that is drastically different from event 1 ( event 2 ≠ event 1 ) , CA1 neurons responding to event 1 are activated at chance level , whereas DGCs that did not respond to event 1 are selected to encode event 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 015 In the DG , distinct populations of DGCs that had limited overlaps were selected to represent two different events that were temporally separated ( Figure 7 ) . Moreover , the utilization of a separated DGC ensemble for encoding newly encountered events could be triggered by small changes in the environmental inputs ( Figures 6F , G and 7 ) . In contrast , CA1 network reactivation was not sensitive to the minor contextual alterations but could be affected by large input changes ( Figures 6F , G and 7 ) . The notion that different populations of DGCs are used to represent different inputs has also been suggested by a computation model based on data obtained by cellular compartment analysis of temporal activity by fluorescence in situ hybridization ( catFISH ) of Arc , another IEG ( Chawla et al . , 2005 ) . Consistent with our results , studies have shown that lesions in the DG but not in CA1 caused a deficit in discrimination of spatial locations of low but not high separations ( Gilbert et al . , 2001; Goodrich-Hunsaker et al . , 2008 ) . Similarly , blocking the plasticity in the DG resulted in a deficit in the discrimination of similar contexts ( McHugh et al . , 2007 ) . Because DGCs are heavily innervated by local and hilar interneurons ( Houser , 2007 ) , inhibition of DGCs by these interneurons can be a potential neural mechanism underlying the population selection; future studies are needed to test this possibility . Our findings seem to disagree with previous results of physiological studies showing that the same ensemble of DGCs was active in multiple different environments despite displaying distinct firing patterns ( i . e . , rate remapping but not global remapping; Leutgeb et al . , 2007; Alme et al . , 2010 ) . One possible explanation for this discrepancy is that the physiological experiments and the experiments described here targeted different groups of neurons in the DG . Despite the fact that the identity of the neurons monitored by in vivo recording cannot be determined by simple histological analysis ( Neunuebel and Knierim , 2012 ) , it is postulated that the recorded neurons are likely to be newly born DGCs that are generated by adult neurogenesis ( Alme et al . , 2010; Neunuebel and Knierim , 2012 ) , because the newborn DGCs are more excitable compared to their mature counterparts and are more likely to be recorded ( Deng et al . , 2010; Aimone et al . , 2011 ) . On the other hand , both mature and newly born DGCs were included in our analysis , with the mature DGCs representing the majority of the population ( >90% ) due to the low rate of adult neurogenesis ( Cameron and McKay , 2001 ) . In a preliminary effort to test this possibility , we measured the distances of LacZ-tagged and EGR1-positive DGCs from the hilus and compared these distances to those of adult-born DGCs because adult-born DGCs tend to be located in the inner third of the granule cell layer ( Mathews et al . , 2010 ) . While the majority of adult-born DGCs labeled by BrdU were located close to the hilus , the LacZ-positive and EGR1-positive DGCs were distributed throughout the granule cell layer and their locations were significantly further from the hilus compared to those of adult-born DGCs ( Figure 8 ) , suggesting that they represented a DGC population different from the adult-born DGCs . Future studies are needed to investigate whether responses of adult-born DGCs in learning and memory are different from those of their mature counterparts , even though it has been shown that the adult-born DGCs are important for spatial discrimination in mice ( Clelland et al . , 2009; Creer et al . , 2010 ) . In addition , the vast difference in kinetics between the in vivo recording studies and our study may also contribute to the inconsistency in the results . Leutgeb et al . ( 2007 ) studied the responses of DGCs to events that occurred minutes apart; however , there was a three-day interval between pre-exposure and re-exposure in our experiments . It is possible that the same group of neurons is recruited to encode for events occurring within a short time interval . Neurons that responded to one event had elevated levels of CREB1 for a short period of time , making them more likely to be recruited by another event occurring in this time window ( Silva et al . , 2009 ) . Finally , although the expression of IEGs can reflect general activation of neurons , it remains unclear what physiological changes the expression of IEGs is corresponding to . It is possible for firing patterns to vary within the IEG positive population . Hence , our findings , together with data from physiological studies , suggest that the DG can carry out pattern separation through both global remapping and rate remapping . 10 . 7554/eLife . 00312 . 016Figure 8 . Comparison of the location of the LacZ-positive and EGR1-positive DGCs with that of adult-born granule cells in the granule cell layer of the DG . The distance of each cell from the hilus was measured using Metamorph . Adult-born granule cells were labeled by treating mice with water containing BrdU for one week . Treated mice were perfused more than 6 weeks later for histological examination of the locations of BrdU-labeled cells in the granule cell layer . ( A ) Frequency distribution showing that the majority of the BrdU-positive cells are located close to the hilus , whereas both LacZ-positive and EGR1-positive populations were distributed across the granule cell layer . ( B ) The distance from the hilus is significantly shorter in BrdU-positive cells compared to that of the EGR1-positive or LacZ-positive cells ( ANOVA: F2 , 392 = 120 . 6 , p<0 . 0001; Bonferroni post hoc test , p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 016 Compared with the situation when the mice experience two different events , when mice encounter a previously experienced event for a second time ( memory recall ) , there is an elevation in the reactivation level of the DGC population that was activated during the initial event learning . Although this level of reactivation did not rise above chance we detected a weak but significant correlation between the reactivation index and behavioral performance ( Figure 9 ) . This observation raises questions regarding which cortical-hippocampal pathway is reinstated by memory recall as well as whether reinstating the DG engram is sufficient and/or necessary for recall . A recent study showed that artificial reactivation of the DGCs involved in the acquisition of contextual fear conditioning was sufficient to induce the expression of fear memory in a neutral context ( Liu et al . , 2012 ) , but the extent of reactivation adequate for memory recall remains unknown . Our findings suggest the possibility that a mild increase in the reactivation by releasing a DGC ensemble from suppression seems enough to trigger the successful memory retrieval and expression . On the other hand , the chance level of reactivation of a learning-induced DGC population by recall suggests an alternative possibility: that preferential reactivation of the DG may not be necessary for memory recall . Because of the existence of multiple parallel pathways between the cortex and hippocampus , it is conceivable that memory retrieval may not necessarily rely on the EC→DG→CA3 pathway . Although this hypothesis remains to be tested directly , several lines of evidence support it . First , lesion of the DG affects encoding but not retrieval of spatial information , as indicated by behavioral studies ( Lassalle et al . , 2000; Lee and Kesner , 2004 ) . Moreover , memory retrieval with the full set of recall cues is not affected by blocking the transmission between the DG and CA3 ( Nakashiba et al . , 2012 ) . Finally , computational studies also suggest that , while the DG inputs to CA3 may be critical during learning , retrieval of memory may rely on direct pathways from EC to CA3 ( Treves and Rolls , 1992; Rolls , 2010 ) . According to this theory , CA3 neurons involved in memory encoding are expected to be preferentially reactivated during memory retrieval . Unfortunately , we were unable to test this hypothesis in the current study due to technical limitations . However , we have demonstrated that CA1 neurons involved in encoding were preferentially reactivated by memory recall . In summary , our results suggest that , in a complex neural network , successful memory recall may not preferentially reactivate all the responsive pathways that are involved in memory formation . 10 . 7554/eLife . 00312 . 017Figure 9 . Correlations between reactivation indexes and behavioral performance in the contextual fear conditioning experiments . A weak but significant correlation between behavioral performance and the reactivation index was detected in the DG ( r = 0 . 3641 , p<0 . 016 ) and no significant correlation was found in CA1 ( r = 0 . 1379 , p>0 . 37 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00312 . 017 In contrast to the DG , CA1 neurons activated during contextual memory formation were preferentially reactivated upon retrieval of the memorized contextual information , even though the context was later associated with an emotional value ( in our case , fear ) . This observation is in agreement with the notion that the hippocampus can automatically encode ongoing events , whereas the association of these events with an emotional value occurs in other brain structures , such as amygdala ( Rudy and O'Reilly , 2001; Stote and Fanselow , 2004 ) . Indeed , similar reactivations of CA1 neurons were observed when animals were re-exposed to a previously experienced environment without a change in emotional value ( Figure 5 , ctxA/A group ) . Similar to our findings , equivalent levels of IEG induction were found in the hippocampus by subjecting mice to contextual fear conditioning training ( context exposure plus foot shocks ) or by exposing mice to the context without foot shocks ( Ramamoorthi et al . , 2011; Liu et al . , 2012 ) . In addition , the role of the amygdala in the association of events with an emotional value is supported by the findings that neurons in the basal lateral amygdala activated during fear conditioning training are preferentially reactivated by the retrieval of contextual fear memory ( Reijmers et al . , 2007 ) and that post-learning elimination of amygdala neurons involved in fear learning erases the fear memory ( Han et al . , 2009 ) . Unlike in the DG , we were not able to detect a linear correlation between the CA1 reactivation and freezing behavior of mice under our experimental conditions ( Figure 9 ) . Particularly , in the experiment involving only small contextual changes , the reactivation index in CA1 did not change accordingly , although the alteration in environmental inputs seemed to be detected by mice , as reflected by their freezing behaviors . It is possible that the reactivation of CA1 may be necessary but not sufficient to drive the behavior under certain circumstance ( e . g . when the input difference is detected by the DG/CA3 network ) . Given that remapping in CA1 is less sensitive to changes in environmental cues compared to CA3 and the DG ( Leutgeb et al . , 2004 , 2007 ) , the small alteration in our experiment may not be significant enough to trigger a global remapping of CA1 neurons , although it remains possible that the firing patterns of activated neurons may be different . In addition , behavioral studies have shown that CA1 is dispensable for spatial pattern separation ( Gilbert et al . , 2001 ) . In summary , our findings are not only consistent with previous reports that CA1 is critically involved in both encoding and retrieval of spatial and contextual information but also suggest that the same CA1 ensemble used for memory formation is likely to be reactivated by recall of the same memory trace .
The TetTag transgenic mice were obtained from Mark Mayford's lab and re-derived into a mixed background of C57BL/6 and balb/c . The mice were bred by intercrossing the hemizygous Fos-tTA:shEGFP line with the hemizygous tetO-tTA*:tau-lacZ line . All mice had food and water ad libitum . The breeding pair and newborn pups were treated with water containing 10 μg/ml dox and 1% sucrose . After weaning , the double transgenic TetTag mice were raised on a 40 mg/kg dox diet . Mice were at least 11 weeks old at the start of the experiments and were group housed until 1 week before the experiments . For BrdU labeling , mice were treated with water containing 2 mg/ml BrdU and 2% sucrose for 1 week . The mice were euthanized >6 weeks later to examine the location of BrdU labeled DGCs . All experimental procedures were approved by the Institutional Animal Care and Use Committee at The Salk Institute for Biological Studies . Mice were sacrificed and brain sections were prepared according to previously reported procedures . Briefly , 1 hr after contextual re-exposure , mice were anesthetized with ketamine ( 100 mg/kg ) and xylazine ( 10 mg/kg ) and were perfused transcardially with saline followed by 4% paraformaldehyde in PBS . The brains of mice were dissected out and post-fixed with 4% paraformaldehyde overnight at 4°C and equilibrated with 30% sucrose . Coronal sections of 40 μm were cut throughout the hippocampal region and stored in the tissue preservation solution at −20°C . Brain sections from a one-in-twelve series were selected for immunostaining . The sections were either double stained with anti-EGR1 and anti-LacZ antibodies or triple stained with anti-FOS , anti-LacZ and anti-RBFOX3 ( aka NeuN ) antibodies . The following primary antibodies were used: mouse anti-LacZ ( 1:10 , 000; Promega , Madison , WI/Fisher , Pittsburgh , PA ) , goat anti-LacZ ( 1:1000; Serotec/Biogenesis , Raleigh , NC ) , rabbit anti-FOS ( 1:800; Santa Cruz , Dallas , TX ) , rabbit anti-EGR1 ( 1:800; Santa Cruz ) and mouse anti-RBFOX3 ( 1:100; Millipore , Billerica , MA ) , rat anti-BrdU ( 1:500; Accurate , Westbury , NY ) . All secondary antibodies were used in 1:250 dilutions and were from Jackson ImmunoResearch . To visualize cell nuclei , all sections were stained with DAPI ( 0 . 5 μg/ml ) . Confocal images were acquired by either a Bio-Rad confocal microscope or a Zeiss LSM 710/780 laser scanning confocal microscope . Images showing the overview of the hippocampus in Figure 1—figure supplement 2 were collected on one z focal plane using a 25× lens with 8 × 4 tiling . Images showing the overview of CA1 and the DG in Figure 2 were collected by a 25× lens with 4 × 2 tiling . For all other images , Z-series ( 10–20 μm for CA1 and 20 μm for the DG ) with a 2-μm interval were acquired using a 40× lens . Images illustrating CA3 in Figure 1—figure supplement 2 were obtained using 2 × 3 tiling . All images used for quantification in the fear conditioning experiments that collected on Zeiss LSM scopes were acquired using 2 × 1 tiling except for those used for DG quantification in Figure 6 ( no tiling ) . Typically , four to five images were analyzed for each animal in each region . The experimenter was blind to the behavioral history of the mice for all quantifications . All statistical analyses were performed using Prism Graphpad software . Data were analyzed with unpaired t-test , one-way ANOVA , two-way ANOVA with repeated measures followed by post hoc Bonferroni tests as indicated . Comparison with chance level was done using one sample t-test using 0 as a theoretical mean . The relationship between the reactivation index and behavioral performance was measured by simple linear correlations ( Pearson correlation ) . All data were presented as mean ± SEM . | Being able to keep memories of similar events separate in your mind is an essential part of remembering . If you use the same carpark every day , recalling where you left your car this morning is challenging , not because you have to remember an event from long ago , but because you have to distinguish between many similar memories . Keeping memories distinct is one of the functions of a subregion of the hippocampus called the dentate gyrus . The process of taking complex memories and converting them into representations that are less easily confused is known as pattern separation . Exactly how the dentate gyrus achieves this , however , is unclear . Computational models predict that a different population of dentate gyrus cells will be active when an animal is in different environments . However , previous experiments have instead shown that the same population of cells is active in multiple environments , and that cells distinguish between environments by firing at different rates . Now , Deng et al . have added to our understanding of pattern separation . The researchers used a type of genetically modified mouse in which it is possible to identify or ‘tag' the activity of a population of hippocampal neurons at multiple time points . They placed each mouse in a box and noted which hippocampal neurons were active as the animal learned about its new environment . After several such learning episodes , the animal received a mild electric shock inside the box . When it was returned to the box the next day , the mouse remembered receiving the shock , enabling the researchers to note which neurons were active during the retrieval process . Deng et al . found that in a subregion of the hippocampus called CA1 , the particular neurons that were active during the initial learning episode were also likely to be active when the animals remembered receiving the shock . However , this was not the case for the dentate gyrus: in this subregion , distinct groups of cells were active during learning and during retrieval . Moreover , exposing the mice to two subtly different environments activated two distinct groups of cells in the dentate gyrus . The work of Deng et al . reveals that memory retrieval does not always involve reactivation of the same neurons that were active during encoding . More importantly , the results indicate that the dentate gyrus performs pattern separation by using distinct populations of cells to represent similar but non-identical memories . Overall the findings add to our understanding of the mechanisms that underpin memory formation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Selection of distinct populations of dentate granule cells in response to inputs as a mechanism for pattern separation in mice |
Behavioral adaptation to environmental threats and subsequent social transmission of adaptive behavior has evolutionary implications . In Drosophila , exposure to parasitoid wasps leads to a sharp decline in oviposition . We show that exposure to predator elicits both an acute and learned oviposition depression , mediated through the visual system . However , long-term persistence of oviposition depression after predator removal requires neuronal signaling functions , a functional mushroom body , and neurally driven apoptosis of oocytes through effector caspases . Strikingly , wasp-exposed flies ( teachers ) can transmit egg-retention behavior and trigger ovarian apoptosis in naive , unexposed flies ( students ) . Acquisition and behavioral execution of this socially learned behavior by naive flies requires all of the factors needed for primary learning . The ability to teach does not require ovarian apoptosis . This work provides new insight into genetic and physiological mechanisms that underlie an ecologically relevant form of learning and mechanisms for its social transmission .
All organisms must acquire and respond to information about their environment . Some changes in the environment are predictable or periodic , like light/dark or seasonal cycles that result in organismal adaptation manifesting as physiological changes in order to optimize survival and fitness in the context of a changing environment ( Baldwin and Meldau , 2013; Cermakian et al . , 2013 ) . This ability to adapt to environmental change is essential for survival , but can such an adaptive response occur in the absence of the direct experience ? Well-defined examples of this phenomenon have been observed in what are considered ‘social’ organisms ( Franks et al . , 2002; Townsend et al . , 2011 ) . Yet , emerging studies are providing mounting evidence to suggest that the use of social cues extend far beyond the traditional notions of social animals: organisms once viewed as asocial in nature are now known to have advanced forms of social communication ( Gariepy et al . , 2014 ) . This social transmission of information can result in distinct behavioral changes , based on another individual's set of experiences . The ability to learn from others influences the choices and behaviors of individuals and allows a group of individuals to share information about a changing environment . It is speculated that social information transmission involves either the ability to feel vicarious reward and punishment or other complex communication strategies to transmit an individual's experience to the community of conspecifics . The potential benefits of adaptive behavior , based on information acquired from others within the community , can give social learners a significant advantage over those that must directly explore and gather environmental information for themselves . Understanding how this information transfer occurs and what the underlying neurological and molecular mechanisms are is critical for a comprehensive view of adaptive behavior across a wide range of taxa . Many species considered as ‘social’ and ‘non-social’ communicate about the environment . Plants have been observed to alter their physiology in response to signaling from another plant ( Baldwin and Schultz , 1983 ) . An example of such communication involves salt stress , which has been shown to trigger the release of volatile organic compounds that induce salt resistance in neighboring plants that have yet to experience any salt stress ( Lee and Seo , 2014 ) . In animals , the process is speculated to be more complex: honeybees are able to fine tune signals directed at individuals within the hive that elicit highly specific behavioral changes in response to specific environmental cues ( Wenner , 1962; Schneider and Lewis , 2004; Richard et al . , 2012 ) . Even Drosophila are prone to social cues , altering their decision making based on the behavior of conspecifics ( Mery et al . , 2009; Sarin and Dukas , 2009; Battesti et al . , 2012 ) . It is clear that the once thought ‘fine line’ between social and non-social organisms is beginning to blur , and that social communication is actually much more fundamental to life than originally considered . In animals , this ability to transmit and process information about the environment has been termed ‘social learning’ ( Gariepy et al . , 2014; Gruter and Leadbeater , 2014 ) . Learning can occur in a social context through olfactory cues , observation and instruction , or by imitation , and thus , is a mechanism for sharing information about a changing environment ( Baldwin and Meldau , 2013; Cermakian et al . , 2013 ) . The potential benefits of adaptive behavior , based on information acquired from others within the community , can give social learners a significant advantage over those that must directly explore and gather environmental information for themselves . However , in general , the underlying molecular mechanisms of social learning are almost entirely mysterious and remain a terra incognita in terms of the strategies for communication , perception , neural plasticity , and the underlying physiological changes that cause changes in behavior . In this study , we use endoparasitoid wasps to explore social learning in the Drosophila model system with the aim of addressing some of these open questions . Endoparasitoid wasps are ubiquitous keystone species in many ecosystems around the world . These wasps prey on immature stages of other insects , using larva and pupa of certain species as hosts for their own offspring . Such wasps pose a serious threat to juvenile Drosophila , with infection rates as high as 90% in natural populations ( Janssen et al . , 1988; Driessen et al . , 1990; Fleury et al . , 2004 ) . Adult Drosophila have evolved complex behavioral changes to protect their offspring from these predatory wasps , including altered food preference and reduced oviposition rates ( Lefevre et al . , 2012; Kacsoh et al . , 2013 ) . Adult Drosophila themselves are not infected by these wasps , thus , making the change in reproductive behavior beneficial only to an anticipated threat to their offspring and not a response to predation itself . A remarkable feature of this altered reproductive behavior is that female Drosophila never having seen this predator can nevertheless robustly and reproducibly respond to it , suggesting an innate recognition of this predator-threat . Here , we use this natural predator system to explore predator threat communication within Drosophila melanogaster and describe the specific learning , memory , and anatomical components necessary for this response . Our findings report the first example of social learning in Drosophila that can be delineated from simple mimicry , through the use of a natural predator . Exposure to the predatory wasp results in a distinct germ line-cell physiological apoptotic response in both flies having seen the wasp ( direct experience ) or flies having been paired with experienced individuals ( social learning ) , which is clearly independent of mimicry . Furthermore , we address the genetic factors , neural circuits , and behavioral changes necessary for the transmission of this socially learned alteration in germ line physiology .
Drosophila melanogaster alters its egg-laying behavior after it encounters parasitoid wasps , which infect fly larvae . This behavioral change entails at least two different and quantifiable behavioral responses . First , if high-ethanol containing food is made available to adult Drosophila , then wasp-exposed females will actively prefer to lay eggs on ethanol-laden food ( Kacsoh et al . , 2013 ) . Second , if ethanol-containing food is not an option , Drosophila females will depress their egg-laying frequency , presumably to allow for time to search and discover an egg-laying environment that is not wasp infested ( Lefevre et al . , 2012 ) . Adult Drosophila are not infected by these wasps , thus , making the change in reproductive behavior beneficial only to an anticipated threat to their offspring . To address the question of whether changes in reproductive behavior could be transferred from exposed teacher flies to naive student conspecifics , we examined the underlying physiological , physical , and genetic components of the exposed teacher and naive student flies and asked if these mechanisms rely on learned reproductive behavior . Drosophila were exposed for 24 hr to wasps in cylindrical 7 . 5-cm long by 1 . 5-cm diameter tubes arrayed into fly condos of 24-tubes where each tube contained five female flies and one male fly , either with three female wasps ( exposed ) or with no wasps at all ( unexposed ) ( Figure 1A , see methods and supporting information for details ) . After 24 hr , food plates were removed and embryos counted . Consistent with previous observations ( Lefevre et al . , 2012 ) , exposed females reduced their oviposition rate significantly ( average unexposed lay ∼65 ± 3 . 2 eggs; average exposed lay ∼13 ± 1 . 98 eggs ) ( Figure 1B ) . We observed this robust response in at least four different genetic backgrounds including Canton-S ( CS ) , Oregon-R ( OreR ) ( unexposed ∼57 ± 2 . 84 eggs compared to exposed 13 ± 1 . 88 eggs on average ) , w1118 ( unexposed ∼25 ± 1 . 54 eggs compared to exposed ∼1 ± 0 . 53 egg on average ) , and transgenic flies carrying Histone H2AvD-GFP ( His-GFP ) ( unexposed ∼108 ± 7 . 69 eggs compared to exposed 18 ± 1 . 97 eggs ) ( Clarkson and Saint , 1999 ) . To test whether this decrease in egg laying can be transmitted from exposed flies to naive females , we exposed Canton-S flies to wasps for 24 hr , then removed the wasps and placed these pre-exposed flies in a new condo with three naive female flies expressing Histone-GFP ( His-GFP ) for an additional 24 hr ( Figure 1A ) . The His-GFP line was ideal for discriminating mixed populations of non-green fluorescent protein ( GFP ) and GFP embryos since this histone is clearly visible by 70 min after oviposition ( embryonic cell cycle 9 ) ( Foe et al . , 1993; Clarkson and Saint , 1999 ) ( Figure 1—figure supplement 1A , B ) . Oviposition in exposed teacher females was significantly reduced during the 24-hr exposure to wasps ( acute depression: 0–24 hr ) ( 53 ± 3 . 35 compared to 14 ± 1 . 59 eggs ) and this depression persisted for an additional 24-hr post wasp exposure ( learned depression: 24–48 hr ) ( 35 ± 2 . 44 compared to 19 ± 1 . 33 eggs ) , relative to age-matched , unexposed sibling controls ( Figure 1C , Figure 1—figure supplement 1C ) . Quantification of total GFP and non-GFP embryos deposited during the 24–48 hr after initial teacher exposure to wasps demonstrated that naive His-GFP student flies had also decreased oviposition , relative to His-GFP siblings mixed with unexposed Canton-S flies ( 33 ± 2 . 34 compared to 6 ± 0 . 86 eggs ) ( Figure 1C , Figure 1—figure supplement 1C ) . In the reciprocal experiment , naive Canton-S student flies mixed with pre-exposed His-GFP teacher flies also exhibited a decrease in oviposition ( 46 ± 2 . 48 compared to 14 ± 1 . 34 eggs , see Supplementary files 6 , 7 for all raw egg numbers ) ( Figure 1—figure supplement 1D , E ) . Thus , naive female flies , never experiencing wasp exposure directly , reduced oviposition when encountering exposed flies . The decrease in oviposition of student flies is not due to an effect of the ratio of teacher to student flies . We tested a 1:1 ratio of 3 exposed female teachers to 3 naive female student flies . This elicited a similar reduction in oviposition ( Figure 1—figure supplement 1F , G ) . Interestingly , when we tested a 1:1 ratio of 3 exposed males to 3 naive female student flies , we found no significant decrease in oviposition for students instructed by exposed males ( Figure 1—figure supplement 1H , I ) . This suggests that , under these conditions , only females can transmit predator-response information . Males are neither necessary nor sufficient for the information transfer . Therefore , for all further experiments , we used a teaching cohort of 5 females and 1 male to 3 female students , unless otherwise noted . 10 . 7554/eLife . 07423 . 003Figure 1 . Flies respond to wasps by decreasing oviposition and are able to confer this information to naive flies . ( A ) Standard exposure setup . ( B and C ) Percent of eggs laid normalized to unexposed . ( B ) Wild-type flies unexposed or exposed to wasps . ( C ) Canton-S teachers and His-GFP students . For ( B ) and ( C ) , error bars represent standard error ( n = 24 biological replicates ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 00310 . 7554/eLife . 07423 . 004Figure 1—figure supplement 1 . Social transmission of information from wasp-exposed female teacher fly to naive female student fly occurs . ( A ) Bright field image of oviposition plate containing Canton-S and His-GFP . ( B ) Green fluorescent protein ( GFP ) channel image of oviposition plate containing Canton-S and His-GFP eggs . ( C to I ) Percent of eggs laid normalized to unexposed shown . ( C ) Second 24 replicates of Canton-S teachers and His-GFP students run in April 2014 . ( D ) First 24 replicates of His-GFP teachers and Canton-S students run in August 2013 . ( E ) Second 24 replicates of His-GFP teachers and Canton-S students run in April 2014 . ( F ) 1:1 ratio of female teacher Canton-S flies to student His-GFP . ( G ) 1:1 ratio of female teacher His-GFP flies to student Canton-S flies . ( H ) 1:1 ratio of male teacher Canton-S flies to student His-GFP flies . ( I ) 1:1 ratio of male teacher His-GFP flies to student Canton-S flies . For ( C ) to ( I ) , error bars represent standard error ( n= 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 004 To test whether the decrease in oviposition can be transmitted from students to a new batch of naive flies , we removed Canton-S pre-exposed teacher females from student His-GFP expressing flies and placed the teacher-instructed student flies in a new chamber with 3 new , naive Canton-S flies ( Figure 2A ) . In teacher-instructed student flies , reduced oviposition behavior persisted for 24 hr after they were separated from teacher flies , indicative of a persisting memory of social learning . Interestingly , we found that our teacher-instructed student His-GFP flies were not able to instruct new students , as the naive Canton-S females did not decrease oviposition ( Figure 2B , Figure 2—figure supplement 1A ) . 10 . 7554/eLife . 07423 . 005Figure 2 . Student flies cannot become teachers . ( A ) Standard exposure setup . ( B ) Teacher exposed primary student His-GFP flies paired with naive secondary student Canton-S flies . Error bars represent standard error ( n = 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 00510 . 7554/eLife . 07423 . 006Figure 2—figure supplement 1 . Student flies cannot become teachers . Teacher exposed primary student Canton-S flies paired with naive secondary student His-GFP flies . Percent of eggs laid normalized to unexposed shown . Error bars represent standard error ( n= 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 006 We postulated that perhaps information transfer could only occur once between wasp-exposed teachers and student flies , leading to the inability of students to further pass on information and become teachers . To test this , we removed the first cohort of student His-GFP expressing flies and placed the Canton-S pre-exposed teacher female flies in a new chamber with a second cohort of 3 new , naive Canton-S flies ( Figure 3A ) . We found that oviposition depression in exposed teacher females was persistent for an additional 24-hr post wasp exposure ( learned depression: 48–72 hr ) , relative to age-matched , unexposed , sibling controls ( Figure 3B ) . Quantification of total GFP and non-GFP embryos deposited during the 48–72 hr after initial teacher exposure to wasps demonstrated that the second cohort of naive His-GFP student flies had also decreased oviposition , relative to His-GFP siblings mixed with unexposed Canton-S flies ( Figure 3B ) . In the reciprocal experiment , a second cohort of naive Canton-S student flies mixed with pre-exposed His-GFP teacher flies also exhibited a decrease in oviposition ( Figure 3C ) . Our results demonstrate that teachers can instruct multiple cohorts of students , thus , the inability of a student to become a teacher is not due to a limitation in amount a teacher can teach . 10 . 7554/eLife . 07423 . 007Figure 3 . Teacher flies can teach multiple batches of students . ( A ) Standard exposure setup for teachers teaching multiple batches of students . ( B and C ) Percent of eggs laid normalized to unexposed . ( B ) Canton-S flies unexposed or exposed to wasps and paired with primary and secondary His-GFP students . ( C ) His-GFP flies unexposed or exposed to wasps and paired with primary and secondary Canton-S students . For ( B ) and ( C ) , error bars represent standard error ( n = 24 biological replicates ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 007 In order to better understand the physiological basis of how a predator-threat leads to changes in oviposition behavior , we examined the status of egg production in exposed female ovaries . Given that poor nutrition or other stressors can cause egg chambers in the ovaries to be eliminated by apoptosis at oogenesis checkpoints in region-2/3 of the germarium or stage 7/8 egg chambers ( the mid-oogenesis checkpoint ) ( Drummond-Barbosa and Spradling , 2001; McCall , 2004 ) , we hypothesized that the presence of parasitoid wasps could similarly reduce oviposition by triggering an oogenesis checkpoint , and thus , account for depressed oviposition . Therefore , we quantified stage-specific apoptosis in ovaries of exposed females . Dissection of ovaries from females having been exposed to wasps for 24 hr revealed a significant increase in the number of egg chambers exhibiting apoptosis relative to unexposed sibling control females ( Supplementary file 1A , B ) . Interestingly , the majority of apoptosis was observed at the stage 7/8 egg chamber checkpoint , with almost no apoptosis in region 2/3 , as visualized by DNA staining with 4' , 6-diamidino-2-phenylindole ( DAPI ) , suggesting that the pathway through which apoptosis was being triggered is fundamentally different from previously described apoptotic events ( Drummond-Barbosa and Spradling , 2001; McCall , 2004 ) ( Supplementary file 1A , B , Figure 4A–F ) . Canton-S and His-GFP fly ovaries were easily distinguishable when stained together , thus , making it possible to score apoptosis levels in ovaries of exposed and unexposed females under completely identical conditions ( Figure 4—figure supplement 1A–D ) . Further confirmation that wasp exposure triggered a true apoptotic event is evidenced by the presence of characteristic DAPI-intense pychnotic nurse cell nuclei , by terminal deoxynucleotide transferase dUTP nick end labeling ( TUNEL ) stain that detects fragmented DNA ( Figure 4G–J and Figure 4—figure supplement 1E , F ) , and activated caspase-3 staining ( Figure 4—figure supplement 1G–J ) : All positive markers of the cell death process ( McCall , 2004 ) . We noted that both DAPI and TUNEL were readily detected in apoptotic stage 12/13 nurse cells in both exposed and unexposed females at similar levels . Developmentally regulated cell death is normally expected to eliminate late-stage nurse cells in maturing oocytes , thus , serving as an internal control for the level of detected apoptosis in exposed and unexposed females ( Supplementary file 1A , B ) . Similar to the reduced oviposition behavior observed , this physiologically triggered apoptosis specifically of stage 7/8 egg chambers persisted well beyond the period of initial wasp exposure ( Figure 4A , B , Supplementary file 1C , D ) . 10 . 7554/eLife . 07423 . 008Figure 4 . Stage-specific apoptosis observed in wasp-exposed teachers and teacher-exposed student flies . ( A and B ) Average percent of apoptotic events for stage 7/8 egg chambers . ( A ) Canton-S exposed and unexposed ovary apoptosis . ( B ) His-GFP exposed and unexposed ovary apoptosis . ( C to D ) Canton-S unexposed/exposed ovariole . ( E to F ) His-GFP unexposed/exposed ovariole . ( G to H ) Canton-S transferase dUTP nick end labeling ( TUNEL ) staining performed on exposed fly ovaries . ( I to J ) His-GFPTUNEL staining . For ( A ) and ( B ) , error bars represent standard error ( n = 3 biological replicates from which 12 ovaries were scored for each group ) ( *p < 0 . 05 ) . Scale bars , 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 00810 . 7554/eLife . 07423 . 009Figure 4—figure supplement 1 . Stage-specific apoptosis is induced following wasp exposure . ( A ) Canton-S and His-GFP ovaries intermingled are easily distinguishable . ( B ) Canton-S and His-GFP egg chambers that are undergoing apoptosis are easily distinguishable by using DAPI or TUNEL . ( C to D ) His-GFP unexposed/exposed ovary chain corresponding to Figure 3E , F , visualized using the GFP channel . ( E to F ) DAPI and His-GFP images of TUNEL-stained egg chambers corresponding to Figure 3I , J . ( G to H ) Canton-S exposed egg chambers stained with DAPI and anti-caspase . ( I to J ) His-GFP exposed egg chambers stained with DAPI and anti-caspase . Scale bars are 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 009 We considered the possibility that exposure to wasps could change fly feeding behavior , and subsequent poor nutrition could trigger the mid-oogenesis checkpoint ( Drummond-Barbosa and Spradling , 2001 ) . We gave both exposed and unexposed flies a high-protein yeast food stained with red food dye to visualize food intake . We found that both wasp-exposed and unexposed flies exhibited a similar amount of high-protein yeast food intake even when given a choice to feed on normal food without yeast by visualizing the red dye in the fly abdomens ( Figure 5A–D , Figure 5—figure supplement 1A–D ) . The red yeast paste was placed on instant Drosophila media , which turns blue upon contact with water , allowing us to visualize whether flies are preferring high ( red ) - or low ( blue ) - nutrient food ( Figure 5—figure supplement 1E–L ) . We found that even in the presence of high-protein yeast food , exposed flies still depressed oviposition when compared to unexposed controls , in addition to having apoptosis induced at the egg chamber stage 7/8 checkpoint ( Figure 5E–G , Figure 5—figure supplement 1M–T , Supplementary file 1E ) . Thus , the mid-oogenesis apoptosis checkpoint triggered in exposed flies is not due to a poor nutrition intake . These data are indicative of a predator-induced neuroendocrine signaling pathway that impinges on a pathway specifically controlling mid-oogenesis specifically ( stage 7/8 but not stage 2/3 ) , and therefore , is likely different from the previously described poor nutrition oogenesis checkpoint . 10 . 7554/eLife . 07423 . 010Figure 5 . Flies continue to eat high-protein diet following wasp exposure but still depress oviposition . Continued oviposition depression cannot be explained by a lack of nutrient intake that normally inactivates insulin signaling . The high-nutrient intake by exposed female flies suggests that an active insulin signaling pathway is inhibited or bypassed downstream of nutrient sensing . ( A ) Exposed and unexposed flies anesthetized immediately after 24-hr exposure period shows red food in abdomens . ( B ) Lateral view of unexposed fly . ( C ) Lateral view of exposed fly . ( D ) Percent of male and female flies with red food in abdomen , error bars are 95% confidence intervals . ( E ) Percent of eggs laid normalized to unexposed following 24-hr exposure period . All eggs on the food plate were counted , including eggs on the yeast paste . ( F ) Representative ovary dissected from unexposed fly . 36 total ovaries were dissected and examined across 3 replicates for each treatment . ( G ) Ovary dissected from exposed fly . Scale bar for ( F ) to ( G ) is 1 . 0 mm . ( H ) Average percent apoptosis in mid-oogenesis checkpoint for unexposed and exposed Canton S . For ( D ) , ( E ) , and ( H ) , error bars represent standard error ( n = 3 biological replicates . For ( D ) , 100 female and 20 male flies were counted per replicate . For ( E ) , 3 egg lay plates were counted per treatment . For ( H ) , 3 biological replicates from which 12 ovaries were scored for each group ) ( *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 01010 . 7554/eLife . 07423 . 011Figure 5—figure supplement 1 . Flies continue to eat high protein diet following wasp exposure but still depress oviposition . Additional evidence demonstrating that exposed flies eat rich nutrient food . ( A ) Unexposed flies after 24-hour period . ( B ) Ventral view of unexposed fly . ( C ) Exposed flies with wasps after 24-hour period . ( D ) Ventral view of exposed fly . ( E ) Unexposed flies after a 24-hour period fed only blue food have blue abdomens . ( F ) Exposed flies with wasps after a 24-hour period fed only blue food have blue abdomens . Percent of male and female flies with red food in abdomen . ( G ) Lateral view of unexposed fly . ( H ) Ventral view of unexposed fly . ( I ) Lateral view of exposed fly . ( J ) Ventral view of exposed fly . ( K ) Food plate from unexposed flies following 24-hour period . ( L ) Food plate from exposed flies following 24-hour period . ( M ) Zoom-in of food plate from unexposed flies showing many eggs . ( N ) Zoom-in of food plate from exposed flies showing few eggs . ( O to Q ) Unexposed Canton S ovaries from individuals fed yeast paste for 24 hours . ( R to T ) Exposed Canton S ovaries from individuals fed yeast paste for 24 hours . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 011 To test whether triggering of the mid-oogenesis check point could be transmitted from experienced , wasp-exposed females to naive females , we mixed teacher and student flies as described above . Naive student flies mixed with exposed teachers showed apoptosis at the stage 7/8 checkpoint , as did their teachers ( Supplementary file 1C , D , F , G , Figure 4A–B ) . Students mixed with unexposed , ‘mock’ teachers did not show significant levels of increased apoptosis in the ovary ( Supplementary file 1C , D , F , G , Figure 4A–B ) . Thus , in naive student flies , transmitted information from exposed teacher flies results in triggering a specific-apoptotic mid-oogenesis checkpoint in students that have learned from teachers' experience . These data indicate that teacher flies transmit instructive cues to student flies that student flies receive these cues and then process them in a manner that leads to apoptosis of egg precursor cells and reduced oviposition . One explanation for social learning could be that student flies instinctively mimic the behavior of more experienced teacher flies . Repeated episodes of imitative behavior could lead to a strengthening of neural circuits that underlie this behavior . We explored this idea by testing if wasp-exposed flies that are genetically unable to suppress oviposition efficiency are still able to successfully act as teacher flies . The Drosophila mid-oogenesis checkpoint is known to activate effector caspases Dcp-1 and drice ( McCall , 2004 ) . Additionally , the caspase-3 staining we performed on wasp-exposed teacher ovaries recognizes effector caspases Dcp-1 and drice ( Figure 4—figure supplement 1G–J ) , leading us to hypothesize that these caspases are important in oviposition depression in teacher and student flies as a response to parasitoid wasps . By using a maternal α-Tubulin > Gal4 driver to express an RNA-hairpin targeting mRNA from each of these genes , we were able to reverse both the decrease in oviposition as well as the increase of stage 7/8 egg chamber apoptosis of wasp-exposed females , while RNAi depletion of these caspases had no effect on oviposition of unexposed females ( Figure 6—figure supplement 2A , B ) . This provides further evidence that the stage 7/8 egg chamber apoptosis and corresponding oviposition decrease is a specific physiological checkpoint , similar to that previously described for poor nutritional intake ( Figure 4A , B , Supplementary file 1H ) ( Drummond-Barbosa and Spradling , 2001 ) . We considered the possibility that ovarian apoptosis could produce secondary signals important for conveying information to naive flies , which in turn triggers apoptosis in student ovaries . To test this , we used teacher flies that were incapable of triggering apoptosis because of RNAi depletion of Dcp-1 or drice , specifically in developing egg chambers . Strikingly , following wasp exposure , flies , depleted of germ line Dcp-1 or drice function , were still excellent teachers capable of cueing naive student flies to decrease their oviposition and induce apoptosis at the stage 7/8 mid-oogenesis checkpoint in the students' ovaries ( Figure 6A , B , Figure 6—figure supplement 1 A-B and G , Supplementary file 1I , J ) . The finding that Dcp-1 and drice deficient females incapable of depressing oviposition can nevertheless convey critical cues to naive students demonstrates that the depressed oviposition response can be decoupled from the process required for teacher–student information transfer . Thus , information transfer in this context is not due to secondary effects of ovarian cell death . Interestingly , Dcp-1- and drice-deficient student females could not depress oviposition in response to exposed , wild-type teachers , suggesting that the same effector caspases activated in exposed teachers are also needed for oviposition depression in students ( Figure 6C–D ) . Control , parental lines were found to behave as wild type as both teachers and students ( Figure 6—figure supplement 1A–F ) . We tested two additional Dcp-1 ( Dcp-12 and Dcp-13 ) ( Etchegaray et al . , 2012 ) mutant lines that displayed the same phenotype as the RNAi result ( Figure 6E–F , Figure 6—figure supplement 1G , H ) . We conclude that the depressed oviposition in student flies cannot be from simple mimicry . 10 . 7554/eLife . 07423 . 012Figure 6 . Socially transmitted oviposition depression in response to wasp exposure acts through the mid-oogenesis checkpoint . ( A to F ) Percent of eggs laid normalized to unexposed . ( A and C ) Drice RNAi-knockdown as teachers and students . ( B and D ) Dcp-1 RNAi-knockdown as teachers and students . ( E to F ) Dcp-12 as teachers and students . For ( A ) to ( F ) , error bars represent standard error ( n = 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 01210 . 7554/eLife . 07423 . 013Figure 6—figure supplement 1 . Socially transmitted oviposition depression acts through the mid-oogenesis checkpoint . ( A to H ) Percent of eggs laid normalized to unexposed . ( A to F ) Teacher and student ability of GAL4 and UAS control parental strains from drice and Dcp-1 RNAi-knockdown crosses . ( G ) to ( H ) Teacher and student ability of Dcp-13 . For ( A ) to ( H ) , error bars represent standard error ( n= 24 biological replicates ) . ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 01310 . 7554/eLife . 07423 . 014Figure 6—figure supplement 2 . Further evidence indicating that oviposition depression acts through the mid-oogenesis checkpoint . ( A ) Average percent of apoptotic events for stage 7/8 egg chambers in exposed and unexposed Dcp-1 RNAi-knockdown flies as teachers and His-GFP as students ( n= 3 biological replicates from which 12 ovaries were scored for each group ) ( *p < 1 . 0e-5 ) . ( B ) Average number of eggs laid for Drice and Dcp-1 RNAi and parental lines , shown with standard error . No significant difference in oviposition for unexposed flies was observed , P > 0 . 05 in all comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 014 Previous work has demonstrated that wasp-exposed females actively prefer to lay eggs on ethanol-laden food through the use of visual cues . These visual cues were important for wasp perception and subsequent behavior change ( Kacsoh et al . , 2013 ) . Therefore , to better understand the mechanism through which information was being transferred from teacher to student flies , we tested the role of both smell and vision in information acquisition in our system by testing these mutations in both teacher and student flies . The gene Orco is known to be expressed in almost all olfactory receptor neurons , and the mutant-lacking Orco is unable to respond to smell stimuli ( Vosshall et al . , 1999 ) . We found that Orco1 flies could respond to wasps and teach student flies ( Figure 7A ) . Additionally , Orco1 flies as naive students could learn normally from teacher flies ( Figure 7B ) . These data suggest that olfaction is not necessary to perceive the wasp threat nor to confer or receive the information during social learning . 10 . 7554/eLife . 07423 . 015Figure 7 . Flies respond to wasps and confer this information to naive flies through visual cues . ( A to D ) Percent of eggs laid normalized to unexposed . ( A to B ) Smell mutants as teachers and students . ( C to D ) Sight mutants as teachers and students . For ( A ) to ( D ) , error bars represent standard error ( n = 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 015 We then analyzed the role of vision in this paradigm with the use of flies mutant for ninaB . ninaB is part of a single enzyme family , which acts as a key component for visual pigment production and vision in Drosophila ( von Lintig et al . , 2001; Voolstra et al . , 2010 ) . The ninaBP315 blind females exhibited no initial response to the presence of wasps and were not able to transmit information to naive flies ( Figure 7C ) . In contrast to Orco1 flies , blind ninaBP315 student flies were unable to learn from teacher flies ( Figure 7D ) . Our ninaBP315 data suggest that visual stimuli are responsible for both the acute and learned response . Therefore , we wanted to further elucidate the role of vision in this system . As in previous studies , we impaired vision of wild-type flies simply by running trials in complete darkness ( Tompkins et al . , 1982; Budick et al . , 2007; Duistermars et al . , 2009; Robie et al . , 2010; Ofstad et al . , 2011 ) . We found that performing the entirety of experiment in darkness using Canton-S or His-GFP teachers yielded no response to the presence of wasps and exposed females were not able to transmit information to naive flies ( Figure 8A , B , Figure 8—figure supplement 1A ) . Similarly , performing only the wasp exposure period in the dark and the social-learning period in the light again yielded no response to the presence of wasps and these exposed females were not able to transmit information to naive flies ( Figure 8C , D , Figure 8—figure supplement 1B ) . Finally , we performed the wasp exposure period in the light , but moved the teachers paired with students for the social-learning period into the dark ( Figure 8E ) . Here , we find teacher flies had both an acute and learned response , but these teachers were not able to transmit information to naive flies , presumably due to the learning period being in the dark ( Figure 8F , Figure 8—figure supplement 1C ) . Consistent with previous studies indicating the necessity of light in visual learning ( Ofstad et al . , 2011 ) , these data suggest that wild-type fly vision can only detect cues from wasps and teachers if there is light present , again demonstrating the role for visual cues for the behavior . 10 . 7554/eLife . 07423 . 016Figure 8 . Acute and teaching response requires light . ( B , D , and F ) Percent of eggs laid normalized to unexposed . ( B , D , and F ) Canton S as teachers and His-GFP as students . ( A ) Exposure setup when both acute and social response occurs in dark . ( B ) Results of experiment as described in ( A ) . ( C ) Exposure setup when acute response occurs in the dark but social response occurs in the light . ( D ) Results of experiment as described in ( C ) . ( E ) Exposure setup when acute response occurs in the light but social response occurs in the dark . ( F ) Results of experiment as described in ( E ) . For ( B ) , ( D ) and ( F ) , error bars represent standard error ( n = 24 biological replicates ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 01610 . 7554/eLife . 07423 . 017Figure 8—figure supplement 1 . Further evidence indicating that learning requires light . ( A to C ) Percent of eggs laid normalized to unexposed . ( A to C ) His-GFP as teachers and Canton-S as students . ( A ) His-GFP as teachers when both acute and social response occurs in dark . ( B ) His-GFP as teachers when acute response occurs in the dark but social response occurs in the light . ( C ) His-GFP as teachers when acute response occurs in the light but social response occurs in the dark . For ( A to C ) error bars represent standard error ( n = 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 017 Finally , we wanted to elucidate if a visual cue alone is sufficient to elicit the behavioral changes . Previous experiments had both teachers and students co-habitating , leading us to speculate whether other stimuli were involved in either the acute- or social-learning response . To test this , we built the Fly Duplex , which we constructed by using three standard 25 mm × 75-mm glass microscope slides that were adhered between two 75 mm × 50 mm × 1-mm glass microscope slides using clear aquarium silicone sealant , making two compartments separated by one 1-mm thick glass slide . This setup allows flies to see other flies or wasps in the neighboring chamber , but do not allow direct contact ( Figure 9A ) . We find that both the acute and learned response are intact when performing the exposure in separate , but adjacent , chambers using the Fly Duplex ( Figure 9B–C ) . We also find that teachers are able to transmit information to naive flies when in separate chambers , yielding depressed oviposition ( Figure 9B–C ) . Both the requirement for light and the use of the Fly Duplex strongly suggest that olfactory , auditory , and tactile information is not likely to be important for this type of social communication . Instead , this demonstrates that visual cues alone are sufficient for acute- , learned- , and social-learning responses . 10 . 7554/eLife . 07423 . 018Figure 9 . Visual cues are necessary and sufficient for learning . ( B and C ) Percent of eggs laid normalized to unexposed . ( A ) Standard exposure setup using the Fly Duplex . The Fly Duplex ensures only visual cues are transferred between groups . ( B ) Canton S as teachers with His-GFP students . ( C ) His-GFP as teachers with Canton S as students . For ( B and C ) error bars represent standard error ( n = 10 biological replicates ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 018 Collectively , our data demonstrate that teacher flies respond to a visual stimulus during wasp exposure and subsequently provide visual cues , which student flies process in a manner that leads to reduced oviposition . In order to elucidate the visual cue used to transmit information from teachers to naive students , we tested flies that were missing wings , either through genetic or mechanical perturbation . We first tested flies mutant in the wingless gene ( wg1 ) . The wingless phenotype in the wg1 stock is not fully penetrant . The progeny of wg1 parents are comprised of flies with two wings , one wing , and no wings ( Figure 10A , B , Figure 10—figure supplement 1A , B ) . Reported segregation patterns suggest that the three phenotypes are genotypically similar and that phenotypic change is a result of incomplete penetrance ( Sharma , 1973 ) . We find that both one-winged and two-winged mutants have an intact acute and learned response following wasp exposure ( Figure 10C , Figure 10—figure supplement 10F ) . However , one-winged wg1 flies are unable to act as teachers , suggesting a role for both wings in communication ( Figure 10C ) . Two-winged wg1 flies behaved as wild-type teachers , demonstrating that the wg1 mutation does not induce impaired teaching ( Figure 10—figure supplement 1F ) . For additional validation of this observation , we mechanically removed the wings of wild-type flies . The wings of wild-type Canton-S flies were cut prior to wasp exposure and tested for oviposition response . These flies displayed an intact acute and learned response , but they were unable to teach ( Figure 10D–F , Figure 10—figure supplement 1C , D ) . Finally , we used the GAL4/UAS system to express the cell death protein reaper ( UAS-Rpr ) in conjunction with a wing driver ( MS1096 ) to ablate proper wing development ( Figure 10G , Figure 10—figure supplement 1E ) . We find that these flies also have an intact acute and learned response , but they were unable to teach ( Figure 10H ) . Flies lacking wild-type wings were able to function as students , demonstrating that wings are not necessary for student learning ( Figure 10—figure supplement 1G ) . 10 . 7554/eLife . 07423 . 019Figure 10 . Teacher–student dynamics require wings to allow for communication to take place . ( For C , F , and H ) Percent of eggs laid normalized to unexposed . ( A ) Dorsal view of wg1 with one wing . ( B ) Dorsal view of wg1 with two wings . ( C ) wg1 one-winged flies as teachers . ( D ) Dorsal view of Canton-S female . ( E ) Dorsal view of Canton-S female with clipped wings . ( F ) Canton-S flies with clipped wings as teachers . ( G ) Dorsal view of a female fly expressing reaper in the wing disc . ( H ) Flies expressing reaper in the wing disc as teachers . Error bars represent standard error ( For ( C ) n = 18 biological replicates . ) ( For [F and H] n = 24 biological replicates ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 01910 . 7554/eLife . 07423 . 020Figure 10—figure supplement 1 . Teacher flies need wings in order to instruct student flies . ( A ) Lateral view of one winged wg1 female . ( B ) Lateral view of two-winged wg1 female . ( C ) Lateral view of Canton-S female . ( D ) Lateral view of Canton-S female with clipped wings . ( E ) Lateral view of female fly expressing reaper in the wing disc . ( F ) wg1 two-winged flies as teachers . For ( F ) and ( G ) , percent of eggs laid normalized to unexposed . ( G ) Flies expressing reaper ( UAS-Rpr ) in the wing disc as students . Error bars represent standard error ( For [F and G] n= 24 biological replicates . ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 020 We hypothesized that perhaps flies whose wings had been genetically ablated or mechanically removed could be experiencing overall mobility impairment , thus , yielding the inability to teach . We decided to perform our assay using flies mutant in the erect wing locus , which encodes a protein , EWG . Loss-of-function erect wing alleles result in embryonic lethality . Viable alleles of erect wing cause severe abnormalities of the indirect flight muscles ( DeSimone et al . , 1996 ) . Flies carrying viable allelic combinations of mutations at the erect wing ( ewg ) locus do not have , or have greatly reduced , indirect flight muscles ( Deak II et al . , 1982; Fleming et al . , 1982 ) . We tested two EWG alleles , ewg1 and ewg2 , and found that these flies displayed an intact acute and learned response , but they were unable to teach . These mutants exhibited a wild-type ability to learn from His-GFP teachers , again demonstrating that wings are not required to learn ( Figure 11A–D ) . EWG is also required in the development of the nervous system ( Fleming et al . , 1982; DeSimone and White , 1993 ) . Given this information , we wanted to examine if nervous system-specific expression of wild-type EWG protein in an ewg mutant background is sufficient to restore teaching ability . This expression does not rescue the muscle phenotype ( DeSimone et al . , 1996 ) . We found that ewgNS4 ( neuronal rescue ) displayed an intact acute and learned response , had the ability to learn from His-GFP teachers , but they were unable to teach ( Figure 11E–F ) . 10 . 7554/eLife . 07423 . 021Figure 11 . Teacher–student dynamics require functional wings to allow for communication to take place . ( For A to F ) Percent of eggs laid normalized to unexposed . ( A and B ) ewg1 as teachers and students . ( C and D ) ewg2 as teachers and students . ( E and F ) ewgNS4 as teachers and students . Error bars represent standard error ( For [A to F] n = 24 biological replicates . ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 021 Through the use of multiple genetic mutants and genetic and mechanical perturbations of wings , we find that both wings and wing movements are necessary for teaching ability . Collectively , these data suggest that teacher flies are using their wings as the visual cue to inform naive student flies . To examine the possibility that the behavioral response to predator-threat requires active learning and associated plasticity in wasp-exposed flies , we asked how predator responses were affected in learning and memory mutants rutabaga ( rut1 , rut2080 ) , dunce ( dnc1 , dncML ) , Adf1 ( Adf1nal ) , amnesiac ( amn1 , amnX8 ) , FMR1 ( Fmr13 , Fmr1B55 ) , and Orb2ΔQ; the last being of particular significance as the ΔQ mutation leaves all essential functions of the Orb2 neuronal regulator intact , but deletes a Gln-rich prion domain exclusively required for persistent long-term memory , possibly by enabling an Orb2 conformational switch that leads to active synaptic translation ( Si et al . , 2003; Keleman et al . , 2007; Majumdar et al . , 2012 ) . Each of these mutants responded acutely to predator presence with a dramatic decrease in oviposition when in the presence of wasps for the first 24 hr ( Figure 12A , C , E , G , I , K and Figure 12—figure supplement 1A , B , E , G ) . This indicates that the acute oviposition depression is independent of these gene functions . However , when wasps were removed and mutant flies were placed in a new tube for an additional 24 hr after wasp exposure , oviposition returned to levels comparable to unexposed flies ( Figure 12A , C , E , G , I , K and Figure 12—figure supplement 1A , B , E , G ) . This indicates that although the acute response to a predator threat does not require memory consolidation , the persistence of decreased oviposition behavior after wasp removal requires a form of long-term memory whose consolidation requires cAMP signaling and translational control mediated at least in part through the prion domain of Orb2 . These results are consistent with other wasp-induced fly memory formation , specifically with respect to seeking ethanol-laden substrates upon wasp exposure ( Kacsoh et al . , 2015 ) . Naive wild-type student flies encountering the pre-exposed mutants also did not respond through oviposition decrease ( Figure 12A , C , E , G , I , K and Figure 12—figure supplement 1A , B , E , G ) . Collectively , the data from multiple alleles of multiple mutants indicated that these mutations yielded flies that did not retain physiological effects of the threat-response necessary to successfully transmit information to naive wild-type student females . 10 . 7554/eLife . 07423 . 022Figure 12 . Learning mutants are unable to teach or be students . ( A to L ) Percent of eggs laid normalized to unexposed . ( A to B ) Orb2ΔQ as teacher and student . ( C to D ) rut1 as teacher and student . ( E and F ) dnc1 as teacher and student . ( G and H ) Adf1nal as teacher and student . ( I and J ) FMR1B55 as teacher and student . ( K and L ) amn1 as teacher and student . For ( M ) , average percent of apoptotic events for stage 7/8 egg chambers . ( M ) Orb2ΔQ exposed and unexposed ovary apoptosis . Error bars represent standard error . ( For [A] to [L] n = 24 biological replicates . ) ( For [M] n = 3 biological replicates from which 12 ovaries were scored for each group ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 02210 . 7554/eLife . 07423 . 023Figure 12—figure supplement 1 . Learning mutants are unable to teach or be students . ( A to H ) Percent of eggs laid normalized to unexposed . ( A and C ) dncML as teacher and student . ( B and D ) rut2080 as teacher and student . ( E and F ) FMR13 as teacher and student . ( G and H ) amnX8 as teacher and student . Error bars represent standard error ( n= 24 biological replicates ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 023 Unexpectedly , socially learned depression of oviposition in naive student flies was defective in rut , dnc , Adf1 , amn , FMR1 , and Orb2 mutants ( Figure 12B , D , F , H , J and Figure 12—figure supplement 1C , D , F , H ) . As these learning mutants show normal acute oviposition depression in response to direct wasp exposure , this suggests that wasp-induced and teacher-induced reductions in oviposition behavior occur through fundamentally different mechanisms . This is consistent with the fact that wasps and teachers must provide different visual signals to initiate learning and must , therefore , be expected to alter behavior through different neural circuit mechanisms . Taken together with the observations of blind ninaBP315 mutants , experiments performed in the dark , and the Fly Duplex , these results demonstrate that during social learning student flies must be able to visually perceive information from teacher flies and then undergo an active-learning process in order to stably respond by depressing oviposition . We further asked how apoptosis in egg chambers was affected in wasp-exposed orb2ΔQ mutant flies . The apoptotic response to acute wasp exposure ( 0–24 hr ) in orb2ΔQ was similar to the wild type , as expected , given that these flies had a normal depressed oviposition in presence of wasps ( Figure 12M , Supplementary file 1K ) . However , in the 24-hr period following removal of wasps ( 24–48 hr ) , orb2ΔQ female flies had increased their egg laying and showed low levels of apoptosis in stage 7/8 egg chambers comparable to control unexposed flies ( Figure 12M , Supplementary file 1L ) . We conclude that Drosophila females depress their egg laying during exposure to predatory wasps through an acute pathway that requires visual perception of wasp presence and leads to active elimination of developing eggs . The persistence of depressed oviposition and apoptosis in the 24-hr period after wasp removal requires an intact orb2 gene , suggesting that maintenance of the initial behavior may require neural consolidation of the memory of wasp presence learned during the exposure period . Both acute and persistent mechanisms indicate that a systemic pathway initiated in photoreceptors and visual systems of female flies , processed centrally through neural circuits that can encode memories , leads to neuroendocrine signaling that impinges on developing egg chambers where it activates caspase-signaling cascades . To test if the reduced oviposition requires continued neuronal input to maintain reduced oviposition and teaching behavior , we mechanically removed neural input of exposed wild-type flies . Following wasp exposure , we surgically removed fly heads and paired them with naive student flies . Decapitated flies are of standard use in behavioral assays , and only decapitated flies that recovered after anesthesia were used ( Cook , 1975; Nilsen et al . , 2004; Clyne and Miesenbock , 2008; Trott et al . , 2012 ) . We found that decapitated flies could not maintain the same level of reduced oviposition as normal flies ( i . e . , decapitation led to an increase in oviposition ) , and they could no longer teach , suggesting a continued input from the brain is needed to elicit these behavioral changes ( Figure 13A–C , Figure 13—figure supplement 1E , F ) . To ask whether the mushroom body ( MB ) specifically plays a role in maintained oviposition reduction and the teaching behavior , we used the GAL4/UAS system to express tetanus toxin light chain ( UAS-TeTx ) in conjunction with a MB driver ( OK-107-GAL4 ) ( Aso et al . , 2009 ) to block synaptic transmission ( Martin et al . , 2002 ) . The tetanus toxin light chain works by catalytically inhibiting synaptic transmission once present in the cytosol by cleaving either synaptobrevin , syntaxin , or SNAP-25 ( Poulain et al . , 1988; Bittner et al . , 1989; Mochida et al . , 1990; Kurazono et al . , 1992; McMahon et al . , 1993 ) . We found that flies expressing UAS-TeTx in the MB exhibited a wild-type acute response , suggesting that the acute response occurs independent of the MB . However , in the learned period , these flies no longer showed reduced oviposition and were unable to teach naive students ( Figure 13D ) . Using a second MB driver ( MB247 ) , this result was recapitulated ( Figure 13—figure supplement 1I ) ( Mao et al . , 2004 ) . Control parental lines functioned as both wild-type students and teachers both as homozygotes and when outcrossed to Canton-S ( Figure 13—figure supplement 1A–H ) . Flies expressing UAS-TeTx in the MB failed to function as students ( Figure 13—figure supplement 1J–K ) . These data suggest that wasp presence is sensed through the visual system , and this information is relayed to the MB to induce a persistent reduction of oviposition , apoptosis , and teaching behavior , all of which are maintained over a time span of days . 10 . 7554/eLife . 07423 . 024Figure 13 . Learning and teaching require a continuous neural input from the brain . ( A and D ) Percent of eggs laid normalized to unexposed . ( A ) Canton-S teachers with heads removed after acute exposure . ( B ) Dorsal view of representative Canton-S female . ( C ) Dorsal view of representative Canton-S female with no head . ( D ) Flies expressing tetanus toxin ( UAS-TeTx ) in mushroom body ( MB ) as teacher . Error bars represent standard error ( For [A] and [D] n = 24 biological replicates . ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 02410 . 7554/eLife . 07423 . 025Figure 13—figure supplement 1 . Blocking synaptic transmission in the MB prevents teacher behavior and student learning . ( A to E ) Percent of eggs laid normalized to unexposed . ( A and B ) OK107 GAL4 as teachers and students . ( C and D ) OK107/+ GAL4 as teachers and students . ( E and F ) UAS-TeTx as teachers and students . ( G and H ) UAS-TeTx/+ as teachers and students . ( I ) Flies expressing tetanus toxin in MB as teachers using MB 247 GAL4 . ( J ) Flies expressing tetanus toxin in MB as students using MB 247 GAL4 ( K ) Flies expressing tetanus toxin in MB as students using OK107 GAL4 . Error bars represent standard error ( For [A] to [K] n = 24 biological replicates . ) ( *P < 0 . 05 , **P < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 025 We found that mutants in orb2 exhibited a defect of oviposition depression as well as teaching and social-learning ability ( Figure 12A–B ) . However , these experiments could not exclude the possibility that orb2 gene product was required in non-neural tissues . Similarly , orb2 may have been necessary for early neuronal development , and mutant phenotypes observed simply reflected developmental defects that precluded proper adult MB functions ( pleiotropic effects ) . Given that inhibiting synaptic transmission in the MB with UAS-TeTx eliminated a long-term behavioral response to wasp exposure , teaching ability , and social learning ( Figure 13D , Figure 13—figure supplement 1I–K ) , we tested the hypothesis that the gene products of known learning and memory genes ( such as orb2 ) may also be required to function in this anatomical region of the brain . To test this , we used the GAL4/UAS system as before: in this case , the MB driver ( OK-107-GAL4 ) drove expression of an RNA-hairpin targeting orb2 mRNA . We found that RNAi depletion of Orb2 in the MB produced the same phenotype as the orb2ΔQ mutant tested ( Figure 12A , B , Figure 14A–B ) . This result highlights that flies deficient in orb2 in the MB are able to perceive and respond to wasps , but not remember exposure , and therefore cannot teach naive students , once wasps are removed . Flies deficient in orb2 in the MB are also unable to learn from wild-type teachers . Control parental lines with either just the OK-107-GAL4 or UAS-Orb2-hairpin transgenes ( but not both ) functioned as wild type as they exhibited no defects in behavior persistence ( Figure 13—figure supplement 1A–D , Figure 14—figure supplement 1A , B ) . Control lines expressing RNA-hairpin targeting the white gene in the MB demonstrated wild-type behavior , demonstrating induction of the RNA-hairpin alone does not induce deficient memory formation , teaching ability , or learning ability ( Figure 14C , D , Figure 14—figure supplement 1C , D ) . This suggests that orb2 is required in MB neuronal circuits in order for maintained wasp-induced oviposition depression , and it further suggests that persistence of this behavior likely requires long-term memory formation is the MB . 10 . 7554/eLife . 07423 . 026Figure 14 . Knockdown of Orb2 in the MB results in defective learning . ( A to D ) Percent of eggs laid normalized to unexposed . ( A to B ) Orb2 RNAi-knockdown as teachers and students . ( C to D ) white RNAi-knockdown as teachers and students . ( For [A] to [D] n = 24 biological replicates . ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 02610 . 7554/eLife . 07423 . 027Figure 14—figure supplement 1 . Expression of an RNAi hairpin in the MB does not induce defective learning and memory . ( A to D ) Percent of eggs laid normalized to unexposed . ( A to B ) Teacher and student ability of UAS control parental strains from Orb2 RNAi-knockdown cross . ( C to D ) White RNAi-knockdown as teachers and students . ( For [A] to [D] n= 24 biological replicates . ) ( **p< 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 027 The above data , however , do distinguish between two possible roles for orb2 . First , the orb2 gene product could be required for normal development of the MB and other parts of the nervous system that interface with the MB . The OK-107-GAL4 driver begins expression of GAL4 in the larvae . Thus , it remains possible that RNAi depletion of Orb2 in the larvae could cause developmental defects that then indirectly cause behavioral phenotypes in adults . A second possibility is that persistence of depressed oviposition and in turn teaching ability requires orb2 function in the adult MB , regardless of its possible function during MB development . In order to address this question , we turned to the GAL4-based Gene-Switch System where the GAL4 transcription factor is fused to the human progesterone ligand-binding domain ( Burcin et al . , 1999 ) . We used flies expressing the Gene-Switch transgene specifically in the MB , where only an administration of the pharmacological Gene-Switch ligand RU486 could activate the GAL4 transcription factor ( Mao et al . , 2004 ) . In order to confirm our feeding protocol could work in The Fly Condo , we used the MB Gene-Switch line to express a nuclear-localized GFP . Flies were placed into condos containing instant Drosophila media hydrated by a mixture of RU486 dissolved in methanol and water . We found that flies placed in the Fly Condo where the food contains RU486 are able to function as wild-type teachers and students ( Figure 15A–B ) . This observation demonstrates that RU486 does not perturb Drosophila's ability to perceive and respond to wasp presence by changing their oviposition behavior , as both flies expressing a Gene-Switch construct and His-GFP flies behaved as wild type . Our data also demonstrate that induction of a protein in MB , in this case GFP , does not perturb learning and memory formation nor teaching ability . When the assay is run with just methanol , therefore lacking RU846 , we find flies are able to function as wild type , similar to when they were fed RU486 ( Figure 15C–D ) . Control MB-Gene switch parental lines behave as wild-type flies as both homozygotes and when outcrossed to Canton-S . In cases when RU486 laden food was fed to flies containing the MB Gene Switch and GFP nuclear localization signal ( nls ) construct , we find that 24 hr is sufficient to induce GFP signal specifically localized to the MB , whereas food lacking RU486 ( methanol only ) does not induce GFP after 24 hr ( Figure 15E–G , Figure 15—figure supplement 1E–G ) . 10 . 7554/eLife . 07423 . 028Figure 15 . Induction of GFP in the MB using the Gene-Switch System does not perturb learning and memory . ( A to D ) Percent of eggs laid normalized to unexposed . ( A to B ) GFP induction with RU486 feeding in the MB as teachers and students . ( C to D ) Lack GFP induction with methanol feeding in the MB as teachers and students . Brains from flies expressing the GeneSwitch construct ( RU486+ ) in the MB along with a GFP nuclear localization signal ( nls ) showing ( E ) DAPI , ( F ) GFP expression , and ( G ) the merged image . Scale bar = 10 μm . ( For [A] to [D] n = 24 biological replicates . ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 02810 . 7554/eLife . 07423 . 029Figure 15—figure supplement 1 . Further evidence demonstrating that induction of GFP in the MB using the GeneSwitch System does not perturb learning and memory . ( A to D ) Percent of eggs laid normalized to unexposed . ( A and B ) MB GeneSwitch GAL4 as teachers and students . ( C and D ) MB GeneSwitch GAL4 /+ as teachers and students . Brains from flies not expressing the GeneSwitch construct ( methanol ) in the MB along with a GFP nls showing ( E ) DAPI , ( F ) GFP expression , and ( G ) the merged image . Scale bar = 10 μm . ( For [A] to [D] n= 24 biological replicates . ) ( **p< 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 029 Given the successful feeding protocol and the MB Gene-Switch construct specificity , we used the MB Gene-Switch to express an RNA-hairpin targeting mRNA for Orb2 . Induction of the RNA-hairpin through RU486 feeding in the MB was expected to occur within the same window of time as the GFP expression ( Figure 15 ) . Flies expressing the MB Gene-Switch and carrying the UAS-Orb2-RNA-hairpin construct , that were not fed RU486 , showed normal , wild-type memory , learning , and teaching ability ( Figure 16C–D ) . Flies expressing the MB Gene-Switch and carrying the UAS-Orb2-RNA-hairpin construct , which were fed RU486 , showed a wild-type acute response , but impaired memory formation , learning , and teaching abilities ( Figure 16A , B ) . These two data points suggest that the UAS-Orb2-RNA-hairpin construct is only driven in flies expressing the MB Gene-Switch when fed RU486 only . When the MB Gene-Switch parental control line was used to express an RNA-hairpin to the white gene , flies elicited wild-type memory formation with and without RU486 feeding , demonstrating that the Gene-Switch ligand ( RU486 ) alone and an RNA-hairpin alone is not responsible for memory , teaching , and learning impairment ( Figure 16E–H , Figure 16—figure supplement 1A–D ) . This observation again demonstrates that RU486 does not perturb Drosophila's ability to perceive and respond to wasp presence and that orb2 function is required for formation of a long-term memory of wasp exposure and not perception of and an acute response to wasps . 10 . 7554/eLife . 07423 . 030Figure 16 . Knockdown of Orb2 in the MB using the GeneSwitch System results in defective learning . ( A to H ) Percent of eggs laid normalized to unexposed . ( A to B ) Orb2 RNAi-knockdown in the MB ( GeneSwitch ) fed RU486 as teachers and students . ( C to D ) Orb2 RNAi-knockdown in the MB ( GeneSwitch ) not fed RU486 ( methanol fed ) . ( E to F ) White RNAi-knockdown in the MB ( GeneSwitch ) fed RU486 as teachers and students . ( G to H ) White RNAi-knockdown in the MB ( GeneSwitch ) not fed RU486 ( methanol fed ) . ( For [A] to [H] n = 24 biological replicates . ) ( *p < 0 . 05 , **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 03010 . 7554/eLife . 07423 . 031Figure 16—figure supplement 1 . Expression of an RNAi hairpin in the MB using the GeneSwitch System does not perturb learning and memory . ( A to D ) Percent of eggs laid normalized to unexposed . ( A to B ) White RNAi-knockdown in the MB ( GeneSwitch ) fed RU486 as teachers and students . ( C to D ) White RNAi-knockdown in the MB ( GeneSwitch ) not fed RU486 ( methanol fed ) . ( For [A] to [D] n= 24 biological replicates . ) ( **p < 1 . 0e-5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 031 Collectively , these data indicate that normal orb2 function is required in the adult MB for normal long-term memory formation and behavioral changes that persist over multiple days , such as the ability to teach . Use of the MB Gene-Switch construct provides strong evidence to delimit temporal and spatial expression requirements for orb2 function in the context of this memory assay . Importantly , Orb2-RNAi knockdown in the MB using either OK107-GAL4 or MB Gene-Switch did not prevent oviposition depression to occur when flies were in the presence of wasps . This also demonstrates that loss/diminution of orb2 function in the MB does not affect perception and acute response to this predator ( Figure 14A , B , Figure 16A–B ) .
In this study we have shown that Drosophila exhibit an acute response to predatory wasp that entails apoptosis of germ line cells within the ovary and corresponding reduced egg-laying behavior . The response persists over multiple days when learning and memory functions are intact . We also find that this behavior can be socially transmitted from experienced teacher females to naive student females: the transfer of information from teachers does not occur as a by-product of apoptosis in the teacher , but rather through an independent pathway , since depressed oviposition is not a necessary condition for social transmission of reduced egg-laying behavior or apoptosis in the student females ( Figure 17 ) . These conclusions are further supported by the unexpected observation that student flies , that had learned to reduce oviposition , could not serve as teachers ( Figure 2B , Figure 2—figure supplement 1 ) . We emphasize that teacher-instructed students continued to exhibit depressed oviposition and stage 7/8 egg chamber apoptosis in the 24-hr period after removal of teachers . This again indicates that depressed oviposition itself is not sufficient for information transfer . However , at a higher level , these observations also indicate that such adaptive information transfer cannot spread throughout a population , since only primary teachers are able to transmit the predator-threat information . 10 . 7554/eLife . 07423 . 032Figure 17 . Pathway model for fly-wasp mediated social learning . Initial oviposition depression during the 0- to 24-hr acute response period and information transmission during social learning 24- to 48-hr periods are not coupled . Sustained oviposition depression requires learning and memory genes in both teachers and students . Alleles tested for indicated genes were ninaBP315 , Orb2ΔQ , Adf1nal , dnc1 , dncML , rut1 , rut2080 , FMR1B55 , FMR13 , amn1 , amnX8 , wg1 , ewg1 , ewg2 , ewgNs4 , and drice-RNAi , Dcp-1-RNAi , Dcp-11 , Dcp-13 . DOI: http://dx . doi . org/10 . 7554/eLife . 07423 . 032 The above findings document a pathway initiated through visual stimulation and results eventually in a dramatic physiological response in the ovary . The discovery of neurally driven control of non-neural germ line cell physiology is conceptually similar to a recent study in Drosophila , which demonstrated that olfactory stimulation was necessary for maintenance of blood progenitor cells ( Shim et al . , 2013 ) , thus , also establishing a link between perception of environmental information and physiological response to specific information . Although learning mutants and flies expressing an RNA-hairpin to orb2 could perceive and respond to predator presence , the observation that egg production completely recovered by 24 hr following removal of the wasp threat ( Figure 12 , Figure 12—figure supplement 1 , Supplementary file 1K , L ) is consistent with previous observations where females switched from a poor to rich food source repress the mid-oogenesis checkpoint via insulin signaling and recover normal egg production within 24 hr ( Drummond-Barbosa and Spradling , 2001 ) . This rapid recovery of oviposition in learning and memory mutants , coupled with removing fly heads and inhibiting synaptic transmission in the MB , suggests that maintenance of the depressed oviposition state requires continued neural signaling mediated by a memory component of the brain . Our observations document and describe a particularly robust form of social learning in Drosophila and establish several fundamental features . First , direct learning and social learning require visual system function but occur through different mechanisms: in particular , the acute response of flies to direct wasp exposure can occur even in classic-learning mutants , while persistence of the predator response and subsequent social learning requires functions of learning genes and continued neural input . Loss of memory gene functions , such as Adf1 , amn , dnc , dFmr1 , rut , and Orb2 , or inhibition of MB synaptic transmission had no effect on the ability to change oviposition behavior in the presence of wasp , however , in each of these cases , persistence of this behavior after wasp removal , and subsequent teaching ability , was abolished . Additionally , inhibition of orb2 using the GAL4/UAS and Gene-Switch systems suggests that maintenance of the change in oviposition state requires neural signaling mediated by a memory component of the adult brain . Second , social learning occurs through a mechanism distinct from mimicry . Information of wasp presence can be transmitted by animals that have encountered wasps but are physiologically unable to display egg retention , which is the normal behavioral output of such learning ( Figure 17 ) . Third , social learning in this context appears to be limited in its spread: being transmitted only from teachers with direct predator experience to students that they encounter . Therefore , students that have learned through social learning cannot become teachers themselves ( Figure 2 ) . This is noteworthy because the inability of primary students to further transfer information to secondary students will limit the time frame and number of individuals in which this knowledge transfer takes place . The spreading of socially learned behavior has been previously postulated to possibly drive local adaptation by maintaining behavioral diversity of groups through self-propagating social learning once initiated in an individual ( Battesti et al . , 2012 ) . With regards to social learning of oviposition depression in response to a predator threat , it seems reasonable that such information would be most useful if limited to nearest neighbors , whose progeny may be similarly vulnerable in time and space by parasitoid wasps . However , the fitness costs of prolonged oviposition depression and/or spreading to conspecifics beyond primary learners could be devastating if it were self propagating , and thus , the degree to which it can spread within a group must be limited by restricting teaching behavior only to individuals having had direct visual experience of the threat , while ensuring memory of the threat in both primary ( teachers ) and secondary ( students ) learners is maintained and then decays over time . In sum , we have shown that visual inputs modify synaptic signaling in the MB of the fly brain to implement a behavioral and physiological change , both of which are transferable through extrinsic inputs to naive student flies , and that experiments based around wasp exposure can serve as a simple and robust learning , social learning , and memory paradigm in future D . melanogaster studies . The learning and memory genes we tested and found to be involved are conserved across many animal species ( Bolduc and Tully , 2009 ) , and thus , serves as an excellent approach to model cellular and neuronal network functions that may be relevant to vertebrate brain function . Even though the vertebrate brain is vastly more complex than that of the fly , additional genes , gene families , and pharmacological effects can be elucidated in Drosophila and may identify core mechanisms that are used in all species . These conserved components provide starting points in vertebrate animals for further vertical integration in the fields of learning and social communication . In this way , mechanisms that are unique to vertebrates can also be inferred , and we suggest that the learning and memory paradigm presented here will prove to be a useful discovery tool . We believe this study establishes a new and robust ecologically relevant model of social learning in Drosophila with possible far reaching implications for neurobiology , Darwinian selection and evolution .
The D . melanogaster strains Canton-S ( CS ) , Oregon-R ( OR ) , w1118 , and transgenic flies carrying Histone H2AvD-GFP ( His-GFP ) were used as wild-type strains for assaying egg retention in the presence of wasps . All subsequent experiments we performed using either CS or His-GFP flies as wild type . Orco1 , ninaBP315 , Dcp-1RNAi , driceRNAi , Histone H2AvD-GFP , and the Matα GAL4 mutant strains were acquired from the Bloomington Drosophila Stock Center ( strain numbers 23129 , 24776 , 28909 , 32403 , 35518 , and 7063 , respectively ) . dnc1 , dncML , rut1 , rut2080 , amn1 , and amnX8 were kindly provided by Leslie Griffith ( Brandeis University ) . Wg1 , ewg1 , ewg2 , ewgNS4 , MS1096 GAL4 , and UAS-Reaper ( UAS-Rpr ) flies were kindly provided by Yashi Ahmed ( Geisel School of Medicine at Dartmouth ) . Dcp-12 and Dcp-13 were kindly provided by Kim McCall ( Boston University ) . The MB Gene-Switch line and the MB-247 were kindly provided by Greg Roman ( Baylor College of Medicine ) ( Supplementary file 2 ) . All flies were maintained on standard cornmeal/yeast/molasses Drosophila medium . For all outcrosses , Canton-S virgin females were mated to males of the appropriate genotype . Flies aged 3–5 days post-eclosion on fresh , molasses-based , Drosophila media were used in all experiments . Stocks were maintained at 25°C in 70% humidity with a 12:12 light:dark cycle . For stocks maintained in vials , 25 females were kept for stocks with 10 males at maximum to prevent over-crowding . Stocks kept in bottles had a maximum of 100 females and 40 males to prevent over-crowding . When flies were close to eclosion , parents were removed from the bottles . Newly eclosed flies were moved to fresh Drosophila media ( in bottles or vials at the same population density ) and aged until they were between 3 and 5 days of age maintained at 25°C in 70% humidity with a 12:12 light:dark cycle , at which point they were used in experiments . We stress the importance of aging the flies on fresh media , as it appears that flies aged on old media ( i . e . , the same media in which they eclosed ) are nutrient deprived and naturally lay very few eggs . The Figitid larval endoparasitoid Leptopilina heterotoma ( strain Lh14 ) was used in all experiments . L . heterotoma strain Lh14 originated from single females collected in Winters , California in 2002 , and was kindly provided by Todd Schlenke ( Schlenke et al . , 2007 ) . In order to culture wasps , adult flies were allowed to lay eggs in standard Drosophila vials containing standard Drosophila medium for 4 days before being replaced by adult wasps ( 10 female , 6 male ) , which then attack the developing fly larvae . Wasp vials were supplemented with approximately 500 µl of a 50% honey/water solution applied to the inside of the cotton vial plugs . Wasps aged 3–7 days post-eclosion were used for all experiments . Fresh wasps were used for all experiments , such that wasps were never reused between experiments . Fly oviposition rates were conducted using The Fly Condo ( Genesee Scientific ( San Diego , CA ) Cat # 59-110 ) ( Figure 1A ) , which contained 24 independent chambers . Each chamber is 7 . 5 cm long by 1 . 5-cm diameter . Each condo/chamber had a bottom 24-well food plate with approximately 2 ml of standard , molasses cornmeal media per chamber . Briefly , bottles containing Drosophila were microwaved for 30 s at maximum heat . This liquid food was allowed to cool before dispensing 2 ml into the Fly Condo plates , where food was allowed to cool for another 30 min before the start of the experiment . All experiments used this food protocol unless otherwise noted ( specifically experiments using instant Drosophila media with RU486 experiments ) . Mesh wire was along the top of the condo , allowing air transfer . In order to assay egg retention of flies in the presence of wasps ( acute exposure ) , 5 female flies and 1 male fly ( prepared and aged as described above ) were placed into one chamber of The Fly Condo in the control , while 3 female Lh14 wasps were placed with the flies in the experimental setting . The oviposition plate from control and experimental condos was made 24 hr later . In order to assay fly communication and the social learning period , 5 female flies and 1 male fly were placed into one chamber of The Fly Condo in the control , while 3 female Lh14 wasps were placed with the flies in the experimental setting for 24 hr . After the 24-hr exposure , wasps were removed by anesthetizing flies and wasps in the condos . Control flies underwent the same anesthetization . Wasps were removed and replaced with 3 female ‘student’ flies . All flies were placed into new clean condos for the second 24-hr period . The oviposition plate from each fly condo was replaced 24 hr after the start of the experiment , and the second plate was removed 48 hr after the start of the experiment . Fly egg counts from each plate were made at the 0–24-and 24–48-hr time points . To control for both seasonal influence and population effects of both flies and wasps used , acute- and social-learning period experiments were repeated in 24-experimental replicate increments in both August 2013 and April 2014 . We found the same effect in both time points tested , suggesting that seasonal changes and population effects were not affecting our results ( Figure 1C and Figure 1—figure supplement 1C–E ) . In order to demonstrate that students cannot become teachers , the same protocol as above was performed ( Figure 2 ) . At the 48-hr time point , exposed teacher flies were removed by anesthetizing teacher and 1° student flies in the condos . Control flies underwent the same anesthetization . The exposed teacher flies were replaced with 3 new naive student flies , termed 2° students . These 2° students were placed with 1° student flies into new condos for the third 24-hr period . Fly egg counts from each oviposition plate were made at the 0–24 , 24–48 , and 48–72 hr time points . In order to demonstrate that teachers could teach more than one cohort of students , the same protocol as above was performed with the exception of teacher removal ( Figure 3 ) . At the 48-hr time point , 1° student flies were removed by anesthetizing students and teachers and teacher flies were placed into new condos . The 1° student flies were replaced with 3 new naive student flies , termed 2° students . These 2° students were placed with teacher flies into new condos for the third 24-hr period . Fly egg counts from each oviposition plate were made at the 0–24 , 24–48 , and 48–72 hr time points . To assay if the ratio of teachers to students impacted the ability for information transfer , 3 female flies were placed into one chamber of The Fly Condo in the control , while 3 female Lh14 wasps were placed with the flies in the experimental setting for 24 hr . Wasps were then removed and replaced with 3 female , naive student flies of the opposite genotype ( either His-GFP or Canton-S depending on teacher identity ) . Flies were then placed into new , clean condos . This provided a 1:1 ratio of teachers to students . The same protocol was performed to see if males participated in transmission of information by having 3 male flies exposed to 3 female Lh14 wasps . Wasps were then removed and replaced with 3 female , naive student flies of the opposite genotype ( either His-GFP or Canton-S depending on teacher identity ) . In order to assay the role of light during initial exposure and the role of light during the learned response , multiple assays were performed where light availability was varied ( Figure 8 ) . For experiments where the acute response occurred in the dark , flies were anesthetized and placed into The Fly Condo with or without wasps as described above . However , they were then immediately placed into a box , taped closed with Duct Tape ( to prevent light leaks ) , and allowed to awaken in the dark . Flies were kept in the dark for either 24 hr , after which wasps were removed and students were added and moved into the light , or kept in the dark for the duration of the experiment ( 48 hr ) including the social-learning period with students . If flies were to be kept in the dark , the only light they were exposed to was just before they were anesthetized and given students . All treatments were run at 25°C at 70% humidity with a 12:12 light:dark cycle in twenty-four replicates unless otherwise noted with both teacher and student flies aged 3–5 days post-eclosion . Food used for Fly Condo plates was the same molasses based Drosophila media used in maintaining fly stocks , unless otherwise noted . Fly condos and oviposition plates were bleached thoroughly with 10% bleach and rinsed with distilled water mixed with Sparkleen after every use ( 1 gallon of water: 1 gram of Sparkleen ) . All egg plates were coded and scoring was blind as the individual counting eggs were not aware of treatments or genotypes . To assay whether flies continued to eat high-nutrient food during wasp exposure , flies were placed into a large embryo collection chamber ( Genesee Scientific ( San Diego , CA ) Cat No . 59-101 ) , which fits a 100-mm Petri dish . Dishes were filled with 5 grams of blue instant drosophila media ( Fisher Scientific ( Pittsburgh , PA ) Cat No . S22315C ) , supplemented with a total of 20 ml of distilled water to hydrate the food . Yeast paste was made with 15 ml distilled water , 5 mL McCormick's red food dye , and 13 mL live yeast . Approximately , 15 mL of the yeast paste solution was added to the center of the petri dish containing the instant Drosophila media . In the egg lay chambers , 100 female Canto- S and 20 male Canton-S flies were added for control conditions . For exposed conditions , 100 female Canton-S and 20 male Canton-S flies were added with the addition of 50 female Lh14 . The experiment was run for 24 hr at 25°C in 70% humidity on a 12:12 light:dark cycle . After 24 hr , flies were anesthetized and were scored for color in their abdomens . A random subset of 12 females was taken after abdominal quantification for ovary dissection and DAPI staining . Three replicates were performed for these experiments . To assay whether or not wings were involved in the information transmission in the social learning period , 5 female and 1 male Canton S were anesthetized and their wings were cut at the base using micro-scissors ( Fine Science Tools ( Foster City , CA ) ; Item No . 15 , 001-08 ) . Following clipping , flies were placed into one chamber of The Fly Condo in the control , while 3 female Lh14 wasps were placed with the flies in the experimental setting for 24 hr . After the 24-hr exposure , wasps were removed by anesthetizing flies and wasps in the condos . Control flies underwent the same anesthetization . Wasps were removed and replaced with 3 female ‘student’ flies . All flies were placed into new clean condos for the second 24-hr period . The oviposition plate from each fly condo was replaced 24 hr after the start of the experiment , and the second plate was removed 48 hr after the start of the experiment . Fly egg counts from each plate were made at the 0–24 and 24–48 hr time points . In order to assay whether a continued input from the brain is needed for flies to remember wasp exposure and to transmit that information , 5 female flies and 1 male fly were placed into one chamber of The Fly Condo in the control , while 3 female Lh14 wasps were placed with the flies in the experimental setting for 24 hr . After the 24-hr exposure , wasps were removed by anesthetizing flies and wasps in the condos . Control flies underwent the same anesthetization . During this anesthetization period , both male and female flies were decapitated using the micro-scissors . Decapitated flies that were not standing after anesthesia recovery were excluded . Wasps were removed and replaced with 3 female ‘student’ flies . All flies were placed into new clean condos for the second 24-hr period . The oviposition plate from each fly condo was replaced 24 hr after the start of the experiment , and the second plate was removed 48 hr after the start of the experiment . Fly egg counts from each plate were made at the 0–24 and 24–48 hr time points . Fly duplexes ( Figure 9 ) were constructed by using three standard 25 mm × 75-mm glass microscope slides ( VWR ( Radnor , PA ) : Item No . 48 , 300-025 ) that were adhered between two 75 mm × 50 mm × 1-mm glass microscope slides ( Fisher: Item No . 12-550C ) . Clear aquarium silicone sealant was used to glue these glass slides together , making two compartments separated by one 1-mm thick glass slide . Sealant was allowed to cure for 48 hr; each duplex was then soaked in water and Sparkleen detergent overnight ( 1 gallon distilled water: 1 gram Sparkleen ) , rinsed in distilled water ( dH2O ) overnight , rinsed with 70% ethanol and air-dried . The interior dimensions of each of the two units measured approximately 23 . 5 mm ( wide ) × 25 mm ( deep ) × 75 mm ( tall ) . For experiments using Fly Duplexes , plates from The Fly Condo ( Genesse Cat . Item No . 59-113 ) were filled to the rim with standard Drosophila media and allowed to cool . Upon cooling , a single Fly Duplex was inserted into the food such that it touched the bottom of the plate . The open end of the Fly Duplex was closed using a cotton plug ( Genesse Scientific ( San Diego , CA ) Cat . Item No . 51-102B ) to prevent insect escape . 10 female flies and 2 male flies were placed into one chamber of the Fly Duplex in the control , while 10 female Lh14 wasps were placed in the compartment adjacent to the flies in the experimental setting for 24 hr . After the 24-hr exposure , flies and wasps were removed by anesthetizing flies and wasps in the Fly Duplexes . Control flies underwent the same anesthetization . Wasps were removed and replaced with 10 female ‘student’ flies . All flies were placed into new clean Duplexes for the second 24-hr period . The oviposition plate from each fly condo was replaced 24 hr after the start of the experiment , and the second plate was removed 48 hr after the start of the experiment . Fly egg counts from each plate were made at the 0–24 and 24–48 hr time points . RU486 ( Mifepristone ) was used from Sigma-Aldrich Corp . ( St . Louis , MO ) ( Lot Item No . SLBG0210V ) . Condos were prepared by measuring 0 . 375 grams of flaky instant blue Drosophila medium into each well of The Fly Condo plates . For all food treatments , we pipetted a total liquid volume of 2250 µl directly onto the instant food . For experiments with RU486 , an RU486 solution was used . This was prepared by dissolving 3 . 575 mg of RU486 in 800 µl methanol ( Fisher Scientific ( Pittsburgh , PA ) Lot number 141313 ) . This solution was added to 15 . 2 ml of distilled water . The total solution ( 16 ml ) was thoroughly mixed and 2250 µl were pipetted onto the instant food into each well . For plates containing no RU486 ( methanol only ) , 800 µl methanol was mixed with 15 . 2 ml of distilled water . The total solution ( 16 ml ) was thoroughly mixed and 2250 µl were pipetted onto the instant food into each well . Ovaries that were prepared for immunofluorescence were fixed in 4% methanol-free formaldehyde in PBS with 0 . 001% Triton-X for approximately 5 min . The samples were then washed in PBS with 0 . 1% Triton-X and blocked with 2% normal goat serum ( NGS ) for 2 hr . The primary antibody , rabbit cleaved caspase-3 ( Cell Signaling ( Beverly , Massachusetts ) 5 A1E ) at a concentration of 1:400 , was incubated overnight at 4°C in 2% NGS . The secondary antibody used was Cy3 conjugated ( Jackson Immunoresearch ( West Grove , PA ) ) and used at a concentration of 1:150 during a 2-hr incubation at room temperature . This was followed by a 10-min nuclear stain by DAPI . In order to assay whether feeding flies RU486 in The Fly Condo would be sufficient to turn on the MB gene switch construct , we placed flies into condos containing RU846+ food . Flies had the MB switch construct as well as a UAS-GFP nls construct , such that if the MB switch is activated , it should fluoresce with GFP . After a 24-hr period in The Fly Condo , adults were removed and fixed in 4% methanol-free formaldehyde in PBS with 0 . 001% Triton X overnight at 4°C . Brains were then dissected out of whole adults in PBS . The samples were then washed in PBS with 0 . 1% Triton X and stained with DNA staining with DAPI , for 10 min and mounted in Vectashield ( Vector Laboratories ( Burlingame , CA ) Item No . H-1000 ) before imaging . Individual ovarioles were dissected and fixed in PBS with 4% methanol-free formaldehyde and 0 . 1% Triton-X for 30 min . Ovarioles were washed and incubated in PBS with 20 µg/ml proteinase K for 10 min . Recombinant terminal transferase ( Tdt ) labeling was conducted with the use of Cy3-conjugated dUTP ( GE Healthcare ( Troy , NY ) PA53032 ) . Tdt reaction mixture ( 200 mM NaCacodylate , 0 . 1 mM DTT , 1 mM CoCl2 , 0 . 05 mM Cy3-dUTP , 0 . 05 mM dTTP ) in Tdt buffer and Tdt enzyme ( Roche ( Basel , Switzerland ) 03333566001 ) was incubated with samples for 3 hr at 37°C in a dark hybridization oven . At the end of the incubation period , 2 μl of ( 0 . 25 M ) EDTA was added to stop the reaction . Samples were counter-stained with DAPI , mounted in Vectashield , and stored at −20°C until imaging . For quantification of egg chamber apoptotic events , ovaries from exposed teachers and exposed students ( in addition to unexposed controls ) were fixed in 4% methanol-free formaldehyde in PBS with 0 . 001% Triton X for approximately 5 min . The samples were then washed in PBS with 0 . 1% Triton X and stained with DAPI for 10 min . Batches of student and teacher flies were stained together in the same wells to prevent stain bias . In all cases , student and teacher ovaries on the same slides could be distinguished based on the Histone H2AvD-GFP marker ( Figure 4—figure supplement 1A , B ) . A Nikon ( Melville , New York ) A1R SI Confocal microscope was used for imaging TUNEL , brain , and caspase staining . Image averaging of 4× during image capture was used for all images unless otherwise specified . A Nikon E800 Epifluorescence microscope with Olympus DP software was used to quantify apoptotic events in egg chambers in addition to the capture of egg images and of whole flies ( Figure 4B , C , F , G , Figure 4—figure supplement 1A–J , M–T ) . Images of The Fly Condo , oviposition plates with red yeast paste , and low-magnification images of exposed and unexposed flies with red abdomens were made using an iPad 2 operating with ISO 64 ( Figure 1A , Figure 4A , Figure 4—figure supplement 1K–L ) . Images of The Fly Condo and the Fly Duplex were color enhanced in iPhoto ( Figure 1A , Figure 9A ) . Statistical tests were preformed in R ( version 3 . 0 . 2 , ‘Frisbee Sailing’ ) . Welch’s two-tailed t-tests were preformed for all egg count data . p-values reported were calculated for comparisons between paired treatment-group and unexposed . A chi square test was preformed to determine significance of feeding experiments for frequency of colored abdomens . Welch's two-tailed t-tests were performed on apoptosis data with each exposure batch treated as a replicate ( n = 3 ) , in instance where both the treatment and control group had 0% apoptosis across all of the three replicates the p-value was not calculable , and is reported as ‘N/A’ ( See Supplementary files 3–5 ) . | Every animal must be able to adapt to threats and changes to their environment that could affect their survival . Some ‘social’ animals , such as honeybees and ants , go further than this , and also transmit information about a threat—and how to survive it—to other members of their species . This helpful behavior is now known to occur to some extent even in animals that have not been considered to be social , like the Drosophila species of fruit fly . Parasitoid wasps lay their eggs in the larvae and pupae of certain insect species . When the wasp eggs hatch , they feed on the host insect , eventually killing it . Drosophila fruit flies have evolved various behaviors to protect their offspring from these wasps . For example , female fruit flies reduce the number of eggs they lay when they are in the presence of a wasp . Kacsoh , Bozler et al . exposed female flies to wasps for a day . These flies produced fewer eggs than flies that were not exposed to wasps and continued to lay fewer eggs for 24 hours after the wasps were removed . Introducing these flies to ‘naive’ flies that had not encountered a wasp caused the naive flies to produce fewer eggs as well . After ruling out several possible ways that the wasp-exposed flies might ‘teach’ the naive flies to produce and lay fewer eggs , Kacsoh , Bozler et al . found that naive flies cannot learn this behavior when they are blind . In addition , exposed flies cannot instruct other flies of the threat if their wings are absent or deformed . These and other findings , therefore , suggest that information about the wasp threat is transmitted through visual cues that involve the wings . Kacsoh , Bozler et al . found that the flies must have certain brain circuits associated with memory and learning to be able to teach others and to reduce the numbers of eggs they lay after the wasp has been removed . This suggests that signals from this brain region must be continually sent out to alter the physiology of the developing eggs in order to maintain the lower rate of egg laying; understanding how flies use visual cues for communication and how the brain signals to the ovary remain key challenges for future work . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2015 | Social communication of predator-induced changes in Drosophila behavior and germ line physiology |
To provide an effective substrate for cognitive processes , functional brain networks should be able to reorganize and coordinate on a sub-second temporal scale . We used magnetoencephalography recordings of spontaneous activity to characterize whole-brain functional connectivity dynamics at high temporal resolution . Using a novel approach that identifies the points in time at which unique patterns of activity recur , we reveal transient ( 100–200 ms ) brain states with spatial topographies similar to those of well-known resting state networks . By assessing temporal changes in the occurrence of these states , we demonstrate that within-network functional connectivity is underpinned by coordinated neuronal dynamics that fluctuate much more rapidly than has previously been shown . We further evaluate cross-network interactions , and show that anticorrelation between the default mode network and parietal regions of the dorsal attention network is consistent with an inability of the system to transition directly between two transient brain states .
The presence of large-scale distributed networks of temporally correlated spontaneous activity is a well-established phenomenon in neuroimaging ( Biswal et al . , 1995; Fox and Raichle , 2007; Raichle et al . , 2001 ) . These so-called resting state networks ( RSNs ) have been consistently identified across different subjects in the absence of any explicit task from covariations in the blood oxygenation level dependent ( BOLD ) signal , as measured by functional magnetic resonance imaging ( fMRI ) ( Beckmann et al . , 2005; Damoiseaux et al . , 2006; Smith et al . , 2009 ) . These networks are known to have functional relevance and clinical significance ( Greicius et al . , 2004; Filippini et al . , 2009 ) . However , an important limitation in our understanding of spontaneous activity is how the low frequency fluctuations typically associated with RSNs are related to the much faster timescales of cognitive and sensory processing ( Heeger and Ress , 2002; Deco and Corbetta , 2011; Raichle , 2011; Siegel et al . , 2012 ) . This is in part due to the fact that the indirect hemodynamic response measured via the BOLD signal precludes studying the rich temporal dynamics of the underlying electrophysiological activity ( Logothetis , 2008 ) . In contrast , non-invasive electrophysiological recordings such as magnetoencephalography ( MEG ) and electroencephalography ( EEG ) provide a direct measure of neuronal activity at high temporal resolution . Recent studies using these modalities have revealed that these networks have an electrophysiological basis ( Laufs et al . , 2003; He et al . , 2008; Jann et al . , 2010; Liu et al . , 2010; Brookes et al . , 2011 ) , and are underpinned by much richer spatiotemporal dynamics that may be better characterized using time-varying measures of interactions ( Brookes et al . , 2014; Chang et al . , 2013; Hutchison et al . , 2013; de Pasquale et al . , 2010 , 2012 ) . These studies have shown evidence that functional connectivity within whole brain networks exhibit temporal variability on a time scale of seconds to tens of seconds . However , to provide an effective substrate for cognitive processes , functional networks should be able to rapidly reorganize and coordinate on a sub-second temporal scale ( Bressler and Tognoli , 2006 ) . Here , we present a study that identifies transient networks of brain activity , with no prior assumptions on the brain areas or time scales involved . This uses a distinct methodology based on a hidden Markov model ( HMM ) , which infers a number of discrete brain states that recur at different points in time . Each inferred state corresponds to a unique pattern of whole-brain spontaneous activity , which is modeled by a multivariate normal distribution and a state time course indicating the points in time at which that state is active . These two outputs are shown schematically in Figure 1 , and allow us to describe both the spatial and temporal characteristics of each inferred state . 10 . 7554/eLife . 01867 . 003Figure 1 . Schematic of the HMM outputs . An HMM with K states is inferred from band-limited amplitude envelopes of source reconstructed MEG data . Each state is characterized by a multivariate normal distribution ( defined by means μk and covariance matrix Σk ) and a state time course , which is a binary sequence that indicates the points in time at which the state is active . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 003 By applying this methodology to band-limited amplitude envelopes of source reconstructed MEG data , we show that the HMM can independently identify brain states in MEG data that correspond to established RSNs , and which fluctuate at time scales two orders of magnitude faster than has previously been shown . We further assess cross-network interactions , and show that the antagonistic behavior between the default mode network ( DMN ) and parietal regions of the dorsal attention network ( DAN ) is consistent with an inability of the system to transition directly between two of these states .
State specific changes in oscillatory amplitude revealed spatial patterns of activity with good similarity to several well-known networks , previously associated with brain wide correlations at slow ( <0 . 1 Hz ) timescales ( Figure 2 ) . Each map represents the partial correlation between the state time courses and the group-concatenated amplitude envelope at each voxel . Accordingly , state specific increases and decreases in amplitude are represented by red/yellow and blue colors respectively . State 1 shows increased activity in nodes of the default mode network ( DMN ) including left and right inferior parietal lobule , medial frontal gyrus and medial temporal lobe ( but notably , not in the posterior cingulate/precuneus ) . States 2–6 show increased activity in the visual cortex ( states 2 and 6 ) , the sensorimotor network ( state 3 ) , and the left and right lateralized temporal lobes ( states 4 and 5 ) . States 7–8 show decreases in activity ( blue in Figure 2 ) in parietal regions including the intraparietal sulcus ( IPS; state 7 ) , and visual cortex ( state 8 ) . 10 . 7554/eLife . 01867 . 004Figure 2 . State specific changes in band-limited amplitude . An 8 state HMM was inferred from temporally concatenated band-limited amplitude time courses ( concatenated over nine subjects , 10 min each ) . The volumes and surface renderings show the partial correlation of each state time course with the envelope data at each voxel . The correlation values have been thresholded between 60% and 100% of the maximum correlation for each state and the color maps represent these ranges . Red/yellow and blue colors indicate positive and negative correlations respectively . See also Figure 2—figure supplement 1 for equivalent results from HMMs inferred with 4–14 states . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 00410 . 7554/eLife . 01867 . 005Figure 2—figure supplement 1 . Maximum intensity projection maps showing the partial correlation computed between each state time course and the envelope data for a k state HMM for k = 4 to k = 14 . Increasing the number of states did not change the topographies of the most prominent RSN-like states nor did it reveal any new RSN-like topographies that were distinct from those inferred with 8 states , but rather resulted in the splitting of states into multiple similar maps . This suggests that there is no advantage to using more than 8 states for the purpose of identifying states corresponding to known RSNs . In fact , it could be argued that fewer than 8 states are required for this purpose . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 00510 . 7554/eLife . 01867 . 006Figure 2—figure supplement 2 . Spatial maps of five of the inferred states alongside a matched RSN derived from application of spatial ICA to resting state fMRI data ( Smith et al . , 2009 ) . For each HMM–RSN pair , the spatial correlation is shown alongside the maps . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 006 The temporal properties of each state were characterized from the state time courses , which indicate the points in time at which each state is active . By inspection it is evident that the states are short lived ( Figure 3A ) . The temporal characteristics of each state may be quantified in terms of their fractional occupancy ( fraction of the total time spent in a state; Figure 3B ) , life times ( time spent in a state before making a transition; Figure 3C ) , and interval lengths ( time between consecutive state visits; Figure 3D ) . Average life times are between 100 ms and 200 ms . These life times are markedly shorter in duration than the time scales typically associated with resting state networks , which have previously been shown to be dominated by frequencies below 0 . 1 Hz ( Cordes et al . , 2001 ) . We also characterized variations in the rate at which these states are visited by computing the fractional occupancy within a 10-s sliding window . These fractional occupancy time courses reveal slower temporal changes in the occurrence of the HMM states . These time courses are shown for all subjects in Figure 3E . It is clear that each state was represented in all subjects . 10 . 7554/eLife . 01867 . 007Figure 3 . Temporal characteristics of the HMM states . ( A ) State time courses showing the most likely state at each time point for the first 10 s of data . ( B ) Fractional occupancy for each inferred state showing the mean and s . e . m . over subjects . The asterisks denote that the fractional occupancy of a state differs significantly from the other states . ( C ) Life times , and ( D ) interval lengths for each inferred state . The filled areas in ( C ) and ( D ) represent the distribution of values and the black crosses show the mean . ( E ) Fractional occupancy of each state as a function of time over all subjects , derived by averaging each state time course within a 10-s sliding window ( 75% overlap between adjacent windows ) . See also Figure 3—figure supplement 1 for a description of how these statistics vary when the HMM is inferred with 4–14 states . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 00710 . 7554/eLife . 01867 . 008Figure 3—figure supplement 1 . Effect of number of states on ( A ) model evidence , approximated by the negative of the free energy , ( B ) minimum fractional occupancy and ( C ) mean life time , computed over all inferred states and 50 realizations of each HMM inference . Error bars show the mean and s . e . m . over all subjects . The free energy monotonically increases up to 15 states suggesting that the Bayes-optimal model may require an even higher number of states ( A ) . However , a larger number of states result in a decrease in minimum fractional occupancy ( B ) . This may provide a more meaningful metric since a sufficient amount of data is required for reliable estimation of the covariance matrix . Increasing the number of states also has an effect on the mean life times of the inferred states . Varying the number of states from 4 to 15 results in mean lifetimes decreasing from 230 ms to 140 ms ( C ) . This suggests that the splitting of states that arises from increasing the number of states does not result in states with fewer occurrences , but rather a splitting into shorter-lived events . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 008 We have shown that whole brain spontaneous activity may be broken down into a set of distinct connectivity patterns that appear to be stable for periods of 100–200 ms . To confirm that these brain states are consistent with coordinated fluctuations at these rapid time scales , we performed a follow up analysis using the inferred state time courses . We reasoned that if there were coordinated fluctuations at the fastest time scales in the state time course , then low pass filtered versions of the state time courses should do worse at explaining fluctuations in the data . Different low pass filtered versions of the state time courses were obtained by computing fractional occupancy time courses using a range of time windows from 0 . 1 to 8 s . Note that for the shortest time windows , the fractional occupancy time course approximates the state time course . These fractional occupancy time courses were then separately regressed onto the amplitude envelope time course from a representative voxel of the corresponding brain state ( the voxel that most correlated with the state time course ) . This analysis reveals a peak in correlation for window widths between 200 ms and 400 ms , demonstrating that when using the fractional occupancy the fastest time scales at which we can detect fluctuations in the amplitude envelopes are slower than those suggested by the HMM state life-times ( ∼100 ms ) , but still much faster than has been previously shown ( Figure 4 ) . 10 . 7554/eLife . 01867 . 009Figure 4 . Analysis of the time scales that best reflect within-network envelope fluctuations . Partial correlation maps and fractional occupancy time window dependency are shown for ( A ) DMN , ( B ) visual network and ( C ) sensorimotor network . The fractional occupancy time window dependency was computed by fitting a GLM to the amplitude envelope at the voxel with highest correlation with the state time course with the fractional occupancy ( computed within different time windows ) as a single regressor . See also Figure 4—figure supplement 1 for a control analysis with surrogate data . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 00910 . 7554/eLife . 01867 . 010Figure 4—figure supplement 1 . Control analysis of the time scales that best reflect within-network envelope fluctuations . Correlation maps , mean life times , and fractional occupancy time window dependency for 8 state HMMs inferred from ( A ) the original group-concatenated envelopes , ( B ) these envelopes low pass filtered below 0 . 5 Hz and ( C ) these low pass filtered envelopes with uncorrelated Gaussian noise added . Results are shown for the default mode network ( top/red ) , visual network ( middle/blue ) and sensorimotor network ( bottom/green ) . The correlation maps were computed by fitting a GLM to the data with the state time courses for all states as regressors . The fractional occupancy time window dependency was computed by fitting a GLM to the data with the fractional occupancy time course ( computed within different time windows ) as a single regressor . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 010 As a control , we repeated the HMM inference after first removing any potential high frequency network interactions from the data . The group-concatenated amplitude envelopes were low pass filtered below 0 . 5 Hz to remove any higher frequency dynamics and an 8 state HMM was inferred from these filtered envelopes . Despite removing the faster envelope fluctuations , a number of states were inferred with similar spatial topographies as the original HMM ( Figure 4—figure supplement 1A , B ) . The life times of these states were longer , at around 1 s , reflecting the slower time scales of these low pass filtered signals . Next , to test whether simply introducing high frequency noise could result in shorter life times , band-limited Gaussian noise ( low pass filtered below 10 Hz , reflecting the spectral content of the original envelopes , and therefore with the same non-independence properties between time points ) , were added to the envelopes prior to inferring the HMM . As before , states with spatial topographies similar to the original HMM states were identified ( Figure 4—figure supplement 1C ) . However , the life times of these states were now much shorter , at around 0 . 3 s . As with the real data , we then fitted a GLM with a single regressor that corresponded to the fractional occupancy time course computed using a range of time windows from 0 . 1 s to 8 s . When applied to the low pass filtered dataset ( without added Gaussian noise; Figure 4—figure supplement 1B ) , this analysis reveals a peak in correlation for a window width of ∼1 s , consistent with the longer state life times . However , for the low pass filtered data set with added Gaussian noise ( Figure 4—figure supplement 1C ) , while the state life times were reduced to ∼100 ms , the peak in correlation remains at a window width of ∼1 s . This demonstrates that in this control there are no detectable within-network amplitude fluctuations faster than 1 s , which is consistent with the applied 0 . 5 Hz low pass filtering . This is in stark contrast to the real data ( Figure 4 ) , where the analysis reveals a peak in correlation for window widths between 200 ms and 400 ms . Networks of whole brain spontaneous activity may be characterized not only in terms of their within-network activity , but also in terms of cross-network interactions . To assess the relationship between different functional networks in the context of the HMM , we examined the relationship between the points in time at which different states are active . An important result from fMRI studies is the observation that the DMN exhibits anticorrelation with networks associated with attention demanding tasks , such as the dorsal attention network ( DAN ) ( Fox et al . , 2005; Smith et al . , 2009 ) . Of particular interest therefore is the relationship between state 1 , which represents the DMN , and state 7 , which we postulate may represent parietal regions of the DAN . While the correspondence between BOLD and electrophysiological activity has yet to be fully understood , it has been shown that alpha and beta power at rest correlate positively with BOLD in the DMN and negatively with BOLD in the DAN ( Mantini et al . , 2007 ) . Accordingly , increased activity in the DMN ( red/yellow in state 1 ) and decreased activity in parietal regions of the DAN ( blue in state 7 ) would both correspond to an increase in the BOLD signal . Based on this reasoning , we hypothesized that the antagonistic behavior of these two networks as shown in multiple fMRI studies would manifest in the inferred HMM states as an anticorrelation between the fractional occupancy time courses of these two states . In other words , periods of time in which the DMN state is frequently visited would coincide with periods of time in which the putative DAN state is rarely visited ( and vice versa ) . To this end , we computed the correlation coefficient between the fractional occupancy time courses for each state ( computed within 10-s sliding windows as shown in Figure 3E ) . Positive correlations between a pair of states indicate that the two states are visited more frequently during similar periods of time . As predicted , we found strong antagonistic behavior between the DMN ( state 1 ) and the putative DAN ( state 7 ) , suggesting an electrophysiological basis to the anticorrelated nature of these networks ( Figure 5A ) . 10 . 7554/eLife . 01867 . 011Figure 5 . Relationship between states . ( A ) Correlation matrix between the fractional occupancy time courses of each state . Positive correlations between a pair of states indicate that the two states are visited more frequently during similar periods of time . ( B ) State transition matrix for the group HMM . The matrix shows the probabilities of transitioning to any particular state given the current state . The probability of remaining in the same state has been excluded from each matrix ( shown in white ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 011 We further assessed the relationship between different functional networks by examining the transitions between the inferred states . These transitions may be represented in the form of a transition matrix , where each row represents the probability of transitioning to any other state given the current state ( Figure 5B ) . There is a clear structure to the matrix showing that transitions between certain pairs of states are more likely than others . A number of these transitions are intuitive , for example the strong probability of transitioning between visual states 2 and 6 . There is a very low probability of transitioning between the DMN ( state 1 ) and the putative DAN ( state 7 ) , raising the intriguing possibility that anticorrelation between these networks may arise from an inability of the system to transition directly between these two transient states . A number of previous studies have assessed the temporal dynamics of resting state connectivity by computing measures of functional connectivity such as band-limited amplitude correlation within sliding time windows of several seconds or longer ( Allen et al . , 2012; Brookes et al . , 2014; Chang and Glover , 2010; de Pasquale et al . , 2012 ) . To investigate whether any relationship exists between temporal variability in amplitude correlation and the occurrence of states inferred from the HMM , we computed the sliding window envelope correlation between different nodes of the DMN and compared the resulting correlation time courses with changes in the fractional occupancy of the inferred DMN state . Regions of interest were defined in the centers of the nodes corresponding to the inferior parietal lobule ( IPL ) , medial frontal gyrus ( MFG ) and medial temporal lobe ( MTL ) in both hemispheres ( Figure 6A ) . The envelope of oscillatory activity was computed at these six locations for each subject . With reference to previous findings ( de Pasquale et al . , 2010 ) , envelope correlation was computed within 10-s sliding windows between all six ipsilateral ROI pairs and all three contralateral ROI pairs . The HMM state corresponding to the DMN was identified and the fractional occupancy was computed within the same 10-s sliding window . The time courses of these two measures are shown for a single subject in Figure 6B , where the DMN pair shown is the right IPL and right MFG . In line with previous findings , it is evident that the envelope correlation between different network nodes alternates between periods of high and low correlation ( de Pasquale et al . , 2010 ) . Interestingly , periods of high envelope correlation are generally associated with an increase in the fractional occupancy of the DMN state , suggesting that fluctuations in interregional functional connectivity represent periods of time in which a particular transient network is frequently visited . This relationship may be quantified as the correlation between the two time series for each subject and node pair . There is a clear positive correlation between the two measures for all inter- and intra-hemispheric pairs ( Figure 6C ) . 10 . 7554/eLife . 01867 . 012Figure 6 . Comparison between HMM state occupancy and sliding window envelope correlation . ( A ) Six nodes identified from the DMN state ( rIPL , lIPL , rMFG , lMFG , rMTG , lMTG ) . ( B ) Time courses of rIPL-rMFG envelope correlation ( blue ) and DMN state fractional occupancy ( green ) for a single subject computed using a 10-s sliding window ( 75% overlap between adjacent windows ) . ( C ) Correlation between the sliding window envelope correlation time course and fractional occupancy time course for each ipsilateral pair ( bilaterally homologous pairs from the left and right hemispheres have been averaged together ) and each contralateral pair ( mean and s . e . m . over subjects ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01867 . 012
By inferring an HMM from the amplitude envelopes of group-concatenated data , we identified spatial maps representing state specific increases in oscillatory amplitude in anatomical locations corresponding to well-known RSNs . These networks include the default mode , visual and sensorimotor networks and parietal regions of the DAN . In addition to inferring the spatial covariance structure , the HMM also infers the time points at which each state is active . Each state was well represented over the group with all occupancies between 5% and 20% of the total time . The state life times were short , between 100 ms and 200 ms on average . These findings demonstrate that RSNs derived from conventional techniques such as ICA or seed–based correlation may be identified from discrete transient epochs that represent only a small fraction of the total recording , confirming recent studies from both fMRI and MEG ( Liu and Duyn , 2013; de Pasquale et al . , 2012 ) . Due to the assumption of mutual exclusivity of the HMM states , we should be cautious in automatically interpreting the fast switching between states in the form of short life times as a property of the underlying spontaneous activity . After all , an attractive hypothesis is that the brain’s resting state activity consists of a finite number of unique networks that can overlap spatially and temporally ( Smith et al . , 2012 ) . However , these relatively weak assumptions render the hypothetical networks unidentifiable using existing decomposition techniques ( to the best of our knowledge ) , and so further constraints are currently needed to proceed . For example , in temporal ICA the assumption of temporal independence will discourage temporal overlap . In the HMM , the mutual exclusion approach will also prohibit temporal overlap . As a result , the HMM states can perhaps be best thought of as representing the most dominant unique configurations of these hypothetical networks . Based on this interpretation , the state time courses provide a meaningful window on the underlying network dynamics by indicating the most dominant state at each point in time . Indeed , we have shown that these time courses can provide an insight into the relationship between different functional networks ( Figure 5 ) . Nonetheless , the rate of occurrence of the inferred states as described by the fractional occupancy allows us to assess the time scales at which within-network amplitude fluctuations are detectable . By modeling these fluctuations using the fractional occupancy time course computed at a range of different window widths , we show that they are best described by sub-second temporal dynamics with coordinated fluctuations on the order of 200–400 ms . Importantly , these time scales were not found for control surrogate data , where the fast network interactions were replaced by uncorrelated noise ( see Figure 4—figure supplement 1 ) , demonstrating that within-network functional connectivity is underpinned by neuronal dynamics that fluctuate much more rapidly than has previously been shown . The HMM assumes that the states themselves are mutually exclusive . It might be thought that this assumption negates the possibility of the HMM providing insight into cross-network ( i . e . , across-state ) interactions . While this is true at the fast ( below 100 ms ) within-state time scales , this does not mean that we cannot assess cross-network interactions at all . In particular , the HMM furnishes us with the probability of transitioning between different networks . This can tell us if there is a preference for the brain to move between two different networks , or an antagonism such that there is rarely a transition between two different networks . Indeed , the relationship between the DMN ( state 1 ) and parietal areas of the DAN ( state 7 ) was shown to be an example of the latter ( Figure 5B ) . We have also demonstrated how cross-network interactions can be assessed using fractional occupancy time courses , which describe how frequently states are visited within 10-s time windows . By computing correlations between these slower time courses , we can assess relationships between networks at the time scales more typically associated with previous investigations into long-range resting state interactions ( Fox et al . , 2005 ) . Of particular interest is the anticorrelation between the fractional occupancy time courses of the DMN state and the putative DAN state ( Figure 5A ) . The fact that cross-network relationships present in the fractional occupancy time courses are maintained when looking at the faster time scales of the state transitions , suggests a link between these two time scales . However , with the current methodology it is not possible to establish a causal link between these time scales ( i . e . , whether infrequent transitions arise because these networks are anticorrelated at longer time scales , or that this anticorrelation arises due to these infrequent transitions ) . The states inferred by the HMM show some consistency with RSNs measured using fMRI . The spatial correspondence between the spatial maps obtained via the HMM and RSNs derived from application of ICA to BOLD are shown in Figure 2—figure supplement 2 . A notable difference between these maps is the absence of the posterior cingulate cortex ( PCC ) or precuneus in the DMN state . One possible explanation is that the PCC may act as a functional ‘hub’ , such that it is not strongly represented in any one state . While the HMM allows spatial overlap , we have visualized the states in terms of state-specific amplitude changes , such that any node that is active during a number of states will be suppressed in the spatial maps . It is possible that , as a hub , the PCC has membership in the majority of states and is therefore poorly identified by mapping state specific activity . However , it is not trivial to identify nodes that are active in multiple states because this is confounded by the spatial leakage due to the ambiguities in the source reconstruction ( Van Veen et al . , 1997; Schoffelen and Gross , 2009 ) . While there is strong evidence for hub-like behavior of the PCC/precuneus in terms of structural ( Sporns et al . , 2007; Hagmann et al . , 2008 ) and functional ( Buckner et al . , 2009; Tomasi and Volkow , 2011 ) connectivity , evidence in MEG is more varied . Graph theoretical analyses of MEG band-limited power have revealed strong hubs in dorsal prefrontal cortex , lateral parietal cortex , and temporal cortex ( in essence those areas represented by state 1 ) , but notably not the PCC ( Hipp et al . , 2012 ) . Furthermore , the PCC was not found to be present in RSNs derived from application of temporal ICA to MEG data ( Brookes et al . , 2011 ) . Interestingly , an MEG-derived DMN comprising the PCC has been shown using seed–based correlation with the PCC as a seed , but only when restricted to time points in which the network was maximally correlated ( de Pasquale et al . , 2012 ) . Another reason why the PCC may not be present in the HMM is due to the relative insensitivity of MEG to deeper sources . This hypothesis is supported by the fact that in the present study , and in Brookes et al . ( 2011 ) and Hipp et al . ( 2012 ) , the sensor array comprised only axial gradiometers , which measure the spatial gradient of the field and are relatively insensitive to deep sources . Conversely , in de Pasquale et al . ( 2012 ) , the array comprised magnetometers that have increased depth sensitivity . It is therefore possible that the PCC is not present simply because it does not generate a measurable signal . Finally , it is worth considering to what extent the DMN or DAN , and the particular spatial nodes they incorporate , are definitive networks . The particular form of such networks is tied to the approaches used to identify them , for example spatial ICA . Indeed when other approaches with different assumptions are made , then different networks can be inferred . One good example of this is that , when using temporal ICA on fMRI data , no single DMN is found ( Smith et al . , 2012 ) . In short , the concept of a ‘network’ and the form they take depends on the assumptions made in the data decomposition approach , and none should be considered necessarily better than the other ( assuming they show similar objective performance , e . g . , Bayesian model evidence and reproducibility ) , but instead they each offer different perspectives on the nature of the brain’s activity . The short life times inferred from the HMM raises the question of whether there exists a relationship between the states shown here and the ‘microstates’ found from EEG studies ( Lehmann et al . , 1998; Koenig et al . , 2005 ) . Microstates are quasi-stable topographies in which the distribution of EEG power over the scalp remains stable for periods of around 100 ms . Clustering of these topographies into a limited number of classes has revealed that relatively few ( typically four ) unique maps are consistently identified across multiple time points and subjects ( Lehmann et al . , 1998 ) . It has also been suggested that EEG microstates may represent the electrophysiological signatures of resting state networks . A number of recent studies have sought to investigate this by correlating fMRI RSNs with simultaneously acquired EEG recordings ( Britz et al . , 2010; Musso et al . , 2010; Yuan et al . , 2012 ) . Segmentation of EEG scalp maps into microstates is based on finding repeating distributions of power across multiple recording sites and therefore captures similar interactions to those that drive the HMM . However , a clear distinction is that the HMM explicitly models the temporal dynamics and is therefore tuned to finding states that repeat in a predictable way . Another difference is that our approach exploits the superior spatial resolution available using MEG , by basing the inferred states on source space projections of MEG data . This makes results more directly interpretable in the context of fMRI resting state networks . While it is possible that the HMM states inferred in this paper relate to the source space counterparts of EEG microstates , without measuring EEG and MEG concurrently , a direct relationship cannot be confirmed . The states inferred using the HMM represent networks of activity that show some consistency with those found previously using spatial ICA on FMRI data ( Beckmann et al . , 2005; Smith et al . , 2009 ) and using temporal ICA on MEG data ( Brookes et al . , 2011; Luckhoo et al . , 2012 ) . In both the ICA and HMM approaches , data from all voxels are used and therefore no prior spatial localization assumptions ( e . g . , seed voxels ) are required . In the case of temporal ICA , two regions will tend to be strongly represented in the same network if their time courses exhibit a strong time-averaged correlation with the same independent component . By contrast , in the case of the HMM , membership of two regions to a particular state ( ‘network’ ) depends only on there being a repeated pattern of covariance at those points in time at which the state is active , that is , it is not time averaged over ‘all’ time points . The results in this paper , for example the fact that the two approaches result in similar networks , are consistent with the idea that ICA RSNs derive from the transient time-varying behavior captured by the HMM states . This idea is supported by the findings that seed-based functional connectivity is increased if evaluated only within those time points at which the rest of the network is synchronized ( de Pasquale et al . , 2010 ) . In this paper , we provide evidence of a relationship between the sliding window correlation computed between nodes of the DMN and the occurrence rate of the DMN state ( Figure 6 ) . Accordingly , changes in within-network envelope correlation may reflect variations in the frequency at which a particular connectivity state is visited . One explanation for this observation is the idea that electrophysiological data are characterized by scale-free dynamics that span from hundreds of milliseconds to tens of seconds ( Van de Ville et al . , 2010 ) . This fractal property of neural dynamics may provide an explanation for the similar spatial correlation structure that exists between signals at different temporal scales . Hence , slow fluctuations in band-limited amplitude correlations that underlie MEG RSNs may capture similar physiological phenomena as the HMM states , but seen through different temporal filters . In the present study , the data for individual subjects were temporally concatenated , yielding a single combined data set from which the HMM was inferred . Group concatenation is widely used in unsupervised analyses of resting state activity , particularly in the case of ICA ( Calhoun et al . , 2009 ) . Nonetheless , it is important to bear in mind that this approach assumes that there is an anatomical correspondence between subjects . In the case of ICA , this assumption means that individuals share common group maps ( Calhoun et al . , 2009 ) . In the case of the HMM approach , this assumption extends to the requirement that the data from different individuals within a particular state are drawn from the same multivariate normal distribution , and that the transitions between these states occur in a repeatable manner over subjects . This being said , there is no requirement that all states should be present in all subjects , however it is clear from the fractional occupancy time courses in Figure 3E that they are . In light of this assumption , states inferred from the group data may best be thought of as representing the spatiotemporal patterns of activity that occur most consistently over the group . An alternative strategy to group concatenation is to infer separate HMMs for each subject individually . This would allow the model to more freely adapt to individual subjects’ patterns of activity and functional connectivity . However , a severe limitation of this approach is that there will not necessarily be a correspondence between states inferred from different subjects , making it difficult to perform subsequent analyses at the group level . A limitation of the proposed technique is that HMM inference requires an a priori specification of the number of states , K . Bayesian inference techniques provide a means to test model order selection , by providing an approximation to the model evidence via the free energy . In theory , it should be possible to pick the optimal number of states by selecting the model with the greatest ( negative ) free energy . In practice however , we observe that the free energy increases monotonically up to K = 15 states , suggesting that the Bayes-optimal model may require an even higher number of states ( Figure 3—figure supplement 1A ) . In the absence of a straightforward data driven approach to model order selection , we opted instead to repeat the analysis for values of K from 4 to 15 and arbitrarily chose 8 states as the case to present here , which we believe represents a good trade-off between richness and redundancy . Results for different model orders are shown in Figure 2—figure supplement 1 . Varying the number of states between 4 and 14 did not change the topographies of the most prominent RSN-like states . It is worth noting that a similar limitation exists for more established data-driven decompositions such as ICA , in which the choice of model order is driven by the application; for example , lower model orders are used to obtain the classic RSNs , and higher model orders are used to obtain finer grained parcellations for use in subsequent network analysis ( Smith et al . , 2011 ) . Finally , the observation model used in this paper corresponded to a multivariate normal distribution . The assumption of a Gaussian observation model allows inference of the HMM to be made tractable to variational Bayesian inference and thus permits its application to large amounts of data ( 40 observations and 1 . 5 hr of time points ) . However , it should be recognized that modeling only the first and second order statistics under a Gaussian assumption is likely to be an oversimplification of the underlying network dynamics . The multivariate normal distribution is just one of many potential observation models that may be used in the context of the HMM . For example , binary observation models have previously been used in the context of modeling interictal spikes ( Ossadtchi et al . , 2005 ) . Future work will focus on implementing a multivariate autoregressive ( MVAR ) model that can model time lagged dependencies between observations . The work presented here rests on the underlying assumption that resting state activity may be broken down into a set of distinct connectivity patterns that repeat over time and where only one functional state may be active at any one time . In other words , the states inferred by the HMM are mutually exclusive . While this assumption may be an oversimplification of the underlying network dynamics , the concept that there exist distinct functional connectivity states that recur at different points in time is compatible with computational models of neuronal connectivity ( Deco et al . , 2011 ) and observations from both fMRI ( Allen et al . , 2012 ) and EEG ( Britz et al . , 2010; Musso et al . , 2010; Yuan et al . , 2012 ) . The idea that RSNs represent states in which distributed cortical areas synchronize transiently is also compatible with the idea of a ‘dynamic repertoire’ of states that are continuously explored in order to more quickly adopt the network configuration optimal for a given impending input ( Deco et al . , 2011 ) . This organization of dynamic activity through transient spatial patterns of coordination may provide the flexibility required to adapt to the rapidly changing computational demands of cognitive processing ( Bressler and Tognoli , 2006 ) .
Resting state MEG data were acquired from nine healthy subjects . The subjects were asked to lie in the scanner with their eyes open while 10 min of data were recorded . The MEG data were acquired using a 275 channel CTF whole-head system ( MISL , Conquitlam , Canada ) at a sampling rate of 600 Hz with a 150 Hz low pass anti-aliasing filter . Synthetic third order gradiometer correction was applied to reduce external interference . Localization of the head within the MEG helmet was achieved using three electromagnetic head position indicator ( HPI ) coils . By periodically energizing these coils their position within the MEG sensor array was identified . Prior to data acquisition , the HPI coil locations , the position of three fiducial points ( the nasion , and left and right preauricular points ) , and the head shape were recorded using a three-dimensional digitizer ( Polhemus Isotrack ) . MR images were acquired using a 3T Phillips Achieva MR system at 1 × 1 × 1 mm3 resolution running an MPRAGE sequence . Each subject’s structural MRI was registered to the MNI152 standard brain such that all subsequent source space analysis was performed in MNI space . The locations of the MEG sensors with respect to the anatomy were determined by registering the digitized head surface to the head surface extracted from the structural MRI . The data were converted to SPM8 http://www . fil . ion . ucl . ac . uk/spm and down sampled to 200 Hz . Each recording was visually inspected to identify channels and/or periods of data containing obvious artifacts or with abnormally high variance , which were discarded . Independent component analysis ( ICA ) was used to decompose the sensor data for each session into 150 temporally independent components ( tICs ) and associated sensor topographies ( http://research . ics . aalto . fi/ica/fastica ) . Artifact components were classified via the following procedure . Eye-blink , cardiac and mains interference components were manually identified by the combined inspection of the spatial topography , time course , kurtosis of the time course and frequency spectrum for all components . Eye-blink artifacts typically exhibited high kurtosis ( >20 ) , a repeated blink structure in the time course and very structured spatial topographies . Cardiac component time courses strongly resembled the typical ECG signals , as well as having high kurtosis ( >20 ) . Mains interference had extremely low kurtosis ( typically <−1 ) and a frequency spectrum dominated by 50 Hz line noise . Following artifact rejection the data were frequency filtered into the 4–30 Hz band . The pre-processed MEG data were projected onto a regular 8-mm grid spanning the entire brain using a custom scalar beamformer implemented in SPM8 ( Van Veen et al . , 1997; Vrba and Robinson , 2001; Woolrich et al . , 2011 ) . Beamforming is an adaptive spatial filter in which the estimated neuronal electrical activity qr , φ ( t ) at a pre-determined brain space location r with orientation φ , and at time t is given by a weighted sum of the N sensor measurementsqr , φ ( t ) =wr , φTm ( t ) where , wr , φ is a ( N × 1 ) set of weights that govern the projection of the ( N × 1 ) sensor data m ( t ) into source space , and superscript T indicates a transpose . Here , we use a linearly constrained minimum variance ( LCMV ) scalar beamformer ( Van Veen et al . , 1997; Robinson and Vrba , 1999; Woolrich et al . , 2011 ) . The weights are determined by an estimate of the ( N × N ) covariance matrix C computed between all sensor pairs , and a set of lead fields hr , φ that describes how a unit current generated by a dipolar source at location r with orientation φ would be measured at each MEG sensor ( Sarvas , 1987; Huang et al . , 1999 ) wr , φT=[hr , φTC−1hr , φ]−1hr , φTC−1where , hr , φ is the ( N × 1 ) projection of the ( N × 3 ) lead fields Hr , Φ to the dipole orientation φ that maximizes the projected signal-to-noise ratio , computed as in Sekihara et al . ( 2001 ) . The sensitivity of the beamformer varies for different locations in the head ( Van Veen et al . , 1997; Vrba and Robinson , 2001 ) . To account for this spatial bias , the projected data qr , φ ( t ) were scaled by an estimate of the projected noise:zr , φ ( t ) = qr , φ ( t ) wr , φTwr , φwhere , wr , φT × wr , φ represents the projection of uncorrelated sensor noise ( i . e . , the noise covariance matrix is the identity matrix ) . Following beamformer projection , the oscillatory amplitude envelope at each voxel was derived by computing the magnitude of the Hilbert transform of the source-reconstructed data . For computational efficiency , the envelopes were down sampled to 40 Hz by temporally averaging within sliding windows with a width of 100 ms and 75% overlap between consecutive windows . The amplitude envelopes were concatenated temporally across all subjects after spatially smoothing with a Gaussian kernel ( FWHM 9 . 4 mm ) . The envelope data for each subject were demeaned and normalized by the global ( over all voxels ) variance prior to concatenation . The group-concatenated envelopes were demeaned and pre-whitened to reduce the data to 40 principal ( temporal ) components with unit variance and zero mean . An HMM with 8 states was inferred ( although see Figure 2—figure supplement 1 and Figure 3—figure supplement 1 for equivalent results with other model orders ) . Each inferred HMM state is associated with a unique multivariate normal distribution over the observations ( principal components ) , defined by a ( M × 1 ) mean vector and a ( M × M ) covariance matrix where M = 40 ( the number of principal components ) ( Rezek and Roberts , 2005; Woolrich et al . , 2013 ) . To account for variations in the inference due to different initializations , 10 realizations were performed for each inference and the model with the lowest free energy was chosen . The most probable a posteriori state ut at each time point and was obtained using the Viterbi algorithm ( Rezek and Roberts , 2005 ) . State time courses were defined for each state as indicator variables that indicate the points in time in which that state is most probable ut==k . The HMM toolbox and example scripts may be downloaded from www . fmrib . ox . ac . uk/∼woolrich/HMMtoolbox . We assume an HMM of length T samples , state space dimension K , hidden state variables s={s1 . . . sT} and observed data y={y1 . . . yT} , where yt are the ( M × 1 ) principal components at time t computed from the group-concatenated envelope data . The full true posterior probability of the model is then given by:P ( y , s , Θ ) =P ( s0|π0 ) ∏tTP ( st|st−1 , πt ) P ( yt|st , θ ) P ( πt ) P ( π0 ) P ( θ ) where , P ( πt ) , P ( π0 ) and P ( θ ) are chosen to be non-informative priors . The HMM parameters Θ={π0 , πt , θ} consist of π0 which parameterize the initial state probability P ( s0 ) , πt which determine the state transition probability P ( st|st-1 ) and θ which describe the observation probabilities P ( yt|st ) . Here , we assume that the probability to transition to another state depends only on the state that the system is in , and not on the path it took to get to its current state , that is , it is Markovian:P ( st|s1…st−1 ) =P ( st | st−1 ) = πtwhere , πt is the ( K × K ) transition probability matrix in which the element ( i , j ) describes the probability of transitioning from state i to state j between time t-1 and time t . The term P ( yt|st , θ ) , is the observation model . In this work , we assume that the observation model for state k is a multivariate normal distribution with θk={μk , Σk} , where μk is the ( M × 1 ) mean vector , and Σk is the ( M × M ) covariance matrix:P ( yt|st=k , θ ) ∼ N ( μk , Σk ) The prior distributions over the HMM parameters Θ={π0 , πt , θ} are chosen to be conjugate distributions . The approximate posterior distributions will then be functionally identical to the prior distributions ( i . e . , a Gaussian prior density is mapped to a Gaussian posterior density ) , making the model tractable to certain kinds of inference . See Rezek and Roberts ( 2005 ) for details . In this work , we use variational Bayes ( VB ) inference on the HMM , as described in Rezek and Roberts ( 2005 ) . This is fully probabilistic and furnishes us with full posterior distributions on the model parameters P ( Θ , s , y ) . Aside from the inferred posterior distribution over the observation model P ( yt|st , θ ) , we are also interested in determining those points in time at which particular states are active . The relevant output is the marginal posterior inference on the state variables P ( st|y ) . This is obtained using Viterbi decoding . For the purpose of computing summary statistics of state life time and occupancy , we have chosen to hard classify the states as being on or off by choosing the most probable a posteriori state ut at each time point:ut= arg maxkP ( st=k|y ) We defined a number of summary statistics to describe the temporal characteristics of the inferred states . For the purpose of computing these statistics , we have chosen to hard classify the states as being on or off by choosing the most probable a posteriori state ut at each time point: Fractional occupancy is defined as the fraction of time spent in each statefractional occupancy ( k ) = 1T∑t ( ut ==k ) where , ut==k is one if ut=k and is zero otherwise , and T is the length of the state sequence in samples . The mean life time is defined as the average amount of time spent in each state before transitioning out of that state:mean life time ( k ) = ∑t ( ut==k ) number of occurences ( k ) The mean interval length is similarly defined as the average amount of time spent between consecutive visits to a particular state:mean interval length ( k ) = T−∑t ( ut==k ) number of occurences ( k ) where , the number of occurrences is given by:number of occurrences ( k ) = ∑t ( ( ( ut==k ) − ( ut−1==k ) ) ==1 ) By performing a principal component analysis prior to inferring the HMM , the dimensionality of the data was reduced to a computationally manageable amount . However , this means that the multivariate normal distributions that define each state span this reduced subspace , and are therefore not readily interpretable in terms of the underlying anatomy . To interrogate those brain areas associated with each state , we mapped state-specific activity by correlating the amplitude envelope at each voxel with the HMM state time courses . This has the advantage of identifying only activity that is unique to each state , and thus reduces components of the signal that are common across states that may obscure state-specific effects . State specific changes in oscillatory activity were identified by computing the partial correlation of the state time courses with the source space data . The partial correlation was computed using a general linear modeling ( GLM ) framework as follows . The maximum a posteriori state ut at each time point was used to construct a ( K × T ) design matrix X , where each column k is an array of indicator variables that indicates whether state k is on or off:X ( t , k ) ={1 , ut=k0 , ut≠k This design matrix , together with the full-rank ( before whitening ) source space data , was used in a GLM analysis ( Friston et al . , 1996; Brookes et al . , 2004; Woolrich et al . , 2009 ) . Specifically , we perform a multiple linear regression at each voxel with the group-concatenated envelope data as the dependent variable . To compute the partial correlation , both the design matrix and the data were normalized to have zero mean and unit variance prior to fitting the GLM . This yields a set of K spatial maps representing estimates of the partial correlation coefficient between each state and the data . | When subjects lie motionless inside scanners without any particular task to perform , their brains show stereotyped patterns of activity across regions known as resting state networks . Each network consists of areas with a common function , such as the ‘motor’ network or the ‘visual’ network . The role of resting state networks is unclear , but these spontaneous activity patterns are altered in disorders including autism , schizophrenia , and Alzheimer’s disease . One puzzling feature of resting state networks is that they seem to last for relatively long times . However , the majority of studies into resting state networks have used fMRI brain scans , in which changes in the level of oxygen in the blood are used as a proxy for the activity of a given brain region . Since changes in blood oxygen occur relatively slowly , the ability of fMRI to detect rapid changes in activity is limited: it is thus possible that the long-lived nature of resting state networks is an artefact of the use of fMRI . Now , Baker et al . have used a different type of brain scan known as an MEG scan to show that the activity of resting state networks is shorter lived than previously thought . MEG scanners measure changes in the magnetic fields generated by electrical currents in the brain , which means that they can detect alterations in brain activity much more rapidly than fMRI . MEG recordings from the brains of nine healthy subjects revealed that individual resting state networks were typically stable for only 100 ms to 200 ms . Moreover , transitions between different networks did not occur randomly; instead , certain networks were much more likely to become active after others . The work of Baker et al . suggests that the resting brain is constantly changing between different patterns of activity , which enables it to respond quickly to any given situation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Fast transient networks in spontaneous human brain activity |
Essential to spatial orientation in the natural environment is a dynamic representation of direction and distance to objects . Despite the importance of 3D spatial localization to parse objects in the environment and to guide movement , most neurophysiological investigations of sensory mapping have been limited to studies of restrained subjects , tested with 2D , artificial stimuli . Here , we show for the first time that sensory neurons in the midbrain superior colliculus ( SC ) of the free-flying echolocating bat encode 3D egocentric space , and that the bat’s inspection of objects in the physical environment sharpens tuning of single neurons , and shifts peak responses to represent closer distances . These findings emerged from wireless neural recordings in free-flying bats , in combination with an echo model that computes the animal’s instantaneous stimulus space . Our research reveals dynamic 3D space coding in a freely moving mammal engaged in a real-world navigation task .
As humans and other animals move in a 3D world , they rely on dynamic sensory information to guide their actions , seek food , track targets and steer around obstacles . Such natural behaviors invoke feedback between sensory space representation , attention and action-selection ( Lewicki et al . , 2014 ) . Current knowledge of the brain’s representation of sensory space comes largely from research on neural activity in restrained animals , generally studied with 2D stimuli ( Van Horn et al . , 2013 ) ; however , far less is known about 3D sensory representation , particularly in freely moving animals that must process changing stimulus information to localize objects and guide motor decisions as they navigate the physical world . Animals that rely on active sensing provide a powerful system to investigate the neural underpinnings of sensory-guided behaviors , as they produce the very signals that inform motor actions . Echolocating bats , for example , transmit sonar signals and process auditory information carried by returning echoes to guide behavioral decisions for spatial orientation ( Griffin , 1958 ) . Work over the past decade has revealed that echolocating bats produce clusters of sonar calls , termed sonar sound groups ( SSGs ) , to closely inspect objects in their surroundings or to negotiate complex environments ( Kothari et al . , 2014; Moss et al . , 2006; Petrites et al . , 2009; Sändig et al . , 2014 ) . We hypothesize that the bat’s sonar inspection behavior sharpens spatio-temporal echo information processed by the auditory system in a manner analogous to the active control of eye movements to increase visual resolution through sequences of foveal fixations ( Hayhoe and Ballard , 2005; Moss and Surlykke , 2010; Tatler et al . , 2011 ) . Importantly , the bat’s acoustic behaviors provide a quantitative metric of spatial gaze , and can thus be analyzed together with neural recordings to investigate the dynamic representation of sensory space . Echolocating bats compute the direction of echo sources using a standard mammalian auditory system ( Wohlgemuth et al . , 2016 ) . The dimension of target distance is computed from the time delay between sonar emissions and echoes ( Simmons , 1973 ) . Neurophysiological investigations of echo processing in bats reveal that a class of neurons shows facilitated and delay-tuned responses to simulated pulse-echo pairs . It has been hypothesized that echo delay-tuned neurons carry information about the distance to objects ( Feng et al . , 1978; O’'Neill and Suga , 1982; Suga and O'Neill , 1979; Valentine and Moss , 1997 ) ; however , the neural representation of target distance in bats listening to self-generated echoes reflected from physical objects has never previously been empirically established . The midbrain superior colliculus ( SC ) has been implicated in sensory-guided spatial orienting behaviors , such as visual and auditory gaze control in primates , cats and barn owls ( Knudsen , 1982; Krauzlis , 2004; du Lac and Knudsen , 1990; Middlebrooks and Knudsen , 1984; Munoz et al . , 1991; Sparks , 1986; Stein et al . , 1989 ) , prey-capture behavior in frog and pit viper ( Grobstein , 1988; Hartline et al . , 1978; Newman and Hartline , 1981 ) , and echolocation in bats ( Valentine and Moss , 1997; Valentine et al . , 2002 ) . Previous work has also demonstrated that the SC is an integral part of the egocentric spatial attention network , specifically for target selection and goal-directed action ( Krauzlis et al . , 2013; Lovejoy and Krauzlis , 2010; McPeek and Keller , 2004; Mysore and Knudsen , 2011; Zénon and Krauzlis , 2012 ) . Work in freely behaving rodents has also demonstrated a more general role of the SC in sensory-guided orienting behaviors ( Duan et al . , 2015; Felsen and Mainen , 2008 ) . Additionally , measures of the local field potential ( LFP ) in the midbrain optic tectum ( avian homologue of the SC ) have shown that increases in the gamma band ( ~40–140 Hz ) correlate with attention to sensory stimuli ( Sridharan and Knudsen , 2015 ) . The research reported here is the first to investigate the behavioral modulation of depth-tuned single unit responses and gamma band oscillations in the SC of a mammal inspecting objects in its physical environment . Prior work on sensorimotor representation in the mammalian SC has been largely carried out in restrained animals performing 2D tasks , leaving gaps in our knowledge about the influence of action and attention on sensory responses in animals moving freely in a 3D physical environment . To bridge this gap , we conducted wireless chronic neural recordings of both single unit activity and LFPs in the SC of free-flying bats that used echolocation to localize and inspect obstacles along their flight path . Central to this research , we developed a novel echo model to reconstruct the bat’s instantaneous egocentric stimulus space , which we then used to analyze echo-evoked neural activity patterns . Our data provide the first demonstration that neurons in the midbrain SC of a freely moving animal represent the 3D egocentric location of physical objects in the environment , and that active sonar inspection sharpens and shifts the depth tuning of 3D neurons .
To measure auditory spatial receptive fields in the bat SC , we first determined the azimuth , elevation and distance of objects , referenced to the bat’s head direction and location in the environment ( Figure 2—figure supplement 1A shows a cartoon of a bat with a telemetry recording device and markers to estimate the bat’s head direction , Figure 2—figure supplement 1B shows a top view of the bat’s head with the telemetry device and head tracking markers , also see Materials and methods ) . In order to determine the 3D direction and arrival time of sonar echoes returning to the bat , we relied on the physics of sound to establish an echo model of the bat’s instantaneous sensory space . The echo model takes into account an estimate of the beam width of the bat’s sonar calls , its 3D flight trajectory , its head direction , as well as physical parameters of sound ( Figure 2—figure supplement 1A and B – schematic , see Materials and methods ) to compute a precise estimate of the time of arrival of echoes at the bat’s ears , as well as the 3D location of the echo sources ( Figure 2A – cartoon explains the echo model , with cones showing the sonar beam pattern , Figure 2B – the time series of call and echoes from the cartoon in Figure 2A; Figure 2C – actual bat flight trajectory with sonar vocalizations , orange circles , and 3D head aim vectors , black lines; Figure 2D and E – the instantaneous solid angles of the head aim with respect to objects and echo arrival times of sonar returns from different objects along the trajectory in 2C; also see Materials and methods ) . The echo model was used to construct the instantaneous acoustic sensory space of the bat each time it vocalized and received echoes from physical objects in its flight path . We first determined the onset of each vocalization produced by the bat , then the 3D position of the bat at the time of each sonar vocalization , and the 3D relative positions of flight obstacles . Past work has demonstrated that the big brown bat’s sonar beam axis is aligned with its head ( Ghose and Moss , 2003; 2006 ) , and the direction of the sonar beam was inferred in our study from the head-mounted markers showing the head aim of the bat . We then referenced the 50 deg −6 dB width of the sonar beam at 30 kHz ( Hartley and Suthers , 1989 ) , and the time at which the sonar beam reflected echoes from flight obstacles in the animal’s path . From this calculation , we computed the direction and time of arrival of all echoes returning to the bat’s ears each time the animal emitted a sonar call . Although it is possible to use a wireless , head-mounted microphone to record the returning echo stream , there are significant limitations to this methodology . First , a single head-mounted microphone has a higher noise floor than the bat’s auditory receiver and therefore does not pick up all returning echoes that the bat may hear . Moreover , a single microphone would add weight to the devices carried by the bat in flight and could only provide information regarding echo arrival time , not sound source direction . A head-mounted microphone is therefore insufficient to compute the 3D locations of echo sources , thus highlighting the importance of the echo model in our study to compute the bat’s instantaneous 3D sensory space . We computed errors in the measurements of head-aim as well as in the estimation of echo arrival times at the bat’s ears ( Figure 2—figure supplement 2 ) . Our measurements indicate that the maximum error in the reconstruction of the bat head-aim does not exceed 5 . 5 degrees , and the error in echo arrival time measurement is between 0 . 35 and 0 . 65 ms ( see Figure 2—figure supplement 2C and D – estimation of errors in head-aim reconstruction , Figure 2—figure supplement 2 – errors in echo arrival time; see Materials and methods ) . To confirm that the echo model accurately calculated the 3D positions of sonar objects , we used echo playbacks from a speaker and microphone pair ( see Materials and methods , Figure 2—figure supplement 2 ) , with additional validation by using a microphone array placed behind the bat’s flight direction . The microphone array recorded the echoes reflected off objects as the bat flew and produced sonar vocalizations , which were analyzed with time of arrival difference ( TOAD ) algorithms to compare the measured echo sources with the calculated echo sources based on our echo model ( see Materials and methods ) . The establishment of the echo model was a critical step in computing 3D spatial tuning of SC neurons recorded from the animals in flight . The spatial acoustic information ( echo arrival times and 3D locations of echo sources ) obtained from the echo model was converted into 3D egocentric coordinates to compute the acoustic stimulus space from the point of view of the flying bat as it navigated the room ( Figure 1—figure supplement 1F and G , see Materials and methods ) . Bats were released from different locations in order to cover the calibrated volume of the flight room ( Figure 3—figure supplement 1A ) , and they continuously produced echolocation calls , which resulted in series of echoes from objects during each recording session ( Figure 3—figure supplement 1A , B and C ) . We also released the bats from multiple locations in the room so that they took a variety of flight paths through the room , and interacted with the flight obstacles from a broad range of directions and distances , which is necessary for computing spatial receptive fields . These data therefore yielded measurements of echoes returning to the animal from objects at many different directions and distances in egocentric space ( Figure 3—figure supplement 1D - range coverage , E - azimuth coverage , and F - elevation coverage ) . The output of the echo model was used to analyze audio/video-synchronized neural recordings from single units ( see Figure 1E , Figure 1—figure supplement 1 and Materials and methods ) taken in the midbrain SC using a 16-channel wireless telemetry system . We recorded a total of 182 single neurons . We then classified neurons as sensory ( n = 67 ) , sensorimotor ( 45 ) , vocal premotor ( n = 26 ) , or unclassified ( n = 44 ) , as described in the Materials and methods section . Here we focus on sensory neurons in the SC of free-flying bats . For all sensory neurons we first calculated the distance , or echo-delay tuning ( Figure 3A and B ) . An example reconstruction of a neuron’s spatial tuning along the distance axis is displayed in Figure 3B , showing neural activity aligned to sonar vocalization times ( red arrows ) , and responses to echoes returning at ~10 ms delay . Arrival time of the first echo at the bat’s ears is indicated with a green arrow , and a second returning echo ( from another , more distant object ) is indicated with a blue arrow . Note that this example neuron does not spike in response to the second echo , nor to echoes arriving very early ( Figure 3C , top panel ) , or late ( Figure 3C , bottom panel ) . Figure 3D shows the computed distance ( echo-delay ) tuning profile of this same example neuron . Using the echo model , we also calculated the tuning profiles of each neuron in azimuth and elevation ( Figure 3—figure supplement 2A – azimuth and B – elevation ) . Once we calculated the azimuth , elevation , and distance tuning of neurons ( Figure 4A ) , we constructed three-dimensional spatial response profiles for each neuron . Figure 4B shows surface plots of the three-dimensional tuning for two other example neurons . Of the 67 single sensory neurons ( Bat 1–28 in green , and Bat 2–39 in brown ) recorded in the SC of two big brown bats , 46 neurons ( Bat 1–19 and Bat 2–27 ) in the data set showed selectivity to stimulus locations in 3D egocentric space ( Figure 4C , see Materials and methods for details about spatial selectivity analysis ) , and these spatial tuning profiles were stable within recording sessions ( Figure 4—figure supplement 1 ) . Additionally , the selectivity of the neurons , in the distance dimension , did not vary as a function of dorsal-ventral location in the SC ( Figure 4—figure supplement 2 ) . Further , three neurons were tuned to both azimuth and range , two were tuned to both range and elevation , and five , three and three neurons were tuned exclusively to range , azimuth and elevation , respectively ( see Figure 4—figure supplement 3 ) . Best echo delays spanned values of 4 to 12 ms , corresponding to the distances of objects encountered by the bat ( ~70–200 cm ) in our flight room ( Figure 4D , E and F show histograms of standard deviations of normal fits to spatial receptive fields , also see Materials and methods ) . Guided by growing evidence that an animal’s adaptive behaviors and/or attentional state can modulate sensory responses of neurons in the central nervous system ( Bezdudnaya and Castro-Alamancos , 2014; Fanselow and Nicolelis , 1999; McAdams and Maunsell , 1999; Reynolds and Chelazzi , 2004; Spitzer et al . , 1988; Winkowski and Knudsen , 2006; Womelsdorf et al . , 2006 ) , we investigated whether the bat’s active sonar inspection of objects in space alters the 3D sensory tuning of SC neurons . We compared the spatial receptive fields of single SC neurons when the bat produced isolated sonar vocalizations ( non-SSGs ) to times when it adaptively increased sonar resolution by producing SSGs ( Figure 5A – an example trial; non-SSGs , blue circles; SSGs , red circles; Figure 5B – spectrograms from the data in 6A , with SSGs again highlighted in red; Figure 5C – a plot showing SSGs can be quantitatively identified , see Materials and methods ) . We discovered that a neuron’s distance tuning is sharper to echo returns from the bat’s production of SSGs , as compared to responses to echoes returning from single ( non-SSG ) calls ( Figure 5D shows an example neuron ) . Figure 5E shows summary data comparing the sharpness of distance tuning to echoes returning from SSG and non-SSG calls ( n = 51 , neurons which met the power analysis criterion , see Materials and methods for details about power analysis; data from Bat 1 is shown in green , Bat 2 in brown ) . Supplementary file 1A – gives details of sharpness of distance tuning comparisons for SSG and non-SSG tuning , using the Brown-Forsyth test , for each of the neurons in Figure 5E . We also found that a neuron’s best echo delay ( target distance ) is often shifted to shorter delays ( closer objects ) when the bat is engaged in the production of SSGs , suggesting that distance tuning is dynamically remapped when the bat actively inspects objects in its environment ( Figure 5D example ) . Figure 5F shows summary data , comparing the mapping of distance tuning of single neurons in response to echoes from SSG and non-SSG calls ( n = 53 neurons which met the power analysis criterion , see Materials and methods for details about power analysis; data from Bat 1 is shown in green , Bat 2 in brown ) . Supplementary file 1B – gives details of mean distance tuning comparisons for SSG and non-SSG echo delay responses , using the Brown-Forsyth test . For each of the neurons in Figure 5E and F; filled circles indicate cells with a significant sharpening ( Figure 5E ) , or a significant decrease in peak distance tuning in response to echoes from SSGs ( Figure 5F ) ; while open circles indicate non-significant comparisons ( rank-sum , p<0 . 05 ) . We also examined the responses to echoes returning from the first sonar vocalization of an SSG versus the last vocalizations of an SSG . We found that there is no difference in spatial tuning profiles computed separately for the first and last echoes of SSGs , but there is a significant increase in spike probability in response to echoes from the last vocalization of an SSG ( Figure 5—figure supplement 1 ) . Similar to foveation , which is a behavioral indicator of visual attention to resolve spatial details ( Reynolds and Chelazzi , 2004 ) , measurements of adaptive sonar behavior have been used as a metric for the bat’s acoustic gaze to closely inspect objects ( Moss and Surlykke , 2010 ) . Previous behavioral research shows that bats increase the production of sonar sound groups ( SSGs ) under conditions that demand high spatial resolution , e . g . in dense acoustic clutter and when tracking erratically moving targets ( Kothari et al . , 2014; Moss et al . , 2006; Petrites et al . , 2009; Sändig et al . , 2014 ) . SSGs are clusters of echolocation calls , often produced at stable rate ( Figure 6A , see Materials and methods ) , which are hypothesized to sharpen acoustic images of objects in the environment ( Moss and Surlykke , 2010 ) , and are distinct from the overall increase in sonar call rate of a bat approaching a target . Previous work in other systems has shown that the gamma frequency band ( 40–140 Hz - Sridharan and Knudsen , 2015 ) of the LFP in the SC increases in power when an animal is attending in space ( Gregoriou et al . , 2009; Gunduz et al . , 2011; Sridharan and Knudsen , 2015 ) , and we investigated whether this conserved indicator of spatial attention also appears during SSG production . Shown in Figure 6B is a comparison of gamma band activity during the bat’s production of SSGs over non-SSGs , demonstrating an increase around the time of SSG production . Displayed is the call triggered average ( ±s . e . m . ) of the gamma band across recording sites , for SSG ( red , n = 539 ) and non-SSG ( blue , n = 602 ) production . Figure 6C illustrates the significant increase in gamma band power during the production of SSGs ( red ) as compared to non-SSGs ( blue ) on a neuron-by-neuron basis ( n = 26 ) , and this finding was consistent across recording depths ( Figure 6—figure supplement 1 ) . Only sites in which neural recordings were unaffected by motion artifact were included in this analysis ( Figure 6—figure supplement 2 , Also see Materials and methods ) . In agreement with past work in other systems and brain areas ( Gregoriou et al . , 2009; Gunduz et al . , 2011; Sridharan and Knudsen , 2015 ) , there was a significant increase in gamma power when the bat produced SSGs , providing further evidence that SSGs indicate times of sonar inspection and spatial attention ( Figure 6C , p<0 . 005 , Wilcoxon sign-rank test ) . Additionally , we analyzed the timing of gamma power increase with respect to echo-evoked neural activity . Because sensing through echolocation temporally separates vocal production time from echo arrival time , we can accurately measure the amplitude of gamma activity with respect to motor production and/or sound reception . The data show that the increase in gamma power occurred specifically around the time of the echo-evoked spike events in SC sensory neurons ( Figure 6D – SSGs and 6E – non-SSGs , vertical white line indicates onset of sensory evoked spikes , horizontal white line separates data from Bat 1 , below , and Bat 2 , above ) , and that the increase in gamma band power is temporally precise , with the peak in gamma power occurring within 10 milliseconds of spike time .
The superior colliculus ( SC ) , a midbrain sensorimotor structure , is implicated in species-specific sensory-guided orienting behaviors , target selection and 2D spatial attention ( Duan et al . , 2015; Knudsen , 2011; Krauzlis et al . , 2013; Lovejoy and Krauzlis , 2010; McPeek and Keller , 2004; Mysore and Knudsen , 2011; Mysore et al . , 2011; Zénon and Krauzlis , 2012 ) . Past research has led to conflicting views as to whether the SC plays a role in orienting in 3D space ( Chaturvedi and Gisbergen , 1998; Chaturvedi and van Gisbergen , 1999; Chaturvedi and Van Gisbergen , 2000; Hepp et al . , 1993; Van Horn et al . , 2013; Leigh and Zee , 1983; Walton and Mays , 2003 ) , but limited evidence from sensory mapping in primates shows response selectivity to binocular disparity ( Berman et al . , 1975; Dias et al . , 1991 ) , and vergence eye movements ( Chaturvedi and Gisbergen , 1998; Chaturvedi and van Gisbergen , 1999; Chaturvedi and Van Gisbergen , 2000; Van Horn et al . , 2013 ) , indicating a role of the SC in 3D visuomotor integration . Here , we present the first direct evidence of 3D egocentric sensory responses to physical stimuli in the midbrain of an animal freely moving through its environment . Our results therefore provide a critical bridge to understanding the brain’s dynamic representation of the 3D physical world . Psychophysical studies have reported that human and non-human primates show increased visual detection and discrimination performance when stimuli are presented at attended locations ( Bichot et al . , 2005; Carrasco , 2011; Posner , 1980; Wurtz and Mohler , 1976; Yeshurun and Carrasco , 1999 ) . Neural recording experiments have corroborated these results by showing that spatial attention modulates firing rates of cortical neurons representing attended locations ( McAdams and Maunsell , 1999; Reynolds and Chelazzi , 2004; Reynolds et al . , 1999; Spitzer et al . , 1988; Womelsdorf et al . , 2006 ) . Other studies report an increase in the gain of tuning curves at an attended location or a selected stimulus feature , while a decrease in neural response occurs for unattended locations or features ( McAdams and Maunsell , 1999; Treue and Martínez Trujillo , 1999; Verghese , 2001 ) . The midbrain SC has been specifically implicated in an attention network through past studies of SC inactivation that produced behavioral deficits ( Lovejoy and Krauzlis , 2017; McPeek and Keller , 2004 ) , but none of these studies measured the spatial selectivity of single SC neurons under conditions in which animals freely inspected objects in the physical environment . Evidence for sharpening of tuning curves and/or remapping spatial receptive fields with attention has been limited to a few studies showing shifts in 2D cortical tuning to artificial visual stimuli in restrained animals ( Spitzer et al . , 1988; Womelsdorf et al . , 2006 ) . And in studies of the auditory system , behavioral discrimination of acoustic stimuli has been shown to influence the response profiles of cortical neurons in restrained ferrets ( Fritz et al . , 2003 , 2007 ) . Here we report for the first time dynamic shifts in 3D sensory tuning with sonar-guided attention in animals engaged in natural orienting behaviors . Our study not only revealed changes in single neuron 3D spatial selectivity with dynamic sonar inspection of objects in the physical scene , but also a corresponding increase in the gamma band of the local field potential ( LFP ) . Past work in humans , non-human primates , other mammals , and birds have reported stimulus driven gamma band modulation when stimuli are presented at attended locations ( Fries et al . , 2001; Goddard et al . , 2012a; Gregoriou et al . , 2009; Sridharan and Knudsen , 2015; Sridharan et al . , 2011 ) . Moreover , changes in the gamma band of the LFP have been shown to occur for stimulus selection and discrimination mediated by touch , vision , and hearing , suggesting that gamma oscillations may reflect multi-modal network activity related to attention ( Bauer et al . , 2006; Canolty et al . , 2006; Gruber et al . , 1999; Senkowski et al . , 2005 ) . Our findings that gamma power increases during epochs of SSG production and echo reception support the hypothesis that the bat’s adaptive sonar behaviors serve as indicators of spatial attention ( Moss and Surlykke , 2010 ) . It is important to emphasize the distinction between our report here on 3D egocentric sensory responses in the midbrain SC of the insectivorous echolocating big brown bat , and 3D allocentric memory-based representation of space in the hippocampus of the echolocating Egyptian fruit bat ( Yartsev and Ulanovsky , 2013 ) . These two distinct frames of reference are used for different suites of natural behaviors . Egocentric sensory representation of space contributes to overt and covert orienting to salient stimuli ( Knudsen , 2011; Krauzlis et al . , 2013; Mysore and Knudsen , 2011 ) and has not previously been described in free-flying bats . By contrast , 3D allocentric ( Geva-Sagiv et al . , 2015; Yartsev and Ulanovsky , 2013 ) and vectorial representations ( Sarel et al . , 2017 ) in the bat hippocampus support spatial memory and navigation . Further , published studies on the Egyptian fruit bat hippocampus have not considered the acoustic sensory space of this species that uses tongue clicks to echolocate ( Yovel et al . , 2010 ) , nor potential modulation of hippocampal activity by sonar signal production . In other words , past work on the Egyptian fruit bat hippocampus shows 3D spatial memory representation; whereas , our study of the big brown bat SC reveals important new discoveries of state-dependent midbrain sensory representation of 3D object location . Finally , and importantly , our results fill a long-standing gap in the literature on the neural representation of target distance in the bat auditory system , which has almost exclusively been studied in passively listening animals ( Dear and Suga , 1995; Feng et al . , 1978; O'Neill and Suga , 1979; Valentine and Moss , 1997 ) , but see Kawasaki et al . , 1988 . Echolocating bats estimate target distance from the time delay between sonar call emission and echo reception , and show behavioral range discrimination performance of less than 1 cm , which corresponds to an echo delay difference of about 60 μsec ( Moss and Schnitzler , 1995; Simmons , 1973 ) . The bat’s sonar signal production is therefore integral to target ranging , and yet , for over nearly four decades of research , scientists have simulated the dimension of target distance in neural recording experiments in restrained bats by presenting pairs of synthetic sound stimuli ( P/E pairs – pulse/echo pairs ) , one mimicking the echolocation call , and a second , delayed and attenuated signal , mimicking the echo . Here , we report the first delay-tuned neural responses to echoes from physical objects in the auditory system of free-flying bats , thus providing a critical test of a long-standing hypothesis that neurons in actively echolocating bats respond selectively to echoes from objects in 3D space . Beetz et al . ( 2016a ) report that distance tuning of neurons in the auditory cortex of passively listening , anesthetized bats ( Carollia perspicillata ) is more precise when neurons are stimulated with natural sonar sequences , such as those produced by echolocating bats in the research reported here . Another study of auditory cortical responses in anesthetized bats ( Phyllostomus discolor ) reports that delay-tuned neurons shift their receptive fields under stimulus conditions that simulate echo flow . ( Bartenstein et al . , 2014 ) . In a related study , Beetz et al . , 2016b show a higher probability of neural firing in cortical neurons of the bat species Carollia perspicillata to the first echo in a sequence , which leads them to hypothesize that global cortical inhibition contributes to the representation of the closest object , without active attention . It is possible that global cortical inhibition is an intrinsic feature , which enables an animal to represent the most salient ( in the above case , closest ) stimulus . Our data also show that sensory neurons respond primarily to the first echo arriving in a neuron’s receptive field , as compared to later echoes , and may depend on a similar mechanism . A mechanism of global inhibition for selective attention has also been demonstrated in the barn owl optic tectum ( Mysore et al . , 2010 ) . Additionally , our data demonstrate a higher probability of auditory responses in the midbrain SC to echoes returning from the last echo of a SSG , a finding , which can only be demonstrated in a behaving echolocating bat , as it involves feedback between sensing and action . And while studies of auditory cortical processing in anesthetized , passively listening animals can shed light on sensory processing mechanisms , ultimately this information must be relayed to sensorimotor structures , such as the midbrain superior colliculus , which serve to orchestrate appropriate motor commands for spatial navigation and goal-directed orientation . Our study reveals the novel finding that auditory neurons in awake and behaving echolocating bats show shifts and sharpening of spatial receptive fields with echolocation call dynamics . Crucially , because bats in our study were engaged in a natural spatial navigation task , we could directly investigate the effects of sonar-guided attention on the 3D spatial tuning of single auditory neurons . Our results demonstrate the dynamic nature of 3D spatial selectivity of single neurons in the SC of echolocating bats and show that active behavioral inspection of objects not only remaps range response areas , but also sharpens depth tuning . Furthermore , our data reveal echo-delay tuning of single SC neurons in response to echoes from actively echolocating bats is sharper than previously reported from recordings in passively listening bats ( Dear and Suga , 1995; Menne et al . , 1989; Moss and Schnitzler , 1989; Simmons et al . , 1979; Simmons et al . , 1990; Valentine and Moss , 1997 ) and bear relevance to a long-standing controversy on the neural basis of fine echo ranging acuity of bats ( Menne et al . , 1989; Moss and Schnitzler , 1989; Simmons , 1979; Simmons et al . , 1990 ) . In summary , our study generated new discoveries in the field of systems neuroscience by integrating chronic neural recordings , multimedia tracking of dynamic animal behaviors in the 3D physical environment , and acoustic modeling . We report here the first empirical demonstration that neurons in a freely moving animal encode the 3D egocentric location of objects in the real world and dynamically shift spatial selectivity with sonar-guided attention . Specifically , we show that single neurons in the actively echolocating , free-flying bat respond selectively to the location of objects over a restricted distance ( echo delay ) , azimuth and elevation . Importantly , we discovered that the sensory response profiles of SC neurons become sharper along the range axis and shift to shorter distances ( echo delays ) when the bat actively inspects physical objects in its environment , as indicated by temporal adjustments in its echolocation behavior . Our discovery of dynamic 3D sensory representations in freely behaving animals call for comparative studies in other species , which can collectively contribute to a more complete understanding of nervous system function in the context of natural behaviors .
Two adult big brown bats , Eptesicus fuscus , served as subjects in this study . Bats were wild caught in the state of Maryland under a permit issued by the Department of Natural Resources and housed in an animal vivarium at the University of Maryland or Johns Hopkins University . Both the University of Maryland’s , and Johns Hopkins University’s Institutional Animal Care and Use Committee approved all of the procedures utilized for the current study . The two big brown bats were tested in related tasks , carried out in a 6 × 6 × 2 . 5 m room , illuminated with IR and equipped with 16 high-speed cameras and an ultrasound microphone array ( Figure 1 , see below ) . The first bat navigated around objects in a large flight room and landed on a platform . In order to ease the task for the second bat , it simply flew around the room , navigated around objects , and landed on any wall . Both bats were fed mealworms at the end of each trial to keep them active , but they were not rewarded for flight . The flight room was illuminated with infrared lighting ( ~850 nm ) to preclude the bat’s use of vision , ERG data show that Eptesicus does not see wavelengths longer than 600 nanometers ( Hope and Bhatnagar , 1979 ) . The room was also equipped with high-speed cameras and an ultrasound microphone array to track the bat’s flight path and record the bat’s echolocation behavior . Bats navigated around obstacles in the room ( explained in detail below ) , and were released at different locations in the room for each trial ( eight positions for Bat 1 , five different positions for Bat2 ) , which required them to use sonar echoes to steer around obstacles rather than a consistent or memorized flight path around objects in the room ( see Figure 3—figure supplement 1A ) . As such , the bats determined the duration and flight path of each trial . The obstacles were four plastic cylinders ( hard plastic as to be acoustically reflective ) , approximately 13 cm in diameter and 30 cm in length . Once the bat flew freely throughout the room and in the case of Bat 1 , learned to land on a platform , a surgery was performed to implant in the midbrain superior colliculus ( SC ) a 16-channel chronic recording silicon probe ( Neuronexus ) mounted on a custom microdrive . The bats’ weights were between 18 and 21 grams , and the weight of the implant , microdrive and transmitter device was 3 . 8 grams . The bat was given several days to rest and acclimate to the implanted device , after which they were able to fly and navigate around objects in the flight room . Data collection began after the animal was able to perform ~30 flight trials per session , which took place twice a day ( morning and afternoon ) in the experimental test room . During experimental sessions , there was no conditional reward; instead the bats were fed mealworms at the end of every trial , that is , when they landed . Bat 1 flew for 12 sessions , and Bat 2 flew for 15 sessions . For each recording session , the positions of the four flight obstacles were varied . Further , across trials the bat was released from different locations in the room . The obstacle configurations and flight start locations were varied to ensure that the bat’s flight trajectories covered the entire room , and the stimulus space sampled by the bat changed from trial to trial . This approach prevented the bats from relying on spatial memory and/or stereotyped flight paths . Figure 3—figure supplement 1A shows the bat’s flight trajectories in a single session and illustrates room coverage . Coverage was restricted in elevation , due to the height of the flight room , with a floor to ceiling dimension of approximately 250 cm . Although the landing behavior of the bats differed slightly ( i . e . landing on a platform vs . a wall ) , neural analysis was focused on the times when the animals were in flight and the data from the two bats are comparable . Additionally , both bats performed natural echolocation and flight behaviors as neural recordings were taken . The flight trajectory of the bat was reconstructed using a motion tracking system with 16 high-speed cameras ( Vicon ) . The motion tracking system was calibrated with a moving wand-based calibration method ( Theriault et al . , 2014 ) , resulting in sub-millimeter accuracy and 3D spatial location information of the bat at a frame rate of 300 Hz . Once the motion tracking system is calibrated , it tracks the bat in a 3D coordinate frame of reference , which we refer to as ‘world coordinates . ’ Affixed on the dorsal side of the transmitter board were three IR reflective markers ( 3 mm round ) that were then tracked with the high-speed motion tracking system ( Vicon ) . By tracking the 3D position of these three markers , we were able to determine the 3D position and head aim of the bat during the experiment . Around the perimeter of the room , at a distance from the walls of about 0 . 5 meters , the motion capture cameras did not provide adequate coverage , and data from the bat at these locations was not used for analysis . In addition to recording the position of the bat , we also recorded the sonar calls of the bat using an array of ultrasonic microphones ( Pettersson Elektronik , Ultrasound Advice , see Figure 1A ) . The microphone recordings were hardware bandpass filtered between 10 KHz and 100 KHz ( Alligator Technologies and Stanford Research Systems ) and were digitized using data acquisition systems ( National Instruments + custom built hardware ) . All three hardware systems ( i . e . neural recording , video-based 3D positioning , and microphone array ) were synchronized using the rising edge of a square pulse generated using a custom circuit . The square pulse was manually triggered at the end of each trial ( i . e . at the end of each individual flight ) when the bat landed on the platform/wall . At the generation of the TTL pulse , each system ( video and audio ) saved 8 s of pre-buffered data into the hard disk of the local computer . To ensure that the bats were not using spatial memory to guide their flight , we randomly released the bats from different spatial locations in the flight room . The average number of flights per session were 22 for Bat 1 and 27 for Bat 2 . Further , we used eight positions ( a-h ) for releasing Bat 1 and 6 positions ( a-f ) for releasing Bat 2 . To evaluate stereotypy in the bats’ flight paths , we used methods previously developed by Barchi et al . , 2013 . Occupancy histograms were created by collapsing the 3D trajectory data to 2D plan projection ( x , y and x , z ) . The number of points across a set of flight paths that fell inside 10 cm2 bins were counted . These points were converted to probabilities by dividing each bin count by the total number of points across each set of flights . After normalization , the occupancy histograms of trials could be compared within each session . The next step was to compute the autocorrelation of each trial and cross-correlation of each trial with every other trial . The maximum value of each 2D cross-correlation was divided by the maximum value of the autocorrelation . This ratio is shown as a matrix for a representative session for both bats in Figure 1—figure supplement 1 . The value of each square along the diagonal is one ( yellow on the color bar ) as it represents the autocorrelation of each flight trajectory . Cooler colors indicate minimum correlation between flight trajectories and warmer colors indicate stereotypy between trajectories . Once the bats were trained on the task , a surgery was performed to implant a 16-channel silicon probe ( Neuronexus ) . The probe consisted of four shanks spaced 100 μm micrometers apart , with four recording sites also spaced 100 μm apart on each shank , resulting in a 300 × 300 square μm grid of recording sites . The silicon probe was connected by a ribbon cable to an electrical connector ( Omnetics ) , and this assembly was then mounted on a custom-made , manual microdrive so that it could be moved through the dorsal/ventral axis ( i . e . across layers ) of the superior colliculus during the experiment . The silicon probe and microdrive assembly was then mounted on the head of the bat over a craniotomy performed above the superior colliculus ( SC ) . The SC sits on the dorsal surface of the brain of the big brown bat ( Valentine and Moss , 1997; Valentine et al . , 2002 ) , allowing for skull surface landmarks to be used in determining the implant location . Once the recording implant was positioned , a cap was made with cyanoacrylate ( Loctite 4013 ) to protect and secure the implant to the skull surface . The bat was allowed several days to recover , and then we started running the neural recording experiment . In order to study neural activity in the superior colliculus during a real-world navigation task , a wireless neural-telemetry system ( Triangle BioSystems International ) was used in conjunction with a multi-channel neural acquisition platform ( Plexon ) . This allowed for chronic neural recordings to be collected from the superior colliculus ( SC ) while the echolocating bat was navigating around obstacles in flight . During the experiment , a wireless RF telemetry board ( Triangle BioSystems International ) was connected to the plug of the silicon probe mounted on top of the bat’s head . Bat 1 flew for 12 sessions while recordings were made in the SC , and Bat 2 flew for 15 sessions . Each session typically lasted 30–45 min , and the microdrive was advanced at the end of each session to collect activity from a new set of neurons in the following recording session . Neural data were sorted offline after filtering between 800 and 6000 Hz using a 2nd order elliptic filter . Filtered neural traces were then sorted using a wavelet based algorithm and clustering technique ( Quiroga et al . , 2004 ) . Furthermore , we determined the Lratio and isolation distance for each wavelet-based cluster in order to provide a traditional measure of the efficacy of our clustering technique . In previous reports , an Lratio less than 0 . 07 , and an isolation distance more than 15 , were used as thresholds for significantly separated spike-waveform clusters ( Saleem et al . , 2013; Schmitzer-Torbert et al . , 2005 ) . For our wavelet-based clustering technique , all Lratio’s were less than 0 . 05 , and isolation distances were greater than 15 ( Figure 1—figure supplement 3 ) , providing a secondary quantitative metric of the significant separation of our single unit clustering . This algorithm also separated movement artifact out of the raw neural traces . If any spike events occurred simultaneously with movement artifact , however , they were not recoverable . Movement artifact rarely occurred across all channels during flight and was mostly confined to times when the bat was landing . We only used data from the bats in flight for analysis . Of all sorted single units ( n = 182 ) , 67 units ( sensory neurons ) were selected for analysis , as described below . The isolated single units were stable throughout the session ( see Figure 1—figure supplement 2 ) . Audio recordings were analyzed using custom Matlab software to extract the relevant sound features , that is , pulse timing , duration , and interval . Combining the pulse timing ( time when sound reached a stationary microph one ) with the 3D flight trajectory data allowed compensating for the sound-propagation delays and calculating the actual call production times at the source ( i . e . the veridical time when the bat produced the sonar sound ) . Sonar sound groups ( SSGs ) are defined as clusters of two or more vocalizations which occur at a near constant PI ( within 5% error with respect to the mean PI of the sound group ) , and are flanked by calls with a larger PI at both ends ( at least 1 . 2 times larger ) ( Kothari et al . , 2014; Moss and Surlykke , 2001; Moss et al . , 2006 ) . SSGs of two vocalizations are also produced by the bat , and our criteria for these SSGs is that surrounding PI’s must be at least 1 . 2 times larger than the PI between the two vocalizations contained within the SSG . Here , we use the same definitions and thresholds as used in prior work ( see Figure 6A for a visual explanation ) . As we use pulse rate in the main text , it is important to note that Pulse Interval = 1/Pulse Rate . The ‘echo model’ is an acoustic model , which takes into account the instantaneous 3D position of the bat , 3D positions of the objects , the bat’s head direction vector , time of production of the sonar sound as well as the physical parameters of sound in air , in order to compute the direction and time of arrival of echoes at the bat’s ears . For this model , each time the bat vocalized , we computed the arrival time and direction of returning echoes . Figure 2—figure supplement 1A shows an outline of a bat with the neural telemetry headstage ( TBSI ) . The headstage is shown as a grey box with a 16-channel Omnetics connector ( male and female ) at the bottom . Three reflective markers ( 4 mm diameter ) , P , Q and R ( black ) , which are tracked by the infrared motion tracking cameras ( Vicon ) are also shown . A top view ( cartoon ) of the bat and telemetry headstage , with markers is shown in Figure 2—figure supplement 1B . In order to classify neurons , we developed an algorithm based on variability in the firing latency distributions of spike times with respect to echo arrival time , previous call production time , and next call production time . In simple terms , this algorithm measures the variability in spike latencies to echo time and call time ( previous and next ) as a way of classifying neurons as sensory , vocal premotor or sensorimotor . This determination was based on the assumption that a neuron’s activity is most temporally coupled with its functionally relevant event . If a neuron’s spike latency distribution was sharpest with respect to echo arrival time , it was classified as sensory; if spike latencies were sharpest with respect to pulse time , the neuron was classified as vocal premotor , and if spike latencies showed clustering around pulse time and echo arrival times , it was classified as sensorimotor . It is important to mention that for sensory neurons we further solved the problem of echo assignment by only considering neurons that fire for the first arriving echo and do not exhibit activity for subsequent echo events ( see Figure 3 ) . This also solves the problem of wall/camera/microphone echoes , as they were the last to arrive . More than 90% of the sensory neurons analyzed in this study responded only to the first echo . For the remaining neurons that responded to a cascade of echoes ( about 10% of those sampled ) , it was not possible to reliably assign their activity to specific echo arrival times and we therefore excluded them from the data reported in this paper . Using this algorithm , the 182 recorded neurons were classified as sensory ( n = 67 ) , vocal premotor ( n = 26 ) and sensorimotor ( n = 45 ) . Classification into sensory , sensorimotor and premotor categories is common for SC neurons ( Mays and Sparks , 1980; Schiller and Koerner , 1971 ) . The remaining 44 neurons were unclassified . Spatial tuning profiles were only constructed for the sensory neurons ( n = 67 ) . Once a neuron was identified as sensory ( see above criterion ) , direction information from the echo model was converted into egocentric coordinates of the bat’s instantaneous position and the X , Y and Z information was converted into azimuth , elevation and range coordinates . Further , we test spatial selectivity based on an ANOVA ( p<0 . 05 ) performed along each dimension ( azimuth , elevation and range ) . Only cells which passed the ANOVA for each dimension were used for further analysis . Neural responses of cells that passed the spatial selectivity test were normalized based on the amount of coverage in each of these dimensions , as explained below . The spatial response profiles ( for neurons which pass the spatial selectivity test ( see above ) were then normalized using the stimulus space , that is , the time spent by the animal , in each dimension ( see Figure 3—figure supplement 1D – range , E – azimuth and F – elevation ) : that is , the spike-count spatial response profile was divided by the time-spent spatial profile , to yield a spiking probability per bin in each dimension ( distance , azimuth , and elevation ) . Regions of the stimulus space with echo events per bin less than one standard deviation from the mean were excluded from the computations ( indicated by open bins in Figure 3—figure supplement 1D , E and F ) . Finally , normalized spatial response profiles in each dimension were then fit to a Gaussian function using the fit function in Matlab . Spatial response profile means , half widths and standard deviations are then taken from the Gaussian fit . Out of the 67 sensory neurons ( see criterion above ) , overlapping populations of neurons showed either 3D , 2D or 1D spatial selectivity . 46 neurons ( Bat 1–19 and Bat 2–27 ) showed spatial selectivity in 3D ( azimuth , elevation and depth ) . Further , 56 , 52 and 51 neurons showed 1D spatial selectivity , for depth , azimuth and elevation , respectively . Figure 4—figure supplement 3 describes the complete distribution of 3D , 2D and 1D neurons . The mean response latencies of single sensory neurons we recorded was 5 . 9 ± 3 . 4 ms . In more detail , the minimum spike latency was 3 ms and the minimum s . d . of latency was 1 ms . The median s . d . of the response latencies for the 67 sensory neurons was 3 . 8 ms . Previous publications have reported a wide range of response latencies in the SC of the passively listening bat , as long as 40 ms , but also as short as 4 ms ( Valentine and Moss , 1997 ) , 3 . 6 ms ( Jen et al . , 1984 ) and 4 ms ( Wong , 1984 ) , and short latency responses are likely mediated through a direct projection from the nucleus of the central acoustic tract to the SC ( Casseday et al . , 1989 ) . Separate range tuning profiles are computed for each cell for SSG and non-SSG vocalizations . Variance ( sharpening ) of SSG and non-SSG tuning profiles was tested using the non-parametric Brown-Forsythe test of variance at the α level of 0 . 05 . The test results for each cell are described in detail in table supplementary table 1 ( also see Figure 5E ) . Also , SSG and non-SSG distance tuning curves were tested using the Wilcoxon rank-sum test . Test statistic details for each cell is given in table supplementary table 2 ( also see Figure 5F ) . The firing of auditory neurons in the echolocating big brown bats is very sparse ( see for example Dear et al . , 1993; Valentine and Moss , 1997 ) . For the SSG and non-SSG analysis ( above ) we separated spiking activity when the bat produced SSGs and nonSSGs . This resulted in some of the data sets containing low spike counts . To ensure that for each comparison , for each neuron , we had enough statistical power , we performed a permutation test . Here , we combined the data for SSG and nonSSG data sets and randomly shuffled and picked spikes ( without repetitions ) . Following this , we performed the Brown-Forsythe test or the Wilcoxon rank-sum test , for the sharpening and shifting groups , respectively . We repeated this procedure 1000 times and each time we collected the value of the test statistic . Finally , we compared the test statistic value of the original sample to the distribution obtained from the shuffled groups and obtained a p-value . We only included in the analysis the cells , which passed the test at the p<0 . 05 criterion level , which excluded 3/56 cells from Figure 5E and 5/56 cells from Figure 5F . The local field potential ( <300 Hz ) was extracted from each channel recording using second order elliptical filters . Further , we analyzed the gamma band ( 50–140 Hz ) ( Goddard et al . , 2012a; Sridharan and Knudsen , 2015 ) to investigate whether the epochs when the bat produced sonar sound groups ( SSGs ) were correlated with gamma band activity . We first identified channels without distortions in the LFP as a result of movement artifact ( Figure 6—figure supplement 2 ) . We then extracted 100 ms spike triggered LFP windows from corresponding recording sites . We separated these into SSG and non-SSG events and averaged these separately to estimate the root mean squared ( RMS ) gamma band power ( Jaramillo and Zador , 2011 ) ( Figure 6A and B ) when the bat produced SSG and non-SSGs . Further , to investigate the timing of the gamma signal , the averaged gamma band amplitude envelope was normalized across SSG and non-SSG trials across each neuron . A Gaussian was fit to each time waveform to estimate the peak ( Figure 6C and D ) . The average of the peaks across all units was taken as the average latency of the LFP following the spike event . We also examined whether movement artifact from the bat’s wing beats could have corrupted the LFP analysis . The bat’s wingbeat is approximately 12 Hz , whereas the frequency range for the Gamma band we analyzed was 50–140 Hz . The third harmonic of the wingbeat , which would be in the frequency range of the Gamma band , was significantly attenuated . To further ensure that movement artifact did not corrupt the analysis of the LFP , we chose channels where the power ratio between the low frequency band ( 10–20 Hz ) and the gamma band was less than 6 dB . We identified 21 low noise channels containing 26 single neuron recordings , ( see Figure 6—figure supplement 2 ) , which were then used for further analysis . The original raw data can be obtained upon request from NBK , MJW or CFM ( cynthia . moss@jhu . edu ) . Given the size of the raw data ( approx . . 2 terabytes ) , the full dataset has not been deposited to a public repository , but partial and processed data sets to generate Figures 5E , 5F , 6C , 6D and E have been made available through an open source license on GitHub ( Kothari et al . , 2018 copy archived at https://github . com/elifesciences-publications/Dynamic-3D-auditory-space-in-bats ) . | Humans and other animals can navigate their natural environments seamlessly , even if there are obstacles in their path . However , it is not well understood how an animal’s brain processes information from the senses to map where it is in relation to these objects , both in terms of distance and direction . Bats can help answer these questions because they use a biological navigation system: echolocation . Bat produce high-pitched squeaks and then listen to the echoes that return when the sound bounces off of nearby objects . A bat can then use this information to estimate both which direction an object is , and how far away it is . Bats can also change their echolocation signals to focus on different objects , and researchers can record and analyze these signals to directly measure what the bat is paying attention to . Kothari , Wohlgemuth and Moss have now investigated how the brain cells of bats process the animals’ movements while flying in three-dimensional space . A wireless probe was inserted into the midbrain region of each bat to detect whenever there was an electrical impulse in the nearby brain cells . The bats were then allowed to fly freely in a large room that contained obstacles , while each bat’s echolocation signals and brain activity were recorded . The experiments revealed a group of brain cells that codes for the position of an object in three-dimensional space . Kothari , Wohlgemuth and Moss noted that these brain cells register the distance to objects more precisely when the bat changed its echolocation behavior to focus on those objects . Moreover , the activity in the bat’s brain also shifted when the bat noticed a closer object . These findings are not only relevant to echolocating bats , but rather reflect the general role that shifts in attention may play when many species map the locations of objects around them . Further similar studies with other species would contribute to a more complete understanding of animals’ nervous systems work under natural conditions . In the future , these findings , and the studies that build upon them , could be applied to other fields of research like medicine or engineering . For example , smart wireless devices , designed to record and transmit physiological measurements based on movement , could efficiently monitor human health , and robots equipped with adaptive sonar could navigate effectively in complex environments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Dynamic representation of 3D auditory space in the midbrain of the free-flying echolocating bat |
New therapeutic targets for oral squamous cell carcinoma ( OSCC ) are urgently needed . We conducted genome-wide CRISPR-Cas9 screens in 21 OSCC cell lines , primarily derived from Asians , to identify genetic vulnerabilities that can be explored as therapeutic targets . We identify known and novel fitness genes and demonstrate that many previously identified OSCC-related cancer genes are non-essential and could have limited therapeutic value , while other fitness genes warrant further investigation for their potential as therapeutic targets . We validate a distinctive dependency on YAP1 and WWTR1 of the Hippo pathway , where the lost-of-fitness effect of one paralog can be compensated only in a subset of lines . We also discover that OSCCs with WWTR1 dependency signature are significantly associated with biomarkers of favorable response toward immunotherapy . In summary , we have delineated the genetic vulnerabilities of OSCC , enabling the prioritization of therapeutic targets for further exploration , including the targeting of YAP1 and WWTR1 .
Head and neck squamous cell carcinoma ( HNSCC ) is a heterogeneous tumor arising from the mucosal surfaces lining the upper aerodigestive tract . The commonest subtype , oral squamous cell carcinoma ( OSCC ) is especially prevalent among Asian countries ( Bray et al . , 2018 ) . OSCC has been associated with distinct risk habits such as betel quid chewing , tobacco smoking and alcohol consumption ( Shield et al . , 2017 ) . The 5-year survival rate for OSCC is about 50% ( Kumar et al . , 2016 ) and surgery remains the mainstay of treatment . Cetuximab , an inhibitor of the epidermal growth factor receptor ( EGFR ) , is used in combination with platinum-based chemotherapy for the treatment of advanced OSCC ( Vermorken et al . , 2008 ) . However , the improvement in survival remains marginal ( Vermorken et al . , 2008 ) . More recently , immune checkpoint inhibitors have been approved for the treatment of advanced and metastatic OSCC ( Cohen et al . , 2019 ) . Although an improvement in patients’ outcome is anticipated with the advancement of immunotherapy , clinical trial outcomes showed an average objective response rate of only 13–36% ( Bauml et al . , 2017; Burtness et al . , 2019; Ferris et al . , 2016 ) , and the factors determining response towards checkpoint inhibitors are still largely unknown . This underscores the need to identify further therapeutic targets for OSCC . Genomic sequencing technology has enabled the delineation of the comprehensive mutational and transcriptomic landscape of cancers , including OSCC ( The Cancer Genome Atlas Network , 2015; Pickering et al . , 2013 ) . However , the functional significance of most of these genetic alterations remains unclear and little is known about their value as therapeutic targets for OSCC . Identifying the genetic dependencies of OSCC will , therefore , be critical for the development of novel therapies . Genome-scale functional genetic screens allow the high-throughput identification of genes that govern cell survival ( Gerhards and Rottenberg , 2018 ) . Previously , such genes were identified using RNA interference ( RNAi ) technology ( McDonald et al . , 2017; Tsherniak et al . , 2017 ) . More recently , essential genes have been identified through the use of CRISPR-Cas9 technology due to its high specificity and efficiency compared to RNAi ( Gerhards and Rottenberg , 2018 ) . Several studies using genome-wide CRISPR-Cas9 screen have already shown promising outcome in identifying novel cancer-specific vulnerabilities that are useful drug targets ( Steinhart et al . , 2017; Wang et al . , 2017 ) , as well as improving the understanding of drug mechanism of action ( Barazas et al . , 2018; Hou et al . , 2017 ) . The Cancer Dependency Map project ( a consortium effort by the Wellcome Sanger Institute and the Broad Institute ) have conducted CRISPR-Cas9 screen on a large number of cell lines including some OSCC models ( Behan et al . , 2019; Meyers et al . , 2017 ) . However , there is a lack of representation of Asians OSCC , such as those associated with betel quid chewing habit , one of the major risk factors of OSCC in many Asian countries ( Shield et al . , 2017 ) . Further , the comparison of genomics data across different populations has revealed distinctive features in the different populations ( Chai et al . , 2020 ) . To identify genetic vulnerabilities in OSCC , we performed genome-wide CRISPR-Cas9 screens on 21 highly annotated OSCC cell lines , most of which are unique models derived from Asian patients ( Fadlullah et al . , 2016 ) . Our study contributes to approximately one-third of the OSCC functional genetic screens currently available globally , expanding the representation of this heterogeneous disease ( Behan et al . , 2019; Meyers et al . , 2017 ) . In addition to finding known genetic vulnerabilities , we also uncover novel candidate genes essential for OSCC survival that can facilitate the development of new targeted therapies for OSCC . We validated the essentiality of Yes-associated protein 1 ( YAP1 ) and WW domain-containing transcription regulator protein 1 ( WWTR1 ) and revealed mutually exclusive dependency and compensable functions of these paralogs in different subsets of OSCC models . We identified OSCC tumors with a gene expression signature similar to cell lines with validated dependencies . Among which , OSCC resembling the WWTR1-dependent cell lines showed significant enrichment of immune-related pathways and are associated with biomarkers of response towards checkpoint inhibitors . In summary , our study demonstrated the robustness of genome-wide CRISPR-Cas9 screen in identifying genetic vulnerabilities in diverse OSCC models , offering new molecular insights into this disease .
In order to identify genetic vulnerabilities of OSCC , particularly those that are more relevant to tumors of Asian origin , we conducted genome-wide CRISPR-Cas9 knockout screens ( Figure 1A and Figure 1—figure supplement 1A ) . We screened a unique set of 14 well-characterized OSCC cell lines termed the ORL-series [ORL-48 , –115 , −136 , –150 , −153 , –156 , −166 , –174 , −188 , –195 , −204 , –207 , -214 , –215] . These were established from the tumors of Malaysian OSCC patients ( Fadlullah et al . , 2016 ) and are comprehensively annotated with whole-exome sequencing ( WES ) and RNA sequencing data . In addition , we screened a further seven OSCC cell lines [BICR10 , Ho-1-u-1 , HSC-2 , HSC-4 , PE/CA-PJ15 , SAS and SCC-9] sourced from commercial cell line repositories . Demographic details of the patients from whom the 21 OSCC cell lines were derived are shown in Figure 1A . The presence of mutations and copy number alterations in the top five significantly mutated genes are indicated . Overall , we find that our selection of cell lines represents the diversity of mutated driver genes observed in OSCC . Fitness genes were identified after an unsupervised computational correction with CRISPRcleanR ( Behan et al . , 2019; Iorio et al . , 2018 ) , followed by mean-variance modeling and systematic ranking of significantly depleted genes using MAGeCK ( Li et al . , 2014; Figure 1—figure supplement 1B ) . At a false discovery rate ( FDR ) of 5% , the number of significantly depleted genes ranged from 525 genes in ORL-156 to 1399 genes in ORL-215 ( Figure 1A and Supplementary file 1 ) . As we aimed to identify genetic vulnerabilities of OSCC that can be safely targeted therapeutically , we filtered the significantly depleted genes to exclude previously defined core fitness genes ( Behan et al . , 2019; Hart et al . , 2014; Hart et al . , 2017; Meyers et al . , 2017 ) , that were found to be essential across the many cell lines from different lineages , and are likely toxic to the cells when targeted . In general , more than 80% of all the MAGeCK hits were found to be core fitness genes and were filtered out ( Figure 1B and Supplementary file 1 ) . Overall , among the 18 , 010 genes screened , 2539 ( 14% ) were found to be significant MAGeCK hits in at least one cell line and following the removal of core fitness genes ( Supplementary file 2 ) , 918 context-specific fitness genes were shortlisted for further prioritization ( Figure 1B ) . About 40% ( 366 genes ) were uniquely essential in a single cell line , while the remaining 60% ( 552 genes ) were essential in at least two cell lines , hence are recurrent context-specific essentialities ( Figure 1C ) . Pathway enrichment analyses on the 918 context-specific fitness genes was conducted on each individual cell line ( Figure 2—figure supplement 1A ) . Consistent with the highly heterogeneous nature of OSCC , diverse pathways were enriched across these cell lines . Pathways enriched among the 918 genes revealed several cancer-related pathways , potentially comprising important cancer-specific targets ( Figure 2—figure supplement 1B and Supplementary file 3A-B ) . Whilst pathways such as ubiquitin-mediated proteolysis and cellular senescence are common across all cell lines , pathways such as NF-kappa B and MAPK signaling pathways are selectively enriched only in subsets of cell lines . Next , we sought to determine which components of the common oncogenic pathways altered in HNSCC ( The Cancer Genome Atlas Network , 2015 ) were required for cancer cell fitness and annotated the genes with the frequency of dependency ( Figure 2A ) . Notably , most of the fitness genes are either existing drug targets or deemed clinically actionable . For example , drugs targeting PIK3CA and CDK6 are already in clinical trials for HNSCC treatment [NCT01816984 , NCT02537223 , NCT03356223 , NCT03356587] . We also examined the dependency profile of 44 cancer genes with driver mutations known to be associated with HNSCC ( Bailey et al . , 2018; Martincorena et al . , 2018; Figure 2B ) and found that more than half of these cancer genes were dispensable for OSCC survival . Oncogene addiction has been the promising source of finding the Achilles heel for successful molecular targeted therapy ( Weinstein and Joe , 2008 ) . Based on the WES data of the 21 OSCC cell lines , we found 43 genes with driver mutations in at least one cell line and plotted the CRISPR score ( measure of sgRNA depletion in the CRISPR screen ) to examine if there is any differential dependency associated with the mutations ( Figure 2C and Supplementary file 4 ) . Dependencies on mutated PIK3CA were observed in four OSCC cell lines with a hotspot mutation in E545K ( BICR10 ) , Q546R ( ORL-150 ) and H1047R ( HSC-2 ) , and to a lesser extent , in ORL-115 with H1047L mutation . Intriguingly , HSC-4 , which harbors the same E545K mutation as BICR10 , did not show any dependency on the mutated PIK3CA , this is consistent with findings from Project Score ( Behan et al . , 2019 ) . A splice site driver mutation in PTEN co-occurred in HSC-4 and may have counteracted the oncogene addiction effect on the mutated PIK3CA , as suggested previously in breast cancer ( Lazaridis et al . , 2019 ) . Dependency on NFE2L2 was observed in Ho-1-u-1 and BICR10 however the mechanism of activating this oxidative pathway differed between these two cell lines . Oncogene addiction is observed for Ho-1-u-1 with a hotspot mutation ( E82D ) in NFE2L2 , that has been shown to enhance its transcriptional activity and promoting cell proliferation ( Shibata et al . , 2008 ) . On the other hand , BICR10 harbors an inactivating mutation on KEAP1 ( R320Q ) a negative regulator of NFE2L2 . The R320Q mutation has been reported to stabilize NRF2 ( encoded by NFE2L2 ) and enhances cell fitness as reported previously in lung cancer ( Hast et al . , 2014 ) . Finally , the only cell line that shows dependency on HRAS is ORL-214 which carries a mutation in HRAS ( G12C ) . Another gene , encoding ZFP36L1 with truncating mutation at S324 in ORL-48 also showed preferential dependency , suggesting that the effect of this mutation should be studied further . Studies on the genomic landscape of Asian and Caucasian OSCC have revealed distinct molecular differences , suggesting that some population-specific risk habits might have contributed to these differences ( The Cancer Genome Atlas Network , 2015; Chai et al . , 2020; Hsieh et al . , 2001; India Project Team of the International Cancer Genome Consortium , 2013; Zanaruddin et al . , 2013 ) . Betel quid chewing is frequently associated with OSCC in Asia ( Guha et al . , 2014; India Project Team of the International Cancer Genome Consortium , 2013; Shah et al . , 2012 ) . In this study , several Asian-derived OSCC models that were associated with betel-quid chewing were included , and we had the opportunity to determine if there are differences in genetic dependencies between these OSCC ( ORL-115 , ORL-136 , ORL-174 , ORL-195 , ORL-204 , ORL-207 , ORL-214 ) ( n = 7 ) , with those that are not associated with betel quid chewing ( n = 14 ) ( Figure 2D ) . Of the 110 fitness genes uniquely seen in betel-quid-associated OSCC , the NF-kB signaling pathway stands out as one of the significantly enriched pathway . The fitness genes from this pathway that are unique to the betel quid-associated OSCC include NFKB2 , TNFAIP3 , CSNK2A1 , and TRIM25 . When cross checking with the DepMap and Project Score data , three out of four ( 75% ) of these genes ( NFKB2 , TNFAIP3 , and TRIM25 ) were not found among the screened OSCC models , which were mostly derived from Caucasians , or from Asians not known to chew betel quid . Interestingly , our findings are coherent with previous studies from Taiwan and India , where betel quid chewing is common , which have demonstrated that extract from the areca nut of betel quid can directly activate the NF-kB signaling pathway , favoring OSCC cells survival ( Chiang et al . , 2008; Islam et al . , 2019; Lin et al . , 2005 ) . Given that many of the reported HNSCC-related cancer genes do not appear to be fitness genes , we sought to determine which of the 918 genes could potentially be tractable using previously defined frameworks ( Behan et al . , 2019; Brown et al . , 2018; Figure 2E and Supplementary file 5 ) . From the 918 genes , 45 genes fall into the tractability group 1 , some examples of genes in this group include EGFR , PIK3CA , CDK4 , and CDK6 , where anticancer drugs targeting these genes are already approved or clinical trials for the treatment of HNSCC are on-going , demonstrating the robustness of our results . When classified based on protein function using PANTHER ( Mi et al . , 2013 ) , most of those in tractable group 1 , are transferase ( kinases ) ( 32% ) and oxidoreductase ( 30% ) , including genes like CDK4 , CDK6 and PIK3CA; and several genes in the family of NADH:ubiquinone oxidoreductase such as NDUFB9 and NDUFC2 ( Figure 2—figure supplement 1C ) . Interestingly , emerging oncology and non-oncology drugs such as the HDAC inhibitors ( Yoon and Eom , 2016 ) and miglustat , an approved drug for Gaucher’s disease ( Barth et al . , 2013 ) , were amongst potential drug repurposing candidates that target fitness genes in tractability group 1 ( targeting HDAC2 and UGCG respectively ) . The only transcription factor that falls within tractability group one is the ESR2 , with several antagonists available owing to its well-studied ligands and structure ( Ho , 2004; Figure 2—figure supplement 1C ) . A further 210 genes ( 23% ) belong to tractability group 2 , which harbors novel targets that have evidence supporting their tractability . Albeit no drugs are currently in clinical trials , these may hold potential for future drug development . For example , several companies are developing drugs that could target YAP1 , SLC2A1 and PTPN11 which are in tractability group 2 . However , about 70% of the 918 genes belong to tractability group 3 ( least tractable or lacking evidence ) where significant efforts in understanding their structure and function would be necessary to evaluate their tractability . Consistent with previous reports ( Behan et al . , 2019 ) , the least tractable group comprises mainly of nucleic acid-binding proteins and transcription factors , such as the Kruppel-like factors ( KLF ) gene families ( KLF4 and KLF5 ) and zinc finger proteins ( ZNF148 and ZNF236 ) ( Figure 2—figure supplement 1C ) . HNSCC belongs to the ‘C class’ tumor , where the landscsape of genomic alterations is dominated by copy number alterations including recurrent chromosomal gains and losses ( Ciriello et al . , 2013 ) . The most frequently reported copy number gains occurr in chromosomes 3q , 5 p , 7 p , 8q , and 11q ( The Cancer Genome Atlas Network , 2015; Salahshourifar et al . , 2014 ) . To identify putative oncogenes that are essential for OSCC within these amplified regions , we evaluated the number of candidate fitness genes before and after CRISPRcleanR correction for copy number bias ( Figure 3—figure supplement 1A ) . After correction , no enrichment/bias was found in the frequently amplified chromosome , demonstrating effective correction of copy number bias ( Figure 3—figure supplement 1B ) . KEGG pathway analysis of the 152 genes from the five amplified regions showed enrichment of several oncogenic signaling pathways such as the small cell lung cancer , Hippo signaling pathway and ErbB signaling pathway ( Figure 3—figure supplement 2A ) . We focused our analysis on the Hippo signaling pathway ( Figure 3—figure supplement 2B ) , which has recently been implicated for major oncogenesis role in squamous cell carcinoma , including OSCC ( Campbell et al . , 2018; Ge et al . , 2011; Hiemer et al . , 2015; Wang et al . , 2018 ) . Furthermore , YAP1 or WWTR1 amplifications occur in approximately 19% of HNSCC ( YAP1–5 . 5% , WWTR1–14 . 3% ) , which puts it among the top five cancers with the highest amplification of these genes amongst 33 cancer types ( Wang et al . , 2018 ) . YAP1 and WWTR1 ( also known as TAZ ) are transcription co-activators , which are the major effectors of the Hippo pathway ( Guo and Teng , 2015; Wang et al . , 2018 ) . YAP1 and WWTR1 are paralogs with ~46–60% similarity in their amino acid sequence ( Guo and Teng , 2015 ) . They were shown to have both overlapping and distinct roles in different contexts ( Guo and Teng , 2015; Plouffe et al . , 2018 ) . In OSCC , overexpression of YAP1 and WWTR1 has been shown to increase proliferation , survival and migration , mainly via interaction with the transcriptional enhanced associate domain ( TEAD ) transcription factors ( Hiemer et al . , 2015 ) . Interestingly , across the 21 OSCC cell lines , there is a subset of lines that show significant dependency on only one of the paralogs , while another subset of lines does not exhibit significant dependency on either YAP1 or WWTR1 ( Figure 3A ) . With the exception of two cell lines , ORL-153 and ORL-215 , YAP1-dependent lines and WWTR1-dependent lines are mutually exclusive . There were seven lines that were highly dependent on YAP1 ( ORL-48 , ORL-136 , ORL-156 , ORL-204 , ORL-207 , SAS , SCC-9 ) , four highly dependent on WWTR1 ( ORL-174 , ORL-188 , ORL-214 , PE/CA-PJ15 ) , while the other eight were not dependent on either ( Figure 3A ) . Notably , WWTR1 gene locus ( 3q25 ) is at close proximity to the locus of PIK3CA , SOX2 and TP63 at 3q26-28 , whereby their focal amplification is frequently reported in HNSCC ( The Cancer Genome Atlas Network , 2015; Figure 3—figure supplement 2C ) . The majority of the WWTR1-dependent cell lines have copy number amplification for these genes on 3q25-28 ( Figure 3A ) , but only WWTR1 CRISPR scores are significantly different between those with and without copy number amplification ( p<0 . 001 ) ( Figure 3—figure supplement 2D ) . Only two of nine cell lines with PIK3CA copy number amplification are dependent on PIK3CA itself . This suggests that copy number amplification of WWTR1 may constitute to a functional oncogenic role of WWTR1 in OSCC , instead of being a passenger gene that is co-amplified with the canonical HNSCC oncogene , PIK3CA . Notably , we also observed an enrichment of PIK3CA mutations ( p=0 . 0003 ) among cell lines that are neither dependent on YAP1 or WWTR1 , whereby five out of six such lines have PIK3CA hotspot mutations ( BICR10 , HSC-2 , HSC-4 , ORL-115 , and ORL-150 ) ( Figure 3A ) . Intriguingly , mutually exclusive copy number gains of chromosome 3q and 11q22 ( where YAP1 is mapped to ) have been reported in squamous cell carcinoma ( Campbell et al . , 2018 ) and consistently , YAP1 and WWTR1 amplification were also found to be mutually exclusive in HNSCC ( Wang et al . , 2018 ) . In our study , despite having cell lines dependent on YAP1 , none of the 21 OSCC cell lines shows copy number amplification of YAP1 or neighbouring genes on the chromosome 11q22 . This suggests that other non-genomic mechanisms could likely be in place to activate YAP1 . To investigate if expression of YAP1 and WWTR1 is associated with the dependency , we examined baseline mRNA and protein expression of YAP1 and WWTR1 in representative OSCC lines ( Figure 3—figure supplement 2E–F ) . Among the YAP1-dependent lines ( ORL-48 , ORL-204 ) , overexpression of YAP1 mRNA and protein levels were observed . Further , these lines have low protein expression of WWTR1 . The mRNA and protein expression of WWTR1 are relatively higher than YAP1 among the WWTR1-dependent cell lines ( ORL-214 and PE/CA-PJ15 ) , and those that were not affected when either YAP1 or WWTR1 is knocked-out ( ‘non-dependent’ ) . To determine if the association between dependency and gene expression is exclusive to OSCC , we computed the differential dependency score of YAP1 and WWTR1 and their gene expression for 273 cancer cell lines from Project Score ( Behan et al . , 2019; Figure 3—figure supplement 3A ) . Generally , cell lines that are dependent on YAP1 have higher YAP1 expression , and similarly , WWTR1-dependent lines have higher expression of WWTR1 compared to YAP1 . WWTR1 gene expression showed significant negative correlation with its dependency in these 273 cancer cell lines ( Pearson R = −0 . 570 , p-value=3e-25 ) and showed non-significant negative correlation among our 21 OSCC lines screened ( Pearson R = −0 . 354 , p-value=0 . 116 ) ( Figure 3—figure supplement 3B–C ) . In other words , the higher dependency on WWTR1 gene is associated with higher WWTR1 gene expression . Interestingly , this observation was also seen in other cancer types , including non-small cell lung carcinoma , squamous cell lung carcinoma , glioblastoma , breast carcinoma , and lung adenocarcinoma ( Figure 3—figure supplement 3D ) , based on the Project Score data . Among these cancers , non-small cell lung carcinoma showed the highest percentage of WWTR1-dependency ( 30% ) , with the strongest correlation ( Pearson’s R = −0 . 934 , p-value=0 . 0021 ) . To validate the differential dependency for YAP1 and WWTR1 , we performed single-gene knockout using two sgRNAs per gene and investigated the growth inhibition effect of gene knockout using co-competition assay , as previously reported ( Behan et al . , 2019 ) . Two cell lines each from the YAP1-dependent ( ORL-48 and ORL-204 ) , WWTR1-dependent ( ORL-214 and PE/CA-PJ15 ) and YAP1/WWTR1 non- dependent ( BICR10 , HSC-2 ) groups were used . The efficacy of protein knockout using individual sgRNAs was assayed with western blotting ( Figure 3B ) . The results obtained from the co-competition assay corroborated our CRISPR screen data ( Figure 3C ) . More than half of the transduced cell population was depleted following YAP1-knockout in ORL-48 and ORL-204 cells , but not when WWTR1 was knocked-out . Likewise , the growth inhibition in ORL-214 and PE/CA-PJ15 was only seen following WWTR1-knockout , but not upon YAP1-knockout . On the other hand , the fraction of transduced cells in BICR10 and HSC-2 did not show any prominent changes upon knockout of either YAP1 or WWTR1 . These experiments were validated using clonogenicity assays ( Figure 3—figure supplement 4A ) . Together , these results support the differential dependency pattern on YAP1 and WWTR1 . Besides , another YAP1-dependent line , SAS has recently been reported to harbor a fusion protein of YAP1 and MAML2 ( Picco et al . , 2019 ) . Our co-competition assay also confirms the dependency on YAP1 in these cells and the differential depletion of sgRNA targeting early and late exons of YAP1 suggested that this oncogenic fusion protein provided a survival advantage ( Figure 3—figure supplement 4B–E ) . In OSCC , CTGF , and CYR61 are two canonical transcriptional targets of YAP1 and WWTR1 ( Hiemer et al . , 2015 ) . Consistent with their known pro-survival properties , more substantial reduction of CTGF and CYR61 gene expression were seen only when the respective upstream fitness genes were knocked-out , as measured by qPCR ( Figure 3D and Figure 3—figure supplement 4F ) . Since YAP1/WWTR1 were both known to regulate proliferation and apoptosis , we next investigated whether the depletion of YAP1 and WWTR1 affects cell proliferation or apoptosis in the selected OSCC cells . Consistently , YAP1 depletion in ORL-48 resulted in significant reduction in viable cells , to a level comparable to the depletion of PLK1 ( Figure 4A ) , and this was reflected in the increase in apoptotic cells ( Figure 4B ) . Further , WWTR1 depletion in ORL-214 showed significantly lower percentage of viable cells when compared with the control , as well as significant increase in apoptotic cells ( Figure 4A–B ) . Overall , these results confirm the dependency on either YAP1 or WWTR1 for survival , whereby depletion of the respective fitness genes resulted in increased apoptosis . By contrast , depletion of either YAP1 or WWTR1 did not affect the survival of HSC-2 , as confirmed by the apoptotic assay ( Figure 4A–B ) . The lack of dependency on either YAP1 or WWTR1 in the non-dependent lines is intriguing , given the importance of these genes in most OSCC cancer cells . As YAP1 and WWTR1 share high structural homology and have common downstream targets , we hypothesized that YAP1 and WWTR1 could provide compensatory functions to maintain the survival of BICR10 and HSC-2 cell lines when either one of the genes was knocked-out . To confirm our hypothesis , we knocked-out both YAP1 and WWTR1 simultaneously by co-transducing the cell lines with lentivirus carrying blue fluorescence protein ( BFP ) -tagged YAP1 sgRNA and mCherry-tagged WWTR1 sgRNA ( Figure 5A ) . The co-competition assays showed that the population of BICR10 and HSC-2 with the double knockout of both YAP1 and WWTR1 depleted drastically ( Figure 5B ) compared to when each gene was knocked-out individually ( Figure 3C ) . This suggests that in this YAP1/WWTR1 compensable subset of cell lines , the paralogs can compensate for the function of one another to activate the downstream mechanisms required to maintain cell fitness . This was substantiated by quantitative-PCR ( qPCR ) of downstream targets CTGF and CYR61 where substantial down-regulation in double knockout cells were observed compared to when each gene is knocked-out individually ( Figure 5C ) . Next , we sought to determine whether the differential dependency on YAP1 and WWTR1 is also relevant in OSCC tumors . We first derived the gene expression signatures representing the three groups using differentially expressed gene ( DEG ) analysis based on the OSCC cell lines with validated dependency ( Figure 6—figure supplement 1A–B and Supplementary file 6 ) . Using the ‘YAP1 dependency signature score’ , ‘WWTR1 dependency signature score’ and ‘Compensable signature score’ [see materials and methods] , OSCC cell lines were clustered into three broad groups based on their dependency on YAP1 , or WWTR1 ( Figure 6A ) . Using the same algorithm , we then computed the dependency signature score for each of the 315 OSCC tumors from the TCGA HNSCC cohort . From the heatmap and clustering analysis based on their dependency signatures , the three groups were also observed among the OSCC tumors ( Figure 6B ) . To define representative ‘core’ samples , we found 41 OSCC tumors ( 13% ) with high YAP1 dependency signature score ( >0 . 5 ) ; 30 OSCC ( 9 . 5% ) with high WWTR1 dependency signature score ( >0 . 5 ) and 34 OSCC ( 11% ) with high Compensable signature score ( >0 . 5 ) . Using these core OSCC samples and cell lines with validated dependency , we then used GSEA to identify hallmark pathways that are enriched in each of these groups ( Figure 6C and Supplementary file 7 ) . YAP1-dependent cell lines and tumors showed enrichment in hallmarks related to cell cycle , such as the E2F targets , G2M checkpoint , MYC targets , and DNA repair pathways ( Figure 6C ) . This is consistent with previous reports that have demonstrated that the transcription factors E2F and MYC are critical downstream regulators of YAP/TEAD-mediated activation of cell cycle genes ( Kapoor et al . , 2014; Pattschull et al . , 2019 ) . While OSCC with high WWTR1 dependency signature score showed high expression of genes in several hallmarks related to immunity , such as the interferon responses , inflammatory responses and the complement pathway . This association is aligned with the recent findings that WWTR1 may play a role in immunity by upregulating PD-L1 expression ( Janse van Rensburg et al . , 2018 ) . On the other hand , cell lines and tumors with high Compensable signature score showed enrichment in several metabolism-related hallmarks , such as fatty acid metabolism and xenobiotic metabolism . Notably , all three YAP1/WWTR1 compensable cell lines ( BICR10 , HSC-2 and HSC-4 ) harbor PIK3CA mutation , and that alterations in the PI3K signaling pathway have been linked to multiple metabolic dysregulations in cancer ( Hao et al . , 2016 ) . Among the plethora of diverse functions for YAP1 and WWTR1 evidence for their critical roles in mediating immune response have recently emerged ( Geng et al . , 2017; Janse van Rensburg et al . , 2018; Pan et al . , 2019 ) . Specifically , WWTR1 but not YAP1 was shown to be essential for TH17 cell differentiation ( Geng et al . , 2017 ) , and constitutively active WWTR1 ( TAZ-S89A ) was shown to induce PD-L1 expression to a much greater extent than YAP1 ( YAP1-S127A ) ( Janse van Rensburg et al . , 2018 ) . Consistently , the results from the enrichment analysis of WWTR1-dependent OSCC were dominated by immune-related hallmarks . To investigate the association of YAP1/WWTR1 dependency with the immunity of OSCC , we mined the previously defined immune landscape of core samples from TCGA for comparison . Using the stromal and immune signature defined previously ( Yoshihara et al . , 2013 ) , ssGSEA enrichment scores revealed that OSCC with high WWTR1 dependency signature score showed significantly lower stromal signature , but higher immune signature , when compared with OSCC of high YAP1 dependency or Compensable signature scores ( Figure 6D ) . This suggests an enrichment of immune cell infiltration among OSCC with high WWTR1 dependency signature score . Next , we also assessed the other immune-related scores from previous work ( Thorsson et al . , 2018 ) and confirmed that these OSCC were associated with significantly enriched interferon-gamma ( IFNγ ) response signatures and reduced transforming growth factor-beta ( TGFβ ) response ( Figure 6E and Figure 6—figure supplement 2A–B ) . Consistent with that , they also possess significantly higher cytolytic T cells ( CD8 ) score than the other two groups ( Figure 6E ) . This interesting observation led us to postulate that OSCC cancers with a gene expression signature associated with WWTR1 dependency might be more vulnerable to checkpoint blockade . To extend this hypothesis , we examined the level of several predictive biomarkers for immune checkpoint inhibitor response that had been tested and validated in the clinical setting ( Ayers et al . , 2017; Cristescu et al . , 2018 ) . Interestingly , the PD-L1 mRNA expression and the 18-gene T-cell inflamed gene expression profile ( GEP ) enrichment scores of the OSCC with high WWTR1 dependency signature score were significantly higher than the other two groups ( Figure 6F–G ) . They were also associated with higher tumor mutational burden ( TMB ) albeit not statistically significant when compared with the other two groups ( Figure 6—figure supplement 2C ) . To ensure that the association with immune signatures seen is specific to WWTR1 but not caused by other co-amplified genes in 3q25-28 ( such as PIK3CA , TP63 , and SOX2 ) , we examined the changes in gene expression level of PD-L1 upon depletion of WWTR1 and other co-amplified genes . qPCR results revealed strongest suppression of PD-L1 gene expression upon WWTR1 knockout , but remain largely unchanged when the other genes were knockout ( PIK3CA , TP63 , and SOX2 ) ( Figure 6—figure supplement 3A ) . We also examined a microarray dataset from Hiemer et al . , 2015 and showed that a significant correlation of WWTR1 gene expression with PD-L1 expression was observed , but no correlation in gene expression was seen between PIK3CA , TP63 or SOX2 and PD-L1 ( Figure 6—figure supplement 3B ) . Together , these results suggested that OSCC resembling the gene signature with WWTR1-dependent cell lines may be associated with better respond to immunotherapy .
New therapeutic targets are urgently needed for the development of OSCC treatment . However , genomics studies have shown that oncogenic mutations in OSCC are largely limited to PIK3CA and HRAS and even in these , mutations are only found in a small subset of patients . Therefore , determining the oncogenic pathways and specific therapeutic targets have not been straight forward for this disease . We performed genome-wide CRISPR-Cas9 screens in a unique collection of OSCC cell lines derived from patients with diverse risk habits to identify genetic vulnerabilities that can serve as a basis for further therapeutic development for OSCC . Adding to existing genomics datasets in OSCC , this approach identifies fitness genes required for the survival of cancer cells where targeting these will result in the killing of these cells . The majority of the cell lines used in this study were from Asian patients where OSCC is more prevalent , and where causative factors include betel quid chewing in addition to smoking that is more often found in Caucasian patients ( Cheong et al . , 2017; Kumar et al . , 2016 ) . The molecular drivers of these patients are under-characterized as genomics data on Asian patients remain limited ( India Project Team of the International Cancer Genome Consortium , 2013; Su et al . , 2017 ) . Here , we identified 918 fitness genes in OSCC . Pathway enrichment analysis revealed that these genes were highly associated with diverse cancers pathways , confirming the robustness of our screen and pipeline in identifying targetable genetic dependencies . These included known candidate genes that are already being investigated in clinical trials for OSCC or other cancers such as CDK6 , PIK3CA , and FGFR1 , as well as novel genes that are yet to be explored as therapeutic targets , including those within the oxidative stress pathway ( KEAP1 , NFE2L2 ) . Notably , we showed that about 5% ( 45/918 ) of these genes are highly tractable with approved drugs , or have drugs that are in late-stage of clinical testing , demonstrating that these screens could help to prioritize drugs that could be repurposed for OSCC treatment . We compared our screens with previous genome-wide RNAi screens and found that candidate genes such as those relating to the cell cycle ( CKAP5 , KPNB1 , RAN , TPX2 , and KIF11 ) were also identified in our dataset ( Martens-de Kemp et al . , 2013 ) . However , due to stringent filtering of core-fitness genes ( Behan et al . , 2019; Hart et al . , 2014; Hart et al . , 2017; Meyers et al . , 2017 ) , these were no longer within the list of OSCC non-core fitness genes . Notably , the unique inclusion of several Asian OSCC models known to be associated with betel-quid chewing in our screen enabled us to identify that the NF-kB signaling pathway is among one of the most significantly enriched dependencies among betel-quid associated OSCC compared to cancers not associated with this risk habit . This finding is in line with the past studies reporting direct activation of the NF-kB signaling , upon treatment of OSCC cell lines with the extract of areca nut which is the main component in the betel quid ( Chiang et al . , 2008; Lin et al . , 2005 ) . By contrast , members of this signaling pathway ( TRIM25 , NFKB2 , TNFAIP3 ) did not seem to be fitness genes in the other cell lines not associated with betel quid chewing , nor those from the Project Score or DepMap ( Behan et al . , 2019; Meyers et al . , 2017 ) . Members of the NF-kB signaling pathways have been proposed as therapeutic targets for inflammatory diseases and cancers , with various types of inhibitors being developed . In the event of the development of NF-kB inhibitosr with promising clinical utililty in the future , it would be of interest to investigate if increase efficacy of this inhibitor would be seen among the betel-quid-associated OSCC . Appreciating this differences in dependencies in Asian OSCC that remains undiscovered in existing large genetic screens could have significant implications , especially when employing precision medicine in the different populations . We acknowledge , however , that further in-depth investigations including a larger sample size and further representative models , are needed in order to confirm our findings and to inform the basis of developing potential targeted therapy against the unique vulnerabilities among the Asian OSCC . Driver mutations are often expected to be robust biomarkers in precision medicine . However , upon analysing the dependency profiles on cancer genes that are commonly mutated in HNSCC and those with driver mutations among our 21 OSCC cell lines , we show that with the exception of some genes with driver mutations leading to oncogene addiction ( PIK3CA , HRAS and NFE2L2 ) , most other driver mutations did not confer preferential gene dependency and their value as a drug target remains unclear . Given the propensity of copy number alterations in driving OSCC , a ‘C’ class tumor ( The Cancer Genome Atlas Network , 2015; Ciriello et al . , 2013 ) , we investigated the commonly amplified genomic regions to look for functionally important candidate genes . One of the pathways that was significantly enriched was the Hippo pathway . Within this pathway , we focused our analyses on YAP1 and WWTR1 , two paralogs that show differential dependency pattern in our 21 OSCC cell lines . Copy number amplifications of 11q22 and 3q25 ( where YAP1 and WWTR1 are located , respectively ) are common events reported in OSCC , often in mutual exclusive manner ( Campbell et al . , 2018; Wang et al . , 2018 ) . These two genes are the major effectors negatively regulated by the Hippo pathway that is increasingly reported to play multiple roles in carcinogenesis , as reviewed comprehensively in recent years ( Dey et al . , 2020; Santos-de-Frutos et al . , 2019 ) . However , the majority of the studies focused on either one of the paralogs or had assumed similar functions between paralogs ( Santos-de-Frutos et al . , 2019; Zanconato et al . , 2016 ) . Emerging evidence demonstrate that YAP1 and WWTR1 have distinct roles where they partner with different transcription factors , drive different downstream effectors and also modulate the tumor microenvironment distinctively ( Callus et al . , 2019; Janse van Rensburg et al . , 2018; Kaan et al . , 2017; Plouffe et al . , 2018 ) . Our study revealed the intricate dominance of dependency on either one of the paralogs , despite the other not being deleted or loss . More intriguingly , the subset of lines thought not to be dependent on either YAP1 or WWTR1 , were actually lines where the remaining paralog is able to compensate for the lost paralog , therefore , enabling the cells to continue surviving . In the context of OSCC/HNSCC , studies demonstrating the distinct roles and regulatory mechanism of YAP1 and WWTR1 are emerging . Analysis of genome-wide transcriptional changes upon knockdown of YAP1 or WWTR1 in OSCC showed that YAP1 had a more prominent role in transcriptional regulation ( Hiemer et al . , 2015 ) . Using a tongue orthotopic mouse model with the deletion of MOB1A/B , Omori et al . , 2020 provided strong evidence that YAP1 acted as a strong driver in OSCC tumor initiation and progression , whereby WWTR1 did not seems to play an equivalent role ( Omori et al . , 2020 ) . The differences between YAP1 and WWTR1 can also be further exemplified in terms of their interaction with upstream/downstream pathways , that would involve other frequently co-amplified genes such as PIK3CA , TP63 , and SOX2 . The co-occurrence of amplifications in these genes that are part of the extended signaling network of the Hippo pathway underscores the critical role of the Hippo pathway in driving the OSCC tumorigenesis . Overexpression of PIK3CA was shown to be correlated with YAP1 activation and associated with poor clinical outcome ( García-Escudero et al . , 2018 ) . Further , the activation of the PI3K through mitogenic signaling inhibits the Hippo pathway leading to YAP1 activation and cell growth ( Fan et al . , 2013 ) . On the other hand , WWTR1 was shown to act upstream of SOX2 , facilitating stemnesses in HNSCC ( Li et al . , 2019 ) . Intriguingly , while WWTR1 knockdown led to reduction of SOX2 mRNA and protein expression , this was not seen when YAP1 was knockdown in the HNSCC cells ( Cal27 and Fadu ) ( Li et al . , 2019; Huang et al . , 2017 ) . In support of that , an inverse relationship between the expression levels of YAP1 and ΔNp63 were reported in lung SCC tumor samples ( Huang et al . , 2017 ) . However , this was not seen in HNSCC tumor samples and cell lines ( Ge et al . , 2011 ) , suggesting a cancer/context-specific regulatory mechanism might be in place . Consistent with the oncogenic roles of YAP1 reported in OSCC ( Hiemer et al . , 2015; Omori et al . , 2020 ) , another study showed that p63 , together with the co-expressing chromotin remodeling factor , ACTL6A , can drive YAP1 activation , suppressing differentiation and promoting cell proliferation in HNSCC ( Saladi et al . , 2017 ) . Similar observations between WWTR1 and TP63 have not been reported thus far . Hence , understanding the context in which the OSCC lines can be either YAP1-dependent/WWTR1-dependent or having compensable YAP1/WWTR1 is important , as YAP1 and WWTR1-dependency appear to be associated with the enrichment of distinct pathways . Current inhibitors of YAP1 such as verteporfin and CA3 that blocks its interactions with the TEAD transcription factors also targets WWTR1 ( Song et al . , 2018; Zhang et al . , 2015 ) ; therefore , this data underscores the need to develop more specific inhibitors to prevent the targeting of many different downstream pathways . Intriguingly , we also observed enrichment of PIK3CA mutant ( p=0 . 0003 ) among OSCC lines that are compensable for YAP1 or WWTR1 . As recent studies have provided evidence that YAP1 and WWTR1 could mediate mutant PIK3CA-induced tumorigenesis ( Zhao et al . , 2018 ) andother studies also suggested crosstalk between these Hippo pathway effectors with the PI3K-Akt pathway ( García-Escudero et al . , 2018 ) , confirmatory and mechanistic studies will be needed to delineate why YAP1 and WWTR1 function can be compensated in these PIK3CA-mutated cell lines , while distinct dependencies on either paralog are observed in PIK3CA wild-type lines . The functional loss of mutated FAT1 has also been reported to be associated with YAP1 activation in head and neck cancer ( Martin et al . , 2018 ) , however , no enrichment of FAT1 mutation was seen among the YAP1-dependent nor WWTR1-dependent models in this study . We also provided tissue-relevant insights of our findings by including analysis of the OSCC tumors from TCGA . Since YAP1 and WWTR1 are transcription co-factors that could regulate a plethora of gene transcriptions , we devised an analysis workflow utilizing the DEGs among the three subsets of OSCC models . We have identified 105 OSCC tumors that show highly similar gene expression signature as the cell lines , which are predicted to share the same dependency pattern . Similarly , comparison of the OSCC tumors based on their gene signatures revealed significant differences in terms of their enriched gene sets and immune signatures . As checkpoint blockade is approved for the treatment of recurrent and metastatic OSCC , understanding how YAP1 and WWTR1 influence the immune microenvironment could provide clues on the combination therapies that could increase the subset of patients responding checkpoint inhibitors . In particular , OSCC with high WWTR1 dependency signature score are significantly associated with various biomarkers that were predicted to show good response toward checkpoint inhibitors . This finding is consistent with the recent discovery that constitutively active WWTR1 induces PD-L1 expression , to a greater extent than YAP1 ( Janse van Rensburg et al . , 2018 ) , and that tumors with YAP1 amplifications have low T-cell infiltration ( Saloura et al . , 2019 ) . These observations have clinical implication as anti-PD1 is an approved therapy for HNSCC and therefore , further investigation and validation will be needed to confirm this observation and its clinical impact . While many companies are developing novel inhibitors targeting YAP1-TEAD transcriptional activity , which should be effective against all other OSCC , combination with checkpoint inhibitor could be considered for those OSCC with WWTR1 dependency signatures . We acknowledge , however , that whilst the dependency observed in cell lines could be recapitulated in OSCC , OSCC tissues would be much more heterogenous and could harbor specific genetic abrogations that could be the dominant driver of tumorigenesis . Therefore , further validation of the association between WWTR1-dependency signature and response to checkpoint inhibitors should be validated particularly in the context of clinical trials involving checkpoint inhibitors . The roles of the Hippo signaling pathway and its effectors YAP1 and WWTR1 in cancer immunity remains unclear . The inactivation of the Hippo pathway through the loss of LATS1/2 was reported to cause the induction of anti-tumor immune response and inhibition of HNSCC tumor growth , via the hyperactivation of YAP1/WWTR1 ( Barth et al . , 2013; Moroishi et al . , 2016 ) . This demonstrates that components of the Hippo signaling pathway could also modulate the host tumor microenvironment in addition to what we have demonstrsated in cancer cells . The design and models used in our study have not been set up to examine this where the inherent limitation of using cell lines do not consider the components of the tumor microenvironment in the in vitro screening . Nonetheless , our findings provided a novel insight linking the intricate dependency on YAP1 and WWTR1 with differential state of the immune microenvironment in OSCC which warrants further investigation with the use of immune-competent mouse models , before further clinical evaluation can be made . In summary , our study is the first large-scale CRISPR-Cas9 screen and focused analysis conducted on large panels of Asian derived OSCC cell lines and provided a cancer-specific overview of the fitness genes landscape , affording opportunities for further therapeutic targets development . The ability to scrutinize the functional genomics of these fitness genes/pathways to a greater detail was also exemplified in this study .
Fourteen OSCC cell lines ( referred to as the ORL- series ) were derived spontaneously from surgically resected OSCC tissue specimens in Cancer Research Malaysia . Briefly , tissues were collected in α-MEM containing 20% ( v/v ) FBS , 200 iu/l penicillin , 200 μg/ml streptomycin and 0 . 1 μg/ml of fungizone . Subsequently , tissues were washed in absolute ethanol for 20–30 s and then washed twice with phosphate-buffered saline ( PBS ) under sterile conditions . Tissues were minced , washed twice in culture media and re-suspended in α-MEM containing 20% ( v/v ) FBS , 200 iu/l penicillin , 200 μg/ml streptomycin , 0 . 4 ng/ml EGF , 2 μg/ml hydrocortisone and 2 mM L-glutamine , and seeded into tissue culture dishes . Cultures were continuously maintained for more than 100 population doublings ( Fadlullah et al . , 2016 ) . HSC-2 , HSC-4 , and SCC-9 were isolated from squamous cell carcinoma of various oral regions , by which the surgically excised tumors were minced and disaggregated to single cells . Epithelial cells proliferated from the explants were then sub-cultured continuously ( Momose et al . , 1989; Rheinwald and Beckett , 1981 ) . No special immortalization methods were detailed for BICR10 , PE/CA-PJ15 , Ho-1-U-1 and SAS ( Edington et al . , 1995 , Berndt et al . , 1997 , Miyauchi et al . , 1985 , Takahashi et al . , 1989 ) nor for all other cell lines used in this study . All these OSCC cells lines were maintained in Dulbecco’s Modified Eagle’s Medium ( DMEM ) /Nutrient Mixture F-12 medium ( Gibco ) supplemented with 10% ( v/v ) heat inactivated fetal bovine serum ( Gibco ) , and 100 IU Penicillin/Streptomycin ( Gibco ) . All lines were incubated in a humidified atmosphere of 5% CO2 at 37°C . The lines were authenticated by STR profiling using Promega PowerPlex16HS Assay ( Promega , Wisconsin , United States ) , with the data giving more than 80% match to the respective donor or reference as deposited in the databases of cell line resources ( such as ATCC , DSMZ , JCRB ) . Cell lines were routinely tested for the presence of mycoplasma with MycoAlert mycoplasma detection kit ( Lonza , Basel , Switzerland ) . Only mycoplasma-free cell lines were used in all experimentation . HEK293 cells were transfected using jetPRIME transfection reagent ( Polyplus Transfection ) according to the manufacturer’s instructions . Briefly , transfection complex consisting of jetPRIME buffer , jetPRIME reagent , vector of interest , pMD2 . G and psPAX2 were prepared and mixed with Opti-MEM ( Gibco ) . Next , the transfection complex medium was added to HEK293 cells with 90% confluency . After overnight incubation at 37°C , 5% CO2 , transfection complex medium was replaced with fresh DMEM high glucose ( Gibco ) complete medium . Medium containing virus was collected at 48- and 72 hr post-transfection and filtered using PVDF 0 . 45 μm syringe filter . To perform virus transduction , selected cell lines were transduced with lentivirus containing the vector of interest , in the presence of 8 μg/ml polybrene . After overnight incubation , medium containing lentivirus was replaced with fresh DMEM/F12 complete medium . Cells were incubated for 48 hr and harvested to evaluate the transduction efficiency via flow cytometry analysis with BD LSR Fortessa X-20 cell analyser ( BD Biosciences ) . Gating strategy for flow cytometry analysis of transduced cells carrying fluorescence marker is exemplified in Supplementary file 8 . Selected cell lines were transduced with lentivirus containing the pKLV2-EF1α-Cas9Bsd-W ( Addgene plasmid # 68343 , gift from Kosuke Yusa ) . Cells stably expressing Cas9 enzyme was established via blasticidin selection 3 days post-transduction . Cas9 enzyme cutting efficiency was routinely checked via lentivirus transduction of the reporter plasmid pKLV2-U6gRNA5 ( gGFP ) -PGKmCherry2AGFP-W ( Addgene plasmid # 67982 , gift from Kosuke Yusa ) . The efficiency of Cas9 cutting activity was accessed using a reporter plasmid and analyzed using flow cytometry analysis . The screening will only be conducted on those cell lines with >80% Cas9 cutting activity , as indicated by the efficiency of GFP knockout in cell lines transduced with the reporter plasmid . The Human Improved Genome-wide Knockout CRISPR Library v1 ( Addgene plasmid #67989 , gift from Kosuke Yusa ) containing 90 , 709 gRNAs targeting a total number of 18 , 010 protein-coding genes was used in the genome-wide CRISPR-Cas9 screening ( Tzelepis et al . , 2016 ) . For each of the OSCC cell lines , a total of 60 million Cas9-expressing cells were transduced with the CRISPR library lentivirus at a multiplicity of infection ( MOI ) of 0 . 3 . Polybrene at 8 μg/ml was added to increase transduction efficiency . All screenings were performed in triplicates . Library representation was evaluated by the percentage of BFP-expressing cells , determined using flow cytometry on day 4 post-transduction . The library representation was minimally kept at 100X coverage of the library , equivalent to 30 million cells total , or 10 million cells expressing BFP before proceeding to puromycin selection ( 2 . 0 μg/ml ) for 3–4 days , to select for successfully transduced cells . Following complete selection with puromycin , a minimum of 75 million cells were maintained throughout the 18 days screen . BFP expression was monitored to ensure selection was adequate . On day 18 post-transduction , 60 million cells were pelleted down for genomic DNA extraction . Genomic DNA was extracted from 60 million post-CRISPR screened cells using QIAGEN blood and cell culture DNA Maxi kit ( Qiagen ) , according to manufacturer’s instruction . Extracted DNA was quantified using Qubit 2 . 0 fluorometer ( Thermo Fisher Scientific ) . To prepare and generate Illumina libraries for deep sequencing , amplification of sgRNA was performed using Q5 Hot Start High-Fidelity 2 × Master Mix and forward/reverse primers pair ( gLibrary-HiSeq_50bp-SE-U1 F and R ) in 50 μl reaction , as previously described4 . For the CRISPR library v1 plasmid , 10 independent PCR reactions were set up using 2 . 0 μg of the plasmid . While for the CRISPR-screen cell lines , 2 . 0 μg of genomic DNA harvested from day-18 post-transduction was used in each of the 36 independent PCR reactions . The PCR conditions were as follows: 98°C for 30 s , 26–28 cycles of 98°C for 10 s , 61°C for 15 s and 72°C for 20 s , and the final extension , 72°C for 2 min . The PCR products were analyzed on 2% agarose gel and additional PCR cycles were added if necessary . About 5 μl of PCR products were pooled from all 36 reactions and QIAquick PCR purification kit ( QIAGEN ) was used to purify the amplified gRNA . Concentration of purified PCR products were quantified with Qubit dsDNA broad-range ( BR ) assay kit ( Thermo Fisher Scientific ) , using the Qubit 2 . 0 fluorometer ( Thermo Fisher Scientific ) . PCR enrichment was then carried out using 200 pg of purified PCR products with 2x KAPA HotStart ReadyMix . 1 μl of forward P5 fusion primer ( PE 1 . 0 p5 Top_PE_C ) and 1 μl of different reverse primers ( indexed iPCRTags ) were used . The PCR conditions were as follows: 98°C for 30 s , 10–12 cycles of 98°C for 10 s , 66°C for 15 s and 72°C for 20 s , and the final extension , 72°C for 5 min . Finally , SPRISelect beads ( Beckman Coulter ) were used to purify the PCR products at a PCR-product-to-bead ratio of 1:0 . 8 . Purified libraries were dissolved in 30 μl nuclease-free water and quantified using Agilent High Sensitivity DNA kit ( Agilent Technologies ) on Agilent 2100 Bioanalyzer ( Agilent Technologies ) . Purified libraries of the triplicate screens tagged with different iPCR tags were pooled and sequenced at about 300x coverage on Illumina HiSeq 2500 with 19 bp single-end ( SE ) deep sequencing at the Wellcome Sanger Institute ( WSI ) . About 30–40 million reads were obtained for each of the three replicates . Sequences of the Read one sequencing primer ( U6-Illumina-seq2 ) and index sequencing primer can be found in Supplementary file 9 . sgRNA raw counts in each triplicate of the CRISPR screen were generated using the in-house script developed at WSI , as previously described ( Behan et al . , 2019; Iorio et al . , 2018 ) . The CRISPRcleanR tool was downloaded as an R package ( https://github . com/francescojm/CRISPRcleanR ) and used for pre-processing of the sgRNA raw counts ( Iorio et al . , 2018 ) . This tool allows for an unsupervised correction for copy number amplification bias and other gene-independent responses when subjected to CRISPR-Cas9 targeting , hence reducing false-positive call for essential genes ( Iorio et al . , 2018 ) . Briefly , the ccr . NormfoldChanges function was used to compute the median-ratio normalization of raw counts and log2 fold-changes for all sgRNAs , averaging from triplicates . The built-in KY_library_v1 . 0 was used for library annotation and sgRNAs with read counts less than 30 in the plasmid were excluded . Then , the ccr . logFCs2chromPos and ccr . GWclean function of the CRISPRcleanR were used to perform genome mapping and sorting of the sgRNAs , followed by the correction of the gene-independent responses to compute corrected log-fold changes . As part of the quality assessment , Pearson correlation test was used to compare sgRNA raw counts between replicates of the same cell line , while precision-recall was assessed as previously described , using sets of known essential and non-essential genes ( Behan et al . , 2019; Supplementary file 10 ) . The corrected gene-level log-fold changes were quantile normalized and corrected for batch effect using ComBat ( Leek et al . , 2012 ) , we refer this CRISPRcleanR corrected , quantile normalized and ComBat corrected log-fold changes as the ‘CRISPR score’ , with a negative value indicating the extent of depletion of sgRNA counts targeting the gene when compared with the initial plasmid library . The function ccr . correctCounts used inverse-transformation method to generate CRISPRcleanR-corrected counts , which is used as input files for MAGeCK analysis ( Iorio et al . , 2018; Li et al . , 2014 ) . For each of the 21 OSCC cell lines , MAGeCK analysis was performed using default parameters , except that normalization is set to ‘none’ , as the input corrected counts had already been normalized using CRISPRcleanR . A false discovery rate cut-off of 5% ( FDR ≤ 0 . 05 ) was applied to identify the significantly depleted genes in each cell line , defined here as MAGeCK hits . To remove potential false-positive hits , RNA-seq expression data of each of the 21 cell lines were utilized to filter out MAGeCK hits with negligibly low or no reported reads ( Fragments Per Kilobase of transcript per Million mapped reads ( FPKM ) <0 . 5 ) . Next , in order to identify and prioritize genes that can be safely targeted for the treatment of OSCC , we curated a list of core fitness essential genes from four different sources and used it to further filter the list of MAGeCK hits from each cell line . The first two sources were the core essential genes ( CEG ) list , published by Hart et al . , 2014 and the subsequent updated CEG2 list , published in 2017 ( Hart et al . , 2017 ) . The third source is the ‘common-essential genes’ , downloaded from Broad’s Institute Cancer Dependency Map database ( 18Q3 release ) ( Meyers et al . , 2017 ) . We also utilized the list of pan-cancer core-fitness genes compiled by Project Score of Cancer Dependency Map at WSI ( Behan et al . , 2019 ) . The full list of all genes from these four sources was tabulated in Supplementary file 2 . All 21 OSCC cell lines were subjected to WES at the WSI using HiSeq2500 . WES data were processed using an established pipeline as previously described to identify driver mutations ( Iorio et al . , 2016 ) . A total of 43 genes with driver mutation in at least one cell line were identified ( Supplementary file 5 ) . KEGG pathway enrichment analysis was performed using the over-representation analysis function at the ConsensusPathDB ( http://ConsensusPathDB . org ) ( Kamburov et al . , 2013 ) . A threshold of minimum two genes overlapping with the gene set of a given pathway and p-value cut-off of 0 . 05 were applied . Enriched pathways were ranked by q-value . OSCC models with known betel quid chewing as the only risk habit were included in this analysis – ( n = 7; ORL-115 , ORL-136 , ORL-174 , ORL-195 , ORL-204 , ORL-207 and ORL-214 ) , to compare with the other OSCCs not associated with betel quid chewing ( n = 14; ORL-48 , ORL-150 , ORL-153 , ORL-156 , ORL-166 , ORL-188 , ORL-215 , BICR10 , Ho-1-u-1 , HSC-2 , HSC-4 , SAS , SCC-9 , PE/CA-PJ15 ) . Venn diagram was used to depict the number of unique and overlapping fitness genes that are found between those OSCCs with or without association with betel quid chewing . To assess the tractability of the 918 non-core fitness genes , we utilized the genome-wide target tractability assessment pipeline as previously described ( Behan et al . , 2019; Brown et al . , 2018 ) . Based on assessment for small molecules tractability , essential genes were assigned into tractability bucket 1 to 10 , with decreasing tractability . Next , essential genes in each tractability group were further classified into protein classes using the PANTHER database online tool ( http://www . pantherdb . org/ ) ( Mi et al . , 2013 ) . To validate the results obtained from the screen , individual targeted genetic knockouts were generated using CRISPR/Cas9 and plasmid expressing sgRNAs targeting the gene of interest . Two sgRNA sequences were used for each target gene , one was selected from the CRISPR library v1 while another sequence was designed using the Genetic Perturbation Platform ( GPP ) web portal ( https://portals . broadinstitute . org/gpp/public/analysis-tools/sgrna-design ) . List of sgRNA used and their sequences can be found in Supplementary file 11 . pKLV2-U6gRNA5 ( BbsI ) -PGKpuro2ABFP-W and pKLV2-U6gRNA5 ( BbsI ) PGKpuro2AmCherry-W were linearized using BbsI enzyme ( NEB R0539S ) and the concentration was adjusted to 20 ng/μL . Target oligos were phosphorylated and annealed using T4 PNK ( NEB M0201 ) . The thermocycler condition used are as follows: 37°C for 30 min , 95°C for 5 min , followed by a ramp down to 25°C at 0 . 1°C /s . The annealed oligos were next diluted twice for prior to ligation: 1st dilution = 139 μL EB buffer + 2 μL of 10 μM double-stranded oligos; 2nd dilution = 57 μL EB buffer + 3 μL of 1st dilution . Following that , overnight ligation at 4°C was carried out using T4 ligase ( NEB M0202S ) and 10X ligase buffer ( NEB M0202S ) . The ligation products were then transformed into DH5α chemically- generated competent cells and plated onto Luria Broth ( LB ) agar plates containing 100 μg/ml Ampicillin . The plasmids were then extracted using QIAprep Spin Miniprep Kit ( QIAGEN ) and the sgRNA sequences were verified by Sanger sequencing prior to use . The relative growth rate of sgRNA-transduced and non-transduced cells was compared using co-competiton assay , as described previously ( Behan et al . , 2019; Tzelepis et al . , 2016 ) . Briefly , in order to achieve single gene-specific knockout , the Cas9-expressing cell lines were transduced at 30–70% transduction efficiency , with lentivirus carrying gene-specific sgRNA in pKLV2-U6gRNA5 ( BbsI ) -PGKpuro2ABFP-W . Using flow cytometry , the percentage of BFP-positive sgRNA-transduced cells was measured between day 4 and day 18 post-transduction . The results obtained from days 8 , 11 , 15 to 18 were normalized to the percentage of BFP-positive transduced cells on day 4 or 6 to investigate the relative growth changes of the transduced population following gene depletion . For each target gene ( YAP1 and WWTR1 ) , two different sgRNA were used , one from the Kosuke Yusa’s CRISPR Library v1 ( ‘Y1K’ – sgRNA targeting YAP1; ‘W1K – sgRNA targeting WWTR1’ ) and another independently designed sgRNA using Broad’s sgRNA-designer tool ( ‘Y2B’ – sgRNA targeting YAP1; ‘W2B’ – sgRNA targeting WWTR1 ) . sgRNA targeting a core fitness gene , Polo-like kinase 1 ( PLK1 ) was included as a positive control , choline acetyltransferase ( CHAT ) -targeting sgRNA was used as a non-fitness gene negative control and a non-targeting ( NT ) sgRNA w also included . To achieve double gene knockout , sgRNAs targeting two different genes were cloned into either one of the plasmids tagged with BFP or mCherry ( pKLV2-U6gRNA5 ( BbsI ) -PGKpuro2ABFP-W or pKLV2-U6gRNA5 ( BbsI ) -PGKpuro2AmCherry-W ) . Thereafter , the changes in the BFP- and mCherry double-positive cell population was measured as mentioned above . To determine the baseline protein expression level , OSCC parental cell lines were seeded in 100 mm3 dish and cultured until they reached 70–80% confluency . To assess the differential expression of the protein of interest after gene knockdown , Cas9-expressing cells were transduced with the target sgRNA at above 90% transduction efficiency . Next day , transduced cells were selected using 2 μg/ml of Puromycin . Day four post-transduction , percentage of BFP-expressing cells were determined using flow cytometry and total cell lysates ( TCL ) were extracted with RIPA buffer ( 50 mM Tris pH8 , 1% ( v/v ) NP-40 , 0 . 5% ( w/v ) sodium deoxycholate , 0 . 1% ( w/v ) SDS , 150 mM NaCl ) supplemented with Halt Protease and Phosphatase Inhibitor ( PI ) Cocktail ( Pierce Biotechnology ) on ice . TCL was collected by centrifugation and quantified using the BCA method ( Thermo Fisher Scientific ) . About 20 µg of the TCL was resolved on SDS-PAGE gel and proteins were transferred onto PVDF membranes ( Millipore ) . Membranes were blocked with 5% ( w/v ) milk in TBST ( 0 . 1% ( v/v ) Tween 20 ) and probed with primary antibodies ( 1:1000 dilution in 1% ( w/v ) BSA ) overnight at 4°C . Horseradish peroxidase ( HRP ) -conjugated secondary antibodies ( 1:10 , 000 dilution in 5% milk ) were probed for one hour at room temperature . For signal development , WesternBright Quantum HRP substrate ( Advansta Inc ) was used and visualized using the FluorChem HD2 imaging systems ( Alpha Innotech ) . To normalize for loading , the blots were re-probed with an anti-tubulin monoclonal antibody ( 1:1000 dilution in 1% BSA ) and processed as described above . List of primary and secondary antibodies used is found in Supplementary file 12 . Uncropped western blot images can be found in Supplementary file 13 . Cas9-expressing cells were transduced with selected sgRNA to achieve single gene or double gene knockout , as described above . On day 4 post-transduction , total RNA was extracted using TRIzol Reagent ( Thermo Fisher Scientific ) . Total RNA ( 1 μg ) was used for reverse transcription to complementary DNA ( cDNA ) using high-capacity cDNA reverse transcription kit ( Applied Biosystems ) . Real-time quantitative PCR was performed using 1 μl of 5x diluted cDNA with PowerUp SYBR Green Master Mix and corresponding primers in 7500 Real-Time PCR System ( Applied Biosystem ) . All reactions were performed in technical triplicates and repeated twice . Cycling conditions used are as follows: 50°C for 2 min , 95°C for 2 min , 40 cycles of 95°C for 15 s , and 60°C for 1 min . A default melt curve stage was included to allow inspection of primer specificity . Ribosomal protein L13 ( RPL13 ) was used as an endogenous reference control for normalization . Sequences of all primers used can be found in Supplementary file 9 . On 4- or 6 days post-transduction of sgRNA-containing lentiviruses , 2000 cells were seeded into six-well plate . After a week , cells were fixed using ice-cold methanol followed by staining with crystal violet solution . The 3- ( 4 , 5-dimethylthiazol-2-yl ) −2 , 5-diphenyltetrazolium bromide ( MTT ) assay was used to access the effect of target gene knockdown/knockout on cell viability . Briefly , 4- or 6 days post-transduction with sgRNA , 2000 cells were seeded in triplicates in 96-well plate . 72 hr later , 50 μl MTT was added to each well and incubated for 4 hr at 37°C . After removing the media , 150 μl dimethyl sulfoxide was added to dissolve the formazan crystal and optical density was measured using Synergy H1M microplate reader ( BioTek Instruments , USA ) at 570 nm . On 4- or 6 days post-transduction of sgRNA-containing lentiviruses , 30 , 000 cells were seeded into 24-well plate and harvested after 72 hr . Cells were pelleted down and washed with PBS . For detection of apoptotic cells , the cell pellet was resuspended in 1x Annexin V buffer containing 2 . 5 μl of Annexin V solution ( BD Biosciences ) and 2 . 5 μl of propidium iodide and incubated for 15 min in the dark . The proportion of apoptotic cells were analyzed using the LSR Fortessa X-20 cell analyser ( BD Biosciences ) and FlowJo ( version 10 . 5 . 3 , BD Biosciences ) , considering all single- and double-stained cells as apoptotic cells . Gating strategy for detection of apoptotic cells using flow cytometry can be found in Supplementary file 8 . Three representative cell lines with validated dependency on YAP1 or WWTR1 were used to derive gene expression signatures . ‘YAP1-dependent’ – ORL-48 , ORL-204 , SAS; ‘WWTR1-dependent’ – ORL-214 , PE/CA-PJ15 , ORL-174; ‘Compensable’ – BICR10 , HSC-2 and HSC-4 . DEGs for each group were computed using the limma package ( Bioconductor ) on the iRAP-processed , ComBat corrected FPKM matrix for these nine cell lines . Non-overlapping DEGs with significant p-value threshold <0 . 01 and log fold change >2 were retained . The final list of DEGs is found in Supplementary file 6 . Gene expression data of the HNSCC cohort in TCGA in the form of RSEM was downloaded from cbioportal ( https://www . cbioportal . org/ ) ( Gao et al . , 2013 ) . Gene expressions of all DEGs were then extracted for the 315 OSCC samples . For all DEGs , Z-score was computed and a ‘dependency signature score’/ ‘compensable signature score’ was generated for each cell line/tumor sample , taking the difference between the average of all Z-score of upregulated DEGs and that of downregulated DEGs . For example , ‘YAP1-dependency signature score’ = ( average of all Z-score of upregulated DEGs among YAP1-dependent cell lines ) - ( average of all Z-score of downregulated DEGs among YAP1-dependent cell lines ) . Subsequently , cell lines and OSCC tumors were analyzed using hierarchical clustering and visualize with heatmap ( generated using Morpheus , Broad Institute: https://software . broadinstitute . org/morpheus ) using the ‘YAP1 dependency signature score’ , ‘WWTR1 dependency signature score’ and ‘Compensable signature score’ . Core OSCC samples with >0 . 5 dependency signature score in one of the three groups were identified . There was a total of 43 OSCC samples with high YAP1 dependency signature score; 31 with high WWTR1 dependency signature score and 30 with high Compensable signature score . Clinical and genomic data of these core OSCC samples were accessed from cBioPortal ( Gao et al . , 2013 ) . Gene set enrichment analyses ( GSEA ) for the three groups of core OSCC samples were performed using the Broad Institute’s Molecular Signatures Database ( MSigDB ) hallmark gene sets as reference database ( Liberzon et al . , 2015 ) . Signatures reflective of the immune landscape of these core OSCC samples were extracted from the supplementary table 1 of Thorsson et al . , 2018 . Single-sample GSEA ( ssGSEA ) was performed using the GenePatterns web-tool ( https://www . genepattern . org/ ) ( Reich et al . , 2006 ) with the 18-genes T-cell inflamed Gene Expression Profile ( GEP ) gene set , which was found to be predictive biomarkers for response to pembrolizumab in HNSCC clinical trial ( Ayers et al . , 2017; Cristescu et al . , 2018 ) . All statistical significance analyses were performed using unpaired parametric two-tailed t-test in GraphPad Prism ( version 8 , GraphPad Software Inc ) unless otherwise stated . Unpaired t-test with Welch’s correction ( Welch’s t-test ) was used for all analyses in Figure 6 and Figure 6—figure supplement 2 due to unequal sample size . For estimation of the Pearson correlation , the cor . test function in Rstudio ( version 1 . 2 . 1335 , Rstudio Inc ) was used . No unreported or custom code was used in this study . Open source softwares were used for data analysis and codes are available upon request . | Many types of cancer now have 'targeted treatments' , which specifically home in on genes cancer cells rely on for survival . But there are very few of these treatments available for the most common type of mouth cancer , oral squamous cell carcinoma , which around 350 , 000 people are diagnosed with each year . Designing targeted treatments relies on detailed knowledge of the genetic makeup of the cancer cells . But , little is known about which genes drive oral squamous cell carcinoma , especially among patients living in Asia , which is where over half of yearly cases are diagnosed . One way to resolve this is to use gene editing technology to find the genes that the cancer cells need to survive . Now , Chai et al . have used a gene editing tool known as CRISPR to examine 21 cell lines from patients diagnosed with oral squamous cell carcinoma . Most of these lines were from Asian patients , some of whom had a history of chewing betel quid which increases the risk of mouth cancer . By individually inactivating genes in these cell lines one by one , Chai et al . were able to identify 918 genes linked to the survival of the cancer cells . Some of these genes have already been associated with the spread of other types of cancer , whereas others are completely unique to oral squamous cell carcinoma . The screen also discovered that some cell lines could not survive without genes involved in a signalling pathway called Hippo , which is known to contribute to the progression of many other types of cancer . Uncovering the genes associated with oral squamous cell carcinoma opens the way for the development of new targeted treatments . Targeted therapies already exist for some of the genes identified in this study , and it may be possible to repurpose them as a treatment for this widespread mouth cancer . But , given that different cell lines relied on different genes to survive , the next step will be to identify which genes to inactivate in each patient . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics",
"cancer",
"biology"
] | 2020 | Genome-wide CRISPR screens of oral squamous cell carcinoma reveal fitness genes in the Hippo pathway |
Adeno-associated virus ( AAV ) -mediated gene therapies are rapidly advancing to the clinic , and AAV engineering has resulted in vectors with increased ability to deliver therapeutic genes . Although the choice of vector is critical , quantitative comparison of AAVs , especially in large animals , remains challenging . Here , we developed an efficient single-cell AAV engineering pipeline ( scAAVengr ) to simultaneously quantify and rank efficiency of competing AAV vectors across all cell types in the same animal . To demonstrate proof-of-concept for the scAAVengr workflow , we quantified – with cell-type resolution – the abilities of naturally occurring and newly engineered AAVs to mediate gene expression in primate retina following intravitreal injection . A top performing variant identified using this pipeline , K912 , was used to deliver SaCas9 and edit the rhodopsin gene in macaque retina , resulting in editing efficiency similar to infection rates detected by the scAAVengr workflow . scAAVengr was then used to identify top-performing AAV variants in mouse brain , heart , and liver following systemic injection . These results validate scAAVengr as a powerful method for development of AAV vectors . This work was supported by funding from the Ford Foundation , NEI/NIH , Research to Prevent Blindness , Foundation Fighting Blindness , UPMC Immune Transplant and Therapy Center , and the Van Sloun fund for canine genetic research .
Gene therapy is a rapidly developing approach for the treatment of inherited disease , and AAV is a leading viral vector candidate for safe and efficient delivery . A growing number of clinical trials are using AAV to treat diseases such as retinal degeneration , neurological disorders , and hemophilia , through gene replacement , genome editing , and optogenetics ( High and Roncarolo , 2019; Sahel and Dalkara , 2019 ) . And , with the recent FDA approval for treatment of Leber congenital amaurosis and spinal muscular atrophy , gene therapies are rapidly becoming a clinical reality . However , significant hurdles prevent the successful , widespread implementation of AAV-mediated gene therapies , including efficient gene delivery and immune response to viral vectors and gene products . Recent efforts to reengineer viral vectors have shown promise for addressing these issues , resulting in AAVs with improved abilities ( Li and Samulski , 2020 ) . The enhanced viral vectors produced by these high-throughput methods still require quantitative validation and comparison , however , currently a challenging and burdensome process . The selection of an optimal vector is essential to the success of the therapy . Sufficient gene expression is critical , while enhanced tropism and greater efficiency of gene delivery reduces the titer of vector required and decreases the likelihood of immune response . Quantitative comparisons of newly engineered vectors , including evaluation of transgene expression levels and cell-type tropism , have in the past required large numbers of animals , and therefore involved significant ethical and financial burden . Additionally , in primates , the large variability between animals , due to differences in anatomy and immune responses , has made comparisons between animals inaccurate . Here , we have developed a single cell RNA-seq AAV engineering ( scAAVengr ) pipeline for rapid , quantitative in vivo comparison of transgene expression from newly engineered AAV capsid variants across all different cell types in a tissue in parallel , and in the same animals . In this work , the scAAVengr pipeline was applied to primates , which are the most physiologically similar animal to humans , and are thus a critical preclinical model . Successful clinical translation of gene therapies depends on highly efficient vectors for human tissue , and vector performance in small animals often does not accurately predict efficiency in primates . For retinal gene therapy in particular , primates are essential , as existing AAV vectors infect the primate retina significantly less efficiently than in rodent retina ( Dalkara et al . , 2013 ) . Furthermore , primates are the only animal model that has a macula and foveal pit ( the region of the retina responsible for high acuity vision in humans ) , making them the most relevant translational model . Notably , the pattern of AAV expression also differs in foveal and in peripheral retina ( Dalkara et al . , 2013 ) , with highest expression in the foveola and in a perifoveal ring of retinal ganglion cells , and punctate expression near blood vessels in the periphery . The scAAVengr single-cell RNA-Seq pipeline allowed us to quantitatively evaluate the clinical potential of multiple lead candidates across all retinal cell types , in the foveal and peripheral retina , in a large animal model with eyes similar to humans . The scAAVengr pipeline can be applied to any species or tissue for which marker genes can be identified , however , as demonstrated here through screening performed in mouse brain , heart , and liver following systemic injections of pooled AAV library .
All procedures were performed in compliance with the ARVO statement for the Use of Animals in Ophthalmic and Vision Research , and for canine studies with approval by the University of Pennsylvania Institutional Animal Care and Use Committee ( IACUC # 803813 ) , and for the NHP and mouse studies with approval from the University of Pittsburgh Institutional Animal Care and Use Committee ( IACUC #18042326 ) . AAV vectors were produced in HEK293T cells ( ATCC ) , or 293AAV cells ( Cell Biolabs ) using a double ( for AAV2-7mer , LoopSwap , AAV2-ErrorProne and SCHEMA libraries ) or triple transfection method ( Grieger et al . , 2006 ) . Short tandem repeat profiling was done by ATTC Cell Line Authentication Service and all cell lines were checked for mycoplasma using Hoechst staining . Directed evolution libraries were packaged using an empirically determined molar ratio of plasmids in the packaging cell line , such that each AAV particle contained the genome encoding its own capsid ( Koerber et al . , 2006 ) . All recombinant AAVs were purified by iodixanol gradient ultracentrifugation , buffer exchanged and concentrated with Amicon Ultra-15 Centrifugal Filter Units ( #UFC8100 ) in DPBS and titered by quantitative PCR relative to a standard curve using ITR-binding primers or by using QuickTiter AAV Quantitation Kit ( Cell Biolabs ) . The relative titer of each variant was confirmed by Illumina MiSeq sequencing . We packaged AAV2 error prone ( Koerber et al . , 2006 ) , AAV2-7mer ( Müller et al . , 2003 ) , loop swap ( Koerber et al . , 2008 ) and SCHEMA libraries ( Ojala et al . , 2018 ) , which were pooled and injected intravitreally into both eyes of wild-type dogs ( Figure 1—figure supplement 1 ) . Intravitreal injections ( 150–250 µL ) were performed with a 30-gauge insulin syringe under general anesthesia delivering the viral solution in the mid-vitreous . Three weeks later , dogs were euthanized by intravenous injection of sodium pentobarbital , and both eyes were flattened by making relief cuts in the globe . Two mm punches of RPE were immediately collected from superior , inferior , temporal , and nasal regions of the retina , as well as from the area centralis , and flash frozen . DNA was extracted from samples using a Qiagen DNeasy blood and tissue kit , according to the manufacturer’s instructions , and AAV cap genes were recovered via PCR from retinal pigment epithelium ( RPE ) punches . AAV genomes were then repackaged and reinjected . Five rounds of selection were performed ( Supplementary file 1 ) , with error prone PCR done following the third round of selection to introduce additional diversity into the library . Following rounds of selection , the AAV2-7mer library was found to give rise to the majority of resulting variants . Each round of selection from the AAV2-7mer was then subjected to deep sequencing in order to analyze the dynamics of each individual variant and overall convergence of the library . A ~ 75–85 base pair region containing the 7mer insertion was PCR amplified from harvested DNA . Primers included Illumina adapter sequences containing unique barcodes to allow for multiplexing of amplicons from multiple rounds of selection ( Supplementary file 4 ) . PCR amplicons were purified and sequenced with a 100-cycle single-read run on an Illumina HiSeq 2 , 500 . DNA sequences were translated into amino acid sequences , and the number of reads containing unique 7mer insert sequences were counted . Read counts were normalized by the total number of reads in the run . Pandas was used to create plots . Best performing variants were chosen as variants with the greatest fold increase in the final round of selection relative to the initial plasmid library ( # reads in final round , normalized to total number of reads in the round / # of reads in plasmid library , normalized to total number of reads in the round ) . A pseudo-count of 1 was added to each variant in every round , in order to mitigate effects of small number increases and allow analysis of variants with a zero count in sequencing of the original library ( Fowler et al . , 2014 ) . Twenty top variants with the largest fold increases during the overall selection were chosen for a head-to-head analysis in canine retina ( the directed evolution subset library ) . To compare the selected variants head-to-head , these 20 vectors , along with an AAV2 control , were packaged individually with a ubiquitous CAG promoter driving expression of GFP fused to a unique DNA barcode ( AAV-barcode ) . Vectors were titer matched , mixed together , and injected intravitreally into both eyes of 3 WT dogs . Six weeks after injection , potent GFP fluorescence was detected by fundus imaging of the canine eyes . GFP expression was present in every layer of the dog retina ( Figure 1—figure supplements 1 and 3 ) . Eyes were harvested , tissue samples were collected from across the retina , the RPE was separated from the neuroretina , and photoreceptors were collected using transverse sectioning on a cryostat ( Byrne et al . , 2020 ) . DNA and mRNA were extracted from retinal samples using a Qiagen Allprep kit . Samples were collected from areas across the retina , and from the outer nuclear layer ( ONL ) or RPE . Following DNA and mRNA extraction , AAV-barcodes were PCR amplified from genomic DNA and from cDNA , from photoreceptors and RPE . cDNA was created from mRNA using Superscript III reverse transcriptase , according to the manufacturer’s recommendations . AAV-barcodes were PCR amplified directly from DNA or cDNA . Primers amplified a ~ 50 bp region surrounding the AAV-barcode and contained Illumina adapter sequences and secondary barcodes to allow for multiplexing of multiple samples ( Supplementary file 4 ) . AAV-barcodes amplified from the ONL , RPE , and the injected AAV libraries were then subjected to Illumina sequencing to quantify the representation of each of the variants . PCR amplicons sequenced with a 100-cycle single-read run on a MiSeq . Read counts were normalized by total number of reads in the run . Analysis of barcode abundance was performed using in-house code written in Python , followed by creation of plots in Pandas . Best performing variants were selected based on the fold increase in the percent of total library , relative to the injected library ( % of total in recovered sample / % of total in injected library ) . Analysis was performed on n = 3 dogs . Variants were ranked on the basis of the normalized change in frequency of their representation in the recovered genomes relative to the injected AAV library ( % of total in recovered AAV library / % of total in injected library ) . Selected variants largely outperformed AAV2 in three dogs and across peripheral , mid-peripheral and central retina ( Figure 1—figure supplement 4 ) . In addition , the most abundant variant ( K91 , LAHQDTTKNA ) , which was overly represented in the original library , did not outperform other variants in canine retina , indicating that the metric of quantity of representation in the final round of selection is not the best indicator of fitness for transgene expression . Rankings based on mRNA and DNA recovery indicated different top-performing variants . Evaluation on the basis of mRNA is a more relevant readout of AAV performance , as it is indicative of transgene expression , rather than persistence in extracellular spaces of the tissue or viral endocytosis without useful mRNA expression . Unique 25 bp DNA barcodes were cloned after the stop codon of eGFP , in an AAV ITR-containing plasmid construct containing a self-complementary CAG promoter driving eGFP expression ( scCAG-eGFP-Barcode-bghPolyA ) . Cap genes were cloned into an AAV rep/cap plasmid ( Addgene #64839 ) for packaging . Individual AAV variants were then packaged separately with constructs containing different barcodes using a triple transfection method . Variants were then titer matched and mixed in equal ratios before injection into dogs . For histology , both retinas were lightly fixed in 4% paraformaldehyde , and transferred to PBS . Retinas were then embedded in 5% agarose and sectioned at 100 µm on a vibratome . Tissue was then examined by confocal microscopy . Antibodies for labeling were anti-GFP ( A11122 , Thermo , 1:250 ) and peanut agglutinin ( PNA , Molecular Probes , 1:200 ) a lectin that specifically binds to the cone photoreceptor extracellular matrix . Packaging constructs , containing scCAG-eGFP-Barcode-bghPolyA were constructed as for the DE subset libraries . Approximately equal quantities of AAV serotypes were packaged and pooled . The total titer of pooled virus was: ~ 2 . 5E + 12–5 . 0E + 12 , see Supplementary file 1 . Deep sequencing was used to quantify the relative abundance of vector in the pooled library , by amplifying using primer/adapters and sequencing on a MiSeq Nano flow cell . Titers for all variants in the pool were determined to be within ±1 log from the average variant in the pool , a range that was found to be compatible with accurate normalization across samples . The NHP retinas were dissected , and regions of interest were isolated ( macula , superior , and inferior periphery ) . For cynomolgus macaque , superior , and inferior periphery were pooled . Retinal tissue was placed in Hibernate solution ( Hibernate A -Ca Solution , BrainBits LLC ) , and cells were then dissociated using Macs Miltenyi Biotec Neural Tissue Dissociation Kit for postnatal neurons ( 130-094-802 ) according to manufacturer’s recommendations . Dissected retina pieces were incubated with agitation at 37 °C and further mechanically dissociated . The dissociated neural retina was filtered using a 70 μm MACS Smart Strainer ( Miltenyi Biotec ) to ensure single-cell suspension . Cells were resuspended in 0 . 1% BSA in D-PBS and processed immediately for scRNA-seq . Brain , heart , and liver of mice were freshly dissected , and cells were dissociated using Macs Miltenyi Adult Brain Tissue Dissociation Kit ( 130-107-677 ) , Multi Tissue Dissociation Kit 2 ( 130-110-203 ) and Liver Dissociation Kit ( 130-105-807 ) according to manufacturer’s recommendations . The cells were resuspended in 0 . 1% BSA in D-PBS and processed immediately for scRNA-seq . Following dissociation using Macs Miltenyi Tissue Dissociation Kits specific for retina , brain , heart , and liver , a Miltenyi MACS Tyto sorter was used to enrich for GFP-positive cells . Cells were resuspended in 0 . 1% BSA in D-PBS and processed immediately for scRNA-seq . Marmoset and cynomolgus macaque samples were prepared for single-cell analysis using a 10x Chromium Single Cell 3’ v3 kit . Briefly , single cells from retina samples were captured using a 10x Chromium system ( 10x Genomics ) , the cells were partitioned into gel beads-in-emulsion ( GEMS ) , mRNAs were reverse transcribed and cDNAs with 10x Genomics Barcodes were created with unique molecular identifiers ( UMIs ) for different transcripts . Purified cDNA was PCR amplified and further purified with SPRIselect reagent ( Beckman Coulter , B23318 ) . Final libraries were generated after fragmentation , end repair , A-tailing , adaptor ligation , and sample index PCR steps according to 10x Single Cell 3’ workflow . An additional targeted sequencing analysis was run on these 10x-prepped cDNA samples , using PCR amplification with Q5 High Fidelity DNA Polymerase to target the GFP sequence and its associated AAV-barcode . 10x libraries were pooled and all samples were submitted for deep sequencing on an Illumina Novaseq S4 flowcell at the UPMC Genome Center . Sequencing depth was targeted at 100 , 000 reads per sample for the standard scRNA-seq analysis . Sequenced samples were processed and analyzed on Bridges and Bridges-2 through the Extreme Science and Engineering Discovery Environment ( XSEDE ) ( Towns et al . , 2014 ) . Samples were also analyzed using resources from the University of Pittsburgh Center for Research Computing . Samples from mouse tissues and cultured 293AAV ( Cell Biolabs ) cells were prepared for single cell analysis using a 10x Chromium Single Cell 3’ v3 . 1 kit . The resulting libraries were pooled , and an additional targeted gene enrichment protocol was performed using 10x Chromium Targeted Gene Expression kit . Samples were submitted for deep sequencing on Illumina Novaseq S2 or SP flow cells . Sequencing data was demultiplexed into sample-level fastq files using Cell Ranger mkfastq ( v3 10x Genomics ) . Alignment and cell demultiplexing were run using STARsolo ( Dobin et al . , 2013 ) ( v2 . 7 ) with default parameters . DropletUtils ( Lun et al . , 2019 ) ( v1 . 4 . 3 ) was used after STARsolo to remove empty droplets ( lower . prop = 0 . 05 ) . Cynomolgus macaque samples were aligned to the Macaca_fascicularis_5 . 0/macFas5 reference obtained from UCSC and marmoset samples were aligned to ASM275486v1 obtained from Ensembl . Gene annotation for the cynomolgus macaque was created by lifting over the pre-mRNA gene annotations from the hg38 Ensembl human genome . ASM275486v1 gene annotation files from Ensembl were used for the marmoset . Mouse samples were aligned to the GRCm38 reference GCA_000001635 . 5 from NCBI and annotated with the GENCODE vM17 basic annotation file . Cell-free RNA contamination in droplets was estimated using SoupX ( Young and Behjati , 2020 ) ( v0 . 3 . 1 ) . We estimated contamination using genes selected from SoupX’s inferNonExpressedGenes method , which identifies genes with highly bimodal expression in the samples . The gene expression in cynomolgus macaque samples was adjusted according to the SoupX estimates , using the ‘adjustCounts’ method . No indication of cell-free RNA contamination was observed in marmoset or mouse samples , based off of global expression of key marker genes , and therefore gene expression was not adjusted . Doublets ( 10x droplets containing two cells instead of one ) were then identified using SCDS ( Bais and Kostka , 2020 ) ( v1 . 0 . 0 ) . Any droplets with a hybrid score >1 . 3 were considered doublets . Size factor normalization of the single-cell gene expression was achieved using Scran ( Lun et al . , 2016 ) ( v1 . 12 . 1 ) , and replicates as well as left/right eyes of the same region were combined for normalization . Finally , imputation strategies were used to denoise the high sparsity that is common in scRNA-sequencing ( ALRA v1 . 0 Linderman et al . , 2018 ) . Scanpy ( Wolf et al . , 2018 ) ( v1 . 4 . 4 . post1 ) was used for the analysis of the scRNA-seq data . First , the top 50 principal components of the gene expression matrix were computed and the Euclidean distance between cells was calculated in this low dimensional space . Then , the distances of 0 . 5 % of the closest neighbors were kept for each cell and embedded into a neighborhood graph using the UMAP algorithm . Finally , Leiden clustering was performed on the single cell neighborhood graph . Batch correction was performed to combine samples within the same species ( including samples across the two marmosets as well as FACS-sorted/non FACS-sorted cynomolgus macaque samples ) using Scanorama ( Hie et al . , 2019 ) ( v1 . 2 ) and clustering was performed on the batch-corrected values . If samples were batch corrected , normalized counts were saved as raw data and used for differential gene expression analysis . Cell types were determined by running a differential gene expression analysis using Scanpy’s ‘rank_gene_groups’ function . We used a hypergeometric test and calculated the significance of the intersection of marker genes from one cluster with the published marker genes of each retinal cell type . A Bonferroni p-value correction was applied to account for multiple-hypothesis test . Each cluster is assigned a cell type based on the most significant marker gene intersection p-value . For clusters where the hypergeometric test could not identify a specific cell type match , we annotated the cell type based on marker gene expression using a known cell type marker database . We used two scRNA-seq retina papers ( Macosko et al . , 2015; Peng et al . , 2019 ) to construct our database of marker genes for the retina as well as a larger aggregated scRNA-seq marker database ( Zhang et al . , 2019 ) . For the mouse samples taken from other organs , we created marker gene sets using the Tabula Muris dataset ( Schaum , 2018 ) as well as information from a combination of organ-specific papers for the brain ( Zeisel et al . , 2018; Zeisel et al . , 2015 ) and heart ( Cui et al . , 2019 ) . Statistical tests were run in R . Normality of the datasets was checked using the Shapiro-Wilk test , and it was found that the datasets were unlikely to be normally distributed ( p-values < 2 . 2E-16 for percent cells infected , p-values < 5 . 86E-13 for mean transcripts in infected cells ) . Friedman’s test was run on 8 samples from the marmoset and cynomolgus macaque comparing the total percent of all cells infected in each sample across the AAV variants . Additional Friedman’s tests were run for each cell type , analyzing the percentage of cells infected across variants on a cell type level . One-sided Wilcoxon signed-rank tests were run on the same datasets ( total cells and individual cell types ) , comparing K912 or NHP26 with the other AAV variants , and the Benjamini-Hochberg method was used to correct p-values . AAV variants were also compared by analyzing the average transcripts in infected cells using the same statistical procedure ( Supplementary files 2 and 3 ) . The performance of AAV variants was analyzed based on quantification of AAV variant-mediated GFP-barcode mRNA expression ( AAV-barcodes ) . For non-human primates , AAV-barcodes were analyzed from ( 1 ) the original scRNA-seq data and ( 2 ) PCR amplification of GFP from the 10x single cell prepped sample library . Mouse samples were analyzed using AAV-barcodes from ( 1 ) the original scRNA-seq data and ( 2 ) targeted gene enrichment against GFP and other marker genes . Targeted gene enrichment samples from the mouse were downsampled to a similar number of reads as the non-human primate GFP PCR-amplified non-human primate samples . AAV-barcodes were identified using Salmon ( Patro et al . , 2017 ) ( v0 . 9 . 1 ) transcript quantification . Only reads with one hit to an AAV-barcode were kept . Using these reads , AAV variants were identified based on the AAV-barcode . 10x barcodes in the reads from the PCR amplification analysis were corrected according to the 10x Cell Ranger count algorithm to mitigate any errors that may have been introduced by multiple rounds of PCR . As each UMI ( unique molecular identifier ) represents a single mRNA molecule captured , only one AAV-barcode should exist for each UMI . Rarely , multiple AAV-barcodes were found per UMI – possibly due to sequencing/PCR-introduced errors – in which case the AAV variant with the highest number of counts for that UMI was kept . The PCR-amplified barcodes resulted in a higher number of AAV variants found and 10x barcodes with AAV that were identified from the original scRNA-seq data were added to this set . Additionally , AAV variants from the scRNA-seq dataset were added to the set if that 10x barcode was present in the PCR-amplified analysis but that AAV variant was not previously reported . After identifying AAV variants for each 10x barcode , the 10x barcodes were mapped to the cell types identified previously during the standard scRNA-seq analysis . Once mapped to their respective cell types , AAV counts were normalized by dividing by the total transcriptome nUMI for that 10x barcode and corrected by the dilution factor for each AAV variant . Variants were then divided by the dilution factor of the variant with the highest percentage of cells infected . 293AAV ( HEK293 ) cells were downsampled to 300 , 000 reads and GFP was quantified as previously described using Salmon . 10x cell barcodes were corrected according to the 10x Cell Ranger count algorithm . Cells were not run through any additional quality control steps traditionally used in whole transcriptome single cell analysis , such as empty droplets , as these steps were not applicable to targeted enrichment data . Therefore , corrected 10x barcodes were used for the final cell count and UMI counts originating from GFP were used to estimate the average number of transcripts per cell . K912 was packaged with an SaCas9 construct ( Addgene; pX601-AAV-CMV::NLS-SaCas9-NLS-3xHA-bGHpA;U6::BsaI-sgRNA , Plasmid #61591 ) . The gRNA was designed to target 285 bp downstream of the RHO start codon . A cynomolgus macaque and a rhesus macaque were both injected intravitreally , and 9 weeks ( cyno ) or 6 weeks ( rhesus ) later , tissue was collected for processing . Genomic DNA was extracted using a Qiagen DNeasy Kit and the target site in the RHO gene was PCR amplified with primers attached to Illumina adapter sequences . Amplicon sequences targeting RHO were sequenced on an Illumina iSeq and ~1 , 000 , 000 reads were recovered for each sample . CRISPResso2 ( Clement et al . , 2019 ) ( v2 . 0 . 34 ) was used to quantify and visualize the edits , using the amplicon sequence and guide sequence as input . Reads were filtered using an average base quality of 30 and single base quality of 20 . Primer sequences are listed in Supplementary file 4 .
In order to develop and validate our pipeline , we first engineered AAV vectors with an enhanced capacity to target the outer retina following intravitreal injection , by implementing directed evolution ( DE ) of AAV in canines ( Figure 1—figure supplements 1–3 , and methods ) . DE , which involves applying a selective pressure to libraries of mutated AAV vectors , and conducting iterative rounds of selection , has been used in mouse to create AAV vectors with new abilities to infect Müller glia ( Klimczak et al . , 2009 ) , to infect photoreceptors ( Dalkara et al . , 2013 ) and in primates to deliver genes to the outer retina ( Byrne et al . , 2020 ) . Here , we used DE to engineer new AAV vectors with the ability to bypass structural barriers and infect retinal cells following intravitreal injection in canine retinas . Canines are the main preclinical large animal model for development of retinal gene therapies , including the landmark gene therapy clinical trials for RPE65-LCA2 , due to similar ocular structure and availability of homologous mutant retinal degeneration strains ( Acland et al . , 2001; Beltran et al . , 2012 ) . Therefore , we hypothesized that canines were a promising model in which to conduct a DE screen . DE was implemented similarly to the screen previously reported in primate retina ( Figure 1—figure supplement 1 Byrne et al . , 2020 ) . AAV2-based DE libraries , including a ~588 peptide insertion library ( which contained a random 7-mer peptide flanked by constant linker sequences LA and A , for a total of 10 amino acids inserted at VP1 position ~588 ) ( Müller et al . , 2003 ) , an AAV2-Loopswap library ( Koerber et al . , 2008 ) and an AAV2-ErrorProne library ( Koerber et al . , 2006 ) were pooled and intravitreally injected into canine eyes ( see Figure 1—figure supplement 2 for a description of each of the AAV libraries and pools used in the study ) . Promising variants were identified from the DE screen , based on the fold increase over five rounds of selection , normalized to their frequencies in the starting plasmid library . Then , a secondary round of screening in bulk tissue was performed to compare 20 top candidate canine DE variants . These 20 top vectors , along with an AAV2 control , were packaged individually with a ubiquitous CAG promoter driving expression of GFP fused to a unique DNA barcode . Vectors were titer matched , mixed together to create a subset library containing these top-performing DE variants ( DE-subset library ) , and injected intravitreally into both eyes of 3 WT dogs ( Figure 1—figure supplement 3 ) . Following DNA and mRNA extraction , AAV genome barcodes were PCR amplified from genomic DNA and from cDNA , from photoreceptors and RPE . Variants were ranked on the basis of the normalized change in frequency of their representation in the recovered genomes relative to the injected AAV library ( % of total in recovered AAV library / % of total in injected library ) ( Figure 1—figure supplement 3 ) . Rankings based on mRNA and DNA recovery indicated different top-performing variants . The top ranked variant based on DNA recovery , K916 , contains a 10 amino acid insertion ( PAPQDTTKKA ) at position ~588 . The top ranked variant based on mRNA recovery , K912 , contains a 10 amino acid insertion ( LAPDSTTRSA ) at position ~588 . The convergent variant from the DE screen , that is , the variant that was most abundant at the end of the screen , K91 , which was also overrepresented in the original library , contains a 10 amino acid insertion ( LAHQDTTKNA ) at position ~588 . We have previously shown that convergent variants from DE screens are not necessarily top performers ( Byrne et al . , 2020 ) . That is , greatest fold increase during selection , rather than greatest frequency in the final pool , is the optimal metric for identifying top-performing variants . An additional top-ranking variant based on DNA and RNA recovery , K94 , with amino acid insertion ~588 LATTSQNKPA , was also chosen for further testing . Once an initial set of vector candidates has been created , the relative fitness of these variants to infect different types of cells and retinal regions , and to generate abundant transgene expression must be precisely quantified and compared in order to identify optimally efficient vectors and the best candidate vectors for clinical translation . In order to evaluate the performance of AAV variants created through DE in canine retina , and to quantitatively compare their performance to previously engineered AAV variants created through DE in primate retina ( Byrne et al . , 2020 ) as well as to tyrosine-mutated AAV vectors , we developed a scRNA-Seq based workflow ( scAAVengr ) ( Figure 1A ) . A set of 17 AAV vectors were packaged individually with GFP constructs fused to unique barcodes ( the scAAVengr library ) . Naturally occurring AAV variants included in the scAAVengr library were: AAV1 , AAV2 ( the parental serotype of the canine and primate DE variants ) , AAV5 , AAV8 , AAV9 , and AAVrh10 . Tyrosine and threonine-mutated versions of AAVs , which have been shown to prevent capsid degradation ( Petrs-Silva et al . , 2009; Zhong et al . , 2008 ) , included in the scAAVengr library were AAV2-4YF , AAV2-4YFTV , AAV8-2YF , and AAV9-2YF . DE variants included in the set were K91 , K912 , K916 , K94 , and primate DE variants NHP9 , NHP26 , and SCH/NHP26 ( Byrne et al . , 2020 ) . In previous work , NHP9 has been shown to be highly fovea specific . NHP26 has been shown to bypass structural barriers in primate retina at decreased titer ( Byrne et al . , 2020 ) . Equal amounts of each GFP-barcoded virus were packaged and pooled . The representation of each variant in the packaged and pooled scAAVengr library was then quantified by deep sequencing . The pooled scAAVengr library was intravitreally injected into the eyes of 3 NHPs ( 2 marmosets and one cynomolgus macaque , Figure 1A , and see Supplementary file 1 ) . Eight weeks after intravitreal injection , samples from GFP-expressing retinas were collected ( Figure 1B–E ) . Retinal tissue from macula and peripheral regions were dissociated into single cell suspensions ( Figure 1F ) , a 10x microfluidics controller was used to create cDNA libraries from single cells ( Figure 1G ) , and the cDNA libraries were then sequenced to a depth of 100 , 000 reads per cell . Raw sequencing reads were aligned ( Dobin et al . , 2013 ) to the marmoset genome ( Ensembl ) or the cynomolgus macaque genome ( UCSC ) and processed with multiple QC methods: empty droplets were identified ( Lun et al . , 2019 ) , ambient RNA was removed ( Young and Behjati , 2020 ) , doublets were removed ( Bais and Kostka , 2020 ) , imputation was performed to remove the effects of sparse sampling from sequencing ( Linderman et al . , 2018 ) , single cell gene expression was normalized ( Lun et al . , 2016 ) , and batch correction was performed ( Hie et al . , 2019 ) . Scanpy ( Wolf et al . , 2018 ) was used in conjunction with the Leiden clustering algorithm to assign individual cells to clusters . The hypergeometric test was then used to quantify the significance of intersection of a clusters’ differentially expressed genes with retinal cell type marker genes identified previously ( Macosko et al . , 2015; Peng et al . , 2019 ) . Each cluster was assigned a cell identity based on the most significant intersection . Clusters of all major retinal cell types were identified in marmosets and macaques , largely in agreement with previous scRNA-seq performed in primate retina ( Peng et al . , 2019 ) . AAV-barcodes were quantified using Salmon ( Patro et al . , 2017 ) and mapped to identified cell types using in-house scripts ( Figure 2 , levels of barcoding in scAAVengr analysis are shown in Figure 2—figure supplement 1 ) . The number of cells analyzed , after filtering , were: marmoset superior: 69 , 799; marmoset inferior: 55 , 941; marmoset macula:65 , 023; macaque peripheral: 21 , 904; macaque central: 33 , 907; macaque all cells: 55 , 811 . Three metrics were used to compare vector performance across cell types: First , the absolute number of cells infected by each serotype was quantified ( Figure 3—figure supplement 1 ) . Second , the percent of total cells infected by each serotype was quantified for each major cell type ( Figure 3A , and Figure 3—source data 1 -8 ) . Third , within infected cells , the level of transgene expression was evaluated , relative to total transcripts recovered from each cell ( Figure 3B , and Figure 3—source data 9 -16 ) . Each of these metrics was corrected by the dilution factors for variants in the injected library , previously determined by deep sequencing . Heat maps of these metrics revealed that variants engineered through DE using canine retinas and primate retinas markedly outperformed AAV2 and AAV2 tyrosine mutants across cell types and in peripheral and macular retina . Statistical analysis revealed a significant difference in the percent of total cells infected ( p < 0 . 001 , Friedman’s test , and see Supplementary files 2 and 3 ) . Of the canine variants , K912 outperformed other engineered serotypes , in agreement with the results observed in bulk analysis performed in dog retina ( Figure 1—figure supplement 4 ) . The convergent variant ( K91 ) did not outperform parental serotypes , underscoring the need for deep sequencing to determine top performing AAV variants from the DE screen . Of the primate variants , NHP26 outperformed other variants , infecting major retinal cell types in inner and outer retina in marmoset and cynomolgus macaque retina . Evaluation of AAV infectivity at cell-type resolution revealed that newly engineered K9 variant AAVs and NHP26 infected inner and outer retinal cells in primate retina ( Figure 3 ) . Infectivity , in terms of percent cells infected , was most efficient in RGCs and Müller glia , particularly in the macula where the inner limiting membrane is less of an anatomical barrier . In the outer retina , rods and cones were also infected . Higher rates of infection and expression levels were seen in marmosets compared to the cynomolgus macaque . In order to rank best performing pan-retinal variants and the best performing variants by cell type , variants were plotted by the mean transcripts per cell in infected cells vs . the percent cells infected for each AAV serotype ( Figure 4 ) . Plots were created with data from all cell types on the same plot ( Figure 4A ) or in individual plots per cell type , for each region tested in each primate ( Figure 4B . ) . These plots revealed that K912 was the overall best performing canine variant across retinal cell types , and NHP26 was the top performing primate-derived variant across cell types . Of all the variants tested , K912 was the top performer across cell types . In order to determine the number of AAV variants infecting a single cell , upset plots were created to show the number and serotypes of AAV particles infecting individual cells . Upset plots show the number of cells infected by a particular combination of AAVs ( the intersection size ) as well as the number of cells infected by a particular serotype ( the set size ) . The majority of infected cells were infected by a single variant ( K912 ) , although many cells were infected by multiple serotypes ( Figure 5A ) . As many as eight serotypes infected a single cell in marmoset retina , while up to three serotypes infected a single cell in macaque retina . Next , in order to further interrogate the dynamics of infection in the context of pooled libraries of AAV variants , and to determine whether the presence of other AAV variants impedes infection of library members , HEK293 cells grown in vitro were infected with either AAV2 alone , or with a pool containing 4 AAV’s ( AAV2 , AAV8 , AAV9 and K912 ) , or with a pool containing 16 AAV’s ( AAV2 , AAV8 , AAV9 , K912 , AAV1 , AAVrh10 , AAV2-4YF , AAV2-4YF-T491V , AAV8-2YF , AAV9-2YF , K91 , K916 , K94 , NHP9 , NHP26 , and SCH NHP9/26 ) ( Figure 5B ) . 1E + 6 HEK293 cells were infected with ( a ) AAV2 ( MOI of ~6E + 3 , 2 technical replicates were performed ) , or ( b ) AAV2 ( MOI of ~6E + 3 ) + AAV8 ( MOI of ~4E + 4 ) , AAV9 ( MOI ~ 2E + 4 ) and K912 ( MOI ~ 4E + 3 ) , or ( c ) a pool of 16 variants ( total combined MOI of the pool ~5E + 3 ) . In all three conditions ( alone , in the presence of three additional variants , or infected in the presence of 16 additional variants ) , the number of cells infected by AAV2 , and the average number of transcripts recovered from infected cells ( averaged across all cells infected ) were stable . For K912 , the number of cells infected and the average number of transcripts recovered from infected cells were stable between the 4-member pool and the 16-member pool . Together these results indicate that competition for receptors , or the presence of additional variants in the library does not impact quantification of AAV performance . Then , we estimated the number of unique AAV variants that could be directly compared through the scAAVengr pipeline . AAV constructs were cloned ( as in Figure 1—figure supplement 1 , Step 8 ) , containing a CAG promoter driving expression of GFP , which was then fused to a barcode either 3 , 6 , 9 , or 14 base pairs in length . Each possible nucleotide was equally represented at each position of the barcodes ( hand mixed , IDT ) , with a maximum possible diversity of 64 ( 3 bp barcode ) , 4096 ( 6 bp barcode ) , 262 , 144 ( 9 bp barcode ) , or 268 , 435 , 456 ( 14 bp barcode ) unique barcodes ( Figure 2—figure supplement 1 ) . These constructs were then packaged into AAV2 and used to infect ~1E + 6 HEK293 cells in vitro at an MOI of 1000 . Following onset of GFP expression , 8000 cells were processed through the scAAVengr pipeline , and the number of unique AAV-barcodes recovered was quantified . The number of barcodes recovered were 64 ( 3 bp barcode ) , 4096 ( 6 bp barcode ) , 109 , 701 ( 9 bp barcode ) , or 78 , 307 ( 14 bp barcode ) , indicating that> E + 5 unique AAV variants could be quantified simultaneously , from a single sample containing 8000 cells . Retinal cell expression with K912 , the overall top performer , was then individually validated by packaging and intravitreally injecting a self-complementary CAG-GFP construct in two primates ( Figure 6A–G , Figure 6—figure supplement 1 ) . Ten weeks after injection , GFP expression was evident in retinal flatmounts and cross sections . Confocal microscopy imaging of PNA ( which labels cone inner segments ) -labeled peripheral retina , imaged at the level of the photoreceptor layer , revealed GFP expression in rods and cones , which was higher in rods than in cones , in agreement with scAAVengr heat maps . Cross sections showed strong expression in RGCs and Müller glia , which was more efficient than in outer retina , particularly in the macula , also in agreement with scAAVengr heat maps . Then , as a functional test performed in a therapeutic context , K912 was also packaged with SaCas9 ( Addgene #61591 ) driven by a ubiquitous CMV promoter and a guide RNA targeting rhodopsin , packaged in a single vector ( Figure 6H–J ) . This vector was injected intravitreally in a cynomolgus macaque and a rhesus macaque . RT-PCR amplifying SaCas9 cDNA showed expression of Cas9 in injected , but not in uninjected primate retinas ( Figure 6H ) . We then used deep sequencing to quantify editing from retinal punches containing all cell types , which revealed that 1 . 8% of reads mapping to the targeted site showed editing events in the cynomolgus macaque and 1 . 7% editing in the rhesus macaque ( Figure 6I and J ) , similar to the percent total of K912- GFP infected cells . Individual reads revealed deletions , insertions and base substitutions in the cynomolgus macaque and rhesus macaque following injection with K912-saCas9-gRNA-RHO . Finally , in order to validate the scAAVengr pipeline in other species and tissues , we screened the same 17-member scAAVengr AAV library in mouse brain , heart and liver following systemic injections ( Figure 7 ) . AAV library was packaged , containing each GFP-barcoded virus , and 50 µL of a 5e + 12 vg/mL titer library was injected via facial vein in P0 mice . The representation of each variant in the packaged and pooled library was quantified by deep sequencing . Three weeks after injection , brain , heart and liver were collected and dissociated into single cell suspensions , and a 10x microfluidics controller was used to create cDNA libraries from single cells . GFP+ cells were enriched using FACS , and AAV infection was quantified across all cell types . cDNA libraries were sequenced to a depth of 100 , 000 reads per cell . Raw sequencing reads were aligned to GRCm38 . Reads were processed with QC methods as previously described in order to identify empty droplets ( Lun et al . , 2019 ) , remove doublets ( Bais and Kostka , 2020 ) , perform imputation to remove the effects of sparse sampling from sequencing ( Linderman et al . , 2018 ) and normalize single cell gene expression ( Lun et al . , 2016 ) . Scanpy ( Wolf et al . , 2018 ) was used in conjunction with the Leiden clustering algorithm to assign individual cells to clusters . The hypergeometric test was then used to quantify the significance of intersection of a clusters’ differentially expressed genes with cell type marker genes identified previously ( Cui et al . , 2019; Schaum , 2018; Zeisel et al . , 2018; Zeisel et al . , 2015 ) . Each cluster was assigned a cell identity based on the most significant intersection . The number of cells analyzed , after filtering , were: brain: 17 , 373; brain FACS: 1 , 213; heart: 980; heart FACS: 2 , 746; liver: 10 , 397; liver FACS: 1 , 688 . The absolute number of cells infected by each serotype was quantified . Then , the percent of total cells infected by each serotype was quantified for each major cell type and used to create heat maps of serotype infectivity ( Figure 8A , and Figure 8—source data 1 -6 ) . Finally , within infected cells , the level of transgene expression was evaluated , relative to total transcripts recovered from each cell ( Figure 8B , and Figure 8—source data 7 -12 ) . Each of these metrics was corrected by the percent total of each of the variants in the injected library , previously determined by deep sequencing . Heat maps of these metrics revealed that variants AAV8 , AAV8-2YF , AAV9 , AAV9-2YF , and AAVrh . 10 infected brain , heart , and liver following neonatal systemic injections . AAV1 and AAV5 also infected liver cells . These results are in agreement with previously published data on the tropism of these AAV serotypes ( Duan , 2016; Foust et al . , 2009; Wang et al . , 2010; Wang et al . , 2005; Yang et al . , 2014; Zhang et al . , 2011; Zincarelli et al . , 2008 ) . SCH/NHP26 , a variant created through DE in primate with a backbone partially based on AAV9 , also infected brain , heart and liver . In contrast , AAV2-based retinal DE variants , including K912 and NHP9 , did not efficiently infect organs outside of the eye .
Together these results validate the scAAVengr pipeline as a platform for simultaneous quantitative evaluation and ranking of new AAV serotypes with cell-type resolution . Quantitative comparisons of newly engineered vectors , including evaluation of transgene expression levels and cell-type tropism , have , in the past , required large numbers of animals . In primates , such studies have been impractical , due to the associated ethical and financial burden . Additionally , in primates , the large variability between animals has made comparisons between animals inaccurate , and evaluation of multiple AAVs in the same animal was not possible . In order to address these problems , we developed a single-cell RNA-seq AAV engineering ( scAAVengr ) pipeline for quantitative head-to-head in vivo comparison of transgene expression from newly engineered AAV capsid variants . By simultaneously quantifying cellular and viral RNA at the single-cell level , this method allows for efficient , direct , and head-to-head comparison of multiple vectors across all cell types in the same animals . Pan-tissue efficiency of transgene expression can be determined , as well as specificity for any cell-type of interest that can be identified by its transcriptome profile . The number and identity of unique AAV serotypes infecting a single cell can be observed , and the efficiency and specificity of a potential gene therapy can be accurately estimated . Here , we have evaluated the AAV tropism of newly engineered AAV capsid variants , created in the context of dog retina , with increased ability to infect all major retinal cell types . These variants were directly compared to variants created through DE in primate eyes , as well as naturally occurring variants and tyrosine modified versions . The overall top-performing AAV , K912 , was identified through screening performed in the context of the canine retina . Canines , which have an even thicker vitreous than primates , may represent a more difficult model , with a higher barrier for transduction , resulting in better performing viruses . Further work will need to be done to determine whether vectors from canine screens consistently out-perform variants screened in other animal models . The overall rate of infection achieved by K912 , around 2% of total retinal cells , suggests that significant improvements in AAV transduction must be achieved for maximal therapeutic benefit to be achieved via intravitreal injections . Additional improvements to AAV screening protocols or library construction may be required in order to find even better performing vectors . Here , the best performing variants were determined by using the cloned AAV DE library ( plasmid used for packaging ) as the denominator for evaluation . Additional work will be required to determine whether the packaged AAV DE library ( Round 0 ) may be a better common denominator to follow DE enrichment . It remains to be seen whether barriers to retinal penetration by AAV vectors from intravitreal injection can be overcome by modification of the AAV capsid alone , or whether alternative methods for vector delivery are required . The scAAVengr pipeline may also be useful in determining absolute rates of infection and direct comparison with other routes of administration or methods for transgene expression . The 17-member scAAVengr library of AAV variants was also analyzed in brain , heart and liver following systemic injection in neonatal mice . Variants including K912 and NHP9 , which performed well in primate retina , did not infect mouse brain , liver and heart , which indicates that the mutations which confer increased ability to cross structural barriers in the retina are not sufficient for infection from systemic injection in neonatal mice . Further analysis could be performed , using the same dataset , to quantify infectivity in specific subtypes of retinal cell types , as subtype marker genes are validated . Additional work is required to determine the maximum number of pooled AAV variants that can be screened simultaneously , though our results indicate that libraries containing at least E + 5 variants can be evaluated from a single sample containing 8000 cells . Quantification revealed that multiple AAV serotypes do infect individual cells , suggesting that competition between variants does not inhibit infection . Analysis of in vitro infection of AAV variants in HEK293 cells also suggests that competition between variants infecting the same cells does not significantly affect the number of infected cells or the level of transgene expression in infected cells , validating infectivity data from pooled AAV variants . However , in vivo confirmation of this finding is warranted , as the barriers affecting the dynamics of infection may differ in a more complex environment , such as the retina . Marmoset eyes , which are significantly smaller than human and macaque eyes , were more easily infected than cynomolgus macaques , with greater percentages of cells infected and more infection events per cell . Marmoset eyes may represent a less challenging target for AAV transduction than cynomolgus macaques due to more efficient diffusion of viral particles in smaller eyes , a less rigid vitreous consistency , or immunological factors . This suggests that large primates may be of more use for accurate prediction of performance of gene delivery efficiencies in humans , particularly in the retina , but potentially in other targets as well , such as infection of organs from systemic injections , where diffusion rates and structural barriers may differ between marmosets and macaques . The rate of Cas9 editing in macaques was similar to the total percent of cells infected by K912 , indicating that editing was efficient in infected cells . Importantly , similar rates of infection and genome editing indicate that scRNA-seq can accurately estimate the efficiency of viral gene delivery . Data from primate retina and mouse brain , heart , and liver validate that the scAAVengr workflow is applicable to any tissue for which cell type marker genes are available , and provides a rapid , quantitative method by which AAV vectors can be rapidly evaluated for their clinical potential . This method enables the definitive ranking of AAV’s , in terms of transgene expression efficiency , and in future studies will enable the identification of vectors with greater cell-type specificity . The quantitative nature and single-cell resolution provided by the scAAVengr pipeline therefore enables identification and development of optimal AAV vectors for clinical translation . | Gene therapy is an experimental approach to treating disease that involves altering faulty genes or replacing them with new , working copies . Most often , the new genetic material is delivered into cells using a modified virus that no longer causes disease , called a viral vector . Virus-mediated gene therapies are currently being explored for degenerative eye diseases , such as retinitis pigmentosa , and neurological disorders , like Alzheimer’s and Parkinson’s disease . A number of gene therapies have also been approved for treating some rare cancers , blood disorders and a childhood form of motor neuron disease . Despite the promise of virus-mediated gene therapy , there are significant hurdles to its widespread success . Viral vectors need to deliver enough genetic material to the right cells without triggering an immune response or causing serious side effects . Selecting an optimal vector is key to achieving this . A type of viruses called adeno-associated viruses ( AAV ) are prime candidates , partly because they can be easily engineered . However , accurately comparing the safety and efficacy of newly engineered AAVs is difficult , due to variation between test subjects and the labor and cost involved in careful testing . Öztürk et al . addressed this issue by developing an experimental pipeline called scAAVengr for comparing gene therapy vectors head-to-head . The process involves tagging potential AAV vectors with unique genetic barcodes , which can then be detected and quantified in individual cells using a technique called single-cell RNA sequencing . This means that when several vectors are used to infect lab-grown cells or a test animal at the same time , they can be tracked . The vectors can then be ranked on their ability to infect specific cell types and deliver useful genetic material . Using scAAVengr , Öztürk et al . compared viral vectors designed to target the light-sensitive cells of the retina , which allow animals to see . First , a set of promising viral vectors were evaluated using the scAAVengr pipeline in the eyes of marmosets and macaques , two small primates . Precise levels and locations of gene delivery were quantified . The top-performing vector was then identified and used to deliver Cas9 , a genome editing tool , to primate retinas . Öztürk et al . also used scAAVengr to compare viral vectors in mice , analysing the vectors’ ability to deliver their genetic cargo to the brain , heart , and liver . These experiments demonstrated that scAAVengr can be used to evaluate vectors in multiple tissues and in different organisms . In summary , this work outlines a method for identifying and precisely quantifying the performance of top-performing viral vectors for gene therapy . By aiding the selection of optimal viral vectors , the scAAVengr pipeline could help to improve the success of preclinical studies and early clinical trials testing gene therapies . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods",
"Results",
"Discussion"
] | [
"medicine",
"neuroscience"
] | 2021 | scAAVengr, a transcriptome-based pipeline for quantitative ranking of engineered AAVs with single-cell resolution |
Splenic dendritic cells ( DCs ) present blood-borne antigens to lymphocytes to promote T cell and antibody responses . The cues involved in positioning DCs in areas of antigen exposure in the spleen are undefined . Here we show that CD4+ DCs highly express EBI2 and migrate to its oxysterol ligand , 7α , 25-OHC . In mice lacking EBI2 or the enzymes needed for generating normal distributions of 7α , 25-OHC , CD4+ DCs are reduced in frequency and the remaining cells fail to situate in marginal zone bridging channels . The CD4+ DC deficiency can be rescued by LTβR agonism . EBI2-mediated positioning in bridging channels promotes DC encounter with blood-borne particulate antigen . Upon exposure to antigen , CD4+ DCs move rapidly to the T-B zone interface and promote induction of helper T cell and antibody responses . These findings establish an essential role for EBI2 in CD4+ DC positioning and homeostasis and in facilitating capture and presentation of blood-borne particulate antigens .
Dendritic cells ( DCs ) play a crucial role in presenting antigens to T cells to initiate adaptive immune responses ( Banchereau and Steinman , 1998 ) . The DC lineage is divided into several subtypes based on transcription factor requirements , surface markers and ability to prime CD4 vs CD8 T cell responses . By surface phenotype , the spleen contains three major populations of conventional DCs: CD4+ , CD8+ and double negative ( DN ) DCs ( Shortman and Liu , 2002; Hashimoto et al . , 2011; Miller et al . , 2012 ) . CD4+ DCs also express the C-type lectin receptor called DC-inhibitory receptor 2 ( DCIR2 ) , identified by the antibody 33D1 , whereas CD8+ DCs express the lectin DEC205 ( Dudziak et al . , 2007 ) . Lymph nodes ( LNs ) contain these DC populations but also contain additional subsets that travel to these organs from peripheral tissues sites ( Hashimoto et al . , 2011 ) . Based on soluble antigen immunization studies and antigen targeting to DC subtypes using antibodies , as well as studies in mice deficient in DC subsets , CD4+ and DN DCs have been most strongly implicated in priming CD4 T cell responses against exogenous antigens whereas CD8+ DCs have a more prominent role in cross-presenting antigens and promoting CD8 T cell responses ( Pooley et al . , 2001; Dudziak et al . , 2007; Hildner et al . , 2008 ) . However , these functional distinctions are not strict with some studies , for example , suggesting that CD8+ DCs contribute to priming of CD4+ T cell responses ( Tamura et al . , 2005; Fukaya et al . , 2012 ) . The spleen , the largest secondary lymphoid organ , has an essential function in supporting rapid B and T cell responses against circulating antigens ( Mebius and Kraal , 2005 ) . Blood entering the spleen is released at the marginal sinus prior to travelling through the marginal zone in a direction away from the lymphoid ( white pulp ) regions , into the red pulp to return to circulation via venous sinuses . The CD4+ 33D1+ DCs in the spleen are enriched in an area of the white pulp known as the marginal zone ( MZ ) bridging channel , where the T zone abuts the blood-rich red pulp ( Mitchell , 1973; Witmer and Steinman , 1984; Steinman et al . , 1997 ) . CD8+ DEC205+ DCs , by contrast , are mostly located within the T zone ( also known as the periarteriolar lymphatic sheath or PALS ) . The factors that promote CD4+ DC positioning in MZ bridging channels have been unclear . The developmental requirements of DCs have been under intense investigation . DCs derive from pre-DCs in the bone marrow ( BM ) and these cells seed the spleen and give rise to the CD4+ , CD8+ and DN DC subsets ( Naik et al . , 2006; Liu et al . , 2007 ) . The transcription factors IRF2 , IRF4 , RelB and RBP-J are needed for CD4+ DC development whereas Batf3 , IRF8 and Id2 are required for CD8+ DC development ( Hashimoto et al . , 2011; Satpathy et al . , 2012 ) . CD4+ DCs require signaling by Notch to induce RBP-J ( Lewis et al . , 2011 ) , and by the cytokine LTα1β2 , the latter coming from B cells ( Kabashima et al . , 2005 ) . How maturing DCs are guided into locations that ensure receipt of appropriate developmental or homeostasis signals is not understood . EBI2 ( GPR183 ) is a Gαi coupled receptor that responds to the oxysterol ligand , 7α , 25-hydroxycholesterol ( OHC ) ( Hannedouche et al . , 2011; Liu et al . , 2011 ) . EBI2 is upregulated by activated B cells and plays an important role in promoting their movement to inter- and outer-follicular regions of lymphoid organs during T-dependent antibody responses ( Gatto et al . , 2009; Pereira et al . , 2009 ) . 7α , 25-OHC is synthesized from cholesterol by the actions of two enzymes , CH25H and CYP7B1 , and these are expressed at high levels by stromal cells in inter- and outer-follicular regions while expression of Ch25h is repressed in the center of follicles ( Yi et al . , 2012 ) . As well as expression by B cells , EBI2 is present on a variety of other hematopoietic cell types , including DCs ( Hannedouche et al . , 2011; Liu et al . , 2011 ) . However , its function in DCs is unknown . Here we show that EBI2- and EBI2-ligand deficient mice have a marked deficiency in CD4+ DCs in spleen and related DCs in LNs . CD4+ DCs migrate strongly to 7α , 25-OHC and the CD4+ DCs remaining in the spleen of EBI2-deficient mice fail to localize correctly in MZ bridging channels . The CD4+ DCs in EBI2-deficient mice show evidence of reduced LTβR engagement and their numbers can be rescued by increased LTβR agonism . When situated in the MZ bridging channels , CD4+ DCs rapidly capture particulate antigen , upregulate CCR7 and relocalize to the T-B zone interface to promote CD4 T cell and antibody responses . These responses are defective when DCs lack EBI2 .
CD11c+ splenic DCs from EBI2-GFP reporter mice showed high GFP abundance in CD8− vs CD8+ DCs ( Figure 1A ) . In agreement with the pattern of reporter expression , CD4+ DCs from wild-type mice had more Ebi2 transcripts than CD8+ DCs , and CD4+ DCs had higher surface expression of EBI2 ( Figure 1B , C ) . This difference in chemoattractant receptor expression was unique to EBI2 as it was not seen for the highly expressed chemokine receptors , CCR7 and CXCR4 ( Figure 1—figure supplement 1A ) . The higher EBI2 expression in CD4+ DCs conferred a strong ability to chemotax in response to 7α , 25-OHC in transwell assays , with the cells exhibiting migratory responses to subnanomolar concentrations of ligand ( Figure 1D ) . By contrast , CD8+ DCs failed to migrate to subnanomolar ligand and migration was weak even at high ligand concentrations ( Figure 1D ) . 10 . 7554/eLife . 00757 . 003Figure 1 . EBI2 expression in DCs and deficiency of CD4+ DCs in mice lacking EBI2 or correct amounts of EBI2 ligand . ( A ) Flow cytometric detection of GFP fluorescence in gated splenic CD11c+MHCII+ cells from Ebi2GFP/+ mice . ( B ) Quantitative PCR analysis of Ebi2 transcript abundance in sorted splenic CD11c+MHCII+CD4+ cells ( CD4+ DCs ) and CD11c+MHCII+CD8+ cells ( CD8+ DCs ) . Expression is shown relative to Hprt ( n = 4 mice ) . ( C ) EBI2 surface staining of gated splenic CD4+ and CD8+ DCs from Ebi2−/− or Ebi2+/+ mice ( one representative of four experiments ) . ( D ) Migration of CD4+ DCs and CD8+ DCs in response to 7α , 25-OHC ( mean from four mice ± SE , combined from two experiments ) . ( E ) – ( I ) Flow cytometry and enumeration of splenocytes ( E , F , and G ) , inguinal LN cells ( H ) , and mesenteric LN cells ( I ) from Ebi2−/− , Ch25h−/− , Cyp7b1−/− , Hsd3b7−/− mice , and their matched littermate controls . Numbers adjacent to outlined areas in E indicate percent cells in each gate . DN DCs are defined as CD4-CD8-CD11c+MHCII+ cells . Each dot in F–I represents an individual mouse and error bars indicate mean ± SE of samples combined from three to five independent experiments . Lymph node migratory DCs are defined as MHCIIhiCD11cint and resident DCs , including the CD8+ and 33D1+ DCs , as MHCIIintCD11chi . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , Student’s T-test . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 00310 . 7554/eLife . 00757 . 004Figure 1—figure supplement 1 . DC properties in EBI2-deficient mice and in mice lacking CYP7B1 in radiation resistant cells . ( A ) Quantitative PCR analysis of Ebi2 , Ccr7 , Cxcr4 , and Cxcr5 transcript abundance in sorted CD11c+MHCII+CD4+ cells ( CD4+ DCs ) , CD11c+MHCII+CD8+ cells ( CD8+ DCs ) , and CD4-CD8-CD11c+MHCII+ cells ( DN DCs ) . Expression is shown relative to Hprt . ( B ) Cyp7b1+/- or Cyp7b1-/- recipients were reconstituted with WT bone marrow cells and analyzed by flow cytometry for the indicated DC markers ( n = 8 , combined from three experiments ) . ( C ) Flow cytometric detection of 33D1 expression in gated splenic DC subsets . One representative of three replicated experiments is shown . ( D ) MHCII , CD86 , CD83 , and CD80 expression on gated CD11c+MHCII+ cells from Ebi2+/− and Ebi2−/− mice . Gating is based on isotype control and mean ± SE frequencies are shown for positive staining cells ( n = 4 ) . ( E ) Sorted CD4+ and CD8+ DCs from Ebi2+/− and Ebi2−/− mice were cultured with purified Cell Trace violet labeled allogeneic BALB/c CD4+ T cells . After three days of culture , T cell proliferation was visualized by flow cytometry . Gated percentage of proliferating cells is shown as mean ± SE ( n = 3 ) . ( F ) Flow cytometric detection of GFP fluorescence in gated lymph node DCs from Ebi2GFP/+ mice . Gating was performed as described for Figure 1H . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 004 Analysis of DC subsets in EBI2-deficient mice revealed a threefold to fourfold deficiency in splenic CD4+ DCs without a change in the number of CD8+ DCs or DN DCs ( Figure 1E , F ) . Quantitation of DCs in mice lacking either of the enzymes needed for 7α , 25-OHC synthesis , CH25H or CYP7B1 , showed a comparable selective loss of CD4+ DCs ( Figure 1G ) . Moreover , mice lacking HSD3B7 , the enzyme that metabolizes 7α , 25-OHC , and that have greatly increased amounts of 7α , 25-OHC in lymphoid organs ( Yi et al . , 2012 ) , had a similar deficiency of CD4+ DCs ( Figure 1G ) . When Cyp7b1-deficient mice were reconstituted with wild-type bone marrow , the mice remained CD4+ DC deficient , indicating that radiation resistant stromal cells were a necessary source of EBI2 ligand ( Figure 1—figure supplement 1B ) . The C-type lectin DCIR2 , detected with the 33D1 antibody ( Witmer and Steinman , 1984; Dudziak et al . , 2007 ) , is present on all CD4+ DCs and on a fraction of DN DCs ( Figure 1—figure supplement 1C ) . Enumeration of 33D1+ DCs showed a significant reduction of positive cells in the spleen , confirming that the reduction in CD4+ DCs is due to a loss of this cell type rather than being due to a reduction in surface marker expression ( Figure 1F ) . The CD4+ DCs remaining in EBI2-deficient mice exhibited normal expression of the surface molecules MHC class II , CD80 , CD83 and CD86 and in vitro they supported a normal mixed lymphocyte reaction ( Figure 1—figure supplement 1D , E ) . Although the DC populations present in LNs are more heterogeneous than within spleen , we detected a similar reduction in 33D1+ DCs in peripheral ( inguinal ) and mucosal ( mesenteric ) LNs , while CD8+ DCs and migratory DCs were present at normal frequencies ( Figure 1H , I ) . As in the spleen , LN 33D1+ DCs expressed high amounts of EBI2 ( Figure 1—figure supplement 1F ) . To test whether EBI2 was required intrinsically in CD4+ DCs we generated Ebi2−/− CD45 . 2: WT CD45 . 1 mixed BM chimeras . This analysis revealed a similar reduction in CD4+ DCs to that seen in fully deficient mice , establishing an intrinsic role for EBI2 in these cells and showing that the phenotype was not increased when the mutant cells had to compete with wild-type cells ( Figure 2A , B ) . All other splenic DC subsets , including pDCs , were unaffected by EBI2-deficiency ( Figure 2A , B ) . As another test of the intrinsic in vivo activity of EBI2 in DCs we reconstituted mice with BM cells that had been transduced with an EBI2 and hCD4-reporter expressing retrovirus or with a truncated NGFR vector control . 8 weeks post reconstitution there was a marked increase in the frequency of DCs in the spleens of mice overexpressing EBI2 and this increase was restricted to the 33D1+ DC subset ( Figure 2C , D ) . These data indicate that EBI2 is necessary for development or maintenance of CD4+33D1+ DC and elevated expression of EBI2 is sufficient to promote increased accumulation of this DC type . 10 . 7554/eLife . 00757 . 005Figure 2 . Intrinsic requirement for EBI2 in CD4+ DCs . ( A ) and ( B ) Wild-type CD45 . 2+ mice were lethally irradiated and reconstituted with mixed BM cells ( 1:1 ratio ) from CD45 . 1+ Ebi2+/+ mice and CD45 . 1+CD45 . 2+ Ebi2−/− mice or their wild-type littermate controls . Eight weeks after reconstitution , splenocytes were analyzed by flow cytometry for CD4+ DCs , CD8+ DCs , DN DCs ( gated as in Figure 1 ) , and pDCs ( defined as CD11c+siglecH+MHCII+ ) . ( A ) Representative flow cytometric data . ( B ) Summary of CD45 . 2+ cell frequencies for data of the type in ( A ) ( n = 5 mice ) . ( C ) and ( D ) Mice were reconstituted with BM transduced with a retroviral construct encoding EBI2 or truncated neural growth factor receptor ( NGFR ) , with an IRES–truncated human CD4 cassette as a reporter . Flow cytometric analysis of splenic DCs of chimeras six weeks after reconstitution ( n = 5–7 mice ) . Plots in ( C ) are shown gated on hCD4+ cells and right plots are further gated on MHCII+CD11c+ cells . Plots in ( D ) are summary of frequency of 33D1+ DCs among total CD11c+MHCII+ cells that were non-transduced ( hCD4− ) and transduced ( hCD4+ ) . ***p<0 . 001 , Student’s T-test . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 005 Given the strong chemoattractant activity of EBI2 ligand ( Figure 1D ) and the demonstration in our studies on B cell positioning that the enzymes required for ligand synthesis are expressed abundantly in interfollicular regions ( Yi et al . , 2012 ) , we asked whether EBI2 deficiency led to alterations in DC positioning . In the spleen , CD4+ DCs are enriched in MZ bridging channels , specialized interfollicular regions that connect the T zone with the red-pulp ( Mitchell , 1973 ) . Locating DCs in tissue sections by staining for CD4 is problematic due to the large abundance of CD4 T cells . We therefore took advantage of the DC restricted expression of DCIR2 , detected with 33D1 ( Witmer and Steinman , 1984; Dudziak et al . , 2007 ) , for this assessment . Strikingly , while 33D1+ DCs were concentrated in bridging channels of control mice , they were markedly under represented in these regions in EBI2-deficient mice ( Figure 3A ) . Although the total number of 33D1+ DCs in the sections was reduced , consistent with the reduced cell numbers detected by flow cytometry , the 33D1+ DCs that remained were mostly distributed in the red pulp or in the T zone ( Figure 3A ) . The distribution of DEC205+ DCs was unaltered in EBI2-deficient mice ( Figure 3B ) . A similar deficit of 33D1+ DCs in bridging channels was seen in Ch25h−/− , Cyp7b1−/− and Hsd3b7−/− mice ( Figure 3A ) confirming the need for correctly distributed 7α , 25-OHC to mediate 33D1+ DC positioning . Although 33D1+ DCs are less numerous in LNs , they were again situated most abundantly in areas between follicles in wild-type mice , and they were reduced in these regions in mice lacking EBI2 ( Figure 3C ) . 10 . 7554/eLife . 00757 . 006Figure 3 . EBI2 is required for DC positioning in MZ bridging channels and LN interfollicular regions . Immunohistochemistry of spleen ( A and B ) and lymph node ( C ) sections from the indicated mice were stained for IgD and 33D1 or DEC205 , as indicated . Data are representative of at least six mice of each type . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 006 To examine the basis for the reduced numbers of CD4+ DCs in EBI2-deficient mice we first examined DC turnover rates using bromodeoxyuridine ( BrdU ) incorporation . These measurements showed that each subset of splenic DCs turned over at similar rates in EBI2-deficient and control mice ( Figure 4A ) . Splenic DCs are short-lived ( Kamath et al . , 2002 ) and their viability can be prolonged by removal of the proapoptotic molecule Bim ( Chen et al . , 2007 ) . However , when Cyp7b1-deficient mice were reconstituted with Bim−/− BM , despite having more total DCs , they continued to suffer a CD4+ DC deficiency ( Figure 4B ) . Consistent with this finding , enumeration of annexin V+ DCs and active Caspase-3+ DCs in spleen cell suspensions failed to show differences in their frequency in wild-type and EBI2-deficient mice ( not shown ) . These observations suggest that more rapid apoptosis does not account for the reduced numbers of CD4+ DCs in mice lacking EBI2 function . 10 . 7554/eLife . 00757 . 007Figure 4 . Normal mature DC turnover and pre-DC function in EBI2-deficient mice . ( A ) BrdU positive cells as percent of CD4+ and CD8+ DCs from Ebi2+/− or Ebi2−/− mice given BrdU in drinking water for 1 , 2 , or 4 days . Data are pooled from two to three experiments with five mice in each group . ( B ) Cyp7b1+/− or Cyp7b1−/− recipients were reconstituted with equal number of BM cells from Bim−/− ( CD45 . 2 ) and Bim+/+ ( CD45 . 1 ) mice and DCs enumerated by flow cytometry ( n = 4 ) . ( C ) Flow cytometric detection of GFP fluorescence in gated splenic pre-DCs or DCs from Ebi2GFP/+ or WT mice ( negative control ) . Pre-DCs are defined as Lin- ( CD19 , B220 , CD3 , NK1 . 1 , Ter119 ) CD11c+MHCII-CD172aintCD135+ . ( D ) Number of pre-DCs from spleen and BM of Ebi2+/− and Ebi2−/− mice ( n = 5 mice ) . ( E ) Sorted pre-DCs ( 1 . 0 × 105 ) from CD45 . 1+ WT mice were adoptively transferred into CD45 . 2+ Cyp7b1+/− or Cyp7b1−/− recipients . ( F ) Sorted pre-DCs ( 1 . 0 × 105 ) from CD45 . 2+ Cyp7b1+/− or Cyp7b1−/− mice were adoptively transferred into CD45 . 1+ WT recipients . 6 days after transfer , different subsets of CD45 . 1+ ( E ) or CD45 . 2+ ( F ) pre-DC derived DCs were quantified by flow cytometric analysis , with the right plots in each pair being pre-gated as shown in each left plot ( n=3 mice , one representative of two experiments ) . Graphs on right show a summary of data from three recipients of each type . ( G ) and ( H ) 1:1 Ratio mixed Ebi2−/− ( CD45 . 2+ ) and Ebi2+/− ( CD45 . 1+CD45 . 2+ ) pre-DCs were labeled with violet tracer and transferred into CD45 . 2+ WT recipients . 6 days after transfer , the appearance of DC subsets ( G ) and the frequency of each DC type that had divided three or more times ( H ) were analyzed by flow cytometry ( n = 6 ) . A description of the experimental scheme and an example of the flow cytometric data is included in the figure supplement . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 00710 . 7554/eLife . 00757 . 008Figure 4—figure supplement 1 . PreDC transfer strategy , Notch and IRF4-induced gene expression and effects of Flt3L on DC frequencies . ( A ) 1:1 Ratio mixed sorted Ebi2−/− ( CD45 . 2+ ) and Ebi2+/− ( CD45 . 1+CD45 . 2+ ) pre-DCs were labeled with violet tracer and transferred into CD45 . 2+ WT recipients . 6 days after transfer , the appearance of DC subsets and their extent of proliferation were analyzed by flow cytometry ( n = 6 mice ) . Representative patterns were gated on CD11c+MHCII+CD45 . 2+Ebi2GFP/+ cells . Numbers in plots indicate mean ( ±SE ) frequency of cells in each gate . ( B ) Flow cytometric analysis of ESAM expression on gated CD4+ and CD8+ DCs from the indicated mice . Percentage of ESAM+ cells is shown as mean ± SE ( n = 7 mice ) . ( C ) Quantitative PCR analysis of Dtx1 and Mmp12 transcripts in sorted CD4+ DCs , CD8+ DCs and DN DCs . Expression is shown relative to Hprt ( n = 6 mice ) . ( D ) Ebi2+/− and Ebi2−/− mice were inoculated 1 × 105 Flt3-ligand expressing or control B16 melanoma cells . 7 days later , splenocytes from the indicated mice were analyzed for DC subsets . One representative of four replicated mice is shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 008 EBI2 is expressed in pre-DCs though at somewhat lower levels than in CD4+ DCs ( Figure 4C ) . However , EBI2-deficient mice contained normal numbers of pre-DCs in BM and spleen ( Figure 4D ) . To test whether splenic pre-DC that had developed in the absence of EBI2 signaling retained normal developmental potential we took advantage of the finding that transfers of pre-DCs into congenically marked recipient animals gives rise to small but traceable numbers of CD4 , CD8 and DN DCs 6 days later ( Naik et al . , 2006; Liu et al . , 2007 ) . When WT pre-DCs were transferred to Cyp7b1−/− recipients , the number of CD4+ DCs generated was reduced , consistent with the need for EBI2 ligand to sustain these DCs ( Figure 4E ) . By contrast , when Cyp7b1−/− pre-DCs were transferred to wild-type recipients , normal numbers of CD4+ DCs were generated ( Figure 4F ) . These data demonstrate that Cyp7b1−/− spleens contain a normal compartment of pre-DCs and indicate that EBI2 is required in DCs after their differentiation to a CD4+ state . In further support of this conclusion , when sorted congenically distinguishable and violet tracer labeled pre-DCs from Ebi2−/− and Ebi2+/− mice were cotransferred into wild-type hosts , fewer Ebi2−/− CD4+ DCs were detectable 6 days later but these cells appeared to have divided a similar number of times as the control cells ( Figure 4G , H and Figure 4—figure supplement 1A ) . CD4+ DC development is strongly dependent on Notch signaling ( Caton et al . , 2007; Lewis et al . , 2011 ) . Flow cytometric analysis showed that the CD4+ DC remaining in Ebi2−/− mice had similar amounts of the Notch induced gene , ESAM , as in controls ( Figure 4—figure supplement 1B ) . They also expressed similar amounts of the Notch target gene Deltex1 ( Dtx1 ) ( Figure 4—figure supplement 1C ) . IRF4 is another factor important for development of splenic CD4+ DCs ( Tamura et al . , 2005 ) but EBI2-deficient CD4+ DCs had normal levels of the IRF4 target Mmp12 ( Figure 4—figure supplement 1C ) , suggesting that this transcription factor had been appropriately activated . Flt3 ligand plays a role in the development and maturation of most splenic DC types ( Hashimoto et al . , 2011 ) . To test if diminished access to Flt3 ligand might account for the reduction in CD4+ DCs mice were inoculated with Flt3 ligand producing cells . This led to the expected expansion of both CD4+ and CD8+ DCs in control mice ( Figure 4—figure supplement 1D ) . There was also expansion of DCs in the EBI2-deficient mice , but the selective deficiency in CD4+ DCs remained ( Figure 4—figure supplement 1D ) . Inoculation of GM-CSF producing tumor cells also expanded total CD11c+ cell numbers but failed to restore the CD4+ DC compartment in EBI2-deficient mice ( not shown ) . These data suggest that insufficient access to Notch ligand , Flt3 ligand or GM-CSF cannot explain the reduced numbers of CD4+ DCs . Another factor that is important for homeostasis of splenic CD4+ DCs is the cytokine LTα1β2 ( Kabashima et al . , 2005; Wang et al . , 2005 ) . In previous work we found that when mice lacked LTα1β2 not only were CD4+ DCs reduced in number , but the level of LTβR on the cells was elevated , consistent with less ligand-mediated down-modulation ( Kabashima et al . , 2005 ) . Flow cytometric analysis of cells from EBI2-deficient mice revealed elevation of LTβR on the remaining CD4+ DCs while there was minimal change in the LTβR surface levels on CD8+ DCs ( Figure 5A ) . Consistent with the DC-deficiency being a consequence of reduced LTβR signaling , treatment for 6 days with an LTβR agonistic antibody ( De Trez et al . , 2008; Stanley et al . , 2011 ) led to a significant rescue in the proportion of CD4+ DCs present in the spleen ( Figure 5B , C ) . In tissue sections , increased numbers of EBI2-deficient 33D1+ DCs were evident in both the T zone and red pulp but not in the bridging channels ( Figure 5D ) . A rescue in the proportion of CD4+ DCs was also seen when Cyp7b1−/− mice were reconstituted with BM from transgenic mice ( Ngo et al . , 2001 ) that overexpress LTα1β2 on B cells ( Figure 5E ) . These data are consistent with the possibility that EBI2-mediated positioning of CD4+ DCs between follicles and near LTα1β2-expressing B cells ( Ansel et al . , 2000 ) is required for maintenance of the CD4+ DC compartment . 10 . 7554/eLife . 00757 . 009Figure 5 . Rescue of EBI2-deficient CD4+ DCs by increased LTβR agonsim . ( A ) Surface staining of LTβR in indicated DC subsets from LTβR−/− , WT , and Ebi2−/− mice ( n=7 mice ) . Right panel shows the Ebi2−/−/WT LTβR staining median fluorescence intensity ( MFI ) ratio . ( B ) and ( C ) 1:1 Mixed Ebi2−/− ( CD45 . 2 ) and Ebi2+/+ chimeric mice were treated with LTβR agonist antibody or saline control . ( B ) Shows flow cytometric analysis for CD4 and CD8 on CD11c+ cells and ( C ) shows percentage of CD4+ DCs among total DCs from three to four mixed chimeras of each type . ( D ) Ebi2+/− or Ebi2−/− mice were treated with LTβR agonist antibody for 6 days . Immunohistochemistry staining for 33D1 and IgD in splenic sections from the indicated mice treated with LTβR agonist antibody . One representative picture of three replicated mice is shown . ( E ) Cyp7b1+/− or Cyp7b1−/− recipients were reconstituted with WT or lymphotoxin ( LTβ10 ) transgenic ( tg ) BM cells and analyzed by flow cytometry for the indicated markers . Plots on right are pre-gated on MHCII+CD11c+ cells ( n = 4 mice ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 by Student’s T-test . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 009 The EBI2-dependent positioning of many CD4+ DCs in the MZ bridging channels , adjacent to the blood-rich MZ and red-pulp , suggests these cells will have enhanced access to particulate antigens compared to DCs situated more deeply in the T zone . To explore this idea we injected mice intravenously with sheep red blood cells ( SRBCs ) that had been labeled with the membrane dye , PKH26 ( Hagnerud et al . , 2006 ) . 1 day later , flow cytometric analysis revealed marked capture of PKH26+ antigen by splenic CD4+ DCs whereas capture by CD8+ DCs was considerably lower ( Figure 6A ) . A similar extent of antigen capture by CD4+ DCs was evident as early as 30 min after SRBC injection and PKH26+ CD4+ DCs were still evident in the spleen at 48 hr ( Figure 6—figure supplement 1A , B ) . Analysis of tissue sections from mice 30 min after injection of PKH26-labeled SRBC and fluorescent CD11c antibody showed the expected high density of SRBC throughout the MZ ( Hagnerud et al . , 2006 ) , including in regions overlapping with clusters of CD11c+ DCs ( Figure 6—figure supplement 1C ) . Control experiments indicated that the antigen capture detected by flow cytometry occurred in vivo and not during tissue preparation ( Figure 6—figure supplement 1D ) . Comparison of the number of PKH26 antigen+ DCs between wild-type and EBI2-deficient mice showed a marked deficiency in the latter ( Figure 6B ) . 10 . 7554/eLife . 00757 . 010Figure 6 . MZ bridging channel positioning facilitates DC capture of blood-borne particulate antigen and rapid movement to T zone . ( A ) Mice were i . v . immunized with SRBC-PKH26 or PBS control and analyzed 24 hr later by flow cytometry for PKH26 in gated DC subsets . Numbers indicate the frequency of cells within gate ( mean ± SE , n = 4 mice ) . ( B ) Ebi2+/+ or Ebi2−/− mice were immunized with SRBC-PKH26 and 24 hr later the number of PKH26+ total DCs or CD4+ DCs were quantified as in ( A ) ( n = 4 mice of each type ) . ( C ) Ebi2+/− and Ebi2−/− mice were immunized with SRBCs and spleens isolated at the indicated time points and stained by immunohistochemistry for 33D1 and IgD . ( D ) Mice were pre-immunized with SRBCs or PBS and 6 hr later , immunized with SRBC-PKH26 . 1 day later , the percentage of PKH26+ DCs was quantified by flow cytometry ( n = 3 mice ) . ( E ) CCR7 surface staining of CD4+ DCs from Ccr7−/− mice or WT mice immunized with or without SRBC for 1 . 5 hr . One representative flow cytometric pattern of four mice is shown . ( F ) Migration of CD4+ DCs from PBS or SRBC immunized mice ( 1 . 5 hr ) in response to medium only ( nil ) , CCL19 ( 0 . 1 μg/ml ) , CCL21 ( 0 . 1 μg/ml ) , or 7α , 25-OHC ( 0 . 5 nM ) . Bars show mean ± SE ( n = 4 mice ) . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 by Student’s T-test . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 01010 . 7554/eLife . 00757 . 011Figure 6—figure supplement 1 . Splenic distribution and DC capture of PKH26-labeled SRBC antigen and induction of DC repositioning by SRBC and LPS . ( A ) and ( B ) Flow cytometric detection of PKH26 signal on gated CD4+ DCs in spleens taken from mice injected with SRBC-PKH26 the indicated number of hours before . Data are representative of at least three mice from two experiments . ( C ) Immunofluorescence picture of mice injected with SRBC-PKH26 and anti-CD11c for 0 . 5 hr ( n = 3 ) . ( D ) Flow cytometric analysis of PKH26 signal on DCs in splenocytes co-isolated from a mouse injected with SRBC-PKH26 1 day earlier and a control GFP transgenic mouse that did not receive PKH26-SRBC . Numbers in gates indicate percent cells in each . Data are representative of six mice from two experiments . ( E ) Immunohistochemistry staining of 33D1 and IgD on splenic sections of WT mice immunized with SRBC or LPS at the indicated time points prior to analysis . ( F ) Spleen sections from chimeric mice reconstituted with CCR7+/+ or CCR7−/− BM mixed with Ebi2−/− BM ( 1:1 ratio ) , immunized 1 day earlier with SRBCs or saline ( PBS ) , stained for CD11c and IgD . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 011 Although MZ bridging channel positioning situates CD4+ DCs well for exposure to blood-borne antigens , this location may be less optimal for interaction with T cells . Previous work has shown that bridging channel DCs move into the T zone in response to LPS exposure ( De Becker et al . , 2000; De Trez et al . , 2005 ) but it has been unclear whether this occurs upon immunization with other antigen types . Remarkably , SRBC injection led to very rapid movement of 33D1+ DCs from bridging channels to the T zone , with a notable bias for positioning at the B-T zone interface ( Figure 6C ) . The cells remained in this location for 1 day , but by 2 days the distribution of 33D1+ DCs resembled more closely the unstimulated state ( Figure 6C ) . Interestingly , the inward movement of 33D1+ DCs following SRBC immunization occurred more rapidly than following injection with LPS and with a greater bias for the T-B boundary ( Figure 6—figure supplement 1E ) . Confirming the importance of bridging channel positioning in favoring access to blood-borne antigen , when mice were pretreated with unlabeled SRBCs for 6 hr to promote DC movement in to the T zone and then injected with PKH26+ SRBC , the extent of PKH26+ antigen capture by CD4+ DCs was selectively diminished ( Figure 6D ) . To define the basis for the rapid repositioning of antigen-engaged DCs we tested whether the cells underwent changes in EBI2 or CCR7 expression . CCR7 was upregulated within 1 . 5 hr of SRBC injection ( Figure 6E ) and the cells responded more vigorously to the CCR7 ligands CCL21 and CCL19 than cells from saline injected mice ( Figure 6F ) . By contrast , EBI2 expression and function was minimally affected following SRBC injection ( Figure 6F and not shown ) . When mice were reconstituted with a mixture of CCR7-deficient and EBI2-deficient BM such that the majority of CD4+ DCs were CCR7-deficient while most other cell types were CCR7 wild-type , injection of SRBC failed to cause DC mobilization from MZ bridging channels to the T-B boundary ( Figure 6—figure supplement 1F ) , confirming the role of CCR7 in this repositioning event . Finally we examined how the reduced CD4+ DC frequency and defective DC positioning in EBI2-deficient mice affected CD4 T cell and T-dependent B cell responses . When EBI2-deficient mice that had received wild-type ovalbumin ( Ova ) -specific OTII TCR transgenic T cells were immunized with Ova-conjugated SRBCs there was a significant defect in the extent of T cell proliferation ( Figure 7A ) . The responding T cells also showed less strong upregulation of ICOS and PD1 ( Figure 7A ) . A similar defect in OTII T cell proliferation was observed when transfers were made into Ebi2−/−: Itgax-DTR ( CD11c-DTR ) mixed BM chimeras that had been treated with diphtheria toxin ( DT ) such that most DCs remaining in the animals were EBI2-deficient whereas all hematopoietic CD11c− cell types were ∼50% EBI2 wild-type ( Figure 7—figure supplement 1A ) . This defect appeared to be selective to responses against particulate antigen as CD4 T cell proliferation following injection with soluble Ova or 33D1 antibody conjugated with Ova ( Dudziak et al . , 2007 ) was unaffected by EBI2-deficiency ( Figure 7—figure supplement 1B , C ) . 10 . 7554/eLife . 00757 . 012Figure 7 . DC deficiency in Ebi2−/− mice is associated with a reduced ability to support CD4 T cell and B cell responses to particulate antigen . ( A ) CD45 . 1+ OTII splenocytes were labeled with cell trace violet and adoptively transferred into Ebi2+/− and Ebi2−/− mice . 1 day after transfer , mice were immunized with SRBC-OVA conjugate . 3 days post immunization , OTII T cell proliferation and expression of ICOS and PD1 was examined by flow cytometry . Left panels are pre-gated on CD45 . 1+CD45 . 2− cells and right three panels are further gated on TCR Vα2+ OTII cells . Numbers on gates indicate mean ( ±SE ) frequencies for seven mice combined from two experiments . ( B ) and ( C ) Cell trace violet labeled Hy10 splenocytes were transferred into Ebi2+/− or Ebi2−/− recipients ( B ) or 1:1 mixed CD11cDTR and Ebi2+/− or Ebi2−/− BM chimeras ( C ) . 1 day after cell transfer , mice were immunized with SRBC-HEL2x and flow cytometric analysis of Hy10 B cell frequency and proliferation was conducted at day 3 after immunization . Hy10 B cells were identified as intracellular HEL-binding cells . For ( C ) , mice were treated with DT at the day of Hy10 B cell transfer and again 2 days after cell transfer . Mean ( ±SE ) cell frequency is shown next to each gate ( n = 5–9 mice combined from two to three replicated experiments ) . ( D ) and ( E ) CD45 . 1+ Hy10 splenocytes were transferred into Ebi2+/− or Ebi2−/− recipients and mice were immunized with SRBC-HEL2x . 5 days after immunization , HEL-binding Hy10 B cells and HEL-specific plasma cells of different Ig isotypes were quantified by flow cytometric analysis . Frequencies of GC B cells and total plasma cells are shown in plots ( middle two panels ) pre-gated on Hy10 B cells ( gate shown in left panels ) . Frequencies of IgG1+ plasma cell are shown in plots ( right panels ) pre-gated on Hy10 plasma cells . Numbers on plots in D indicate mean ( ±SE ) cell frequencies in the indicated gates and graph in ( E ) provides a summary of total HEL-binding cell numbers of the indicated types . Data are representative of three replicated experiments with four to six mice in each experiment . Each dot represents an individual mouse . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , by Student’s T-test . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 01210 . 7554/eLife . 00757 . 013Figure 7—figure supplement 1 . EBI2 deficiency on DCs results in defective CD4+ T cell response to particulate antigen but not soluble and 33D1-coupled OVA . ( A ) Cell trace violet labeled CD2-dsRed OTII splenocytes were transferred into 1:1 mixed CD11cDTR and Ebi2+/− or Ebi2−/− BM chimeras . 1 day after cell transfer , mice were immunized with SRBC-HEL2x and flow cytometric analysis of OTII cell proliferation and ICOS expression was conducted at day 3 after immunization . Patterns are pre-gated on OTII cells identified as dsRed+TCRVα2+CD4+ cells . Recipients were treated with DT at the day of Hy10 B cell transfer and again 2 days after cell transfer . Mean ( ±SE ) frequencies are shown next to each gate ( n = 3 mice ) . ( B ) and ( C ) Ebi2+/− and Ebi2−/− mice ( CD45 . 2 ) were transferred with cell trace violet labeled OTII splenocytes ( CD45 . 1 ) and immunized with soluble OVA ( A ) or 33D1-OVA conjugate ( B ) . 3 days post immunization , flow cytometric analysis was performed to determine the OTII cell frequency ( defined as TCRVα2+CD45 . 2− ) and proliferation in splenocytes from the indicated mice . Numbers next to gates indicate mean ( ±SE ) frequency of cells in each ( n > 6 mice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 013 To test for effects on induction of a T-dependent antibody response , EBI2-deficient or control mice were given transfers of wild-type HEL-specific Hy10 B cells and then immunized with HEL-SRBC . 3 days later the Hy10 B cells had undergone reduced proliferation in EBI2-deficient compared to control hosts ( Figure 7B ) . A similar defect in Hy10 B cell proliferation was observed when transfers were made into DT-treated Ebi2−/−:Itgax-DTR mixed BM chimeras ( Figure 7C ) . When the Hy10 B cell response in EBI2-deficient hosts was examined at day 5 , there was an ∼14-fold reduction in total plasma cells and a ∼7-fold reduction in germinal center B cells ( Figure 7D , E ) . Analysis of isotype switched ( IgG1 and IgG2b ) plasma cell numbers showed they were more strongly affected ( 15- to 30-fold reduced ) than IgM plasma cells ( ∼7-fold reduced , Figure 7D , E ) . These findings provide strong evidence that EBI2 function in DCs is needed for mounting normal CD4 T cell-dependent B cell responses .
The high density of DCs in splenic MZ bridging channels has been appreciated for 30 years ( Witmer and Steinman , 1984 ) yet the factors controlling this localization have been unknown . The above findings establish a crucial role for EBI2 and its oxysterol ligand , 7α , 25-OHC , in positioning CD4+ 33D1+ DCs in MZ bridging channels and in LN interfollicular regions . EBI2 is also required for the maintenance of a CD4+ 33D1+ DC compartment of normal size . These roles of EBI2 in DCs are necessary for supporting normal CD4 T cell and B cell proliferative responses , and early plasma cell and germinal center responses to particulate blood-borne antigens . These observations are in close agreement with a study that appeared online at the time this work was submitted ( Gatto et al . , 2013 ) . Our study additionally provides evidence that EBI2-mediated positioning of CD4+ DCs in splenic MZ bridging channels promotes two important processes: ( 1 ) encounter with B cell-derived LTα1β2 that engages the DC LTβR , delivering a signal necessary for maintaining the homeostasis of the population; and , ( 2 ) efficient exposure to blood-borne particulate antigens and an ability to promptly access the T-B zone interface to stimulate T-dependent B cell responses ( Figure 8 ) . 10 . 7554/eLife . 00757 . 014Figure 8 . Model for the role of EBI2 in mediating marginal zone ( MZ ) bridging channel positioning of CD4+ DCs in the spleen . The major zones and cell types are labeled . EBI2 ligand is suggested to be produced in high amounts in the MZ bridging channel by stromal cells ( not depicted ) and acts to promote positioning of EBI2hiCD4+ DCs ( that are also 33D1+ ) in this region . This in turn juxtaposes them to MZ and follicular B cells , resulting in exposure to B cell-derived LTα1β2 ( not depicted ) . LTβR engagement on the DC transmits a homeostasis-promoting signal . This positioning also exposes the cells to particulate antigens traveling with blood flow through the marginal sinus and the MZ . Following particulate antigen exposure , DCs migrate in a CCR7-dependent manner from the bridging channel in to the T zone where they present antigen to helper T cells . DOI: http://dx . doi . org/10 . 7554/eLife . 00757 . 014 Our finding of a similar disruption in CD4+ DC numbers and bridging channel positioning in mice incapable of making 7α , 25-OHC ( Ch25h- and Cyp7b1-deficient ) and in mice unable to properly degrade 7α , 25-OHC ( Hsd3b7-deficient ) ( Russell , 2003; Yi et al . , 2012 ) provides strong evidence that the critical function of EBI2 in CD4+ DCs is to mediate their correct positioning . The alternative possibility that EBI2 is providing a maintenance signal that is independent of its positioning role seems unlikely as Hsd3b7−/− mice with elevated 7α , 25-OHC would not be expected to have the same defect in EBI2 signaling as mice that lack 7α , 25-OHC . Cyp7b1 and Ch25h are abundantly expressed in the outer follicle and in fibroblastic reticular cells ( Yi et al . , 2012 ) and our present findings suggest the necessary source of 7α , 25-OHC for bridging channel positioning and maintenance of CD4+ DCs is stromal . Further experiments will be needed to determine the nature of the stromal cells expressing these enzymes in MZ bridging channels and LN interfollicular regions . In previous work that established a role for LTα1β2 in maintaining normal numbers of CD4+ DCs within the spleen , a slight reduction in the rate of Ltbr−/− DC turnover was observed ( Kabashima et al . , 2005; Wang et al . , 2005 ) . The basis for our inability to observe a statistically significant reduction in DC turnover in EBI2-deficient mice is not clear but may indicate that the cells continue to receive low level LTβR engagement . Precisely how a reduction in LTβR engagement leads to a decrease in CD4+ DC numbers needs further investigation . The sufficiency of increased EBI2 expression to lead to greater numbers of CD4+ DCs may be a consequence of increased LTβR signaling since increased LTβR engagement is adequate to increase CD4 DC numbers ( Kabashima et al . , 2005; Wang et al . , 2005 ) . To account for our inability to observe changes in the CD4+ DC turnover rate or rescuing effects of antagonizing apoptosis when EBI2 function was lacking , we suggest that CD4+ DCs are lost in EBI2-deficient mice at a similar rate irrespective of how long it has been since they were generated from pre-DCs or last proliferated . It seems possible that the cells suffer from the combined effect of mispositioning due to EBI2-deficiency and lack of expression of molecules downstream of LTβR signaling that control interactions with other cell types , with the outcome that the CD4+ DCs are more frequently engulfed by phagocytes or become caught in blood flow and lost from the spleen . In this regard , it is notable that mice lacking CD47 or its partner protein SIRPα , have a similar deficiency in splenic CD4+ DCs ( Hagnerud et al . , 2006; Van et al . , 2006; Saito et al . , 2010 ) . Although we did not observe alterations in CD47 or SIRPα expression by DCs in EBI2-deficient mice ( T . Yi and J . Cyster , unpublished observation ) , it is possible that the function of these molecules in mediating cell–cell interactions or in regulating engulfment by phagocytic cells is somehow altered in EBI2-deficient DCs . The marked defect in T cell proliferative responses in EBI2-deficient mice following injection with antigen-conjugated to SRBCs , but intact response to soluble Ova , observed here and in the recent study of Brink and coworkers ( Gatto et al . , 2013 ) , suggests that EBI2 function in splenic DCs is most important for responses against particulate antigens . We provide evidence that one reason for this differing dependence on EBI2 is that particulate antigen is not able to freely access DCs already positioned within the T zone . This finding is consistent with other studies showing an inverse relationship between the size of molecules and their ability to diffusively access the white pup ( Nolte et al . , 2003 ) . The intact response of EBI2-deficient mice to 33D1-Ova was unexpected but suggests that the defective response to SRBC-antigen is not solely a consequence of reduced CD4+ DC numbers . Instead , these data suggest that appropriate positioning in MZ bridging channels , and thus in close proximity to the marginal sinus and blood-rich MZ , is critical for supporting the response against particulate antigens . This requirement might reflect both the improved efficiency with which DCs in this region can capture blood-borne particles and the ability to be triggered for rapid movement into the T zone ( Figure 8 ) . Although 33D1 coupling ensured efficient delivery of antigen to CD4+ DCs it did not promote their movement into the T zone ( Chappell et al . , 2012 and data not shown ) . The reduction in CD4 T cell responses in mice lacking EBI2+ DCs described here and by Gatto et al . ( 2013 ) might be sufficient to account for the reduced plasma cell and germinal center responses observed in both studies . However , a recent report showed that when antigen was targeted to 33D1+ DCs as an antibody conjugate , it promoted some B cell activation events in a manner that did not depend on T cell help ( Chappell et al . , 2012 ) . A number of studies have shown that DCs can directly augment B cell responses ( MacPherson et al . , 1999; Balazs et al . , 2002; Jego et al . , 2005; Qi et al . , 2006 ) . Our finding that 33D1+ DCs move to the B-T zone interface following particulate antigen immunization also seems consistent with the possibility that these DCs interact with activated B cells as well as T cells . Future studies will be needed to determine whether EBI2-dependent DC–B cell interactions contribute to driving early B cell activation events during responses to particulate antigens . The surface markers ICOS and PD1 are highly expressed by T follicular helper ( Tfh ) cells ( Vinuesa and Cyster , 2011 ) . The less effective induction of ICOS and PD1 on T cells responding to SRBC-Ova in EBI2-deficient hosts suggests that antigen presentation by CD4+ DCs may be important in favoring induction of an early B-helper phenotype in CD4 T cells . In addition to efficiently presenting antigen in the context of MHC class II ( Pooley et al . , 2001; Dudziak et al . , 2007 ) it is possible that appropriately activated CD4+ DCs bias T cell differentiation in a manner that favors provision of B cell help . Consistent with early induction of Tfh cell properties in activated T cells augmenting B cell responses , when T cells lack the ‘master regulator’ of Tfh cell differentiation , Bcl6 , they are poorly able to support extrafollicular plasmablast responses ( Lee et al . , 2011 ) . The reduced induction of Tfh-phenotype cells likely also contributes to the reduced germinal center response . Our findings indicate that CD4+ DC positioning is controlled by a balance of EBI2 and CCR7 expression and ligand responsiveness . Under homeostatic conditions , EBI2 has a dominant influence and promotes positioning in MZ bridging channels . However , following activation by antigen exposure the balance shifts in favor of CCR7 and the cells move promptly into the T zone ( Figure 8 ) . While we suggest that direct exposure to antigen triggers DC movement , we do not exclude the possibility that CCR7 upregulation is promoted indirectly as a consequence of cytokine production by other antigen-exposed cells . The basis for the antigen activated DCs favoring the B-T zone interface compared to the central T zone is not yet clear but this does not seem to depend on EBI2 expression . These data add to a series of findings showing how differential responsiveness to chemoattractants emanating from adjacent zones determines cell position ( Reif et al . , 2002; Cyster , 2005; Bromley et al . , 2008 ) . Although LPS causes CCR7 upregulation on splenic DCs and promotes their movement to the T zone ( De Becker et al . , 2000; De Trez et al . , 2005; Seul et al . , 2012 ) we found that LPS-mediated repositioning happened more slowly and with a more uniform distribution of DCs through the T zone than occurred following SRBC immunization . The DC sensor triggered following SRBC immunization is not clear but in preliminary experiments we do not find a requirement for Myd88 ( unpublished observation ) . SRBCs rapidly become opsonized by complement and presumably also by natural antibody and it is possible that these factors contribute to SRBC capture by CD4+ DCs and to triggering CCR7 upregulation . It seems possible that detection of damaged endogenous RBCs may be a trigger for DC maturation as a number of important pathogens , most notably the malaria parasite , propagate inside RBCs , insert foreign proteins in their membranes and alter their functional properties .
C57BL/6NCr and C57BL/6-cBrd/cBrd/Cr ( CD45 . 1 ) mice at age 7–9 weeks were from National Cancer Institute ( Frederick , MD ) . Ebi2−/− mice ( Pereira et al . , 2009 ) were backcrossed to C57BL/6J for eleven generations . These mice carry an eGFP gene inserted in place of the Ebi2 open reading frame . Ch25h−/− mice ( Bauman et al . , 2009 ) were backcrossed 10 generations to C57BL/6 , Cyp7b1−/− mice ( Rose et al . , 2001 ) were backcrossed to C57BL/6J for 6 generations . Hsd3b7−/− mice were backcrossed to C57BL6/J for two generations and maintained on chow containing 0 . 5% cholic acid and pan-vitamin supplemented water ( Shea et al . , 2007 ) . HEL-specific Hy10 mice , OVA-specific OTII TCR-transgenic mice , Itgax-DTR transgenic mice and LTβ line 10 ( LTβ10 ) -transgenic mice were on a C57BL/6J background . For BrdU labeling , mice were given BrdU containing ( 0 . 5 mg/ml ) drinking water ad lib . For bone marrow chimeras , mice were lethally irradiated by exposure to 1300 rads of g-irradiation in two doses 3 hr apart and bone marrow cells ( 2–5 × 106 ) were transferred through the tail vein . Chimeric mice were analyzed 6–10 weeks after reconstitution . Animals were housed in a specific pathogen-free environment in the Laboratory Animal Research Center at the University of California , San Francisco , and all experiments conformed to ethical principles and guidelines approved by the Institutional Animal Care and Use Committee . For antibody responses , 1 × 105 Hy10 B cells , OT II T cells , or TCR7 T cells were adoptively transferred into Ebi2−/− or matched control recipients . 1 day after cell transfer , recipients were i . p . immunized with 2 × 108 SRBCs conjugated with low affinity HEL2× or OVA . 33D1-OVA is produced by transfecting 293T cells with 33D1-OVA plasmid ( kindly provided by Michel Nussenzweig ) and further purified through protein G column . To visualize cell proliferation , cells were labeled with CellTrace violet cell proliferation kit ( Invitrogen; Grand Island , NY ) . For LTßR agonist treatment , mice were treated with agonistic anti-LTβR antibody ( 3C8 , provided by Carl F Ware ) by i . p . injection of 100 μg of antibody every 3 days for 6 days . Sheep blood was obtained from Colorado Serum Company ( Denver , CO ) . For conjugation of SRBCs with PKH26 ( Sigma-Aldrich; St Louis , MO ) , the procedure was conducted following the manual instruction with 25 μM PKH26 per 250 μl SRBC alsevers . For conjugation of SRBC-HEL2× , SRBCs were first washed with PBS for three times , mixed with HEL2× ( 10 μg/ml SRBC alsevers ) , crosslinked with EDCI ( 1-ethyl-3- ( 3-dimethylaminopropyl ) carbodiimide , Sigma-Aldrich ) for 30 min , and washed three times to remove the free HEL . For SRBC-OVA conjugation , OVA protein ( Sigma-Aldrich ) was first cross-linked with HEL with glutaraldehyde and further conjugated to SRBC in a same manner as SRBC-HEL2× conjugation . For EBI2 surface staining , cells were incubated with a goat anti-EBI2 ( A20 ) polyclonal antibody ( Santa Cruz Biotechnology; Dallas , TX ) , then a biotinlynated anti-goat antibody , and streptdavidin Alex647 ( Jackson Immunoresearch; West Grove , PA ) . Antibodies against CD11c , DCIR2 ( 33D1 ) , DEC205 , CD4 , CD8α , MHCII , CD172α , CD24 , CD135 , Siglet H , CD45 . 1 , CD45 . 2 were obtained from Biolegend ( San Diego , CA ) or eBioscience ( San Diego , CA ) . For intracellular plasma cell staining , spleens were perfused and digested with RPMI1640 containing 2% fetal bovine serum , collagenase type 4 ( 0 . 25 mg/ml ) , and DNase I ( 20 μg/ml ) for 45 min in a 37°C incubator . Immediately after digestion , enzymes were inactivated by adding 2 mM EDTA and splenocytes were isolated through a 70-µm cell strainer . Cells were first stained with the fixable viability dye eFluor450 ( eBioscience ) , blocked with anti-FcR for 10 min ( UCSF Cell Culture Core ) , and stained with anti-CD45 . 1 ( Biolegend ) , anti-CD45 . 2 ( Biolegend ) , anti-B220 ( Biolegend ) , anti-CD138 ( BD Biosciences; San Jose , CA ) for 20 min on ice . The cells were then washed twice and fixed with a BD Cytofix/Cytoperm fixation kit ( BD Biosciences ) . Fixed cells were stained with Alexa647-conjugated HEL and anti-Ig isotype on ice for 15 min . For DC sorting , spleens were digested with 1 . 6 mg type II collagenase ( Worthington Biochemical; Lakewood , NJ ) and DNase I for 30 min at 37°C . Digested spleens were meshed into single cell suspension through a 100-μm cell strainer in PBS buffers containing 2% FCS and 2 mM EDTA . DCs were pre-enriched with anti-CD11c microbeads ( Miltenyi Biotec ) and further sorted on a FACSAira III with a 100-µm nozzle to purities of over 99% . For pre-DC sorting , Type II collagenase and DNase I digested splenocytes were stained for anti-CD11c Biotin and anti-Biotin microbeads ( Miltenyi Biotec; Auburn , CA ) . Cells were pre-enriched with LS-column and further sorted on Moflow to purities of over 95% . For pre-DCs transfer experiments , 1 . 0 × 105 sorted pre-DCs ( define as Lin-CD11c+CD135+ SirpαintMHCII− ) were adoptively transferred into non-irradiated congenic 3- to 4-week-old mice . 6 days after transfer , enriched CD11c cells from the whole mouse spleen was analyzed by flow cytometry with 0 . 7∼3 . 0 × 106 events collected . Cryosections of 7 µm were fixed and stained as described ( Pereira et al . , 2009 ) with: FITC-conjugated anti-IgD ( 11-26c . 2a; BD Biosciences ) , biotin conjugated DEC205 ( Biolegend ) , bio-conjugated 33D1 ( Biolegend ) . For staining of 33D1 , a tyramide kit was used ( TSA Biotin System; Perkin Elmer; Waltham , MA ) . Images were captured with a Zeiss AxioOberver Z1 inverted microscope . For EBI2 transduction of BM-derived cells , BM cells were harvested 4 days after 5-flurouracil ( Sigma ) injection and cultured in the presence of recombinant IL-3 , IL-6 , and mouse stem cell factor ( SCF ) ( 100 ng/ml; Peprotech; Rocky Hill , NJ ) . BM cells were spin-infected twice with a retroviral construct expressing EBI2 or truncated nerve growth factor receptor ( NGFR ) and an IRES–truncated human CD4 cassette as a reporter . 1 day after the last spin infection , hCD4 positive cells were FACS sorted and injected into lethally irradiated C57BL/6 recipients . Around 50% of the hematopoietic cells in recipients were hCD4 positive 6 weeks after reconstitution . 7 , 25-OHC was synthesized as previously described ( Hannedouche et al . , 2011 ) . Mouse recombinant CCL19 and CCL21 were obtained from R&D systems . The chemotaxis assay was conducted as previously described ( Yi et al . , 2012 ) and 5 μM transwell plates were obtained from Corning Life Sciences . Total RNA was isolated from ∼2 . 0 × 104 double-sorted cells with the RNEasy kit ( Qiagen; Hilden , Germany ) and in-column DNA digestion . Real-time PCR was performed using SYBR Green PCR Mix ( Roche; Mannheim , Germany ) and an ABI prism 7300 sequence detection system ( Applied Biosystems; Foster City , CA ) . Hprt mRNA levels were used as internal controls . Sequences for PCR primers were as previously described ( Yi et al . , 2012 ) . | One of the main roles of the spleen is to make the antibodies that protect the body against viruses , bacteria and other microorganisms . Antibodies are made by B cells , which are a type of white blood cell , after they have been exposed to antigens . For most antibody responses , it is also necessary for the B cells to get help from other white blood cells called T cells that have been exposed to antigens . Specialized cells called dendritic cells have a central role in bringing the antigens—which are usually fragments of the infectious agents that have invaded the body—to the T cells . One subset of dendritic cells , called CD4+ dendritic cells , are found in large numbers in a part of the spleen called the bridging channel , but the process by which these cells become localized in this channel has not been fully understood . Now , Yi and Cyster show that a receptor called EBI2 , which is found on the surface of CD4+ dendritic cells , binds to a type of organic molecule called an oxysterol that is produced in the bridging channel . In mice that had been genetically engineered to lack EBI2 or the enzymes needed to make this particular oxysterol—which is known as 7α , 25-dihydroxycholesterol , or 7α , 25-OHC for short—the CD4+ dendritic cells were no longer clustered in the bridging channel and their number was markedly decreased . This showed that the interaction between EBI2 and the oxysterol was essential for ensuring that the CD4+ dendritic cells were in the right place . The correct positioning of the CD4+ dendritic cells was , in turn , necessary for maintaining cell numbers . Moreover , these mice had a weakened immune response because of the very low number of antigens that were being presented to the T cells . A number of autoimmune diseases , such as lupus , are caused by the body developing an immune response to its own cells and tissues . One implication of the work of Yi and Cyster is that if small molecule inhibitors of EBI2 could be designed , they might be able to suppress the onset of such autoimmune responses . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2013 | EBI2-mediated bridging channel positioning supports splenic dendritic cell homeostasis and particulate antigen capture |
Animal behavior has been studied for centuries , but few efficient methods are available to automatically identify and classify it . Quantitative behavioral studies have been hindered by the subjective and imprecise nature of human observation , and the slow speed of annotating behavioral data . Here , we developed an automatic behavior analysis pipeline for the cnidarian Hydra vulgaris using machine learning . We imaged freely behaving Hydra , extracted motion and shape features from the videos , and constructed a dictionary of visual features to classify pre-defined behaviors . We also identified unannotated behaviors with unsupervised methods . Using this analysis pipeline , we quantified 6 basic behaviors and found surprisingly similar behavior statistics across animals within the same species , regardless of experimental conditions . Our analysis indicates that the fundamental behavioral repertoire of Hydra is stable . This robustness could reflect a homeostatic neural control of "housekeeping" behaviors which could have been already present in the earliest nervous systems .
Animal behavior is generally characterized by an enormous variability in posture and the motion of different body parts , even if many complex behaviors can be reduced to sequences of simple stereotypical movements ( Berman et al . , 2014; Branson et al . , 2009; Gallagher et al . , 2013; Srivastava et al . , 2009; Wiltschko et al . , 2015; Yamamoto and Koganezawa , 2013 ) . As a way to systematic capture this variability and compositionality , quantitative behavior recognition and measurement methods could provide an important tool for investigating behavioral differences under various conditions using large datasets , allowing for the discovery of behavior features that are beyond the capability of human inspection , and defining a uniform standard for describing behaviors across conditions ( Egnor and Branson , 2016 ) . In addition , much remains unknown about how the specific spatiotemporal pattern of activity of the nervous systems integrate external sensory inputs and internal neural network states in order to selectively generate different behavior . Thus , automatic methods to measure and classify behavior quantitatively could allow researchers to indetify potential neural mechanisms by providing a standard measurement of the behavioral output of the nervous system . Indeed , advances in calcium imaging techniques have enabled the recording of the activity of large neural populations ( Chen et al . , 2013; Jin et al . , 2012; Kralj et al . , 2011; St-Pierre et al . , 2014; Tian et al . , 2009; Yuste and Katz , 1991 ) , including whole brain activity from small organisms such as C . elegans and larval zebrafish ( Ahrens et al . , 2013; Nguyen et al . , 2016; Prevedel et al . , 2014 ) . A recent study has demonstrated the cnidarian Hydra can be used as an alternative model to image the complete neural activity during behavior ( Dupre and Yuste , 2017 ) . As a cnidarian , Hydra is close to the earliest animals in evolution that had nervous systems . As the output of the nervous system , animal behavior allows individuals to adapt to the environment at a time scale that is much faster than natural selection , and drives the rapid evolution of the nervous system , providing a rich context to study nervous system functions and evolution ( Anderson and Perona , 2014 ) . As Hydra's nervous system evolved from that present in the last common ancestor of cnidarians and bilaterians , the behaviors of Hydra could also represent some of the most primitive examples of coordination between a nervous system and non-neuronal cells . This could make Hydra particularly relevant to our understanding of the nervous systems of model organisms such as Caenorhabditis elegans , Drosophila , zebrafish , and mice , as it provides an evolutionary perspective to discern whether neural mechanisms found in those species represent a specialization or are generally conserved . In fact , although Hydra behavior has been study for centuries , it is still unknown whether Hydra possesses complex behaviors such as social interactions and learning , how its behavior changes under environmental , physiological , nutritional or pharmacological manipulations , or what are the underlying neural mechanisms of these potential changes . Having an unbiased and automated behavior recognition and quantification method would therefore enable such studies with large datasets . This could allow high-throughput systematic pharmacological assays , lesion studies , environmental and physiological condition changes in behavior , or alternations under activation of subsets of neurons , testing quantitative models , and linking behavior outputs with the underlying neural activity patterns . Hydra behavior was first described by Trembley ( 1744 ) , and it consists of both spontaneous and stimulus-evoked movements . Spontaneous behaviors include contraction ( Passano and McCullough , 1964 ) and locomotion such as somersaulting and inchworming ( Mackie , 1974 ) , and can sometimes be induced by mechanical stimuli or light . Food-associated stimuli induce a stereotypical feeding response that consists of three distinct stages: tentacle writhing , tentacle ball formation and mouth opening ( Koizumi et al . , 1983; Lenhoff , 1968 ) . This elaborate reflex-like behavior is fundamental to the survival of Hydra and sensitive to its needs: well-fed animals do not appear to show feeding behavior when exposed to a food stimulus ( Lenhoff and Loomis , 1961 ) . In addition , feeding behavior can be robustly induced by small molecules such as glutathione and S-methyl-glutathione ( GSM ) ( Lenhoff and Lenhoff , 1986 ) . Besides these relatively complex behaviors , Hydra also exhibits simpler behaviors with different amplitudes and in different body regions , such as bending , individual tentacle movement , and radial and longitudinal contractions . These simpler behaviors can be oscillatory and occur in an overlapping fashion and are often hard to describe in a quantitative manner . This , in turn , makes complex behaviors such as social or learning behaviors , which can be considered as sequences of simple behaviors , hard to quantitatively define . Indeed , to manually annotate behaviors in videos that are hours or days long is not only extremely time-consuming , but also partly subjective and imprecise ( Anderson and Perona , 2014 ) . However , analyzing large datasets of behaviors is necessary to systematically study behaviors across individuals in a long-term fashion . Recently , computational methods have been developed to define and recognize some behaviors of C . elegans ( Brown et al . , 2013; Stephens et al . , 2008 ) and Drosophila ( Berman et al . , 2014; Johnson et al . , 2016 ) . These pioneer studies identify the movements of animals by generating a series of posture templates and decomposing the animal posture at each time points with these standard templates . This general framework works well for animals with relatively fixed shapes . However , Hydra has a highly deformable body shape that contracts , bends and elongates in a continuous and non-isometric manner , and the same behavior can occur at various body postures . Moreover , Hydra has different numbers of tentacles and buds across individuals , which presents further challenges for applying template-based methods . Therefore , a method that encodes behavior information in a statistical rather than an explicit manner is desirable . As a potential solution to this challenge , the field of computer vision has recently developed algorithms for deformable human body recognition and action classification . Human actions have large variations based on the individual’s appearance , speed , the strength of the action , background , illumination , etc . ( Wang et al . , 2011 ) . To recognize the same action across conditions , features from different videos need to be represented in a unified way . In particular , the Bag-of-Words model ( BoW model ) ( Matikainen et al . , 2009; Sun et al . , 2009; Venegas-Barrera and Manjarrez , 2011; Wang et al . , 2011 ) has become a standard method for computer vision , as it is a video representation approach that captures the general statistics of image features in videos by treating videos as ‘bags’ of those features . This enables to generalize behavior features in a dataset that is rich with widely varied individual-specific characteristics . The BoW model originated from document classification and spam-detection algorithms , where a text is represented by an empirical distribution of its words . To analyze videos of moving scenes , the BoW model has two steps: feature representation and codebook representation . In the first step , features ( i . e . ‘words’ such as movements and shapes ) are extracted and unified into descriptor representations . In the second step , these higher order descriptors from multiple samples are clustered ( i . e . movement motifs ) , usually by k-means algorithms , and then averaged descriptors from each cluster are defined as ‘codewords’ that form a large codebook . This codebook in principle contains representative descriptors of all the different movements of the animal . Therefore , each clip of the video can be represented as a histogram over all codewords in the codebook . These histogram representations can be then used to train classifiers such as SVMs , or as inputs to various clustering algorithms , supervised or unsupervised , to identify and quantify behavior types . BoW produces an abstract representation compared to manually specified features , and effectively leverages the salient statistics of the data , enabling modeling of large populations . Doing so on a large scale with manually selected features is not practical . The power of such a generalization makes the BoW framework particularly well suited for addressing the challenge of quantifying Hydra behavior . Inspired by previous work on C . elegans ( Brown et al . , 2013; Kato et al . , 2015; Stephens et al . , 2008 ) and Drosophila ( Berman et al . , 2014; Johnson et al . , 2016; Robie et al . , 2017 ) as well as by progress in computer vision ( Wang et al . , 2011 ) , we explored the BoW approach , combining computer vision and machine learning techniques , to identify both known and unannotated behavior types in Hydra . To do so , we imaged behaviors from freely moving Hydra , extracted motion and shape features from the videos , and constructed a dictionary of these features . We then trained classifiers to recognize Hydra behavior types with manual annotations , and identified both annotated and unannotated behavior types in the embedding space . We confirmed the performance of the algorithms with manually annotated data and then used the method for a comprehensive survey of Hydra behavior , finding a surprising stability in the expression of six basic behaviors , regardless of the different experimental and environmental conditions . These findings are consistent with the robust behavioral and neural circuit homeostasis found in other invertebrate nervous systems for "housekeeping" functions ( Haddad and Marder , 2017 ) .
Our goal was to develop a method to characterize the complete behavioral repertoire of Hydra under different laboratory conditions . We collected a Hydra behavior video dataset ( Han , 2018a ) using a widefield dissecting microscope , allowing Hydra to move freely in a culture dish ( Figure 1a ) . We imaged 53 Hydra specimens at a rate of 5 Hz for 30 min , and we either allowed each of them to behave freely , or induced feeding behavior with glutathione , since feeding could not be observed without the presence of prey ( which would have obscured the imaging ) . From viewing these data , we visually identified eight different behaviors , and manually annotated every frame of the entire dataset with the following labels for these eight behavioral states: silent ( no apparent motion ) , elongation , tentacle swaying , body swaying , bending , contraction , somersaulting , and feeding ( Figure 1b; Figure 1e-l; Videos 1–7 ) . Overall , we acquired an annotated Hydra behavior dataset with 360 , 000 fames in total . We noticed that most behaviors in our manual annotation lasted less than 10 s ( Figure 1c ) , and that , within a time window of 5 s , most windows contained only one type of behavior ( Figure 1d ) . A post-hoc comparison of different window sizes ( 1–20 s ) with the complete analysis framework also demonstrated that 5 s windows result in the best performance ( Figure 2—figure supplement 1a ) . Therefore , we chose 5 s as the analysis length of a behavior element in Hydra . Due to the large shape variability of the highly deformable Hydra body during behavior , methods that construct postural eigenmodes from animal postures are not suitable . Therefore , we designed a novel pipeline consisting of four steps: pre-processing , feature extraction , codebook generation , and feature encoding ( Han , 2018b ) ( Figure 2 ) , in line with the BoW framework . Pre-processing was done to exclude the variability in size and rotation angle during imaging , which introduces large variance . To do so , we first defined a behavior element as a 5 s time window , splitting each behavior video into windows accordingly . Then we fitted the body column of Hydra into an ellipse , and centered , rotated , and scaled the ellipse to a uniform template ellipse in each element window . We then encoded spatial information into the BoW framework by segmenting the Hydra area through the videos with an automated program , dividing it into a tentacle region , an upper body region , and a lower body region ( Materials and methods; Video 8 ) . After this encoding , in a feature extraction step we applied a dense trajectory method in each 5 s window element ( Wang et al . , 2011 ) . This dense trajectory method represents video patches by several shape and motion descriptors , including a Histogram of Oriented Gradient ( HOG ) ( Dalal and Triggs , 2005 ) , which is based on edge properties in the image patch; and a Histogram of Optical Flow ( HOF ) as well as a Motion Boundary Histogram ( MBH ) ( Dalal et al . , 2006 ) , based on motion properties . With the dense trajectory method , we first detected and tracked points with prominent features throughout the videos . Then , for each feature point , we analyzed a small surrounding local patch and computed the motion and shape information therein represented by HOF , HOG and MBH descriptors ( Video 9 ) . Thus , each video window element was captured as motion and shape descriptors associated with a set of local video patches with distinguished visual features . To quantize the ‘bags’ of features from each element time window , we collected a uniform feature codebook using all the dense trajectory features . Intuitively , the elements in the codebook are the representative features for each type of motion or shape in a local patch , therefore they can be regarded as standard entries in a dictionary . Here , we generate the codebook in a ‘soft’ manner , where the codebook contains information of the centroid of clusters and their shape . We fitted the features with k Gaussian mixtures . Because each Gaussian is characterized not only by its mean , but also by its variance , we preserved more information than with other ‘hard’ methods like k-means . The next step was to encode the features with the codebook . For this , ‘hard’ methods where one encodes the features by assigning each feature vector to its nearest Gaussian mixture , lose information concerning the shapes of the Gaussians . To avoid this , we encoded the features using Fisher vectors , which describe the distance between features and the Gaussian mixture codebook entries in a probabilistic way , encoding both the number of occurrence and the distribution of the descriptors ( Perronnin et al . , 2010 ) ( Figure 2—figure supplement 1b ) . Since each element window was split into tentacle , upper body and lower body region , we were able to integrate spatial information by encoding the features in each of the three body regions separately ( Figure 2—figure supplement 1b ) . Finally , we represented the behavior in each element window by the concatenated Fisher vector from the three regions . Like all animals , Hydra exhibits behaviors at various time scales . Basic behaviors such as elongation and bending are usually long and temporally uniform , while tentacle swaying , body swaying and contraction are usually short and executed in a burst-like manner . Feeding and somersaulting are more complex behaviors that can be broken down into short behavior motifs ( Videos 6–7 ) ( Lenhoff and Loomis , 1961 ) . Feeding is apparently a stepwise , fixed action pattern-like uniform behavior , with smooth transitions between tentacle writhing , ball formation , and mouth opening ( Video 6 ) . Somersaulting represents another fixed action pattern-like behavior and typically consists of a sequence of basic behaviors with elongation accompanied by tentacle movements , contraction , bending , contraction , elongation , and contraction; completing the entire sequence takes a few minutes ( Video 7 ) . The time spent during each step and the exact way each step is executed vary between animals . Thus , to study Hydra behavior , it is essential to accurately recognize the basic behavior types that comprise these complex activities . We aimed to capture basic behaviors including silent , elongation , tentacle swaying , body swaying , bending , contraction , and feeding , using the Fisher vector features that encode the video statistics . These features were extracted from 5 s element windows and exhibited stronger similarity within the same behavior type , but were distinguished from features of different behavior types ( Figure 3a ) . We then trained support vector machine ( SVM ) classifiers with manual labels on data from 50 Hydra , and tested them on a random 10% withheld validation dataset . We evaluated classification performance via the standard receiver operating characteristic ( ROC ) curve and area under curve ( AUC ) . In addition , we calculated three standard measurements from the number of true positives ( TP ) , true negatives ( TN ) , false positives ( FP ) , and false negatives ( FN ) : accuracy , defined as ( TP + TN ) / ( TP + TN + FP + FN ) ; precision , defined as TP/ ( TP + FP ) ; and recall , defined as TP/ ( TP + FN ) . We achieved perfect training performance ( AUC = 1 , accuracy 100% ) , while on the validation data the overall accuracy was 86 . 8% , and mean AUC was 0 . 97 ( Figure 3b and c; Table 1 ) . This classification framework was easily generalized to new data . With data from three Hydra that were not involved in either codebook generation or classifier training , we extracted and encoded features using the generated codebook , and achieved classification accuracy of 90 . 3% for silent ( AUC = 0 . 95 ) , 87 . 9% for elongation ( AUC = 0 . 91 ) , 71 . 9% for tentacle swaying ( AUC = 0 . 76 ) , 83 . 4% for body swaying ( AUC = 0 . 75 ) , 93 . 9% for bending ( AUC = 0 . 81 ) and 92 . 8% for contraction ( AUC = 0 . 92 ) . All the classifiers achieved significantly better performance than chance levels ( Figure 3b , c and d; Table 1; Video 10 ) . Interestingly , the variability in classifier performance with new data matched human annotator variability ( Figure 1—figure supplement 1 ) . This demonstrates that the codebook generated from training data efficiently captured Hydra behaviors and that trained classifiers can robustly identify the basic behaviors of Hydra and predict their occurrence automatically from the data . Hydra can exhibit overlapping behaviors at the same time . For example , a Hydra specimen could be moving its tentacles while bending , or swaying its body while elongating . In such cases , it would be imprecise to allow only a single behavior label per time window . To capture this situation , we allowed a ‘soft’ classification strategy , taking up to three highest classification types that have a classifier probability within a twofold difference between them . With joint classifiers , we achieved 86 . 8% overall accuracy on the validation data ( 81 . 6% with hard classification ) , and 59 . 0% with new test data ( 50 . 1% with hard classification ) . Soft classification improved classification performance by allowing a realistic situation when Hydra transitions between two behaviors , or executes multiple behaviors simultaneously . In addition to optimally classifying the seven basic behaviors described above , classifying somersaulting video clips with basic behavior classifiers showed a conserved structure during the progression of this behavior ( Figure 3e; Video 11 ) . Somersaulting is a complex behavioral sequence that was not included in the seven visually identified behavior types . This long behavior can typically be decomposed into a sequence of simple behaviors of tentacle swaying , elongation , body swaying , contraction , and elongation . Indeed , in our classification of somersaulting with the seven basic behavior types , we noticed a strong corresponding structure: the classified sequences start with tentacle swaying , elongation , and body swaying , then a sequence of contraction and elongation before a core bending event ( Figure 3e ) ; finally , elongation and contraction complete the entire somersaulting behavior . This segmented classification based on breaking down a complex behavior into a sequence of multiple elementary behaviors agrees with human observations , indicating that our method is able to describe combined behaviors using the language of basic behavior types . Manual annotation identifies behavior types on the basis of distinct visual features . However , it is subjective by nature , especially when the Hydra exhibits multiple behaviors simultaneously and can be affected by the individual biases of the annotator . Therefore , to complement the supervised method described above , where classifiers were trained with annotated categories , we sought to perform unsupervised learning to discover the structural features of Hydra behaviors . Since the Fisher vector representation of video statistics is high-dimensional , we applied a nonlinear embedding technique , t-Distributed Stochastic Neighbor Embedding ( t-SNE ) , to reduce the feature vector dimensionality ( Berman et al . , 2014; Van Der Maaten , 2009 ) . This also allowed us to directly visualize the data structure in two dimensions while preserving the local structures in the data , serving as a method for revealing potential structures of the behavior dataset . Embedding the feature vectors of training data resulted in a t-SNE map that corresponded well to our manual annotation ( Figure 4a ) . Generating a density map over the embedded data points revealed cluster-like structures in the embedding space ( Figure 4b ) . We segmented the density map into regions with a watershed method , which defined each region as a behavior motif region ( Figure 4c and e ) . We evaluated the embedding results by quantifying the manual labels of data points in each behavior motif region . We then assigned a label to each region based on the majority of the manually labeled behavior types in it . Using this approach , we identified 10 distinct behavior regions in the map ( Figure 4d ) . These regions represented not only the seven types we defined for supervised learning , but also a somersaulting region , and three separate regions representing the three stages of feeding behavior ( Figure 4d ) . Embedding with continuous 5 s time windows , which exclude the effect of the hard boundaries of separating the behavior elements , revealed the same types of behaviors ( Figure 4—figure supplement 1 ) . The generated embedding space could be used to embed new data points ( Berman et al . , 2014 ) . We embedded feature vectors from a withheld validation dataset , as well as from three Hydra that were involved neither in generating the feature codebook , nor in the embedding space generation ( Figure 4f ) . Quantitative evaluation of embedding performance with manual labels showed that all behavior types were accurately identified by embedding in the validation data . In test samples , embedding identification of elongation , tentacle sway , body sway , contraction , and the ball formation stage of feeding , all agreed with manual labels ( Figure 4g ) . Therefore , embedding of feature vectors can identify the same behavior types that are identified by human annotation . We wondered if Hydra has any spontaneous behaviors under natural day/night cycles that were not included in our manually labeled sets . We mimicked natural conditions by imaging a Hydra polyp for 3 days and nights with a 12 hr dark/light cycle ( Figure 5a ) , keeping the Hydra in a 100 µm thick coverslip covered chamber to constrain it within the field of view of the microscope ( Figure 5b ) ( Dupre and Yuste , 2017 ) . This imaging approach , although constraining the movement of Hydra , efficiently reduced the complexity of the resulting motion from a three-dimensional to a two-dimensional projection , while still allowing the Hydra to exhibit a basic repertoire of normal behaviors . Using this new dataset , we generated a t-SNE embedding density map from the feature vectors as previously described , and segmented it into behavior motif regions ( Figure 5c ) . Among the resulting 260 motif regions , we not only discovered previously defined behavior types including silent , elongation , bending , tentacle swaying , and contraction , but also found subtypes within certain classes ( Videos 12–19 ) . In elongation , for example , we found three different subtypes based on the state of the animal: slow elongation during the resting state of the animal , fast elongation after a contraction burst , and inter-contraction elongation during a contraction burst ( Videos 13–15 ) . In contraction , we found two different subtypes: the initial contraction of a contraction burst , and the subsequent individual contraction events when the animal is in a contracted state ( Videos 18–19 ) . Interestingly , we also discovered one region in the embedding map that showed a previously unannotated egestion behavior ( Figure 5c; Video 20 ) . Egestion behavior ( also known as radial contraction ) has been observed before ( Dupre and Yuste , 2017 ) , and is typically a fast , radial contraction of the body column that happens within 1 s and empties the body cavity of fluid . Although this behavior happens with animals in their natural free movement , its fast time scale and the unconstrained movement make it hard to identify visually during human annotation . In addition , another t-SNE region showed a novel hypostome movement associated with egestion , characterized by a regional pumping-like movement in hypostome and lower tentacle regions ( Video 21 ) . We evaluated the reliability of the identification of this newly discovered egestion behavior from the embedding method by detecting egestion with an additional ad-hoc method . We measured the width of the Hydra body column by fitting it to an ellipse , and low-pass filtered the width trace . Peaks in the trace then represent estimated time points of egestion behavior , which is essentially a rapid decrease in the body column width ( Figure 5d ) . Detected egestion time points were densely distributed in the newly discovered egestion region in the embedding map ( Figure 5e ) , confirming that our method is as an efficient way to find novel behavior types . Although basic Hydra behaviors such as contraction , feeding and somersaulting have been described for over two centuries , the quantitative understanding of Hydra behaviors has been limited by the subjective nature of human annotation and by the amount of data that can be processed by manual examination . To build quantitative descriptions that link behaviors to neural processes and to explore behavior characteristics of Hydra , we used our newly developed method to compare the statistics of behavior under various physiological and environmental conditions . In its natural habitat , Hydra experiences day/night cycles , food fluctuations , temperature variations , and changes in water chemistry . Therefore , we wondered whether Hydra exhibit different behavioral frequencies or behavioral variability under dark and light conditions , as well as in starved and well-fed conditions . Since we did not expect Hydra to exhibit spontaneous feeding behavior in the absence of prey , we only analyzed six basic behavior types using the trained classifiers: silent , elongation , tentacle swaying , body swaying , bending , and contraction . Lighting conditions ( light vs . dark ) did not result in any significant changes in either the average time spent in each of the six behavior types ( Figure 6a ) or the individual behavior variability defined by the variation of the percentage of time spent in each behavior in 30 min time windows ( Figure 6b ) . Also , compared with starved Hydra , well-fed Hydra did not show significant changes in the percentage of time spent in elongation behavior ( Figure 6c ) , but showed less variability in it ( Figure 6d; starved: 8 . 95 ± 0 . 69% , fed: 5 . 46 ± 0 . 53% , p=0 . 0047 ) . As Hydra polyps vary significantly in size depending on the developmental stage ( e . g . freshly detached buds vs . fully grown animals , ) and nutrition status ( e . g . Hydra that has been starved for a week vs . well-fed Hydra ) , we also explored whether Hydra of different sizes exhibit different behavioral characteristics . For this , we imaged behaviors of Hydra with up to a threefold difference in sizes . Large Hydra polyps had similar silent , body swaying , and contraction patterns , but spent slightly less time in elongation , and more in tentacle swaying ( Figure 6e; elongation small: 22 . 42 ± 1 . 35% , large: 17 . 00 ± 0 . 74% , p=0 . 0068; tentacle swaying small: 34 . 24 ± 1 . 24% , large: 41 . 06 ± 2 . 70% , p=0 . 03 ) . The individual behavior variability remained unchanged ( Figure 6f ) . Finally , we further inquired if different Hydra species have different behavioral repertoires . To answer this , we compared the behaviors of Hydra vulgaris , and Hydra viridissima , ( i . e . green Hydra ) , which contains symbiotic algae in its endodermal epithelial cells ( Martínez et al . , 2010 ) . The last common ancestor of these two species was at the base of Hydra radiation . Indeed , we found that Hydra viridissima exhibited statistically less silent and bending behaviors , but more elongations ( Figure 6g; elongation vulgaris: 15 . 74 ± 0 . 50% , viridissima: 18 . 63 ± 0 . 87% , p=0 . 0303; bending vulgaris: 2 . 31 ± 0 . 27% , viridissima: 1 . 35 ± 0 . 17% , p=0 . 0177 ) , while individual viridissima specimens also exhibit slightly different variability in bending ( Figure 6h; vulgaris: 2 . 17% ± 0 . 26% , viridissima: 1 . 33 ± 0 . 20% , p=0 . 0480 ) . We concluded that different Hydra species can have different basic behavioral repertoires .
Interdisciplinary efforts in the emerging field of computational ethology are seeking novel ways to automatically measure and model natural behaviors of animals ( Anderson and Perona , 2014 ) ( Berman et al . , 2014; Branson et al . , 2009; Brown et al . , 2013; Creton , 2009; Dankert et al . , 2009; Johnson et al . , 2016; Kabra et al . , 2013; Pérez-Escudero et al . , 2014; Robie et al . , 2017; Stephens et al . , 2008; Swierczek et al . , 2011; Wiltschko et al . , 2015 ) . Most of these approaches rely on recognizing variation of the shapes of animals based on fitting video data to a standard template of the body of the animal . However , unlike model organisms like worms , flies , fishes and mice , Hydra differs dramatically from these bilaterian organisms in having an extremely deformable and elastic body . Indeed , during contraction , Hydra appears as a ball with all tentacles shortened , while during elongation , Hydra appears as a long and thin column with tentacles relaxed . Moreover , these deformations are non-isometric , that is , different axes , and different parts of the body , change differently . The number of tentacles each Hydra has also varies . These present difficult challenges for recognizing Hydra behaviors using preset templates . To tackle the problem of measuring behavior in a deformable animal , we developed a novel analysis pipeline using approaches from computer vision that have achieved success in human action classification tasks ( Ke et al . , 2007; Laptev et al . , 2008; Poppe , 2010; Wang et al . , 2009; Wang et al . , 2011 ) . Such tasks usually involve various actions and observation angles , as well as occlusion and cluttered background . Therefore , they require more robust approaches to capture stationary and motion statistics , compared to using pre-defined template-based features . In particular , the bag-of-words ( BoW ) framework is an effective approach for extracting visual information from videos of humans or animals with arbitrary motion and deformation . The BoW framework originated from document classification tasks with machine learning . In this framework , documents are considered ‘bags’ of words , and are then represented by a histogram of word counts using a common dictionary . These histogram representations are widely used for classifying document types because of their efficiency . In computer vision , the BoW framework considers pictures or videos as ‘bags’ of visual words , such as small patches in the images , or shape and motion features extracted from such patches . Compared with another popular technique in machine vision , template matching , BoW is more robust against challenges such as occlusion , position , orientation , and viewing angle changes . It also proves to be successful in capturing object features in various scenes , and thus has become one of the most important developments and cutting edge methods in this field . For these reasons , BoW appears ideally suited for the problem behavior recognition tasks of deformable animals , such as Hydra . We modified the BoW framework by integrating other computational methods , including body part segmentation ( which introduces spatial information ) , dense trajectory features ( which encode shape and motion statistics in video patches ) and Fisher vectors ( which represent visual words in a statistical manner ) . Our choice of framework and parameters proved to be quite adequate , considering both its training and validation accuracy , as well as its generalizability on test datasets ( Figure 2—figure supplement 1 ) . Indeed , the robust correspondence between supervised , unsupervised and manual classification that we report provides internal cross-validation to the validity and applicability of our BoW machine learning approach . Our developed framework , which uses both supervised and unsupervised techniques , is in principle applicable to all organisms , since it does not rely on specific information of Hydra . Compared with previously developed methods , our method would be particularly suitable for behaviors in natural conditions that involve deformable body shapes , as a first step to developing more sophisticated behavioral methods in complex environment for other species . Our goal was to describe all possible Hydra behavior quantitatively . Because of this , we used the BoW framework to capture the overall statistics with a given time frame . We defined the length of basic behavior elements to be 5 s , which maximizes the number of behaviors that were kept intact while uncontaminated by other behavior types ( Figure 1c–d ) . However , it should be noted that our approach could not capture fine-level behavior differences , for example , single tentacle behavior . This would require modeling the animal with an explicit template , or with anatomical landmarks , as demonstrated by deformable human body modeling with wearable sensors . Our approach also does not recover transition probabilities between behavior types , or behavioral interactions between individual specimens . In fact , since our method treats each time window as an independent ‘bag’ of visual words , there was no constraint on the temporal smoothness of classified behaviors . Classifications were allowed to be temporally noisy , therefore they could not be applied for temporal structure analysis . A few studies have integrated state-space models for modeling both animal and human behavior ( Gallagher et al . , 2013; Ogale et al . , 2007; Wiltschko et al . , 2015 ) , while others have used discriminative models such as Conditional Random Field models for activity recognition ( Sminchisescu et al . , 2006; Wang and Suter , 2007 ) . These methods may provide promising candidates for modeling behavior with temporal structure in combination with our approach ( Poppe , 2010 ) . In our analysis pipeline , we applied both supervised and unsupervised approaches to characterize Hydra behavior . In supervised classifications ( with SVM ) , we manually defined seven types of behaviors , and trained classifiers to infer the label of unknown samples . In unsupervised analysis ( t-SNE ) , we did not pre-define behavior types , but rather let the algorithm discover the structures that were embedded in the behavior data . In addition , we found that unsupervised learning could discover previously unannotated behavior types such as egestion . However , the types of behaviors discovered by unsupervised analysis are limited by the nature of the encoded feature vectors . Since the BoW model provides only a statistical description of videos , those features do not encode fine differences in behaviors . Due to this difference , we did not apply unsupervised learning to analyze behavior statistics under different environmental and physiological conditions , as supervised learning appeared more suitable for applications where one needs to assign a particular label to a new behavior video . Once we established the reliability or our method , we quantified the differences between six basic behaviors in Hydra under different experimental conditions with two different species of Hydra and found that Hydra vulgaris exhibits essentially the same behavior statistics under dark/light , large/small and starved/fed conditions . Although some small differences were observed among experimental variables , the overall dwell time and variance of the behavioral repertoire of Hydra were unexpectedly very similar in all these different conditions . Although we could not exclude the possibility that there were differences in the transition probabilities between behaviors , our results still show that , from the six basic behaviors analyzed , Hydra possess a surprisingly robust behavioral frequencies and similarities across environmental and physiological conditions , while interspecies differences introduce stronger behavior differences . Passano and McCullough ( 1964 ) reported that Hydra littoralis , a close relative with our Hydra vulgaris AEP strain ( Martínez et al . , 2010 ) , showed fewer contraction bursts in the evenings and nights than in the day , and feeding every third or fourth day resulted in fewer contraction bursts than was seen with daily feeding . However , they detected contraction bursts by electrical recording of epithelial cell activity , and defined coordinated activity as a contraction event . In our method , we did not measure the number of such events , but instead measured the number of time windows that contain such contractile behavior . This is essentially a measurement of the time spent in contractions instead of frequency of individual events . Using natural light instead of lamp light could also lead to a difference in the observation results . Interestingly , we observed that Hydra vulgaris exhibits different behavior statistics compared with Hydra viridissima . The split leading to Hydra vulgaris and Hydra viridissima is the earliest one in the Hydra phylogenetic tree ( Martínez et al . , 2010 ) , thus these two species are quite divergent . Hydra viridissima also possesses symbiotic algae , and requires light for normal growth ( Lenhoff and Brown , 1970 ) . These differences in genetics and growth conditions could help explaining the observed behavioral differences . Given the similarity in statistics of basic behaviors in different conditions across different animals within the same species , we naturally wondered if our approach might not be effective or sensitive enough to detect significant behavioral differences among animals . However , the high accuracy of the classification of annotated behavior subtypes ( Figure 3 ) and also the method reproducibility , with small variances when measuring different datasets , rules out the possibility that this machine learning method is insensitive , in which case the results of our behavioral analysis would have been noisy and irreproducible . This conclusion was corroborated by the statistical differences in behavior found across two different Hydra species . We had originally expected to observe larger variability of behaviors under different experimental conditions and we report essentially the opposite result . We interpret the lack of behavioral differences across individuals as evidence for robust neural control of a basic behavioral pattern , which appears unperturbed by different experimental conditions . While this rigidity may not seem ideal if one assumes that behavior should flexibly adapt to the environment , it is possible that the six behaviors we studied represent a basic ‘house keeping’ repertoire that needs to be conserved for the normal physiology and survival of the animal . Our results are reminiscent of work on the stomatogastric ganglion of crustaceans that has revealed homeostatic mechanisms that enable central pattern generators to function robustly in different environmental conditions , such as changes in temperature ( Haddad and Marder , 2017 ) . In fact , in this system , neuropeptides and neuromodulators appear to be flexibly used to enable circuit and behavioral homeostasis ( Marder , 2012 ) . Although we do not yet understand the neural mechanisms responsible for the behavioral stability in Hydra , it is interesting to note that the Hydra genome has more than one hundred neuropeptides that could play neuromodulator roles ( Chapman et al . , 2010; Fujisawa and Hayakawa , 2012 ) . This vast chemical toolbox could be used to supplement a relatively sparse wiring pattern with mechanisms to ensure that the basic behavior necessary for the survival of the animal remains constant under many different environmental conditions . One can imagine that different neuromodulators could alter the biophysical properties of connections in the Hydra nerve net and thus keep a stable operating regime of its neurons in the physiological states . In addition , a possible reason for the behavioral similarity among different specimens of Hydra could be their genetic similarities . We used animals derived from the same colony ( Hydra AEP strain ) , which was propagated by clonal budding . Thus , it is likely that many of the animals were isogenic , or genetically very similar . The lack of genetic variability , although it does not explain the behavioral robustness , could partly be a reason behind our differences across species , and it would explain a relatively small quantitative variability across animals of our H . vulgaris colony , as opposed to a larger variability in specimens from the wild . Finally , it is also possible that the behavioral repertoire of cnidarians , which represents some of the simplest nervous systems in evolution in structure and probably also in function , could be particularly simple and hardwired as compared with other metazoans or with bilaterians . From this point of view , the robustness we observed could reflect a ‘passive stability’ where the neural mechanisms are simply unresponsive to the environment , as opposed to a homeostatic ‘active stability’ , generated perhaps by neuromodulators . This distinction mirrors the difference between open-loop and closed-loop control systems in engineering ( Schiff , 2012 ) . Thus , it would be fascinating to reverse engineer the Hydra nerve net and discern to what extent its control mechanisms are regulated externally . Regardless of the reason for this behavioral stability , our analysis provides a strong baseline for future behavioral analysis of Hydra and for the quantitative analysis of the relation between behavior , neural and non-neuronal cell activity . Revisiting Hydra as a model system with modern imaging and computational tools to systematically analyze its behavior provides a unique opportunity to image the entire neural network in an organism and decode the relation between neural activity and behaviors ( Bosch et al . , 2017 ) . With recently established GCaMP6s transgenic Hydra lines ( Dupre and Yuste , 2017 ) and the automated behavior recognition method introduced in this study , it should now be possible to identify the neural networks responsible for each behavior in Hydra under laboratory conditions . With this method , we demonstrate that we are able to recognize and quantify Hydra behaviors automatically , and to identify novel behavior types . This allows us to investigate the behavioral repertoire stability under different environmental , physiological and genetic conditions , providing insight into how a primitive nervous system adapt to its environment . Although our framework does not currently model temporal information directly , it serves as a stepping-stone toward building more comprehensive models of Hydra behaviors . Future work that incorporates temporal models would allow us to quantify behavior sequences , and to potentially investigate more complicated behaviors in Hydra such as social and learning behaviors . As a member of the phylum Cnidaria , Hydra is a sister to bilaterians , and its nervous system and bilaterians nervous systems share a common ancestry . As demonstrated by the analysis of its genome ( Chapman et al . , 2010 ) , Hydra is closer in gene content to the last common ancestor of the bilaterian lineage than some other models systems used in neuroscience research , such as Drosophila and C . elegans . In addition , comparative studies are essential to discern whether the phenomena and mechanisms found when studying one particular species are specialized or general and can thus help illuminate essential principles that apply widely . Moreover , as was found in developmental biology , where the body plan of animals is built using the same logic and molecular toolbox ( Nüsslein-Volhard and Wieschaus , 1980 ) , it is possible that the function and structure of neural circuits could also be evolutionarily conserved among animals . Therefore , early diverging metazoans could provide an exciting opportunity to understand the fundamental mechanisms by which nervous systems generate and regulate behaviors .
The Hydra behavior dataset consisted of 53 videos from 53 Hydra with an average length of 30 min . The AEP strain of Hydra was used for all experiments . Hydra polyps were maintained at 18°C in darkness and were fed with Artemia nauplii once or more times a week by standard methods ( Lenhoff and Brown , 1970 ) . During imaging , Hydra polyps were placed in a 3 . 5 cm plastic petri dish under a dissecting microscope ( Leica M165 ) equipped with a sCMOS camera ( Hamamatsu ORCA-Flash 4 . 0 ) . Videos were recorded at 5 Hz . Hydra polyps were allowed to behave either undisturbed , or in the presence with reduced L-glutathione ( Sigma-Aldrich , G4251-5G ) to induce feeding behavior , since Hydra does not exhibit feeding behavior in the absence of prey . Each video in the Hydra behavior dataset was examined manually at a high playback speed , and each frame in the video was assigned a label in the following eleven classes based on the behavior that Hydra was performing: silent , elongation , tentacle swaying , body swaying , bending , contraction , somersaulting , tentacle writhing of feeding , ball formation of feeding , mouth opening of feeding , and a none class . These behaviors were labeled as 1 through 11 , where larger numbers correspond to more prominent behaviors , and the none class is labeled as 0 . To generate manual labels for a given time window , the top two most frequent labels , L1 and L2 , within this time window were identified . The window was assigned as L2 if its count exceed L1 by three-fold and if L1 is more prominent than L2; otherwise , the window was assigned as L1 . This annotation method labels time windows as more prominent behaviors if behaviors with large motion , e . g . contraction , happens in only a few frames , while the majority of frames are slow behaviors . Prior work has shown that the bag of words methods for video action classification perform better when encoding spatial structure ( Taralova et al . , 2011; Wang et al . , 2009 ) . Encoding spatial information is especially important in our case because allowing the animal to move freely produces large variations in orientation , which is not related to behavior classification . Therefore , we performed a basic image registration procedure that keeps the motion information invariant , but aligns the Hydra region to a canonical scale and orientation . This involves three steps: background segmentation , registration , and body part segmentation . In brief , the image background was calculated by a morphological opening operation , and the background was removed from the raw image . Then , image contrast was adjusted to enhance tentacle identification . Images were then segmented by clustering the pixel intensity profiles to three clusters corresponding to Hydra body , weak-intensity tentacle regions and background by k-means , and the largest cluster from the result was treated as background , and the other two clusters as foreground , that is Hydra region . Connected components that occupied less than 0 . 25% of total image area in this binary image were removed as noise , and the resulting Hydra mask was then dilated by three pixels . To detect the body column , the background-removed image was convolved with a small 3-by-3 Gaussian filter with sigma equals one pixel , and the filtered image was thresholded with Otsu’s segmentation algorithm . The binarization was repeated with a new threshold defined with Otsu’s method within the previous above-threshold region , and the resulting binary mask was considered as the body column . The body column region was then fitted with an ellipse; the major axis , centroid , and orientation of the ellipse were noted . To determine the orientation , two small square masks were placed on both ends of the ellipse along the major axis , and the area of the Hydra region excluding the body column under the patch was calculated; the end with the larger area was defined as the tentacle/mouth region , and the end with the smaller area was defined as the foot region . To separate the Hydra region into three body parts , the part under the upper body square mask excluding the body column was defined as the tentacle region , and the rest of the mask was split at the minor axis of the ellipse; the part close to the tentacle region was defined as the upper body region , and the other as the lower body region . This step has shown to improve representation efficiency ( Figure 2—figure supplement 1b ) . Each 5-s video clip was then centered by calculating the average ellipse centroid position and centering it . The average major axis length and the average orientation were also calculated . Each image in the video clip was rotated according to the average orientation to make the Hydra vertical , and was scaled to make the length of the Hydra body 100 pixels , with an output size of 300 by 300 pixels , while only keeping the region under the Hydra binary mask . Video features including HOF , HOG and MBH were extracted using a codebase that was previously released ( Wang et al . , 2011 ) . Briefly , interest points were densely sampled with five pixels spacing at each time point in each 5 s video clip and were then tracked throughout the video clip with optical flow for 15 frames . The tracking quality threshold was set to 0 . 01; the minimum variation of trajectory displacement was set to 0 . 1 , the maximum variation was set to 50 , and the maximum displacement was set to 50 . The neighboring 32 pixels of each interest point were then extracted , and HOF ( 8 dimensions for eight orientations plus one extra zero bin ) , HOG ( eight dimensions ) and MBH ( eight dimensions ) features were calculated with standard procedures . Note that MBH was calculated for horizontal and vertical optical flow separately , therefore two sets of MBH features , MBHx and MBHy were generated . All features were placed into three groups based on the part of body they fall in , that is tentacles , upper body column , and lower body column . All parameters above were cross-validated with the training and test datasets . A Gaussian mixture codebook and Fisher vectors were generated using the code developed by Jegou et al . for each feature type ( Jégou et al . , 2012 ) , using 50 Hydra in the behavior dataset that includes all behavior types . Features from each body part were centered at zero , then PCA was performed on centered features from all three body parts , keeping half of the original dimension ( five for HOF , four for HOG , MBHx and MBHy ) . Whitening was performed on the PCA data as following , which de-correlates the data and removes redundant information:xwhite , i=xi√λiwhere x denotes principal components , and λ denotes eigenvalues . K=256 Gaussian mixtures were then fitted with the whitened data using a subset of 256 , 000 data points . We then calculated the Fisher vectors as following:zX=Lλ ∇λ LXλ ) where X={xt , t=1 … T} is a set of T data points that were assumed to be generated with Gaussian distributions uλx=∑i=1Kwiui ( x ) , with λ={wi , μi , σi , i=1 , … , K} denotes the Gaussian parameters , and Lλ is the decomposed Fisher Information Matrix:F λ-1≡Ex~uλ∇λloguλ ( x ) ∇λloguλxT=LλTLλ Fisher vectors then represent the normalized gradient vector obtained from Fisher kernel KX , X':KX , X'= ∇λ L X λ ) T Fλ-1 ∇λ L X' λ ) =zXTzX Comparing with hard-assigning each feature to a code word , the Gaussian mixtures can be regarded as probabilistic vocabulary , and Fisher vectors encode information of both the position and the shape of each word with respect to the Gaussian mixtures . Power normalization was then performed on the Fisher vectors to improve the quality of representation:f ( z ) = signzzαwith α=0 . 5 , followed by l2 normalization , which removes scale dependence ( Perronnin et al . , 2010 ) . The final representation of each video clip is a concatenation of Fisher vectors of HOF , HOG , MBHx and MBHy . In this paper , the GMM size was set to 128 with cross-validation ( Figure 2—figure supplement 1c ) . PCA was first performed on the concatenated Fisher vectors to reduce the dimensions while keeping 90% of the original variance . A random 90% of samples from the 50 training Hydra were selected as training data , and the remaining 10% were withheld as validation data . Another three Hydra that exhibit all behavior types were kept as test data . Because each behavior type has different numbers of data points , we trained SVM classifiers using the libSVM implementation ( Chang and Lin , 2011 ) by assigning each type a weight of wi= ( ∑iNi ) /Ni , where i=1 , … , 7 denotes the behavior type , and Ni denotes the number of data points that belong to type i . We trained SVM classifiers with a radial basis kernel , allowing probability estimate , and a fivefold cross-validation testing the cost parameter c with a range of log2c∈-5:2:15 , and the g in the kernel function with a range of log2g∈-5:2:15 , where -5:2:15 denotes integers ranging from −5 to 15 with a step of 2 . The best parameter combination from cross-validation was chosen to train the SVM classifiers . To classify test data , features were extracted as above and were encoded with Fisher vectors with the codebook generated from the training data . PCA was performed using the projection matrix from training data . A probability estimate for each behavior type was given by the classifiers , and the final assigned label is the classifier with the highest probability . For soft classifications , we allowed up to three labels for each sample if the second highest label probability is >50% of the highest label , and the third is >50% of the second highest label . To evaluate classification performance , true positives ( TP ) , false positives ( FP ) , true negatives ( TN ) and false negatives ( FN ) were calculated . Accuracy was defined as Acc= ( TP+TN ) / ( TP+TN+FP+FN ) ; precision was defined as Prc=TP/ ( TP+FP ) ; recall was defined as Acc=TN/ ( TN+FP ) . Two other measurements were calculated: true positive rate TPR=TP/ ( TP+FN ) , and false-positive rate FPR=FP/ ( FP+TN ) . Plotting TPR against FPR gives the standard ROC curve , and the area under curve ( AUC ) reflects the performance of classification . In this plot , a straight line TPR = FPR with AUC = 0 . 5 represents random guess; the upper left quadrant with AUC >0 . 5 represents better performance than random . Embedding was performed with the dimension-reduced data . A random 80% of the dataset from the 50 training Hydra were chosen to generate the embedding map , and the remaining 20% were withheld as validation dataset . Three other Hydra were used as test dataset . We followed the procedures of Berman et al . ( 2014 ) , with a slight modification that uses Euclidean distance as the distance measurement . Embedding perplexity was chosen as 16 . To generate a density map , a probability density function was calculated in the embedding space by convolving the embedded points with a Gaussian kernel; σ of the Gaussian was chosen to be 1/40 of the maximum value in the embedding space by cross-validation with human examination to minimize over-segmentation . In the 3-day dataset , σ was chosen to be 1/60 of the maximum value in order to reveal finer structures . To segment the density map , peaks were found in the density map , a binary map containing peak positions was generated , and peak points were dilated by three pixels . A distance map of the binary image was generated and inverted , and the peak positions were set to be minimum . Watershed was performed on the inverted distance map , and the boundaries were defined with the resulting watershed segmentation . Estimated egestion time points were calculated by first extracting the width profile of Hydra from the pre-processing step , then filtering the width profile by taking the mean width during 15 min after each time point t , and the mean width during 15 min before time t , and subtracting the former from the latter . Peaks were detected on the resulting trace and were regarded as egestion behaviors , since they represent a sharp decrease in the thickness of the animals . All Hydra used for experiments were fed three times a week and were cultured at 18°C . On non-feeding days , the culture medium was changed . Hydra viridissima was cultured at room temperature under sunlight coming through the laboratory windows . For imaging , animals were placed in a petri dish under the microscope without disturbance to habituate for at least 30 min . Imaging typically started between 7 pm and 9 pm , and ended between 9 am and 11 am except for the large/small experiments . All imagings were done excluding environmental light by putting a black curtain around the microscope . For dark condition , a longpass filter with a cutoff frequency of 650 nm ( Thorlabs , FEL0650 ) was placed at the source light path to create ‘Hydra darkness’ ( Passano and McCullough , 1962 ) . For starved condition , Hydra were fed once a week . For the large/small experiment , Hydra buds that were detached from their parents within 3 days were chosen as small Hydra , and mature post-budding mature Hydra polyps were chosen as large Hydra . There was a two- to threefold size difference between small and large Hydra when they were relaxed . However , since the Hydra body was constantly contracting and elongating , it was difficult to measure the exact size . Imaging for this experiment was done during the day time for 1 hr per Hydra . All statistical analyses were done using Wilcoxon rank-sum test unless otherwise indicated . Data is represented by mean ± S . E . M unless otherwise indicated . The code for the method developed in this paper is available at https://github . com/hanshuting/Hydra_behavior . A copy is archived at https://github . com/elifesciences-publications/hydra_behavior ( Han , 2018b ) . The annotated behavior dataset is available on Academic Commons ( dx . doi . org/10 . 7916/D8WH41ZR ) . | How do animals control their behavior ? Scientists have been trying to answer this question for over 2 , 000 years , and many studies have analysed specific behaviors in different animals . However , most of these studies have traditionally relied on human observers to recognise and classify different behaviors such as movement , rest , grooming or feeding . This approach is subject to human error and bias , and is also very time consuming . Because of this , reseachers normally only study one particular behavior , in a piecemeal fashion . But to capture all the different actions an animal generates , faster , more objective methods of systematically classifying and quantifying behavior would be ideal . One promising opportunity comes from studying a small freshwater organism called Hydra , one of the most primitive animals with a nervous system . Thanks to Hydra’s transparent body , modern imaging techniques can be used to observe the activity of their whole nervous system all at once , while the animal is engaged in different actions . However , to realise this potential , scientists need a quick way of automatically recognising different Hydra behaviors , such as contracting , bending , tentacle swaying , feeding or somersaulting . This is particularly difficult because Hydra’s bodies can change shape in different situations . To address this , Han et al . borrowed cutting-edge techniques from the field of computer vision to create a computer program that could automatically analyse hours of videos of freely-moving Hydra and classify their behavior automatically . The computer algorithms can learn how to recognise different behaviors in two ways: by learning from examples already classified by humans ( known as ‘supervised learning’ ) or by letting it pick out different patterns by itself ( known as ‘unsupervised learning’ ) . The program was able to identify all the behaviors previously classified by humans , as well as new types that had been missed by human observation . Using this new computer program , Han et al . discovered that Hydra’s collection of six basic behaviors stays essentially the same under different environmental conditions , such as light or darkness . One possible explanation for this is that its nervous system adapts to the environment to maintain a basic set of actions it needs for survival , although another possibility is that Hydra just does not care and goes along with its basic behaviors , regardless of the environment . Han et al . ’s new method is useful not only for classifying all behavioral responses in Hydra , but could potentially be adapted to study all the behaviors in other animal species . This would allow scientists to systematically perform experiments to understand how the nervous system controls all animal behavior , a goal that it is the holy grail of neuroscience . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Comprehensive machine learning analysis of Hydra behavior reveals a stable basal behavioral repertoire |
Future therapeutic use of engineered site-directed nucleases , like zinc-finger nucleases ( ZFNs ) and transcription activator-like effector nucleases ( TALENs ) , relies on safe and effective means of delivering nucleases to cells . In this study , we adapt lentiviral vectors as carriers of designer nuclease proteins , providing efficient targeted gene disruption in vector-treated cell lines and primary cells . By co-packaging pairs of ZFN proteins with donor RNA in ‘all-in-one’ lentiviral particles , we co-deliver ZFN proteins and the donor template for homology-directed repair leading to targeted DNA insertion and gene correction . Comparative studies of ZFN activity in a predetermined target locus and a known nearby off-target locus demonstrate reduced off-target activity after ZFN protein transduction relative to conventional delivery approaches . Additionally , TALEN proteins are added to the repertoire of custom-designed nucleases that can be delivered by protein transduction . Altogether , our findings generate a new platform for genome engineering based on efficient and potentially safer delivery of programmable nucleases .
The capacity of designed nucleases , like ZFNs and TALENs , to generate DNA double-stranded breaks ( DSBs ) at desired positions in the genome has created optimism for therapeutic translation of locus-directed genome engineering . ZFNs and TALENs are chimeric nucleases composed of a custom-designed DNA binding domain fused to the DNA-cleavage domain from the FokI endonuclease that upon dimer formation cleaves the DNA . ZFN- and TALEN-induced DSBs trigger genome editing through cellular repair mechanisms involving either error-prone non-homologous end joining ( NHEJ ) or homologous recombination ( HR ) with an available DNA donor template . Designer nucleases have broad applications in biological experimentation ( Urnov et al . , 2010; Bogdanove and Voytas , 2011 ) and have been successfully utilized for the production of gene knockout model animals ( Doyon et al . , 2008; Geurts et al . , 2009; Tesson et al . , 2011 ) and in emerging gene therapies ( Perez et al . , 2008; Li et al . , 2011 , 2013; Sun et al . , 2012 ) . The safety of designer nucleases is of major concern in relation to their use in treatment of human diseases . Thus far , ZFNs and TALENs have been administered to cells by transfection or electroporation of nucleic acids , DNA or RNA , encoding a pair of nuclease proteins ( Urnov et al . , 2005; Miller et al . , 2011; Carlson et al . , 2012 ) or by exploiting viral gene vehicles such as integrase-deficient lentiviral vectors ( IDLVs ) ( Lombardo et al . , 2007 ) , adeno-associated virus-derived vectors ( AAV vectors ) ( Ellis et al . , 2013 ) , or adenoviral vectors ( Holkers et al . , 2013 ) . Successful administration of ZFN- or TALEN-encoding genes leads to high intracellular levels of nucleases and furthermore imposes a risk of random insertion in the genome , resulting potentially in prolonged nuclease expression and accumulating events of off-target cleavage . Ideally , ZFNs and TALENs are provided in a ‘hit-and-run’ fashion allowing short-term and dose-controllable nuclease activity without losing the effectiveness of creating locus-directed DSBs . Towards this goal , ZFNs have been fused to destabilizing domains regulated by small molecules to attenuate ZFN toxicity ( Pruett-Miller et al . , 2009 ) . Moreover , by exploiting the cell-penetrating capability of ZFNs , targeted gene disruption has recently been achieved by direct cellular delivery of purified ZFN proteins ( Gaj et al . , 2012 ) . Although such approach may require multiple treatments due to the reduced cellular uptake of proteins ( Mellert et al . , 2012 ) , recent findings suggest that ZFN uptake may be further improved by ligand-mediated endocytosis ( Chen et al . , 2013 ) . However , for gene correction by homology-directed repair such strategies would need to be combined with other means of delivering the donor template . It has been known for decades that retroviruses can tolerate the incorporation of heterologous proteins ( Jones et al . , 1990; Weldon et al . , 1990 ) . Lentiviral particles ( LPs ) have been engineered to carry foreign proteins for the purpose of visualizing the intracellular behavior of the virus during infection ( McDonald et al . , 2002; Jouvenet et al . , 2008 ) and altering the viral integration profile ( Bushman , 1994; Goulaouic and Chow , 1996; Bushman and Miller , 1997 ) , as well as for ferrying antiviral ( Okui et al . , 2000; Ao et al . , 2008 ) and antitumor ( Link et al . , 2006; Miyauchi et al . , 2012 ) protein therapeutics . As the delicate structural composition of HIV-1-derived lentiviral particles is easily disturbed by an inappropriate load of nonviral proteins , leading to suboptimal vector yields and/or reduced transduction capability , various strategies for transducing heterologous protein cargo have been scrutinized . In early strategies , the accessory HIV-1 protein Vpr was adapted as a carrier of fused proteins ( Wu et al . , 1995 ) . Recently , Vpr fusions have been shown also to ferry Cre recombinase ( Michel et al . , 2010 ) and I-SceI meganuclease ( Izmiryan et al . , 2011 ) into transduced cells . However , HIV-1 virions incorporate relatively few copies of Vpr ( estimated 700 copies Vpr per virion [Swanson and Malim , 2008] ) , and the therapeutic potential of such approach may be hampered further by the known toxicity of the Vpr protein ( Tachiwana et al . , 2006 ) . Alternatively , nonviral proteins may be packaged in LPs as part of the Gag polypeptide , as was previously shown for reporter proteins like GFP ( Aoki et al . , 2011 ) and the apoptosis-inducing caspase 3 protein ( Miyauchi et al . , 2012 ) . During virion maturation , Gag is processed by the viral proteins into shorter proteins constituting the structural—and most abundant—proteins of the virus particle . It is estimated that each virion contains 5000 copies of Gag and 250 copies of GagPol ( Swanson and Malim , 2008 ) . We recently adapted LPs for the delivery of the piggybac DNA transposase ( Cai et al . , 2014 ) . The transposase was released from Gag in the virus particles in a protease-dependent manner and found to be able to facilitate efficient DNA transposition in transduced cells . In yet another strategy , heterologous proteins fused to the integrase in the Pol region of the GagPol polypeptide were successfully delivered by protein transduction ( Schenkwein et al . , 2010 ) . In this study , we describe the use of lentivirus-derived particles as carriers of designer nucleases for safe administration of ZFN and TALEN proteins fused to lentiviral Gag precursors . We produce ZFN-loaded lentiviral particles that induce high-efficiency gene disruption with a favorable on-target/off-target ratio in safe genomic harbors like the CCR5 locus . Also , gene disruption and repair is evident in cells treated with particles carrying TALEN proteins . Successful incorporation of nuclease proteins within lentiviral particles allows co-delivery of nucleases and the donor template for homology-directed repair . Our findings demonstrate targeted and programmable gene repair in the human genome by delivery of both ‘scissors’ and ‘patch’ in a single combined protein and gene vehicle .
To incorporate ZFN proteins in LPs , we fused ZFNs to the N-terminus of Gag containing also an intervening heterologous phospholipase C-δ1 pleckstrin homology ( PH ) domain thought to promote the recruitment of Gag and GagPol to the membrane ( Urano et al . , 2008 ) . We began by incorporating ZFNs targeting the egfp reporter gene ( Urnov et al . , 2005 ) into the gag gene of a packaging construct harboring the IN D64V mutation ( Figure 1A ) , rendering the viral integrase ( IN ) incapable of catalyzing vector insertion . An HIV-1 protease cleavage site ( SQNY/PIVQ ) was included between the ZFN and the PH domain to allow the release of functional ZFN proteins during particle maturation . LPs harboring HA-tagged left and right egfp-targeting ZFNs , respectively , were produced separately and analyzed for their particle content by Western blot analysis ( Figure 1B , C ) . For both ZFN-loaded LPs , we detected HA-tagged ZFNs of the predicted , full-length size ( ∼37 . 5 kDa ) , demonstrating that the majority of the virion-associated ZFNs was correctly released from the Gag and GagPol polypeptides during particle maturation . The identity of this band was confirmed by control analyses including plasmid-encoded HA-tagged ZFN protein ( Figure 1—figure supplement 1 ) . Nevertheless , both longer and shorter forms of the ZFNs were evident ( indicated by ZFN* in Figure 1B , C ) , suggesting that processing was not complete and that proteolytic cleavage had also occurred outside the inserted cleavage site , despite the fact that neither the ZFNs nor the PH domain contained naive HIV-1 cleavage sites . Release of ZFN and p24 ( Figure 1B , C , left and right panels , respectively ) from ZFN-PH-GagPol polypeptides was inhibited by treatment of the virus-producing cells with the HIV-1 protease inhibitor Saquinavir ( SQV ) , confirming that polypeptide cleavage , and the resulting release of ZFN proteins , was dependent on the HIV-1 protease . 10 . 7554/eLife . 01911 . 003Figure 1 . Targeted egfp gene disruption by lentiviral delivery of ZFN proteins . ( A ) Schematic representation of the composition of Gag and GagPol polypeptides encoded by ZFN-encoding packaging constructs ( top ) and the production of ZFN-loaded LPs ( bottom ) . Gag is composed of the N-terminal ZFN domain , an HIV-1 cleavage site ( SQNY/PIVQ ) , the phospholipase C-δ1 pleckstrin homology ( PH ) domain , matrix ( MA ) , capsid ( CA ) , nucleocapsid ( NC ) , and p6 , whereas Pol consists of protease ( PR ) , reverse transcriptase ( RT ) , and integrase harboring the D64V mutation ( IN-D64V ) . LPs harboring two types of ZFNs ( indicated by blue and green dots inside the virion ) are produced by co-transfecting 293T cells with pMD . 2G ( encoding VSV-G surface protein ) and pZFNL-PH-gagpol-D64V and pZFNR-PH-gagpol-D64V encoding ZFNL and ZFNR , respectively . ( B ) and ( C ) Analysis of the contents of LPs by Western blot using HA- and p24-specific antibodies . HA-tagged ZFNs were incorporated in this LP batch , allowing detection of ZFNs and ZFN derivatives ( left panel ) . ZFNs originating from non-intentional processing or cleavage at cryptic HIV-1 cleavage sites are indicated by arrows labeled with ZFN* . The same membrane was stripped and re-used for detection of p24 ( right panels ) . It is indicated below each panel whether 0 . 2 μM of the protease inhibitor Saquinavir ( SQV ) was included during LP production . ( D ) egfp gene disruption by protein transduction of ZFNs in HEK293-eGFPmut reporter cells as measured by Surveyor nuclease-based detection of DNA mismatches . Cells were harvested for analysis 24 hr posttransduction . Arrowheads point to the specific cleavage products . Quantified locus modification rates ( indel % ) are indicated below relevant lanes . ( E ) egfp gene disruption by ZFN proteins at different time points after transduction . HEK293-eGFPmut reporter cells were transduced with 300 ng p24 LP-ZFNLR ( gfp ) . Locus modification rates ( indel % ) are provided below the gel . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 00310 . 7554/eLife . 01911 . 004Figure 1—figure supplement 1 . Validation of the release of full-length ZFN protein within ZFN-loaded LPs . Western blot analysis of the cleavage pattern of ZFN proteins was carried out on cellular protein , as control , and on protein prepared from ZFN-loaded virions . For the control , 293T cells were co-transfected with pcDNA3 . 1-HA-ZFNL ( gfp ) and pcDNA3 . 1-HA-ZFNL ( gfp ) expressing HA-tagged left and right ZFNs , respectively ( left lane ) . Cell protein was purified for Western blot analysis 1 day after transfection . Right lanes show protein incorporated in LPs packaged with both HA-tagged left and right ZFNs . A short exposure of the blot is shown on the left , whereas a longer exposure of the blot is shown on the right . ZFNs originating from non-intentional processing are indicated by an asterisk ( * ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 004 Next , we transduced HEK293-eGFPmut reporter cells harboring a mutated egfp gene with increasing doses of LPs , designated LP-ZFNLR ( gfp ) , harboring both left and right egfp-targeting ZFNs ( ZFNL ( gfp ) and ZFNR ( gfp ) ) . ZFN protein transduction caused dose-dependent cleavage of egfp as measured by detection of NHEJ-induced sequence alterations using the Surveyor nuclease assay ( Figure 1D ) . The presence of indels inside egfp was confirmed by sequence analysis of cloned PCR fragments obtained from cells treated with 600 ng p24 LP-ZFNLR ( gfp ) ( Table 1—source data 1 ) . Among a total of 42 analyzed clones , 8 were found to contain egfp gene disruptions , providing an unbiased disruption frequency of 19% ( Table 1 ) . Notably , ZFN-directed cleavage and subsequent error-prone repair by NHEJ was detectable as early as 12 hr posttransduction ( Figure 1E ) . 10 . 7554/eLife . 01911 . 005Table 1 . Summary of the targeting rates obtained by sequence-based identification of locus-targeted indels after LP-directed delivery of ZFN proteinsDOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 00510 . 7554/eLife . 01911 . 006Table 1—source data 1 . Sequences of egfp gene disruption by ZFN protein transduction in the HEK293-eGFPmut reporter cells . Genomic DNA of HEK293-eGFPmut reporter cells transduced with 600 ng p24 LP-ZFNLR ( gfp ) was used as PCR template for amplification and subsequent cloning of the part of the egfp gene encompassing the region recognized by the two ZFNs . The wild-type sequence is shown at the top . The net change of length caused by indels is indicated to the right of each sequence . Green dashes represent deleted nucleotides , red lower case letters represent nucleotide substitutions , whereas blue lower case letters illustrate inserted nucleotides . If one particular sequence appeared in more than one clone , the exact number of clones with this particular sequence is provided in parenthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 00610 . 7554/eLife . 01911 . 007Table 1—source data 2 . Sequences of CCR5 gene disruption by ZFN protein transduction of HEK293 , NHDFs and HKs . Genomic DNA of cells transduced with 200 ng p24 LP-ZFNLR ( CCR5 ) was used as PCR template for amplification and subsequent cloning of a CCR5 amplicon encompassing the region recognized by the two ZFNs . The wild-type sequence is shown at the top . Types of indels are indicated as described in the legend to Table 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 00710 . 7554/eLife . 01911 . 008Table 1—source data 3 . Sequences of AAVS1 gene disruption by ZFN protein transduction in HEK293 , NHDFs and HKs . Genomic DNA of cells transduced with 200 ng p24 LP-ZFNLR ( AAVS1 ) was used as PCR template for amplification and subsequent cloning of an AAVS1 amplicon encompassing the region recognized by the two ZFNs . The wild-type sequence is shown at the top . Types of indels are indicated as described in the legend to Table 1—source data 1 . If one particular sequence appeared in more than one clone , the exact number of clones with this sequence is provided in parenthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 008Cell typeTarget locusHEK293Normal human dermal fibroblastsPrimary human keratinocytesegfp8/42 ( 19% ) *N/AN/ACCR57/42 ( 17% ) 8/46 ( 17% ) 11/45 ( 24% ) AAVS16/46 ( 13% ) 7/43 ( 16% ) 8/40 ( 20% ) Provided ratios indicate the number of alleles with indels out of the total number of analyzed alleles after treatment with 600 ng p24 LP-ZFNLR ( gfp ) , 200 ng p24 LP-ZFNLR ( CCR5 ) or 200 ng p24 LP-ZFNLR ( AAVS1 ) , respectively . *indicates that data were obtained in HEK293-eGFPmut reporter cells; N/A , not available . To evaluate the versatility of this approach , we produced LPs loaded with ZFNs targeting the endogenous CCR5 and AAVS1 loci ( designated as LP-ZFNLR ( CCR5 ) and LP-ZFNLR ( AAVS1 ) , respectively ) and exposed HEK293 cells , normal human dermal fibroblasts ( NHDFs ) , and primary human keratinocytes ( HKs ) to increasing dosages of these LPs . For both loci and in all cell types , we observed significant levels of disruption and found that such disruption occurred in an LP dose-dependent manner ( Figure 2 ) . In HEK293 cells and HKs , we consistently observed an extra band ( indicated by * in Figure 2 , left panels ) that was confirmed by sequencing to originate from the CCR5 Δ32 allele present in these cells . To evaluate the extent of cleavage and error-prone repair at the CCR5 and AAVS1 loci induced by LP-delivered pairs of ZFNs , we cloned and sequenced PCR products encompassing the targeted regions . For both loci and in all three cell types , numerous different indels could be identified in the targeted region of the genome ( Table 1—source data 2 and 3 ) . The rates of gene disruption by both CCR5- and AAVS1-targeting ZFNs ( using an amount of LPs corresponding to 200 ng p24 ) are summarized in Table 1 . For the three cell types , the percentage of sequenced clones harboring targeted disruptions in the CCR5 locus ranged from 17% in HEK293 cells and NHDFs to 24% in primary HKs , whereas from 13% to 20% of the analyzed AAVS1 loci , depending on the cell type , harbored indels . In summary , our findings demonstrate potent targeted gene disruption and knockout by lentiviral delivery of ZFN pairs targeting predetermined loci in the human genome . 10 . 7554/eLife . 01911 . 009Figure 2 . Targeted disruption of endogenous genes by protein transduction of ZFNs . HEK293 , normal human dermal fibroblasts ( NHDFs ) and primary human keratinocytes ( HKs ) were transduced with increasing amounts of LPs containing ZFNs targeting the CCR5 ( left ) and AAVS1 ( right ) loci . Cells were harvested for analysis 24 hr posttransduction . Arrowheads indicate specific cleavage products , whereas fragments marked with * were generated due to the presence of the CCR5 Δ32 allele in the analyzed cells . Quantified locus modification rates are indicated below relevant lanes . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 009 We next set out to engineer ZFN-loaded lentiviral vectors with the capacity to not only disrupt but also edit targeted genes through HR . By packaging a vector genome harboring a HR donor sequence into LPs , we reasoned that ZFNs and donor could be co-delivered in such ‘all-in-one’ lentiviral vectors ( Figure 3A ) . We generated , therefore , a donor vector plasmid carrying a 2 . 1-kb long sequence with homology to the egfp cassette present in the HEK293-eGFPmut cell line . This vector did not itself express eGFP , but served as donor for robust repair of egfp ( leading to eGFP production ) in HEK293-eGFPmut cells that were transfected with donor-containing plasmid DNA and treated with increasing dosages of LP-ZFNLR ( gfp ) ( Figure 3B ) . Hence , at a dose of 300 ng p24 , LP-ZFNLR ( gfp ) induced a level of repair that was comparable to that obtained by the efficient co-transfection of plasmid DNA encoding the two ZFNs ( Figure 3B , compare two right columns ) . Next , the donor sequence was delivered by integrase-defective lentiviral vectors ( IDLV/donor ) , allowing the reverse-transcribed vector to serve as a recombination donor during HR-directed editing . Co-delivery of IDLV/donor and LP-ZFNLR ( gfp ) to the HEK293-eGFPmut cells induced gene repair , whereas neither IDLV/donor nor LP-ZFNLR ( gfp ) alone induced significant levels of eGFP expression ( Figure 3C ) . 10 . 7554/eLife . 01911 . 010Figure 3 . Targeted gene editing by ‘all-in-one’ IDLVs . ( A ) Schematic representation of the production and intracellular action of IDLVs carrying two ZFNs ( indicated by blue and green dots ) and a vector with the donor sequence for HR-directed repair ( indicated in red as an RNA homodimer ) . Upon endosomal escape and uncoating , the donor sequence is reverse-transcribed to double-stranded DNA that is imported to the nucleus , where it serves as a donor for repair either in the form of linear DNA or as 1-LTR or 2-LTR circles ( only HR between linear DNA and the target is shown ) . Homologous sequences are highlighted in light yellow . The egfp gene harboring internal mutations ( indicated by a red box ) is repaired through ZFN-mediated cleavage and HR using the reverse-transcribed vector as a recombination donor . ( B–E ) Correction of the egfp gene by lentiviral delivery of ZFN proteins . Flow cytometric analysis was performed 4 days after transduction or transfection . In ( B ) , donor plasmid ( pLV/egfp-donor-fw ) was transfected 6 hr prior to ZFN protein transduction . Co-transfection of donor plasmid and ZFN-encoding plasmid DNA ( pZFN ) served as a positive control . In ( C ) , the donor sequence was provided by IDLV/donor ( MOI of 46 ) co-transduced with LP-ZFNLR ( gfp ) , whereas in ( D ) correction was achieved by co-transduction with two IDLVs ( IDLV-ZFNL ( gfp ) /donor and IDLV-ZFNR ( gfp ) /donor , respectively ) , both at an MOI of 9 , loaded each with one of the two egfp-targeting ZFNs . In ( E ) , ‘all-in-one’ IDLVs ( IDLV-ZFNLR ( gfp ) /donor ) induced potent gene correction . Gene editing was measured with virus loads ranging from an MOI of 2 ( corresponding to 37 ng p24 ) to an MOI of 34 ( corresponding to 600 ng p24 ) . IDLVs without the VSV-G surface protein as well as reporter cells pretreated with of 1 μM Bafilomycin A1 ( Baf A1 ) served as negative controls . ( F ) Targeted editing at endogenous CCR5 and AAVS1 loci . Schematic representation of PCR-based assay used for detection of gene editing at CCR5 and AAVS1 loci ( left panel ) . Primers are indicated above the edited target sequence . LS , linker sequence . Gene editing at CCR5 and AAVS1 loci in HEK293 cells and NHDFs , as confirmed by PCR ( right panel ) , was obtained with an MOI of 34 . PCR fragments amplified from the AAVS1 locus and the CCR5 locus served as controls for CCR5- and AAVS1-directed LS insertion , respectively . The error bars represent ±SD from three independent replicates of the experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 01010 . 7554/eLife . 01911 . 011Figure 3—figure supplement 1 . Determination of multiplicity of infection ( MOI ) of IDLVs carrying left and right ZFNs . Quantitative PCR analysis was carried out on total DNA isolated from 105 HEK293-eGFPmut reporter cells treated in three independent experiments with a load of IDLV/donor and IDLV-ZFNLR ( gfp ) /donor corresponding to 30 ng p24 . Genomic DNA was extracted 2 days after transduction . 48 ng genomic DNA was used for qPCR and each reaction was performed in triplicate . ( A ) Number of lentiviral DNA copies as measured by qPCR amplification of the WPRE element located in the lentiviral vector carrying the donor . ( B ) Determination of the number of copies of the albumin gene in the transduced cells . ( C ) Calculated multiplicity of infection ( MOI ) obtained in cells treated with a number of IDLVs corresponding to 30 ng p24 . The MOI was calculated as follows: 2 × lentiviral DNA copy number/albumin gene copy number . ( D ) Determination of the number of infectious units ( IU ) per ng p24 . The number of IU per ng p24 was calculated as follows: cell numbers × MOI/ng p24 . The error bars represent ±SD from three independent replicates of the experiments using for each IDLV three separate vector preparations . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 011 To investigate whether gene editing could be obtained with viral particles containing both donor RNA and ZFN proteins , HEK293-eGFPmut cells were co-transduced first with two separately produced IDLVs carrying either ZFNL ( gfp ) or ZFNR ( gfp ) . Both IDLVs carried the vector RNA containing the egfp donor sequence . Induced eGFP expression was observed only in cells treated with both IDLVs , indicative of active editing in about 4% of the treated cells , whereas neither of the two vectors alone caused gene editing ( Figure 3D ) . We then created ‘all-in-one’ IDLVs ( IDLV-ZFNLR ( gfp ) /donor ) carrying the egfp donor vector as well as both ZFNs , containing thus all the components necessary to facilitate ( i ) vector transfer , ( ii ) targeted formation of DSBs , and ( iii ) editing by HR . Treatment of the HEK293-eGFPmut cells with this vector resulted in potent induction of eGFP expression in a vector dose-dependent manner , leading to editing at most in more than 8% of the cells treated with ZFN-loaded IDLVs corresponding to 600 ng p24 ( Figure 3E ) . To provide an indication of the number of IDLV particles that was required to obtain such efficient gene correction , we determined the multiplicity of infection ( MOI ) of donor-containing IDLVs with and without virally incorporated ZFNs ( IDLV/donor and IDLV-ZFNLR ( gfp ) /donor , respectively ) . Copy numbers of lentiviral DNA and the genomic albumin gene in transduced cells were quantified by quantitative PCR ( Figure 3—figure supplement 1A , B ) on total DNA prepared from 1 × 105 HEK293-eGFPmut reporter cells transduced with IDLVs corresponding to 30 ng p24 . Using this approach , we determined MOIs of 9 . 2 and 1 . 7 for IDLV/donor and IDLV-ZFNLR ( gfp ) /donor , respectively ( Figure 3—figure supplement 1C ) , demonstrating that the overall transduction capacity of ZFN-loaded IDLVs was reduced more than fivefold relative to standard IDLVs . Hence , the most effective repair rate by IDLV-ZFNLR ( gfp ) /donor ( Figure 3E ) was obtained with an MOI of approximately 34 . It is directly deduced from these analyses that IDLV/donor samples contain 3 . 1 × 104 infectious units ( IU ) /ng p24 , whereas preparations of IDLV-ZFNLR ( gfp ) /donor contain an estimated 5 . 7 × 103 IU/ng p24 ( Figure 3—figure supplement 1D ) . Importantly , editing of the egfp gene was absent when the particles were not pseudotyped with VSV-G and therefore incapable of transducing the cells ( Figure 3E ) . Furthermore , pretreatment of the reporter cells with Bafilomycin A1 ( Baf A1 ) , an inhibitor of endosomal escape , markedly reduced the level of editing ( Figure 3E ) . We conclude that ‘all-in-one’ lentiviral vectors containing both ZFN proteins and the donor sequence facilitate high levels of targeted gene repair in a manner that depends on the dose and VSV-G-mediated endocytosis . To investigate the versatility of such engineered ‘all-in-one’ lentiviral virions as tools in genomic editing , we generated IDLVs containing pairs of ZFN proteins targeting the CCR5 and AAVS1 loci , respectively , and transduced HEK293 cells and NHDFs at an MOI of 34 . In these IDLVs , we packaged a vector genome containing a donor sequence with an internal linker sequence ( LS ) flanked by two homology arms matching either CCR5 or AAVS1 ( Figure 3F , left panel ) . By PCR amplification using primers recognizing the LS and sequences outside the homology arm , we observed site-directed editing facilitated by HR of both the CCR5 and AAVS1 loci in HEK293 cells and NHDFs ( Figure 3F , right panel ) . Notably , signs of editing were not evident in cells treated with preparations of IDLVs harboring only one of the ZFNs ( ZFNL ( CCR5 ) or ZFNL ( AAVS1 ) ) , indicating that homology-directed insertion of the linker was achieved only after delivery of both left and right ZFN proteins . To expand the repertoire of designed nucleases that are compatible with lentiviral protein transduction , we set out next to explore the possibility of delivering TALEN proteins incorporated into lentiviral particles . We designed a shuttle plasmid that is compatible with the Golden Gate assembly method ( Cermak et al . , 2011 ) and equivalent to pTAL3 , allowing us to isolate the DNA segment carrying the assembled TALEN sequence and insert it into GagPol in the correct reading frame ( Figure 4A ) . HEK293-eGFPmut reporter cells were transduced with increasing amounts of LP-TALENLR ( gfp ) carrying both the left and right TALEN proteins designed to recognize sequences flanking the mutations in the egfp gene . Gene disruption was identified by Surveyor nuclease-directed detection of mismatches in re-annealed PCR products ( Figure 4B ) and confirmed by sequencing of cloned PCR products . Hence , 3 out of 46 sequenced alleles ( 6 . 5% ) were found to contain indels as a result of the treatment with LP-delivered TALEN proteins ( Figure 4C ) . Moreover , by transfecting the cells with plasmid DNA containing the donor sequence prior to treatment of the cells with LP-TALENLR ( gfp ) , we found that 0 . 2% of the cells were eGFP positive ( as measured 15 days after transduction ) , whereas control treatments without LPs and donor DNA , respectively did not trigger eGFP expression ( Figure 4D ) . These findings demonstrated the capacity of lentivirally delivered TALEN proteins to facilitate targeted gene repair . Notably , Western blot analysis of viral particles harboring an HA-tagged version of one of the two TALEN proteins fused to the GagPol polypeptide unveiled a cleavage pattern indicative of substantial proteolytic cleavage within the TALEN domain of the fusion protein , leading to truncated versions of the protein ( indicated by TALEN* in Figure 4E ) . Although a low level of full-length TALEN protein with the same mobility as HA-tagged TALEN protein expressed in cells ( Figure 4E , upper band marked by ‘TALEN’ ) was indeed evident in the particles , these data indicated the presence of cryptic HIV-1 protease cleavage sites internally in the TALEN protein . This suggests that TALEN-directed gene disruption and repair may be further optimized by localizing and eliminating internal protease cleavage sites . 10 . 7554/eLife . 01911 . 012Figure 4 . Targeted egfp gene editing by lentiviral delivery of TALEN proteins . ( A ) Schematic representation of the construction of the TALEN-GagPol polypeptide expression construct . GoldyTALEN was assembled as the Golden Gate assembly method into a shuttle plasmid pC-Goldy-TALEN-PH and was then cut out and cloned into pGFP-PH-gagpol-D64V to get the destination construct pTALEN-PH-gagpol-D64V expressing polypeptides composed of GoldyTALEN and GagPol connected by the HIV-1 protease cleavage site SQNY/PIVQ . NLS , SV40 nuclear localization signal , Δ152 , 152 amino acids deletion from the wild-type TALE protein , RVDs , repeat variable di-residues , +63 , 63 amino acids following the last repeat , PH , phospholipase C-δ1 pleckstrin homology domain . ( B ) egfp gene disruption by protein transduction of TALENs in HEK293-eGFPmut reporter cells as measured by Surveyor nuclease assay 24 hr posttransduction . ( C ) Sequences of egfp gene disruption by TALEN protein transduction in the HEK293-eGFPmut reporter cell line . Genomic DNA of HEK293-eGFPmut reporter cells transduced with 160 ng p24 LP-TALENLR ( gfp ) was used as PCR template for amplification and subsequent cloning of the part of the egfp gene encompassing the region recognized by the two TALENs . The wild-type sequence is shown at the top . The net change of length caused by the indels is indicated to the right of each sequence . Green dashes represent deleted nucleotides , whereas blue lower case letters illustrate inserted nucleotides . Three alleles out of 46 sequenced clones were found to be disrupted ( disruption frequency: 6 . 5% ) . ( D ) Targeted egfp gene repair in HEK293-eGFPmut reporter cells . Cells were transfected with 1 . 8 μg donor plasmid ( pLV/egfp-donor-fw ) and transduced with 160 ng p24 of LP-TALENLR ( gfp ) 6 hr later . Cells treated only with LP-TALENLR ( gfp ) or donor served as negative controls . eGFP expression was analyzed by flow cytometry 15 days posttransduction . ( E ) Analysis of the contents of LP-HA-TALENR ( gfp ) by Western blot using HA-specific antibody . The left lane shows protein derived from 293T cells expressing right HA-TALEN ( gfp ) from transfected plasmid pcDNA3 . 1-Goldy-HA-TALENR ( gfp ) . The expected size of the full-length TALEN is indicated , and truncated TALEN derivatives originating from non-intentional cleavage at cryptic , internal HIV-1 cleavage sites are indicated by an asterisk ( * ) . The error bars represent ±SD from three independent replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 012 Nuclease protein transduction facilitates both coordinated delivery of the two proteins and a short boost of activity in transduced cells . As other delivery techniques based on intracellular nuclease production may lead to prolonged nuclease activity , we hypothesized that ZFNs delivered in LPs would have low off-target relative to on-target activity . To test this notion , we compared the activity of CCR5-directed ZFNs in the neighboring CCR2 locus after ZFN delivery into HEK293 cells by protein transduction and plasmid DNA transfection , respectively . The CCR2 locus has high homology to CCR5 and has been identified as a major off-target site of CCR5-targeting ZFNs ( Perez et al . , 2008; Lee et al . , 2010 ) . For the CCR5-targeting ZFNs employed in this study , only a single nucleotide differs between the CCR5 and CCR2 recognition sites of each ZFN . In initial examinations of the potential off-target activity at the CCR2 locus , we designed a PCR method that would detect NHEJ between fragments created by ZFN activity at both the CCR5 and CCR2 locus leading to deletion of the intervening sequence ( Figure 5A ) . Indeed , we observed off-target cleavage at the CCR2 locus by both protein transduction and plasmid transfection but not in the non-treated cells ( Figure 5B ) , indicating that both LP-delivered and DNA-encoded ZFNs induced simultaneous double-stranded DNA breaks within the CCR2 and CCR5 loci . 10 . 7554/eLife . 01911 . 013Figure 5 . On- and off-target activity by CCR5-targeting ZFNs delivered by LP protein transduction or plasmid transfection . ( A ) Schematic representation of the PCR-based assay used for detection of simultaneous on-target and off-target cleavage at the CCR5 and CCR2 loci , respectively . Small blue and red boxes above each locus indicate the location of the two ZFN recognition sites in the CCR5 locus and the highly homologous sequences located within the CCR2 locus . Horizontal arrows indicate the location of sequences recognized by the primers used . ( B ) Detection of off-target cleavage by PCR-based detection of CCR2-CCR5 fusion fragments . HEK293 cells were seeded at a density of 1 × 105 cells/well on day 1 and were on day 2 either transduced with increasing amounts of LP-ZFNLR ( CCR5 ) ( 50 , 100 , 200 ng p24 , respectively; indicated by ‘LP’ ) or co-transfected with increasing amounts of the two ZFN-encoding plasmids ( 50 ng + 50 ng , 100 ng + 100 ng , 200 ng + 200 ng , respectively; indicated by ‘plasmid’ ) . On day 3 , cells were harvested for genomic DNA purification . ( C ) Disruption within the CCR5 and CCR2 loci after LP- and plasmid-directed ZFN delivery . HEK293 cells were treated with increasing amounts of LP-ZFNLR ( CCR5 ) and ZFN-encoding plasmids as in ( B ) . Sequence changes introduced by NHEJ-directed repair were identified by Surveyor nuclease-directed detection of mismatches in re-annealed PCR products . Analyses of CCR5- and CCR2-directed cleavage were performed on the same genomic DNA samples . Arrowheads indicate specific cleavage products . Quantified locus modification rates are indicated below each gel . ( D ) Distinct levels of CCR2 gene disruption using experimental conditions that support similar levels of CCR5 disruption after LP- and plasmid-directed ZFN delivery . HEK293 cells were treated by LP transduction ( 80 ng p24 ) or plasmid transfection ( 300 ng + 300 ng ) , essentially as described in ( B ) , facilitating similar levels of gene disruption in the CCR5 locus . Surveyor nuclease assays were performed as in ( C ) except for the use of an alternative CCR5 primer set . Quantified locus modification rates are indicated below relevant lanes . DOI: http://dx . doi . org/10 . 7554/eLife . 01911 . 013 Next , we compared the on-target activity in the CCR5 locus achieved by the two delivery methods . As shown in Figure 5C ( left panel ) by Surveyor nuclease-directed detection of mismatches in re-annealed PCR products , protein transduction by lentiviral particles induced under the given conditions significantly higher rates of on-target gene disruption than plasmid transfection . These findings also confirmed the dose-dependent activity of LP-delivered ZFNs . Similarly , when we analyzed for mismatches at the CCR2 locus , an increasing dose of ZFN-loaded LPs was seen to trigger increased off-target activity ( Figure 5C , right panel ) . Importantly , however , a similar level of off-target activity was observed for all doses of transfected plasmid DNA , indicating that LP-delivered ZFNs even at levels that induced markedly higher levels of on-target disruption did not induce increased levels of off-target activity at the CCR2 locus . These findings showed a reduced relative off-target activity of lentivirus-delivered ZFNs . To simplify and strengthen the comparison between LP-delivered and plasmid-encoded ZFNs , we established experimental conditions under which the efficiency of on-target gene disruption by LP transduction and plasmid transfection was similar ( Figure 5D , left panel ) . By subsequent analyses of off-target disruption within the CCR2 locus of the same cells , we detected disruption ( 3% ) induced by plasmid-encoded ZFNs , whereas indels and mismatches were not evident after ZFN protein transduction . In summary , our findings lend strong support to the notion that locus targeting by LP-directed protein transduction is both effective and favored by an improved on-target/off-target cleavage ratio .
Engineered nucleases have become driving forces in genomic engineering ( Gaj et al . , 2013 ) . To improve the safe use of such nucleases , it is crucial to scrutinize novel delivery techniques . Lentiviral virions have previously been engineered to incorporate foreign proteins to study different aspects of HIV-1 biology and antiviral strategies ( Wu et al . , 1995; Fletcher et al . , 1997; Hermida-Matsumoto and Resh , 2000; Hubner et al . , 2007 ) , demonstrating the capability of ferrying protein effectors into target cells . Indeed , delivery of Cre recombinase and I-SceI meganuclease has been achieved by fusing the heterologous protein-of-interest to the HIV-1 accessory protein Vpr ( Michel et al . , 2010; Izmiryan et al . , 2011 ) . However , none of these enzymes are programmable and therefore can only edit a single artificial target , which would need to be introduced in the target cells prior to editing . Also , as both Cre and I-SceI function as monomers they may be easily adaptable for alternative delivery strategies . In consideration of the increasing focus on engineered nucleases , it is crucial to investigate alternative nuclease delivery strategies . With the goal of effectively co-delivering two different nuclease proteins and the donor for repair , we turned to a strategy based on transferring the heterologous protein as part of the lentiviral Gag polypeptide . Derivatives of Gag polypeptides are the most abundant proteins in HIV-1-derived particles , and each virion has been estimated to contain approximately 5000 copies of Gag-derived proteins like p24 ( Swanson and Malim , 2008 ) . Hence , such strategy for transduction of proteins is superior in terms of cargo load . Though IDLVs have been exploited to deliver ZFN-encoding genes ( Lombardo et al . , 2007 ) , there are some clear inherent limitations to this approach . Firstly , random integration of the vector can lead to insertional mutagenesis and sustained expression of ZFNs may cause cytotoxicity or chromosomal aberrations . Secondly , two separate IDLVs are necessary to provide the two ZFNs and a third IDLV needs to be included if a donor template is required for homologous recombination . Single IDLVs encoding both ZFNs have recently been described , but these do not perform as well as two separate IDLVs ( Joglekar et al . , 2013 ) . Thirdly , due to the highly repetitive sequences in TALENs there are major difficulties in delivering these by IDLVs due to an inherent propensity of lentiviral vectors to undergo recombination during reverse transcription ( Mikkelsen and Pedersen , 2000; Holkers et al . , 2013 ) . In this study , we have shown the feasibility of exploiting lentivirus-derived particles as vehicles for different sets of designer nuclease proteins that are released from the Gag polypeptide upon virion maturation . This approach facilitates a short boost of protein activity that drives immediate targeted gene disruption ( within 12 hr ) in up to near one-fourth of the target alleles in a population of cells without the risks of permanently inserting copies of ZFN- or TALEN-encoding genes into the genome . Moreover , as an alternative to protocols based on transfection of in vitro-transcribed RNA ( Meng et al . , 2008 ) or the cell-penetrating capability of purified ZFNs ( Gaj et al . , 2012 ) , a lentiviral protein delivery approach benefits from effective lentiviral transduction and can easily be further customized by adapting different lentiviral pseudotypes to direct protein transduction to specific cell types of interest . The ability to edit predetermined genomic loci , rather than disrupt loci , may be crucial for future therapeutic applications of designer nucleases . We have shown that ZFN proteins ( left and right ) and HR donor sequences can be incorporated into a single preparation of IDLVs and trigger effective gene repair in transduced cells . At an MOI of approximately 34 , such ‘all-in-one’ vectors edited the reporter gene in up to 8% of treated cells . We also provide proof-of-principle that ‘all-in-one’ vectors can target the insertion of additional genetic sequences in the form of a linker sequence , suggesting that HR-mediated insertion facilitated by transduced ZFN proteins has a potential as a tool for targeted lentiviral insertion . The level of gene repair obtained with IDLVs carrying both ZFNs and the donor template is higher than previously reported repair rates obtained with the same pair of ZFNs under similar conditions using an analogous reporter system ( Urnov et al . , 2005 ) . Although a direct comparison is not possible , we reason that a high rate of repair after protein transduction reflects the combined capacity of LPs to ( i ) coordinately deliver pairs of nucleases by protein transduction and ( ii ) provide a donor for homologous recombination which is associated with the pre-integration complex and therefore is actively and efficiently translocated to the nucleus ( Coluccio et al . , 2013 ) . We have previously observed that the uptake of LP-delivered heterologous proteins is extremely limited relative to the very high levels of nuclear protein that can be achieved by standard plasmid DNA transfection ( Cai et al . , 2014 ) . Based on these findings , we currently favor a model by which editing is further supported by the co-localization of ZFN proteins and the donor template within the pre-integration complex . In several aspects , such model mimics the ability of conventional lentiviral vectors to carry the integrase protein along with the reverse-transcribed vector DNA into the nucleus . In comparison to ZFNs , TALENs have higher modularity and are becoming more routinely used due to the ease of production . However , TALENs are relatively large proteins ( ∼100 kDa ) , and this may challenge their delivery . In addition , lentiviral delivery of TALEN-encoding genes is further complicated by frequent recombination between the highly similar sequence repeats encoding the DNA-binding domain ( Holkers et al . , 2013 ) , and lentiviral delivery of genes encoding re-coded TALE variants has been reported so far only for transcriptional activation ( Yang et al . , 2013 ) . In this study , we show that TALENs can be delivered as proteins by LPs , facilitating gene disruption or , in the presence of a donor template , repair in a reporter cell line . However , the majority of virally incorporated TALEN subunits are cleaved internally by the HIV-1 protease , which is suspected to limit the overall efficiency . Further work is necessary to identify and eliminate the cryptic viral protease cleavage site to improve the delivery of virus-incorporated TALEN proteins . It is likely that efficient gene targeting can be achieved in a narrow window of ZFN exposure and that activity outside this time window can cause toxicity without improving gene targeting efficiencies ( Porteus and Baltimore , 2003; Pruett-Miller et al . , 2009 ) . To analyze this aspect , we compared ZFN-directed disruption in the CCR5 target locus and the nearby and highly similar CCR2 locus . Interestingly , LPs loaded with CCR5-directed ZFNs caused disruptions in the CCR5 locus more effectively than ZFNs expressed from transfected plasmid DNA . In the same cells , disruptions in the off-target CCR2 locus were evident with all concentrations of plasmid but detectable only with the highest and most effective dose of LPs . Importantly , under experimental conditions that allowed similar rates of disruption using LP-directed ZFN delivery and plasmid-encoded ZFN production , off-target activity within the CCR2 locus was detected only after ZFN-encoding plasmid delivery . These findings support the notion that LP-delivered ZFNs target a safe harbor in the human genome with an improved on-target/off-target cleavage ratio . We reason that such improvement is a combined effect of efficient transduction and the short-term exposure to transduced ZFN proteins . With ZFNs and TALENs as early driving forces , designer nucleases stand out as key tools in genomic editing with implications for the development of future gene repair therapies . We anticipate that lentiviral protein transduction of nuclease proteins represents a versatile alternative to current nuclease delivery techniques and will support continued efforts to promote safe genome editing therapies . Uniquely , this approach allows the delivery of the tools for targeted gene editing in a single combined gene and protein vehicle . Likely , genomic engineers will strive to adapt LPs as carriers of Cas9 proteins for RNA-guided genome editing .
Throughout this work , the annotation ‘LP-ZFNLR ( target ) ’ was used to designate lentiviral particles that contain left and right ZFN proteins but do not carry a lentiviral vector genome . Same nomenclature applies to TALEN . A similar type of particle harboring , for example , only the left ZFN was accordingly designated LP-ZFNL ( target ) . When a lentiviral genome was included for example as a carrier of a HR donor sequence , we referred to this vector as an IDLV ( due to presence of the IN D64V mutation , rendering the vector integrase-defective ) . These vectors were named with the specification of the incorporated ZFNs and the donor that were included . IDLV-ZFNLR ( gfp ) /donor , for example , contains both egfp-directed ZFNs ( left and right ) and RNAs of which reverse transcription products carry a donor sequence with homology to egfp . Of note , we have consistently observed that virions harboring Gag fused N-terminally to foreign proteins are strongly reduced in their capability of transferring vector RNA ( most likely due to complications during reverse transcription ) , but that this capacity can be restored by including unfused GagPol ( encoded by pMDlg/pRRE-D64V ) in the virions . Preparations of LPs and IDLVs were produced as follows . On day 1 , 293T cells were plated at a density of 6 × 104/cm2 . On day 2 , cells were transfected with calcium phosphate precipitates of plasmid DNA . To produce LPs , 293T cells in 15-cm dishes were transfected with 10 μg pMD . 2 G , 30 μg pZFNL ( target ) -PH-gagpol-D64V ( targeting either gfp , CCR5 , or AAVS1 ) and 30 μg pZFNR ( target ) -PH-gagpol-D64V ( targeting either gfp , CCR5 , or AAVS1 ) . LPs containing TALENs were produced accordingly except using pTALENL ( gfp ) -PH-gagpol-D64V and pTALENR ( gfp ) -PH-gagpol-D64V instead of corresponding ZFN constructs . To produce IDLVs that did not incorporate foreign protein , cells in 15-cm dishes were transfected with 9 . 07 μg pMD . 2G , 7 . 26 μg pRSV-Rev , 31 . 46 μg pMDlg/pRRE-D64V , and 31 . 46 μg pLV/egfp-donor-re . To produce IDLVs incorporating ZFN proteins , 293T cells plated in 15-cm dishes were transfected with 9 . 07 μg pMD . 2G , 7 . 26 μg pRSV-Rev , 15 . 73 μg pMDlg/pRRE-D64V , 7 . 8 μg pZFNL ( target ) -PH-gagpol-D64V ( targeting either gfp , CCR5 , or AAVS1 ) , 7 . 8 μg pZFNR ( target ) -PH-gagpol-D64V ( targeting either gfp , CCR5 or AAVS1 ) , and 31 . 46 μg donor-containing vector ( either pLV/egfp-donor-re , pLV/CCR5-donor-LS , or pLV/AAVS1-donor-LS ) . After transfection , the medium was refreshed on day 3 , and supernatants were harvested on day 4 and day 5 , passed through a 0 . 45-μm filter ( Millipore , Billerica , Massachusetts ) , and ultracentrifuged at RPM 25 , 000 at 4°C for 2 hr . Pellets were re-suspended in PBS and stored at −80°C . Concentrations of HIV-1 p24 were measured by ELISA ( Zeptometrix , Buffalo , New York ) according to the manufacturer’s protocol . pT2/CMV-egfp-mut . SV40-neo containing mutated egfp was constructed by inserting a BsrGI/SacII-digested overlap PCR product into pT2/CMV-egfp ( s ) -SV40 . neo vector ( Staunstrup et al . , 2012 ) . The overlap PCR product was generated by amplifying first two fragments from pT2/CMV-egfp ( s ) . SV40-neo ( using primer sets 4684-YJ001R and YJ002F-BGHpA ) and fusing these by overlap PCR with primers 4684 and BGHpA ( primers are listed in Supplementary file 1 ) . The resulting egfp cassette harbored both a stop mutation and a frameshift mutation in the sequence flanked by the two ZFN recognition sequences . The lentiviral vector plasmid harboring the egfp donor sequence in the forward orientation ( pLV/egfp-donor-fw ) was created by inserting a PCR product amplified from pT2/CMV-egfp ( s ) . SV40-neo with primers YJ003F and YJ004R into ApaI/XhoI-digested pLV/RSV-SB100X ( Moldt et al . , 2011 ) . By this procedure , the RSV-SB100X cassette was completely removed and exchanged with the donor sequence . A lentiviral plasmid vector with the egfp donor in the reverse orientation ( pLV/egfp-donor-re ) was created by amplifying the donor sequence from pLV/egfp-donor-fw with the primer set YJ150F-YJ151R and inserting the donor sequence into ApaI/XhoI-digested pLV/egfp-donor-fw . Each resulting donor vector contained a truncated egfp gene ( with an upstream 37-bp deletion from the start codon ) , the SV40 promoter , and a shortened neo gene ( with a 46-bp deletion as counted from the third position of the stop codon ) together constituting a 2100-bp cassette with homology to its target sequence within genomically inserted pT2/CMV-egfp-mut . SV40-neo . Sequences encoding the zinc-finger ( ZF ) of known ZFNs targeting the CCR5 and AAVS1 loci were synthesized by GenScript according to sequences that have been provided in the literature ( Lombardo et al . , 2007; Hockemeyer et al . , 2009 ) . To introduce these sequences into a ZFN context , AflII/BamHI-fragments containing the ZF sequence were inserted into AflII/BamHI-digested pcDNA3 . 1-ZFNL ( gfp ) ( Urnov et al . , 2005 ) which was kindly provided by Michael C Holmes ( Sangamo Biosciences , Richmond , California ) . The resulting plasmids , in which the ZF sequences were fused to the FokI nuclease domain , were designated pcDNA3 . 1-ZFNL ( CCR5 ) , and pcDNA3 . 1-ZFNR ( CCR5 ) , pcDNA3 . 1-ZFNL ( AAVS1 ) , and pcDNA3 . 1-ZFNR ( AAVS1 ) . We subsequently fused the ZFN-coding sequences to the 5′-end of gag gene in pGFP-PH-gagpol , which was kindly provided by Jun Komano , National Institute of Infectious Diseases , Tokyo , Japan . Prior to insertion of the ZFN sequences , the D64V mutation was introduced in the HIV-1 integrase sequence in pGFP-PH-gagpol ( creating pGFP-PH-gagpol-D64V ) to abolish the conventional lentiviral integration capability . Sequences encoding the left and right gfp-targeted ZFNs were amplified from pcDNA3 . 1-ZFNL ( gfp ) and pcDNA3 . 1-ZFNR ( gfp ) , respectively , with primers YJ112F and YJ113R . Sequences encoding HA-tagged versions of these ZFNs , HA-ZFNL ( gfp ) and HA-ZFNR ( gfp ) , were amplified from pcDNA3 . 1-ZFNL ( gfp ) and pcDNA3 . 1-ZFNR ( gfp ) with primers YJ168F and YJ113R . ZFNL ( CCR5 ) - and ZFNR ( CCR5 ) -encoding sequences were amplified from pcDNA3 . 1-ZFNL ( CCR5 ) and pcDNA3 . 1-ZFNR ( CCR5 ) , respectively , with primers YJ177F and YJ113R . ZFNL ( AAVS1 ) - and ZFNR ( AAVS1 ) -encoding sequences were amplified from pcDNA3 . 1-ZFNL ( AAVS1 ) and pcDNA3 . 1-ZFNR ( AAVS1 ) with the primer sets YJ175F-YJ113R and YJ176F-YJ113R , respectively . The resulting PCR products were digested with XmaI and cloned into AgeI/AccIII-digested pGFP-PH-gagpol-D64V to create pZFNL ( gfp ) -PH-gagpol-D64V , pZFNR ( gfp ) -PH-gagpol-D64V , pHA-ZFNL ( gfp ) -PH-gagpol-D64V , pHA-ZFNR ( gfp ) -PH-gagpol-D64V , pZFNL ( CCR5 ) -PH-gagpol-D64V , pZFNR ( CCR5 ) -PH-gagpol-D64V , pZFNL ( AAVS1 ) -PH-gagpol-D64V , and pZFNR ( AAVS1 ) -PH-gagpol-D64V . To construct pcDNA3 . 1-HA-ZFNL ( gfp ) and pcDNA3 . 1-HA-ZFNR ( gfp ) for cellular production of HA-tagged ZFNs , HA-ZFN fragments were amplified from pHA-ZFNL ( gfp ) -PH-gagpol-D64V and pHA-ZFNR ( gfp ) -PH-gagpol-D64V , respectively , using primer pair YJ211F-YJ212R and inserted into AflII/XhoI-digested pcDNA3 . 1-ZFNL ( gfp ) . A lentiviral vector plasmid , pLV/CCR5-donor-LS , with a CCR5-targeted donor sequence ( consisting of two 550-bp homology arms ) and an internal linker sequence ( LS ) was constructed by PCR amplification of the homology arms from genomic DNA of NHDFs with the primer sets YJ193F-YJ206R and YJ195F-YJ196R . The two arms were fused by overlap PCR using primers YJ193F and YJ196R , generating a PCR product with the internal 24-bp LS sequence . The ApaI/XhoI-digested PCR product was inserted into ApaI/XhoI-digested pLV/RSV-SB100X ( Moldt et al . , 2011 ) , allowing substitution of the RSV-SB100X cassette with the CCR5 donor sequence . pLV/AAVS1-donor-LS containing two 600-bp AAVS1 homology arms flanking an internal LS was constructed by a similar method by which two PCR products were generated with the primer pairs YJ200F-YJ201R and YJ202F-YJ203R . The MluI/XhoI-digested overlap PCR product ( generated with primers YJ200F and YJ203R ) was inserted into MluI/XhoI-digested pLV/egfp-donor-fw . To generate pC-Goldy-TALEN-PH , YJ213F-YJ186R-amplified fragments from pC-GoldyTALEN ( Bedell et al . , 2012 ) were digested by SpeI/XmaI and inserted into XbaI/XmaI digested pC-GoldyTALEN . pC-Goldy-HA-TALEN-PH was generated by the same way except using primer YJ218F instead of YJ213F . The repeats of TAL effector were assembled as described except using pC-Goldy-TALEN-PH or pC-Goldy-HA-TALEN-PH to replace pTAL3 . The reading frames of the gfp-targeting TALENs were cut out by AgeI/XmaI and inserted into the destination vector pGFP-PH-gagpol-D64V digested with AgeI/AccIII . The resulting constructs , pTALENL ( gfp ) -PH-gagpol-D64V , pTALENR ( gfp ) -PH-gagpol-D64V , pHA-TALENL-PH- ( gfp ) -gagpol-D64V , and pHA-TALENR-PH- ( gfp ) -gagpol-D64V encode polypeptides containing both the TALEN and lentiviral GagPol connected by an HIV-1 cleavage site SQNY/PIVQ . All these constructs contain 63 amino-acid residues at the C-terminus of the TALEN repeat region . pTALEN-PH-gagpol-D64V and a version with an HA-tagged TALEN , pHA-TALEN-PH-gagpol-D64V , were assembled ( Cermak et al . , 2011 ) and cloned as shown in Figure 4A . The repeat variable di-residue sequences of GoldyTALEN for egfp targeting were ‘NN HD NG NG HD NI NN HD HD NN HD NG NI HD HD’ and ‘NN NN HD NN NN NI HD NG NG NN NI NI NN NI NI’ , respectively . pYJ63-HA-PH-gagpol-D64V was constructed by inserting fragments that were amplified from pC-Goldy-HA-TALEN-PH with primer pair YJ256F-YJ257R into pGFP-PH-gagpol-D64V . Inserts and vector were digested by AccIII and AgeI/AccIII , respectively . pcDNA3 . 1-Goldy-HA-TALEN was generated by inserting fragments amplified from pYJ63-HA-PH-gagpol-D64V with YJ553F-YJ554R into pcDNA3 . 1-ZFNL ( gfp ) . Inserts and vector were cut by AflII/SalI and AflII/XhoI , respectively . HA-tagged TALEN , pcDNA3 . 1-Goldy-HA-TALENR ( gfp ) , was assembled as Golden Gate cloning by using pcDNA3 . 1-Goldy-HA-TALEN in place of pTAL3 . To determine gene disruption frequencies , we utilized the mismatch detection assay based on the Surveyor nuclease ( Transgenomic , Omaha , Nebraska ) according to manufacturer’s instructions . Briefly , cells treated with ZFN-loaded LPs were harvested 24 hr after transduction unless specific time points were indicated . Genomic DNA was extracted by saturated NaCl and precipitated by absolute ethanol . egfp , CCR5 , AAVS1 and CCR2 fragments were amplified with primer sets 4684-YJ170R , YJ207F-YJ208R ( or YJ207F-YJ225R ) , YJ222F-YJ223R and YJ220F-YJ359R , respectively , using the Phusion High-Fidelity PCR Master Mix ( Thermo Scientific , Waltham , Massachusetts ) . After denaturation and re-annealing , each sample was digested by 1 μl Surveyor nuclease plus 1 μl enhancer ( Transgenomic ) . Cleavage products were separated by gel electrophoresis in 1 . 5% agarose gel and stained by ethidium bromide . Quantification was based on relative band intensities . Indel percentage was determined by the formula 100 × ( 1− ( 1−fraction cleaved ) 1/2 ) , wherein the fraction cleaved is the sum of the cleavage product peaks divided by the sum of the cleavage product and parent peaks . To analyze gene disruption by sequencing , we PCR-amplified and cloned target fragments allowing for separate sequencing of plasmid preparations from separate bacterial clones . Fragments encompassing the egfp , CCR5 and AAVS1 target sites were amplified using primer sets YJ180F-YJ181R , YJ237F-YJ238R , and YJ240F-YJ244R , respectively , from genomic DNA of cells transduced with 600 ng p24 in case of LP-ZFNLR ( gfp ) and with 200 ng p24 in the case of LP-ZFNLR ( CCR5 ) and LP-ZFNLR ( AAVS1 ) . The resulting PCR products were digested with EcoRI and BamHI ( for egfp and CCR5 ) or EcoRI and BglII ( for AAVS1 ) , and cloned into EcoRI/BamHI-digested pUC57 . Sequence analysis was performed on individual cloned transformants using primer YJ226F . Cells transduced with ‘all-in-one’ vectors IDLV-ZFNLR ( CCR5 ) /donor-LS and IDLV-ZFNLR ( AAVS1 ) /donor-LS were harvested 9 days posttransduction . Genomic DNA was purified as described above . PCR primer sets were designed to identify gene targeting as shown in Figure 3F ( YJ191F-YJ208R for IDLV-ZFNLR ( CCR5 ) /donor-LS and YJ191F-YJ205R for IDLV-ZFNLR ( AAVS1 ) /donor-LS ) . Fragments of AAVS1 and CCR5 loci were amplified by primer pair YJ222F-YJ223R and YJ224F-YJ225R , respectively , as loading controls . HEK293 ( human embryonic kidney cells ) , 293T , and NHDFs ( normal human dermal fibroblasts ) were cultured in Dulbecco’s Modified Eagle’s Medium ( Lonza , Basel , Switzerland ) . Culture medium was supplemented with 10% fetal calf serum , 100 U/ml penicillin , 100 μg/ml streptomycin , and 250 μg/ml L-glutamine . Primary human keratinocytes ( HKs ) were grown in serum-free keratinocyte medium ( Gibco BRL-Life Technologies , Carlsbad , California ) supplemented with bovine pituitary extract ( 25 μg/ml ) and recombinant epidermal growth factor ( 0 . 1–0 . 2 ng/ml ) . All cells were cultured at 37°C and 5% ( vol/vol ) CO2 . To determine the activity of LP-incorporated ZFNs , we established a reporter system in HEK293 cells ( HEK293-eGFPmut reporter cells ) based on the genomic insertion of a Sleeping Beauty DNA transposon ( carried by the plasmid pT2/CMV-egfp-mut . SV40-neo ) harboring an egfp cassette gene from which expression was inhibited by stop- and frameshift-mutations flanked by the two ZFN recognition sequences . To examine LP-ZFNLR ( gfp ) -induced homologous repair , HEK293-eGFPmut reporter cells were seeded in six-well plates ( 1 × 105 cells/well ) , and on the following day , cells were co-transfected with 1 . 8 μg pLV/egfp-donor-fw , 100 ng pcDNA3 . 1-ZFNL ( gfp ) , 100 ng pcDNA3 . 1-ZFNR ( gfp ) as a positive control , or only transfected with 1 . 8 μg pLV/donor-egfp-fw and 6 hr later transduced with LP-ZFNLR ( gfp ) . To measure homologous repair induced by IDLV-ZFNLR ( gfp ) /donor , HEK293-eGFPmut reporter cells were seeded in six-well plates at a density of 1 × 105 cells/well and the next day transduced with VSV-G-positive or VSV-G-negative lentiviral vectors . Reporter cells pretreated with 1 μM Bafilomycin A1 ( Baf A1 ) for 30 min before transduction served as an additional negative control . 4 days after transduction , cells were washed with PBS before being harvested and fixed with 4% paraformaldehyde . Data were collected on a FACSCalibur ( Becton Dickinson , Franklin Lakes , New Jersey ) and analyzed with FlowJo ( Tree Star , Ashland , Oregon ) . To analyze the protein contents of engineered LPs , LP-containing supernatants were centrifuged through a 20% ( wt/vol in PBS ) sucrose cushion . LPs were lysed in the presence of a protease inhibitor . The viral proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to the PVDF membrane . Membranes were blocked by 5% fat-free milk dissolved in TBS/0 . 05% Tween-20 for 1 hr and incubated with an HA monoclonal antibody ( Covance , Princeton , New Jersey ) overnight at 4°C . The membranes were incubated with anti-mouse secondary antibodies ( Dako , Glostrup , Denmark ) and visualized by enhanced chemiluminescence ( ECL ) using a HRP substrate ( Thermo Scientific ) . The HA monoclonal antibody was washed away by stripping buffer ( Thermo Scientific ) , and the PVDF membrane was re-used for incubation with a HIV-1 p24 polyclonal antibody ( Thermo Scientific ) and later peroxidase-conjugated anti-rabbit secondary antibody ( Dako ) . Quantitative PCR ( qPCR ) on lentiviral DNA was performed as previously described ( Bak et al . , 2013 ) . Briefly , qPCR reactions were run using Maxima Probe qPCR Master Mix ( Fisher Scientific , Waltham , Massachusetts ) with primers and probe specific for the WPRE sequence present in the lentiviral vector . Primers and probe specific for the albumin gene were used to quantify the cell number in each qPCR reaction assuming an albumin gene copy number of two per genome . WPRE and albumin DNA copy numbers were determined using standard curves with serially diluted plasmid containing the two targets . For WPRE , the pCCL-PGK-Puro-H1-MCS plasmid was used ( Bak et al . , 2011 ) and for albumin we used pAlbumin kindly provided by Didier Trono ( Addgene plasmid #22037 ) . The multiplicity of infection ( MOI ) was calculated as: 2 × copy numbers of lentiviral DNA/copy numbers of albumin . The infectious units/ng p24 was calculated as: number of cells × MOI/ng p24 . | Altering the genetic code of a living organism to produce certain desirable outcomes is the goal of genetic engineering . The field builds on a long history of human attempts to alter genetics , from selective breeding of crops and livestock to genetically modified organisms and gene therapies . Researchers routinely use gene editing to create ‘knock-out’ mice in which a particular gene is turned off: the researchers can learn more about the function of this gene by watching what happens when it is absent . As gene editing techniques have grown more sophisticated , they have become an increasingly promising tool for treating diseases that are caused by gene mutations . The aim of this work is to replace faulty genes with genes that work properly . However , it has been difficult to adapt genetic engineering techniques so that they can be used safely in humans . Scientists have created customized enzymes called nucleases that can remove specific genes , but it has been a challenge to get these nucleases into cells in the first place . A virus can be used to deliver the genes that encode these nucleases into the DNA of a cell , but this approach can lead to the production of too many nucleases and to the removal of more genes than intended . Now Cai et al . have developed a ‘hit-and-run’ method for getting the nucleases into cells and making them active only for a short period of time . This method involves using a virus to deliver two different nucleases to a cell . Once inside the cell , the viruses released the nucleases , which were able to remove up to one-quarter of their gene targets , with relatively few errors , in the time that they were active . Next , Cai et al . added gene patches—new genes to replace those removed by the nucleases—to the viruses . This ‘cut and patch’ strategy was successful in up to 8% of the treated cells . The results also suggest that this approach is safer than other gene-editing techniques . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression"
] | 2014 | Targeted genome editing by lentiviral protein transduction of zinc-finger and TAL-effector nucleases |
Membrane proteins with multiple transmembrane domains play critical roles in cell physiology , but little is known about the machinery coordinating their biogenesis at the endoplasmic reticulum . Here we describe a ~ 360 kDa ribosome-associated complex comprising the core Sec61 channel and five accessory factors: TMCO1 , CCDC47 and the Nicalin-TMEM147-NOMO complex . Cryo-electron microscopy reveals a large assembly at the ribosome exit tunnel organized around a central membrane cavity . Similar to protein-conducting channels that facilitate movement of transmembrane segments , cytosolic and luminal funnels in TMCO1 and TMEM147 , respectively , suggest routes into the central membrane cavity . High-throughput mRNA sequencing shows selective translocon engagement with hundreds of different multi-pass membrane proteins . Consistent with a role in multi-pass membrane protein biogenesis , cells lacking different accessory components show reduced levels of one such client , the glutamate transporter EAAT1 . These results identify a new human translocon and provide a molecular framework for understanding its role in multi-pass membrane protein biogenesis .
The human genome encodes thousands of integral membrane proteins , which play critical roles in nearly all aspects of cell physiology . Membrane proteins of the cell surface and most intracellular compartments are first assembled at the endoplasmic reticulum . Most of these are inserted by the evolutionarily conserved Sec61 complex , which guides hydrophobic transmembrane domains ( TMDs ) into a central aqueous channel that opens laterally to allow TMD entry into the bilayer ( Voorhees and Hegde , 2016; Li et al . , 2016; Pfeffer et al . , 2015 ) . How this process is elaborated to facilitate the insertion and folding of membrane proteins containing multiple TMDs is not well understood . The human genome encodes ~2500 multi-pass proteins , including GPCRs , solute carriers , ion channels , and ABC transporters . These show considerable biophysical and topological complexity , including TMDs of variable length and hydrophobicity , closely spaced TMD hairpins , and re-entrant loops that span only part of the membrane ( Cymer et al . , 2015; Foster et al . , 2000 ) . These features are often critical for function , but they pose a significant challenge for the biosynthetic machinery ( Foster et al . , 2000; Tector and Hartl , 1999 ) . The ‘translocon’ is a poorly defined and dynamic ensemble that coordinates the insertion , folding , modification and assembly of most membrane proteins . The eukaryotic translocon comprises the core Sec61 channel in association with different accessory factors . The best studied of these factors include the OST ( Chavan et al . , 2005 ) and TRAP complexes ( Fons et al . , 2003 ) , TRAM ( Görlich and Rapoport , 1993; Voigt et al . , 1996 ) , Sec62/63 ( Conti et al . , 2015 ) , and the signal peptidase complex ( Kalies et al . , 1998 ) . A significant challenge to studying the translocon is that many accessory factors only transiently or sub-stoichiometrically associate with the core machinery , contributing to difficulties in isolating intact complexes ( Görlich and Rapoport , 1993; Wang and Dobberstein , 1999 ) . As a result , the composition and stoichiometry of ribosome-bound translocons , and their structures , functions and clientele , remain poorly defined . We previously identified TMCO1 as a eukaryotic member of the Oxa1 superfamily , whose members are linked to membrane protein biogenesis ( Anghel et al . , 2017 ) . These proteins , including the EMC3 subunit of the ‘ER membrane complex’ ( EMC ) ( Chitwood et al . , 2018; Guna et al . , 2018; Shurtleff et al . , 2018; Volkmar et al . , 2019 ) , the Get1 subunit of the Get1/2 complex ( Schuldiner et al . , 2008; Mariappan et al . , 2011; Wang et al . , 2014 ) , and the Oxa1/Alb3/YidC proteins ( Shanmugam and Dalbey , 2019 ) , function in different contexts as TMD insertases and/or as intramembrane chaperones to facilitate membrane protein folding and assembly ( Shurtleff et al . , 2018; Nagamori et al . , 2004; Serdiuk et al . , 2016; Klenner et al . , 2008 ) . The function of TMCO1 is not yet known , but consistent with a role in a co-translational process at the ER membrane , it can be natively isolated in association with ribosome-Sec61 complexes ( Anghel et al . , 2017 ) .
To identify components of TMCO1-ribosome complexes we solubilized microsomes isolated from 3xFlag-TMCO1 HEK293 cells , affinity purified via the Flag tag on TMCO1 , isolated the ribosome-bound fraction by sedimentation , and identified co-purifying proteins by quantitative mass spectrometry ( Figure 1A , B ) . Ribosomal proteins , subunits of the Sec61 complex , and TMCO1 were enriched relative to control cells lacking the Flag tag , whereas known translocon accessory factors—including subunits of the OST and TRAP complexes , TRAM , Sec62/63 and the signal peptidase complex—were either weakly enriched or absent . We also observed strong enrichment of three poorly studied proteins: the single-pass membrane protein CCDC47 ( calumin ) and two subunits of the Nicalin-TMEM147-NOMO transmembrane complex ( Dettmer et al . , 2010 ) . We confirmed recovery of Sec61 , TMCO1 , Nicalin , TMEM147 , NOMO and CCDC47 in the ribosome-associated fraction following 3xFlag-TMCO1 immunopreciptation ( Figure 1C , D ) . Notably , the catalytic OST subunit STT3A was not detected , consistent with the absence of OST from the TMCO1-ribosome complexes . Affinity purification via a Flag tag on Nicalin recovered TMCO1 , CCDC47 and NOMO in the ribosome-bound fraction ( Figure 1—figure supplement 1A ) , indicating that these proteins can be isolated as a single , ribosome-associated complex . In the absence of ribosomes , however , only components of the Nicalin-TMEM147-NOMO complex remained intact ( Figure 1—figure supplement 1B ) , suggesting that TMCO1 , CCDC47 and the pre-formed Nicalin-TMEM147-NOMO complex assemble in the context of the ribosome . CCDC47 , Nicalin , TMEM147 and NOMO are abundant ER-localized proteins , conserved across eukaryotes , widely expressed in human tissues , and associated with several human diseases ( Itzhak et al . , 2016; Burdon et al . , 2011; Sharma et al . , 2012; Xin et al . , 2010; Caglayan et al . , 2013; Alanay et al . , 2014; Li et al . , 2019; Reuter et al . , 2017; Morimoto et al . , 2018 ) . Although their functions remain obscure , CCDC47 has been linked to various membrane-associated processes ( Morimoto et al . , 2018; Zhang et al . , 2007; Konno et al . , 2012; Thapa et al . , 2018; Yamamoto et al . , 2014 ) , and the Nicalin-TMEM147-NOMO complex has been proposed to regulate subunit assembly and localization of several cell surface receptors and ion channels ( Almedom et al . , 2009; Gottschalk et al . , 2005; Kamat et al . , 2014; Rosemond et al . , 2011 ) . More recently , all four genes were identified in a genome-wide screen for factors that impair surface expression of a mutant TRP channel ( Talbot et al . , 2019 ) . That these proteins can be stably isolated with TMCO1-bound ribosome-Sec61 complexes suggests a link between these observations and a co-translational process at the ER . We sought to clarify the role of these ribosome-associated membrane components by examining their arrangement relative to key functional domains of the ribosome and the Sec61 complex . Using chemical cross-linking and mass spectrometry ( XL-MS ) , we identified 1229 unique , high-confidence intra- and inter-protein cross-links in the affinity-purified complexes ( Figure 2—figure supplement 1A , B ) . Multiple cross-links between 60S ribosomal subunits and the cytosolic-facing regions of Sec61 , TMCO1 , CCDC47 , TMEM147 and NOMO confirmed their predicted membrane topologies , and placed them in the vicinity of the ribosome exit tunnel ( Figure 1D and Figure 2—figure supplement 1C , D ) . We next used single particle cryo-electron microscopy ( cryo-EM ) to directly visualize the natively purified complexes ( Figure 2 , Figure 2—figure supplements 2–5 and Table 1 ) . Density for the ribosome is well-defined , revealing hybrid state A/P and P/E tRNAs , and a mixture of nascent polypeptides in the exit tunnel ( Figure 2—figure supplement 5A ) . Additional density is visible surrounding the ribosome exit tunnel . Local resolution within the translocon varies from ~3 . 5–4 . 5 Å in Sec61 and regions contacting the ribosome , to ~5 . 5–7 . 5 Å for most of the membrane region , and ~10–15 Å in peripheral and luminal regions ( Figure 2—figure supplement 4 and Figure 2—figure supplement 5B , C ) . Sec61 is in a conformation similar to that observed in the ribosome-Sec61-OST complex ( Braunger et al . , 2018 ) , with a closed lateral gate and the plug helix occluding the central pore ( Figure 2—figure supplement 5D ) . A cluster of eight TMDs visible near the Sec61 hinge were unambiguously assigned to the Nicalin-TMEM147 sub-complex using a homology model based on the APH1-Nicastrin sub-complex of human γ-secretase ( Figure 2—figure supplement 6; Dettmer et al . , 2010; Bai et al . , 2015; Haffner et al . , 2004 ) . The distinctive arrangement of the seven TMEM147 TMDs could be docked into the density as a rigid body with only minor adjustments ( Figure 2C and Figure 2—figure supplement 5E ) . This enabled assignment of the remaining helical density to the single Nicalin TMD , which packs against TM1 of TMEM147 in its evolutionarily predicted position . Here , the large luminal domain of Nicalin extends into low-resolution density directly below the translocon ( Figure 2D ) . Notably , the cytosolic end of TM3 in TMEM147 is ten residues shorter than the corresponding TM3 in APH-1 ( Figure 2—figure supplement 6E ) . This allows TMEM147 to bind Sec61 despite limited space in the ribosome-translocon junction , and positions the short cytosolic regions of TMEM147 in contact with uL24 and rRNA H7 , in agreement with the XL-MS ( Figure 2C and Figure 2—figure supplement 1C , D ) . Adjacent to the Nicalin-TMEM147 sub-complex is a cluster of three TMDs , which were assigned to TMCO1 ( Figure 2E ) . We generated a model of TMCO1 using RaptorX-Contact ( Wang et al . , 2017; Xu , 2019 ) , which employs co-evolutionary data and deep learning for distance-based structure prediction ( Figure 2—figure supplement 7 ) . The model recapitulates the conserved three TMD core , N-out/C-in topology , and cytosolic-facing coiled-coil found in members of the Oxa1 superfamily ( Anghel et al . , 2017; Borowska et al . , 2015; Kumazaki et al . , 2014 ) , and could be placed into density with only minor adjustments . The conserved and positively charged coiled-coil of TMCO1 extends out of the membrane into the cytosolic vestibule where it packs against a surface on the ribosome that includes rRNA H19 , H24 and uL24 . This agrees with the proposed ribosome-binding mode of bacterial YidC ( Kedrov et al . , 2016 ) , satisfies numerous inter- and intra-molecular cross-links observed by XL-MS ( Figure 2E and Figure 2—figure supplement 1C , D ) , and rationalizes the previously reported ribosome-binding activity of TMCO1 ( Anghel et al . , 2017 ) . The most prominent feature in the cytosolic vestibule is a globular density that curls out of the membrane near TMCO1 and terminates in a long helical extension that traces along the ribosome surface ( Figure 2F ) . Using RaptorX-Contact we generated a model of the large cytosolic region of CCDC47 , revealing a long and flexible C-terminal coiled-coil extending from a globular N-terminal domain ( Figure 2—figure supplement 8 ) . After placing the globular domain as a rigid body , the coiled-coil was adjusted to fit the extended helical density . The globular domain of CCDC47 contacts eL6 and rRNA H25 , while the conserved and positively charged coiled-coil wedges between Sec61 and rRNA H24 , before terminating at the mouth of the exit tunnel . This satisfies multiple intra- and inter-molecular cross-links to Sec61α , to the flexible N-terminus of Sec61β , to uL22 , eL31 and eL32 , and to the TMCO1 coiled-coil and C-terminal helix ( Figure 2F and Figure 2—figure supplement 1C , D ) . The translocon extends ~90 x 120 x 140 Å , with Sec61 and its accessory factors arranged near the ribosome exit tunnel and significant mass on both sides of the membrane ( Figure 3A ) . Ribosome binding is mediated by multiple protein-protein and protein-RNA contacts ( Figure 3A , C ) . These position the cytosolic domains of TMCO1 and CCDC47 near the nascent chain as it emerges from the ribosome exit tunnel , and place the luminal Nicalin domain near translocated segments of the nascent chain . Notably , the long C-terminal coiled-coil of CCDC47 extends to the nascent chain at the ribosome exit tunnel ( Figure 3A , B ) . Truncating this conserved motif causes a developmental disorder in humans ( Morimoto et al . , 2018 ) , suggesting that this interaction is functionally important . A prominent feature of the complex is a large ( ~25 x 30 x 30 Å ) , lipid-filled cavity formed at the center of the translocon by Sec61 , TMEM147 and TMCO1 ( Figure 3B ) . Like other structurally characterized members of the Oxa1 superfamily ( Anghel et al . , 2017; Kumazaki et al . , 2014; Borowska et al . , 2015 ) , the transmembrane helices of TMCO1 form a funnel that extends from the cytosol into the lipid bilayer ( Figure 4A ) . In bacterial YidC , this funnel operates as a transient binding site for TMDs , which are then released into the membrane ( Kumazaki et al . , 2014; Borowska et al . , 2015 ) . TMCO1 is located on the ‘back’ side of Sec61 in the TMCO1 translocon ( Figure 4B ) . Here , the TMCO1 funnel lines the lipid-filled cavity at the center of the translocon , suggesting that a hydrophobic segment could be inserted from the cytosol into a protected membrane environment . TMEM147 also lines the lipid-filled cavity . Here , its seven TMDs form a funnel that extends from the lumen partway across the membrane . Within the bilayer , the Sec61 hinge ( located between TM5 and TM6 ) contacts TM2 , TM3 and TM4 inside the TMEM147 funnel . A similar intra-membrane interaction is observed in γ-secretase , where the presenilin C-terminus fills the hydrophobic APH-1 funnel ( Figure 4—figure supplement 1B; Bai et al . , 2015 ) . Unlike in γ-secretase , however , the Sec61 hinge only partially occupies the TMEM147 funnel , laterally sealing it in membrane , but leaving it open to the lumen ( Figure 4C , D ) . This is reminiscent of the Hrd1 protein conducting ERAD channel , in which a structurally similar hydrophilic funnel , proposed to transport transmembrane segments from the bilayer to the cytosol , opens to the cytosol , and is laterally sealed by a neighboring Hrd1 subunit ( Figure 4—figure supplement 1C; Schoebel et al . , 2017 ) . By analogy , TMEM147 could insert a hydrophobic segment from the lumen into the central membrane cavity in a process gated by Sec61 . Taken together , these structural observations suggest that the TMCO1 translocon may be specialized for membrane protein biogenesis . We sought to test this possibility by sequencing the mRNAs associated with ribosomes recovered after affinity purification via the Flag tag on TMCO1 ( RIP-seq ) . Remarkably , we observed strong enrichment for transcripts encoding secretory pathway transmembrane proteins ( Figure 5A ) . Of these , single-pass proteins—by far the most abundant type of membrane protein in the human genome—were strongly depleted ( Figure 5B ) . By contrast , transcripts encoding multi-pass membrane proteins with four or more TMDs were enriched ( Figure 5B ) . These include numerous transporters , receptors , transferases and hydrolases ( Figure 5C ) . Consistent with selective enrichment of TMCO1-linked transcripts , we observed enrichment across three orders of magnitude of transcript abundance in the input sample ( Figure 5D ) , and this was independent of protein length ( Figure 5—figure supplement 1A ) . These data directly link the TMCO1 translocon to a co-translational process involving hundreds of different multi-pass clients . To evaluate the role of the TMCO1 translocon in biogenesis , we monitored the endogenous protein levels of the ‘Excitatory amino acid transporter 1’ ( EAAT1; SLC1A3; GLAST-1 ) in HEK293 cells lacking different accessory components . EAAT1 is a member of the large solute carrier ( SLC ) transporter superfamily , more than one-third of which were enriched by RIP-seq . EAAT1 functions as a homotrimer , and its structure contains multiple TMDs of marginal hydrophobicity and re-entrant helical loops on both sides of the membrane , all of which are required for function ( Canul-Tec et al . , 2017 ) . Compared to wild-type cells , the steady-state expression level of EAAT1 was reduced by ~3 fold in TMCO1 knockout cells , but was unaffected in cells lacking the auxiliary translocon component TRAM ( Figure 5E–G ) . Similar reductions were observed in Nicalin ( 2 . 4-fold ) , TMCO1/Nicalin ( 2 . 8-fold ) and TMCO1/CCDC47 ( 3 . 8-fold ) single- and double-knockout cells , while a single CCDC47 ( 1 . 6-fold ) knockout showed only a modest reduction ( Figure 5F , G ) . By contrast , the steady-state expression levels and glycosylation patterns of two single-pass membrane proteins , integrin α5 and TRAPα , were unchanged , demonstrating that TMCO1 disruption does not lead to a general defect in membrane protein biogenesis ( Figure 5—figure supplement 1B ) . We also observed little change in EAAT1 mRNA levels in TMCO1 , Nicalin and CCDC47 single- and double-knockout cells . Together with the structural analysis , these data implicate the TMCO1 translocon in multi-pass membrane protein biogenesis .
Our data identify TMCO1 , CCDC47 and the Nicalin-TMEM147-NOMO complex as conserved components of an ER translocon that functions co-translationally with Sec61 during biogenesis of multi-pass membrane proteins . The biochemical function of the TMCO1 translocon is currently unclear . Although we do not formally exclude a role in client-specific targeting to the ER , we propose that the TMCO1 translocon functions as an insertase and intramembrane chaperone . Our structural model for TMCO1 is consistent with its evolutionary relationship to members of the Oxa1 superfamily , including YidC , Get1 , EMC3 and Ylp1 ( Anghel et al . , 2017 ) . These proteins have evolved to function in different contexts , but their ability to move transmembrane segments into the membrane appears to be conserved ( Wang et al . , 2014; Klenner et al . , 2008; Pleiner et al . , 2020; Yu et al . , 2008 ) . By analogy , we propose that hydrophobic segments of the nascent chain that inefficiently engage with Sec61 could access the membrane through the conserved cytosolic TMCO1 funnel . In addition , hydrophobic segments that have translocated across the bilayer through the canonical Sec61 channel might access the membrane through the luminal TMEM147 funnel . As segments integrate , the central cavity of the translocon could shield the nascent chain to minimize misfolding and degradation . Organizing these putative functions in a single translocon might increase the efficiency with which different biophysical and topological features of the nascent chain are accommodated during multi-pass membrane protein biogenesis ( Cymer et al . , 2015; Lu et al . , 2000; Skach , 2009 ) . High resolution structures of the TMCO1 translocon and analysis of its interactions with substrate will be important for testing this model . More broadly , our data support a general view of the translocon as a dynamic assembly whose subunit composition varies temporally to meet the demands of a particular client ( Conti et al . , 2015; Johnson and van Waes , 1999 ) . TMCO1 , CCDC47 and the Nicalin-TMEM147-NOMO complex are abundant ( Itzhak et al . , 2016 ) , which presumably allows them to compete with other translocon-associated factors for access to the nascent chain . Indeed , while many multi-pass clients of the TMCO1 translocon harbor N-terminal STT3A glycosylation sites ( Cherepanova et al . , 2019 ) , we see little biochemical or structural evidence for an associated OST complex . This is consistent with the substantial steric overlap observed between the membrane and luminal regions of OST and the TMCO1 accessory factors ( Figure 4—figure supplement 2 ) , which likely dictates that they alternately access the nascent chain during synthesis . Notably , TMCO1 , CCDC47 and the Nicalin-TMEM147-NOMO complex do not stably associate in the absence of ribosomes ( Figure 1—figure supplement 1 ) . An intriguing possibility is that these are modular components , capable of acting independently in additional contexts . A general role for the TMCO1 translocon in multi-pass membrane protein biogenesis is consistent with the wide expression and conservation of its subunits , and the numerous cellular and organismal phenotypes associated with their dysfunction . In humans , TMCO1 has been linked to glaucoma ( Burdon et al . , 2011; Sharma et al . , 2012 ) , and loss of either TMCO1 ( Xin et al . , 2010; Caglayan et al . , 2013; Alanay et al . , 2014 ) or CCDC47 ( Morimoto et al . , 2018 ) causes rare autosomal recessive developmental disorders . Similarly , a mutation in TMEM147 has been linked to a rare neurodevelopmental disorder manifesting with severe intellectual disability and impaired vision ( Reuter et al . , 2017 ) . At the cellular level , disrupting TMCO1 , CCDC47 , Nicalin , TMEM147 or NOMO leads to reduced fitness ( Wang et al . , 2015 ) . Cells lacking CCDC47 show attenuated ERAD ( Yamamoto et al . , 2014 ) and impaired Ca2+ signaling ( Zhang et al . , 2007; Konno et al . , 2012 ) , while the Nicalin-TMEM147-NOMO complex is linked to Nodal signaling ( Haffner et al . , 2004 ) and altered localization and subunit composition of some multi-pass membrane proteins ( Almedom et al . , 2009; Kamat et al . , 2014; Rosemond et al . , 2011 ) . Cells lacking TMCO1 show defects in Ca2+ handling , which has led to the proposal that TMCO1 functions as a Ca2+-channel ( Wang et al . , 2016 ) . Our data reconcile these different observations , which likely result from biogenesis defects in hundreds of different multi-pass proteins . As the folding capacity of the cell must be robust to mutations and other stresses that affect folding efficiency , it is likely that other ER chaperones and accessory factors can partially compensate for loss of TMCO1 translocon components . In this regard , it will be important to define the functional relationship between the TMCO1 translocon and the ER membrane complex ( EMC ) , each of which harbors a subunit belonging to the Oxa1 superfamily ( Anghel et al . , 2017 ) and facilitates multi-pass membrane protein biogenesis ( Chitwood et al . , 2018; Guna et al . , 2018; Shurtleff et al . , 2018; Volkmar et al . , 2019; Tian et al . , 2019 ) .
Antibodies against human TMCO1 , Sec61β and TRAPα were characterized previously ( Fons et al . , 2003; Anghel et al . , 2017; Görlich et al . , 1992 ) . Additional antibodies were obtained from the following sources: anti-EAAT1 ( Santa Cruz , sc-515839 ) , anti-Sec61α ( Thermo Fisher , PA5-21773 ) , anti-uL22 ( Abgent , AP9892b ) , anti-STT3A ( Novus , H00003703-M02 ) , anti-Tubulin ( Abcam , ab7291 ) , anti-Integrin α5 ( Cell signaling , 4705 ) anti-Nicalin ( Bethyl , A305-623A-M ) , anti-TMEM147 ( Thermo Fisher , PA5-95876 ) , anti-NOMO ( Thermo Fisher , PA5-47534 ) , anti-CCDC47 ( Bethyl , A305-100A ) , anti-TRAM1 ( Abcam , ab190982 ) , anti-Mouse rabbit HRP ( Abcam , ab6708 ) , anti-Rabbit donkey HRP ( Sigma , SAB3700863 ) , anti-Goat rabbit HRP ( Sigma , A5420 ) . Flp-In T-REx 293 cells containing a 3xFlag-Cas9 construct were maintained in DMEM supplemented with 10% FBS ( Gemini Foundation ) and penicillin/streptomycin mixture ( Invitrogen ) . TMCO1 knockout and 3xFlag-TMCO1 HEK293 cell lines have been described and characterized previously ( Anghel et al . , 2017 ) . Nicalin and CCDC47 knockout cell lines were generated using the CRISPR/Cas9 system , in both parental and TMCO1 knockout backgrounds . Cas9 expression was induced by addition of 10 ng/mL doxycycline followed by transfection of sgRNAs targeting either Nicalin ( ACGGAATGCAGTGCTGAACA ) or CCDC47 ( TCAGTGATTATGACCCGTT ) . Cells were grown for 48 hr , followed by single cell sorting into 96 well plates for clonal isolation . Nicalin and CCDC47 knockouts in parental and TMCO1 knockout backgrounds were verified by western blot and genomic DNA sequencing . A TRAM1 knockout cell line was generated using the CRISPR/Cas9 system in Flp-In T-Rex 293 cells ( Thermo Fisher ) by transfecting a modified pX330 plasmid ( Addgene ) expressing human codon-optimized Cas9 and an sgRNA targeting TRAM1 ( TTTGATGCCATAGTAATAAA ) . Single cells were isolated by sorting and allowed to grow clonally . The final TRAM1 knockout was verified by western blot and genomic DNA sequencing . To scale up the sample preparation for crosslinking mass spectrometry ( XL-MS ) and cryo-EM , Flp-In TRex 293 cells expressing 3xFlag-TMCO1 from the endogenous TMCO1 promoter were cultured in suspension . Cells were grown in 1 L PETG square media bottles ( Fisher , 09-923-16C ) containing 250 ml Freestyle 293 media ( Gibco ) supplemented with 10 mM Hepes pH 7 . 5 ( Invitrogen ) , 10 mM L-glutamine ( Invitrogen ) , 0 . 3 μg/ml penicillin ( Gemini ) , 0 . 5 μg/ml streptomycin ( Gemini ) , and 0 . 5% FBS ( Gemini Foundation ) , at 37°C , 5% CO2 , and 135 rpm , to a final density of ~1×106 cells/ml . Stable cell lines overexpressing N-terminally 3xFlag tagged TMCO1 and Nicalin were generated by transfecting the respective knockout cell lines with a modified pEGFP-n1 plasmid ( Addgene ) encoding N-terminally 3xFlag-tagged TMCO1 or Nicalin ( tag inserted after the signal peptide ) , under the control of a CMV promoter . Cells were transfected using the TransIT-293 transfection reagent ( Mirus ) and selected for 14 days by treatment with 0 . 7 mg/ml G418 ( Invitrogen ) , with selection media changed every 3 days . Selected cells were maintained in DMEM supplemented with 10% FBS , penicillin/streptomycin , and 0 . 3 mg/ml G418 . Expression was verified by western blot . Cells were checked approximately every three months for mycoplasma contamination using the Universal Mycoplasma Detection Kit ( ATCC ) , and were found to be negative . For mass spectrometry , approximately 2 × 108 of wild-type ( control ) and 3xFlag-TMCO1 cells were pelleted , resuspended in ice cold hypotonic lysis buffer ( 10 mM Hepes pH 7 . 4 , 10 mM potassium acetate , 1 mM magnesium chloride ) and incubated on ice for 15 min . Unless otherwise noted , all buffers included emetine at a final concentration of 50 µg/ml . Cells were lysed with 25 strokes of a pre-chilled dounce tissue grinder with a tight-fitting pestle , then 250 mM sucrose and 1 mM PMSF was added to the lysate . Nuclei were pelleted by centrifugation at 700 x g for 3 min . The membrane-containing supernatant was removed and put on ice . The pellet was washed with 1 ml ice cold assay buffer ( 50 mM Hepes pH 7 . 4 , 250 mM sucrose , 250 mM potassium acetate , 10 mM magnesium chloride ) and centrifuged again , and the supernatant combined with the membrane fraction . Membranes were sedimented at 10 , 000 x g for 10 min at 4°C and resuspended in assay buffer to an A260 of ~50 . Monosomes were generated by treating the resuspended membranes with 1 mM calcium acetate and 10 , 000 U of micrococcal nuclease ( NEB , M0247S ) , and incubating at 25°C for 10 min . Nuclease activity was stopped by adding EGTA to a final concentration of 2 mM . Membranes were solubilized in ice cold assay buffer supplemented with 2 . 5% digitonin ( Calbiochem 11024-24-1 ) for 15 min on ice , and insoluble material was removed by centrifugation at 10 , 000 x g for 10 min at 4°C . TMCO1-ribosome complexes were affinity purified by incubating solubilized material with M2 Flag affinity gel ( Sigma , A2220 ) for 1 hr at 4°C with gentle end-over-end mixing . Unbound material was removed by centrifugation , the resin was washed twice with five bed volumes of ice-cold wash buffer ( 50 mM Hepes pH 7 . 4 , 250 mM sucrose , 350 mM potassium acetate , 10 mM magnesium chloride , 0 . 25% digitonin ) , and twice with five bed volumes of assay buffer supplemented with 0 . 25% digitonin . Bound material was eluted in two successive 30 min incubations with two bed volumes of ice-cold assay buffer supplemented with 0 . 25% digitonin and 0 . 5 mg/ml 3xFlag peptide , at 4°C with gentle end-over-end mixing . The ribosome containing fraction was obtained by sedimenting the IP elutions through a 1 mL sucrose cushion ( 1 M sucrose , 150 mM potassium chloride , 50 mM Tris pH 7 . 5 , 5 mM magnesium chloride , 0 . 1% digitonin ) at 250 , 000 x g for 2 hr in a TLA100 . 3 rotor . Ribosome pellets were resuspended in 50 mM Hepes pH 7 . 4 , 100 mM sodium chloride , 1% SDS . Proteins were then methanol-chloroform extracted , FASP trypsin-digested , TMT-labeled and analyzed in a single 180 min LC-MS/MS run at the Proteomics and Mass Spectrometry Facility at Harvard University . Enrichment ratios were calculated as Flag IP/control IP for all peptides identified more than once . Small-scale IPs were performed similarly , using microsomes isolated from stably integrated 3xFlag-TMCO1 or 3xFlag-Nicalin HEK293 cells . Affinity purification of TMCO1-ribosome complexes for XL-MS and cryo-EM was done as described above with the following changes . Typically , XL-MS samples were produced from ~4×109 cells , and cryo-EM samples from ~7×108 cells . Emetine was not used . To remove any contaminating DNA , isolated membranes were treated with 5 U/ml RNase Free DNase ( Promega , M6101 ) for 15 min at room temperature . Following affinity purification , TMCO1-ribosome complexes were isolated via sedimentation through a 300 µl sucrose cushion ( 0 . 5 M sucrose , 150 mM potassium acetate , 50 mM Hepes pH 7 . 4 , 5 mM magnesium chloride , 0 . 25% digitonin ) at 355 , 000 x g for 45 min in a TLA120 . 1 rotor . Pellets were resuspended in 150 mM potassium acetate , 50 mM Hepes pH 7 . 4 , 5 mM magnesium chloride , 0 . 25% digitonin , and concentration determined by A260 . Approximately 85 μg of purified TMCO1-ribosome complexes were resuspended in 150 mM potassium acetate , 50 mM Hepes pH 7 . 4 , 5 mM magnesium chloride , 0 . 25% digitonin to a concentration of 0 . 5 mg/ml . Crosslinking was performed by adding disuccinimidyl suberate ( DSS , Thermo Fisher , 21555 ) ( prepared as a fresh 10 mM stock in DMSO ) to a final concentration of 0 . 5 mM and incubating for 30 min at 35°C . Crosslinking reactions were mixed by lightly agitating the tube every 5 min during incubation , and quenched by adding 100 mM Tris pH 8 . Reactions were TCA precipitated before processing for mass spectrometry . Pellets were washed with ice cold acetone to remove excess lipid and detergent , and then pelleted again . For mass spectrometry , the TCA precipitated material was resuspended in 8 M Urea and 10 mM TCEP , heated at 56°C for 20 min , alkylated with 15 mM iodoacetamide ( 30 min at room temperature ) , and then quenched with 15 mM dithiothreiotol ( 15 min at room temperature ) . The sample was then diluted to 2 M Urea and digested overnight with 1 µg trypsin ( Promega Gold ) for 4 hr at 37°C . A second aliquot of 1 µg trypsin was then added and digestion was allowed to proceed overnight . The digestion mixture was acidified to 0 . 5% TFA and diluted 6-fold prior to desalting on a Peptide C18 MacroTrap column ( Michrom Bioresources ) controlled by Akta Purifier ( GE Healthcare Life Sciences ) and evaporated to dryness . Crosslinked products were brought up in 10 µl of SEC buffer ( 70:30 H2O:ACN with 0 . 1% TFA ) and enriched by size-exclusion chromatography ( Superdex Peptide , GE Healthcare Life Sciences ) as in Leitner et al . , 2012 . 100 µl fractions eluting between 0 . 9 and 1 . 4 ml were dried , resuspended in 0 . 1% formic acid for MS analysis . The fractions starting at 0 . 9 ml and 1 . 3 ml were combined prior to evaporation to make four MS fractions . Samples were reconstituted in 5 µl of 0 . 1% formic acid for mass spectrometry . LC-MS analysis was performed with an Orbitrap Fusion Lumos mass spectrometer ( Thermo Scientific ) coupled with a nanoelectrospray ion source ( Easy-Spray , Thermo ) and M-Class NanoAcquity UPLC system ( Waters ) . Crosslink enriched fractions were separated on a 50 cm x 75 μm ID PepMap C18 column ( Thermo ) . 2 . 5 µl of sample was loaded onto the column and eluted running a gradient from 3 . 5% solvent B ( A: 0 . 1% formic acid in water , B: 0 . 1% in ACN ) to 25% B in 175 min followed by a second gradient to 30% B over 10 min . Precursor scans were acquired in the Orbitrap from 375 to 1500 m/z ( resolution: 120000 , AGC Target: 4 . 0e5 , max injection time: 50 ms ) . Precursor ions were selected for dissociation using the following criteria: peptide monoisotopic precursor determination , charge state between 3–9 , intensity greater than 5e4 , and a 30 s dynamic exclusion window . Each precursor that passed the selection criteria was subjected to subsequent HCD and ETD MS2 scans ( resolution: 30000 , quadrupole isolation window: 1 . 6 m/z units , HCD NCE: 28% , HCD AGC Target: 1 . 0e5 , HCD max injection time: 150 ms , ETD collision time: calibrated charge dependent ETD parameters , ETD supplemental activation: 10% EThcD , ETD AGC Target: 2 . 0e5 , ETD max injection time: 200 ms ) . Nine product ion scans of each type were performed for each precursor scan . Separate peaklists were generated for ETD and HCD scans using Proteome Discoverer 2 . 2 ( Thermo ) and searched using Protein Prospector 5 . 23 . 0 ( 66 ) . The search database consisted of the sequences of 82 human ribosomal protein components in addition to 10 sequences corresponding to the membrane associated components: TMCO1 , Nicalin , NOMO1 , NOMO2 , NOMO3 , Sec61A1 , Sec61A2 , Sec61β , Sec61γ , TMEM147 and CCDC47 . The sequence of TMCO1 contained the N-terminal 3xFlag tag ( reported crosslinked residue numbers reference the endogenous sequence ) . The three NOMO isoforms are highly homologous and in most cases cross-links to NOMO could not be assigned a specific isoform . These proteins were confirmed to be the dominant components of the sample by MS analysis of late eluting SEC fractions ( corresponding to linear peptides ) . The 92 target proteins were concatenated with a decoy database consisting of 10 randomized amino acid sequences of for each target sequence ( 1012 total protein sequences searched ) . ETD peaklists were searched using Prospector instrument type ESI-ETD-high-res and HCD peaklists were searched using ESI-Q-high-res . Other search parameters were: mass tolerance of 7 ppm ( precursor ) and 15 ppm ( product ) ; fixed modifications of carbamidomethylation on cysteine; variable modifications of peptide N-terminal glutamine conversion to pyroglutamate , oxidation of methionine , and ‘dead-end’ modification of lysine and the protein N-terminus by semi-hydrolyzed BS3 , protein N-terminal acetylation , protein N-terminal methionine loss , and incorrect monoisotopic precursor selection ( neutral loss of 1 Da ) ; crosslinking reagent was DSS/BS3; trypsin specificity was used with three missed cleavages and three variable modifications per peptide were allowed . The top 85 product ion signals were used for the search . Searches were performed using 64 cores on an HPC cluster and took about 4 hr to complete . Cross-link spectral matches ( CSM ) were initially kept with peptide scores above 20 , score difference above 0 , and length of each peptide between 4–25 residues . A linear support vector machine ( SVM ) model was constructed to classify CSMs between decoy and target classes ( Trnka et al . , 2014 ) . Features selected for the SVM classifier were: score difference , percent of ions matched , precursor charge state , rank of each peptide , and length of each peptide . Models were trained on half of the dataset and parameters were chosen to give a specificity of 90% tested on the other half of the data . Separate classifiers were built for ETD and HCD results . The best scoring CSM per unique cross-linked residue pair was selected and the ETD and HCD results were merged . The distribution of cross-linked residue pairs with one and two incorrectly identified peptides was modeled using essentially the same logic as ( Fischer and Rappsilber , 2017 ) , but extending their analysis to account for the 10x increased size of decoy database . The number of target-target hits with one wrong peptide is given by:tf ( TT ) = ( 1/k ) ∗TD− ( 2/k2 ) ∗DDand the number with both wrong is given by:ff ( TT ) = ( 1/k2 ) ∗DDwhere TT , TD , and DD are the number of target-target , target-decoy , and decoy-decoy hits , and k is the scaling factor describing the ratio in size of the decoy database to size of the target database . In this case k = 10 . The final list of cross-links was reported at an SVM score of 1 . 5 which corresponded to a 0 . 55% FDR . Distance analysis was performed by measuring the Cα-Cα distances between all ribosome cross-links against an EM reconstruction of the human 80S ribosome ( PDB ID 4ug0 ) . At the reporting threshold of 1 . 5 , the violation rate ( fraction of mappable cross-links > 35 Å ) was 8 . 7% . Quantifoil 1 . 2/1 . 3 200 mesh grids coated in a 2 nm carbon film ( Ted Pella , Inc ) were glow discharged for 30 s immediately before use . Using an FEI Vitrobot , 2 . 5 μl of ~100 nM sample was applied to each grid , which was then incubated for 15 s at 22°C and 100% humidity , blotted for 11 s , and flash frozen in liquid ethane . Data were collected on an FEI Titan Krios at 300 KV using Latitude S ( Gatan ) software , targeting defocus values from −2 . 5 to −1 . 0 μm . Exposure movies were recorded using a Gatan K3 energy filter and direct electron detector in super resolution mode at 64 , 000x magnification ( super resolution pixel size of 0 . 68 Å ) and a total exposure of 50e-/Å2 fractionated over 40 frames . 5562 super resolution movies were summed and motion corrected using Motioncor2 ( Zheng et al . , 2017 ) with 2x binning , generating corrected micrographs with a pixel size of 1 . 36 Å . Contrast transfer function ( CTF ) parameters were estimated using GCTF ( Zhang , 2016 ) . 1 , 049 , 128 Particles were picked using the semi-autonomous particle picking algorithm in Relion3 . 1 ( 70 ) . All 2D classification , 3D classification , and 3D refinement steps were performed in Relion3 . 1 . Reference-free 2D classification was used to discard non-ribosome containing particles . An initial round of 3D classification using a reference 80S ribosome ( EMD-5592 ) low pass filtered to 60 Å as an initial model was used to isolate particles with clear ribosomal features . Particles in the best classes from this initial round of classification were further examined via a second round of 3D classification against the same initial model . Five classes from this second round of 3D classification showed clear density for 40S and 60S ribosomal subunits , tRNAs , and luminal density below the micelle . Particles from these classes were used for an initial 3D refinement ( 286 , 091 particles ) . CTF refinement was used to estimate beamtilt across the dataset and refine per-particle defocus values . The ribosome density in this initial map was further refined by focused refinement using local angular searches and a mask around the ribosome density ( Figure 2—figure supplement 3 , yellow mask ) . Non-translocon density was then removed from the particle set using signal subtraction and a mask surrounding the translocon ( Figure 2—figure supplement 3 , cyan mask ) . These signal subtracted particles were subjected to 3D classification without alignment . One of the classes from this classification showed strong density for the TMCO1 translocon ( 82 , 684 particles , 28 . 8% ) . Particles from this class were refined using a translocon mask and local angular searches , producing a 3 . 8 Å reconstruction ( Map 1 , EMD-21426 ) . We also reverted these particles to their original , ribosome-containing state for further analysis without signal subtraction . This produced a 3 . 4 Å reconstruction of the entire ribosome-translocon complex ( Map 2 , EMD-21427 ) . Focused refinement with a translocon mask ( Figure 2—figure supplement 3 , magenta mask ) and local angular searches produced a 3 . 8 Å reconstruction with improved translocon density ( Map 3 , EMD-21435 ) . Where noted , maps were sharpened by applying a B-factor determined by the automated methods implemented in Relion3 . 1 ( 70 ) . Additionally , local resolution estimation and filtering was performed using automated methods implemented in Relion3 . 1 . We used the 60S ribosomal subunit , A/P and P/E tRNAs and the nascent chain from the human 80S ribosome-nascent chain complex structure ( PDB ID 6OM0 ) , and the Sec61 complex from the mammalian 80S ribosome-Sec61-OST structure ( PDB ID 6FTI ) as starting points for model building . Homology models for TMEM147 and Nicalin were generated in iTasser ( Yang et al . , 2015 ) , using the γ-secretase subunits , APH-1 and Nicastrin , as templates ( PDB ID 5A63 ) . TMCO1 and CCDC47 models were generated with RaptorX-Contact ( Wang et al . , 2017; Xu , 2019 ) . All three maps were used for model building . The 60S ribosomal subunit ( with tRNAs and poly-Ala nascent chain ) was initially fitted as a rigid body into the 3 . 4 Å globally refined map ( Map 2 ) ( sharpened and low-pass filtered by local resolution ) using UCSF Chimera ( Pettersen et al . , 2004 ) , and then manually adjusted with rigid-body and real-space refinement using COOT 0 . 9-pre ( Emsley et al . , 2010 ) . The Sec61 and TMEM147-Nicalin complexes , TMCO1 and CCDC47 were placed into focused ( Map 3 ) and signal subtracted ( Map 1 ) maps ( unsharpened , and low-pass filtered by local resolution ) as rigid bodies , and then adjusted using tightly restrained real-space refinement in COOT . No density was assigned to NOMO or the CCDC47 lumenal and transmembrane domains . Real-space refinement of the model ( 60S and translocon ) was done with PHENIX ( Afonine et al . , 2018 ) , against the focused map ( Map 3 ) ( unsharpened , and low-pass filtered by local resolution ) . Three rounds of global minimization and group B-factor refinement were performed with tight secondary structure , reference model , rotamer , and Ramachandran restraints applied . Secondary structure- and reference model restraints were determined from the starting models . Hydrogen-bonding and base-pair and stacking parallelity restraints were applied to the rRNA . Final model statistics are provided in Table 1 . Structure figures were generated with UCSF Chimera and PyMOL ( http://www . pymol . org ) . Affinity purified TMCO1-ribosome complexes were isolated as described above ( ‘Isolation of TMCO1-ribosome complexes for interaction analysis’ ) , with the following changes . ~ 108 cells were processed for each of three biological replicates . All buffers were made using DEPC-treated RNase free water . Solubilized membranes were incubated with M2 Flag affinity gel ( Sigma , A2220 ) for 2 hr at 4°C with end-over-end mixing . To remove contaminating DNA , 1 U RNase-Free DNase ( Promega , M6101 ) was added to the sample during resin binding . Unbound material was removed and the resin was washed four times with five column volumes of wash buffer to remove contaminating ribosomes . TMCO1-ribosome complexes were isolated by centrifugation as before . After centrifugation , the final pellet was resuspended in 250 mM sucrose , 300 mM potassium acetate , 50 mM Hepes pH 7 . 4 , 5 mM magnesium acetate , 50 µg/mL emetine and 0 . 1% digitonin , flash frozen and stored at −80°C until ready for sequencing . All mRNA sequencing was performed at the University of Chicago Genomics Facility . For each of three biological replicates , RNA was extracted , ribosomal RNA was removed by RiboZero and cDNA libraries were prepared . Fragment sizes were determined by Bioanalyzer , and samples were pooled for sequencing on an Illumina HiSeq 4000 . 50 bp single-end sequence reads were aligned to the human GRCh38 reference transcriptome using STAR ( v2 . 6 . 1 ) ( Dobin et al . , 2013 ) and gene transcript abundance was quantified by featureCounts using the Subread package ( v1 . 6 . 3 ) ( Liao et al . , 2013 ) . Possible batch effects were adjusted using the SVA package ( Leek et al . , 2012 ) in R . An IP enrichment score was calculated as follows: IP enrichment = ( Flag IP abundance - Control IP abundance ) /Total membrane abundance , where ‘Total membrane abundance’ was determined by mRNA sequencing the total membrane fraction in HEK293 TRex cells , as described below . Only genes with mean CPM higher than 0 . 5 were considered confidently identified and used in the analysis . For mRNA sequencing of total membrane-associated mRNAs , membrane suspensions from three biological replicates of parental HEK293 TRex cells were prepared as above , with the inclusion of 1 U/mL SuperaseIn and 50 µg/mL emetine at all times . Membranes were washed twice with 250 mM sucrose , 150 mM potassium acetate , 50 mM Hepes pH 7 . 4 , 5 mM magnesium acetate , 50 µg/mL emetine , and then RNA was Trizol extracted . For each biological replicate , ribosomal RNA was removed by Oligo-dT affinity purification and cDNA libraries were prepared , sequenced , and analyzed as described above . For each replicate of the expression analysis , 750 , 000 cells were plated on poly-L-lysine coated plates and grown overnight . Cells were harvested by centrifugation , lysed using RIPA buffer ( 1% triton , 0 . 5% deoxycholate , 0 . 1% SDS and 1X protease inhibitor cocktail ) , and EAAT1 protein levels analyzed by SDS-PAGE and western blotting . Immunoblots were quantified using ImageJ ( Schneider et al . , 2012 ) . For EAAT1 , Integrin α5 and TRAPα glycosylation analysis , RIPA cell lysates were reduced with 2% β−mercaptoethanol and denatured by heating for 10 min at 65°C . Samples were then incubated with EndoH ( NEB ) or PNGaseF ( Promega ) for 7 hr at 37°C , and analyzed by SDS-PAGE and western blotting . For mRNA quantitation , total RNA was Trizol extracted ( Ambion ) . cDNA ( 1000 ng ) was synthesized using gDNA Clear cDNA Synthesis Kit ( Bio-rad ) . qPCR was performed using iTaq Universal SYBR Green Supermix ( Bio-rad ) via CFX96 Touch Real-Time PCR Detection System ( Bio-rad ) . Primers used for mRNA quantification were: EAAT1 fwd 5’-TTCCTGGGGAACTTCTGATG-3’ , EAAT1 rev 5’-CCATCTTCCCTGATGCCTTA-3’ , GAPDH fwd 5’-ACAACTTTGGTATCGTGGAAGG-3’ , and GAPDH rev 5’-GCCATCACGCCACAGTTTC-3’ . | Cell membranes are structures that separate the interior of the cell from its environment and determine the cell’s shape and the structure of its internal compartments . Nearly 25% of human genes encode transmembrane proteins that span the entire membrane from one side to the other , helping the membrane perform its roles . Transmembrane proteins are synthesized by ribosomes – protein-making machines – that are on the surface of a cell compartment called the endoplasmic reticulum . As the new protein is made by the ribosome , it enters the endoplasmic reticulum membrane where it folds into the correct shape . This process is best understood for proteins that span the membrane once . Despite decades of work , however , much less is known about how multi-pass proteins that span the membrane multiple times are made . A study from 2017 showed that a protein called TMCO1 is related to a group of proteins involved in making membrane proteins . TMCO1 has been linked to glaucoma , and mutations in it cause cerebrofaciothoracic dysplasia , a human disease characterized by severe intellectual disability , distinctive facial features , and bone abnormalities . McGilvray , Anghel et al . – including several of the researchers involved in the 2017 study – wanted to determine what TMCO1 does in the cell and begin to understand its role in human disease . McGilvray , Anghel et al . discovered that TMCO1 , together with other proteins , is part of a new ‘translocon’ – a group of proteins that transports proteins into the endoplasmic reticulum membrane . Using a combination of biochemical , genetic and structural techniques , McGilvray , Anghel et al . showed that the translocon interacts with ribosomes that are synthesizing multi-pass proteins . The experiments revealed that the translocon is required for the production of a multi-pass protein called EAAT1 , and it provides multiple ways for proteins to be inserted into and folded within the membrane . The findings of McGilvray , Anghel et al . reveal a previously unknown cellular machinery which may be involved in the production of hundreds of human multi-pass proteins . This work provides a framework for understanding how these proteins are correctly made in the membrane . Additionally , it suggests that human diseases caused by mutations in TMCO1 result from a defect in the production of multi-pass membrane proteins . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2020 | An ER translocon for multi-pass membrane protein biogenesis |
Perinatal brain injuries , including hippocampal lesions , cause lasting changes in dopamine function in rodents , but it is not known if this occurs in humans . We compared adults who were born very preterm with perinatal brain injury to those born very preterm without perinatal brain injury , and age-matched controls born at full term using [18F]-DOPA PET and structural MRI . Dopamine synthesis capacity was reduced in the perinatal brain injury group relative to those without brain injury ( Cohen’s d = 1 . 36 , p=0 . 02 ) and the control group ( Cohen’s d = 1 . 07 , p=0 . 01 ) . Hippocampal volume was reduced in the perinatal brain injury group relative to controls ( Cohen’s d = 1 . 17 , p=0 . 01 ) and was positively correlated with striatal dopamine synthesis capacity ( r = 0 . 344 , p=0 . 03 ) . This is the first evidence in humans linking neonatal hippocampal injury to adult dopamine dysfunction , and provides a potential mechanism linking early life risk factors to adult mental illness .
More than 10% of babies born in the USA are born preterm ( born before 37 weeks of gestation ) , and about 2% are born very preterm ( VPT , before 32 weeks of gestation ) ( Hamilton et al . , 2015 ) . Premature birth is a risk factor for cognitive impairment ( Anderson , 2014 ) and a number of psychiatric disorders , including schizophrenia and affective disorders ( Nosarti et al . , 2012 ) . The second and third trimesters of gestation are critical periods for neurodevelopment , particularly for axon and synapse formation , glial proliferation and the development of neurotransmitter systems including the dopaminergic system ( de Graaf-Peters and Hadders-Algra , 2006 ) . Thus , VPT birth occurs during a critical time for the development of a number of neural systems , when the brain is particularly susceptible to exogenous and endogenous insults ( Volpe , 2009 ) . VPT babies are at risk of sustaining a variety of perinatal brain injuries , including periventricular haemorrhage , ventricular dilatation and periventricular leukomalacia that are often associated with hypoxic-ischaemic events ( Huang and Castillo , 2008 ) . The sequelae of VPT birth include long-lasting and widespread structural brain alterations , with hippocampal and prefrontal cortical development consistently affected ( Nosarti and Froudist-Walsh , 2016 ) . There is substantial evidence from animal models that perinatal brain injury due to hippocampal lesions ( Lipska et al . , 1993 ) or obstetric complications ( Boksa and El-Khodor , 2003 ) can lead to long-term alterations in the dopamine system , which remain evident in adulthood . Several animal models of schizophrenia have linked hippocampal lesions at different life stages to altered dopaminergic function . Neonatal ventral hippocampal lesions lead to behavioural alterations normally associated with increased dopaminergic activity ( Lipska et al . , 1993 ) despite a reduction , or no change in presynaptic dopamine activity ( Lillrank et al . , 1999; Wan et al . , 1996 ) . In contrast , both adult hippocampal lesions and pre-natal injection of the mitotoxin methylazoxymethanol acetate ( MAM ) into the ventral hippocampus lead to similar behavioural effects and increased presynaptic dopaminergic activity ( Lodge and Grace , 2007; Wilkinson et al . , 1993 ) . This may mirror the increased dopamine synthesis and release seen in human schizophrenia ( Howes et al . , 2012 ) , a condition that has long been associated with obstetric complications ( Cannon et al . , 2002 ) . In rodents , neonatal hippocampal lesions lead to disrupted development of the prefrontal cortex ( Flores et al . , 2005; Tseng et al . , 2008 ) . We have previously demonstrated structural and functional cortico-striatal connectivity alterations following very preterm birth ( Karolis et al . , 2016; White et al . , 2014 ) , which could have significant effects on dopamine transmission ( Cachope and Cheer , 2014; Zhang and Sulzer , 2003 ) . However , it is not known if perinatal brain injury is associated with dopaminergic alterations in adulthood in humans , or how this relates to hippocampal and prefrontal structural alterations . We aimed to disentangle the preclinical , post-mortem and indirect clinical evidence regarding the effects of early brain insults on later dopamine function by directly comparing two contrasting hypotheses , namely that early brain injury leads to hyper- , or alternatively hypo-dopaminergia in the striatum . Moreover , in view of the preclinical findings showing that perinatal hippocampal lesions can lead to lasting alterations to the dopamine system ( Lipska and Weinberger , 2000 ) , and the vulnerability of the hippocampus to perinatal brain injury ( Liu et al . , 2004 ) , we hypothesised that hippocampal volume and striatal dopaminergic function would be related . In an exploratory analysis we further investigated whether dorsolateral prefrontal cortex ( dlPFC ) volume was associated with striatal dopamine synthesis , or whether it mediated the relationship between hippocampal volume and striatal dopamine .
Seventeen individuals from the VPT-perinatal brain injury group , fourteen from the VPT-no diagnosed injury group and fourteen from the term-born control group were recruited . One VPT-perinatal brain injury participant was excluded from both PET and MRI analysis as a diagnosis of hypothyroidism was discovered at assessment . Incomplete PET data were acquired in one subject from the VPT-no diagnosed injury group because the participant felt unwell and finished the PET scan early . This participant was also excluded from further analysis . In addition to the two participants ( one perinatal brain injury , one very preterm no diagnosed injury ) excluded from the PET study , four further participants ( three perinatal brain injury , one control ) were not included in the MRI study due to contraindications to scanning . Thus , thirteen individuals from the VPT-perinatal brain injury group , thirteen from the VPT-no diagnosed injury group and thirteen from the term-born control group had complete PET- and MRI-derived measures . VPT-perinatal brain injury participants had a lower gestational age and birth weight than VPT-no diagnosed injury participants ( Table 1 ) . This was expected as lower gestational age at birth and birth weight are strongly associated with increased risk of perinatal brain injury ( Vollmer et al . , 2003 ) . There were no group differences in age at scanning , IQ , injected dose , gender , alcohol consumption , smoking or socio-economic status between the groups in the PET sample ( Table 1 ) . There was a significant effect of group on Kicer corresponding to a partial eta-squared of 0 . 233 ( a large effect size , Table 2 ) . Post-hoc tests showed Kicer was significantly reduced in the VPT-perinatal brain injury group compared to the VPT-no diagnosed injury group ( p=0 . 023 , Cohen’s d = 1 . 36 ) and controls ( p=0 . 010 , Cohen’s d = 1 . 07 ) in the whole striatum with large effect sizes ( Figure 1 Table 2; see also associated Figure 1 source data and Create Figure 1 script ) . There was no significant difference in Kicer between the VPT-no diagnosed injury group and controls ( Figure 1 , Table 2 ) . The reduction in dopamine synthesis capacity was significant in the caudate nucleus and the nucleus accumbens , but not the putamen ( see Table 2 ) . Additional sensitivity analyses showed that the reduction in Kicer in the VPT-perinatal brain injury group remained significant when removing all participants who had a history of psychiatric diagnosis ( VPT-perinatal brain injury group n = 4 , VPT-no diagnosed injury group n = 2 , control group n = 1 ) in the whole striatum ( F = 4 . 825 , p=0 . 023 ) and the caudate nucleus ( F = 5 . 608 , p=0 . 023 ) but not the nucleus accumbens ( F = 3 . 047 , p=0 . 061 ) . Furthermore , when just including the participants who also took part in the MRI study ( and hence had individual FreeSurfer-based striatal segmentations ) , reduced Kicer in the VPT-perinatal brain injury group remained significant in the whole striatum ( F = 5 . 708 , p=0 . 018 ) , the caudate nucleus ( F = 10 . 130 , p=0 . 003 ) and in the nucleus accumbens ( F = 4 . 306 , p=0 . 034 ) . The reduction in Kicer in the VPT-perinatal brain injury group remained significant when co-varying for age , IQ , region-of-interest ( i . e . whole striatum , caudate , putamen or nucleus accumbens ) volume and intracranial volume in the whole striatum ( F = 7 . 113 , p=0 . 005 ) , the caudate nucleus ( F = 7 . 083 , p=0 . 005 ) and in the nucleus accumbens ( F = 3 . 663 , p=0 . 037 ) . The two VPT groups differed not only on perinatal brain injury status , but also on gestational age at birth and birth weight ( Table 1 ) . Furthermore , younger gestational age and lower birth weight are both common risk factors for perinatal brain injury ( Vollmer et al . , 2003 ) . When we combined these three neonatal risk factors into a single model to predict whole striatal Kicer , perinatal brain injury remained a significant predictor of dopamine synthesis capacity ( F = 9 . 23 , p=0 . 006 ) , but neither gestational age at birth ( F = 0 . 01 , p=0 . 929 ) , nor birth weight ( F = 0 . 01 , p=0 . 925 ) significantly predicted dopamine synthesis capacity . In order to further probe whether group differences in Kicer varied across striatal subregions , we performed a repeated-measures ANOVA with striatal subregion as the within-subjects factor , group as the between subjects factor and Kicer as the dependent variable . There was no significant subregion-by-group interaction ( F = 1 . 03 , p=0 . 398 ) . As expected , there were significant effects of subregion ( F = 81 . 26 p<0 . 001 ) and group ( F = 6 . 95 , p=0 . 003 ) . There was a significant difference in hippocampal volumes across the three groups ( Table 3 ) . The VPT-perinatal brain injury group had significantly lower volumes than controls , while the VPT-no diagnosed injury group did not differ significantly from either group ( Table 3 ) . The group differences in hippocampal volume remained significant after controlling for intracranial volume ( ICV ) ( F = 7 . 19 , p=0 . 002 ) . On assessing striatal volume with repeated-measures ANOVA , with striatal sub-region volume as a within-subjects factor and group as a between-subjects factor , we found no significant main effect of group ( p=0 . 081 ) and no significant group*subregion interaction ( p=0 . 123 ) . Analysing the whole striatum and each sub-region separately using one-way ANOVAs and post-hoc t-tests confirmed that there were no significant between-group volumetric differences in the striatum ( Table 3 ) . We additionally analysed the estimated striatal volumes for all individuals with PET scans ( i . e . including those without MRI ) , and again found that there were no significant between group differences in striatal volume as a whole ( F = 0 . 77 , p=0 . 628 ) or in any striatal subregion after FDR correction for multiple comparisons ( caudate , F = 0 . 17 , p=0 . 841; putamen , F = 2 . 73 , p=0 . 154 ) , although there was a trend for differences between groups in the volume of the nucleus accumbens , which did not reach significance ( F = 4 . 38 , p=0 . 076 ) . There were no statistically significant differences between the control group and both VPT groups in dlPFC volume ( Raw volumes , F = 0 . 711 , p=0 . 499; Relative volumes , F = 1 . 169 , p=0 . 324 ) . A significant correlation was observed between hippocampal volume and Kicer in the caudate ( r = 0 . 34 , p=0 . 032 , Figure 2A; see also associated Figure 2—source data 1 and Create Figure 2 script ) and in the nucleus accumbens ( r = 0 . 32 , p=0 . 049 , Figure 2B ) across the whole sample . These associations remained significant when controlling for ICV ( caudate Kicer - hippocampal volume , r = 0 . 39 , p=0 . 017; nucleus accumbens Kicer , r = 0 . 34 , p=0 . 036 ) . In order to test the interaction between hippocampal volume and striatal subregion , we again performed a repeated-measures ANOVA , with subregion as the within-subject factor , hippocampal volume and intracranial volume as covariates and and Kicer as the dependent variable . We found a significant effect of hippocampal volume ( F = 4 . 90 , p=0 . 033 ) , but no hippocampal volume by striatal subregion interaction ( F = 0 . 88 , p=0 . 420 ) . Additionally , there was no significant effect of ICV on Kicer ( F = 1 . 11 , p=0 . 299 ) . We then examined whether the relationship between hippocampal volume and striatal Kicer varied significantly by group , again by using region as a within-subjects factor , and this time having group and the group-by-hippocampal-volume interaction term as between-subjects factors . Again we found a significant main effect of group ( F = 4 . 794 , p=0 . 015 ) , but no group by hippocampal volume interaction ( F = 0 . 41 , p=0 . 747 ) . We recently conducted a large-scale structural analysis in an overlapping sample ( Karolis et al . , 2017 ) . In that study , we found evidence of accelerated maturation of the prefrontal cortex , and slower maturation of the caudate nucleus in adults born very preterm . Within the prefrontal cortex , the dlPFC ( anatomically the caudal middle frontal gyrus ) shows consistent grey matter reductions in schizophrenia ( Glahn et al . , 2008 ) , and reduced activation during working memory in adults born very preterm with perinatal brain injury ( Froudist-Walsh et al . , 2015 ) . We therefore examined the relationship between dlPFC volume , hippocampal volume and striatal dopamine synthesis capacity . Using a general linear model , we used as dependent variables Kicer in the two subregions of the striatum which previously showed significant associations with hippocampal volume , namely the caudate nucleus and the nucleus accumbens , and used hippocampal and dlPFC volumes as independent variables . In the caudate model , hippocampal volume was still a significant predictor of dopamine synthesis capacity ( F = 4 . 45 , p=0 . 043 ) , but dlPFC volume was not ( F = 0 . 05 , p=0831 ) . This was also the case when using relative , instead of raw dlPFC volumes ( hippocampus: F = 4 . 79 , p=0 . 03; dlPFC: F = 0 . 33 , p=0 . 569 ) . In the nucleus accumbens model , neither the hippocampus ( F = 1 . 81 , p=0 . 187 ) nor the dlPFC ( F = 1 . 56 , p=0 . 221 ) significantly predicted dopamine synthesis . Again , using relative dlPFC volumes did not alter the result ( hippocampus: F = 2 . 20 , p=0 . 148; dlPFC: F = 0 . 99 , p=0 . 328 ) . We recently demonstrated that in adulthood individuals born very preterm continue to display impairments in executive function , which are associated with less real-life achievement ( Kroll et al . , 2017 ) . Striatal dopamine synthesis capacity has previously been associated with executive function ability , showing an inverted U-shaped relationship such that striatal dopamine synthesis capacity at both the upper and lower ends of the normal range are associated with poorer executive performance ( Cools and D'Esposito , 2011 ) . We performed an additional exploratory analysis to assess whether dopamine synthesis capacity was associated with performance on the Hayling Sentence Completion Test ( Burgess and Shallice , 1997 ) , Controlled Oral Word Association Test ( COWAT ) ( Ruff et al . , 1996 ) , the Stockings of Cambridge and the Intra-Extra Dimensional Shift tasks from the Cambridge Neuropsychological Test Automated Battery ( CANTAB ) ( Fray et al . , 1996 ) and part B of the Trail Making Test ( Tombaugh , 2004 ) . The relationship between each of these measures of executive function and dopamine synthesis capacity was examined using Spearman correlations . We found no significant associations between striatal dopamine synthesis capacity and executive function in any group ( closest association found in controls between striatal dopamine synthesis and performance on the COWAT: r = 0 . 575 , p=0 . 05 ) . We recently showed in an expanded sample , including the subjects from the present study , that adults born very preterm are more likely to exhibit subclinical symptoms across a range of symptom dimensions ( Kroll et al . , 2017b ) , as assessed using the Comprehensive Assessment of At Risk Mental States ( CAARMS ) ( Yung et al . , 2005 ) compared to controls . In the subsample used in the present study , there were no significant between group differences in subclinical symptoms on any CAARMS subscale ( max F = 1 . 188 , min p=0 . 318 ) . Nonetheless , it is possible that the presence of subclinical symptoms is associated with alterations to the dopamine system . We thus performed an exploratory analysis , to identify potential relationships between subclinical symptom expression and regional striatal dopamine synthesis . We found that , across the entire study sample , there was a negative correlation between dopamine synthesis capacity in the nucleus accumbens , and cognitive symptoms identified by the CAARMS ( r = −0 . 92 , p=0 . 020 ) . At a group level , there was a significant negative correlation between nucleus accumbens dopamine synthesis capacity in the VPT-PBI group , and both cognitive ( r = −0 . 57 , p=0 . 032 ) and negative symptoms ( r = −0 . 57 , p=0 . 035 ) . There were no significant correlations between dopamine synthesis capacity and subclinical symptoms in the other two groups , or in other striatal subregions ( all p>0 . 06 ) .
Adults with a history of macroscopic perinatal brain injury associated with VPT birth had reduced dopamine synthesis capacity in the striatum compared to controls born VPT and those born at term , and reduced hippocampal volume compared to individuals born at term . Individuals born similarly preterm but without evidence of macroscopic brain injury showed no significant differences in presynaptic dopamine synthesis capacity from controls , suggesting that preterm birth in the absence of macroscopic brain injury is not sufficient to disrupt striatal dopaminergic function in adult life . It is possible that perinatal brain insults resulted in a long-lasting reduction in the number of dopaminergic neurons ( Burke et al . , 1992; Chen et al . , 1997 ) or caused a down-regulation in dopamine synthetic enzyme levels , in line with post-mortem findings showing reduced tyrosine hydroxylase expression in dopaminergic neurons following prolonged hypoxia ( Pagida et al . , 2013 ) . One alternative possibility is that a common genetic or environmental cause predisposes to both low striatal dopamine synthesis and the direct causes of perinatal brain injury . We also found that reduced striatal dopamine synthesis capacity was associated with reduced hippocampal volume . Several preclinical models , including the MAM model ( Lodge and Grace , 2007 ) and adult hippocampal lesions ( Wilkinson et al . , 1993 ) , have linked hippocampal damage to increased striatal dopaminergic synthesis and release , and behavioral effects including hyper-responsiveness to stress and amphetamine , which are traditionally associated with hyper-dopaminergia ( Kelly et al . , 1975; Pijnenburg and van Rossum , 1973 ) . The MAM model involves injection of the mitotoxin methylazoxymethanol acetate ( MAM ) into the ventral hippocampus of the rat at gestational day 17 . This primarily affects parvalbumin-expressing interneurons , and the resulting reduced inhibitory control leads to increased hippocampal activity , which is sufficient to increase dopaminergic input to the striatum ( Floresco et al . , 2001; Legault et al . , 2000 ) . The neonatal ventral hippocampal lesion model is of particular relevance to the present study due to the vulnerability of the hippocampus to perinatal brain injury . In perhaps the best known result from this animal model , Lipska and colleagues showed that rats that received neonatal excitotoxic lesions of the hippocampus developed hyper-responsiveness to stress and amphetamine , but only after adolescence . Furthermore , these symptoms were successfully treated with haloperidol , a dopamine D2 receptor antagonist ( Lipska et al . , 1993 ) . Later investigation of dopamine synthesis and release in this model by the same group surprisingly found relatively reduced dopamine release , and lower dihydrophenylacetate ( DOPAC ) concentrations indicating reduced dopamine synthesis in response to stress and amphetamine in the lesioned group compared to controls ( Lillrank et al . , 1999 ) . Another study , examining the same lesion model found similar behavioural effects in response to amphetamine , but no alterations to presynaptic dopaminergic function , and led the authors to conclude that ‘presynaptic release of DA had no major contribution to lesion-enhanced DA transmission in the mesolimbic DA system’ ( Wan et al . , 1996 ) . This suggests that similar behavioural symptoms can be evoked by either increased presynaptic dopamine synthesis and release or other mechanisms , such as increased postsynaptic D2 receptor sensitivity . The present study suggests that the first mechanism is not present in humans who were born very preterm or suffered perinatal brain injury . It should also be recognized that reduced presynaptic dopamine synthesis could be a secondary consequence of increased autoregulatory feedback ( Jauhar et al . , 2017 ) , potentially due to increased tonic synaptic dopamine levels in the striatum . Whether increased postsynaptic D2 receptor sensitivity or increased synaptic dopamine levels are seen in humans born very preterm should be tested in further studies . Dopamine also has effects on neurodevelopment , influencing neuronal migration , neurite outgrowth and synapse formation ( Money and Stanwood , 2013 ) , and these effects are particularly marked during the second half of a typical pregnancy ( Kostović and Jovanov-Milosević , 2006 ) , indicating that dopaminergic changes could also influence hippocampal development . Untangling the timing of dopaminergic or hippocampal alterations would seemingly require serial measurements of both systems over the perinatal period , which likely requires post-mortem or preclinical studies . We did not find evidence of a link between dlPFC volume and presynaptic striatal dopamine synthesis in the present sample . It is possible that measures of fronto-striatal connectivity may be more sensitive to detect the effects of prefrontal cortex on striatal dopamine transmission than volumes ( Tziortzi et al . , 2014 ) . Alternatively , other striatal dopaminergic mechanisms , such as dopamine release , may be more directly affected by prefrontal input to the striatum ( Cachope and Cheer , 2014 ) . These results may have implications for cognitive function in people born preterm . While the current group of study participants were not cognitively impaired , cognitive deficits are commonly found in individuals born VPT , and are exacerbated following perinatal brain injury ( Nosarti et al . , 2011 ) . Both longitudinal studies of individuals born preterm and preclinical studies have suggested a link between neonatal hippocampal injury and later working memory impairments ( Beauchamp et al . , 2008; Lipska et al . , 2002; Nosarti and Froudist-Walsh , 2016 ) . The dopaminergic system is crucial for cognitive functions such as reward-based learning ( Schultz et al . , 1997 ) and working memory ( Williams and Goldman-Rakic , 1995 ) , and both hypo- and hyper-dopaminergic function lead to suboptimal cognitive performance ( Cools and D'Esposito , 2011 ) . In the present study , we did not find an association between striatal dopamine synthesis and several measures of executive function . An important limitation of this finding is that our battery of cognitive tests did not include a comprehensive assessment of working memory . Working memory is a particularly common deficit in children born with perinatal brain injury ( Anderson et al . , 2010; Ross et al . , 1996 ) , and is associated with academic outcome in this population ( Mulder et al . , 2010 ) . Individuals with lower presynaptic dopamine synthesis in the caudate nucleus tend to have worse working memory performance ( Cools et al . , 2008; Landau et al . , 2009 ) and respond better to dopamine agonists as cognitive enhancers than individuals with higher baseline dopamine synthesis ( Cools et al . , 2009 ) . VPT individuals with perinatal brain injury who experience working memory deficits could benefit from dopamine agonists as cognitive enhancers , perhaps by dopamine’s role in enhancing intrinsic plasticity mechanisms ( Calabresi et al . , 2007 ) that have been observed in this population ( Froudist-Walsh et al . , 2015; Froudist-Walsh et al . , 2017 ) . Reduced dopamine synthesis capacity is also associated with substance dependence ( Ashok et al . , 2017; Bloomfield et al . , 2014 ) , major depression ( Martinot et al . , 2001 ) and Parkinson’s disease ( Pavese et al . , 2011 ) . Our findings thus suggest that people with perinatal brain injury could be at increased risk for a number of neuropsychiatric disorders . We recently found that individuals born very preterm experience elevated subclinical psychiatric symptoms across a broad range of symptom dimensions ( Kroll et al . , 2017b ) . Here , in an exploratory analysis , we found a negative correlation between striatal dopamine synthesis capacity and subclinical cognitive and negative symptoms in adults born very preterm with perinatal brain injury . ‘Cognitive symptoms’ refer to subjective experience of cognitive change , including concentration , memory and attention problems , whereas ‘negative symptoms’ refer to items such as social isolation , anhedonia and depression . The reduced dopamine synthesis in this group may provide a biological explanation for cognitive and internalising aspects of the ‘preterm behavioural phenotype’ ( Johnson and Marlow , 2011 ) . In contrast , dopamine synthesis capacity is increased in the majority of people with schizophrenia ( Howes et al . , 2012 ) and people at risk of schizophrenia ( Howes et al . , 2011 ) . As yet there have been no PET studies specifically of those people with schizophrenia who have had severe obstetric complications , although it is known that they are especially likely to have small left hippocampi ( Stefanis et al . , 1999 ) . Nevertheless , it is not clear how our results fit with findings that obstetric complications increase the risk of schizophrenia , where interaction with genetic risk factors is likely to be involved ( Howes et al . , 2017; Nicodemus et al . , 2008 ) . In contrast to the increased dopamine synthesis capacity seen in most schizophrenia patients , those who develop schizophrenia-like psychoses following abuse of drugs ( Thompson et al . , 2013 ) , and those with treatment resistant schizophrenia do not share this increased synthesis capacity ( Demjaha et al . , 2012 ) . It is thus possible that the relationship between VPT birth , perinatal brain injury and increased risk for psychosis does not depend on presynaptic dopamine synthesis capacity . It may be important to closely monitor the condition of those individuals born VPT with perinatal brain injury who are treated with antipsychotic medication , as reducing an already-reduced dopaminergic system could lead to unintended extrapyramidal and cognitive effects . Alternatively , it is possible that hypersensitive postsynaptic dopaminergic D2 receptors could unite the seemingly discordant findings of reduced presynaptic dopamine synthesis and increased psychosis risk , as appears to be the case in substance-dependent patients with schizophrenia ( Thompson et al . , 2013 ) . If such disruption were to occur during development , it could have dramatic effects on the developing brain ( Abi-Dargham , 2017 ) , with pre-frontal dependent cognitive functions such as working memory being particularly vulnerable ( Simpson and Kellendonk , 2017 ) . Our finding that there are not marked alterations in dopamine synthesis capacity in the VPT-no diagnosed injury group is also important for the large numbers of people born preterm , as it indicates that the development of the dopamine system , or at least those aspects related to dopamine synthesis , is not disrupted long-term in the absence of macroscopic perinatal brain injury . The VPT-perinatal brain injury and VPT-no diagnosed injury groups in the present study also differed in gestational age , and birth weight , as these neonatal risk factors tend to co-occur ( Vollmer et al . , 2003 ) . Nonetheless , when all three factors were introduced in the same model , only perinatal brain injury was a significant predictor of adult dopamine synthesis capacity . This suggests that reduced striatal dopamine synthesis capacity in adulthood is specific to those individuals with perinatal brain injury . From a methodological perspective , it is possible that between-group differences in the accuracy of image registration may contribute to the apparent reduction in dopamine synthesis capacity seen in the VPT-perinatal brain injury group . However , we used the subject’s own MRI to define the PET region of interest which should mitigate , although not entirely avoid , this risk . Moreover , the results remained significant after controlling for both striatal and total intracranial volume or excluding subjects without MRI scans , suggesting that volume reductions or normalisation differences do not account for the findings . The postnatal ultrasound scans exclude macroscopic brain injury in the VPT-no diagnosed injury group but do not exclude a variety of other microscopic alterations . However , this would not explain our results , as it would , if anything , reduce group differences . Lastly , the final sample size for individuals with combined PET and MRI data of 13 individuals per group is not large . However , PET studies of presynaptic dopamine synthesis with clinical samples have consistently been able to detect group differences with group sizes of between 5 and 12 individuals ( Hietala et al . , 1999; Howes et al . , 2009; Lindström et al . , 1999; Meyer-Lindenberg et al . , 2002; Reith et al . , 1994 ) . Nonetheless , further studies with larger samples investigating pre- and post-synaptic dopamine function in the striatum and other brain areas may help to identify the precise mechanism that links perinatal brain injury with psychiatric risk in adulthood . In summary , we found reduced presynaptic dopamine synthesis capacity in the striatum in individuals born VPT with macroscopic perinatal brain injury . This may help to guide pharmacological interventions for cognitive deficits in this group . We additionally found significant associations between dopaminergic function and reduced hippocampal volume . These results indicate there are long-term neurochemical and structural consequences of perinatal brain injury .
We assessed a group of individuals born VPT who were admitted to the Neonatal Unit of University College Hospital , London in 1979–1985 . These individuals were enrolled in a longitudinal study and have been studied periodically for their entire lives . Macroscopic perinatal brain injury was qualitatively assessed in all participants born VPT and diagnosis of perinatal brain injury was made after consensus between at least two neuroradiologists with a special interest in neonatology . Hemorrhage into the germinal matrix , and those extending to the lateral ventricles or brain parenchyma was labeled as periventricular hemorrhage ( Stewart et al . , 1983 ) , with the grade defined according to the criteria described by Papile and colleagues ( Papile et al . , 1978 ) . Ventricular dilatation was defined as visible dilatation of the lateral ventricles with cerebrospinal fluid while being insufficient to meet the criteria for hydrocephalus . We compared the perinatal brain injury group to: ( 1 ) a group of VPT individuals who were similarly assessed at birth but not diagnosed as having perinatal brain injury ( to control for the effects of preterm birth ) and ( 2 ) healthy controls without a history of perinatal brain injury or preterm birth ( control group ) . Participants who gave consent at previous study time-points to be contacted regarding the study were recruited using the contact details provided previously , and control participants were recruited via advertisements in the local community . Exclusion criteria for all groups were history of post-natal head injury , neurological condition ( including stroke , meningitis , multiple sclerosis , and epilepsy ) or significant physical illness ( such as endocrine or metabolic disorder requiring treatment ) , substance dependence or abuse , psychotic disorder , current antipsychotic use , and pregnancy . The study was undertaken with the understanding and written informed consent and consent to publish of each subject , with the approval of the London Bentham Research Ethics Committee ( Study 11/LO/0732 ) , and in compliance with national legislation and the Code of Ethical Principles for Medical Research Involving Human Subjects of the World Medical Association ( Declaration of Helsinki ) . Birth weight was recorded for all VPT participants and socio-economic status measured in all subjects using the Standard Occupational Classification ( Her Majesty’s Stationary Office , 1991 ) . In adulthood , all participants underwent a 3 , 4-dihydroxy-6-[18F]-fluoro-/-phenylalanine ( [18F]-DOPA ) scan in a Biograph 6 PET/CT scanner with Truepoint gantry ( SIEMENS , Knoxville , TN ) . Subjects were asked to fast from midnight and abstain from smoking tobacco and consuming food and liquids ( except for buttered toast and water ) from midnight before the day of imaging to ensure there were no group differences in amino acid consumption prior to the scan . On the day of the PET scan , a negative urinary drug screen was required and a negative pregnancy test was required in all female subjects . Subjects received carbidopa 150 mg and entacapone 400 mg orally 1 hr before imaging to reduce the formation of radiolabeled [18F]-DOPA metabolites ( Cumming et al . , 1993; Guttman et al . , 1993 ) . Head position was marked and monitored via laser crosshairs and a camera , and minimized using a head-strap . A transmission CT scan was performed before radiotracer injection for attenuation and scatter correction . Approximately 150 MBq of [18F]-DOPA was administered by bolus intravenous injection 30 s after the start of PET imaging . We acquired emission data in list mode for 95 min , rebinned into 26 time frames ( 30 s background frame , four 60 s frames , three 120 s frames , three 180 s frames , and fifteen 300 s frames ) . On a separate day an MRI scan was performed on a 3 Tesla GE Signa MR scanner ( GE Healthcare ) . T1-weighted images were acquired ( TR/TE/TI: 7 . 1/2 . 8/450 ms , matrix: 256 × 256 ) , allowing for 196 slices with no gap and an isotropic resolution of 1 . 1 × 1 . 1 × 1 . 1 mm3 . To correct for head movement , nonattenuation-corrected dynamic images were denoised using a level 2 , order 64 Battle-Lemarie wavelet filter ( Turkheimer et al . , 1999 ) , and individual frames were realigned to a single frame acquired 10 min after the [18F]-DOPA injection using a mutual information algorithm ( Studholme et al . , 1996 ) . Transformation parameters were then applied to the corresponding attenuation-corrected frames , and the realigned frames were combined to create a movement-corrected dynamic image ( from 6 to 95 min following [18F]-DOPA administration ) for analysis . Automatic reconstruction of the hippocampus , caudate nucleus , putamen , nucleus accumbens and cerebellum was performed in the native space of each of the participants with MRI data , allowing for both individual masks and regional volume information extraction , using FreeSurfer version 5 . 1 ( Fischl et al . , 2002 ) . FreeSurfer assigns an atlas label to voxels via use of a probabilistic atlas of region location , which was previously created from a manually labelled training set . Importantly in order to register the atlas and the structural input image , a registration procedure is used that is robust to ventricular enlargement ( Fischl et al . , 2002 ) . The accuracy of the FreeSurfer segmentations of the striatal structures , hippocampus and cerebellum , was assessed by visual comparison with the intensity-corrected t1-weighted scan , which has high grey-white matter contrast around the structures of interest . The primary striatal region of interest was the whole striatum ( nucleus accumbens , caudate and putamen combined ) but we also report the sub-regions separately to determine if there were sub-regional variations . A linear transformation was created between each participant’s T1-weighted structural scan and their individual PET image using FSL FLIRT ( Jenkinson et al . , 2002 ) . This transformation was then applied to each of the previously specified regions of interest in order to obtain individually defined masks of the striatum on the PET scan . Intra-subject registration is generally more accurate than between-subject registration , as there is no between-subject anatomical variability to take into account . In addition to the two participants ( one perinatal brain injury , one very preterm no diagnosed injury ) excluded from the PET study , four further participants ( three perinatal brain injury , one control ) were not included in the MRI study due to contraindications to scanning . In order allow for the inclusion of these participants’ data in the PET analysis , we created a study-specific PET template using Advanced Normalization Tools ( ANTs ) ( Avants et al . , 2011 ) . The template we created was an average of each individual summed PET scan , after mapping onto a common space . We mapped each individual’s FreeSurfer regions-of-interest ( ROIs ) to this custom template again using ANTs . These ROIs were binarised and summed together before being thresholded in order to include only voxels in which the striatum was present in more than 50% of participants . This custom striatum mask was then warped back into the native PET space for those subjects who did not have MRI scans using the inverse ( template-to-native ) transformation that was generated using ANTs . All PET ROIs were visually inspected for accuracy . Once the ROIs were defined in native PET space , we determined [18F]-DOPA uptake [Kicer ( min–1 ) ] , for each ROI using the Gjedde-Patlak graphic analysis adapted for a reference tissue input function ( Patlak and Blasberg , 1985 ) . The cerebellum region was used as the reference region as it represents non-specific uptake ( Kumakura and Cumming , 2009 ) . We additionally undertook exploratory analyses in order to investigate the relationship between presynaptic dopamine synthesis capacity in the striatum and dorsolateral prefrontal cortex volume , executive function abilities and subclinical psychiatric symptoms . Relative and absolute dlPFC volumes were taken from our recent large scale study of volumetric alterations following preterm birth ( Karolis et al . , 2017 ) . Ten individuals from the VPT-perinatal brain injury , 13 individuals from the VPT-no diagnosed group , and 12 controls were included in both studies . Briefly , in that study grey matter volume was analysed at three hierarchical levels , global , modular , and regional . We analysed both raw dorsolateral prefrontal cortex volume , and relative volume ( after regressing out global and module-specific grey matter volumes ) . Measures of executive function were taken from our recent study of cognitive outcome and real-life function ( Kroll et al . , 2017 ) . 16 individuals from the VPT-perinatal brain injury , 13 individuals from the VPT-no diagnosed group , and 12 controls were included in both studies . Briefly , the Hayling Sentence Completion Test ( HSCT ) ( Burgess and Shallice , 1997 ) assessed initiation and suppression responses . The Controlled Oral Word Association Test ( COWAT ) ( Benton and Hamsher , 1976 ) measured verbal fluency . Two subtests from the Cambridge Neuropsychological Test Automated Battery ( CANTAB ) ( Fray et al . , 1996 ) were included . The Stockings of Cambridge ( SOC ) is a task that assesses spatial planning . The Intra-Extra Dimensional Set Shift ( IED ) is a task involving maintaining attention to a reinforced stimulus and then shifting attention to a previously irrelevant stimulus . The Trail Making Test part B ( Tombaugh , 2004 ) measured visual attention , set shifting , and cognitive flexibility . Assessment of subclinical psychiatric symptoms was taken from a larger study ( Kroll et al . , 2017b ) using the Comprehensive Assessment of At-Risk Mental States ( CAARMS ) ( Yung et al . , 2005 ) . 14 individuals from the VPT-perinatal brain injury , 11 individuals from the VPT-no diagnosed group , and 10 controls were included in both studies . ANOVA was used to test the primary hypotheses that there was an effect of group on whole striatal dopamine synthesis capacity and hippocampal volume . p-values from the ANOVAs were adjusted using FDR correction across striatal subregions ( appropriate for positively correlated samples ) ( Benjamini and Hochberg , 1995 ) . Additional sensitivity analyses were conducted using an ANCOVA with Kicer as the dependent variable , group as the independent variable and possible confounds ( age , IQ , intra-cranial and striatal ROI volume ) as covariates . Separately , in those participants born very preterm , we tested for the independent effects of three neonatal risk factors ( perinatal brain injury , gestational age at birth and birth weight ) on dopamine synthesis capacity in the whole striatum using an ANCOVA , with Kicer as the independent variable , group ( VPT-perinatal brain injury vs VPT-no diagnosed injury ) as an independent variable and gestational age at birth and birth weight as covariates . In order to test for regional differences in the effect of VPT birth and perinatal brain injury on dopamine synthesis capacity ( analyzing the entire sample ) , we performed a repeated-measures ANOVA with striatal subregion as the within-subjects factor , group as the between-subjects factor and Kicer as the dependent variable . To test for a hippocampal volume by striatal subregion interaction , we again performed a repeated-measures ANOVA , with subregion as the within-subject factor , hippocampal volume and intracranial volume as covariates and and Kicer as the dependent variable . To examine whether the relationship between hippocampal volume and striatal Kicer varied significantly by group , we used region as a within-subjects factor , with group and the group-by-hippocampal-volume interaction term as between-subjects factors . A two tailed p value<0 . 05 was taken as significant . In our exploratory analyses the following methods were used . To assess the relationship between dlPFC volume , hippocampal volume and striatal dopamine synthesis capacity , we used a general linear model , with dlPFC volume and hippocampal volume as dependent variables , and either caudate nucleus or nucleus accumbens dopamine synthesis as independent variables . The relationship between dopamine synthesis capacity and both executive function measures was assessed with Spearman correlations . Statistical analysis was performed in MATLAB 9 . 2 ( RRID:SCR_001622 ) and SPSS Version 23 ( RRID:SCR_002865 ) . Supporting data are available on request: please contact: oliver . howes@kcl . ac . uk | Thirteen million infants are born too early every year . Improved care allows many to survive , but these “preterm infants” still face an increased risk of death and many other complications . Infants born very early , before 32 weeks , are at risk of brain injury because the brain is normally still developing in the later stages of pregnancy . They also have an increased risk of developing mental health problems later in life . Early-life brain injuries in rats cause changes in the production of a chemical called dopamine . Dopamine is a chemical messenger in the brain that reinforces rewarding behaviour . People with schizophrenia and attention deficit hyperactivity disorder ( ADHD ) have abnormal levels of dopamine . Changes in brain dopamine levels may explain why early-life brain injury is linked to later mental illness . But first scientists must study whether similar changes occur in humans with an early-life brain injury . Now , Froudist-Walsh et al . use brain imaging to show that people born very early who suffered a brain injury have lower dopamine levels than other adults . Imaging techniques were used to scan the brains of 13 adults who were born before 32 weeks and who had a brain injury around birth , 13 adults born before 32 weeks without a brain injury , and 13 adults born at “full term” ( around 39 to 40 weeks ) . Individuals with low dopamine levels reported difficulty concentrating and a lack of motivation and enjoyment in their lives . Both can be warning signs of mental health problems . People born prematurely without a brain injury had normal dopamine levels and did not report such symptoms . More studies may help scientists understand how early brain injuries may cause brain chemical differences later in life , and how these brain changes affect individual’s mental health . They may also help scientists develop treatments to prevent or treat mental illness in people who experienced a brain injury after a very early birth . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | The effect of perinatal brain injury on dopaminergic function and hippocampal volume in adult life |
Myosins play essential roles in the development and function of auditory organs and multiple myosin genes are associated with hereditary forms of deafness . Using a forward genetic screen in Drosophila , we identified an E3 ligase , Ubr3 , as an essential gene for auditory organ development . Ubr3 negatively regulates the mono-ubiquitination of non-muscle Myosin II , a protein associated with hearing loss in humans . The mono-ubiquitination of Myosin II promotes its physical interaction with Myosin VIIa , a protein responsible for Usher syndrome type IB . We show that ubr3 mutants phenocopy pathogenic variants of Myosin II and that Ubr3 interacts genetically and physically with three Usher syndrome proteins . The interactions between Myosin VIIa and Myosin IIa are conserved in the mammalian cochlea and in human retinal pigment epithelium cells . Our work reveals a novel mechanism that regulates protein complexes affected in two forms of syndromic deafness and suggests a molecular function for Myosin IIa in auditory organs .
Mechanosensory receptor cells have organelles derived from modified cilia or microvilli that contain protein complexes dedicated to the detection of , and adaptation to , mechanical force . Myosins , a family of eukaryotic actin-dependent motor proteins , play key roles in the assembly and function of mechanosensory protein complexes . In humans , pathogenic variants of six different myosin genes cause syndromic and non-syndromic deafness , and in many cases these myosins regulate either the assembly of the mechanotransduction apparatus of sensory hair cells , or constitute an integral part of the mechanotransduction complex itself ( Petit and Richardson , 2009 ) . For example , Myosin VIIa is a motor protein present in the tips of hair cell stereocilia where mechanotransduction occurs but it is also present in the cuticular plate that is important for the growth and stability of the stereociliary hair bundle ( Ahmed et al . , 2013 ) . Pathogenic variants of MYO7A , the human homologue of myosin VIIa , can cause Usher syndrome , the leading cause of deaf-blindness ( Bonnet and El-Amraoui , 2012 ) , as well as the non-syndromic forms of deafness DFNA11 ( Liu et al . , 1997 ) and DFNB2 ( Weil et al . , 1997 ) . Dominant mutations in MYH9 , which encodes Myosin IIa , cause a number of syndromes which are now grouped as 'MYH9-related disorders' ( Seri et al . , 2003 ) . Many MYH9-related disorder patients exhibit sensorineural deafness , and variants of MYH9 have also been reported in non-syndromic deafness DFNA17 ( Lalwani et al . , 2000 ) . However , the cellular basis of deafness in pathogenic variants of MYH9 is unclear as MYH9 is widely expressed within the inner ear ( Etournay et al . , 2010; Lalwani et al . , 2000; Meyer Zum Gottesberge and Hansen , 2014; Mhatre et al . , 2006 ) . One approach to identifying new genes that regulate the development and function of mechanosensory organs is to exploit the power of Drosophila to conduct forward genetic screens . The auditory organ of Drosophila , Johnston’s organ , is localized in the second antennal segment . Johnston’s organ responds to near-field sound , gravity and wind flow transduced by motion of the third antennal segment ( Boekhoff-Falk and Eberl , 2014; Gopfert and Robert , 2001; Kamikouchi et al . , 2009; Yorozu et al . , 2009 ) . Although the organs and cells that mediate hearing in vertebrates and Drosophila are morphologically different , they share a striking evolutionary conservation of molecular and functional properties ( Albert and Gopfert , 2015; Boekhoff-Falk and Eberl , 2014 ) . The transcriptional cascades that control key aspects of chordotonal development in flies and hair cell development in vertebrates are regulated by conserved transcription factors , such as the Atonal/Atoh1 family proteins ( Jarman et al . , 1993; Wang et al . , 2002 ) . In addition , myosins such as Myosin VIIa , encoded by the gene crinkled in Drosophila , that function in mammalian hair cell mechanotransduction , are also conserved in Drosophila and are required for hearing ( Todi et al . , 2005b , 2008 ) . Therefore , other molecular pathways and regulatory protein partners that function in hearing are also likely to be shared between insects and vertebrates . Here , we describe a novel ubiquitination pathway in Drosophila that functions to regulate the activity and physical interaction of two proteins implicated in deafness , Myosin II and Myosin VIIa . We identified an E3 ubiquitin ligase , ubr3 , from a collection of lethal mutations on the Drosophila X chromosome ( Haelterman et al . , 2014; Yamamoto et al . , 2014 ) , whose loss of function causes morphological defects in the Johnston’s organ . Ubr3 negatively regulates the mono-ubiquitination of Myosin II and modulates Myosin II-Myosin VIIa interactions , which are required for normal development of Johnston’s organ . We show that ubr3 mutations are phenotypically similar to known pathogenic variants of Myosin II and that Ubr3 physically and genetically interacts with Drosophila homologues of the Usher syndrome proteins Protocadherin 15 ( Pcdh15 ) and Sans . We also show that Myosin IIa interacts with Myosin VIIa in the mouse cochlea and human retinal pigment epithelial cells . Our study reveals a novel conserved ubiquitination pathway in the auditory organs of flies and mammals .
Johnston’s organ is a large chordotonal organ located in the second antennal segment of Drosophila ( Figure 1A ) . Each organ consists of more than 200 functional units or scolopidia ( Kamikouchi et al . , 2006 ) , containing 2~3 sensory neurons , each bearing a single specialized mechanosensitive cilium ( Figure 1B–B’ ) ( Eberl and Boekhoff-Falk , 2007 ) . The neuronal cilium is enveloped by a tube-like scolopale cell , which forms septate junctions with a cap cell that attaches the scolopidium to the cuticle of the third antennal segment . A ligament cell attaches the other end of the scolopidium to the cuticle of the second antennal segment ( Figure 1A , B ) . Each scolopidium is thus suspended between the second and third antennal segments , and rotation of the third antennal segment leads to flexion of the scolopidia and stimulation of the sensory neurons ( Boekhoff-Falk and Eberl , 2014 ) . 10 . 7554/eLife . 15258 . 003Figure 1 . ubr3 regulates auditory organ development in Drosophila . ( A ) The structure of the Drosophila auditory organ , Johnston’s organ . The tips of the neuronal cilia are anchored to the cuticle of the third antennal segment by a dendritic cap containing an extracellular glycoprotein , NompA ( B ) A single scolopidium ( corresponding to the box in A ) shows the markers used to label various structures and cells in the scolopidium . ( B’ ) Immunolabeling of Johnston’s organ with NompA ( red ) and phalloidin ( actin , blue ) . ( C–D ) Pupal Johnston’s organs bearing ubr3 mutant clones were stained with phalloidin to label the actin bundles of scolopale cells . Some ubr3 mutant cells ( labeled by GFP ) exhibit scolopidia detached from the apical junction of Johnston’s organ ( arrows ) . ( E ) Extracellular electrophysiological recordings in flies bearing ubr3 mutant clones in Johnston’s organ ( left 2 columns ) and in flies bearing ubr3 cDNA rescued mutant clones ( right column ) . The data are normalized to flies heterozygous for the corresponding mutations . Numbers of flies recorded are shown in the columns . Error bars show SEM . Statistical symbols show results of t-tests with Welch’s correction as needed ( *p<0 . 05; **p<0 . 01; ns , not significant ) . ( F ) Schematic diagram showing the conserved domains of the Ubr3 protein and the molecular lesions ( Phe949 > Leu and Leu788 > STOP ) identified in the ubr3A and ubr3B alleles respectively . The red bar shows the epitope used to generate anti-Ubr3 antibody . ( G–G’ ) A single confocal cross-section shows co-immunolabeling of Johnston’s organ with anti-Ubr3 ( red ) and HRP ( neurons , in blue ) . ubr3B/B mutant clones were generated and labeled with GFP . Ubr3 protein is localized to neuronal cilia marked by an arrowhead . ( H ) A longitudinal section of Johnston’s organ labeled by anti-Ubr3 ( red ) and HRP ( blue ) . Ubr3 localizes not only to neuronal cell bodies but faint expression is also seen in cilia . Arrowhead labels enriched Ubr3 proteins in apical ciliary tips . Scolopale cell bodies ( Sp ) are outlined by dashed lines . ( I ) Diagram shows distribution of Ubr3 proteins in Johnston’s organ , including its enrichments in the apical tips of the neuronal cilia and the scolopale cells ( arrowhead ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 00310 . 7554/eLife . 15258 . 004Figure 1—figure supplement 1 . Ubr3 is required for normal development of Johnston’s organ . ( A ) Immunolabeling of wild type pupal Johnston’s organs with different markers . ( B ) Sample traces of sound evoked potential ( SEPs ) from flies of indicated genotypes . ( C ) Immunolabeling of the second and third antenna segments of a pupal wild type fly with anti-Ubr3 antibody ( red ) and HRP ( neurons , blue ) . ( D ) Immunolabeling of a pupal Johnston’s organ bearing Ubr3 over-expressing clones ( labeled by GFP , green ) with anti-Ubr3 antibody ( red ) and anti-HRP ( neurons , blue ) . Ubr3 proteins are present in neuronal cell bodies ( N ) , indicated by arrow . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 004 To identify genes required for auditory organ development and function , we screened a collection of X-chromosome induced lethal mutations ( Haelterman et al . , 2014; Yamamoto et al . , 2014 ) . We generated mutant clones in Johnston’s organ through FLP/FRT-mediated mitotic recombination using an eyeless-FLP driver . We assessed morphological defects in the constituent cell types of the scolopidia by co-labeling with cell type-specific markers ( neurons: ELAV and HRP; scolopale cells and scolopale space: Prospero and Eyes shut; actin bundles in scolopale cells: phalloidin; ligament cells: Repo; Figure 1B–B’ and Figure 1—figure supplement 1A ) and identified seven complementation groups that showed a range of different morphological defects in Johnston’s organ . One complementation group , ubr3 ( Zanet et al . , 2015 ) , exhibits a specific detachment of the scolopidia from the third antennal segment of mutant clones ( arrows in Figure 1C , D ) . Extracellular electrophysiological recordings in flies with ubr3 mutant clones in Johnston’s organ showed significantly reduced auditory transduction ( Figure 1E and Figure 1—figure supplement 1B ) . The incomplete reduction in sound-evoked potentials was due to the small size of mutant clones in Johnston’s organ ( Figure 1C and D ) . ubr3 encodes a 2219 amino acid protein homologous to the mammalian RING-type E3 ubiquitin ligase n-recognin 3 ( UBR3 ) ( Figure 1F ) ( Huang et al . , 2014; Meisenberg et al . , 2012; Tasaki et al . , 2007; Yang et al . , 2014; Zanet et al . , 2015; Zhao et al . , 2015 ) . Ubr3 contains a UBR substrate binding domain , a RING E3 ligase domain and a C-terminal auto-inhibitory ( AI ) domain . To determine the expression pattern and protein localization of Ubr3 , we stained antennae at 50% of pupariation , when the scolopidia mature , with a specific Ubr3 antibody ( Zanet et al . , 2015 ) ( Figure 1F ) . Ubr3 is broadly expressed in the second and third segments of the antenna ( Figure 1—figure supplement 1C ) , containing Johnston’s organ and olfactory neurons respectively . A prior study reported expression of mouse UBR3 in multiple sensory tissues , including the inner ear and olfactory epithelium ( Tasaki et al . , 2007 ) . In Johnston’s organ , Ubr3 is enriched in the apical tips of neurons and scolopale cells ( Figure 1G–I , arrowheads , and Figure 1—figure supplement 1D ) . Loss of ubr3 in Johnston’s organ leads to detachment of scolopidia from the hinge of the second and third antennal segment ( Figure 1C , D ) . This phenotype has previously been reported for only one other Drosophila gene , crinkled ( also known as myosin VIIa or myoVIIa ) ( Todi et al . , 2005b , 2008 ) ( Figure 2A ) . The detachment of scolopidia in myoVIIa mutants is accompanied by a severely altered distribution of a glycoprotein , NompA , that links the tip of the neuronal cilium with the antennal cuticle ( Figure 2A–C ) ( Todi et al . , 2005b , 2008 ) . NompA is the homologue of vertebrate tectorins , a glycoprotein family present in the tectorial membrane of the cochlea ( Chung et al . , 2001 ) . We observed a similar change of NompA distribution in ubr3 mutant cells ( Figure 2D , E ) , which can be rescued by over-expressing wild type Ubr3 proteins with an actin-Gal4 driver ( Figure 2—figure supplement 1A–B ) . However , over-expression of an E3 enzymatic inactive form of Ubr3 ( Li et al . , 2016 ) did not rescue the detachment of scolopidia in ubr3B/Bmutant cells ( Figure 2—figure supplement 1A–B ) , suggesting that Ubr3 regulates apical attachment of scolopidia through its E3 ligase activity . To confirm if the ubiquitination function of Ubr3 is necessary for its role in Johnston’s organ , we knocked down ubcD6 using RNAi . ubcD6 encodes the E2 enzyme that interacts with Ubr3 in mammals ( Tasaki et al . , 2007; Zanet et al . , 2015 ) . As shown in Figure 2F , knock down of ubcD6 also causes scolopidial detachment ( Figure 2F ) , suggesting that the phenotype in ubr3 mutant cells is caused by failure of ubiquitination of one or more target proteins . 10 . 7554/eLife . 15258 . 005Figure 2 . Ubr3 genetically interacts with MyoVIIa . ( A ) The normal filamentous structure of NompA in the apical junction of wild-type cells ( white box ) collapses into puncta in the detached scolopidia in flies in which myoVIIa is knocked down ( yellow box ) . Arrow indicates detached scolopidia . ( B ) A diagram shows actin ( cyan ) and NompA ( red ) in a single scolopidium . ( C ) A diagram illustrates the detachment of scolopidia and altered NompA pattern . ( D–E ) The normal filamentous structure of NompA in the apical junction of wild-type cells ( white boxes ) collapses into puncta in the detached scolopidia in ubr3 mutant cells ( labeled by GFP ) ( yellow boxes ) . ( F ) The normal filamentous structure of NompA in the apical junction of wild-type cells ( white box ) collapses into puncta in the detached scolopidia in cells over-expressing ubcD6 RNAi construct ( labeled by GFP ) ( yellow box ) . ( G ) Immunolabeling of Johnston’s organ with ubr3 mutant clones ( marked by GFP , green ) by anti-HRP ( neurons , blue ) and anti-MyoVIIa antibody ( red ) . Arrows indicate detached ubr3 mutant scolopidia . ( H ) A diagram shows localization of MyoVIIa ( red ) in neuronal cilia and scolopale cells . ( I ) Quantification of detached scolopidia in the ubr3A/A and ubr3B/B clones , ubcD6 RNAi clones , and wild type or mutant clones over-expressing myoVIIa . Error bars show SEM . Numbers of flies quantified are shown in the columns . ( ***p<0 . 001 ) DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 00510 . 7554/eLife . 15258 . 006Figure 2—figure supplement 1 . ubr3 mutants phenocopy myoVIIa mutants . ( A ) Immunolabeling of pupal Johnston’s organs bearing ubr3 mutant clones over-expressing wild type Ubr3 or E3 ligase-dead form of Ubr3 ( labeled by GFP , green ) with anti-NompA antibody ( red ) and phalloidin ( actin , blue ) . ( B ) Quantification of detached scolopidia in cells with indicated genotypes in Johnston’s organ . Numbers of flies quantified are shown in the columns . ( C ) Immunolabeling of a pupal Johnston’s organ bearing myoVIIa RNAi expressing clones ( labeled by GFP , green ) with anti-MyoVIIa antibody ( red ) and anti-HRP ( neurons , blue ) . ( D–D’ ) Enlargement of the ciliary region of the scolopidia from a wild type Johnston’s organ , labeled by anti-MyoVIIa ( red ) and anti-HRP ( cyan ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 006 MyoVIIa is an unconventional myosin expressed in hair cells in the vertebrate inner ear and has been shown to be localized to the tip of the stereocilia close to the proposed sites of mechanotransduction ( Grati and Kachar , 2011 ) . In Drosophila , we found that MyoVIIa is abundant in the scolopale cells of Johnston’s organ ( Figure 2G , H and Figure 2—figure supplement 1C , D–D’ ) , being especially enriched at their apical tips ( arrows in Figure 2—figure supplement 1D–D’ ) . The specific and unique phenotype observed in ubr3 and myoVIIa mutant cells suggests that Ubr3 and MyoVIIa may function in the same genetic pathway . To test if ubr3 interacts genetically with myoVIIa , we first generated ubr3B/B; myoVIIaRNAi cells . However , we observed nearly 100% detachment of scolopidia ( data not shown ) , similar to that seen with the myoVIIaRNAiknockdown alone ( Figure 2A ) . We then over-expressed MyoVIIa in ubr3 mutant cells and observed a strong enhancement of the ubr3 mutant phenotype ( Figure 2I ) whereas over-expression of MyoVIIa in wild-type cells did not cause any detachment of scolopidia ( Figure 2I ) . These observations indicate a specific genetic interaction between ubr3 and myoVIIa . Our data suggest that ubiquitination by Ubr3 regulates the apical attachment of scolopidia in the Drosophila hearing organ , and the similarity of ubr3 and myoVIIa mutants raises the possibility that it may regulate the abundance , localization or function of MyoVIIa . However , mutant clones of ubr3 in Johnston’s organ do not exhibit altered protein levels or subcellular localization of MyoVIIa ( Figure 2I ) , suggesting that Ubr3 instead regulates MyoVIIa function . To determine if Ubr3 regulates ubiquitination of MyoVIIa , we purified GFP-MyoVIIa fusion protein from wild type or ubr3 mutant clone cells in eye-antennal discs in third instar larvae followed by western blotting ( Figure 3A , B , arrow ) . We then assessed the ubiquitination of MyoVIIa with an anti-poly- and mono-ubiquitin antibody or an anti-poly-ubiquitin antibody ( Figure 3B ) . Interestingly , although we did not observe ubiquitination of MyoVIIa , we detected mono-ubiquitination of a MyoVIIa-interacting protein , which migrates lower than MyoVIIa-GFP as a 220 kDa protein ( arrowhead in Figure 3B ) . The mono-ubiquitination of this 220 kDa protein is increased in ubr3 mutant cells ( Figure 3B , C ) , suggesting that Ubr3 indirectly regulates the function of MyoVIIa by ubiquitinating an unknown , interacting partner . 10 . 7554/eLife . 15258 . 007Figure 3 . Ubr3 negatively regulates the mono-ubiquitination of MyoII and MyoII-MyoVIIa interaction . ( A ) A GFP amino terminal tagged MyoVIIa construct , GFP-MyoVIIa , is expressed in wild type ( control ) or ubr3 mutant clones in larval eye-antennal discs . The lysate of eye-antenna discs and brains containing ubr3 mutant cells expressing GFP-MyoVIIa protein was immunoprecipitated with GFP nanobody-conjugated beads and examined on western blots . ( B ) Western blots with anti-GFP , anti-poly & mono-ubiquitin , anti-mono-ubiquitin and anti-MyoII antibodies . ( C ) Quantification of mono-ubiquitination of MyoII normalized by total amount of MyoII proteins from ( B ) . ( D–D’’ ) Immunolabeling of GFP ( green ) , MyoVIIa ( red ) and HRP ( neurons , in blue ) in Johnston’s Organ from a myoII-GFP-myoII transgenic fly . D’ and D’’ show magnified images of the region shown by white box in D . ( E ) Distribution of MyoII proteins in Johnston’s organ . ( F , G ) Wild type MyoII over-expressing cells show normal apical structures of scolopidia , whereas ubr3B/B mutant cells expressing wild type MyoII exhibit enhanced detachment of scolopidia . ( H ) Quantification of detached scolopidia in ubr3B/B mutant cells , ubr3B/B mutant cells over-expressing MyoII and wild type cells over-expressing MyoII . Error bars show SEM . Numbers of flies quantified are shown in the columns . ( I ) Diagram shows HA-MyoII-Ub in which MyoII is fused to a Ub coding sequence on the carboxyl terminal . ( J–K’ ) Johnston’s organ containing HA-MyoII-Ub expressing clones ( labeled by GFP , green ) is immunolabeled by anti-NompA ( red ) and phalloidin ( actin , in blue ) . Arrow marks detached scolopidia . ( K–K’ ) One MyoII-Ub expressing scolopidium exhibits accumulated NompA at the tips , but stays attached to the cuticle from the third segment ( arrowheads ) . This may be a defective scolopidium just before detaching , suggesting that NompA mis-localization happens prior to apical detachment , as opposed to being a consequence of detachment . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 00710 . 7554/eLife . 15258 . 008Figure 3—figure supplement 1 . Ubr3 regulates MyoVIIa through Cul1 . ( A , A’ ) Loss of ubr3 ( GFP ) causes a strong up-regulation of Cul1 in auditory sensory neurons ( blue ) in Johnston’s organ . ( B , B’ ) Clones over-expressing Cul1 produce similar apical detachment of scolopidia as that observed in ubr3 mutant cells . ( C , C’ ) Down-regulation of Cul1 through RNAi causes apical detachment of scolopidial cells . ( D–E’ ) Down-regulation of skpA or roc1a , SCF E3 ligase components , leads to scolopidial detachment ( arrows ) . ( F ) Quantifications of detached scolopidia shown in B–E . Error bars show SEM . Numbers of flies quantified are shown in the columns . ( G ) An amino terminal tagged MyoVIIa construct , GFP-MyoVIIa , is expressed in clones that are wild type ( control ) , skpA or ubr3 mutant cells in larval eye-antennal discs . Eye-antennal discs and brain tissues were dissected from third instar larvae and homogenized in tube . Lysate from the eye-antennal discs and brains was immunoprecipitated with GFP nanobody conjugated beads . Western blots were performed with anti-GFP , anti-poly & mono-ubiquitin , anti-mono-ubiquitin and anti-MyoII antibodies . ( H ) Quantification of mono-ubiquitination of MyoII normalized by total amount of MyoII from G . Error bars show SEM . ( I ) A simple working model shows that Ubr3 negatively regulates mono-ubiquitination of MyoII through Cul1 ( SCF ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 008 To identify this unknown protein , we performed mass spectrometry from the 220 kDa band and found the Drosophila homologue of the heavy chain of non-muscle Myosin II ( MyoII ) , encoded by the gene zipper . To verify that MyoII is the target of Ubr3 , we probed the membrane with an anti-MyoII antibody . The MyoII antibody detected a band at the same molecular weight as the ubiquitinated proteins ( Figure 3B , shown by anti-MyoII antibody ) . Interestingly , we observed more MyoII co-precipitating with MyoVIIa in ubr3 mutant cells compared to wild-type cells ( Figure 3B ) , suggesting that loss of ubr3 leads to a stronger MyoVIIa-MyoII interaction . To test if MyoII has a similar function as MyoVIIa in Johnston’s organ , we generated a myoII-GFP-myoII knock-in line by integrating an artificial exon that encodes GFP with flexible linkers into an intron of myoII ( Nagarkar-Jaiswal et al . , 2015a , 2015b; Neumüller et al . , 2012; Venken et al . , 2011 ) . This GFP-tagged protein is fully functional , as it complements a deletion spanning the myoII gene . We detected abundant MyoII protein in both neurons and scolopale cells ( Figure 3D , E ) . We observed an enrichment of MyoII proteins at the apical tips of cilia , in the vicinity of apically-enriched MyoVIIa protein ( Figure 3D’–D’’ , E ) . In addition , over-expression of MyoII in ubr3 mutant cells increased the penetrance of detached scolopidia from 30% to 60% ( Figure 3G , H ) , while over-expression of MyoII in wild-type cells did not affect scolopidial structure ( Figure 3F , H ) . Thus , our data show that MyoII and MyoVIIa interact with ubr3 , and that over-expression of either MyoII or MyoVIIa in Johnston’s organ can cause scolopidial detachment provided ubr3 is also mutated . To understand the function of mono-ubiquitination of MyoII , we fused a ubiquitin coding sequence to the carboxyl terminal of MyoII cDNA ( MyoII-Ub ) to artificially mimic the mono-ubiquitinated MyoII ( Figure 3I ) . When we over-expressed this MyoII-Ub construct in clones in Johnston’s organ , we again observed detached scolopidia and an altered NompA pattern ( Figure 3J–K’ ) . The penetrance of the phenotype is lower than that observed in ubr3 mutant cells ( ~5% ) , probably because carboxyl terminal fused ubiquitin does not function as optimally as those at the normal ubiquitination sites in MyoII . We were surprised to observe that loss of the E3 ubiquitin ligase ubr3 caused increased mono-ubiquitination of MyoII ( Figure 3B , C ) . One potential mechanism is that Ubr3 may negatively regulate a second ubiquitin ligase complex that in turn mono-ubiquitinates Myo II . We tested the expression of widely expressed E3 ligase , Cullin1 , a component of the SCF E3 ubiquitin ligase complex ( Deshaies , 1999; Wu et al . , 2005 ) , and found that it is strongly up-regulated in ubr3 mutant clones in imaginal discs . We observed a similar up-regulation of Cul1 in ubr3 mutant cells in Johnston’s organ ( Figure 3—figure supplement 1A ) , suggesting that Ubr3 negatively regulates Cul1 . To test if Cul1 was also implicated in scolopidial attachment , we over-expressed Cul1 in wild-type cells and found that it recapitulated the apical detachment of scolopidia that was seen in ubr3 and myoVIIa mutant clones ( Figure 3—figure supplement 1B , F ) . This suggests that increased Cul1 expression is likely to cause scolopidial detachment in ubr3 mutant clones . In addition , RNAi-mediated down-regulation of Cul1 produced a similar detachment of scolopidia ( Figure 3—figure supplement 1C , F ) . The observation that both gain- and loss-of-function of Cul1 produce the same specific detachment phenotype in Johnston’s organ implies that a critical range of Cul1 activity or level is necessary for its proper function . To test if other components of the SCF E3 ubiquitin ligase complex , SkpA and Roc1a ( Murphy , 2003; Noureddine et al . , 2002 ) , affect scolopidial attachment we performed RNAi experiments and also observed detachment ( Figure 3—figure supplement 1D–F ) . These data indicate that the phenotypes associated with Cul1 over-expression and down-regulation are mediated by the SCF E3 ligase complex . However , the mono-ubiquitination of MyoII is increased in skpA mutant cells , ( Figure 3—figure supplement 1G–H ) , again showing that the SCF is not the direct E3 ligase that mono-ubiquitinates MyoII ( Figure 3—figure supplement 1I ) . Together , our data show that Ubr3 negatively regulates Cul1 ( SCF ) E3 ligase , and that both E3 ligases control the mono-ubiquitination of MyoII and apical attachment of scolopidia . Dominant pathogenic variants in MYH9 , one of three human paralogues of myoII , cause MYH9-related disease in human ( Ma and Adelstein , 2014; Seri et al . , 2003 ) and affect hearing to varying degrees ( Pecci et al . , 2014; Verver et al . , 2015 ) . We over-expressed four mutant forms of Drosophila myoII in Johnston’s organ that contain variants commonly found in individuals with MYH9-related disorders who have sensorineural hearing loss ( Franke et al . , 2007 ) ( Figure 4—figure supplement 1A ) and observed scolopidial detachment with variable penetrance in all variant forms of myoII ( Figure 4A–E ) . In addition , we tested whether these variants alter the localization of MyoII . Over-expression of all four MyoII variants led to the formation of puncta in neurons , in contrast to a diffuse localization of wild type MyoII ( Figure 4F–J ) . We observed similar puncta formation when the constitutively ubiquitinated form of MyoII , MyoII-Ub , was over-expressed in neurons ( Figure 4K ) . However , the MyoII variants and MyoII-Ub exhibited comparable localization with wild type MyoII when expressed in the scolopale cells ( Figure 4F’–J’ ) . In addition , we performed immunoprecipitation of GFP-tagged wild type or variants of MyoII expressed in eye-antenna discs using ey-Gal4 . We found that MyoIID1847K exhibits an increased interaction with MyoVIIa ( Figure 4—figure supplement 1B ) . The data for the other three are either not changed or there is a decrease in their interaction . This suggests that a mis-regulated MyoII-MyoVIIa interaction may be present in some patients with MYH9-related disorders . Our data suggest that pathogenic variants of MyoII exhibit dominant toxic effects in Johnston’s organs and lead to similar phenotypes to those seen in ubr3 mutants . 10 . 7554/eLife . 15258 . 009Figure 4 . Over-expression of pathogenic variants of MyoII in Johnston’s organ leads to similar defects as ubr3 mutants . ( A–D ) Johnston’s organs over-expressing four different GFP tagged MyoIImut in clones ( labeled with GFP in green ) are immunolabeled with anti-NompA ( red ) and phalloidin ( actin , in blue ) . Arrows mark detached scolopidia . ( E ) Quantification of detached scolopidia in the clone cells expressing the four MyoII mutant forms shown in A . Error bars show SEM . Numbers of flies quantified are shown in the columns . ( F–K’ ) Johnston’s organs expressing GFP-MyoII , MyoII mutant forms or MyoII-Ub in neurons using nsyb-Gal4 driver ( F–K ) or in scolopale cells using nompA-Gal4 ( F’–K’ ) are immunolabeled with anti-GFP ( green ) and HRP ( neurons , in red ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 00910 . 7554/eLife . 15258 . 010Figure 4—figure supplement 1 . Pathogenic mutations of MyoII . ( A ) Structure of MyoII protein and the most common pathogenic mutations used in this study . ( B ) GFP-tagged wild type MyoII or pathogenic variants of MyoII were expressed in eye-antenna discs using ey-Gal4 . Immuno-precipitation was performed using lysate from eye-antenna discs against GFP tags , followed by western blots . ( C ) RT-PCR results from ARPE-19 cells transfected with different siRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 010 To examine if mammalian UBR3 regulates MyoIIa ( encoded by MYH9 ) and its interaction with MyoVIIa , we turned to a human retinal pigment epithelial cell line ARPE-19 , one of the few cell lines that express MyoVIIa ( Soni et al . , 2005 ) as well as MyoIIa and Ubr3 ( see below ) . In wild type ARPE-19 cells , MyoIIa predominantly localizes to stress fiber-like structures ( Figure 5A , B , arrows ) . A small proportion of MyoIIa is present in puncta in the cytoplasm ( arrowheads ) . Interestingly , most MyoVIIa protein in ARPE-19 cells co-localize with MyoIIa in both stress fibers and cytoplasmic puncta , although some MyoVIIa does not overlap with MyoIIa in the peri-nuclear region ( empty arrowhead ) . 10 . 7554/eLife . 15258 . 011Figure 5 . The function of Ubr3 is conserved in vertebrate cells . ( A–B ) Cultured ARPE-19 human cells transfected with indicated siRNAs are co-immunolabeled by anti-MyoIIa and anti-MyoVIIa antibodies and DAPI ( white ) . Arrows: stress fibers . Arrowheads: MyoIIa-MyoVIIa co-localized puncta . Empty arrowheads: MyoVIIa positive , MyoIIa negative puncta . ( C–D ) Quantifications of stress fiber number and MyoIIa-MyoVIIa puncta shown in A–B . Error bars show SEM . Numbers of cells quantified are shown in the columns . ( E , F ) ARPE-19 cells treated with indicated concentrations of blebbistatin for 30 min followed by immunolabeling with anti-MyoIIa ( red ) and anti-MyoVIIa ( green ) antibodies and DAPI . Low concentration ( 2–4 μM ) of blebbistatin treatment resulted in elongated ARPE-19 cells with protrusions , similar as UBR3 siRNA-treated cells . Further increasing the dosage of blebbistatin ( 8–16 μM ) resulted in cells with a more branched , tree-like morphology . The number of puncta correlated with the concentration of blebbistatin , suggesting a specific change in MyoII-MyoVIIa interactions . ( G–G’ ) Immunolabeling of a cochlear section from a neonatal mouse with anti-MyoIIa ( red ) and phalloidin ( green ) . Hair cells are outlined by dashed lines . ( H–H’ ) Surface view of whole mount cochlea from a neonatal mouse immunolabeled with anti-MyoIIa ( red ) and phalloidin ( green ) . Arrowheads mark V-shaped stereocilia ( labeled by phalloidin , green ) . ( I–I’ ) Immunolabeling of cochlear section from P6 pup with anti-MyoIIa ( red ) and phalloidin ( green ) . Arrowheads mark stereocilia ( shown by phalloidin staining in green ) . ( J ) Co-immunoprecipitation with anti-MyoIIa antibody from P5 cochlear lysate followed by western blotting . ( K ) MyoIIa was purified from ARPE-19 cells through immuno-precipitation , followed by western blot . Arrowheads indicate ubiquitinated MyoIIa ( shown by FK2 antibody ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 01110 . 7554/eLife . 15258 . 012Figure 5—figure supplement 1 . MyoII proteins localize close to cell membrane in the hair cells of mouse cochlea . ( A ) Cultured ARPE-19 human cells transfected with indicated siRNAs or treated with 4 μM blebbistatin are co-labled with phalloidin ( Actin , red ) and DAPI ( white ) . ( B ) Single section of confocal image shows distribution of MyoIIa ( red ) in hair cells ( labeled by phalloidin , green ) . MyoIIa proteins localize near cell membrane of hair cells ( arrowheads ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 012 To test whether UBR3 regulates MYH9 in ARPE-19 cells , we knocked down UBR3 with three independent siRNAs , all of which down-regulated UBR3 mRNA to ~30% ( Figure 4—figure supplement 1C ) . Upon down-regulation of UBR3 , the ARPE-19 cells become elongated with long protrusions ( Figure 5A ) . We observed a consistent decrease of stress fiber-localized MyoIIa and MyoVIIa and an increase of MyoIIa-MyoVIIa co-localized puncta in all siRNA transfections ( Figure 5B–D ) . To test whether these changes are caused by defective MyoIIa , we treated the cells with different doses of blebbistatin , an inhibitor for MyoIIa ( Bond et al . , 2013 ) . Interestingly , cells treated with low doses of blebbistatin ( 2–4 μM in Figure 5E , F ) mimic UBR3 knock-down cells ( Figure 5A , B ) . The stress fibers are decreased in the UBR3 knocked-down cells or in the cells treated with blebbistatin ( Figure 5—figure supplement 1A ) , suggesting that the formation of stress fibers in these cells are mis-regulated . These results imply that loss of UBR3 leads to defects in MyoIIa function . To further examine the MyoIIa interaction with MyoVIIa in mammals , we examined the localization of MyoIIa protein in the neonatal mouse cochlea . MyoIIa protein is expressed weakly in both hair cells and supporting cells ( Figure 5G–H ) . In the cell body of hair cells , MyoIIa localizes at the apical surface in neonatal mice ( Etournay et al . , 2010 ) and faintly at junctions with supporting cells ( Ebrahim et al . , 2013 ) ( Figure 5—figure supplement 1B , arrowheads ) . However , MyoIIa is restricted to hair cell stereocilia in six day old mice ( Figure 5I–I' , arrowheads ) . To assess whether MyoIIa physically interacts with MyoVIIa , we performed an immunoprecipitation assay from inner ear lysates of five day old mice . Similar to what we observed with Drosophila MyoII and MyoVII , MyoVIIa co-immunoprecipitated MyoIIa ( Figure 5J ) . In addition , we detected a ubiquitinated form of MyoIIa in ARPE-19 cells ( Figure 5K ) . Therefore , the MyoIIa-MyoVIIa inetraction and ubiquitination of MyoIIa are conserved in mammals . Previous studies have shown that at least five members of the USH1 protein family , including MyoVIIa , can interact to form a complex ( Adato et al . , 2005; Boeda et al . , 2002; Kazmierczak et al . , 2007; Reiners et al . , 2005; Senften et al . , 2006 ) and localize to the tips of hair cell stereocilia in mice ( Figure 6A ) ( Grati and Kachar , 2011; Hilgert et al . , 2009 ) . These protein complexes are thought to interact with and regulate hair cell mechanotransduction channels ( Gillespie and Muller , 2009 ) . Although Drosophila homologues of several USH proteins have been identified ( D'Alterio et al . , 2005; Demontis and Dahmann , 2009; Todi et al . , 2005b ) , only MyoVIIa has been shown to be expressed in the scolopidia of Johnston’s organ and to be required in Drosophila for hearing ( Todi et al . , 2005b , 2008 ) . To test whether other components of the Usher protein complex are conserved in the fly hearing organ and whether Ubr3 can interact with other Drosophila USH proteins in addition to MyoVIIa , we characterized the expression , phenotypes and protein interactions of several USH1 homologues . 10 . 7554/eLife . 15258 . 013Figure 6 . Ubr3 , Cul1 and MyoVIIa interact with Drosophila homologues of Usher proteins . ( A ) Diagram of a vertebrate hair cell and localization of USH1 proteins in stereocilia . ( B–B’ ) Johnston’s organ of a fly carrying a homozygous GFP knock-in allele of Cad99C is labeled with HRP ( blue , neurons ) , phalloidin ( actin , in green , scolopale cells ) and anti-GFP ( red , Cad99C proteins ) . ( C ) Localization of Cad99C proteins in neuronal cilia and the tip region of scolopale cells . ( D ) Johnston’s organ of Cad99C57A mutant is stained with phalloidin ( actin , blue ) and NompA ( red ) . Arrows indicate two detached scolopidia . ( E ) S2 cells transfected with the indicated constructs were lysed and immunoprecipitated with GFP nanobody conjugated beads . Western blots were performed with various antibodies . In the input fraction for the immunoprecipitation , two proteins can be detected: a short 37 kDa carboxyl terminal domain ( arrow ) and a full length 250 kDa protein ( arrowhead ) . MyoVIIa-GFP ( empty arrowhead ) , Cul1-GFP ( square ) , UBR-GFP ( empty square ) , and untagged GFP proteins ( dot ) . UBR-GFP was used because we could not detect expression of Ubr3 full length protein here . ( F ) Johnston’s organs containing ubr3B/B , ubr3B/B Cad99C57A/+ or ubr3B/B sans245/+ clone cells ( GFP , green ) are stained with phalloidin ( actin , blue ) and NompA ( red ) . Arrows mark detached GFP+ scolopidia and arrowheads mark un-detached GFP+ scolopidia . ( G ) Quantification of detached scolopidia . Error bars show SEM . Numbers of flies quantified are shown in the columns . ***p<0 . 001DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 01310 . 7554/eLife . 15258 . 014Figure 6—figure supplement 1 . Ubr3 , Cul1 and MyoVIIa interact with Drosophila homologues of Usher proteins . ( A ) Diagram shows the protein structure of Cad99C . ( B ) HA-Cad99C-V5 fusion proteins ( shown in the right bottom ) were expressed in S2 cells . Anti-HA immunoprecipitation or anti-V5 immunoprecipitation was performed with lysate from these cells , followed by western blots with indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 014 The Drosophila homologue of the vertebrate Usher syndrome gene PCDH15 is Cad99C ( D'Alterio et al . , 2005; Schlichting et al . , 2006 ) . To examine expression of Cad99C , we integrated an artificial exon that encodes GFP with flexible linkers into an intron of Cad99C using the MiMIC technique ( Neumüller et al . , 2012; Venken et al . , 2011 ) . Cad99C was enriched in the apical part of the actin-rich scolopale cells , in a similar manner to MyoVIIa and MyoII ( Figure 6B , C ) . To examine if Ubr3 , Cul1 or MyoVIIa physically interact with Cad99C , we performed co-immunoprecipitation assays in S2 cells . Cad99C encodes a transmembrane cell adhesion protein with 11 cadherin repeats in the extracellular domain and a short intracellular domain ( Figure 6—figure supplement 1A ) . Interestingly , the 37 kDa intracellular fragments of Cad99C ( arrowheads ) co-precipitated with MyoVIIa ( Figure 6E , empty arrowhead ) , Cul1 ( square ) and UBR ( empty square ) , whereas the full length Cad99C protein ( arrow ) did not co-precipitate . To verify the specific interaction with the cleaved , intracellular fragments of Cad99C , we performed the reciprocal co-immunoprecipitation assay ( Figure 6—figure supplement 1B ) and observed consistent results . This indicates that only the short carboxyl-terminal fragments of Cad99C bind to Ubr3 . These results are consistent with previous finding that Cad99C interacts with MyoVIIa through carboxyl-terminal domain in Drosophila ovary ( Glowinski et al . , 2014a ) . We also tested whether a second Usher syndrome homologue , Sans , interacts with Ubr3 and Cul1 . As shown in Figure 6E , MyoVIIa , Cul1 , and the UBR domain all interact with Sans . Altogether , these results demonstrate that Ubr3 and Cul1 physically interact with Cad99C and Sans . To test if the altered MyoVIIa-MyoII interactions seen in ubr3 mutants also affect the function of Drosophila USH protein homologues , we examined mutants for sans and Cad99C in the presence or absence of ubr3 mutations . sans245 , a null mutant for sans ( Demontis and Dahmann , 2009 ) , does not display detachment of scolopidia ( Figure 6G ) . However , Cad99C57A , a null allele ( Schlichting et al . , 2006 ) , exhibits detached scolopidia , albeit at a much lower frequency ( 1% ) than ubr3 or myoVIIa mutants ( Figure 6D , G ) . However , removal of one copy of Cad99C or sans in ubr3 mutant clones increased the penetrance of scolopidial detachment from 30% to 90% or 60% , respectively ( Figure 6F–G ) . These data show that several Drosophila homologues of USH1 proteins co-operate in the Johnston’s organ and suggest that ubr3 genetically interacts with and may regulate a number of Usher complex proteins . This suggests that altering the strength of interactions between MyoII and MyoVIIa also affects the function of other USH1 proteins .
Functional and molecular homologies between hearing organs of Drosophila and mammals have allowed the use of Drosophila to identify new genes involved in hearing ( Boekhoff-Falk and Eberl , 2014; Senthilan et al . , 2012 ) . Using a forward genetic screen based on the phenotype associated with the loss of MyoVIIa ( Todi et al . , 2005a , 2008 ) , we identified a set of proteins that affect ubiquitination and regulate the function of MyoVIIa . We discovered that ubr3 mutations increase the mono-ubiquitination of MyoII and the association of MyoII and MyoVIIa . This increased association causes defects consistent with a reduction of MyoVIIa function , namely the detachment of scolopidia from the cuticle of Johnston’s organ . Significantly , over-expressing either myosin in Johnston’s organ has no effect unless ubr3 is also mutated , suggesting that it is not the absolute levels of either myosin that are important for their function in Johnston’s organ , but rather the level of their ubiquitin-dependent interaction . Myosins are known to be regulated by phosphorylation through their regulatory light chain ( Ito et al . , 2004 ) . However , far less is known about the post-translational regulation of myosin heavy chains . Ubiquitination as a regulator of myosin function has not yet been reported . When we knocked down UBR3 in human ARPE-19 cells , we observed defects similar to inhibition of MyoII by blebbistatin , suggesting that a certain threshold of mono-ubiquitination is sufficient to attenuate the activity of MyoII . Thus , it is possible that phosphorylation and ubiquitination function respectively as an 'accelerator' and 'brake' for MyoII activity . Our data suggests that ubiquitination of MyoII is regulated by multiple E3 ligases ( Figure 7A–A’ ) , in which Ubr3 negatively regulates the Cul1-SCF complex directly or indirectly . This complex probably regulates an E3 ligase that directly ubiquitinates MyoII . 10 . 7554/eLife . 15258 . 015Figure 7 . A novel ubiquitination pathway regulates MyoII-MyoVIIa interactions in the auditory sensory organs of Drosophila and mammals . ( A ) In Drosophila , MyoVIIa and MyoII are present in the apical regions of scolopidia of Johnston’s organ and are enriched in the tips of the scolopale cells where they contact the cap cell . Ubiquitination of MyoII promotes its interaction with MyoVIIa , the precise level of which is crucial for anchoring the apical junction complexes of the scolopidia . It is possible that the motor activity of either myosin is necessary to transport the complex to the tips of the scolopale cell . Both MyoVIIa and MyoII likely bind to the actin bundles in the scolopale cells and regulate apical attachment of scolopidia . Two Drosophila homologues of Usher syndrome type I proteins , Cad99C ( Pcdh15 ) and Sans , interact with MyoVIIa , Ubr3 and Cul1 in a protein complex . It is not clear whether Cad99C mediates attachment to the cap cells as a homodimer or as a heterodimer with another adhesion molecule . ( A’ ) Ubr3 negatively regulates the level of Cul1 protein . Both Ubr3 and Cul1 inhibit ubiquitination of MyoII indirectly through a pathway involving a third unknown E3 ligase . ( A’’ ) MyoVIIa , Cad99C , Sans , Ubr3 and Cul1 interact as a protein complex . ( B ) In mammalian hair cells , an USH1 protein complex which includes MyoVIIa , Sans , Harmonin and Cadherin-23 is present close to the stereocilia tips . We speculate that MyoII interacts with MyoVIIa , and that this interaction is promoted by ubiquitination of MyoII . The motor activity of MyoII or MyoVIIa may be required for transport of the MyoVIIa-MyoII-USH1 protein complex to the stereocilia tips . DOI: http://dx . doi . org/10 . 7554/eLife . 15258 . 015 We were surprised to find that MyoII ubiquitination appears to be regulated by a series of E3 ligases including Ubr3 and Cul1 . Either these E3 ligases function in a sequential order , or alternatively , function collaboratively ( Metzger et al . , 2013 ) . A previous study showed that Fbx2 , an F-box protein that binds Skp1 in SCF E3 ligase complex , is specifically expressed in mouse cochlea , and that Fbx2 deficient mice exhibit selective cochlear degeneration ( Hartman et al . , 2015; Nelson et al . , 2007 ) . Our results argue that the SCF E3 ligase regulates auditory function through MyoIIa and MyoVIIa , two proteins associated with deafness and expressed in hair cells . In addition , multiple E3 ligases , including Ubr3 and a yet unknown E3 ligase that directly ubiquitinates MyoIIa , function in this pathway . It is interesting that Ubr3 and Cul1 can bind to a complex containing MyoVIIa , Cad99C and Sans in Drosophila S2 cells . It is possible that Ubr3 , Cul1 and the unknown E3 ligase that mono-ubiquitinates MyoII all interact in a protein complex , and that the Usher proteins Cad99C or Sans may facilitate ubiquitination of MyoII or may be ubiquintinated themselves . It is currently unclear whether MyoII and MyoVIIa interact directly or indirectly as part of a protein complex in Johnston’s organ . Since both MyoII and MyoVIIa are localized in adjacent zones in the tips of scolopale cells in Johnston’s organ ( Figure 7A ) , it is possible that either myosin is required for the subcellular transport or localization of the MyoVIIa – MyoII complex . The precise mechanical function performed by these two myosins in the scolopale cells of Johnston’s organ involves other homologues of Usher syndrome type I genes given that the morphological phenotype of ubr3 mutants is enhanced by loss of either Sans or Cad99C ( Figure 7A , A’’ ) . We have shown that homologues of three Usher syndrome genes interact together in Drosophila to cause a mechanical failure phenotype which appears to be orthologous to that seen in the mechanosensory hair cells of vertebrates ( Figure 7B ) . Moreover , by revealing a physical interaction between MyoII and MyoVIIa in the hearing organs of insects and mammals and by recapitulating morphological defects in Johnston’s organ by overexpressing Drosophila versions of known pathogenic human MYH9 variants , our study offers a potential mechanism for the hearing deficits associated with MYH9-related disorders .
y w ubr3A FRT19A/FM7c Kr-Gal4 , UAS-GFP and y w ubr3B FRT19A/FM7c , Kr-Gal4 , UAS-GFP flies ( Li et al . , 2016; Zanet et al . , 2015 ) were crossed to tub-Gal80 , y w , eyFLP , FRT19A; Act-Gal4 , UAS-CD8-GFP/CyO to generate GFP-labeled ubr3 homozygous mutant clones . All UAS transgenic flies were generated through φC31-mediated transgenesis ( Venken et al . , 2006 ) . Additional strains used in the study are as follows: Cad99C57A ( Schlichting et al . , 2005 ) , sans245 ( Demontis and Dahmann , 2009 ) , UAS-ubr3 ( Zanet et al . , 2015 ) , UAS-GFP-myoVIIa ( Todi et al . , 2005b ) , UAS-cul1/CyO ( Ou et al . , 2002 ) . UAS-GFP-myoII , UAS-GFP-myoIID1847K , UAS-GFP-myoIID1430N , UAS-GFP-myoIIR1171C , UAS-GFP-myoIIR1939X are kind gifts from Dr . Kiehart ( Franke et al . , 2007 ) . UAS-myoVIIaRNAi ( P{GD1408}v9265 ) , UAS-roc1aRNAi ( P{GD8596} ) ( Dietzl et al . , 2007 ) , UAS-UbcD6RNAi ( TRiP . HMS02466 ) , UAS-cul1RNAi ( TRiP . HM05197 ) , UAS-skpARNAi ( Ni et al . , 2009 ) ( TRiP . HM05185 ) ( Bloomington Drosophila Stock Center , Bloomington , IN ) , nsyb-Gal4 ( FlyBase ID: FBst0051635 , generated by Dr . Julie Simpson , unpublished ) , nompA-Gal4 ( Chung et al . , 2001 ) . All flies were maintained at room temperature and crossed on standard food at 25°C . Fly tissues were dissected in PBS in room temperature and fixed with 3 . 7% formaldehyde in PBS for 20 min , followed by permeabilization with 0 . 2% Triton-X100 in PBS . For whole-mount mouse cochlear staining , cochleas from neonatal mice and 6-day old mice were dissected in PBS , with the spiral ganglia and Reissner’s membrane removed to expose the organ of Corti . For sections of neonatal ear tissue , animal heads were fixed 1–2 hr in 4% PFA at room temperature , cryoprotected in 30% sucrose in PBS at 4°C , embedded in OCT compound , and cryosectioned at 14 µm . For sections of 6-day old mice cochlea , inner ears were dissected in PBS , fixed 1 hr in 4% PFA at room temperature , and decalcified in 0 . 5 M EDTA pH8 . 0 for three days at 4°C . Then the decalcified inner ears were cryoprotected in 30% sucrose in PBS at 4°C , embedded in OCT compound , and cryosectioned . The immunohistochemistry procedure followed standard protocols with some minor modifications . The primary antibodies and secondary fluorescently-labeled antibodies used in this paper were: chicken anti-GFP ( 1:1000 , Abcam , United Kingdom ) , rat anti-ELAV ( 1:1000 , 7E8A10 , DSHB , Iowa City , IA ) ( Robinow and White , 1991 ) , rabbit anti-HRP ( 1:1000 , Jackson Immunoresearch Laboratories Inc . , West Grove , PA ) , mouse anti-Pros ( 1:100 , MR1A , DSHB ) ( Spana and Doe , 1995 ) , mouse anti-Repo ( 1:100 , 8D12 , DSHB ) ( Alfonso and Jones , 2002 ) , rabbit anti-NompA ( 1:250 ) ( Chung et al . , 2001 ) , mouse anti-Futsch ( 1:100 , 22C10 , DSHB ) ( Fujita et al . , 1982; Zipursky et al . , 1984 ) , guinea pig anti-Ubr3 ( 1:1000 ) ( Zanet et al . , 2015 ) , rabbit anti-Cul1 ( 1:250 ) ( Wu et al . , 2005 ) , guinea pig anti-MyoVIIa ( GP6 1:2000 , used in Figure 2G and Figure 2—figure supplement 1C–D ) ( Glowinski et al . , 2014b ) , Mouse anti-MyoVIIa ( 1:10 , 138–1 , DSHB , used in Figure 5A , B , E , F ) , rabbit anti-MyoIIa ( 1:500 , Gene Tex ( Irvine , CA ) GTX113236 , used for ARPE-19 cell and mouse cochlear experiments shown in Figure 5 ) , Alexa488- ( 1: 1000 , Life Technologies , Carlsbad , CA ) , Cy3- and Cy5- conjugated affinity purified donkey secondary antibodies ( 1: 1000 , Jackson ImmunoResearch Laboratories ) . Images were acquired using LSM510 and LSM710 confocal microscopes ( Zeiss , Germany ) and examined and processed using LSM viewer ( Zeiss ) , ZEN ( Zeiss ) and Photoshop ( Adobe ) software . To clone the pUASTattB-HA-MyoII-Ub construct , myoII ( zipper ) cDNA was amplified from genomic DNA of UAS-GFP-myoII transgenic flies . The human Ub sequence was cloned from a construct from Dr . Janghoo Lim from Yale University . myoII cDNA and human Ub sequences were then cloned into pUASTattB vector through EcoRI , NotI and KpnI . An HA sequence was inserted in the primer . To clone pUASTattB-Cad99C-V5 and pUASTattB-sans-V5 constructs , Cad99C and sans cDNAs were amplified from BDGP Gold cDNA clones LP14319 and LD20463 ( Drosophila Genomics Resource Center , Bloomington , IN ) and sub-cloned into pUASTattB vector through NotI and KpnI . A V5 sequence was inserted into the primer at the carboxy terminal of Cad99C or sans in frame . All constructs were verified through Sanger sequencing before use . S2 cells were cultured at 25°C in Schneider’s medium ( Life Technologies ) plus 10% heat-inactivated fetal bovine serum ( Sigma , St . Louis , MO ) , 100 U/mL penicillin ( Life Technologies ) , and 100 μg/mL streptomycin ( Life Technologies ) . Cells were split every 3 days and plated at a density of 106 cells/well in 12-well cell culture plates for experiments . Transfections were carried out using Effectene transfection reagent ( Qiagen , Germany ) . Cells were harvested 48 hr after transfection for biochemical assays . ARPE-19 cells were cultured in at 37°C in 5% CO2 in air in DMEM: F-12 medium ( ATCC ) with 10% fetal bovine serum , as described by the ATCC ( http://www . atcc . org/ ) . Cells were split when reaching 80–90% confluence . siRNAs were transfected using Lipofectamine RNAiMAX Transfection Reagent ( Life Technologies ) . Two days after transfections , cells were lysed for biochemical asssays or fixed for immunolabeling assays . UBR3 siRNAs ( Sigma MISSION Predesigned siRNA ) used in this paper: UBR3 siRNA791: GUUAGAAGGCGCUCUUACA; UBR3 siRNA793: GUACUUAAGAGAAGGCUAU; UBR3 siRNA795: CCGAAAUGUUGUUAGAUAU S2 cells were lysed 48 hr after transfection with plasmids in lysis buffer ( Tris-HCl 25 mM , pH 7 . 5 , NaCl 150 mM , EDTA 1 mM , NP-40 1% , Glycerol 5% , DTT 1 mM ) plus complete protease inhibitor ( Roche , Switzerland ) for 30 min on ice , followed by centrifugation . The supernatant was then immunoprecipitated with agarose beads conjugated to antibodies recognizing different epitope tags , which had been previously equilibrated with lysis buffer , overnight at 4°C . The beads were then washed 3 times in washing buffer ( Tris-HCl 10 mM , pH 7 . 5 , NaCl 150 mM , EDTA 0 . 5 mM ) before boiling in loading buffer . Western blotting was then performed with each sample . To purify MyoVIIa-GFP proteins from clone cells in the eye-antennal discs , tissues were homogenized in RIPA buffer ( Tris-HCl 50 mM , pH 7 . 5 , NaCl 150 mM , sodium deoxycholate 0 . 25% , NP-40 1% , SDS 0 . 1% ) . Immunoprecipitation was performed using the same conditions as above , except for the experiment shown in Figure 5—figure supplement 1C , in which MyoIIa was purified through immuno-precipitation in a denatured condition ( Bloom and Pagano , 2005 ) to avoid pulling down interacting proteins . The following affinity beads were used for immunoprecipitation: Chromotek-GFP-Trap Agarose Beads ( Allele Biotechnology , San Diego , CA ) , monoclonal anti-HA−agarose antibody ( Sigma ) , anti-V5 agarose affinity gel ( Sigma ) , monoclonal anti-MyoIIa ( abcam ab55456 , 1:100 , used in Figure 5J and K ) . The antibodies used in western blot analysis are as following: rabbit anti-GFP ( 1:1000 , Life Technologies ) , rabbit anti-MyoII ( 1:1000 , a gift from Dr . Dan Kiehart , used in Figure 3B and Figure 3—figure supplement 1G ) ( Kiehart and Feghali , 1986 ) , mouse anti-MyoVIIa ( 1:10 , DSHB ) 138–1 , used in Figure 5J ) ( Soni et al , 2005 ) , mouse FK1 ( 1:1000 , Enzo , Farmingdale , NY ) , mouse FK2 ( 1:1000 , Enzo ) , anti-HA ( 1:5000 , Santa Cruz Biotechnology ( Santa Cruz , CA ) , F7 or ( 1:1000 , 16B12 , Covance , Princeton , NJ ) , guinea pig anti-Ubr3 ( 1:1000 ) ( Zanet et al . , 2015 ) , mouse anti-V5 ( 1:5000 , Life Technologies ) , rabbit anti-MyoIIa ( Gene Tex GTX113236 , and 1:1000 for western blot in Figure 5J ) . The intensities of the bands in Figure 3B were quantified using Image J software . For MARCM-mediated knockdown or mutant experiments , the percentage of detached scolopidia was calculated as the number of detached scolopidia in GFP clones / the total number of scolopidia in GFP clones . For Cad99C57A homozygous mutants , detached scolopidia in the each Johnston’s organ were counted and the percentage was calculated by dividing the detached scolopidia number by 230 , the average number of scolopidia per organ ( Kamikouchi et al . , 2006 ) . Statistical calculations were computed using Prism 3 . 0 software . Electrophysiological recordings were performed with electrolytically sharpened tungsten electrodes inserted into the joint between the first and second antennal segments ( recording electrode ) and penetrating the head cuticle near the posterior orbital bristle ( reference electrode ) , in response to near-field playback of computer-generated pulse song , as described in ( Eberl and Kernan , 2011 ) . The signals were subtracted and amplified with a differential amplifier and digitized at 13 kHz . Sound evoked potentials ( SEPs ) were measured as the max-min values in the averaged trace from 10 consecutive presentations of the pulse song , as described . Total RNA was isolated from ARPE-19 cells using Trizol ( Life Technologies ) . Reverse transcription was performed using the Applied Biosystems High-Capacity cDNA Reverse Transcription Kit . RT-PCR was performed using iQ SYBR Green Supermix from BIO-RAD ( Hercules , CA ) and CFX96 Touch Real-Time PCR Detection System . Primers used for the RT-PCR were: UBR3-F ( 5’-TGGCTGTTCAAGGTTTCATAGG-3’ ) and UBR3-R ( 5´- GGTGCCACTGCTTAGTTTTACC-3´ ) , GAPDH-F ( 5’-AATCCCATCACCATCTTCCA-3’ ) and GAPDH-R ( 5’-TGGACTCCACGACGTACTCA-3’ ) . RT-PCR was done with 3 PCR replicates for each biological sample , 3 biological replicates ( 3 independent biological samples in the same experiment ) and was repeated twice ( 2 independent experiments ) . | Over half of the world’s population has hearing loss by the age of 65 , and inherited forms of deafness are responsible for many of the hearing impairments in newborn children . Because the auditory organs that enable insects and mammals to hear work in similar ways , we have learnt a lot about genetic forms of deafness from identifying faulty genes in humans and mice , and studying their effects in model organisms . Sensory cells in the inner ear respond to sound by detecting vibrations in the air and converting them into electrical impulses . A family of motor proteins called myosins play key roles in this conversion process . Mutations in the gene that produces one of these proteins , called myosin VIIa , cause an inherited deaf-blind disorder called Usher syndrome . Mutations in the gene for another type of myosin protein , called myosin II , also cause disorders associated with hearing loss , but it is not clear how they produce such effects . Li et al . have used Drosophila fruit flies to explore the role of myosin proteins in hearing by looking for genes that prevent the insect's auditory organ from developing or working properly . The search identified one gene called E3 ubiquitin ligase ( ubr3 ) , which is required for the auditory organ to develop normally and had not previously been implicated in deafness . Mutating the ubr3 gene caused a defect similar to that seen for mutations in the gene that produces the fruit fly equivalent of myosin VIIa . Through genetic and biochemical studies , Li et al . found that in the fruit flies , myosin VIIa interacts with myosin II . This interaction is regulated by a chemical modification of myosin II that is controlled by ubr3 . Li et al . showed that the equivalent mammalian proteins display the same behaviour in the cells of mammals . Therefore , mutations that affect myosin II alter how the protein interacts with myosin VIIa , which explains why myosin II is associated with deafness in humans . In addition , Li et al . found that three other proteins that have been shown to cause Usher syndrome in humans have equivalents in flies and play a role in fly hearing . This will allow the Drosophila auditory organ to be further developed as a model system for future studies of deafness genes , and should provide insights into how specific genes are required for proper hearing in mammals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2016 | The E3 ligase Ubr3 regulates Usher syndrome and MYH9 disorder proteins in the auditory organs of Drosophila and mammals |
Although aging-regulating pathways were discovered a few decades ago , it is not entirely clear how their activities are orchestrated , to govern lifespan and proteostasis at the organismal level . Here , we utilized the nematode Caenorhabditis elegans to examine whether the alteration of aging , by reducing the activity of the Insulin/IGF signaling ( IIS ) cascade , affects protein SUMOylation . We found that IIS activity promotes the SUMOylation of the germline protein , CAR-1 , thereby shortening lifespan and impairing proteostasis . In contrast , the expression of mutated CAR-1 , that cannot be SUMOylated at residue 185 , extends lifespan and enhances proteostasis . A mechanistic analysis indicated that CAR-1 mediates its aging-altering functions , at least partially , through the notch-like receptor glp-1 . Our findings unveil a novel regulatory axis in which SUMOylation is utilized to integrate the aging-controlling functions of the IIS and of the germline and provide new insights into the roles of SUMOylation in the regulation of organismal aging .
The view that aging is solely driven by stochastic events has changed as mounting evidences showed that certain , apparently independent , genetic and metabolic modulations , slow aging and extend lifespans of various organisms . Dietary restriction ( DR ) , reduced activity of the insulin/IGF signaling pathway ( IIS ) or of the mitochondrial electron transport chain ( ETC ) , and removal of germ cells ( Kenyon , 2005 ) , all slow the pace of aging . Among these , most prominent is IIS reduction , which extends lifespan and elevates stress resistance of worms ( Kenyon et al . , 1993 ) , mice ( Holzenberger et al . , 2003 ) , and presumably humans ( Suh et al . , 2008 ) . In the nematode Caenorhabditis elegans ( C . elegans ) , the sole insulin/IGF receptor , DAF-2 , initiates a signaling cascade that negatively regulates the activity of at least three transcription factors by modulating phosphorylation . Direct phosphorylation of DAF-16/FOXO ( Lee et al . , 2001 ) and of SKN-1/NRF ( Tullet et al . , 2008 ) prevents these factors from entering the nucleus and from regulating their target genes . Similarly , the IIS inhibits the phosphorylation of DDL-1 , which retains the Heat Shock Factor 1 ( HSF-1 ) in the cytosol ( Chiang et al . , 2012 ) . Thus , IIS reduction by daf-2 RNA interference ( RNAi ) or by mutation , hyperactivates its downstream transcription factors , creating long-lived worms ( Kenyon , 2005 ) . IIS reduction also elevates resistance to a variety of stresses including heat ( Lithgow et al . , 1995 ) , ultraviolet ( UV ) radiation ( Murakami and Johnson , 1996 ) , and pathogenic bacteria ( Singh and Aballay , 2006 ) . In addition , IIS reduction protects worms and mice from toxic aggregation ( proteotoxicity ) of various neurodegeneration-causing proteins ( reviewed in Carvalhal Marques et al . , 2015 ) . Finally , IIS reduction also modulates reproduction and egg-laying patterns , as knocking down daf-2 by RNAi , reduces the worm’s brood size but extends the reproduction period ( Dillin et al . , 2002 ) . Although it was shown that the IIS is locally neutralized in germ cells ( Narbonne et al . , 2015 ) and that DAF-2 responds to food availability by modulating oogenesis through RAS-ERK signaling ( Lopez et al . , 2013 ) , whether changes in post-translational modifications are involved in the IIS-mediated control of reproduction , is only partially understood . Moreover , despite the evidences that the ablation of germ cells extends lifespan ( Hsin and Kenyon , 1999 ) and promotes proteostasis in C . elegans ( Shemesh et al . , 2013 ) , it is unclear how the aging-regulating mechanisms downstream of the IIS and those that are activated by the reproduction system are linked , and whether post-translational modifications play roles in the orchestration of these mechanisms . SUMOylation is a post-translational modification involving a reversible covalent attachment of a small ubiquitin-like modifier ( SUMO ) to specific lysine residues of proteins ( Melchior , 2000 ) . While mammals ubiquitously express three forms of SUMO ( SUMO-1 , 2 , and 3 ) , C . elegans expresses only one SUMO-encoding gene , smo-1 that encodes a polypeptide of 91 amino acids with a predicted molecular weight of 10 . 2 kDa ( Choudhury and Li , 1997 ) . SUMOylation controls various biological processes and plays important roles in development and survival ( Johnson , 2004 ) . Among other functions , smo-1 is critically needed for germline development and fertility of the nematode ( Broday , 2017 ) . Here , we examined whether IIS activity controls SUMOylation of C . elegans’ proteins and if this post-translational modification plays roles in aging-associated functions of this pathway . To address this , we compared global SUMOylation patterns of proteins that were extracted from untreated and from daf-2 RNAi-treated animals , and found that among other modulations , IIS reduction lowers the SUMOylation rate of the protein CAR-1 ( Cytokinesis/Apoptosis/RNA-binding protein 1 ) but has no effect on the expression level of car-1 . CAR-1 is an RNA-binding protein , which acts in association with the RNA helicase , CGH-1 in the germline ( Audhya et al . , 2005 ) . The knockdown of car-1 increases the levels of GLP-1 during late oogenesis ( Noble et al . , 2008 ) , thereby leading to germ cell death and to defective embryonic cytokinesis ( Boag et al . , 2005; Squirrell et al . , 2006 ) . We show that knocking down car-1 shortens lifespan and enhances proteotoxicity in model worms . On the contrary , the expression of a mutant CAR-1 , which cannot be SUMOylated on lysine residue 185 ( K185 ) , extends lifespan and promotes proteostasis . These effects are conferred , at least partially , through the GLP-1 axis in a DAF-16-dependent manner , but probably also through an additional , DAF-16-independent pathway . Interestingly , we found that GLP-1 positively controls the expression of car-1 establishing a regulatory circuit . Our findings unveil a novel link between the reproductive system and the IIS , demonstrating that one downstream arm of this pathway regulates certain aspects of aging through the SUMOylation of CAR-1 .
In order to test whether IIS reduction affects global protein SUMOylation in C . elegans , we employed three worm strains: wild-type animals ( strain N2 ) and two conditionally sterile nematode strains: CF512 and CF1903 , all exhibit natural IIS activity . CF512 animals become sterile when exposed to 25°C during development , as they cannot produce sperm . CF1903 animals harbor a temperature-sensitive glp-1 mutant that renders them sterile upon exposure to 25°C during development ( Arantes-Oliveira et al . , 2002 ) . Using these conditionally sterile worm strains , we could compare protein SUMOylation in adult tissues with no background from developing embryos . Eggs of all worm strains were extracted from animals that were grown in 15°C , and placed on plates that were seeded with either control bacteria , harboring the empty RNAi vector ( EV ) , or with daf-2 RNAi expressing bacteria . The plates were incubated at 25°C for 48 hr , transferred to 20°C for additional 24 hr and the worms were harvested at day 1 of adulthood . As expected , wild-type worms were fertile while CF512 and CF1903 worms were sterile ( Figure 1—figure supplement 1 ) . Global protein SUMOylation patterns were determined by western blot ( WB ) analysis , using an anti SUMO antibody . Our results indicated that IIS reduction modulates the patterns of SUMOylation in all three worm strains ( Figure 1A ) , as the SUMOylation levels of several proteins were increased ( arrowheads ) and of others were decreased ( arrows ) upon treatment with daf-2 RNAi . Differences in SUMOylation patterns between strains suggest strain-specific protein SUMOylation . To identify proteins which are differentially SUMOylated upon IIS reduction , we used worms that were deprived of the endogenous smo-1 gene and express a dually tagged smo-1 transgene instead ( His-Flag-smo-1 , strain NX25 [Pferdehirt and Meyer , 2013] ) . NX25 worms were treated from hatching to day 1 of adulthood with daf-2 RNAi or left untreated ( EV ) , harvested and SUMOylated proteins were pulled-down by tandem-purification procedure ( Figure 1B and Figure 1—figure supplement 2 ) . Sediment proteins were analyzed by quantitative Mass Spectrometry ( Supplementary file 1 , full data set can be accessed at http://www . ebi . ac . uk/pride project ID: PXD010011 ) . Our analysis showed that , among other affected proteins , the SUMOylation of CAR-1 is approximately threefold lower in daf-2 RNAi-treated worms compared to the levels observed in untreated NX25 animals . To further examine whether the IIS governs the rate of CAR-1 SUMOylation , we utilized worms that express CAR-1 fused to the green fluorescent protein ( GFP ) under the regulation of the pie-1 promoter ( strain WH346 , GFP-CAR-1 [Squirrell et al . , 2006] ) . These nematodes were used due to the high efficiency and specificity of GFP pulldown . The animals were developed on EV or daf-2 RNAi bacteria , harvested at day 1 of adulthood and GFP-CAR-1 was immuno-precipitated using a GFP antibody , and blotted by an anti SUMO antibody . The intensities of two bands were remarkably higher in homogenates of untreated worms ( EV ) compared to homogenates of daf-2 RNAi-treated worms ( Figure 1C , arrows ) . One band migrated as a ~ 60 kDa protein , the size corresponding to the mono SUMOylated GFP-CAR-1 . The other band migrated as a protein of approximately 250 kDa , suggesting that SUMOylated GFP-CAR-1 is a component of a highly stable protein complex , perhaps with the RNA helicase CGH-1 ( Boag et al . , 2005 ) . This complex appears to be less abundant , or less SUMOylated , in worms that exhibit low IIS activity . To compare the total amounts of GFP-CAR-1 in this pulldown experiment , we re-exposed the blot to an anti GFP antibody and found no difference in the quantities of GFP-CAR-1 molecules which migrated as a protein of approximately 50 kDa ( Figure 1D ) . We next tested whether IIS reduction destabilizes the GFP-CAR-1 protein . WH346 worms were cultured from hatching on EV or daf-2 RNAi bacteria , homogenized at day 1 of adulthood and WB analysis using an anti-GFP antibody was utilized to compare the relative levels of GFP-CAR-1 . Our results showed similar amounts of GFP-CAR-1 in daf-2 RNAi-treated and untreated worms ( Figure 1E ) . Quantification of the GFP-CAR-1 signals in three independent repeats of this WB experiment confirmed that IIS reduction does not significantly changes the levels of this chimeric protein ( Figure 1F ) . Finally , we tested the possibility that the lower level of SUMOylated CAR-1 , observed in daf-2 RNAi-treated worms ( Figure 1C ) , stems from the regulation of car-1 expression by the IIS . To address this , we employed CF512 worms , quantitative real-time PCR ( qPCR ) and car-1-specific primers and found no significant difference in the expression levels of car-1 in daf-2 RNAi-treated and untreated worms ( Figure 1G ) . Taken together , our observations indicate that the IIS modulates the SUMOylation of a sub-population of CAR-1 molecules and show that this signaling pathway affects neither the level of car-1 expression nor the amounts of CAR-1 protein within the worm population . Previous observations regarding the role of CAR-1 as a negative regulator of glp-1 expression in germ cells ( Noble et al . , 2008 ) , the well-documented effects of glp-1 on aging ( Arantes-Oliveira et al . , 2002 ) , and our findings of IIS-mediated SUMOylation of CAR-1 ( Figure 1 ) , have led us to speculate that CAR-1 is involved in the regulation of lifespan . To test this hypothesis , we compared the lifespans of wild-type worms and of nematodes that are car-1 null . To obtain nematodes that lack car-1 , we used worms that carry only one copy of the gene ( strain WH377 ) and selected for progeny that lack both copies of car-1 ( car-1 knockout worms are sterile ) . Lifespans of car-1 knockout worms were found to be significantly shorter than these of wild-type animals ( strain N2 ) ( Figure 2A , Supplementary file 2 , mean lifespans ( LS ) of 14 . 81 ± 0 . 41 and 17 . 56 ± 0 . 52 days , respectively , p<0 . 001 ) . A parallel experiment , using CF512 worms and RNAi towards car-1 or daf-16 , showed similar lifespan shortening by car-1 RNAi ( Figure 2—figure supplement 1A , and Supplementary file 3 , mean LS of 14 . 87 ± 0 . 48 ( car-1 RNAi ) and 18 . 01 ± 0 . 63 ( EV ) days , p<0 . 001 ) . Nevertheless , the car-1 RNAi-mediated lifespan shortening effect was less prominent than that of daf-16 RNAi ( Figure 2—figure supplement 1A , mean LS of 12 . 19 ± 0 . 44 days ) . Next , we asked whether CAR-1 is needed for the full longevity phenotype of nematodes that carry a weak daf-2 allele ( e1370 , strain CB1370 ) . The worms were either grown from hatching on daf-16 , car-1 RNAi , or left untreated ( EV ) and their lifespans were recorded . The knockdown of car-1 significantly shortened the lifespan of daf-2 mutant worms , compared to untreated animals ( Figure 2B , Supplementary file 2 , mean LS of 40 . 02 ± 1 . 40 ( car-1 RNAi ) and 50 . 52 ± 1 . 38 ( EV ) days , respectively , p<0 . 001 ) . This effect was less prominent than that of daf-16 knockdown ( mean LS of 17 . 56 ± 0 . 56 days ) . A similar trend was seen when an additional daf-2 mutant worm strain ( e1368 ) was used ( Figure 2—figure supplement 1B and Supplementary file 2 ) . Yet , the lifespan reduction that we observed among untreated and car-1 RNAi-treated daf-2 ( e1370 ) mutant worms , which was similar to the difference observed among wild-type and car-1 knockout animals ( Figure 2A ) , questioned the notion that CAR-1 is involved in IIS-mediated regulation of lifespan . Thus , these results suggest that the SUMOylation of CAR-1 by the IIS may be involved in other functions of this signaling pathway . To further characterize the roles of car-1 as a regulator of lifespan , we tested whether the knockdown of car-1 affects the lifespan of daf-16 mutant animals ( strain CF1038 ) , and found that car-1 RNAi had no effect on the lifespans of these worms ( Figure 2C and Supplementary file 2 ) . This observation implies that the lifespan regulatory functions of car-1 are DAF-16 dependent . However , since DAF-16 is involved in several longevity-controlling mechanisms ( Hsin and Kenyon , 1999 ) , we further tested whether CAR-1 is mechanistically linked to IIS . Since IIS reduction lowers CAR-1 SUMOylation levels ( Figure 1C ) , we sought to test whether the SUMOylation state of CAR-1 affects lifespan . To directly address this , we used the computational tool GPS-SUMO ( Zhao et al . , 2014 ) , for identifying SUMOylation consensus motifs in the sequence of CAR-1 . Two lysine residues , K185 and K257 were found to be located within predicted SUMOylation motifs . Among the two , K185 is more likely to serve as a SUMOylation site ( Figure 2—figure supplement 2 ) . If SUMOylation of K185 or K257 reduces CAR-1 activity and shortens lifespan , it was expected that the overexpression of a SUMOylation-resistant CAR-1 mutant would extend lifespan . To test this hypothesis , we created worms that over-express mutated car-1 genes in which either K185 or K257 were substituted with arginine ( CAR-1 K185R ( EHC118 ) and CAR-1 K257R ( EHC121 ) , both strains also express the endogenous , wild-type car-1 ) . These mutations prevent potential SUMOylation but maintain the hydrophobicity of the protein . To control for the effect of car-1 over-expression on lifespan , we also created worms that over-express the wild-type car-1 gene ( strain EHC117 ) . The three exogenous car-1 genes were fused to an N-terminal double HA tag , and their expression levels were controlled by the car-1 promoter . To compare the levels of SUMOylated CAR-1 in these worm strains , we homogenized young adult worms of the three strains and subjected equal amounts of protein to IP . Using an HA antibody , we pulled down CAR-1 , separated total proteins of each worm strain and blotted SUMOylated CAR-1 by a SUMO antibody . Our results ( Figure 2—figure supplement 3 ) show that worms that express the WT CAR-1 contain much higher levels of SUMOylated CAR-1 compared to their counterparts that overexpress either CAR-1 K185R or CAR-1 K257R , indicating that these are SUMOylation sites . Our results also show that CAR-1 is SUMOylated on more than one site , as the substitution of either K185 or of K257 with arginine , did not abolish SUMOylation of the protein . Performing lifespan assays we found that the over-expression of CAR-1 K185R significantly extended the worms’ lifespans compared to those of wild-type animals ( Figure 2D , Supplementary file 2 , mean LS of 22 . 51 ± 0 . 60 and 14 . 70 ± 0 . 59 , p<0 . 001 ) . Three independent repeats confirmed the significance of this phenotype ( Figure 2E , Figure 2—figure supplement 4 , A and B , and Supplementary file 2 ) . In contrast , no lifespan extension was observed in worms that overexpress CAR-1 K257R ( Figure 2F and Supplementary file 3 ) or wild-type CAR-1 ( Figure 2—figure supplement 4C and Supplementary file 3 ) , indicating that the SUMOylation of CAR-1 on lysine 185 , but not on lysine 257 , plays a role in lifespan determination . To examine whether the lifespan-extending mechanism that is activated by CAR-1 K185R is DAF-16-dependent , we utilized CAR-1 K185R worms ( of a second clone ) . The worms were either treated with daf-16 RNAi or left untreated ( EV ) , and their lifespans were monitored . Surprisingly , our results ( Figure 2G ) show that daf-16 RNAi-treated CAR-1 K185R worms and wild-type ( N2 ) animals , exhibited indistinguishable lifespans ( see also Supplementary file 2 , 3 ) . In contrast , daf-16 RNAi-treated CAR-1 K257R worms ( Figure 2H ) and animals that over-express the wild-type CAR-1 and fed with daf-16 RNAi bacteria ( Figure 2—figure supplement 4C and Supplementary file 3 ) had shorter lifespans compared to their wild-type ( N2 ) counterparts . These results suggest that CAR-1 also regulates lifespan by a DAF-16-independent mechanism . Taken together , our observations indicate that CAR-1 is needed for wild-type worms to live their natural lifespan and for daf-2 mutant animals to exhibit their full longevity phenotype . They also indicate that SUMOylation of K185 plays a role in the regulation of lifespan . Interestingly , DAF-16 is needed for CAR-1 K185R to extend lifespan; however , the knockdown of daf-16 reduces the lifespans of CAR-1 K185R-expressing animals to be similar to these of wild-type worms , but not shorter as expected . One possible explanation to the lifespan shortening effect of car-1 RNAi , and the longevity conferred by CAR-1 K185R , suggests that CAR-1 modulates lifespan by negatively regulating the activity of GLP-1 . Accordingly , knocking down car-1 by RNAi is expected to hyperactivate GLP-1 and shorten lifespan , whereas the expression of the SUMOylation-resistant , hyperactive CAR-1 K185R , is expected to lower the activity of GLP-1 , thereby extending lifespan . To scrutinize this hypothesis , we utilized CF1903 worms . If the lifespan shortening effect of car-1 RNAi is mediated by hyper-activating GLP-1 , it was expected that the knockdown of car-1 would not shorten the long lifespans of these worms , which lack functional GLP-1 . CF1903 worms were either grown throughout life on control bacteria ( EV ) , or treated with RNAi towards daf-16 or car-1 , and lifespans were recorded ( in this experiment the worms were developed at 25°C and transferred to 20°C at day 1 of adulthood ) . While daf-16 RNAi-treated animals had shorter lifespans ( Figure 3A , Supplementary file 2 , mean LS of 10 . 71 ± 0 . 33 days , p<0 . 001 ) , untreated and car-1 RNAi-treated worms had very similar lifespans ( mean LS of 18 . 13 ± 0 . 85 ( EV ) and 17 . 76 ± 0 . 72 ( car-1 RNAi ) days , respectively , p=0 . 37 ) . These results show that car-1 RNAi does not affect lifespan in the absence of functional GLP-1 , and support the theme that CAR-1 modulates lifespan by controlling the activity of GLP-1 . To further assess the hypothesis that CAR-1 affects lifespan through the GLP-1 axis , we utilized worms that carry a mutant kri-1 gene ( strain CF2052 ( ok1251 ) ) . kri-1 is essential for the mediation of longevity by germ cell ablation but not by IIS reduction ( Berman and Kenyon , 2006 ) . Thus , if car-1 affects lifespan through the modulation of GLP-1 activity , car-1 RNAi is expected not to affect the lifespan of kri-1 mutant worms . CF2052 nematodes were treated with daf-16 or car-1 RNAi and lifespans were recorded . Although daf-16 RNAi-treated CF2052 animals exhibited a short mean lifespan ( 13 . 37 ± 0 . 26 days , p<0 . 001 ) , untreated and car-1 RNAi-treated worms had similar mean lifespans of 16 . 59 ± 0 . 43 and 15 . 96 ± 0 . 39 days , respectively ( Figure 3B , p=0 . 14 , and Supplementary file 2 ) . Together these observations support the theme that car-1 governs lifespan , at least partially , through a glp-1-controlled mechanism . The possible involvement of car-1 in the regulation of lifespan through an additional , daf-16-independent mechanism has led us to ask whether the knockdown of car-1 is linked to the aging-regulating pathway downstream of the transforming growth factor β ( TGF-β ) . This pathway converges with the IIS at the nuclear hormone receptor DAF-12 , whose activation is regulated by the cytochrome P450 enzyme , DAF-9 ( Gerisch et al . , 2001 ) . To test this , we followed the lifespans of daf-9 ( strain CF2531 ) and of daf-12 ( strain AA86 ) mutant worms that were either treated with daf-16 , car-1 RNAi or left untreated , and found that the lifespans of both worm strains were shortened by daf-16 RNAi as well as by the knockdown of car-1 ( Figure 3—figure supplement 1 , A and B , Supplementary file 4 ) . The similar rates of lifespan shortening that resulted from the knockdown of car-1 in wild-type worms ( Figure 2A ) , daf-9 and daf-12 mutant worms , strongly suggest that car-1 RNAi shortens lifespan by a daf-9 and daf-12-independent mechanism . Beside its roles in lifespan determination , GLP-1 is also involved in germ cells proliferation and reproduction ( Austin and Kimble , 1987 ) . Thus , we asked whether the modulation of car-1 expression and activity modifies the amount of germ cells . To test this , we compared the numbers of germ cells in gonads of four groups of worms: ( i ) untreated , wild-type worms ( strain N2 ) , ( ii ) car-1 RNAi-treated wild-type animals , ( iii ) worms that over-express the natural car-1 ( EHC117 ) , and ( iv ) nematodes that over-express the CAR-1 K185R mutant ( EHC118 ) . Nuclei were stained with DAPI and germ cells were counted . We found that the knockdown of car-1 by RNAi had no significant effect on the number of germ cells as untreated and car-1 RNAi-treated animals had similar numbers of germ cells ( Figure 3 , C and D; average of 724 ± 45 . 25 and 719 ± 32 . 36 cells , respectively ) . In contrast , worms that over-express the natural CAR-1 ( EHC117 ) had significantly less germ cells compared to untreated or car-1 RNAi-treated worms ( 414 . 2 ± 6 . 3 cells , p<0 . 001 ) . The over-expression of CAR-1 K185R resulted in further reduction in the number of germ cells ( average of 324 . 8 ± 21 . 37 germ cells/gonad , p<0 . 01 compared to EHC117 ) . The observation that the knockdown of car-1 shows no effect on the number of germ cells may emanate from an efficient SUMOylation-mediated inactivation of CAR-1 in wild-type worms . Thus , CAR-1 is less active and its knockdown has a small effect on the number of germ cells . In contrast , the over-expression of CAR-1 ( in EHC117 worms ) , may exceeds the capacity of the SUMOylation mechanism and thus , hyper-activates CAR-1 , which in turn lowers the activity of GLP-1 , thereby reducing the number of germ cells . According to this explanation , the over-expression of the hyper-active CAR-1 K185R ( EHC118 ) further suppresses the activity of GLP-1 , resulting in an even lower number of germ cells . The knockdown of car-1 by RNAi was reported to enhance physiological apoptosis in hermaphrodites ( Boag et al . , 2005 ) . Thus , we examined how knocking down car-1 , or over expressing wild-type or the mutant CAR-1 K185R , affect the rate of apoptosis in the gonads . Our results ( Figure 3E ) indicate that , as shown previously ( Boag et al . , 2005 ) , the knockdown of car-1 elevates the number of apoptotic nuclei in the gonad by approximately 2 . 5 fold . A similar increase in the rate of apoptosis was seen in EHC118 worms , but not in EHC117 animals . These observations raise the question of how the effects of CAR-1 and its SUMOylation on the number of germ cells and of apoptotic nuclei , affect reproduction . To address this we tested how the expression of wild-type or of CAR-1 K185R affects brood size by comparing the egg laying capabilities of N2 , EHC117 and EHC118 worms . The animals were grown from hatching on control bacteria . At L4 larval stage , 12 animals of each strain were transferred onto new plates , one animal per plate . The worms were transferred onto new plates in 12 hr intervals and viable progeny were counted 48 hr thereafter . We found that EHC118 animals lay fewer eggs than control worms ( Figure 3 , F and G ) . While in total , N2 worms had an average of 271 . 1 progeny , each CAR-1 K185R worm had an average of 104 . 8 living offspring ( Figure 3G ) . A small but not significant difference was observed among EHC117 and N2 worms , as on average EHC117 worms had 171 . 4 viable offspring and N2 worms had 201 . 9 ( Figure 3H ) . Our observations indicate that SUMOylation of CAR-1 on residue 185 is involved in the modulation of reproduction and suggest that this phenotype may be associated with the effects of IIS reduction on egg laying ( Dillin et al . , 2002 ) . Thus , we examined how the knockdown of car-1 affects the egg-laying pattern of worms that exhibit impaired IIS . To test this , we employed daf-2 ( e1370 ) mutant and daf-16 mutant ( mu86 , strain CF1038 ) worms . e1370 worms were treated with either car-1 RNAi , daf-16 RNAi , or were left untreated ( EV ) , while CF1038 animals were treated with car-1 RNAi or fed on control bacteria ( EV ) . Egg-laying patterns were followed as described above . Our results ( Figure 3—figure supplement 2 , A and B ) confirmed that untreated e1370 worms had a much longer reproductive period compared to wild-type animals . However , the total number of progeny per untreated daf-2 mutant worm was on average 95 . 7 , much lower than that of wild-type nematodes ( Figure 3G ) . Both phenotypes were largely rescued by the knockdown of daf-16 , which shifted egg laying to early adulthood ( Figure 3—figure supplement 2A ) and restored the total number of eggs to the average of 255 . 4 ( Figure 3—figure supplement 2B ) . Surprisingly , RNAi-mediated knockdown of car-1 resulted in nearly complete sterility of both daf-2 and daf-16 mutant worms ( Figure 3—figure supplement 2 , A and B ) . These observations show that worms that have impaired IIS activity are much more sensitive to the knockdown of car-1 than wild-type animals , implying that this gene is involved in the control of reproduction by the IIS . Yet , this reduction in brood size may be partially due to additive effects of IIS reduction and knocking down car-1 . We also tested whether knocking down car-1 influences the egg-laying patterns of kri-1 mutant ( CF2052 ) nematodes . As shown previously ( Dillin et al . , 2002 ) , untreated wild-type worms laid the highest number of eggs between day 1 and 1 . 5 of adulthood ( Figure 3F ) . In contrast , N2 worms that were treated with car-1 RNAi laid the highest number of eggs at the beginning of their reproductive stage ( day 0 to 0 . 5 ) , and the number of eggs declined thereafter ( Figure 3—figure supplement 2 , C and D ) . While untreated N2 worms laid on average 244 eggs in total , their car-1 RNAi-treated counterparts laid only 133 eggs ( a reduction of ~ 45% ( Figure 3—figure supplement 2D ) ) . The reduced reproduction was surprising , as the knockdown of car-1 by RNAi had no effect on the number of germ cells ( Figure 3 , C and D ) . Nevertheless , this reduction , which is consistent with a previous report ( Boag et al . , 2005 ) , may be explained by the increase in the number of apoptotic cells in the gonads of these animals ( Figure 3E ) . kri-1 mutant worms ( CF2052 ) laid fewer eggs than N2 animals , but car-1 RNAi treatment further reduced the number of progeny of both strains . While untreated kri-1 mutant worms laid an average of 116 eggs , car-1 RNAi-treated worms of the same strain laid merely 10 eggs ( Figure 3—figure supplement 2 , C and D ) . Unexpectedly , the knockdown of car-1 affects neither egg-laying patterns , nor brood size of daf-9 and daf-12 mutant worms ( Figure 3—figure supplement 2 , E and F ) . These observations show that the egg-laying modulation that resulted from the knockdown of car-1 by RNAi is dependent on the presence of functional daf-9 and daf-12 , linking CAR-1 also with the DAF-9/DAF-12 pathway . Altogether , despite the similarity in the number of germ cells in car-1 RNAi-treated and untreated N2 worms , the reduction in brood size of wild-type worms by car-1 RNAi , strongly suggests that CAR-1 regulates reproduction by GLP-1-dependent and GLP-1-independent mechanisms . In addition , our observation that the knockdown of car-1 leads to nearly complete sterility of worms carrying weak kri-1 or daf-2 alleles or nonfunctional daf-16 , suggests that CAR-1 is also a component of the reproduction-regulating mechanism downstream of the IIS . To directly test whether car-1 affects the transcriptional activity of the GLP-1 pathway , we asked how the knockdown of car-1 affects the expression levels of the glp-1-target genes sygl-1 and lst-1 ( Shin et al . , 2017 ) . First , we used N2 and CF1903 to confirm the regulatory roles of GLP-1 on the transcription of these genes . Worms of both strains were either grown at 15 or 25°C ( to inactivate GLP-1 in CF1903 animals ) and the expression levels of sygl-1 and lst-1 were determined by qPCR . While an increase in the expression levels of both genes was observed in N2 worms upon exposure to 25°C ( a significant increase for sygl-1 and a non-significant trend for lst-1 ) , the inactivation of GLP-1 in CF1903 worms , by exposing them to 25°C during development , resulted in a significant reduction in the expression of both , sygl-1 and lst-1 ( Figure 3—figure supplement 3 , A and B ) . These results indicate that GLP-1 positively regulates the expression of these two genes . We next utilized daf-2 ( e1370 ) mutant animals to test how the knockdown of car-1 affects the expression of sygl-1 and lst-1 in these nematodes . The worms were grown from hatching on EV or on car-1 RNAi bacteria , harvested at day 1 of adulthood and gene expression levels were compared by qPCR . If CAR-1 is a negative regulator of GLP-1 that is negatively controlled by IIS-mediated SUMOylation , it is expected that CAR-1 is hyperactive in daf-2 mutant worms . Accordingly , the knockdown of car-1 by RNAi is predicted to activate GLP-1 and elevate the expression of sygl-1 and lst-1 . Indeed , we observed significantly elevated levels of sygl-1 in e1370 animals ( Figure 3I ) . A non-significant elevation in the expression of sygl-1 was also seen in car-1 RNAi-treated N2 worms ( Figure 3—figure supplement 4 ) . This lack of significance may be explained by the low activity of CAR-1 in these worms , due to its SUMOylation . These results indicate that CAR-1 negatively controls the activity of GLP-1 as a transcriptional regulator of sygl-1 . Unexpectedly , the knockdown of car-1 reduced the expression level of lst-1 in e1370 worms ( Figure 3I ) showing that CAR-1 could play opposing roles on the expression levels of GLP-1 target genes . This observation is consistent with the finding that in some cases , transcriptional co-factors affect the expression of some target genes but not of others ( Volovik et al . , 2014b ) , and show the complex regulatory relations between car-1 and glp-1 . Since the IIS controls the expression levels of some of its components ( Alic et al . , 2011 ) , we asked whether GLP-1 controls the expression of car-1 . Using qPCR , CF1903 , and wild-type worms , we found that CF1903 animals that were developed at 25°C and thus , lack functional GLP-1 , have reduced car-1 levels compared to their counterparts that were grown at 15°C . No significant difference in the expression of car-1 was observed in wild-type worms ( Figure 3J ) . This reduction of approximately 65% in the levels of car-1 , shows that GLP-1 positively regulates the expression of car-1 , and raises the question of whether CAR-1 also plays roles in another feature of the GLP-1-controlled mechanism , the maintenance of proteostasis . The known regulatory roles of glp-1 on proteostasis ( Shemesh et al . , 2013 ) has led us to examine whether CAR-1 also controls proteotoxicity . To address this , we utilized worms that express the Alzheimer's disease associated , human Aβ3-42 peptide ( McColl et al . , 2009 ) , in their body wall muscles ( strain CL2006 , Aβ worms ) ( Link , 1995 ) . The expression of Aβ causes progressive paralysis within the worm population , a phenotype that can be alleviated by the knockdown of daf-2 ( Cohen et al . , 2006 ) . Eggs of Aβ worms were placed on plates seeded with daf-2 or car-1 RNAi bacteria , or left untreated ( EV ) . Rates of paralysis were followed up until day 12 of adulthood . While the knockdown of daf-2 protected the worms from Aβ-mediated toxicity , animals that were treated with car-1 RNAi exhibited higher rate of paralysis than untreated worms ( Figure 4A ) . Five independent repeats confirmed the significance of this phenotype ( Figure 4B ) . We next examined whether CAR-1 is needed for daf-2 RNAi-conferred protection from proteotoxicity , by performing paralysis assays using Aβ worms that were grown on different mixtures of RNAi bacterial strains . First , we checked if the dilution of bacteria expressing either car-1 or daf-2 RNAi with control bacteria ( EV ) reduces the effects of these treatments on paralysis . Worms that were solely treated with car-1 RNAi and their counterparts that were fed with a mixture of car-1 RNAi and EV bacteria exhibited very similar rates of paralysis over time . Likewise , similar protection from paralysis was seen in worms that were exclusively fed with daf-2 RNAi bacteria and those which were fed with a mixture of daf-2 RNAi and EV bacteria ( Figure 4 , C and D ) . These results show that the dilution of car-1 and of daf-2 RNAi bacteria with another bacterial strain does not significantly changes the effects of these treatments on proteostasis . We next examined whether the mixture of daf-2 and car-1 RNAi treatments prevents IIS reduction from promoting its full counter-proteotoxic effect , and found that concurrent knockdown of these genes resulted in an enhanced rate of paralysis compared to the rate observed in animals that were treated solely with daf-2 RNAi ( Figure 4 , C and D , blue , p<0 . 04 ) . These findings imply that CAR-1 is needed for proteostasis-maintenance , downstream of the IIS however , the reduced paralysis rates that we observed in worms that were concomitantly treated with daf-2 and car-1 RNAi imply that the knockdown of daf-2 protects from proteotoxicity by additional , CAR-1-independent mechanisms . Yet , it is also possible that knocking down car-1 , may inflict damage by an IIS-independent mechanism . According to this notion , the observed cumulative paralytic effect results from both , daf-2 RNAi-mediated protection and car-1 RNAi-promoted damage . To directly address this hypothesis , we expressed CAR-1 K185R in Aβ worms ( strain EHC124 ) . If SUMOylation on K185 reduces the activity of CAR-1 , it was expected that the expression of the SUMOylation-resistant CAR-1 K185R mutant would protect the animals from Aβ toxicity . Indeed , we found that Aβ worms expressing CAR-1 K185R are largely ( Figure 4E ) and significantly ( Figure 4F , from day 7 p<0 . 004 ) protected from proteotoxicity . The protective effect of CAR-1 K185R is DAF-16-dependent , since no protection was seen when EHC124 worms were treated with daf-16 RNAi ( Figure 4—figure supplement 1A ) . These daf-16 RNAi-treated worm population exhibited similar rates of paralysis to these of CL2006 worms that were treated with the same RNAi ( Cohen et al . , 2006; Cohen et al . , 2010 ) . The observation that Aβ worms expressing CAR-1 K257R ( strain EHC125 ) and Aβ worms ( CL2006 ) exhibit indistinguishable rates of paralysis ( Figure 4—figure supplement 1B ) indicates that the counter-proteotoxic effect that is conferred by CAR-1 , is suppressed by the SUMOylation of lysine 185 . Finally , the more efficient protection from Aβ proteotoxicity that is conferred by daf-2 RNAi ( Figure 4C ) , supports the idea that the protective mechanisms that are activated by IIS reduction and by CAR-1 K185R only partially overlap . We further examined whether CAR-1 modulates proteotoxicity by employing worms that express fluorescently tagged , poly-glutamine stretches of 67 repeats in their neurons ( polyQ67-YFP , strain AM716 ) . Expanded glutamine stretches cause various human neurodegenerative maladies , including Huntington's disease ( Bates , 2003 ) , and lead to impaired neuronal activity in worms ( Vilchez et al . , 2012 ) . AM716 worms were grown on EV , daf-2 or car-1 RNAi bacteria and placed in a drop of liquid at days 1 and 4 of adulthood . To measure proteotoxicity , the number of body bends per 30 s were counted ( Volovik et al . , 2014a ) . As anticipated , knocking down daf-2 protected from proteotoxicity at both day 1 ( p<0 . 001 ) and day 4 ( p<0 . 001 ) of adulthood . In contrast , car-1 RNAi treatment decreased the number of body bends at day 4 of adulthood ( p<0 . 001 ) but not at day 1 ( Figure 5—figure supplement 1 ) . These results confirm the roles of CAR-1 as a modulator of age-onset proteotoxicity . Our results suggest that CAR-1 modulates proteostasis , at least partially , by negatively regulating glp-1 activity . To examine this possibility , we used worms that express polyQ35-YFP in their body wall muscles ( strain AM140 ) , and thus exhibit a motility defect ( Morley et al . , 2002 ) , and harbor a temperature-sensitive glp-1 mutant ( strain ABZ21 ) . These animals are protected from polyQ-mediated paralysis when grown at 25°C ( Shemesh et al . , 2013 ) . The rates of paralysis of polyQ35-YFP and of ABZ21 worms that were developed at 25°C and were either grown on EV , daf-2 , or car-1 RNAi bacteria , were compared . While car-1 RNAi-treated polyQ35-YFP worms ( that express a functional GLP-1 ) showed an increased rate of paralysis compared to untreated animals , daf-2 RNAi provided nearly complete protection from paralysis ( Figure 5 , A and B ) . In contrast , both untreated and car-1 RNAi-treated ABZ21 worms were protected from paralysis . This set of experiments shows that knocking down car-1 has no deleterious effect on proteostasis when glp-1 is inactive ( p<0 . 05 ) , indicating that CAR-1 controls proteostasis through GLP-1 . Since the CAR-1-associated helicase CGH-1 , which is expressed in meiotic germ cells , modulates lifespan ( Figure 5—figure supplement 2 , Supplementary file 5 and ( Seo et al . , 2015 ) ) , we tested whether this helicase also controls proteostasis . To address this , we let Aβ worms develop on EV bacteria or treated them with RNAi towards daf-2 , car-1 , or cgh-1 , followed their rates of paralysis and found that the knockdown of cgh-1 as well as of car-1 , significantly ( p<0 . 035 ) increase the rates of paralysis compared to untreated worms ( Figure 5 , C and D ) . Similar results were obtained when worms that harbor a metastable perlecan ( unc-52 ( ts ) ) , that misfolds and causes paralysis of worms that are grown at 25°C ( strain HE250 , [Shemesh et al . , 2013] ) , were treated with either car-1 or cgh-1 RNAi ( Figure 5—figure supplement 3 , A and B ) . These results further support the view that CAR-1 and CGH-1 promote proteostasis by modulating germ cell activity , plausibly by negatively regulating GLP-1 . The roles of CAR-1 in lifespan determination and proteostasis maintenance have led us to ask whether it also controls stress resistance . To examine whether car-1 influences heat stress resistance we used two worm strains , N2 and CF512 ( the exposure of CF512 worms to 25˚C during development does not activate the heat shock response [Volovik et al . , 2012] ) . The worms were treated from hatching with car-1 RNAi or left untreated , and exposed at day 1 of adulthood to 35°C for 11 hr . We observed no significant effect of car-1 RNAi on resistance to heat ( Figure 6 , A and B ) . Similarly , car-1 RNAi did not abolish the elevated heat resistance of daf-2 mutant worms ( Figure 6C ) . Surprisingly , the expression of CAR-1 K185R elevated the survival after heat shock compared to control worms ( N2 , p<0 . 04 ) , however a trend but not a significant effect was observed when the natural car-1 was over-expressed ( Figure 6D ) . We next used CF512 worms to analyze resistance to pathogenic bacteria . The worms were treated throughout development with daf-2 , daf-16 , or car-1 RNAi or left untreated ( EV ) . At day 1 of adulthood , the nematodes were transferred onto plates seeded with the pathogenic bacteria Pseudomonas aeruginosa . As expected ( Singh and Aballay , 2006 ) , the knockdown of daf-2 extended , whereas daf-16 RNAi shortened the mean survival of the worms compared to control animals ( mean survival of 11 . 96 ± 0 . 42 , 4 . 28 ± 0 . 11 and 5 . 88 ± 0 . 15 days , respectively , p<0 . 001 ) . Interestingly , car-1 RNAi had a small but significant protective effect from pathogenic bacteria ( mean survival of 6 . 77 ± 0 . 21 days , p<0 . 001 , Figure 6E , Supplementary file 6 ) . To further test this effect , we conducted the reciprocal experiment asking whether the over-expression of the wild-type CAR-1 ( strain EHC117 ) or the K185R CAR-1 ( strain EHC118 ) , shortens the survival rates of worms that were cultured on P . aeruginosa . Our results ( Figure 6F ) show a small but significant lifespan shortening effect that stemmed from the over-expression of the wild-type ( p<0 . 01 ) and a non-significant trend in worms that express the K185R CAR-1 . These results confirm that CAR-1 is deleterious when the worms are exposed to these pathogenic bacteria . We also assessed whether CAR-1 is involved in protection from UV radiation by following the survival of CF512 worms that were treated with RNAi as above , and exposed to a sub-lethal dose of UV . The knockdown of daf-2 protected the worms from UV and the survival rate of the worms treated with car-1 RNAi was also increased compared to control animals ( mean survival rates of 10 . 45 ± 0 . 22 ( daf-2 RNAi ) , 9 . 23 ± 0 . 22 ( car-1 RNAi ) and 7 . 97 ± 0 . 21 ( EV ) days , respectively , p<0 . 001 ) ( Figure 6G , Supplementary file 6 ) . Similarly , the over-expression of the wild-type CAR-1 or of the CAR-1 K185R mutant significantly shortened the lifespans of worms that were exposed to UV radiation ( mean survivals of 12 . 78 ± 0 . 33 ( EV ) , 8 . 89 ± 0 . 19 ( wt CAR-1 ) and 10 . 87 ± 0 . 27 ( CAR-1 K185R ) days , p<0 . 001 ) ( Figure 6H and Supplementary file 6 ) . Together , these results show that CAR-1 plays minor roles in resistance to heat as well as in survival after exposure to pathogenic bacteria and UV radiation . The data obtained in this work culminate to suggest the following model ( Figure 7 ) : besides regulating the cellular localization of its downstream transcription factors ( 7-I ) , the IIS also governs aging-associated functions by SUMOylating CAR-1 on lysine 185 . This post-translational modification inhibits CAR-1’s function ( 7-II ) , thereby activating GLP-1 ( 7-III ) . Accordingly , the expression of the hyper-active , SUMOylation-resistant CAR-1 K185R , efficiently represses GLP-1 thereby , mimicking one aspect of IIS reduction , and promotes proteostasis . Interestingly , GLP-1 regulates the expression of car-1 to create a regulatory circuit ( 7-IV ) . Importantly , CAR-1 may also affects lifespan , proteostasis , stress resistance , and reproduction by a DAF-16-independent mechanism ( 7 V ) .
Post-translational modifications are known to have various biological functions , including the regulation of aging . Phosphorylation regulates the activities of DAF-16 , SKN-1 , and HSF-1 downstream of the IIS ( Lee et al . , 2001; Tullet et al . , 2008; Chiang et al . , 2012 ) and SUMOylation controls the localization of the IGF-1 receptor and its signaling activity in mammalian tissues ( Sehat et al . , 2010 ) . SUMOylation is also critical for aging-associated modulation of mevalonate biosynthesis ( Sapir et al . , 2014 ) , a metabolite that has been implicated in the development of clinical conditions ( Mokarram et al . , 2017 ) . In this study , we unveiled a novel role of SUMOylation in the regulation of aging , serving as a functional switch of CAR-1 , which is governed by the IIS . This raises the question of how the IIS controls the SUMOylation state of CAR-1 . One possible explanation stems from the correlation between AKT function , the stability of SUMO , and the SUMO-conjugating enzyme UBC-9 in mammals ( Lin et al . , 2016 ) . According to this theme , reduced IIS lowers AKT activity , resulting in SMO-1 destabilization and in reduction of UBC-9 activity . This cascade of events may lower the rate of global protein SUMOylation . This possibility appears less likely as our results ( Figure 1A ) indicate that while daf-2 RNAi leads to lower SUMOylation of some proteins others proteins exhibit increased SUMOylation upon IIS reduction . Alternatively , the expression of specific genes that encode for proteins involved in CAR-1 SUMOylation may be positively regulated by the IIS , which lowers the expression of these genes and reduces the rate of CAR-1 SUMOylation . Future research is needed to clarify this issue . Although SUMOylation appears to be a pivotal post-translational modification that influences aging , a simultaneous knockdown of car-1 and daf-2 only partially protects worms from proteotoxicity and only partially shortens lifespans of daf-2 mutant worms . On one hand , these observations suggest that IIS reduction also protects from proteotoxicity and extends lifespan by additional , CAR-1-independent mechanisms . However , on the other hand , the knockdown of daf-16 in CAR-1 K185R-expressing worms did not shorten lifespan below those of wild-type nematodes ( Figure 2G ) , suggesting that CAR-1 also governs lifespan by an additional , DAF-16-independent mechanism ( the lack of functional daf-16 shorten the lifespan of N2 worms by approximately 30% [Kenyon et al . , 1993] ) . Nevertheless , the knockdown of car-1 has not further shortened the lifespan of daf-16 mutant animals ( Figure 2C ) . The lack of additive effect may be a result of the very short lifespan of daf-16 mutant worms , which does not allow the knockdown of car-1 to further shorten lifespan or perhaps by the limited efficiency of RNAi-mediated knockdown of car-1 . An additional interesting aspect of this study is the differential effects of car-1 on distinct environmental insults . While the knockdown of car-1 has no effect on heat stress resistance ( Figure 6 , A-C ) , the expression of CAR-1 K185R , mildly but significantly elevates the survival of heat-stressed worms ( Figure 6D ) . In contrast , the knockdown of car-1 has a small but reproducible protective effect on resistance to UV radiation ( Figure 6G ) and to pathogenic bacteria ( Figure 6E ) . In agreement , the over expression of either the wild-type CAR-1 or the mutated CAR-1 K185R is deleterious to worms that were exposed to these insults ( Figure 6 , F and H ) . The observed protection from UV radiation , conferred by car-1 RNAi , is consistent with a previous report that CAR-1 and CGH-1 negatively regulate DNA-damage-mediated apoptosis ( Tomazella et al . , 2012 ) . Nevertheless , despite its protective effect when the worm is exposed to stress conditions , the knockdown of car-1 shortens lifespan ( Figure 2B ) . These results support the theme that the ability to resist stresses such as heat ( Maman et al . , 2013; Volovik et al . , 2014b ) and oxidation ( Van Raamsdonk and Hekimi , 2012 ) are not necessarily coupled with lifespan determination . They also coincide with the reports that IIS-regulated factors may be involved in the regulation of certain stress resistance mechanisms but not of others . For instance , the transcription factor SMK-1 is needed for the worm to resist UV radiation and pathogenic bacteria but is dispensable for coping with heat ( Wolff et al . , 2006 ) . Similarly , we recently reported that the knockdown of caveolin-1 extends lifespan and provides partial protection from pathogenic bacteria , but has no role in heat stress resistance ( Roitenberg et al . , 2018 ) . The car-1 RNAi-mediated protection from certain stresses appears to be contradictive to the observation that knocking down this gene exposes the animal to proteotoxicity . However , it has been already shown that abolishing the nematode’s ability to resist heat by knocking down neuronal components , provides the worm with partial protection from proteotoxicity ( Prahlad and Morimoto , 2011; Volovik et al . , 2014b ) . Our findings show that manipulating the activity of a germline protein can also confer opposing effects on stress resistance and proteostasis , and raise the question of how CAR-1 promotes these opposing effects . One possible explanation suggests that CAR-1 may differentially affect the expression levels of different GLP-1-controlled genes . Such differential effects of transcriptional co-regulators have been reported . For instance , the knockdown of the DAF-16 transcriptional co-factor nhl-1 , lowers the expression level of sod-3 and of sip-1 , but has no effect on the expression of mtl-1 , which are all known target genes of DAF-16 ( Volovik et al . , 2014b ) . The opposing effects of car-1 RNAi on the expression levels of sygl-1 and lst-1 propose a similar mechanism of differential effects on the expression of distinct genes . How this mechanism functions and what cellular components are involved in the mediation of proteostasis by the SUMOylation-resistant CAR-1 , are questions that require further elucidation .
N2 ( wild-type , Bristol ) , CB1370 ( daf-2 ( e1370 ) mutant worms ) , CL2006 ( unc-54p::human Aβ3-42 ) , CF512 ( fer-15 ( b26 ) II; fem-1 ( hc17 ) IV ) , CF1903 ( glp-1 ( e2141 ) III . ) , AM140 ( Punc54::Q35::YFP ) , HE250 ( unc-52 ( e669su250 ) II . ) , WH377 ( car-1 ( tm1753 ) I/hT2 ) , WH346 ( unc-119 ( ed3 ) III . ojIs34 [GFP::car-1+unc-119 ( + ) ] , CF2052 kri-1 ( ok1251 ) ( I ) , AA86 daf-12 ( rh61rh411 ) X . , CF2531 daf-9 ( rh50 ) X . , CF1038 ( daf-16 ( mu86 ) I ) were obtained from the Caenorhabditis Genetic Center ( CGC , Minneapolis , MN ) . Worms that over-express the K185R or K257R mutated car-1 ( strains EHC118 and EHC121 respectively ) or WT CAR-1 tagged to 2xHA tag ( strain EHC117 ) were generated by injecting a plasmid that carries the gene downstream of the natural car-1 promoter region ( 1074 bp upstream of the ORF ) into N2 worms . rol-6 driven by the unc-54 promoter or gfp driven by the elt-2 promoter were used as selection markers . Worms expressing Aβ in their body wall muscle and CAR-1 K185R or CAR-1 K257R were generated by injecting the same plasmids into CL2006 worms ( strains EHC124 and EHC125 , respectively ) . AM716 ( rmIs284[pF25B3 . 3::Q67::YFP] ) worms were obtained from Dr . Richard I Morimoto ( Northwestern , IL ) . ABZ21 animals ( Punc54::Q35::YFPxglp-1 ( CF1903 ) ) were a gift of Dr . Anat Ben-Zvi ( Ben-Gurion , Israel ) . NX25 ( smo-1 ( ok359 ) ;tvEx25[psmo-1::His-FLAG-SMO-1; rol-6] ) were obtained from Dr . Limor Broday ( TAU , Israel ) . CF512 ( fer-15 ( b26 ) II; fem-1 ( hc17 ) IV ) , CF1903 ( glp-1 ( e2144 ) III . ) nematodes are heat-sensitive sterile and were thus , grown at 15°C . To avoid egg lying , these worms were developed at 25°C and transferred at day 1°C to 20°C until harvested . To achieve sterility , ABZ21 worms were grown at 25°C until harvesting . Other strains were synchronized and grown on the indicated RNAi bacteria at 20°C until day 1 of adulthood . To reduce gene expression , we used bacterial strains expressing dsRNA: empty vector ( pAD12 ) , daf-2 ( pAD48 ) , daf-16 ( pAD43 ) . car-1 and cgh-1 dsRNA-expressing bacteria were obtained from the Vidal RNAi library . RNAi bacteria were grown at 37°C in LB with 100 μg/ml ampicillin and then seeded on NG-ampicillin plates with the addition of 100 mM Isopropyl β-D-1-thiogalactopyranoside ( IPTG ~ 1 mM final concentration ) . To isolate SUMOylated proteins , 300 , 000 NX25 worms were grown on EV or daf-2 RNAi bacteria until day 1 of adulthood , collected and froze in liquid nitrogen ( Figure 1 , B , C and D ) . The worms were then homogenized and equilibration buffer ( 1XPBS , 8M UREA , 5 mM NEM and protease inhibitors ) was added prior to centrifugation ( 10 min , 9 , 391 g ) . Protein concentrations were measured and equalized by Bradford reagent . First , a His-tag purification using HisPur Ni-NTA Resin ( Thermo Scientific , #88221 ) was performed . Lysates were incubated with the resin for 50 min at RT and washed ( 1XPBS ( pH7 . 4 ) , 8M UREA , 250 mM Imidazole ) . SUMOylated proteins were eluted using buffer 1 ( 1XPBS , 2M UREA , 250 mM Imidazole ) followed by elution buffer 2 ( 1XPBS , 1M UREA , 250 mM Imidazole ) . Aliquots of the samples were blotted by WB for validation . The remaining eluted samples were used for Flag-tag purification with the Red ANTI-FLAG M2 Affinity Gel ( Sigma-Aldrich , #F2426 ) . Samples were diluted with RIPA buffer ( 50 mM Tris HCl pH7 . 5 , 150 mM NaCl , 5 mM EDTA , 1% TritionX-100 , 0 . 1% SDS , 1X Protease Inhibitor ( Calbiochem set III #539134 , 1 mM BetaME ) ) and incubated over night at 4°C with the Anti-Flag beads . The beads were washed with RIPA buffer followed by elution ( 100 mM Glycine pH3 . 5 , 150 mM NaCl ) . To detect the SUMOylation state of CAR-1 ( Figure 1 , C and D ) , WH346 worms were treated as described above . The homogenized worms were dissolved in RIPA buffer and the GFP immunoprecipitation was performed using GFP-Trap_A ( #gta-100 , Chromotek , Germany ) according to the manufacturer’s instructions . The beads were incubated with the lysates over night at 4°C and the trapped proteins were eluted and analyzed by WB . To test the SUMOylation state of the CAR-1 ( WT ) , CAR-1 K185R and CAR-1 K257R mutants ( strains EHC117 , EHC118 and EHC121 , respectively ) 120 , 000 worms were harvested as described above . CAR-1 was purified by performing an immunoprecipitation using the anti-HA . 11 Epitope Tag antibody and the Pierce Crosslink Immunoprecipitation Kit . The crosslinked beads and the worm lysates were incubated over night at 4°C , bound proteins were eluted and analyzed by WB . Total RNA was isolated from synchronized worm populations using QIAzol reagent ( QIAGEN , Hilden Germany #79306 ) and NucleoSpin RNA kit ( MACHEREY-NAGEL , #740955 . 50 ) . cDNA was synthesized using iScript cDNA Synthesis Kit ( Biorad , #170–8891 ) . Quantitative real-time PCR reactions were performed with EvaGreen supermix ( Biorad , #172–5204 ) . Quantities were normalized to levels of act-1 and of cdc-42 cDNA . To blot SUMOylated proteins ( Figure 1A ) , N2 , CF512 , and CF1903 worms that were grown at 15°C , were bleached to obtain synchronized eggs . The eggs were placed on plates that were seeded with control bacteria ( EV ) or daf-2 RNAi bacteria and incubated for 48 hr at 25°C ( to sterilize the CF512 and CF1903 worms ) . The worms were transferred thereafter to 20°C for additional 24 hr . For the experiment displayed at Figure 1E , the worms were hatched and grown at 20°C ( strain WH346 ) . At day 1 of adulthood , the worms were washed twice with M9 , and homogenized using a bullet grinder ( full speed , 10 s , three times ) . The worm homogenates were spun for 3 min at 850 g ( 3000 rpm in a benchtop Qiagen centrifuge ) to sediment debris . The post debris supernatants were collected , protein amounts were measured by a BCA kit ( Thermo Fisher #23225 ) , supplemented with loading buffer ( 10% glycerol , 125 mM Tris base , 1% SDS ) and heated at 95°C for 10 min . For each treatment , equal protein quantities were loaded and separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS-PAGE ) , transferred onto a PVDF membrane ( Millipore , Billerica MA ) and probed with the indicated antibody: GFP antibody ( Cell Signaling , Danvers , MA cat #2956 ) , anti-SUMO-1 antibody ( Millipore , #09–409 ) , anti-HA . 11 antibody ( BioLegend , San Diego , CA , #901501 ) or anti-actin antibody ( Simga , #A5441 ) . HRP-conjugated secondary antibody and a luminescent image analyzer ( ChemiDoc XRS + BioRad ) were used to detect protein signals . Synchronized worm eggs were placed on master NG-Ampicillin plates seeded with the indicated RNAi bacterial strain and supplemented with 100 mM IPTG . The eggs were incubated at 20°C until transferred onto small NG- Ampicillin plates , 12 animals per plate ( CF1903 and CF512 were incubated throughout development at 25°C , to induce sterility ) . Adult worms were transferred onto freshly seeded plates every 3 days . Worms that failed to move their noses when tapped twice with a platinum wire were scored as dead . Dead worms were scored daily . Lifespan analyses were conducted at 20°C . For all stress assays synchronized eggs were placed on NG plates seeded with the RNAi bacteria ( as indicated ) . For heat-stress assays , 120 day one adult animals were transferred onto fresh plates ( 12 animals per plate ) spotted with RNAi bacteria and exposed to 35°C ( N2 , EHC117 , EHC118 , EHC121 and CF512 worms for 11 hr and CB1370 worms for 19 hr ) and survival rates were recorded . To assess resistance to ultra-violet ( UV ) radiation , day 1 adult CF512 worms were exposed to sub-lethal UV dose ( 800 j/cm2 ) . Survival rates were scored daily . To evaluate resistance to pathogenic bacteria ( innate immunity ) , eggs of CF512 worms were placed on plates seeded with the indicated RNAi bacteria , grown to day 1 of adulthood , and transferred onto plates seeded with P . aeruginosa . Survival rates were followed daily . 20–24 hr post L4 worms were dissected in egg buffer ( 0 . 025 mM Hepes pH 7 . 4 , 118 mM NaCl , 48 mM KCl , 2 mM MgCl2 , 2 mM CaCl2 , 0 . 1% Tween 20 ) , transferred to superfrost plus slide and freeze cracked . Gonads were fixed in −20°C MeOH for 1 min , and 4% PFA for 30 min . Slides were washed twice in PBST ( PBS with 0 . 1% Tween 20 ) for 5 min , and incubated in PBST with 0 . 5 µg/ml DAPI for 10’ . Finally , the slides were washed for 10 min in PBST , again in 10 mM Tris 7 . 5% and 0 . 1% Tween 20 for 5 min , and sealed with Vectashield ( Vector Laboratories , # H-1000 ) . Imaging was done with Olympus IX81 inverted fluorescent microscope , and 3D images we collected and deconvolved with AutoQuant X3 . Germ cells nuclei were manually counted from the mitotic tip to the end of pachytene . Germ cell corpses were scored in 20 hr post-L4 adult hermaphrodites using acridine orange ( AO ) , as described in Melchior ( 2000 ) . A minimum of 23 gonads were scored for each genotype . Statistical analyses were performed using the two-tailed Mann–Whitney test ( 95% C . I . ) Statistical significance of the results was performed using the Student T-test , two-tailed distribution and two-sample equal variance . The analyses were done using at least three independent biological repeats of each experiment , as indicated . Statistical information of lifespan experiments is presented in Supplementary files 2–6 as mean LS ± SEM . | Aging may seem inescapable , but there are many factors , from diet to genetic mutations , that can affect this process . In fact , scientists have started to uncover the mechanisms that control and influence this slow decline . For example , in the small worm Caenorhabditis elegans , removing the germs cells – which give rise to eggs – extends the lifespan . Similarly , interfering with the activity of the Insulin/IGF-1 signaling ( IIS ) pathway leads to a longer life for the animals . However , it is unclear whether these two mechanisms work together , or if they operate in parallel . To explore this , Moll , Roitenberg et al . first looked at how the IIS pathway regulates a type of protein modification known as SUMOylation in C . elegans . Reducing the activity of the IIS pathway slowed down aging in the worms . It also decreased the levels of SUMOylation of certain proteins , including CAR-1 , which is found in the structures that produce germ cells . Further experiments showed that stopping the SUMOylation of CAR-1 extended the lifespan of the animals . In fact , replacing the protein with a mutated version of CAR-1 that cannot accept the SUMO element makes the worms live longer and resist a toxic protein that causes Alzheimer’s disease in humans . These results therefore show that , in C . elegans , the IIS pathway and a mechanism that involves CAR-1 in germ cells work together to determine the pace of aging . Further studies are now needed to dissect how the IIS pathway influences SUMOylation , and whether the findings hold true in mammals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2018 | The insulin/IGF signaling cascade modulates SUMOylation to regulate aging and proteostasis in Caenorhabditis elegans |
Cyanobacteria are photosynthetic bacteria with a unique CO2 concentrating mechanism ( CCM ) , enhancing carbon fixation . Understanding the CCM requires a systems level perspective of how molecular components work together to enhance CO2 fixation . We present a mathematical model of the cyanobacterial CCM , giving the parameter regime ( expression levels , catalytic rates , permeability of carboxysome shell ) for efficient carbon fixation . Efficiency requires saturating the RuBisCO reaction , staying below saturation for carbonic anhydrase , and avoiding wasteful oxygenation reactions . We find selectivity at the carboxysome shell is not necessary; there is an optimal non-specific carboxysome shell permeability . We compare the efficacy of facilitated CO2 uptake , CO2 scavenging , and HCO3− transport with varying external pH . At the optimal carboxysome permeability , contributions from CO2 scavenging at the cell membrane are small . We examine the cumulative benefits of CCM spatial organization strategies: enzyme co-localization and compartmentalization .
We present our mathematical model , which captures all aspects of the CCM as described above . This model is an expansion of previously developed models ( Reinhold et al . , 1989; Fridlyand et al . , 1996; Reinhold et al . , 1991 ) . Our three dimensional model of the CCM solves for both the HCO3− and CO2 concentration throughout a spherical cell . We solve this model numerically and analytically at steady state for three different spatial organizations of carbonic anhydrase and RuBisCO in the cell ( See Figure 6 ) : enzymes distributed evenly throughout the cell , enzymes localized to the center of the cell but not encapsulated ( as they would be on a scaffold ) , enzymes encapsulated in a carboxysome . We compare the effects of these scenarios in the discussion section , and for now consider a spherical cell of radius Rb = 0 . 5 μm with a single spherical carboxysome of radius Rc = 50 nm containing RuBisCO and carbonic anhydrase . Numerical computations are carried out with finite difference methods in MATLAB . The details of analytic solutions are given in the Supplementary file 1 . We include the effects of diffusion , active transport and leakage through the cell membrane , and reactions with carbonic anhydrase and RuBisCO . In the carboxysome ( r < Rc ) , the equations governing the HCO3− and CO2 , here H and C respectively , are ( 1 ) ∂tC=D∇ 2C+RCA−RRub ( 2 ) ∂tH=D∇2H−RCA , where here D is the diffusion constant , and RCA is the carbonic anhydrase reaction , and RRub is the RuBisCO reaction . The carbonic anhydrase reaction follows reversible Michaelis–Menten kinetics ( Kaplan & Reinhold , 1999; Price et al . , 2007 ) , ( 3 ) RCA ( H , C ) =VbaKcaH−VcaKbaCKbaKca+KcaH+KLbaC , where Vca and Vba are hydration and dehydration rates , proportional to the local carbonic anhydrase concentration . Kca and Kba are the concentration at which hydration and dehydration are half maximum . The RuBisCO reaction follows Michaelis–Menten kinetics with competitive binding with O2 , RRub=VmaxC/ ( C+Km ) , where km=km′ ( 1+O/ki ) . Here Vmax is the maximum rate of carbon fixation and Km is the apparent half maximum concentration value , which has been modified to include competitive binding with O2 , O . Ki is the dissociation constant of O2 with the RuBisCO and Km′ is the half maximum concentration with no O2 present . RuBisCO also requires ribulose-1 , 5-bisphosphate , the substrate which CO2 reacts with to produce 3-phosphoglycolate . Under CO2 limiting conditions it has been shown that there is sufficient ribulose-1 , 5-bisphosphate to saturate all RuBisCO active sites , and the reaction rates are independent of ribulose-1 , 5-bisphosphate concentrations ( Mayo et al . , 1989; Whitehead et al . , 2014 ) . In the cytosol there is no carbonic anhydrase or RuBisCO activity , so RCA = 0 and RRub=0 , and there is only diffusion of CO2 and HCO3− . We do not include the natural , but slow , interconversion of CO2 and HCO3− in the cytosol . This assumption is a good one given that the HCO3− concentration is known to be held out of equilibrium in the cell ( Volokita et al . , 1984; Price & Badger , 1989 ) . In agreement with this experimental observation , we find that all the other processes effecting the concentration of HCO3− in the cytosol happen much faster than the natural interconversion . Boundary conditions prescribe the inorganic carbon fluxes into the cell and the diffusion across the carboxysome boundary . We treat the inorganic carbon fluxes at cell and thylakoid membranes together . At this cell boundary , there is passive leakage of both CO2 and HCO3−: the velocity of CO2 across the cell membrane , kmc is about 1000-fold higher than that of HCO3− , kmH , due to the lower permeability of the membrane to charged molecules . To account for active import of HCO3− , we combine the total HCO3− flux , jc , from all HCO3− transporters . These transporters include BCT1 ( encoded by cpm ) , which is thought to be powered by ATP; and BicA and SbtA which are thought to be symporters between HCO3− and Na+ , driven by the highly controlled electrochemical gradient for Na+ ( Price et al . , 2008 , 2004; Omata et al . , 1999 ) . Additionally , there are two complexes NDH-13 and NDH-14 responsible for converting CO2 to HCO3− . This conversion is thought to either decrease CO2 , creating a gradient across the membranes and ‘facilitating uptake’ of CO2 , or ‘scavenge’ CO2 which has escaped from the carboxysome . These are localized to the thylakoid and possibly the plasma membrane . They have been linked to the photosynthetic linear and cyclic electron transport chain ( Price et al . , 2008; Maeda et al . , 2002; Shibata et al . , 2001 ) . It has been proposed that electron transport drives the formation of local alkaline pockets where CO2 more rapidly converts to HCO3− . We more simply describe the conversion with a maximal reaction rate α , and concentration of half maximal activity of Kα . Combining these effects , the boundary condition setting diffusive flux of HCO3− and CO2 at the cell membrane is ( 4 ) D∂rC=−αCcytosolKα+Ccytosol+kmC ( Cout−Ccytosol ) ( 5 ) D∂rH=jcHout+αCcytosolKα+Ccytosol+kmH ( Hout−Hcytosol ) where the subscript cytosol and out indicate we are taking the concentration immediately inside and outside the cell boundary respectively . The diffusion constant times partial derivatives with respect to the radial coordinate , r , define the diffusive flux at the membrane . The carboxysome shell is composed of proteins with a radius Rc ≈ 50 nm . As of yet , there have been no direct measurements of the carboxysome permeability to small molecules . Using the carboxysome geometry , we can calculate an upper bound for the diffusive velocity across the carboxysome shell , which is directly related to the carboxysome permeability . Crystal structures ( Yeates et al . , 2008; Cheng et al . , 2008; yeates et al . , 2007 ) show the surface has approximately Npores = 4800 small pores with radius rpore≈0 . 35 nm , and thickness l = 1 . 8 nm . If kc is the characteristic velocity that small molecules pass through the shell , these numbers imply the upper bound for diffusive transport kc<πrpore2D4πRc2l ( Npores ) ≈0 . 02cms . This calculation is done by taking the probability that a molecule will encounter a pore on the carboxysome shell ( Npores×pore surface areacarboxysome surface area ) and multipling it by the speed a small molecule will diffuse through the length of the pore ( D/l ) . Since it does not take into account any charge effects , which would add an additional energy barrier , it is an upper bound . Although there has been much speculation that the positively charged pores might enhance diffusion of negatively charged HCO3− ( Yeates et al . , 2008; Dou et al . , 2008; Cheng et al . , 2008 ) , here we explore the simplest assumption , that both HCO3− and CO2 have the same permeability . Namely , the boundary conditions at the carboxysome shell are ( 6 ) D∂rC=kc ( Ccytosol−Ccarboxysome ) ( 7 ) D∂rH=kc ( Hcytosol−Hcarboxysome ) . We will vary kc ( henceforth called carboxysome permeability , although it is a velocity ) within our model and see that there is a range of kc where the CCM is effective even with kc identical for CO2 and HCO3− .
Now that we have defined our model , we wish to find the range of parameters where efficient carbon fixation occurs . In what follows , we fix the enzymatic rates , cell membrane permeability , and diffusion constant as reported in the literature ( Jordan & Ogren , 1981; Missner et al . , 2008; Gutknecht et al . , 1977; Heinhorst et al . , 2006 ) ( see Table 1 and Table 2 ) . Note that full analytic solutions are available in Supplementary file 1 sections S3 and S4 , so the effect of varying other parameters can be analyzed . We consider the efficacy of the CCM as a function of jc , the flux of HCO3− into the cell , kc , the carboxysome permeability , and the parameters ( α , Kα ) governing the CO2 facilitated uptake mechanism . Both α and jc can be regulated by the organism and vary depending on environmental conditions , whereas the carboxysome permeability , kc , is the parameter with the largest uncertainty and debate ( Cannon et al . , 2001; Yeates et al . , 2008; Cheng et al . , 2008 ) . 10 . 7554/eLife . 02043 . 005Table 1 . Parameter values chosen for main set of simulations , unless otherwise indicatedDOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 005ParameterDefinitionValueReferenceHoutconcentration of bicarbonate outside the cell14 μM* ( Price et al . , 2008 ) Coutconcentration of carbon dioxide outside of cell0 . 14 μM* ( Price et al . , 2008 ) Ddiffusion constant of small molecules , CO2 and HCO3−10−5 cm2s ( Fridlyand et al . , 1996 ) kmcpermeability of cell membrane to CO20 . 3 cms ( Missner et al . , 2008; Gutknecht et al . , 1977 ) kmHpermeability of cell membrane to HCO3−3×10−4cms ( Missner et al . , 2008; Gutknecht et al . , 1977 ) Rcradius of carboxysome5×10−6 cm ( Cheng et al . , 2008; Schmid et al . , 2006 ) Rbradius of bacteria5 × 10−5 cm ( Savage et al . , 2010 ) jcHCO3− transport rate resulting in 30mM cytosolic HCO3− pool0 . 6 cms*calculated herekcoptimal carboxysome permeability10−3 cms*calculated hereVcellcell volume5 . 2×10−10μLcalculatedSAcellcell surface area3×10−8cm2calculated*these parameters are varied in the text , but these values are use unless noted otherwise . 10 . 7554/eLife . 02043 . 006Table 2 . Table comparing enzymatic rates ( Woodger et al . , 2005; Sultemeyer et al . , 1995; Heinhorst et al . , 2006 ) DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 006Enzyme reactionactive siteskcat [1s]Vmax in ‘cell’ [μMs]Vmax in carboxysome [μMs]K1/2 [μM]carbonic anhydrase hydration808 × 1048 . 8 × 1031 . 5 × 1073 . 2 × 103carbonic anhydrase dehydration804 . 6 × 1041 . 5 × 1048 . 8 × 1069 . 3 × 103RuBisCO carboxylation2160261781 . 8 × 105270Vmax in cell and carboxysome refer to the volumetric reaction rate assuming the enzymes are distributed throughout the entire cell or only carboxysome . Vba ( Vmax for carbonic anhydrase dehydration ) is estimated by assuming Keq = 5 and using parameters found in ( Heinhorst et al . , 2006 ) . Vca is Vmax for carbonic anhydrase hydration . Similarly , Kba , and Kca are K1/2 for dehydration and hydration respectively . For any given pair of kc and jc , we ask whether the CO2 concentrating mechanism is effective , using the criteria of saturating RuBisCO , reducing oxidation reactions , and not increasing the HCO3− concentration beyond carbonic anhydrase saturation . Our central result is presented in Figure 2 , which shows the range of kc and jc where these conditions are met , assuming that there is no facilitated uptake , α = 0 . The blue shaded region shows where RuBisCO is unsaturated , and the red shaded region shows where carbonic anhydrase is saturated . There is a crescent shaped region between these regions , where the CCM is effective according to our criteria . In the white region oxygenation reactions happen at a rate of greater than 1% . In the green shaded region oxygenation reactions occur at a rate of less than 1% . Within the white and green regions the CO2 concentration in the carboxysome varies greatly . 10 . 7554/eLife . 02043 . 007Figure 2 . Phase space for HCO3− transport , jc , and carboxysome permeability kc . Plotted are the parameter values at which the CO2 concentration reaches some critical value . The left most line ( dark blue ) indicates for what values of jc and kc the CO2 concentration in the carboxysome would half-saturate RuBisCO ( Km ) . The middle line ( light blue ) indicates the parameter values which would result in a CO2 concentration where 99% of all RuBisCO reactions are carboxylation reactions and only 1% are oxygenation reactions when O2 concentration is 260 μM . The right most ( red ) line indicates the parameter values which result in carbonic anyhdrase saturating . Here α = 0 , so there is no CO2 scavenging or facilitated uptake . The dotted line ( grey ) shows the kc and jc values , where the HCO3− concentration in the cytosol is 30 mM . The HCO3− concentration in the cytosol does not vary appreciably with kc in this parameter regime , and reaches 30 mM at jc≈0 . 6cms . All other parameters , such as reaction rates are held fixed and the value can be found in the Table 1 and Table 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 00710 . 7554/eLife . 02043 . 008Figure 2—figure supplement 1 . Phase space for HCO3− transport and carboxysome permeability . Solid lines show lines of constant CO2 concentration in the carboxysome for Dc=1e−5cm2s , or the diffusion constant of small molecule in water . Dashed lines show the same lines of constant CO2 concentration , bur for Dc=1e−7cm2s , or the diffusion constant of a small molecule in a 60% sucrose solution . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 00810 . 7554/eLife . 02043 . 009Figure 2—figure supplement 2 . Phase space for HCO3− transport , jc , and carboxysome permeability , kc . Plotted are the parameter values at which CO2 concentration reaches some critical value . The left most line ( dark blue ) indicates for what values of jc and kc the CO2 concentration in the carboxysome would saturate RuBisCO . The middle line ( light blue ) indicates the parameter values which would result in a CO2 concentration where 99% of all RuBisCO reactions are carboxylation reactions and only 1% are oxygenation reactions when O2 concentration is 260 µM . The right most ( red ) line indicates the parameter values which result in carbonic anyhdrase saturating . Here α=0cms ( solid lines ) and α=0cms ( dashed line ) , showing the effect of CO2 scavenging or facilitated uptake on the phase space . All other parameters , such as reaction rates are held fixed and the value can be found Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 009 The lines dividing the regions in Figure 2 are lines of constant carboxysomal CO2 concentration in jc and kc parameter space . The dark blue line is where CO2 = Km , the CO2 concentration for half-maximum RuBisCO reactions . The light blue line indicates parameter values resulting in the CO2 concentration ( C99% ) where rate of oxygenation reactions is 1% for O2 concentration of 260 μM . Varying carboxysome permeability , kc values , require more or less HCO3 transport , jc , to achieve the same carboxysomal CO2 concentration . We can calculate an amplification factor for the C99% carboxysomal CO2 concentration as Ac=CcarboxysomeCout+Hout=133 . Any combination of jc and kc which produce C = C99% , make 133 times more CO2 available in the carboxysome than there is total inorganic carbon outside the cell . Generally , increasing HCO3− transport , below the carbonic anhydrase saturation point results in higher CO2 concentration in the carboxysome . The basic physics of the phase diagram Figure 2 follows from examining how CO2 and HCO3− in the carboxysome change as jc is varied . Figure 3 shows the response to varying jc , with kc=10−3cms ( the optimal value in Figure 4 ) . 10 . 7554/eLife . 02043 . 010Figure 3 . Numerical solution ( diamonds and circles ) and analytic solutions ( carbonic anhydrase unsaturated , solid lines , and saturated , dashed lines ) correspond well . HCO3− transport is varied , and all other system parameters are held constant . The CO2 concentration above which RuBisCO is saturated is Km ( grey dashed line ) . The CO2 concentration where the oxygen reaction error rate will be 1% is C99% ( grey dash-dotted line ) . The transition between carbonic anyhdrase being unsatruated and saturated happens where the two analytic solutions meet ( where the dashed and solid red lines meet ) . Below a critical value of transport , jc≈10−3cms the level of transport is lower than the HCO3− leaking through the cell membrane . A value of kc=10−3 cm/s for the carboxysome permeability was used for these calculations . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 01010 . 7554/eLife . 02043 . 011Figure 3—figure supplement 1 . No effect of localizing carbonic anhydrase to the shell of the carboxysome . We assume the same amount of carbonic anhydrase and RuBisCO activity for each simulation and compare the case with the enzymes evenly distributed throughout the carboxysome to the case where the carbonic anhydrase is localized to the inner carboxysome shell . The ( - . - ) lines are for no organization and ( x ) for localization with Dc=10−5cm2s . The ( … ) lines are for no organization and ( o ) are for localization with Dc=10−7cm2s . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 01110 . 7554/eLife . 02043 . 012Figure 4 . Concentration of CO2 in the carboxysome with varying carboxysome permeability ( A ) . Numerical solution ( diamonds and circles ) and analytic solutions ( carbonic anhydrase unsaturated , solid lines , and saturated , dashed lines ) correspond well . On all plots CO2 ( red circle ) < HCO3− ( blue diamond ) . Concentration in the cell along the radius , r , with carboxysome permeability kc=10−5cms ( B ) , kc=10−3cms ( C ) , kc=1cms ( D ) . Grey dotted lines in ( B ) , ( C ) , ( D ) indicate location of the carboxysome shell boundary . The transition from low CO2 at high permeability ( D ) to maximum CO2 concentration at optimal permeability ( C ) occurs at kc∗=DRc=2cms . At low carboxysome permeability ( B ) HCO3− diffusion into the carboxysome is slower than consumption . For all subplots α=0cms and jc=0 . 6cms . Qualitative results remain the same with varying jc , increasing α will increase the gradient of CO2 across the cell as CO2 is converted to HCO3− at the cell membrane . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 012 When jc is low , the ratio of CO2 and HCO3− is constant , set by the chemical equilibrium at a given pH . In this case the rate of the carbonic anhydrase reaction is much faster than diffusion within the carboxysome , so that VbaKcaH=VcaKbaC . Unlike the uncatalyzed interconversion of CO2 and HCO3− in the cytosol , carbonic anhydrase brings the concentrations in the carboxysome to equilibrium very quickly . The chemical equilibrium is Keq=H/C= ( KbaVca ) / ( KcaVba ) ≈5 , for pH around 7 [Heinhorst et al . , 2006; DeVoe & Kistiakowsky , 1961 ) ] , so that HCO3− > CO2 in the carboxysome . Increased pH would increase Keq and the proportion of HCO3− , while decreased pH would decrease Keq and the proportion of HCO3− . Such variations do not substantially effect the subsequent discussion and mechanisms , although they will change the absolute values of CO2 concentration in the carboxysome . The Km dashed line in Figure 3 shows the CO2 level above which RuBisCO reaction is saturated: this gives the RuBisCO saturated ( blue ) boundary in Figure 2 . We have similarly marked the concentration C99% where there is a 1% oxygen reaction error rate with a dash-doted line . At higher levels , the CO2 concentration no longer increases with increasing jc , because the carbonic anhydrase is saturated . The saturated regime occurs in Figure 3 when Hcarboxysome>Kba , so that increasing Hcarboxysome ( controlled directly by jc ) no longer increases the rate of production of Ccarboxysome . This transition from unsaturated to saturated carbonic anhydrase defines the line for the carbonic anhydrase saturated region in Figure 2 . For each line of constant concentration in Figure 2 there is an optimal permeability value , where the least HCO3− transport is required to achieve the same CO2 concentration . The optimal permeability value shifts downward with increasing CO2 concentration ( compare light and dark blue curves ) . For C99% the optimal permeability is kc=10−3cms , below the calculated upper bound: kc<0 . 02cms obtained above from the carboxysome structure . To further understand the effect of permeability , we examine the CO2 concentration in the carboxysome for varying carboxysome permeabilities and a fixed HCO3− transport rate in Figure 4 . Figure 4A , shows that there is a broad range of kc where the CCM has maximal efficacy . Figure 4 shows the distribution of inorganic carbon throughout the cell when the permeability is low ( B ) , optimal ( C ) , and high ( D ) . At high permeability , the CO2 produced in the carboxysome rapidly leaks out of the carboxysome , and the CO2 concentration in the cytosol , shown in Figure 4D , is relatively high . When the carboxysome permeability decreases to near the optimal value , Figure 4C , the carboxysome traps CO2 , and the cytosolic levels are lower , decreasing leakage out of the cell . This transition occurs when diffusion across the cell ( and carboxysome ) takes a shorter time than diffusion through the carboxysome shell; or the CO2 in the carboxysome is effectively partitioned from the CO2 in the cell . If the carboxysome permeability is below optimal , diffusion of HCO3− into the carboxysome cannot keep up with consumption from RuBisCO , Figure 4B . The existence of an optima requires RuBisCO consumption to be low enough that there is a kc where the cytosol and carboxysome are partitioned , but HCO3− diffusion in can keep up . When such an optima exists , the carboxysome permeability can improve the CO2 concentration in the carboxysome without any special selectivity between HCO3− and CO2 . The location and concentrating power of the optimal regime , is dependent on the size of the cell and the membrane permeabilities to CO2 and HCO3− .
While we have solved our model to describe a vast parameter space it is instructive to compare the fluxes and concentrations we find within our optimal parameter space ( the green region in Figure 2 ) to actual numbers . At low external inorganic carbon conditions , internal inorganic carbon pools due to CCM activity are regularly measured as high as Ci = 30 mM . The inorganic carbon is predominantly in the form of HCO3− , and measurements do not distinguish between the cytosol and carboxysome ( Kaplan & Reinhold , 1999; Woodger et al . , 2005; Sultemeyer et al . , 1995; Price et al . , 1998 , 2008 ) . In our model , we find that the cytosolic HCO3− concentration is 30 mM when jc=0 . 6cms , over a wide range of the carboxysome permeability ( indicated as the dashed grey line in Figure 2 ) . From Figure 4 we can also see that the cytosolic HCO3− concentration is the dominate form of inorganic carbon in the cell at jc=0 . 6cms . We examine the fate of the HCO3− transported into the cell in terms of the HCO3− leaking out , CO2 leaking out , CO2 fixation or carboxylation , and O2 fixation or oxygenation ( Table 3 ) . 10 . 7554/eLife . 02043 . 013Table 3 . Fate of carbon brought into the cell for jc = 0 . 6cm/s and kc = 10–3cm/sDOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 013formula[picomoles ( cell s ) ]% of fluxHCO3− transportjcHout3 . 26 × 10−4HCO3− leakagekmH ( Hout−Hcytosol ( Rb ) ) 3 . 2 × 10−498 . 6%CO2 leakagekmC ( Cout−Ccytosol ( Rb ) ) 4 . 5 × 10−41 . 4 %carboxylationVmaxCC+Km ( 1+OKO ) 8 . 2 × 10−80 . 03 %oxygenationVmaxOCC+KmO ( 1+CKC ) 6 . 7 × 10−102 × 10−4 % For cells grown under low inorganic carbon conditions net HCO3− fluxes ( transport - leakage ) are measured 105pmolmgChls , with CO2 net flux being slightly lower but the same order of magnitude ( Whitehead et al . , 2014; Badger et al . , 1994 ) . High external inorganic carbon conditions produce slightly higher net HCO3− rates ( Tchernov et al . , 1997 ) . Assuming chlorophyll per cell volume of around 10−11mgChlcell for cells of our size we can convert this into a flux of 10−6pmol ( cells ) ( Whitehead et al . , 2014; Keren et al . , 2004 , 2002 ) . The net HCO3− flux for our model cell is 6×10−6pmol ( cells ) , so we are about an order of magnitude too high . If we choose a HCO3− transport value one order of magnitude smaller , we will get net fluxes of the same order of magnitude as the measurements at the cost of slightly lower carboxylation rates and higher oxygenation rates ( Table 4 ) . This would also mean a lower internal HCO3− pool . Alternatively , the same internal HCO3− could be reached at a lower flux rate , if the external HCO3− is higher . Since the majority of the HCO3− transport is balanced by HCO3− leakage , we can find the transport rate needed to sustain a particular amplification by the simple formula: jc=kmH ( Hout−Hcytosol ( Rb ) ) /Hout . 10 . 7554/eLife . 02043 . 014Table 4 . Fate of carbon brought into the cell for jc = 0 . 06 cm/s and kc = 10–3cm/sDOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 014formula[picomoles ( cell s ) ]% of fluxHCO3− transportjcHout2 . 8 × 10−5HCO3− leakagekmH ( Hout−Hcytosol ( Rb ) ) 2 . 7 × 10−596 . 6 %CO2 leakagekmC ( Cout−Ccytosol ( Rb ) ) 8 . 8 × 10−73 . 2 %carboxylationVmaxCC+Km ( 1+OKO ) 5 . 4 × 10−80 . 2 %oxygenationVmaxOCC+KmO ( 1+CKC ) 2 . 3 × 10−98 × 10−3 % While we can compare the net fluxes , we have not found direct experimental evidence for the absolute HCO3− uptake rate . To determine whether this HCO3− transport rate is reasonable we perform a back of the envelope calculation . Our simulated cell has a flux of 2 × 108 molecules/s . Assuming the rate of transport per transporter of 103moleculess and our cell's surface area this requires about 104transportersμm2 . This is about an order of magnitude higher than the number of ATP synthase complexes on the thylakoid membrane in spinach , 700 complexesμm2 ( Miller and Staehelin , 1979 ) . According to our calculation only around 1 % of the carbon transported into the cell is fixed into 3-phosophoglycerate . Even in this highly CO2 concentrating regime , 5 × 104 2-phosophoglycolate produced per second . Cyanobacteria have been shown to have multiple pathways for recycling 2-phosophoglycolate ( Hackenberg et al . , 2009 ) . Our system fixes CO2 at a rate of 0 . 14 pg/hour . Given the volume of our cell , and the fact that between 115–300 fg/μm3 of carbon are needed to produce a new cyanobacterial cell ( Mahlmann et al . , 2008 ) we need between 0 . 1 and 0 . 3 picograms of carbon per cell . At the higher flux rate ( Table 3 ) this means that a cell could replicate every 7–21 hr and the lower flux rate ( Table 4 ) allows replication every 11–35 hr . Both are consistent with the division times of cyanobacteria . At jc=0 . 6cms , varying the carboxysome permeability changes how the available inorganic carbon is partitioned between the carboxysome and cytosol , thereby setting the carboxysomal CO2 concentration as shown in Figure 4 . Strikingly , the HCO3− concentration is constant across the cytosol . This is because the cell membranes have low permeability to HCO3−; thus , the rate of escape is slow and HCO3− equilibrates across the cell . A consequence of this flat HCO3− profile is that the carboxysome experiences the same HCO3− concentration , independent of its position in the cell . This means the incoming inorganic carbon source for the carboxysome system is invariant with the position of the carboxysome in the cell . In contrast , there is a gradient in CO2 concentration across the cell when the carboxysome permeability is at or above the optimum ( Figure 4C , D ) . The cell membrane is more permeable to CO2 . The gradient means that the CO2 leakage out of the cell affects the CO2 leakage out of the carboxysome . Moving the carboxysome close to the cell membrane increases the leakage rate of CO2 out of the carboxysome . Notably , in S . elongatus the carboxysomes are located along the center line of the cell , away from the cell membranes ( Savage et al . , 2010 ) . The spatial profiles of HCO3− and CO2 give no hint as to why the carboxysomes are spaced apart from one another . Since the gradient in HCO3− is flat , there is no competition between the carboxysomes for HCO3− ( the main incoming source of inorganic carbon ) . In fact , since the local concentration of CO2 is higher near a carboxysome , nearby carboxysomes would ‘feed’ each other CO2 . As has been shown , such clumping would reduce the probability of distributing carboxysomes equitably to daughter cells , possibly counteracting any benefit ( Savage et al . , 2010 ) . The concentration across the carboxysome is basically constant , because the carboxysome is so small that diffusion across it takes very little time . A consequence of this is that the organization of the reactions in the carboxysome does not effect the CO2 concentration in the carboxysome ( Figure 3—figure supplement 1 ) . Therefore , the localization of the carbonic anhydrase to the inner carboxysome shell seems to have no effect on the CCM . It has been suggested that diffusion in the carboxysome should be slower , since the carboxysome is packed with RuBisCO . One proposed consequence of slower diffusion in the carboxysome is that it could trap CO2 , making a low carboxysome permeability unnecessary . We have tested this hypothesis ( Figure 2—figure supplement 1 ) , and find that assuming the diffusion constant one would expect for small molecules in a 60% sucrose solution ( Dc=10−7cm2s ) , does reduce the optimal carboxysome permeability . However , for any carboxysome permeability a higher HCO3− transport rate is needed to achieve the same carboxsomal CO2 concentration . So if the diffusion is indeed slower in the carboxysome it does not aid the CCM . Even at this slower diffusion , the CO2 concentration across the carboxysome is flat . We investigate the effect of CO2 to HCO3− conversion at the thylakoid and cell membranes ( combined in the model ) . Increasing conversion , α > 0 , can facilitate uptake of CO2 from outside the cell and scavenge CO2 escaped from the carboxysome . Facilitated uptake results in saturating both carbonic anhydrase and RuBisCO at a lower level of HCO3− transport . Scavenging broadens the range of carboxysome permeability which will effectively separate the inorganic carbon pools in the carboxysome and outside . Scavenging decreases the concentration of CO2 in the cytosol , so a more permeable carboxysome can still result in a low leakage rate of inorganic carbon out of the cell ( more of the inorganic carbon in the cytosol is in the form of HCO3− which leaks out less readily ) . However , neither of these effects is particularly strong in our current range of reaction rates , and cell membrane permeability ( Figure 2—figure supplement 2 ) . The relative effect of these two mechanisms depends on the external CO2 and HCO3− concentrations . In saltwater environments the pH is near 8 and HCO3− is the predominant inorganic carbon source . While external pH is not explicitly treated in our model , we can account for changes to pH through the external inorganic carbon concentration . To be consistent with oceanic environment , thus far we have shown results for low external inorganic carbon concentrations of [CO2] = 0 . 1 μM and [HCO3−] = 14 . 9 μM . The effect of facilitated uptake , under these assumptions , is very small . In freshwater or under conditions of ocean acidification , where the pH could fall to 6 or lower , there can be a much larger proportion of CO2 ( >50% ) . Figure 5 shows the absolute contribution of HCO3− transport , facilitated CO2 uptake , and CO2 scavenging for varying proportions of external CO2 . Even though we assume the same velocity of facilitated uptake and HCO3− transport ( jc=αKα=1 ) , facilitated uptake contributes less because it is limited by CO2 diffusion across the membrane . At the same rates of transport the facilitated uptake mechanism only contributes more than active HCO3− if the CO2 concentration is greater than 80% of external inorganic carbon . This is consistent with observations that oceanic cyanobacteria such as Prochlorococcus only seem to possess gene homologs for HCO3− transport systems , while other freshwater and estuarine cyanobacteria have gene homologs for both constitutive ( NDH-14 ) and inducible ( NDH-13 ) CO2 uptake systems as well as inducible HCO3− transport systems ( BicA , SbtA , and BCT1 ) ( Price , 2011 ) . 10 . 7554/eLife . 02043 . 015Figure 5 . Size of the HCO3− flux in one cell from varying sources , as the proportion of CO2 to HCO3− outside the cell changes changes . We show results for three carboxysome permeabilities , kc , and only the scavenging is effected . Total external inorganic carbon is 15μM , jc=1cms and αKα=1cms . Scavenging is negligibly small for all values of kc shown . Unless there is very little HCO3− in the environment , HCO3− transport seems to be more efficient than CO2 facilitated uptake . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 015 Scavenging is negligibly small for all values of kc shown . There is very little CO2 in the cytosol , so there is very little CO2 to scavenge , Figure 5 . The effect of scavenging is dependent on the cell membrane permeability to CO2 and HCO3− . Given that scavenging has no obvious affect on HCO3− concentrations , it is reasonable to wonder why this mechanism exists at all . One might assume that scavenging prevents leakage , but if the energy to bring a ‘new’ HCO3− molecule from outside the cell is the same as the energy required to save an ‘old’ CO2 molecule from leaking out , there is no obvious advantage of preventing the leakage . It is possible that since the scavenging mechanism is associated with the electron transport chain of the light reactions of photosynthesis scavenging can be ramped up more easily when there is excess light energy . If this were the case , a comparison of jc=1cms and αkα=1cms is deceiving and αKα could be much larger . Indeed it has been suggested that the cell uses this mechanism as a way to dissipate excess light energy ( Tchernov et al . , 1997 , 2003 ) . The most striking aspect of the CCM is the way that spatial organization is used to increase the efficacy of the reactions . Figure 6 compares the effect of different enzymatic reaction organizations . Concentrating carbonic anhydrase and RuBisCO to a small region in the center of the cell , on a scaffold for example , leads to an order of magnitude increase in the concentration of CO2 . Localizing the carbonic anhydrase to a small volume concentrates it , increasing the maximum reaction rate per volume , Vca and Vba . A larger Vba increases the HCO3− concentration at which carbonic anhydrase is saturated allowing the mechanism to take advantage of a larger HCO3− flux , jc . A small increase can be gained from encapsulating the enzymes in a permeable carboxysome shell and another order of magnitude is gained at the optimal permeability . At optimal carboxysome permeability , the CO2 is effectively partitioned into the carboxysome and conversion can act only as facilitated uptake as shown in Figure 5 . 10 . 7554/eLife . 02043 . 004Figure 6 . Concentration of CO2 achieved through various cellular organizations of enzymes , where we have selected the HCO3− transport level such that the HCO3− concentration in the cytosol is 30 mM . O2 concentration is 260 μM . The oxygenation error rates , as a percent of total RuBisCO reactions are indicated on the concentration bars . The cellular organizations investigated are RuBisCO and carbonic anhydrase distributed throughout the entire cytosol , co-localizing RuBisCO and carbonic anhydrase on a scaffold at the center of the cell without a carboxysome shell , RuBisCO and carbonic anhydrase encapsulated in a carboxysome with high permeability at the center of the cell , and RuBisCO and carbonic anhydrase encapsulated in a carboxysome with optimal permeability at the center of the cell . DOI: http://dx . doi . org/10 . 7554/eLife . 02043 . 004 Another advantage of localizing the enzymes in a small region at the center of the cell is separating carbonic anhydrase from the α ( CO2 to HCO3− ) conversion mechanism , preventing a futile cycle . The futile cycle is most detrimental when the enzymes are distributed through out the cytosol , and increases the oxygenation error rate ( data not shown ) . Concentrating the enzymes away from the cell and thylakoid membranes , where conversion happens , removes this effect . On a scaffold the oxygenation rate is almost exactly the same with and without the α conversion mechanism . This is consistent with the previously shown detrimental effect of having active carbonic anhydrase free within the cytosol ( Price & Badger , 1989 ) . It would be impossible to keep the cytosol completely free from carbonic anhydrase enzyme , so there must be a way of activating it within the carboxysome only . Carbonic anhydrase is inactivated under reducing conditions ( Peña et al . , 2010 ) . Recently it was shown that carboxysomes oxidize after assembly , providing a way to keep carbonic anhydrase inactive until fully enclosed in a carboxysome ( Chen et al . , 2013 ) . Cyanobacteria must regulate pH as almost all biochemical reactions are pH sensitive . We do not attempt to model this regulation or potential pH variation within the cell , however pH may be included implicitly in a couple ways . We have already explored the effect of varying external pH , and the effects of pH on carbonic anhydrase . Cytosolic pH would have little direct effect on the CO2 and HCO3− levels since the non-enzymatic interconversion is very slow as previously discussed . The effect of internal pH could also be explored by varying the reaction rate of RuBisCO , which is pH sensitive . Varying the reaction rate of RuBisCO greatly could change the range where a non-specific carboxysome permeability can increase the concentration of CO2 in the carboxysome . It would be unexpected that the RuBisCO rate be much faster than we assume , as we have assumed a rate on the high end . A lower RuBisCO rate would increase the range of effective carboxysome permeabilities . As previously mentioned the CO2 facilitated uptake mechanism functions by creating local alkaline pockets . Diffusion of hydrogen ions across the cell would be very fast , so such pockets would require a massive reduction from the light reactions to maintain local alkalinity . Whether such pH gradients are possible , is certainly a subject of future interest . We have described and analyzed a model for the CO2 concentrating mechanism in cyanobacteria . There exists a broad range of HCO3− transport and carboxysome permeability values which result in effective CO2 concentration in the carboxysome . This effective concentration parameter space is defined by CO2 levels high enough to saturate RuBisCO and produce a favorable ratio of carboxylation to oxygenation reactions , but not so high as to saturate carbonic anhydrase ( after which increasing HCO3− transport will not increase the CO2 concentration ) . An optimal carboxysome permeability exists , where HCO3− diffusion into the carboxysome is not substantially inhibited , but CO2 leakage is minimal . HCO3− concentrations across the cell are flat and are predominately set by the transport rate in , and leakage out . We quantitatively compare the transport rates and concentrations we predict in our optimal parameter space , and find them to be in good agreement with experiment . We also comment on the effects of external pH on CO2 versus HCO3− uptake mechanisms . Finally we describe the cumulative benefits of co-localization , encapsulation , and optimal carboxysome permeability on the CCM . Further comparison of this model to experimental flux measurements , especially to determine the quantitative contributions of different transporters under different physiological conditions would be very interesting . Current solutions are for steady state at constant external concentration , but most gas exchange measurements , by necessity , measure the fluxes as the inorganic carbon is depleted in the media . The model could be modified to solve the time dependent problem with varying external inorganic carbon . As of yet the carboxysome permeability has not been measured directly , and it would be quite interesting to see how close it is to our ‘optimal’ prediction . | Cyanobacteria are microorganisms that live in water and , like plants , they capture energy from the sun to convert carbon dioxide into sugars and other useful compounds . This process—called photosynthesis—releases oxygen as a by-product . Cyanobacteria were crucial in making the atmosphere of the early Earth habitable for other organisms , and they created the vast carbon-rich deposits that now supply us with fossil fuels . Modern cyanobacteria continue to sustain life on Earth by providing oxygen and food for other organisms , and researchers are trying to bioengineer cyanobacteria to produce alternative , cleaner , fuels . Understanding how cyanobacteria can be as efficient as possible at harnessing sunlight to ‘fix’ carbon dioxide into carbon-rich molecules is an important step in this endeavor . Carbon dioxide can readily pass through cell membranes , so instead cyanobacteria accumulate molecules of bicarbonate inside their cells . This molecule is then converted back into carbon dioxide by an enzyme found in specials compartments within cells called carboxysomes . The enzyme that fixes the carbon is also found in the carboxysomes . However , several important details in this process are not fully understood . Here , Mangan and Brenner further extend a mathematical model of the mechanism that cyanobacteria use to concentrate carbon dioxide in order to explore the factors that optimize carbon fixation by these microorganisms . Carbon fixation is deemed efficient when there is more carbon dioxide in the carboxysome than the carbon-fixing enzyme can immediately use ( which also avoids wasteful side-reactions that use oxygen instead of carbon dioxide ) . However , there should not be too much bicarbonate , otherwise the enzyme that converts it to carbon dioxide is overwhelmed and cannot take advantage of the extra bicarbonate . Mangan and Brenner's model based the rates that carbon dioxide and bicarbonate could move in and out of the cell , and the rates that the two enzymes work , on previously published experiments . The model varied the location of the enzymes ( either free in the cell or inside a carboxysome ) , and the rate at which carbon dioxide and bicarbonate could diffuse in and out of the carboxysome ( the carboxysome's permeability ) . Mangan and Brenner found that containing the enzymes within a carboxysome increased the concentration of carbon dioxide inside the cell by an order of magnitude . The model also revealed the optimal permeability for the carboxysome outer-shell that would maximize carbon fixation . In addition to being of interest to researchers working on biofuels , if the model can be adapted to work for plant photosynthesis , it may help efforts to boost crop production to feed the world’s growing population . | [
"Abstract",
"Introduction",
"Results",
"Discussion"
] | [
"ecology",
"cell",
"biology"
] | 2014 | Systems analysis of the CO2 concentrating mechanism in cyanobacteria |
Mutualisms can be promoted by pleiotropic win-win mutations which directly benefit self ( self-serving ) and partner ( partner-serving ) . Intuitively , partner-serving phenotype could be quantified as an individual’s benefit supply rate to partners . Here , we demonstrate the inadequacy of this thinking , and propose an alternative . Specifically , we evolved well-mixed mutualistic communities where two engineered yeast strains exchanged essential metabolites lysine and hypoxanthine . Among cells that consumed lysine and released hypoxanthine , a chromosome duplication mutation seemed win-win: it improved cell’s affinity for lysine ( self-serving ) , and increased hypoxanthine release rate per cell ( partner-serving ) . However , increased release rate was due to increased cell size accompanied by increased lysine utilization per birth . Consequently , total hypoxanthine release rate per lysine utilization ( defined as ‘exchange ratio’ ) remained unchanged . Indeed , this mutation did not increase the steady state growth rate of partner , and is thus solely self-serving during long-term growth . By extension , reduced benefit production rate by an individual may not imply cheating .
Mutualisms , mutually beneficial interactions between species , are widely observed between microbes ( Goldford et al . , 2018; Morris et al . , 2013; Morris et al . , 2012; Seth and Taga , 2014 ) and between microbes and their hosts ( Seth and Taga , 2014 ) . Often , mutualisms involve the release and consumption of essential metabolites such as vitamins and amino acids ( Beliaev et al . , 2014; Carini et al . , 2014; Helliwell et al . , 2011; Jiang et al . , 2018; Rodionova et al . , 2015; Zengler and Zaramela , 2018 ) . Extensive metabolic interactions between microbes have been thought to contribute to the difficulty of culturing microbes in isolation ( Kaeberlein et al . , 2002 ) . Under certain conditions , microbial metabolic exchanges may even promote community growth ( Tasoff et al . , 2015 ) . In communities , a mutation can exert direct effects on the individual itself as well as on the partner ( see Figure 1A for definition of “direct effects” ) . For example , a mutation can increase benefit supply to the partner wtihout affecting self's growth rate ( Figure 1A , ii ) . We classify this mutation as "strictly partner-serving" . As another example , a mutation can increase self's growth rate without affecting benefit supply to the partner ( Figure 1A , iii ) . By growing better , the mutant stimulates partner growth . However , we classify this mutation as "strictly self-serving" because promoting partner growth is mediated indirectly by improved self-growth . Qualitatively , the direct fitness effect of a mutation on self and on partner can be positive , neutral , or negative , giving rise to 3 × 3 = 9 types . Distinguishing mutation types is important for predicting their evolutionary successes . Consider the general case of microbial mutualisms without any partner choice mechanisms . That is , an individual is not capable of discriminating or ‘choosing’ among spatially-equivalent partners ( Sachs et al . , 2004; Shou , 2015 ) . Then , a well-mixed environment will favor mutations with a positive direct fitness effect on self ( i . e . selfish , strictly self-serving , and win-win; Figure 1B , ‘mixed’ ) . This is because in a well-mixed environment , benefits from mutualistic partners are uniformly distributed , and thus how much an individual contributes to mutualistic partners is irrelevant . In contrast , in a spatially-structured environment , mutations exerting a positive direct effect on the mutualistic partner ( i . e . win-win , strictly partner-serving , and altruistic ) can be favored , while selfish mutations can be disfavored ( Chao and Levin , 1981; Doebeli and Knowlton , 1998; Hamilton , 1964; Harcombe , 2010; Momeni et al . , 2013b; Nowak , 2006; Sachs et al . , 2004; Shou , 2015 ) ( Figure 1B , ‘spatial’ ) . This is because in a spatially-structured environment , interactions are localized and repeated between neighbors . If an individual does not aid its mutualistic neighbor , the individual will eventually suffer as its mutualistic neighbor perishes . Win-win mutations are particularly intriguing because they directly promote both sides of a mutualism . Here , we analyze a mutation arising during the evolution of an engineered yeast mutualistic community in a well-mixed environment . This community CoSMO ( Cooperation that is Synthetic and Mutually Obligatory ) ( Shou et al . , 2007 ) consists of two S . cerevisiae strains that interact via metabolite cross-feeding ( Figure 1C ) . L-H+ requires lysine ( L ) and overproduces and releases the adenine derivative hypoxanthine ( H ) ( Hart et al . , 2019a ) . The complementary H-L+ requires H , and overproduces and releases L . The two yeast strains are reproductively isolated , and thus can be regarded as two species . This mutualism is ‘cooperative’ in the sense that metabolite over-production is costly to both strains ( Figure 1—figure supplement 1; Waite and Shou , 2012 ) . Mutualisms modeled by CoSMO are widely observed in natural communities ( Beliaev et al . , 2014; Carini et al . , 2014; Helliwell et al . , 2011; Jiang et al . , 2018; Rodionova et al . , 2015; Zengler and Zaramela , 2018 ) , including those in the gut and the oral microbiota ( Palmer et al . , 2001; Rakoff-Nahoum et al . , 2014 ) . A simplified community allows us to gain mechanistic insights into basic questions in microbial ecology and evolution ( Momeni et al . , 2011 . ) Indeed , principles learned from CoSMO , including how fitness effects of species interactions affect the composition and spatial patterning of member species , and mechanisms that protect mutualisms from exploiters , have been found to operate in communities of non-engineered microbes ( references in Momeni et al . , 2013a; Momeni et al . , 2013b; Waite and Shou , 2012 ) . CoSMO offers an ideal system for examining the evolution of mutualistic cooperation , especially given the genetic tractability of budding yeast and validated phenotype quantification methods that we have developed ( Hart et al . , 2019a; Hart et al . , 2019b ) . In this study , we demonstrate that an intuitive definition of partner-serving phenotype in mutualism can lead to erroneous conclusions . We will conclude by discussing how to quantify important theoretical concepts such as ‘benefit’ , ‘cost’ , and ‘partner-serving phenotype’ , especially for microbial mutualisms where interactions span multiple generations .
We randomly isolated evolved L-H+ clones from independent communities . Since the community environment was lysine-limited ( Hart et al . , 2019a; Waite and Shou , 2012 ) , improved growth under lysine limitation would be self-serving . Indeed , while the ancestral strain failed to grow into micro-colonies on agar with low lysine ( 1 . 5 µM ) , all tested ( >20 ) evolved clones could ( Figure 2—figure supplement 1; Materials and methods ‘Microcolony assay’ ) , consistent with our previous findings ( Hart et al . , 2019a; Waite and Shou , 2012 ) . Thus , evolved L-H+ clones displayed self-serving phenotypes . Since hypoxanthine is also scarce in the community ( Hart et al . , 2019a ) , a mutant L-H+ cell with increased hypoxanthine release rate should allow the partner to grow faster and is thus partner-serving . We randomly chose several evolved L-H+ clones , and measured their hypoxanthine release rate ( Hart et al . , 2019a ) ( Materials and methods , ‘Release assay’; Figure 3—figure supplement 1 ) . Whereas two of the evolved clones released hypoxanthine at a similar rate as the ancestor , three increased release rates in the sense that each cell released more hypoxanthine per hour than the ancestor ( Figure 3—figure supplement 2 ) . We then sequenced genomes of both types of clones ( Materials and methods , ‘Whole-genome sequencing’; Figure 2—figure supplement 2 ) . Chromosome 14 duplication ( DISOMY14 ) occurred in all three clones that exhibited faster-than-ancestor hypoxanthine release rate ( Figure 3—figure supplement 2 , blue ) , and not in the other two clones that exhibited ancestral release rate ( Figure 3—figure supplement 2 , orange ) . When we back-crossed evolved clones harboring DISOMY14 to the ancestral background ( Materials and methods , 'Strains and medium' ) , only meiotic segregants containing DISOMY14 showed increased hypoxanthine release rate compared to the ancestor ( Figure 3—figure supplement 3 ) . Thus , DISOMY14 genetically co-segregated with increased release rate ( partner-serving ) . DISOMY14 repeatedly rose to a detectable frequency when L-H+ evolved with H-L+ in CoSMO ( 3 out of 3 lines ) , or when L-H+ evolved alone in lysine-limited chemostats ( 5 out of 5 lines ) ( Supplementary file 3 ) . Thus , DISOMY14 is likely adaptive in lysine limitation . Indeed , while ancestors failed to form micro-colonies on low-lysine plate , meiotic segregants containing DISOMY14 could ( self-serving ) ( Figure 2—figure supplement 1 ) . Taken together , we hypothesized DISOMY14 to be both self-serving and partner-serving , that is ‘win-win’ . Chromosome 14 harbors the high-affinity lysine permease LYP1 . To test whether LYP1 duplication might improve the growth rate of L-H+ in limited lysine , we inserted an extra copy of LYP1 into the ancestral L-H+ strain ( Materials and methods , 'Gene knock-in and knock-out’ ) , and quantified cell growth rate under various concentrations of lysine using a microscopy assay ( Materials and methods , ‘Microscopy growth assay’ ) ( Hart et al . , 2019b ) . LYP1 duplication indeed significantly increased the growth rate of L-H+ in low lysine ( Figure 2 , compare green with magenta ) . Similarly , deleting the duplicated copy of LYP1 from DISOMY14 cells abolished the self-serving phenotype of DISOMY14 ( Figure 2 , compare blue and cyan with magenta ) . Taken together , duplication of LYP1 is responsible for the self-serving phenotype of DISOMY14 . To identify which duplicated gene ( s ) might be responsible for the increased hypoxanthine release rate of DISOMY14 , we systematically deleted various sections of the duplicated Chromosome 14 ( Figure 3A; Materials and methods , ‘Chromosome truncation’ ) , and quantified hypoxanthine release rate ( Figure 3B ) . Duplication of the region between YNL193W and GCR2 was necessary for the increased release rate ( Figure 3B , orange ) . This region contains six genes , including WHI3 . Integrating an extra copy of WHI3 into the ancestor increased release rate to near that of DISOMY14 , while deleting one copy of WHI3 from DISOMY14 restored the ancestral release rate ( Figure 4A; Materials and methods , ‘Gene knock-in and knock-out’ ) . WHI3 encodes an inhibitor of the cell division cycle ( Garí et al . , 2001; Nash et al . , 2001 ) . When WHI3 is overexpressed , cell division becomes less frequent even though biomass grows at the same rate , resulting in larger cells . Consistent with this notion , deletion of WHI3 results in smaller cell size , whereas extra copies or overexpression of WHI3 increases cell size ( Garí et al . , 2001; Nash et al . , 2001 ) . Indeed , DISOMY14 cells are bigger than ancestral cells , as quantified by the Coulter counter ( Figure 4—figure supplement 1A; Materials and methods , ‘Cell size measurements’ ) . Integrating an extra copy of WHI3 into the ancestor increased mean cell size , while deleting the extra copy of WHI3 from DISOMY14 restored ancestral cell size ( Figure 4—figure supplement 1A ) . Consistent with DISOMY14 cells being bigger than ancestral cells , DISOMY14 cells utilized more lysine per birth than the ancestor ( Figure 4B; Materials and methods , ‘Metabolite utilization in batch culture’ ) . Integrating an extra copy of WHI3 into the ancestor increased lysine utilization per newborn cell , while deleting the extra copy of WHI3 from DISOMY14 reduced lysine utilization per newborn cell ( Figure 4—figure supplement 1B ) . Thus , although each DISOMY14 cell released more hypoxanthine , each also utilized more lysine . When we normalized hypoxanthine release rate per L-H+ cell ( rH ) by lysine utilization per L-H+ birth ( uL ) , which we define as the ‘H-L exchange ratio’ ( rH/uL ) , DISOMY14 is indistinguishable from the ancestor ( Figure 4C ) . In other words , a fixed amount of lysine can be converted to ancestral L-H+ cells which are more numerous but lower-releasing ( Figure 5B , i ) , or DISOMY14 cells which are fewer but higher-releasing ( Figure 5B , ii ) . As long as the ancestor and DISOMY14 displayed a similar total hypoxanthine release rate per lysine utilized , the partner H-L+ would not benefit more from one than the other during long-term community growth . In Discussion , we will describe how exchange ratio links to concepts such as ‘inclusive fitness’ . If DISOMY14 is not partner-serving , then fixed lysine supply would lead DISOMY14 and ancestor to release identical total hypoxanthine . To test this hypothesis , we cultured ancestral and DISOMY14 L-H+ cells in lysine-limited chemostats with a fixed rate of lysine supply ( Materials and methods , ‘Chemostat’ ) . Indeed , the steady state hypoxanthine concentrations in chemostats were indistinguishable between ancestor and DISOMY14 ( Figure 4D ) . Similar to batch culture results , in chemostats DISOMY14 released hypoxanthine at a higher rate than the ancestor on a per cell basis , but utilized more lysine per birth , leading to an identical H-L exchange ratio as the ancestor ( Figure 4—figure supplement 2B ) . Consistent with DISOMY14 not being partner-serving , partner H-L+ grew at the same steady state rate whether co-cultured with ancestral or DISOMY14 L-H+ ( Figure 4E , right panel ) . This also suggests that no new beneficial interactions have evolved between DISOMY14 L-H+ and partner H-L+ . Note that CoSMO grew at a steady state rate after an initial lag ( Figure 4—figure supplement 3 ) ( Hart et al . , 2019a ) . Here we do not consider the duration or growth rate during the lag phase , but rather , long-term steady state growth rate . Taken together , despite having a higher hypoxanthine release rate per cell , DISOMY14 is not partner-serving or win-win during steady state community growth . Rather , DISOMY14 is strictly self-serving .
General theories have been developed for mutualisms ( Archetti et al . , 2011; Doebeli and Knowlton , 1998; Foster and Wenseleers , 2006; Frank , 1994; Jones et al . , 2015; Sachs et al . , 2004; Trivers , 1971; West et al . , 2002; Yamamura et al . , 2004 ) . In mathematical models of mutualisms ( e . g . Doebeli and Knowlton , 1998; Foster and Wenseleers , 2006; Frank , 1994; Yamamura et al . , 2004 ) , exchanged goods ( investments ) were linked to fitness effects on the focal individual . For example , the net fitness gain of a focal individual = fitness gain per investment made * average investment made within the group – fitness loss per investment made * investment by the focal individual ( Frank , 1994 ) . For a focal H-L+ , the fitness loss term ( fitness loss per lysine released*total lysine released by the focal cell ) is fixed whether the interaction partner is ancestral or DISOMY14 L-H+ . The term of fitness gain per investment made ( how much faster H-L+ would grow per unit of lysine supplied to L-H+ ) does embody the spirit of exchange ratio ( total hypoxanthine return rate per unit of lysine invested ) . However , measuring fitness gain per investment can be difficult since this value is unlikely to be a constant . For example , the growth rate of L-H+ is not a linear function of lysine concentration ( Figure 2 ) . In contrast in exchange ratio , ‘benefit’ and ‘investment’ are defined in physical terms of the goods exchanged , and fitness ‘cost’ for making the investment is separately considered ( see Equation 1 below ) . A focal L-H+'s exchange ratio is equivalent to partner's benefit-to-investment ratio . There are in fact two perspectives for this equivalence . In the ‘individualistic’ perspective , we can define the benefit gained by partner H-L+ as the hypoxanthine release rate by the focal L-H+ cell ( rH ) , while investment made by H-L+ as the amount of lysine required to make the focal L-H+ cell ( uL ) . Then , H-L+'s benefit-to-investment ratio is identical to L-H+'s exchange ratio . In the ‘population’ perspective , we can define the investment made by H-L+ as a unit of lysine released for L-H+ , and the benefit received by H-L+ as the resultant total rate of hypoxanthine reciprocation from not only the focal L-H+ cells but also its offspring ( Figure 5B ) . This benefit-to-investment ratio is also identical to the exchange ratio , since total hypoxanthine release rate/fmole lysine = ( hypoxanthine release rate per cell * total number of L-H+ cells ) /fmole lysine =rH/ ( fmole lysine/total number of L-H+ cells ) =rH/uL . Why would an individual cooperate – paying a fitness cost to provide a benefit that can aid the reproduction of other individuals ( e . g . sterile ant workers aiding the reproduction of the queen ) ? Social evolution theories offer explanations for the evolution of cooperative traits ( Frank , 1998; Hamilton , 1964; Kerr , 2009; Lehmann and Keller , 2006; Maynard Smith , 1964; Price , 1972; Price , 1970; Queller , 1985; Sachs et al . , 2004; Traulsen and Nowak , 2006; West et al . , 2007 ) . A central concept in social evolution theories is ‘inclusive fitness’ . Inclusive fitness considers the fitness impact from social interactions , and when properly formulated , natural selection leads organisms to become adapted as if to maximize their inclusive fitness ( Grafen , 2006 ) . For example , Hamilton’s rule states that cooperation can evolve as long as rb-c >0 , where c is the fitness cost to the focal cooperator , b is the benefit to focal cooperator's partner , and r is their ‘relatedness’ – the similarity of an actor to its recipient relative to the population ( Damore and Gore , 2012; Fletcher and Doebeli , 2009; Fletcher and Doebeli , 2009; Hamilton , 1964; Queller , 1992; van Veelen , 2009 ) . Note that ‘similarity’ can broadly refer to action type ( e . g . cooperation versus no cooperation ) , even if the actor and the recipient are genetically unrelated as in mutualisms . A mathematically-equivalent , individual-centric version of inclusive fitness ( also known as ‘direct fitness’ ) of a focal individual is the sum of its basal fitness w0 in an asocial environment plus rb benefit received from other cooperators in the social environment minus the cost of cooperation c ( Damore and Gore , 2012; Foster and Wenseleers , 2006; Frank , 1998; West et al . , 2007 ) . If a focal individual’s inclusive fitness is greater than its basal fitness ( w0+rb-c>w0 or rb-c >0 ) , then the individual would grow faster by cooperating than by not cooperating . Inclusive fitness relies on assumptions such as additivity of fitness effects . However , such assumptions are often not valid in microbial communities . For example , fitness often increases in a nonlinear ( e . g . sigmoidal or saturating ) fashion as the benefit increases ( Figure 2; Niehaus et al . , 2019 ) , and therefore inclusive fitness described above is thought to be over-simplified and sometimes even misleading ( Damore and Gore , 2012; Grafen , 2006; Nowak et al . , 2010; van Veelen , 2009 ) . However , despite these criticisms , inclusive fitness has served as a useful conceptual framework for many ( Abbot et al . , 2011; Birch , 2017 ) . In obligatory mutualisms between clonal populations ( i . e . relatedness = 1 ) , we arrive at a physical definition of inclusive fitness ( or direct fitness ) based on exchange ratios and costs of making investments . Specifically , the fitness of H-L+ and L-H+ in monoculture is negative due to death rate and the cost of making investments . When growing in communities , divergent strain ratios rapidly converge to a fixed value ( Momeni et al . , 2013a; Shou et al . , 2007 ) . This means that the two strains must grow at an identical rate which equals to the community growth rate gcomm . In other words , the growth rate of the two strains are identical to each other and to community growth rate ( Hart et al . , 2019a; Materials and methods , 'Community growth rate' ) : ( 1 ) gcomm=- ( dH+cH+dL+cL ) 2+rHrLuLuH+ ( dH+cH-dL-cL ) 24 Here , dH and dL are respectively H-L+ and L-H+’s death rates , cH and cL are respectively H-L+ and L-H+’s fitness costs of overproducing metabolites , rH and rL are respectively hypoxanthine and lysine release rates , and uH and uL are respectively hypoxanthine and lysine utilization amount per birth . Note that rHrLuLuH is the geometric mean of the two exchange ratios rH /uL and rL /uH , and cH and cL represent the fitness costs of making investments for partner . A corollary to Equation 1 is that despite DISOMY14’s better affinity for metabolite ( Figure 2 ) , DISOMY14 and ancestral L-H+ have identical fitness when cocultured with H-L+ in separate CoSMO communities ( Figure 4E ) . This is because in CoSMO , the growth of an individual is limited by partner-supplied metabolites . Thus , there is no point of eating faster if meals arrive slowly . Obviously , if we were to culture DISOMY14 L-H+ , ancestral L-H+ , and H-L+ in a single community in a well-mixed environment , then DISOMY14 should outcompete ancestor due to its better affinity for lysine ( Figure 2 ) . This brings up an important point: fitness metric based on exchange ratio only works for cases where each mutualistic partner is of one genotype . It does not capture changes in genotype frequency within a species . Because of this restriction , it is best suited for sub-communities in a spatially-structured environment . In a spatially-structured environment , if the initial population densities are low , then interactions will mainly take place between one genotype from each species . Individuals in the local sub-community with the highest gcomm are predicted to grow the fastest . In the case of spatial CoSMO , would an H-L+ cell grow faster if it had landed next to an ancestral or a DISOMY14 L-H+ cell ? During the initial encounter , H-L+ next to DISOMY14 will benefit more from DISOMY14’s faster release rate ( Figure 5A ) . However , after this initial stage , DISOMY14 and ancestor are identical due to identical exchange ratio ( Figure 5B , i and ii ) . Indeed , H-L+ grew equally fast when co-cultured with ancestral or DISOMY14 L-H+ ( Figure 4E ) . Biological market theory posits that the exchange of goods among organisms can be analyzed in market terms , where individuals attempt to maximize their gains ( Noë and Hammerstein , 1994; Noë and Hammerstein , 1995; Werner et al . , 2014 ) . For example , a male insect that offers more nuptial gifts to a female is regarded by the female as being more partner ( female ) -serving and is thus chosen by the female ( Noë and Hammerstein , 1994; Noë and Hammerstein , 1995; Werner et al . , 2014 ) . Although biological market theory does not apply here since our yeast strains lack partner choice capability , our work can prove useful for other systems . For example , consider the legume-rhizobia mutualism where a legume host provides photosynthates to rhizobia while rhizobia reciprocate fixed nitrogen . Application of biological market theory leads to statements such as "…whether plant hosts can detect variation in resources or services provided ( by rhizobia ) and respond accordingly . Such discrimination mechanisms have been found in legumes , with some species preferentially supporting rhizobial symbionts that provide more fixed N2 for hosts" ( Werner et al . , 2014 ) . Since legume-rhizobia mutualisms last over multiple generations of rhizobia , we suggest that ‘fixed N2’ should be quantified in terms of exchange ratio: a focal rhizobium’s nitrogen release rate normalized by photosynthate utilized to make the rhizobium , or total nitrogen release rate per photosynthate utilized . Obviously , when comparing non-nitrogen fixers with nitrogen fixers , fixers are more mutualistic than non-fixers . But when comparing quantitative variants of nitrogen fixers , this distinction could become important . By explicitly quantifying the exchanged goods , exchange ratio captures the spirit of market economy . In summary , when considering microbial mutualisms that span multiple generations , a focal individual’s partner-serving phenotype can be quantified as the exchange ratio of benefit production rate by the focal individual for partner divided by benefit utilized to make the focal individual . This is equivalent to total benefit production rate for partner per intake benefit . By the same token , an individual with reduced benefit production rate may not be a cheater if the individual also utilizes less benefit from the partner . For a lay audience behind-the-scenes story , see 'Foresight , hindsight , insight , and the blur in-between' ( Supplementary file 1; https://medium . com/@wenying . shou/foresight-hindsight-insight-and-the-blur-in-between-9dd0505b2918 ) .
Genetic manipulations and growth medium for the yeast S . cerevisiae are explained in Guthrie and Fink ( 1991 ) . Protocols and technical details that we have used can be found in Waite and Shou ( 2014 ) . Briefly , we used autoclaved rich medium YPD ( 10 g/L yeast extract , 20 g/L peptone , 20 g/L glucose ) in 2% agar plates for isolating single colonies . Saturated YPD overnight liquid cultures from these colonies were then used as inocula to grow exponential cultures . YPD overnight cultures were stored at room temperature for no more than 4 ~ 5 days prior to experiments . We used defined minimal medium SD ( 6 . 7 g/L Difco yeast nitrogen base with ammonium sulfate without amino acids , 20 g/L glucose ) for all experiments , with supplemental metabolites as noted ( Guthrie and Fink , 1991 ) . To achieve higher reproducibility , we sterilized SD media by filtering through 0 . 22 µm filters . To make SD plates , we autoclaved 20 g/L Bacto agar or agarose in H2O , and after autoclaving , supplemented equal volume of sterile-filtered 2XSD . The ancestral strain was WY1335 , described in detail in Hart et al . ( 2019a ) . All strains are in Supplementary file 2 . The DISOMY14 strain we used for analysis WY2261 ( refrozen as WY2348 and WY2349 ) was obtained in the following manner . The evolved strain WY1584 was back-crossed twice into the ancestral background to get rid of mutations in genes ECM21 and YPL247C . The first cross with WY1521 resulted in ‘38-1D’ , which was then crossed with WY1335 to result in WY2261 ( ‘E2’ ) . To genotype spores , we PCR amplified the mutated regions in ECM21 and YPL247C , and subjected the purified PCR product to Sanger sequencing . For those spores that contained no mutations in ECM21 and YPL247C , we subjected them to restriction-site associated DNA sequencing ( Materials and methods , 'RADseq' ) to determine ploidy . When we modified our sequence analysis pipeline , we realized that WY2261 contained other mutations ( Supplementary file 3 ) . However , the presence of other mutations does not affect our conclusions , since integrating an extra copy of LYP1 or WHI3 into the ancestral background respectively increased growth rate under lysine limitation ( Figure 2 ) and per cell hypoxanthine release rate ( Figure 4A ) . L-H+ ( WY1335 ) and H-L+ ( WY1340 ) were grown separately to exponential phase in minimal SD medium supplemented with lysine ( 164 . 3 μM ) or adenine sulfate ( 108 . 6 μM ) , respectively ( Guthrie and Fink , 1991 ) . Cells were washed free of supplements , counted using a Coulter counter , and mixed at 1000:1 ( Line A ) , 1:1 ( Line B ) , or 1:1000 ( Line C ) at a total density of 5 × 105/ml . The different initio ratios did not noticeably affect evolutionary outcomes . Three 3 ml community replicates ( replicates 1 , 2 , and 3 ) per initial ratio were initiated . Communities were grown at 30°C in glass tubes on a rotator to ensure well-mixing . Community turbidity was tracked by measuring the optical density ( OD600 ) in a spectrophotometer once to twice every day . In this study , 1 OD was found to be 2 ~ 4×107 cells/ml . We diluted communities periodically to maintain OD at below 0 . 5 to avoid additional selections due to limitations of nutrients other than hypoxanthine or lysine . The fold-dilution was controlled to within 10 ~ 20 folds to minimize severe population bottlenecks . Note that no mutagens were used during evolution . Coculture generation was calculated from accumulative population density by multiplying OD with total fold-dilutions . For each coculture at every 10 ~ 20 generations , cell pellet of ~1 ml coculture was resuspended in 1 ml rich medium YPD ( Guthrie and Fink , 1991 ) +10% trehalose , cooled at 4°C for several hours , and frozen at −80°C . Cells frozen this way revived much better than if frozen in SD medium supplemented with a final of 15% glycerol . CoSMO could engage in self-sustained growth only if its initial total cell density was sufficiently high ( Shou et al . , 2007 ) . Thus , to revive a coculture , ~20 μl was scooped from the frozen stock using a sterile metal spatula , diluted ~10 fold into SD , and allowed to grow to moderate turbidity . The coculture was further expanded by adding 3 ml of SD . To isolate clones , cocultures were plated on rich medium YPD , and clones from the two strains were distinguished by their fluorescence colors or drug resistance markers . The detailed protocol of Gibson assembly ( Gibson et al . , 2009 ) for assembling DNA fragments with end homology was obtained from Eric Klavins lab ( University of Washington ) . 1 ml 5xISO buffer: 1M Tris-HCl ( pH 7 . 5 ) 500 µl; 2M MgCl2 25 µl; 100 mM dGTP , dATP , dCTP , and dTTP 10 µl each ( total 40 µl ) ; 1M DTT 50 µl; 100 mM NAD 50 µl; PEG-8000 0 . 25 g; H2O: 145 µl . 5xISO was frozen in 100 µl aliquots at −20°C . The assembly master mix ( 375 µl total ) included: H2O 216 . 75 µl; 5XISO Buffer 100 µl; 1 U/µl T5 Exonuclease 2 µl; 2 U/µl Phusion Polymerase 6 . 25 µl; 40 U/µl Taq DNA Ligase 50 µl . 15 μl aliquots were stored at −20°C . This master mix is ideal for DNA molecules with 20 ~ 150 bp overlapping homology . To carry out the assembly reaction , 15 μl assembly master mix is mixed with a total of 5 μl DNA ( e . g . 125 nM ) , and incubated at 50°C for 1 hr . We constructed plasmid WSB175 to contain G418 resistance ( KanMX ) and a telomeric sequence . Briefly , WSB174 ( from Dan Gottschling lab ) containing a URA3 marker was digested with HindIII and BamHI to remove the URA3 marker and yield the vector backbone ( 4 . 5 kb ) . The KanMX gene was amplified from WSB26 using primers WSO433 and WSO434 , each containing a 25 bp overhang homologous to the vector backbone . The vector backbone and the KanMX PCR product were circularized via Gibson assembly to yield WSB175 . To perform Gibson assembly , we used 5 µl DNA ( including 55 ng vector and 0 . 5 µl of 125 nM insert ) . Gibson mixture was transformed into E . coli , and DNA was extracted from several colonies . DNA was checked via restriction digestion with HindIII and BamHI . To carry out chromosome truncation , we PCR amplified ~600 bp fragments from various genomic locations on Chromosome 14 ( Figure 3B ) . The PCR reaction consisted of: genomic DNA ( 0 . 5 µl out of 25 µl where 0 . 3 ml overnight yeast culture was harvested and DNA extracted ) , 10xPCR buffer ( 5 µl ) , 25 mM MgCl2 ( 3 µl ) , 10 mM dNTP mix ( 1 µl ) , 50 µM forward primer ( 0 . 5 µl ) , 50 µM reverse primer ( 0 . 5 µl ) , Taq polymerase ( 0 . 5 µl ) , H2O ( 39 µl ) . Cycling conditions are as following: 94°C 3 min; [94°C 30 s +56 . 9°C 60 s + 72°C 60 s] x30 cycles; 72°C 10 min . We used Qubit ( Thermofisher ) to quantify DNA concentration . We then assembled PCR fragment with vector backbone containing KanMX and telomere ( 2 . 5 µl of each at 125 nM , corresponding to ~100 ng insert and 350 ng vector ) to yield the assembled DNA . The assembled DNA ( 100 ng ) was PCR amplified again . PCR reaction contained: Gibson product ( 4 . 1 µl ) , 5xPhusion HF buffer ( 10 µl ) , 10 mM dNTP mix ( 1 µl ) , 50 µM forward primer ( 1 µl ) , 50 µM reverse primer ( WSO157 , 1 µl ) , Phusion polymerase ( 0 . 5 µl ) , H2O ( 32 . 4 µl ) . Cycling conditions are as following: 98°C 30 s; [98°C 10 s +59 . 9°C 30 s + 72°C 90 s] x30 cycles; 72°C 10 min . Expected length was ~2 . 4 kb . The PCR product was used to transform the DISOMY14 yeast strain ( WY2261 ) using lithium acetate yeast transformation ( Waite and Shou , 2014 ) . Transformants were selected on rich medium supplemented with G418 ( YPD + G418 ) plate . Transformants were screened for correct integration using PCR amplification across the chromosome integration site ( one primer homologous to the genome , and the other primer homologous to KanMX ) . To introduce an extra copy of LYP1 in the ste3::HygMX locus of the ancestor , we assembled and transformed the following . We amplified a 527 bp homology region upstream of STE3 , the LYP1 gene ( including 333 bp upstream and 466 bp downstream of ORF ) , and the KanMX resistance cassette ( loxP-TEFp-KanMX-TEFt-loxP ) from WSB118 , all using PCR . Note that the KanMX resistance cassette contains 3' homology to the ste3::HygMX locus . Primers used contained 20 bp homology when appropriate to allow us to compile these sequences in the order listed using Gibson assembly . We then amplified this assembly further using PCR and transformed this 4 . 8 kB product into the ancestor ( WY1335 ) , screening for successful integration by G418 resistance , loss of hygromycin resistance , and checking PCR . We obtained WY2254 ~2255 . To knockout duplicated LYP1 from DISOMY14 , we amplified a 532 bp region upstream of LYP1 , and a 739 bp region downstream of LYP1 . We also amplified KanMX resistance . We assembled the three pieces via Gibson assembly ( LYP1 upstream , KanMX , LYP1 downstream ) , PCR amplified the assembled molecule , and transformed DISOMY14 cells with the PCR product . Transformants were plated on YPD + G418 plate , and colonies were screened for correct integration via PCR . We obtained WY2262 ~2263 . We used a similar methodology to introduce an extra copy of WHI3 in the ste3::HygMX locus of the ancestor ( WY1335 ) . We first digested WSB185 ( pBSKII ) with XmaI and HindIII ( HF ) , yielding 2 . 9 kb backbone . We amplified a 428 bp homology region upstream of STE3 , the WHI3 gene ( including 570 bp upstream and 700 bp downstream of ORF;~3 . 3 kb ) , and KanMX resistance ( 1 . 6 kb ) , and assembled all these with the vector backbone . The Gibson product was used to transform E . coli , and colonies were mini-preped and screened via restriction digest with Kpn1 and BamH1 . The correct liberated fragment ( 5 . 4 kb ) was transformed into ancestral cells ( WY1335 ) and plated on YPD + G418 plate . Transformants were checked for loss of hygromycin resistance and via PCR . We obtained WY2357 ~2359 . To delete WHI3 from DISOMY14 , we amplified a 528 bp region upstream of WHI3 , and a 489 bp region downstream of WHI3 . We assembled the two pieces with KanMX cassette via Gibson assembly , PCR amplified the assembled molecule , and transformed DISOMY14 cells with the PCR product . Transformants were plated on YPD + G418 plate , and colonies were screened for correct integration via PCR . We obtained WY2350 ~2352 . For whole-genome sequencing via tagmentation , we extracted yeast genomic DNA using QIAGEN Genomic-tip 20G ( Cat . No . 10223 ) , YeaStar Genomic DNA kit ( Zymo Research ) , or a protocol modified from Sergey Kryazhimskiy and Andrew Murray lab . DNA from the last protocol is suitable for tagmentation but not for RADseq . To extract DNA , we used the following procedure: "Spin down cells ( 0 . 5 ml saturated culture; microfuge at highest speed in 2 ml v-bottom tubes for 2’ ) . Thoroughly discard supernatant . To the pellet add 252 µl Digestion mix ( 50 µl of 0 . 5M EDTA , pH = 7 . 5 or 8 . 0; 200 µl of ddH20; 2 . 5 µl of Zymolyase ( stock: 5 U/ul ) ) . Mix by inversion and incubate at 37°C for 24 hr on rotator . Add 50 µl of miniprep mix ( 0 . 2M EDTA , 0 . 4M Tris , 2% SDS , pH = 8 . 0 ) . Mix by inversion and incubate at 65°C for 30 min . Add 63 µl ( about ⅕ of volume ) of 5M KAc . Mix by inversion and incubate on ice for 30 min . Add 250 µl of chloroform , vortex vigorously for 1 min . This helps to precipitate proteins and lipids . Spin down sample for 10 min on max speed ( tabletop centrifuge ) . The DNA is in the supernatant . Transfer supernatant to a new tube ( max 300 μl ) . Add 720 µl ( >2 x volume ) of 100% EtOH . Mix by inversion . At this point you should see the DNA clots in your tube . Spin down on max speed for 20 min . The DNA is now in the pellet . Thoroughly discard supernatant . Add 50 µl H20 +1 µl of RNAase A ( 10 mg/ml ) to undried pellet . Allow DNA to resuspend by incubating at 37°C for 1 hr . Add 2 µl of Proteinase K ( 20 mg/ml ) . Incubate for 2 hr at 37°C . Add 130 µl of 100% isopropanol . Mix by inversion and spin down at max speed for 10 min . The DNA is in the pellet . Thoroughly discard supernatant . Add 500 µl of 70% EtOH . Mix by inversion and spin down at max speed for 10 min . The DNA is in the pellet . Discard supernatant . Allow pellet to air dry overnight . Resuspend pellet in 100 µl of 10 mM Tris , pH = 8 . 0 . " For RADseq , yeast genomic DNA was extracted from 2 × 108 ~ 109 cells using , for example , the DNeasy Blood and Tissue Kit ( Qiagen ) . High-quality DNA is required for optimal restriction endonuclease digestion and is of utmost importance for the overall success of the protocol . The samples were treated with RNase A following manufacturer’s instructions to remove residual RNA , and then quantified using Qubit . The optimal concentration after elution is 25 ng/μl or greater . The whole genome sequencing protocol was slightly modified from that of Sergey Kryazhimskiy v . 2 . 1 ( 2013-06-06 ) ( Kryazhimskiy et al . , 2014 ) . Indexing primer design followed ( Adey et al . , 2010 ) ( Supplementary file 4 ) . For an illustration of Nextera V2 Illumina sequencing molecular biology , see Figure 2—figure supplement 2 . To tagment genomic DNA , we prepared gDNA at concentration at or below 2 . 5 ng/µl . For n samples = r rows and c columns , make the Tagmentation Master Mix ( TMM ) by mixing n x 1 . 06 × 1 . 25 µl of TD Buffer ( Tagment DNA Buffer ) and n x 1 . 06 × 0 . 25 µl of TDE1 ( Tagment DNA Enzyme ) in a PCR tube . Mix thoroughly by gently pipetting the mixture up and down 20 times . Distribute TMM into r tubes ( or a PCR strip ) , c x 1 . 03 × 1 . 5 µl into each tube . With a multichannel pipette , distribute TMM into all wells of a fresh plate ( ‘tagmentation plate’ ) , 1 . 5 µl per well . With a multichannel pipette , transfer 1 µl of gDNA into the tagmentation plate ( total volume = 2 . 5 µl per well ) . Mix by gently pipetting up and down 10 times . Cover plate with Microseal ‘B’ ( Biorad , MSB-1001 ) . Give the plate a quick spin to collect all liquid at the bottom ( Sorvall or Allegra centrifuges , 1000 rpm for 1 min ) . Place the plate in the thermocycler and run the following program: 55°C for 5 min; hold at 10°C . Next , PCR amplification is performed to add the index adaptors to tagmented DNA . Make the adaptor PCR reaction final volume to be 7 . 5 µl ( 2 . 5 µl of tagmented DNA from above , 3 . 75 µl of 2x KAPA master mix ( KAPA amplification kit KK2611/KK2612 ) , 0 . 625 µl of Index Adapter 1 , 0 . 625 µl of Index Adapter 2 ) . For convenience , we have pre-mixed index primers where index adapter one is always the same ( NexV2ad1noBC; Supplementary file 4 ) and index adapter two is one of the 96 Index adaptors ( NexV2ad2**; Supplementary file 4 ) , and each is at 5 μM in H2O . Mix the entire mix by gently pipetting up and down 10 times . Cover plate with Microseal ‘A’ ( Biorad , MSB-5001 ) . Make sure to press well on each well , especially edge wells . Give the plate a quick spin to collect all liquid at the bottom at 1000 rpm for 1 min . Place the tubes in the thermocycler and run the following program: 72°C for 3 min; 98°C for 2:45 min; [98°C for 15 s; 62°C for 30 s; 72°C for 1:30 min]x8; Hold at 4°C . Ensure that the lid is tight and that it is heated during incubation . Make Reconditioning PCR Master Mix ( RMM ) by mixing n*1 . 06*8 . 5 µl of 2xKAPA polymerase mix , n*1 . 06*0 . 5 µl of primer P1 ( 10 µM; WSO380 AATGATACGGCGACCACCGA ) , and n*1 . 06*0 . 5 µl of primer P2 ( 10 µM; WSO381 CAAGCAGAAGACGGCATACGA ) . Mix thoroughly by gently pipetting the mixture up and down 20 times . With a multichannel pipette , transfer 9 . 5 µl of RMM into each well of the plate ( final PCR volume 17 µl ) . Mix by gently pipetting up and down 10 times . Cover plate with Microseal ‘A’ . Give the plate a quick spin to collect all liquid at the bottom at 1000 rpm for 1 min . Place the tubes in the thermocycler and run the following program: 95°C for 5 min; [98°C for 20 s; 62°C for 20 s; 72°C for 30 s]x4; 72°C for 2 min; Hold at 4°C . PCR clean-up used magnetic beads . Centrifuge the plate to collect all liquid ( 1000 rpm for 1 min ) . Vortex AMPure XP beads ( Beckman Coulter A63880 ) for 30 s to ensure that they are evenly dispersed . Transfer c x 1 . 05 × 1 x V µl ( V = PCR volume = 17 µl ) of beads into r PCR tubes or a PCR strip . Using a multichannel pipette , transfer V µl of beads into each well containing the PCR product . Mix well by gently pipetting up and down 20 times . The color of the mixture should appear homogeneous after mixing . Incubate at room temperature for 5 min so that DNA is captured by beads . Place the plate on the magnetic stand ( Life Technologies , Cat . #123-31D ) and incubate for about 1 min to separate beads from solution . Wait for the solution to become clear . While the plate is on the magnetic stand , aspirate clear solution from the plate and discard . Do not disturb the beads . If beads are accidentally pipetted , resuspend them back , wait for the solution to clear up , and repeat . While the plate is on the magnetic stand , dispense 200 µl of 70% ethanol into each well and incubate for 30 s at room temperature . Aspirate out ethanol without disturbing the beads and discard . Repeat for a total of 2 washes . Remove the remaining ethanol with P10 pipette . Let the plate air dry for approximately 5 min . Do not overdry the beads . Take the plate off the magnetic stand . Add 33 µl of 10 mM Tris-HCl ( pH 8 ) to each well of the plate . Carefully resuspend the beads by mixing 10–15 times . Incubate for 2 min at room temperature . DNA is now in the solution . Place the plate back onto the magnetic stand and incubate for about 1 min to separate beads from solution . Wait for the solution to become clear . While the plate is on the magnetic stand , aspirate clear solution from the plate and transfer to a fresh plate . Do not disturb the beads . If beads are accidentally pipetted , resuspend them back , wait for the solution to clear up , and repeat . Qubit quantify samples using 1 µl of eluate . We get about 6 ng/µl . Send 3 µl for High Sensivitity tapestation ( 75–1000 pg/µl ) to get average length . If Qubit reading was >0 . 38 ng/µl , then sequencing would generally work . Pool samples at equal molarity , with the final pool ideally being at least 2 nM ( although 1 nM seemed fine as well ) . If we have 100 indexes , then each sample needs to be diluted to 0 . 02 nM . Qubit the pooled sample and submit 30 µl at 2 nM for sequencing on Illumina HiSeq 2000 ( paired-end; 50 ~ 150 cycles; Nextera sequencing primers ) . RADseq ( restriction site-associated DNA sequencing ) protocol was obtained from Aimee Dudley lab based on Etter et al . ( 2012 ) . The design scheme is in Figure 3—figure supplement 4 , and primer sequences are in Supplementary file 5 . Briefly , genomic DNA was digested with the six-cutter Mfe1 and the four-cutter Mbo1 ( see below ) . Our desired DNA fragment would be flanked by Mfe1 and Mbo1 sites . To the digested DNA , we ligate annealed primers ( P1 top annealed with P1 bottom containing a 4 bp barcode and Mfe1 overhang , and P2 top annealed with P2 bottom containing a 6 bp barcode and Mbo1 overhang ) . The dual barcode system allows many samples ( e . g . 900 samples ) to be sequenced simultaneously . P1 also contains Illumina Read One sequencing primer which will read 4 bp barcode and genomic DNA adjacent to the Mfe1 site , and P2 also contains Illumina Read Two sequencing primer and Index sequencing primer which will read 6 bp barcode and genomic DNA adjacent to the Mbo1 site . The ligation product can be PCR amplified using WSO381 and WSO398 ( Figure 3—figure supplement 4 ) . P2 top primer has a stretch of sequences ( lower case ) that does not anneal with P2 bottom . Thus , in PCR round 1 , only WSO398 is effective . In PCR round two or later , both WSO398 and WSO381 are effective . Note that the non-annealing sequence is designed for the following reason: Because most genomic DNA fragments are flanked on both sides by the 4-cutter Mbo1 site and these fragments would ligate with P2 on both sides , the non-annealing DNA segment ( lower case ) ensures that fragments lacking P1 cannot be amplified with WSO381 . To anneal P1 and P2 , 100 µM stock plates of P1 ( top and bottom ) and P2 ( top and bottom ) primers were obtained from the Dudley Lab at the Pacific Northwest Research Institute . The P1 bottom primer and the P2 top primer include a 5’-phosphate modification required by ligation . For the P1 annealing reaction , the following was mixed: 10 µl 5M NaCl , 100 µl 1M Tris ( pH 8 . 0 ) , 888 µl H2O , and 1 µl each 100 µM P1 top and P1 bottom primers ( final concentration 100 nM ) . For the P2 annealing reaction , the following was mixed: 7 µl 5M NaCl , 70 µl 1M Tris ( pH 8 . 0 ) , 483 µl H2O , and 70 µl each 100 µM P1 top and P1 bottom primers ( final concentration 10 µM ) . The P1 and P2 top/bottom primer mixtures were aliquoted into PCR tubes at 100 µl per tube . The samples were heated at 95°C for 1 min , then allowed to cool to 4°C ( at a rate of 0 . 1 °C/sec ) to allow annealing of the top and bottom primers . After cooling , the tubes were spun down and immediately put on ice . The P1 aliquots were consolidated into a single tube , and P2 aliquots were consolidated into a single tube . The P1 and P2 annealed primers were combined as follows . 700 µl of P2 was diluted into 4 . 55 ml of H2O ( diluted concentration at 1 . 33 µM ) . To 48 wells of a 96 well plate , 83 µl of this diluted annealed P2 was added to each well . 27 . 5 µl of the annealed P1 primers was added to each well . The final P1 +P2 vol is 110 . 5 µl ( P2 final concentration 1 µM; P1 final concentration of 25 nM ) . It is very important to keep adapters cold after annealing . Specifically , keep at 4°C , mix on ice , thaw on ice after retrieving from −20°C storage . RNase-treated yeast genomic DNA was subjected to restriction digestion by Mfe1 and Mbo1 ( NEB ) . Per reaction , 5 . 25 µl of the purified genomic DNA ( ~125 ng , Qubit quantified ) was mixed with 0 . 625 µl 10X NEB Buffer 4 ( or CutSmart buffer ) , 0 . 25 µl MboI ( 2 . 5 units ) , and 0 . 125 µl MfeI-HF ( 2 . 5 units ) in PCR tubes ( final volume of 6 . 25 ul ) . The mixture was set up on ice and incubated at 37°C for 1 hr , and was heat inactivated by incubating at 65°C for 20 min . Custom adapters were annealed to the digested DNA using NEB T4 DNA ligase . Specifically , the following components were added to a tube in this order: 2 . 5 µl of the combined annealed P1 ( 25 nM ) +P2 ( 1 µM ) adaptor mix , the entire 6 . 25 µl digestion reaction , and 3 . 75 µl of a T4 ligase master mix ( 0 . 1 µl T4 ligase at 2000 units/µl , 1 . 25 µl T4 ligation buffer , and 2 . 4 µl H2O ) . The reaction was combined on ice and mixed by pipetting up and down . Ligation was carried out at room temperature for 20 min and heat inactivated in thermocycler at 65°C for 20 min . The samples were allowed to cool slowly to room temperature ( 30 min ) . Every 24 samples were pooled together and concentrated using the QIAGEN MinElute PCR purification kit , and the MinElute column was eluted in 10 µl EB . The concentrated samples were subjected to gel electrophoresis using 2% low range ultra agarose gel ( 48 samples/lane ) . Sufficient band separation is achieved when the loading dye is approximately halfway down the gel , which ensures that the 150 band is separated from the 100 bp primer dimer . The size range 150 to 500 bp was excised out of the gel under a long-wavelength UV lamp and purified using the QIAGEN MinElute Gel Extraction Kit , eluting with 20 µl EB . The gel extracted DNA was amplified using the NEBNext PCR Master Mix ( NEB#M0541S ) with custom primers ( WSO398 and WSO381; Supplementary file 5 ) . The mixture contained: 25 µl NEBNext PCR Master Mix; 1 µl WSO398 ( 10 µM ) ; 1 µl WSO381 ( 10 µM ) ; 1 µl DNA ( 5 ~ 10 ng ) , and 22 µl H2O ( total 50 µl ) . PCR cycling was: 98°C ( 1 min ) ; [98°C ( 10 s ) +60°C ( 30 s ) +72°C ( 30 s ) ]x14 cycles + 72°C ( 4 min ) +4°C hold . The PCR reaction was cleaned and concentrated using the QIAquick PCR Purification Kit ( QIAGEN ) , eluting in 30 µl H2O . The expected concentration is ~30–40 ng/µl . The library quality ( fragment size distribution ) was ascertained using Tapestation . The resultant DNA was subjected to paired-end 25 cycles on Illumina HiSeq 2000 using TruSeq Dual Index Sequencing Primers . To analyze whole genome sequencing , a custom Perl script incorporating bwa ( Li and Durbin , 2009 ) and SAMtools ( Li et al . , 2009 ) written by Robin Green was used to align paired-end reads to the S . cerevisiae RM-11 reference genome . Mutations were identified via GATK for single-nucleotide variants and indels , and cn . MOPs for local copy-number variant calling . A custom Perl script incorporating vcftools was used to automate comparison between ancestral versus evolved strains . All genetic changes were visually inspected using the Integrated Genome Viewer ( IGV ) environment for quality inspection and validation . Ploidy was calculated using custom python and R scripts wherein read depth was counted for each base . These read depths were averaged within successive 1000 bp windows; each window average is normalized by the median of all window averages across the genome . The normalized values for each window are log2 transformed and plotted versus the respective genomic position ( chromosome/supercontig ) for ease in graphical inspection of ploidy changes . Sequence analysis code can be publicly accessed at https://github . com/robingreen525/ShouLab_NGS_CloneSeq ( Green , 2019; copy archived at https://github . com/elifesciences-publications/ShouLab_NGS_CloneSeq ) . RADseq analysis was performed as described in Etter et al . ( 2012 ) using custom python and R scripts . Samples were split by their respective barcode and aligned to the RM11 reference genome using bwa . Up to six mismatches were allowed per read/marker . Next , reads with Phred quality scores below 20 or with a median coverage of less than two per sample were discarded . To ensure that each respective marker was representative of a properly digested MfeI-MboI , the expected length of each fragment based on a theoretical digestion of the RM11 genome was compared to the length of the actual marker as determined by read alignment to that marker ( i . e finding reads that fell within the expected coordinates of a MfeI-MboI digest product ) . Next , for each marker , the proportion of reads aligning to that marker was normalized against total read alignment to the genome . To ensure that only high quality markers were used , the CV of each marker across all tested strains were analyzed and markers with a CV of >= 0 . 6 were discarded . Additionally , only markers that were within the expected gel cut size of >125 bp and <400 bp were used . This still allowed >2000 markers to be used for downstream analysis . To assess ploidy for RADseq , the same analysis was performed on a panel of 10 euploid strains . For each strain and for each marker , the relative proportion of that maker of the total reads for the strain of interest was compared against the median proportion of the total reads for the euploid panel . A supercontig ( the RM11 assembly does not have full chromosomes but supercontigs ) was called as duplicated if the average proportion of all makers on that supercontig in the backcrossed strain was 2-fold greater than the euploid panel . All disomy 14 calls for a tetrad segregated 2:2 as expected . This method has been described in Hart et al . ( 2019b ) . Briefly , to assay for self-serving phenotype of an L-H+ mutant , we diluted a saturated overnight 1:6000 into SD +164 µM lysine , and allowed cultures to grow overnight at 30°C to exponential phase . We washed cells 3x with SD , starved them for 4–6 hr to deplete vacuolar lysine stores , and diluted each culture so that a 50 µl spot had several hundred cells . We spotted 50 µl on SD plate supplemented with 1 . 5 µM lysine ( 10 spots/plate ) , and allowed these plates to grow overnight . When observed under a 10x objective microscope , evolved cells with increased lysine affinity would grow into ‘microcolonies’ of ~20 ~ 100 cells , while the ancestral genotype would fail to grow ( Figure 2—figure supplement 1 ) . DISOMY14 exhibited an intermediate phenotype where smaller microcolonies with variable sizes formed . Beads ( ThermoFisher Cat R0300 , 3 μm red fluorescent beads ) were autoclaved in a factory-clean glass tube , diluted into sterile 0 . 9% NaCl , and supplemented with sterile-filtered Triton X-100 to a final 0 . 05% ( to prevent bead clumping ) . The mixture was sonicated to eliminate bead clusters and was kept at 4°C in constant rotation to prevent settling and re-clumping . Bead density was quantified via hemacytometer and Coulter counting ( 4−8 × 106 beads/ml final ) . The prepared bead mixture served as a density standard . Culture samples of interest were diluted to OD 0 . 01 ~ 0 . 1 ( 7 × 105 - 7 × 106 cells/ml ) in filtered water . Bead-cell mixtures were prepared by mixing 90 ul of the diluted culture sample , 10 µl of the bead stock , and 2 µl of 1 µM ToPro 3 ( Molecular Probes T-3605 ) , a nucleic acid dye that only permeates cell membranes of dead cells . Triplicate cell-bead mixtures were prepared for each culture in a 96-well format for high-throughput processing . Flow cytometry of the samples was performed on Cytek DxP Cytometer equipped with four lasers , ten detectors , and an autosampler . GFP ( H-L+ ) , mCherry ( L-H+ ) , and ToPro ( dead cells ) are respectively detected by 50 mW 488 nm laser with 505/10 ( i . e . , 500–515 nm ) detector , 75 mW 561 nm Laser with 615/25 detector , and 25 mW 637 nm laser with 660/20 detector . Each sample was individually analyzed using FlowJo software to identify the number of beads , dead cells , and live fluorescent cells . Live and dead cell densities were calculated from the respective cell:bead ratios , corrected for the initial culture dilution factor . The mean cell density from triplicate measurements was used ( coefficient of variation within 10% ) . See Hart et al . ( 2019b ) for details on microscopy and experimental setup , method validation , and data analysis . Briefly , cells were diluted to low densities to minimize metabolite depletion during measurements . Dilutions were estimated from culture OD measurement to result in 1000 ~ 5000 cells inoculated in 300 µl SD medium supplemented with different metabolite concentrations in wells of a transparent flat-bottom microtiter plate ( e . g . Costar 3370 ) . We filled the outermost wells with water to reduce evaporation . Microtiter plates were imaged periodically ( every 0 . 5 ~ 2 hr ) under a 10x objective in a Nikon Eclipse TE-2000U inverted fluorescence microscope . For each well , four adjacent positions were imaged . The microscope was connected to a cooled CCD camera for fluorescence and transmitted light imaging . The microscope was enclosed in a temperature-controlled chamber set to 30°C . The microscope was equipped with motorized stages to allow z-autofocusing and systematic xy-scanning of locations in microplate wells , as well as motorized switchable filter cubes capable of detecting a variety of fluorophores . Image acquisition was done with an in-house LabVIEW program , incorporating bright-field autofocusing ( Hart et al . , 2019b ) and automatic exposure adjustment during fluorescence imaging to avoid saturation . Condensation on the plate lid sometimes interfered with autofocusing . Thus , we added a transparent ‘lid warmer’ on top of our plate lid ( Hart et al . , 2019b ) , and set it to be 0 . 5°C warmer than the plate bottom , which eliminated condensation . We used an ET DsRed filter cube ( Exciter: ET545/30x , Emitter: ET620/60 m , Dichroic: T570LP ) for mCherry-expressing strains . Time-lapse images were analyzed using an ImageJ plugin Bioact ( Hart et al . , 2019b ) . Bioact measured the total fluorescence intensity of all cells in an image frame after subtracting the background fluorescence from the total fluorescence . A script plotted background-subtracted fluorescence intensity over time for each well to allow visual inspection . If the dynamics of four positions looked similar , we randomly selected one to inspect . In rare occasions , all four positions were out-of-focus and were not used . In a small subset of experiments , a discontinuous jump in data appeared in all four positions for unknown reasons . We did not calculate rates across the jump . Occasionally , one or two positions deviated from the rest . This could be due to a number of reasons , including shift of focal plane , shift of field of view , black dust particles , or bright dust spots in the field of view . The outlier positions were excluded after inspecting the images for probable causes . If the dynamics of four positions differed due to cell growth heterogeneity at low concentrations of metabolites , all positions were retained . We normalized total intensity against that at time zero for each position , and then averaged across positions . We calculated growth rate over three to four consecutive time points , and plotted the maximal net growth rate against metabolite concentration . If maximal growth rate occurred at the end of an experiment , then the experimental duration was too short and data were not used . For L-H+ , the initial stage ( 3 ~ 4 hr ) residual growth was excluded from analysis since residual growth was supported by vacuolar lysine storage . 75 µl sample filtered through a 0 . 2 µm filter was mixed with an equal volume of a master mix containing 2xSD ( to provide fresh medium ) as well as tester cells auxotrophic for the metabolite of interest ( ~1×104 cells/ml , WY1340 over-night culture ) in a flat-bottom 96-well plate . We then wrapped the plate with parafilm and allowed cells to grow to saturation at 30°C for 48 hr . We re-suspended cells using a Thermo Scientific Teleshake ( setting #5 for ~1 min ) and read culture turbidity using a BioTek Synergy MX plate reader . Within each assay , SD supplemented with various known concentrations of metabolite were used to establish a standard curve that related metabolite concentration to final turbidity ( e . g . Figure 3—figure supplement 5 ) . From this standard curve , the metabolite concentration of an unknown sample could be inferred . Detailed description of the release assay during lysine starvation can be found in Hart et al . ( 2019a ) . Briefly , L-H+ strain was pre-grown in synthetic minimal media SD supplemented with high lysine ( 164 µM ) to exponential phase . The cultures were washed in lysine-free media and allowed to starve for 2 hr at 30°C to deplete vacuolar lysine stores . Following starvation , the culture was periodically sampled ( approximately every 6 hr for 24 hr ) upon which live/dead cell densities were measured via flow cytometry ( Materials and methods , ‘Flow cytometry’ ) , and culture samples were sterile filtered and supernatants were frozen . The supernatants were subjected to bioassay to measure hypoxanthine concentrations ( Materials and methods , ‘Bioassay’ ) . Hypoxanthine release rate can be inferred by the slope of the linear function relating integrated live cell density over time ( cells/ml*hr ) versus measured hypoxanthine concentration ( µM ) . For an example , see Figure 3—figure supplement 1 . To increase the throughput of this assay for screening chromosome truncation mutants ( Figure 3B ) , we made the following modifications . After measuring time zero cell densities by flow cytometry , we loaded 200 µL of OD ~0 . 05 cells in SD per well and tracked fluorescence every 2 hr using automated 96-well plate fluorescence microscopy imaging ( Hart et al . , 2019b ) . The rest of each culture was treated as in the normal assay for sterile filtering at each sampling . Plate preparation , imaging , and images analysis were done as described in Materials and methods , ‘Microscopy growth assay’ . Fluorescence scales with live cell density ( Hart et al . , 2019b ) , so we were able to estimate live cell densities at each time point t by taking ( fluorescence intensity at time t ) * ( initial cell density ) / ( initial fluorescence intensity ) . Both Coulter counter and flow cytometry forward scattering can be used to compare the cell size distributions of yeast strains , with Coulter counter providing a direct measurement of cell size . We used the Z2 Coulter counter ( Beckman ) , with the following settings: Gain = 128; Current = 0 . 5; Preamp Gain = 224 . We diluted cultures to OD600 ~ 0 . 01 to 0 . 3 ( 1 OD ~ 7×107 cells/ml ) when necessary , sonicated cells ( horn sonicator at low setting for three quick pulses or bath sonicator for 1 min ) , and placed 100 µl culture into Coulter cuvette . We then added 10 ml sterilized isotone down the wall of the titled cuvette to avoid splashing , and analyzed the sample . We measured metabolite utilization after cells fully saturated the culture . We starved exponentially-growing cells ( 3–6 hr for L-H+ , 24 hr for H-L+ ) to deplete initial intracellular stores and inoculated ~1×105 cells/ml into various concentrations of the required metabolite up to 25 µM . We incubated for 48 hr and then measured cell densities by flow cytometry . We performed linear regression between input metabolite concentrations ( horizontal axis ) and final total cell densities ( vertical axis ) within the linear range , forcing the regression line through origin . Utilization per birth in a saturated culture was quantified from 1/slope . Since release rate and metabolite utilization were measured in independent experiments , their errors were uncorrelated . For ratio f=A/B , suppose that A and B have standard deviations of σA and σB , respectively . Then σf is calculated as f ( σA/A ) 2+ ( σB/B ) 2 . We have constructed an eight-vessel chemostat with a design modified from Takahashi et al . ( 2015 ) . For details of construction , modification , calibration , and operation , see Skelding et al . ( 2017 ) . A detailed discussion on using chemostats to quantify release and utilization phenotypes can be found in Hart et al . ( 2019a ) . A summary is presented here . For L-H+ , due to rapid evolution , we devised experiments so that live and dead populations quickly reached steady state . We first calculated the expected steady state cell density by dividing the concentration of lysine in the reservoir ( 20 µM ) by fmole lysine utilized per new cell ( Figure 4 ) . We washed exponentially growing cells to remove extracellular lysine and inoculated 50% ~ 75% of the vessel volume at 1/3 of the expected steady state density . We filled the rest of the 19 ml vessel with reservoir media ( resulting in less than the full 20 µM of starting lysine , but more than enough for maximal initial growth rate , ~5–10 µM ) . We set the pump flow rate to achieve the desired doubling time T ( 19 ml culture volume*ln ( 2 ) /T ) . We collected and weighed waste media for each individual culturing vessel to ensure that the flow rate was correct ( i . e . total waste accumulated over time t was equal to the expected flow rate*t ) . We sampled cultures periodically to track population dynamics using flow cytometry ( Materials and methods , ‘Flow cytometry’ ) , and filtered supernatant through a 0 . 45 µm nitrocellulose filter and froze the supernatant for metabolite quantification at the conclusion of an experiment ( Materials and methods , ‘Bioassay’ ) . At the conclusion of an experiment , we also tested input media for each individual culturing vessel to ensure sterility by plating a 300 µl aliquot on an YPD plate and checking for growth after two days of growth at 30°C . If a substantial number of colonies grew ( >5 colonies ) , the input line was considered contaminated and data from that vessel was not used . For most experiments , we isolated colonies from end time point and checked percent evolved ( Materials and methods , ‘Microcolony assay’ ) . For L-H+ , we only analyzed time courses where >90% of population remained ancestral . In a lysine-limited chemostat , live cell density [L−H+]live is increased by growth ( at a rate g ) , and decreased by dilution ( at a rate dil ) : ( S1 ) d[L−H+]live/dt= ( g−dil ) [L−H+]live L , lysine concentration in the culturing vessel , is increased by the supply of fresh medium ( at concentration L0 ) , and decreased by dilution and utilization ( with birth of each new cell utilizing uL amount of lysine ) . ( S2 ) dL/dt=L0⋅dil−L⋅dil−uLg[L−H+]live Finally , hypoxanthine concentration H is increased by release ( from live cells at rH per live cell per hr , Hart et al . , 2019a ) , and decreased by dilution . ( S3 ) dH/dt=rH⋅[L−H+]live−dil⋅H Note that at the steady state ( denoted by subscript ‘ss’ ) , growth rate is equal to dilution rate ( setting Equation S1 to zero ) : ( S4 ) gss=dil To measure metabolite utilized per birth at steady state , we set Equation S2 to zero and also apply Equation S4 ( S5 ) uL= ( L0⋅dil−Lss⋅dil ) / ( gss[L−H+]live , ss ) ∼L0/[L−H+]live , ss Here , the approximation holds because the concentration of lysine in chemostat ( Lss ) is much smaller than that in reservoir ( L0 ) and thus Lss can be ignored . To measure release rate at steady state , we can set Equation S3 to zero and obtain: ( S6 ) rH=dil⋅Hss/[L−H+]live , ss Thus , the exchange ratio can be quantified from ( S7 ) rHuL=dil⋅Hss/[L-H+]live , ssL0/[L-H+]live , ss=HssL0dil Again , Hss is the steady state hypoxanthine concentration in the chemostat culture vessel , L0 is the lysine concentration in the reservoir , and dil is the dilution rate . This derivation is adapted from Hart et al . ( 2019a ) . If we culture L-H+ with H-L+ , we have ( S8 ) d[L−H+]dt= ( bL ( L ) −dL−cL ) [L−H+] ( S9 ) d[H−L+]dt= ( bH ( H ) −dH−cH ) [H−L+] ( S10 ) dLdt=rL[H−L+]−uLbL ( L ) [L−H+] ( S11 ) dHdt=rH[L−H+]−uHbH ( H ) [H−L+] Equation S8 states that the clonal population density L-H+ increases at birth rate bL which in turn depends on the concentration of lysine L , and decreases at death rate dL and cost of metabolite overproduction cL . Equation S9 describes how clonal population density H-L+ changes over time . Equation S10 states that the concentration of lysine L increases due to releaser H-L+ releasing at a rate rL and decreases as uL amount is utilized per birth of consumer L-H+ . Equation S11 describes how the concentration of H changes over time . All parameters are non-negative . Note that only a single genotype per species is considered . We can calculate the steady state growth rate gcomm . Since strain ratio becomes fixed ( Figure 4—figure supplement 3 ) , both strains must grow at the same rate as the community . This also means that L and H concentrations do not change . bL−dL−cL=gcommbH−dH−cH=gcommrL[H−L+]=uLbL[L−H+]=uL ( gcomm+dL+cL ) [L−H+]rH[L−H+]=uHbH[H−L+]=uH ( gcomm+dH+cH ) [H−L+] Multiply the last two equations , we getrHrL=uHuL ( gcomm+dL+cL ) ( gcomm+dH+cH ) Solving this , we get gcomm=− ( dH+cH+dL+cL ) 2+rHrLuHuL+ ( dH+cH−dL−cL ) 24 . When dH and dL and cH and cL are small compared to rHrLuLuH , which is the case for CoSMO , we have gcomm≈rHrLuHuL . Briefly , exponentially growing L-H+ and H-L+ were washed free of lysine and hypoxanthine supplements , respectively . H-L+ cell were further starved for 24 hr to reduce CoSMO growth lag phase ( Hart et al . , 2019a ) . The two strains were then mixed at approximately 1:1 ratio , and 15 µl of 4 × 104 total cells were spotted on the center of agarose pads ( 1/6 of a petri dish pie ) , forming an inoculum spot of radius ~4 mm . Periodically , cells from pads were washed off into water and subjected to flow cytometry . The agarose pad generally contained 0 . 7 µM lysine , although including or not including this low concentration of lysine did not make a difference in the steady state community growth rate . | Many organisms – including microbes – have mutually beneficially relationships . Often , the exchanged goods , such as nutrients , are costly to make . But what happens when individuals evolve to help themselves more ? Can they also evolve to be more helpful to others ? Hart , Pineda et al . studied a community of two genetically modified yeast strains that had to exchange essential nutrients to survive . One strain overproduced the molecule hypoxanthine , but depended on the second strain to provide the nutrient lysine , and vice versa . The communities were then allowed to evolve . The lysine-requiring strain frequently ended up with a mutation that initially seemed to be win-win: helping self to grow faster and at the same time , releasing more hypoxanthine to the partner . However , closer examination showed that the mutation also made these cells bigger , and bigger cells had to consume more lysine . Consequently , releasing more hypoxanthine was accompanied by consuming more lysine . Since the 'give' to 'take' ratio stayed the same , the partner strain did not benefit more from the mutant than from the ancestor . This suggests that an individual should not be considered helpful solely based on how much it gives to a partner , but also , on how much it takes . In the case of the mutant yeast strain , it produced 30 percent more nutrients , but also consumed 30 percent more , and was therefore not more helpful to the partner than the ancestor . Similarly , releasing less may not imply cheating . Beneficial interactions are very common in natural communities , such as among microbes living in the mouth cavity and the gut . Therefore , a better understanding of how they benefit from and affect each other may provide scientists with more insight into diseases linked to problems with microbial communities , such as tooth decay , inflammation of the gut , or obesity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2019 | Disentangling strictly self-serving mutations from win-win mutations in a mutualistic microbial community |
Lattices abound in nature—from the crystal structure of minerals to the honey-comb organization of ommatidia in the compound eye of insects . These arrangements provide solutions for optimal packings , efficient resource distribution , and cryptographic protocols . Do lattices also play a role in how the brain represents information ? We focus on higher-dimensional stimulus domains , with particular emphasis on neural representations of physical space , and derive which neuronal lattice codes maximize spatial resolution . For mammals navigating on a surface , we show that the hexagonal activity patterns of grid cells are optimal . For species that move freely in three dimensions , a face-centered cubic lattice is best . This prediction could be tested experimentally in flying bats , arboreal monkeys , or marine mammals . More generally , our theory suggests that the brain encodes higher-dimensional sensory or cognitive variables with populations of grid-cell-like neurons whose activity patterns exhibit lattice structures at multiple , nested scales .
Our grid-cell construction has one obvious degree of freedom , the length scale or grid size of the lattice L , that is , the width of the fundamental domain L . For a module with signature ς= ( Ω , ρ , L ) and for arbitrary scaling factor λ > 0 , the rescaled construction λς:= ( Ω ( λr ) , ρ ( λx ) , λ⋅L ) is a grid module too . The corresponding tuning curve satisfies ( Ω∘λ ) λL ( x ) =ΩL ( λx ) and is thus merely a scaled version of the former . Indeed , as we show in the ‘Material and methods’ section , the FI of the rescaled module is λ−2 Jς ( 0 ) . The Cramér-Rao bound ( Equation 4 ) implies that the local resolution of an unbiased estimator could thus rapidly improve with a finer grid size , that is , decreasing λ . However , for any grid module ς= ( Ω , ρ , L ) the posterior probability , that is , the likelihood of possible positions given a particular spike count vector K = ( k1 , … , kN ) , is also periodic . This follows from Bayes rule: ( 8 ) P ( x|K ) =P ( K|x ) ⋅P ( x ) P ( K ) ∝P ( x ) ∏i=1NPi ( ki|τ ΩiL ( x ) ) . Since the right hand side is invariant under operations of L on x , so is the left hand side of this equation . Thus , the multiple firing fields of a grid cell cannot be distinguished by a decoder , so that for λ → 0 the global resolution approaches the a priori uncertainty ( Mathis et al . , 2012a , 2012b ) . By combining multiple grid modules with different spatial periods one can overcome this fundamental limitation , counteracting the ambiguity caused by periodicity and still preserving the highest resolution at the smallest scale . Thus , one arrives at nested populations of grid modules , whose spatial periods range from coarse to fine . The FI for an individual module at one scale determines the optimal length scale of the next module ( Mathis et al . , 2012a , 2012b ) . The larger the FI per module , the greater the refinement at subsequent scales can be ( Mathis et al . , 2012a , 2012b ) . This result emphasizes the importance of finding the lattice that endows a grid module with maximal FI , but also highlights that the specific scale of the lattices can be fixed for this study . We now calculate the FI for a grid module with signature ς= ( Ω , ρ , L ) . For cells whose firing is statistically independent ( Equation 1 ) , the joint probability factorizes; therefore , the population FI is just the sum over the individual FI contributions by each neuron , Jς ( x ) =∑i=1M JΩiL ( x ) . The individual neurons only differ by their spatial phase ci , thus JΩiL ( x ) =JΩL ( x−ci ) . Consequently , Jς ( x ) =∑i=1M JΩL ( x−ci ) , depends only on the function JΩL ( r ) and the deviations x − ci , where ci is the closest lattice point of ci+L to x . If the grid-cell density ρ is uniform across L , then for all x∈ℝD: Jς ( x ) ≈ Jς ( 0 ) . It therefore suffices to only consider the FI at the origin , which can be written as: ( 9 ) Jς ( 0 ) =∑i=1MJΩL ( ci ) =∫L JΩL ( c ) ρ ( c ) dc . For uniformly distributed spatial phases ci and increasing number of neurons M , the law of large numbers implies ( 10 ) limM→∞|det ( L ) MJς ( 0 ) −∫L JΩL ( c ) dc|=0 . Here , det ( L ) denotes the volume of the fundamental domain . Thus , for large numbers of neurons M=∫L ρ ( c ) dc we obtain ( 11 ) Jς ( 0 ) ≈Mdet ( L ) ∫L JΩL ( c ) dc . This means that the population FI at 0 is approximately given by the average FI within the fundamental domain L times the number of neurons M . Let us now assume that supp ( Ω ) = [0 , R] for some positive radius R . Outside of this radius , the tuning shape is zero and the firing rate vanishes . So the spatial phases of grid cells that contribute to the FI at x = 0 lie within the ball BR ( 0 ) . If we now also assume that this ball is contained in the fundamental domain , BR ( 0 ) ⊂L , we get ( 12 ) ∫L JΩL ( c ) dc=∫BR ( 0 ) JΩL ( c ) dc . This result implies that any grid code ς= ( Ω , ρ , L ) , with large M , supp ( Ω ) = [0 , R] , and BR ( 0 ) ⊂L , satisfies ( 13 ) Jς ( 0 ) ≈Mdet ( L ) ∫BR ( 0 ) JΩL ( c ) dc . The FI at the origin is therefore approximately equal to the product of the mean FI contribution of cells within a R-ball around 0 and the number of neurons M , weighted by the ratio of the volume of the R-ball to the area of the fundamental domain L . Due to the radial symmetry of ΩL , the FI matrix JΩL ( c ) is diagonal with identical entries , guaranteeing the spatial resolution's isotropy . The error for each coordinate axis is bounded by the same value , that is , the inverse of the diagonal element 1/Jς ( 0 ) ii , for such a population . Instead of considering the FI matrix Jς ( 0 ) , we can therefore consider the trace of Jς ( 0 ) , which is the sum over the diagonal of Jς ( 0 ) . According to Equation 4 , 1/trJς ( 0 ) bounds the mean square error summed across all dimensions ε ( x^|x ) . For two lattices L1 , L2 , with BR ( 0 ) ⊂ L1∩L2 we consequently obtain ( 14 ) trJΩL1trJΩL2=det ( L2 ) det ( L1 ) , which signifies that the resolution of the grid module is inversely proportional to the volumes of their fundamental domains . The periodic structure L thus has a direct impact on the resolution of the grid module . This result implies that finding the maximum FI translates directly into finding the lattice with the highest packing ratio . The sphere packing problem is of general interest in mathematics ( Conway and Sloane , 1992 ) and has wide-ranging applications from crystallography to information theory ( Barlow , 1883; Shannon , 1948; Whittaker , 1981; Gray and Neuhoff , 1998; Gruber , 2004 ) . When packing R-balls BR in ℝD in a non-overlapping fashion , the density of the packing is defined as the fraction of the space covered by balls . For a lattice L , it is given by ( 15 ) vol ( BR ( 0 ) ) det ( L ) , which is known as the packing ratio Δ ( L ) of the lattice . For a given lattice , this ratio is maximized by choosing the largest possible R , known as the packing radius , which is defined as the in-radius of a Voronoi region containing the origin ( Conway and Sloane , 1992 ) . Figure 2 depicts the disks with the largest in-radius for the hexagonal and the square lattice in blue and illustrates the packing ratio . 10 . 7554/eLife . 05979 . 005Figure 2 . Periodified grid-cell tuning curve ΩL for two planar lattices , ( A ) the hexagonal ( equilateral triangle ) lattice H and ( B ) the square lattice Q , together with the basis vectors v1 and v2 . These are π/3 apart for the hexagonal lattice and π/2 for the square lattice . The fundamental domain , that is , the Voronoi cell around 0 , is shown in gray . A few other domains that have been generated according to the lattice symmetries are marked by dashed lines . The blue disk shows the disk with maximal radius R that can be inscribed in the two fundamental domains . For equal and unitary node-to-node distances , that is , |v1|=|v2|=1 , the maximal radius equals 1/2 for both lattices . The packing ratio Δ is Δ ( H ) =π/12 for the hexagonal and Δ ( Q ) =π/4 for the square lattice; the hexagonal lattice is approximately 15 . 5% denser than the square lattice . DOI: http://dx . doi . org/10 . 7554/eLife . 05979 . 005 We now come to the main finding of this study: among grid modules with different lattices , the lattice with the highest packing ratio leads to the highest spatial resolution . To derive this result , let us fix a tuning shape Ω with supp ( Ω ) = [0 , R] , lattices Lj such that BR ( 0 ) ⊂ Lj for 1 ≤ j ≤ K , and uniform densities ρ for each fundamental domain of equal cardinality M . Any linear order on the packing ratios , ( 16 ) Δ ( L1 ) ≤…≤Δ ( Lj ) ≤…≤Δ ( LK ) , is translated by Equation 14 into the same order for the traces of the FI ( 17 ) trJΩL1≤…≤trJΩLj≤…≤trJΩLK , and thus the resolution of these modules: the higher the packing ratio , the higher the FI of a grid module . The condition supp ( Ω ) = [0 , R] with BR ( 0 ) ⊂ L , although restrictive , is consistent with experimental observations that grid cells tend to stop firing between grid fields and that the typical ratio between field radius and spatial period is well below 1/2 ( Hafting et al . , 2005; Brun et al . , 2008; Giocomo et al . , 2011 ) . Generally , the tuning width that maximizes the FI does not necessarily satisfy this condition; see Figures 3 , 4 , in which the optimal support radius of the tuning curve θ2 is greater than the in-radius R = 1/2 of L . The same observation will hold in higher dimensions ( D > 2 ) , consistent with the finding that the optimal tuning width for Gaussian tuning curves increases with the number of spatial dimensions , whether space is infinite ( Zhang and Sejnowski , 1999 ) or finite ( Brown and Bäcker , 2006 ) . When the radius R of the support of the tuning curve exceeds the in-radius , the optimal lattice can be different from the densest one as we will show numerically for specific tuning curves and Poisson noise . However , with well separated fields , like those observed experimentally , the densest lattice provides the highest resolution for any tuning shape Ω , as we just demonstrated . 10 . 7554/eLife . 05979 . 006Figure 3 . Fisher information for modules of two-dimensional grid cells . ( A ) Top: Periodified bump-function Ω and square lattice L , for various parameter combinations θ1 and θ2 . Here , θ1 modulates the decay and θ2 the support . Middle: Average trace trJL of the Fisher information ( FI ) for uniformly distributed grid cells ΩL . Hexagonal ( H ) and square ( Q ) lattices are considered for different θ1 and θ2 values . The FI of the hexagonal grid cells outperforms the quadratic grid when support is fully within the fundamental domain ( θ2 < 0 . 5 , see main text ) . Bottom: Ratio trJH/trJQ as a function of the tuning parameter θ2 . For θ2 < 0 . 5 , the hexagonal population offers 3/2 times the resolution of the square population , as predicted by the respective packing ratios . ( B ) Average trJL for grid cells distributed uniformly in lattices generated by basis vectors separated by an angle φ ( basis depicted above graph ) . trJL behaves like 1/sin ( φ ) and has its maximum at π/3 . ( C ) Distribution of 5000 realizations of trJLM/M at 0 for a population of M = 200 randomly distributed neurons . For both the hexagonal and square lattice , parameters are θ1 = 1/4 and θ2 = 0 . 4 . The means closely match the average values in ( A ) . However , due to the finite neuron number the FI varies strongly for different realizations , and in about 20% of the cases a square lattice module outperforms a hexagonal lattice . DOI: http://dx . doi . org/10 . 7554/eLife . 05979 . 00610 . 7554/eLife . 05979 . 007Figure 3—figure supplement 1 . The firing rate and Fisher information of the bump tuning shape . Upper left panel: Tuning shape Ω ( r ) with parameters θ2 = 0 . 5 and varying θ1 . Lower left panel: Corresponding Fisher information ( FI ) integrand ℱ ( r ) . Upper right panel: Tuning shape Ω ( r ) with parameters θ1 = 0 . 25 and varying θ2 . Lower right panel: Corresponding FI integrand ℱ ( r ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05979 . 00710 . 7554/eLife . 05979 . 008Figure 4 . Fisher information for modules of 3D grid cells . ( A ) The three lattices considered: face-centered cubic ( FCC ) , body-centered cubic ( BCC ) , and cubic ( C ) . ( B ) trJL for the periodified bump-function Ω for the three lattices and various parameter combinations θ1 and θ2 . The Fisher information ( FI ) of the FCC grid cells outperforms the other lattices when the support is fully within the fundamental domain ( θ2 < 0 . 5 , see main text ) . For larger θ2 the best lattice depends on the relation between the Voronoi cell's boundary and the tuning curve . ( C ) Ratio trJL/trJC as a function of θ2 for L∈{FCC , BCC , C} . For θ2 < 0 . 5 , the hexagonal population has 3/2 times the resolution of the square population , as predicted by the packing ratios . ( D ) Average trJLφ , ψ for uniformly distributed grid cells within a lattice Lφ , ψ generated by basis vectors separated by angles φ and ψ ( as shown above; θ1 = θ2 = 1/4 ) . trJLφ , ψ behaves like 1/ ( sinφ⋅sinψ ) and has its maximum for the lattice with the smallest volume . ( E ) Distribution of 5000 realizations of trJLM/M at 0 for a population of M = 200 randomly distributed neurons . Parameters: θ1 = 1/4 , θ2 = 0 . 4 . The means closely match the averages in ( B ) . Due to the finite neuron number , the FI varies strongly for different realizations . DOI: http://dx . doi . org/10 . 7554/eLife . 05979 . 008 The optimal packing ratio of lattices for low-dimensional space is well known . Having established our main result , we can now draw on a rich body of literature , in particular Conway and Sloane ( 1992 ) , to discuss the expected firing-field structure of grid cells in 2D and 3D environments . With a packing ratio of π/12 , the hexagonal lattice is the densest lattice in the plane ( Lagrange , 1773 ) . According to Equation 14 , the hexagonal lattice is the optimal arrangement for grid-cell firing fields on the plane . For example , it outperforms the quadratic lattice , which has a density of π/4 , by about 15 . 5% ( see Figure 2 ) . Consequently , the FI of a grid module periodified along a hexagonal lattice outperforms one periodified along a square lattice by the same factor . To provide a tangible example , we calculated the trace of the average FI per neuron trJς/∫L ρ for signature ς= ( Ω , ρ , L ) and chose the lattice L to either be the hexagonal lattice H or the quadratic lattice Q . We denote the trace of the average FI per neuron as: trJL = trJς/∫L ρ; trJH and trJQ are similarly defined . We considered Poisson spike statistics and used a bump-like tuning shape Ω ( Equation 26 , ‘Materials and methods’ section ) . The tuning shape Ω depends on two parameters θ1 and θ2 , where θ1 controls the slope of the flank in Ω and θ2 defines the support radius . The periodified tuning curve ΩQ is illustrated for different parameters in the top of Figure 3A and in Figure 3—figure supplement 1 . Figure 3A depicts trJH and trJQ for various values of θ1 and θ2 . Quite generally , the FI is larger for grid modules with broad tuning ( large θ2 ) and steep tuning slopes ( small θ1 ) . Figure 3A also demonstrates that as long as θ2 ≤ 1/2 , trJH consistently outperforms trJQ . But how large is this effect ? As predicted by our theory , the grid module with the hexagonal lattice outperforms the square lattice by the relation of packing ratios 3/2 , as long as the support radius θ2 is within the fundamental domain of the hexagonal and the square lattice of unit length , that is , θ2 ≤ 1/2 ( bottom of Figure 3A ) . As the support radius becomes larger , the FI of the hexagonal lattice is no longer necessarily greater than that of the square lattice; the specific interplay of tuning curve and boundary shape determines which lattice is better: for θ1 = 1/4 , trJH/trJQ drops quickly beyond θ2 = 0 . 5 , even though , for θ1 = 1 , the ratio stays constant up to θ2 = 0 . 6 . Next we calculated the FI per neuron for a larger family of planar lattices generated by two unitary basis vectors with angle φ . Figure 3B displays trJL for φ ∈ [π/3 , π/2] , slope parameter θ1 = 1/4 , and different support radii θ2 . For the lattice to have unitary length , the value φ cannot go below π/3 . The trJL decays with increasing angle φ . Indeed , according to Equation 13 , the FI falls like 1/det L=1/sin ( φ ) so that the maximum is achieved for the hexagonal lattice with π/3 . The FIs trJL are averages over all phases , under the assumption that the density of phases tends to a constant; but are these values also indicative for small neural populations ? To answer this question , we calculated the FI for populations with 200 neurons , as some putative grid cells are found in patches of this size ( Ray et al . , 2014 ) . For M = 200 randomly chosen phases ( Figure 3C ) , the mean of the normalized FI trJLM/M over 5000 realizations is well captured by the FI per neuron calculated in Figure 3A . Because of fluctuations in the FI , however , the square lattice is better than the hexagonal lattice in about 20% of the cases . Our theory implies that for radially symmetric tuning curves the hexagonal lattice provides the best resolution among all planar lattices . This conclusion agrees with earlier findings: Wei et al . considered a notion of resolution defined as the range of the population code per smallest distinguishable scale and then demonstrated that a population of nested grid cells with hexagonal tuning is optimal for a winner-take-all and Bayesian maximum likelihood decoders ( Wei et al . , 2013 ) . Guanella and Verschure numerically compared hexagonal to other regular lattices based on maximum likelihood decoding ( Guanella and Verschure , 2007 ) . Gauss proved that the packing ratio of any cubic lattice is bounded by π/ ( 32 ) and that this value is attained for the face-centered cubic ( FCC ) lattice ( Gauss , 1831 ) illustrated in Figure 4A . This implies that the optimal 3D grid-cell tuning is given by the FCC lattice . For comparison , we also calculated the average population FI for two other important 3D lattices: the cubic lattice ( C ) and the body-centered cubic lattice ( BCC ) , both shown in Figure 4A . Keeping the bump-like tuning shape Ω and independent Poisson noise , we compared the resolution of grid modules with such lattices ( Figure 4B ) . Their averaged trace of FI is denoted by trJFCC , trJBCC , and trJC , respectively . As long as the support θ2 of Ω is smaller than 1/2 , the support is a subset of the fundamental domain of all three lattices . Hence , the trace of the population FI of the FCC outperforms both the BCC and C lattices . As the ratios of the trace of the population FI scales with the packing ratio ( Figure 4C ) , FCC-grid cells provide roughly 41% more resolution for the same number of neurons than do C-grid cells . Similarly , FCC-grid cells provide 8 . 8% more FI than BCC-grid cells . Next we calculated the FI per neuron for a large family of cubic lattices Lφ , ψ generated by three unitary basis vectors with spanning angles φ and ψ . Figure 4D displays trJLφ , ψ for θ1 = θ2 = 1/4 and various φ and ψ . The resolution trJL decays with increasing angles and has its maximum for the lattice with the smallest volume as predicted by Equation 13 . To study finite-size effects , we simulated 5000 populations of 200 grid cells with random spatial phases . Qualitatively , the results ( Figure 4E ) match those in 2D ( Figure 3C ) . Despite the small module size , FCC outperformed the cubic lattice C in all simulated realizations . Fruit is often arranged in an FCC formation ( Figure 5A ) . One arrives at this lattice by starting from a layer of hexagonally placed spheres . This requires two basis vectors to be specified and is the densest packing in 2D . To maximize the packing ratio in 3D , the next layer of hexagonally arranged spheres has to be stacked as tightly as possible . There are two choices for the third and final basis vector achieve this packing , denoted as γ1 and γ2 in Figure 5B ( modulo hexagonal symmetry ) . If one chooses γ1 , then two layers below there is no sphere with its center at location γ1 , but instead there is one at γ2 ( and vice versa ) . This stacking of layers is shown in Figure 5C and generates the FCC lattice . 10 . 7554/eLife . 05979 . 009Figure 5 . Lattice and non-lattice solutions in 3D . ( A ) Stacking of spheres as in an FCC lattice . In this densest lattice in 3D , each sphere touches 12 other spheres and there are four different planar hexagonal lattices through each node . ( B ) Over a layer of hexagonally arranged spheres centered at γ0 ( in black ) one can put another hexagonal layer by starting from one of six locations , two of which are highlighted , γ1 and γ2 . ( C ) If one arranges the hexagonal layers according to the sequence ( … , γ1 , γ0 , γ2 , … ) one obtains the FCC . Note that spheres in layer I are not aligned with those in layer III . ( D ) Arranging the hexagonal layers following the sequence ( … , γ0 , γ1 , γ0 , … ) leads to the hexagonal close packing HCP . Again , each sphere touches 12 other spheres . However , there is only one plane through each node for which the arrangement of the centers of the spheres is a regular hexagonal lattice . This packing has the same packing ratio as the FCC , but is not a lattice . ( E ) trJL for bump-function Ω with L=FCC and HCP for various parameter combinations θ1 and θ2; θ1 modulates the decay and θ2 the support . The two packings have the same packing ratio and for this tuning curve also provide identical spatial resolution . FI: Fisher information . DOI: http://dx . doi . org/10 . 7554/eLife . 05979 . 009 One could achieve the same density by choosing γ1 for both the top layer and the layer below the basis layer . Yet as this arrangement , called hexagonal close packing ( HCP ) , cannot be described by three vectors , it does not define a lattice ( see Figure 5D ) , even though it is as tightly packed as the FCC . Such packings , defined as an arrangement of equal non-overlapping balls ( Conway and Sloane , 1992; Hales , 2012 ) , generalize lattices . While one can define a grid module for any lattice , as we showed above , one cannot define a grid module in a meaningful way for an arbitrary packing , due to the lack of symmetry . But for any given packing P of ℝD by balls B1 of radius 1 , one can define a ‘grid cell’ by generalizing the definition given for lattices ( Equation 7 ) . To this end , consider the Voronoi partition of ℝD by P . For each location x∈ℝD there is a unique Voronoi cell Vp with node p∈P . One defines the grid cell's tuning curve ΩP ( x ) by assigning the firing rate according to Ω ( ∥p−x∥2 ) for tuning shape Ω and distance ∥p−x∥ . Depending on the specific packing , this tuning curve ΩP may or may not be periodic . Because a packing P often has fewer symmetries than a lattice L , the ‘grid cells’ in an arbitrary P cannot generally be used to define a ‘grid module’ . To explain why , consider an arbitrary packing and the unique Voronoi cell V0 that contains the point 0 . Choose M uniformly distributed phases c1 , … , cM within V0 . Locations within V0 will then be uniformly covered by shifted tuning curves Ωi ( x ) :=ΩP ( x−ci ) . However , typically the different Voronoi cells will neither be congruent , nor have similar volumes . Thus , the Ωi will typically not cover each Voronoi cell with the same density and will therefore fail to define a proper grid module . This problem does not exist for lattices . Here , the equivalence classes ci+L cover each cell with the same density . Highly symmetric packings , on the other hand , do permit the definition of grid modules . For example , the hexagonal close packing HCP can be used to define a grid cell ΩHCP ( x ) . Using the same symmetry argument from Equations 9–11 , implies for the FI: ( 18 ) J ( Ω , ρ , HCP ) ( x ) ≈J ( Ω , ρ , HCP ) ( 0 ) ≈Mvol ( V0 ) ∫V0 JΩHCP ( c ) dc . The maximal in-radius R for the HCP with grid size λ = 1 is equal to 1/2 . Like for lattices , we assume that supp ( Ω ) = [0 , R] and BR ( 0 ) ⊂ V0 . Then the integrand vanishes for distances larger than 1/2 from 0 . Hence , we obtain: ( 19 ) J ( Ω , ρ , HCP ) ( 0 ) ≈Mvol ( V0 ) ∫B1/2 ( 0 ) JΩHCP ( c ) dc . Considering the same tuning shape Ω and number of phases M for an FCC lattice , which also has maximal in-radius 1/2 , Equation 13 gives us the following expression for the FCC lattice: ( 20 ) J ( Ω , ρ , FCC ) ( 0 ) ≈Mdet ( FCC ) ∫B1/2 ( 0 ) JΩFCC ( c ) dc . Since both fundamental domains have the same volumes , that is , det ( FCC ) =vol ( V0 ) , and the integrands restricted to these balls are identical , that is , JΩFCC|B1/2 ( 0 ) =JΩHCP|B1/2 ( 0 ) , we can conclude that grid modules comprising FCC or HCP-like symmetries have the same FI . We also numerically calculate the trace of the average FI for a module of HCP grid cells and compare it to the FCC case . For bump-like tuning curves Ω , both FIs are identical ( Figure 5E ) as expected from the radial symmetry of Ω . As a consequence , grid cells defined by either HCP or FCC symmetries provide optimal resolution . Figure 5D , E shows that the cyclic sequences ( γ0 , γ1 ) and ( γ1 , γ0 , γ2 ) lead to HCP and FCC , respectively . The centers γ0 , γ1 , and γ2 can also be used to make a final point on packings: there are infinitely many distinct packings with the same density π/ ( 32 ) . They can be constructed by inequivalent words , generated by finitewalks through the triangle with letters γ0 , γ1 , and γ2 ( Hales , 2012 ) , with each letter representing one of three orientations for the layers . For instance , ( γ0 , γ1 , γ0 , γ2 ) describes another packing with the same density . All packings share one feature: around each sphere there are exactly 12 spheres , arranged in either HCP or FCC lattice fashion ( Hales , 2012 ) . These packings can also be used to define a grid module , because the density of phases will be uniform in all cells . Furthermore , as in the calculation of the FI for the HCP and FCC ( Equation 18–20 ) only local integration was necessary , such mixed packings will have equally large , uniform FI as the pure HCP or FCC packings . Only in recent years has it been proven that no other arrangement has a higher packing ratio than the FCC , a problem known as Kepler's conjecture ( Hales , 2005 , 2012 ) . Based on these results and our comparison of trJHCP and trJFCC ( Figure 5E ) , we predict that 3D grid cells will correspond to one of these packings . While there are equally dense packings as the densest lattice in 3D , this is not the case in 2D . Thue proved that the hexagonal lattice is unique in being the densest amongst all planar packings ( Thue , 1910 ) ; grid cells in 2D should possess a hexagonal lattice structure .
Grid cells are active when an animal is near one of any number of multiple locations that correspond to the vertices of a planar hexagonal lattice ( Hafting et al . , 2005 ) . We generalize the notion of a grid cell to arbitrary dimensions , such that a grid cell's stochastic activity is modulated in a spatially periodic manner within ℝD . The periodicity is captured by the symmetry group of the underlying lattice L . A grid module consists of multiple cells with equal spatial period but different spatial phases . Using information theory , we then asked which lattice offers the highest spatial resolution . We find that the resolution of a grid module is related to the packing ratio of L—the lattice with highest packing ratio corresponds to the grid module with highest resolution . Well-known results from mathematics ( Lagrange , 1773; Gauss , 1831; Conway and Sloane , 1992 ) then show that the hexagonal lattice is optimal for representing 2D , whereas the FCC lattice is optimal for 3D . In 3D , but not in 2D , there are also non-lattice packings with the same resolution as the densest lattice ( Thue , 1910; Hales , 2012 ) . A common feature of these highly symmetric optimal solutions in 3D is that each grid field is surrounded by 12 other grid fields , arranged in either FCC lattice or hexagonal close packing fashion . These solutions emerge from the set of all possible packings simply by maximizing the resolution , as we showed . However , resolution alone , as measured by the FI , does not distinguish between optimal packing solutions with different symmetries . Whether a realistic neuronal decoder , such as one based on population vector averages , favors one particular solution is an interesting open question . As we have demonstrated , using the FI makes finding the optimal L analytically tractable for all dimensions D and singles out densest lattices as optimal tuning shapes under assumptions that are restrictive , but are consistent with experimental measurements ( Hafting et al . , 2005; Brun et al . , 2008; Giocomo et al . , 2011 ) . The assumption that the tuning curves must have finite support within the fundamental domain of the lattice corresponds to grid cells being silent outside of the firing field . Indeed , our numerical simulations also showed that for broader tuning curves , grid modules with quadratic lattices can provide more FI than the hexagonal lattice ( Figure 3A , θ2 ≈ 0 . 6 and θ1 = 1/4 ) and that grid cells with a C or BCC lattice can provide more FI than the FCC ( Figure 4B , θ2 > 0 . 65 and θ1 = 1/4 ) . For the planar case , Guanella and Verschure ( 2007 ) show numerically that triangular tessellations yield lower reconstruction errors under maximum-likelihood decoding than equivalently scaled square grids . Complementing this numerical analysis , Wei et al . ( 2013 ) provide a mathematical argument that hexagonal grids are optimal . To do so , they define the spatial resolution of a single module representing 2D space as the ratio R = ( λ/l ) 2 , where λ is the grid scale and l is the diameter of the circle in which one can determine the animal's location with certainty . For a fixed resolution R , the number of neurons required is N = d sin ( φ ) R in their analysis , where d is the number of tuning curves covering each point in space . As φ ∈ [π/3 , π/2] for the lattice to have unitary length ( Figure 3B ) , minimizing N for a fixed resolution R yields φ = π/3; thus , hexagonal lattices should be optimal . Furthermore , Wei et al . show that this result also holds when considering a Bayesian decoder ( Wei et al . , 2013 ) . While Wei et al . minimize N for fixed l , we minimize l ( in their notation ) . Like Wei et al . , we assume that the tuning curve Ω is isotropic ( notwithstanding the fact that the lattice has preferred directions ) ; unlike these authors , we show that there are conditions under which the firing fields should be arranged in a square lattice , and not hexagonally . Using the FI gives a theoretical bound for the local resolution of any unbiased estimator ( Lehmann , 1998 ) . In particular , this local resolution does not take into account the ambiguity introduced by the periodic nature of the lattice . Our analysis is restricted to resolving the animal's position within the fundamental domain . For large neuron numbers N and expected peak spike counts fmaxτ the resolution of asymptotically efficient decoders , like the maximum likelihood decoder , or the minimum mean square estimator , can indeed attain the resolution bound given by the FI ( Seung and Sompolinsky , 1993; Bethge et al . , 2002; Mathis et al . , 2013 ) . Thus , for these decoders and conditions the results hold . In contrast , for small neuron numbers and peak spike counts , the optimal codes could be different , just as it has been shown in the past that the optimal tuning width in these cases cannot be predicted by the FI ( Bethge et al . , 2002; Yaeli et al . , 2010; Berens et al . , 2011; Mathis et al . , 2012 ) . Maximizing the resolution explains the observed hexagonal patterns of grid cells in 2D , and predicts an FCC lattice ( or equivalent packing ) for grid-cell tuning curves of mammals that can freely explore the 3D nature of their environment . Quantitatively , we demonstrated that these optimal populations provide 15 . 5% ( 2D ) and about 41% ( 3D ) more resolution than grid codes with quadratic or cubic grid cells for the same number of neurons . Although better , this might not seem substantial , at least not at the level of a single grid module . However , as medial entorhinal cortex harbors a nested grid code with at least 5 and potentially 10 or more modules ( Stensola et al . , 2012 ) , this translates into a much larger gain of 1 . 1555 … 10≈2 . 1 … 4 . 2 and 25 … 10≈5 . 7 … 32 , respectively ( Mathis et al . , 2012a , 2012b ) . Because aligned grid-cell lattices with perfectly periodic tuning curves imply that the posterior is periodic too ( compare Equation 8 ) , information from different scales would have to be combined to yield an unambiguous read-out . Whether the nested scales are indeed read out in this way in the brain remains to be seen ( Mathis et al . , 2012a , 2012b; Wei et al . , 2013 ) . An alternative hypothesis , as first suggested by Hafting et al . , is that the slight , but apparently persistent irregularities in the firing fields across space ( Hafting et al . , 2005; Krupic et al . , 2015; Stensola et al . , 2015 ) are being used . Future experiments should tackle this key question . We considered perfectly periodic structures ( lattices ) and asked which ones provide most resolution . However , the first recordings of grid cells already showed that the fields are not exactly hexagonally arranged and that different fields might have different peak firing rates ( Hafting et al . , 2005 ) . More recently , deviations from hexagonal symmetry have gained considerable attention ( Derdikman et al . , 2009; Krupic et al . , 2013 , 2015; Stensola et al . , 2015 ) . Such ‘defects’ modulate the periodicity of the tuning and consequently affect the symmetry of the likelihood function . This might imply that a potential decoder might be able to distinguish different unit cells even given a single module , which is not possible for perfectly periodic tuning curves ( compare Equation 8 ) . The local resolution , on the other hand , is robust to small , incoherent variations as the FI is a statistical average over many tuning curves with different spatial phases . At a given location , Equation 9 becomesJς ( x ) =∑i=1MJΩiL ( x−ci ) ≈∫x−LJΩL¯ ( c ) ρ ( c ) dc , where ΩL¯ is the average of the variable tuning curves ΩiL . Small variations in the peak rate and grid fields will therefore average out , unless these variations are coherent across grid cells . Thus , resolution bounded by the FI is robust with respect to minor differences in peak firing rates and hexagonality . Similar arguments hold in higher dimensions . In this study , we focused on optimizing grid modules for an isotropic and homogeneous space , which means that the resolution should be equal everywhere and in each direction of space . From a mathematical point of view , this is the most general setting , but it is certainly not the only imaginable scenario; future studies should shed light on other geometries . Indeed , the topology of natural habitats , such as burrows or caves , can be highly complicated . Higher resolution might be required at spatial locations of behavioral relevance . Neural representations of 3D space may also be composed of multiple 1D and 2D patches ( Jeffery et al . , 2013 ) . However , the mere fact that these habitats involve complicated low-dimensional geometries does not imply that an animal cannot acquire a general map for the environment . Poincaré already suggested that an isotropic and homogeneous representation for space can emerge out of non-Euclidean perceptual spaces , as one can move through physical space by learning the motion group ( Poincaré , 1913 ) . An isotropic and homogeneous representation of 3D space facilitates ( mental ) rotations in 3D and yields local coordinates that are independent of the environment's topology . On the other hand , the efficient-coding hypothesis ( Barlow , 1959; Atick , 1992; Simoncelli and Olshausen , 2001 ) would argue that surface-bound animals might not need to dedicate their limited neuronal resources to acquiring a full representation of space , as flying animals might have to do , so that representations of 3D space will be species-specific ( Las and Ulanovsky , 2014 ) . Desert ants represent space only as a projection to flat space ( Wohlgemuth et al . , 2001; Grah et al . , 2007 ) . Likewise , experimental evidence suggests that rats do not encode 3D space in an isotropic manner ( Hayman et al . , 2011 ) , but this might be a consequence of the specific anisotropic spatial navigation tasks these rats had to perform . Data from flying bats , on the other hand , demonstrate that , at least in this species , place cells represent 3D space in a uniform and nearly isotropic manner ( Yartsev and Ulanovsky , 2013 ) . The 3D , toroidal head-direction system in bats also suggests that they have access to the full motion group ( Finkelstein et al . , 2014 ) . Our theoretical analysis assumes that the same is true for bat grid cells and that they have radially symmetric firing fields . From these assumptions , we showed the grid cells' firing fields should be arranged on an FCC lattice or packed as HCP . Interestingly , such solutions also evolve dynamically in a self-organizing network model for 3D ( Stella et al . , 2013; Stella and Treves , 2015 ) that extends a previous 2D system which exhibits hexagonal grid patterns ( Kropff and Treves , 2008 ) . Experimentally , the effect of the arena's geometry on grid cells' tuning and anchoring has also been a question of great interest ( Derdikman et al . , 2009; Krupic et al . , 2013 , 2015; Stensola et al . , 2015 ) . First , let us note that even though the environment might be finite , the grid-cell representation need not be constrained by it . In particular , the firing fields are not required to be contained within the confines of the four walls of a box—experimental observations show that walls can intersect the firing fields ( so that one measures only a part of the firing field ) . On the other hand , the borders clearly distort the hexagonal arrangement of nearby firing fields in 2D environments ( Stensola et al . , 2015 ) , whereas central fields are more perfectly arranged . Deviations are also observed when only a few fields are present in the arena ( Krupic et al . , 2015 ) . One might expect similar deviations in 3D , such as for bats flying in a confined space . Our mathematical results rely on symmetry arguments that do not cover non-periodic tuning curves . Given that the resolution is related to the packing ratio of a lattice , extensions of the theory to general packings might allow one to draw on the rich field of optimal finite packings ( Böröczky , 2004; Toth et al . , 2004 ) , thereby providing new hypotheses to test . Many spatially modulated cells in rat medial entorhinal cortex have hexagonal tuning curves , but some have firing fields that are spatially periodic bands ( Krupic et al . , 2012 ) . The orientation of these bands tends to coincide with one of the lattice vectors of the grid cells ( as the lattices for different grid cells share a common orientation ) , so band cells might be a layout ‘defect’ . In this context , we should point out that the lattice solutions are not globally optimal . For instance , in 2D , a higher resolution can result from two systems of nested 1D grid codes , which are aligned to the x and y axis , respectively , than from a lattice solution with the same number of neurons . The 1D cells would behave like band cells ( Krupic et al . , 2012 ) . Similar counterexamples can be given in higher dimensions too . The anisotropy of the spatial tuning in grid cells of climbing rats when encoding 3D space ( Hayman et al . , 2011 ) might capitalize on this gain ( Jeffery et al . , 2013 ) . Radial symmetry of the tuning curve may also be non-optimal . For example , two sets of elliptically tuned 2D unimodal cells , with orthogonal short axes , typically outperform unimodal cells with radially symmetric tuning curves ( Wilke and Eurich , 2002 ) . Why experimentally observed place fields and other tuning curves seem to be isotropically tuned is an open question ( O'Keefe and Dostrovsky , 1971; Yartsev and Ulanovsky , 2013 ) . Grid cells which represent the position of an animal ( Hafting et al . , 2005 ) have been discovered only recently . By comparison , in technical systems , it has been known since the 1950s that the optimal quantizers for 2D signals rely on hexagonal lattices ( Gray and Neuhoff , 1998 ) . In this context , we note that lattice codes are also ideally suited to cover spaces that involve sensory or cognitive variables other than location . In higher-dimensional feature spaces , the potential gain could be enormous . For instance , the optimal eight-dimensional ( 8D ) lattice is about 16 times denser than the orthogonal 8D lattice ( Conway and Sloane , 1992 ) and would , therefore , dramatically increase the resolution of the corresponding population code . Advances in experimental techniques , which allow one to simultaneously record from large numbers of neurons ( Ahrens et al . , 2013; Deisseroth and Schnitzer , 2013 ) and to automate stimulus delivery for dense parametric mapping ( Brincat and Connor , 2004 ) , now pave the way to search for such representations in cortex . For instance , by parameterizing 19 metric features of cartoon faces , such as hair length , iris size , or eye size , Freiwald et al . showed that face-selective cells are broadly tuned to multiple feature dimensions ( Freiwald et al . , 2009 ) . Especially in higher cortical areas , such joint feature spaces should be the norm rather than the exception ( Rigotti et al . , 2013 ) . While no evidence for lattice codes was found in the specific case of face-selective cells , data sets like this one will be the test-bed for checking the hypothesis that other nested grid-like neural representations exist in cortex .
How is the resolution of a grid module affected by dilations ? Let us assume we have a grid module with signature ς= ( Ω , ρ , L ) , as defined in the main text , and that λ > 0 is a scaling factor . Then λς:= ( Ω ( λr ) , ρ ( λx ) , λ⋅L ) is a grid module too , and the corresponding tuning curve ( Ω∘λ ) λL satisfies: ( 21 ) ( Ω∘λ ) λL ( x ) =ΩL ( λx ) . Thus , the tuning curve ( Ω∘λ ) λL is a scaled version of ΩL . What is the relation between the FI of the initial grid module and the rescaled version ? Let us fix the notation: ρ ( c ) =∑iNδ ( c−ci ) . From the definition of the population information ( Equation 9 ) , we calculate ( 22 ) Jλς ( 0 ) =∑i=1MJ ( Ω∘λ ) λL ( λci ) =∑i=1MJΩL ( ci ) ⋅1λ2=1λ2Jς ( 0 ) , where in the second step we used the re-parameterization formula of the FI ( Lehmann , 1998 ) . This shows that the FI of a grid module scaled by a factor λ is the same as the FI of the initial grid module times 1/λ2 . In the ‘Results’ section , we give a concrete example for Poisson noise and the bump function . Here we give the necessary background . Equation 13 states thatJς ( 0 ) ≈Mdet ( L ) ∫BR ( 0 ) JΩL ( c ) dc . One would like to know ∫BR ( 0 ) JΩL ( c ) dc for various tuning shapes Ω with supp ( Ω ) ≤ R . Consider x ∈ L and α ∈ {1 , … , D} . Then: ( 23 ) ∂lnP ( K|x ) ∂xα=∂lnP ( K , s ) ∂s|s=ΩL ( x ) ⋅Ω′ ( ∥x∥2 ) fmaxτ 2xα . Together with the definition of the FI Equation 13 , this yields ( 24 ) JΩL ( x ) αβ=4xαxβfmax2τ2Ω′ ( ∥x∥2 ) 2 . ∑K ( ∂∂slnP ( K , s ) |s=ΩL ( x ) ) 2⋅P ( K , ΩL ( x ) ) ︸=:N ( ∥x∥2 ) . Note that for α ≠ β this function is odd in x . Thus , when averaging these individual contributions over a symmetric fundamental domain L: ∫L JΩL ( c ) αβdc=0 for α ≠ β . Thus , the diagonal entries are all identical . This also holds for any fundamental domain L when BR ( 0 ) ⊂ L , because BR ( 0 ) is symmetric . For Poisson spiking N ( ∥c∥2 ) has a particularly simple form , namely N ( ∥c∥2 ) =1/ ( fmaxτ Ω ( ∥c∥2 ) ) . The trace of the FI matrix becomes ( 25 ) trJς ( 0 ) =4fmaxτ∫BR ( 0 ) ∥c∥2Ω′ ( ∥c∥2 ) 2Ω ( ∥c∥2 ) ︸=:F ( c ) dc . Thus , the trace only depends on the tuning shape Ω and its first derivative . In the main text , we use the following specific tuning shape: ( 26 ) Ω ( r ) ={exp ( θ1θ22−θ1θ22−r2 ) if |r| <θ20otherwise . This type of function is often called ‘bump function’ in topology , as it has a compact support but is everywhere smooth ( i . e . , infinitely times continuously differentiable ) . In particular , the support of this function is [0 , θ2 ) , and is therefore controlled by the parameter θ2 . The other parameter θ1 controls the slope of the bump's flanks ( see upper panels of Figure 3—figure supplement 1 ) . For the bump-function Ω and radius r=∑αDxα2 the integrand for the FI is given by ( 27 ) F ( r ) ={4θ12r2 ( θ22−r2 ) 4 exp ( θ1θ22−θ1θ22−r2 ) if |r| <θ20otherwise . The lower panels of Figure 3—figure supplement 1 depict the integrand of Equation 25 , defined as F ( r ) . This function shows how much FI a cell at a particular distance contribute to the location 0 . By integrating the FI over the fundamental domain L for a lattice L one gets Jς ( 0 ) , that is , the average FI contributions from all neurons ( as shown in Figures 3 , 4 , 5E ) . | The brain of a mammal has to store vast amounts of information . The ability of animals to navigate through their environment , for example , depends on a map of the space around them being encoded in the electrical activity of a finite number of neurons . In 2014 the Nobel Prize in Physiology or Medicine was awarded to neuroscientists who had provided insights into this process . Two of the winners had shown that , in experiments on rats , the neurons in a specific region of the brain ‘fired’ whenever the rat was at any one of a number of points in space . When these points were plotted in two dimensions , they made a grid of interlocking hexagons , thereby providing the rat with a map of its environment . However , many animals , such as bats and monkeys , navigate in three dimensions rather than two , and it is not clear whether these same hexagonal patterns are also used to represent three-dimensional space . Mathis et al . have now used mathematical analysis to search for the most efficient way for the brain to represent a three-dimensional region of space . This work suggests that the neurons need to fire at points that roughly correspond to the positions that individual oranges take up when they are stacked as tight as possible in a pile . Physicists call this arrangement a face-centered cubic lattice . At least one group of experimental neuroscientists is currently making measurements on the firing of neurons in freely flying bats , so it should soon be possible to compare the predictions of Mathis et al . with data from experiments . | [
"Abstract",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2015 | Probable nature of higher-dimensional symmetries underlying mammalian grid-cell activity patterns |
Hsp90 is a conserved chaperone that facilitates protein homeostasis . Our crystal structure of the mitochondrial Hsp90 , TRAP1 , revealed an extension of the N-terminal β-strand previously shown to cross between protomers in the closed state . In this study , we address the regulatory function of this extension or ‘strap’ and demonstrate its responsibility for an unusual temperature dependence in ATPase rates . This dependence is a consequence of a thermally sensitive kinetic barrier between the apo ‘open’ and ATP-bound ‘closed’ conformations . The strap stabilizes the closed state through trans-protomer interactions . Displacement of cis-protomer contacts from the apo state is rate-limiting for closure and ATP hydrolysis . Strap release is coupled to rotation of the N-terminal domain and dynamics of the nucleotide binding pocket lid . The strap is conserved in higher eukaryotes but absent from yeast and prokaryotes suggesting its role as a thermal and kinetic regulator , adapting Hsp90s to the demands of unique cellular and organismal environments .
Hsp90 is a highly conserved molecular chaperone essential for protein and cellular homeostasis . Although molecular chaperones generally promote protein folding and prevent aggregation , Hsp90 is unique in that it interacts with substrate ( ‘client’ ) proteins that are already in a semi-folded state to facilitate downstream protein–protein interactions and promote client function in diverse biological pathways ( Jakob et al . , 1995; Taipale et al . , 2012 ) . Hsp90 interacts with nearly 10% of the eukaryotic proteome ( Zhao et al . , 2005 ) , and its client proteins vary significantly in sequence , structure , and size ( Echeverria et al . , 2011 ) . In most eukaryotes , there are four different Hsp90 homologs: Hsp90α and Hsp90β in the cytoplasm , Grp94 in the endoplasmic reticulum ( ER ) , and TRAP1 in mitochondria , with each homolog contributing unique biological functions ( Chen et al . , 2006; Johnson , 2012 ) . Deregulation of Hsp90 protein levels and function has been linked to multiple human diseases and for this reason Hsp90 is a target for biochemical characterization , structural studies , and drug discovery ( Luo et al . , 2010; Taipale et al . , 2010 ) . Despite such importance , little is known about the biochemical characteristics that regulate client interaction and specificity . Hsp90 exists as a homodimer , with each protomer consisting of three major domains . The N-terminal domain ( NTD ) binds to ATP , the C-terminal domain ( CTD ) provides a dimerization interface between protomers , and the middle domain ( MD ) provides a stabilizing γ-phosphate contact to help facilitate ATP hydrolysis ( Cunningham et al . , 2012 ) . Together with the CTD , the MD has been shown to aid in the formation of client interactions ( Street et al . , 2011 , 2012; Genest et al . , 2013 ) . Large , rigid body motions about each of the domain interfaces give rise to an ensemble of remarkably diverse conformational states that dictate the functional Hsp90 cycle ( Ali et al . , 2006; Shiau et al . , 2006; Dollins et al . , 2007; Southworth and Agard , 2008; Lavery et al . , 2014 ) ( Video 1 ) and are linked to client maturation in vivo ( Panaretou et al . , 1998 ) . Work from numerous labs has demonstrated conservation of the underlying conformational cycle and mechanism; however , each Hsp90 homolog has a distinct conformational equilibrium and catalytic rate ( Panaretou et al . , 1998; Richter et al . , 2008; Southworth and Agard , 2008 ) . Binding of ATP to the NTD nucleotide-binding pocket ultimately leads to stabilization of an NTD-dimerized state . Key steps in this transformation include ATP binding , closure of a mobile structure ( lid ) over the nucleotide , and a subsequent 90o rotation of the NTD relative to the MD ( Krukenberg et al . , 2011 ) . Dimer closure is the rate-limiting step for Hsp90 ATPase activity and mutations that either subtly increase or decrease ATPase rates compromise viability in yeast ( Nathan and Lindquist , 1995; Hessling et al . , 2009 ) . However , our understanding of the sequence of events that regulate these structural rearrangements is limited . 10 . 7554/eLife . 03487 . 003Video 1 . Conformational dynamics of the Hsp90 cycle . A morph between known conformations throughout the activity cycle of Hsp90 ( PDB codes with no order dictated: 2O1V , 2CG9 , 2IOP , 2IOQ , 4IPE , 4IVG ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 003 Recently , we solved a series of full-length crystal structures of TRAP1 bound to different ATP analogs ( Lavery et al . , 2014 ) , providing new insights into the structure , dynamics , and mechanism of Hsp90 . Of particular note was the marked asymmetry between protomers of the homodimer , primarily at the interface between the MD and CTD . This asymmetry was sampled in solution , proved essential for catalytic turnover , and provided a new model for coupling the energy of ATP hydrolysis to client remodeling . A second feature highlighted by the TRAP1 crystal structure was an ordered 14-residue extension ( out of 26 total additional residues ) of the N-terminal β-strand previously shown to cross over ( ‘swap’ ) between protomers in the closed state ( Ali et al . , 2006 ) . While absent in yeast and bacteria , this extension , or ‘strap , ’ is found in most eukaryotic Hsp90 proteins including the cytosolic and organellar forms ( Chen et al . , 2006 ) and can extend for as many as 122 residues as recently found in a splice variant of Hsp90α in higher eukaryotes ( Tripathi and Obermann , 2013 ) . Structure based point mutations and complete removal of the TRAP1 strap ( Δstrap ) resulted in a sixfold increase in ATPase activity in zebrafish TRAP1 ( zTRAP1 ) ( Lavery et al . , 2014 ) , evidence that the strap plays a regulatory role . Similarly , deletion of the strap in Grp94 ( residues 22-72 , referred to as the ‘pre-N domain’ ) resulted in a fivefold increase in ATPase ( Dollins et al . , 2007 ) , indicative of a conserved regulatory role , although the mechanism remains unclear . In this study , we explore the conformational cycle of TRAP1 and demonstrate that the strap is responsible for a large thermal barrier between the apo ( open ) and ATP bound ( closed ) states . Using negative-stain electron microscopy ( EM ) and Small-Angle X-ray Scattering ( SAXS ) , we demonstrate that removal of the strap results in a profound reduction in the temperature sensitivity observed in multiple TRAP1 homologs , indicating that the strap is responsible for this unique behavior . Additionally , we develop fluorescence resonance energy transfer ( FRET ) and continuous-wave EPR ( CW-EPR ) assays to show that the strap regulates the rate-limiting conformational transitions that precede NTD dimerization , including NTD rotation and lid closure over the ATP-binding pocket . These results indicate that the strap must stabilize both the apo state and the closed state , providing a unique evolutionary strategy for modulating different phases of the kinetic landscape and optimizing in vivo function of diverse Hsp90s .
With all previously studied Hsp90s , incubation with slowly- or non-hydrolyzable ATP analogs favors accumulation of a closed , NTD-dimerized state . However , the extent of closed-state accumulated and the rate of closure differentiated the Hsp90s with the individual values positively correlating with the ATP hydrolysis rate ( Richter et al . , 2008; Southworth and Agard , 2008; Hessling et al . , 2009 ) . Specifically , we used EM to demonstrate that the large variability in observed ATPase rates of cytosolic bacterial ( bHsp90 ) , yeast ( yHsp90 ) , and human Hsp90 ( hHsp90 ) , directly correlated with the ability of each homolog to reach a closed conformation in the presence of non-hydrolyzable ATP ( AMPPNP ) ( Southworth and Agard , 2008 ) . Here , negative-stain EM was again used to monitor the ability of human TRAP1 ( hTRAP1 ) to transition from the apo conformation to the closed conformation in the presence of AMPPNP . While hTRAP1 has an ATPase rate similar to the E . coli Hsp90 ( ∼0 . 5 min−1 ) ( Cunningham et al . , 2012 ) , surprisingly hTRAP1 remained in the open conformation despite incubation with saturating AMPPNP ( Figure 1A ) . Noting that the discrepancy might be related to different incubation temperatures between the two experiments , we monitored the ability of hTRAP1:AMPPNP to close as a function of temperature . After 1 hr ( Figure 1A ) or overnight ( Figure 1B ) incubation at room temperature ( RT , ∼23°C ) hTRAP1 remained in the open state . However , after a single hour of incubation at increasing temperatures , the closed state was increasingly populated ( Figure 1A ) . These results correlate well with the temperature sensitive steady-state hydrolysis rates of hTRAP1 that increases by nearly 200-fold between 25°C and 55°C ( Leskovar et al . , 2008 ) and are consistent with closure being rate-limiting for hydrolysis . Importantly , the equilibrium reached at each temperature ( Figure 1A ) remains fixed after subsequent incubation at RT overnight ( Figure 1B ) . These data suggest both a large , unusually thermally sensitive kinetic barrier to closure and a highly stable closed state . 10 . 7554/eLife . 03487 . 004Figure 1 . A temperature-dependent barrier separates the apo and closed state of TRAP1 . ( A ) Negative stain electron microscopy ( EM ) images of hTRAP1 in the presence of AMPPNP at increasing temperatures for 1 hr . While the population at equilibrium appears to remain in an apo conformation at room temperature ( RT ) , conversion to the closed state appears to be intermediate at 30°C and nearly complete at 37°C and 42°C . ( B ) Negative stain EM images of reactions incubated at 23°C and 37°C from A after returning the sample to RT and incubating overnight . Both populations remain apo and closed ( respectively ) demonstrating the large kinetic barrier that limits the conformational transition from apo to the closed state . Scale bar is 100 nm . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 004 To better measure the equilibrium between conformational states as a function of temperature , we used SAXS , which can directly quantify the solution distribution of open and closed states ( Krukenberg et al . , 2008 ) . As demonstrated by a shift towards a more compact pair-wise inter-atomic distance distribution , P ( r ) , there was a strong correlation between temperature and dimer closure , ( Figure 2A ) . By fitting the distributions as a linear combination of open and closed states , the fraction of closed state can be accurately estimated ( ‘Materials and methods’ ) . After 1 hr at 20°C only 0 . 4% of the molecules have reached the closed conformation , while at 43°C roughly 84% of the molecules are closed ( Figure 2C and Table 1 ) . In agreement with our EM data , the equilibrium does not revert back to the apo state when the temperature is lowered ( Figure 2D ) . Interestingly , TRAP1 from zebrafish ( zTRAP1 ) displays a shifted temperature-dependent conformational equilibrium that correlates with its higher basal ATPase rate ( Figure 2 and Table 2 ) and the lower physiological temperature of zebrafish ( ∼29°C ) . 10 . 7554/eLife . 03487 . 005Figure 2 . A large energy barrier to the closed state is modulated by the NTD-strap . ( A ) SAXS distributions at equilibrium for hTRAP1 ( left ) and zTRAP1 ( right ) ( 84% identical to hTRAP1 ) in apo and in the presence of saturating AMPPNP at indicated temperatures for 1 hr . The closed-state population substantially increases at and above 36°C for hTRAP1 while zTRAP1 maintains a higher level of % closed at even lower temperatures , consistent with the differences in physiological temperatures of the two species . ( B ) SAXS distributions of Δstrap in matching conditions from A showing that removal of the strap mitigates the temperature-dependent barrier between the apo and closed states . ( C ) Quantification of percent closed for both TRAP1 species ± the strap region . Apparent is the different temperature dependence of hTRAP1 and zTRAP1 and the loss of temperature response of the chaperone in the case of Δstrap . ( D ) A plot of percent closed state verses temperature of WT hTRAP1 ( left ) and Δstrap hTRAP1 after closure has completed at each given temperature ( solid bars as in ( C ) . These samples were then cooled for 2 hr at 20°C ( stripped bars ) . The data suggest a highly stable closed state . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 00510 . 7554/eLife . 03487 . 006Table 1 . Quantification of percent closed using SAXS data for both TRAP1 species ± the strapDOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 006Temperature ( °C ) Protein% Closed stateR20WT hTRAP10 . 50 . 04223WT hTRAP120 . 04230WT hTRAP1310 . 02432WT hTRAP1410 . 01936WT hTRAP1740 . 01140WT hTRAP1830 . 01043WT hTRAP1840 . 01220WT zTRAP1360 . 01523WT zTRAP1480 . 01130WT zTRAP1750 . 02732WT zTRAP1810 . 03336WT zTRAP1800 . 03043WT zTRAP1690 . 01620hTRAP1 Δstrap660 . 01523hTRAP1 Δstrap680 . 01430hTRAP1 Δstrap690 . 01432hTRAP1 Δstrap680 . 01536hTRAP1 Δstrap670 . 01840hTRAP1 Δstrap690 . 01643hTRAP1 Δstrap710 . 01520zTRAP1 Δstrap600 . 01623zTRAP1 Δstrap640 . 01430zTRAP1 Δstrap610 . 01532zTRAP1 Δstrap620 . 01436zTRAP1 Δstrap550 . 01643zTRAP1 Δstrap760 . 01110 . 7554/eLife . 03487 . 007Table 2 . Steady-state ATP hydrolysis rates at temperatures and buffer conditions of assay specified ( i . e . , EPR is under EPR buffer and temperature conditions ) . If not noted ( top four reactions ) , conditions are the same as reference ( Lavery et al . , 2014 ) DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 007ProteinzTRAP1 ATPase ( min−1 ) hTRAP1 ATPase ( min−1 ) WT ( 30°C ) 1 . 36 ± 0 . 120 . 463 ± 0 . 003salt bridge point mutants ( 30°C ) ( Lavery et al . , 2014 ) ( E-A ) 3 . 57 ± 0 . 62 ( H-A ) 5 . 08 ± 0 . 90Δstrap ( 30°C ) 5 . 84 ± 0 . 4713 . 3 ± 0 . 5Δ60-69 ( 30°C ) 0 . 47 ± 0 . 02WT FRET ( 30°C ) 0 . 21 ± 0 . 0 . 01Δstrap FRET ( 30°C ) 11 . 9 ± 2 . 1CFree WT EPR ( 23°C ) 0 . 88 ± 0 . 05CFree Δstrap EPR ( 23°C ) 5 . 65 ± 0 . 22CFree WT ( Inter FRET ) ( 30°C ) 0 . 79 ± 0 . 0 . 03CFree Δstrap ( Inter FRET ) ( 30oC ) 7 . 6 ± 0 . 36CFree WT ( Intra FRET ) ( 30°C ) 0 . 35 ± 0 . 0 . 01Red text indicates WT or strap truncated protein with native cysteine and label free , while Blue indicates labeled protein used in FRET and EPR experiments ( each in indicated buffer conditions ) . EPR samples are cysteine free except for the desired probe position and are spin-labeled . Inter FRET and Intra FRET samples are cysteine free except for the desired probe position and are labeled with Alexa Fluor dyes ( Life Technologies , see ‘Materials and methods’ ) . Note that ‘Intra FRET’ refers to both probe positions on the same promoter , whereas ‘Inter FRET’ refers to one probe position per promoter . Errors represent the standard deviation of triplicate experiments . Previous studies by Richter et al . had demonstrated that removal of the initial β-strand in yHsp90 ( corresponding to post-strap residues in TRAP1 ) increased ATPase activity and facilitated N-terminal dimerization ( Frey et al . , 2007 ) . The ordered strap extension observed in the zTRAP1 structure is also kinetically important , as Δstrap ( deletion of zTRAP1 residues 73–100 ) and a single point mutant aimed at disrupting a conserved , stabilizing salt bridge at the beginning of the strap , accelerated hydrolysis by sixfold and fourfold , respectively . Together , these raised the possibility that the strap might also be responsible for the unusual temperature-regulated energy landscape observed in TRAP1 homologs . As a first step , we show that in hTRAP1 , strap removal ( lacking residues 60–85 ) has an even more profound impact on ATPase activity ( ∼30-fold ) than on zTRAP1 ( Figure 3 , Table 2 ) . The larger increase in ATPase activity for hTRAP1 correlates with the more significant temperature dependence ( Figure 2C ) and thus a higher kinetic barrier for hTRAP1 at the experimental temperature of 30°C . Notably , a smaller truncation lacking residues 60–69 , that preserved the conserved His71:Glu142 salt-bridge in hTRAP1 , did not have an effect on ATPase activity ( Figure 3 , Table 2 ) . 10 . 7554/eLife . 03487 . 008Figure 3 . NTD-strap regulates ATP hydrolysis rates . WT and strap mutants for hTRAP1 . Removal of the strap ( Δstrap ) results in a ∼30-fold increase in ATPase rate , while truncations before the previously reported salt bridge contact Δ60–69 ( Lavery et al . , 2014 ) show no change in activity . Average steady-state hydrolysis rates ( min−1 ) above each bar , standard deviation of triplicate measurements can be found in Table 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 008 Strikingly , SAXS revealed that at every temperature examined in both hTRAP1- and zTRAP1-Δstrap immediately equilibrated to a closed conformation after adding AMPPNP , indicating a loss of temperature sensitivity ( at least at temperatures ≥20°C ) ( Figure 2B ) . This indicates that beyond a role in stabilizing the closed conformation through trans-protomer interactions , the strap must also be involved in apo interactions that inhibit a transition towards the closed state . These data together with previously solved crystal structures of other Hsp90 N-terminal domains displaying cis-contacts of the initial β-strand suggest that the strap likely makes equivalent contacts with the same NTD ( cis ) in the apo state that it forms with the trans-NTD in the closed state ( Shiau et al . , 2006; Dollins et al . , 2007; Li et al . , 2012 ) . The above results predict that the rate of closure should be proportional to temperature . Fluorescence Resonance Energy Transfer ( FRET ) provides a more convenient method than SAXS to directly measure the rate of closure ( Hessling et al . , 2009; Mickler et al . , 2009; Street et al . , 2011 ) . Two FRET constructs were designed to probe distinct aspects of the closure reaction , relying on a Cys-free version of hTRAP1 ( Lavery et al . , 2014 ) . The first construct modeled from FRET positions previously designed for yHsp90 placed a single Cys residue on each protomer ( E140C and E407C , ‘Inter FRET’ ) so as to give an increase in FRET upon closure ( Hessling et al . , 2009 ) . The second construct modeled on previous work with bHsp90 ( Street et al . , 2012 ) , adds two Cys residues to a single protomer ( S133C and E407C , ‘Intra FRET’ ) , and is designed to track the ∼90o NTD rotation ( relative to the MD ) that occurs upon closure . After forming heterodimers , closure reactions were initiated with AMPPNP over a temperature range mirroring our SAXS experiments and the change in FRET was monitored . Pre- and post-reaction fluorescent scans showed a predicted FRET change indicative of closure for each FRET construct ( Figure 4A ) . As expected , the rate of closure correlated with increasing temperature ( Figure 4B , Figure 4—figure supplement 1 , Table 3 ) for both dimer closure and NTD:MD rotation measurements . To measure the contribution of the strap to the kinetics of closure , we truncated the strap region of either one or both protomers in each FRET construct ( although the dimeric Δstrap construct used to measure NTD rotation proved too unstable to obtain reliable data ) . In both cases , a large acceleration of closure was apparent ( Figure 4C ) with the largest acceleration ( 16-fold ) observed for the double-strap deletion . 10 . 7554/eLife . 03487 . 009Figure 4 . The NTD-strap regulates closure rate of TRAP1 . ( A ) Steady-state FRET scans at 23°C for apo and AMPPNP reactions after closure with AMPPNP reached completion illustrating the anti-correlated change in FRET upon closure as measured by ‘dimer closure’ between protomers ( left , Inter FRET ) and rotation of the NTD from apo to the closed state within one protomer ‘NTD:MD Rotation’ ( right , Intra FRET ) . ( B ) Temperature-dependent closure rates for WT hTRAP1 measured by both the dimer closure and NTD rotation FRET probes from A . Closure rates are comparable between these two sets of FRET probes as indicated in the table to the right . The predicted increase in rate at higher temperatures is apparent . ( C ) Closure at 30°C of WT compared to heterodimers lacking one or both NTD strap residues measured by dimer closure FRET ( left ) and NTD rotation FRET ( right ) . Closure rates are found in the table for each experiment . ( D ) Temperature-dependent closure rates of Δstrap protein measured using the dimer closure probes from A ( Inter FRET ) illustrating both a rate acceleration and a dramatic loss of temperature dependence compared to WT ( B , left panel ) . ( E ) Arrhenius plot of WT and Δstrap plotted using data from panels ( B ) ( left ) and ( D ) . From the difference in activation energies Ea between WT and Δstrap , the strap contributes approximately 60% of the measured Ea for WT hTRAP1 ( 48 . 8 kcal/mol Ea for WT; 29 kcal/mol Δstrap ) . These data are consistent with the steady-state SAXS and ATPase and show that removal of the strap region lowers the energy barrier between apo and the closed state . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 00910 . 7554/eLife . 03487 . 010Figure 4—figure supplement 1 . Alternative view of curve fits for Figure 4B , D . ( A ) Kinetics of FRET closure at lower temperatures ( 23°C , 30°C , 32°C ) with fits shown for full measured curve . All reactions were taken to completion . ( B ) Data from Figure 4D plotted to focus on the smaller difference in closure rate for Δstrap at increasing temperatures . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 01010 . 7554/eLife . 03487 . 011Figure 4—figure supplement 2 . Arrhenius plots for Hsp90 homologs plotted using data from reference ( Frey et al . , 2007 ) . Calculated Ea for each homolog is listed in figure legend parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 01110 . 7554/eLife . 03487 . 012Table 3 . Kinetics of conformational changes as measured by FRET . Errors represent the standard deviation of triplicate experimentsDOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 012ProteinTemperature ( °C ) FRET probe positionKclose ( min−1 ) Kreopen ( min−1 ) Khyd ( min−1 ) CFree WT hTRAP1 ( Intra FRET ) 23S133C . E407C0 . 0011 ± 0 . 0000630S133C . E407C0 . 0066 ± 0 . 0000732S133C . E407C0 . 021 ± 0 . 00136S133C . E407C0 . 073 ± 0 . 00242S133C . E407C0 . 36 ± 0 . 01130S133C . E407C0 . 073 ± 0 . 005CFree WT hTRAP1 ( Inter FRET ) 23E140C/ E407C0 . 00330E140C/ E407C0 . 02 ± 0 . 0020 . 00210 ± 0 . 0000332E140C/ E407C0 . 03936E140C/ E407C0 . 11842E140C/ E407C0 . 431CFree Δstrap single ( Inter FRET ) 30E140C/ E407C0 . 21 ± 0 . 013CFree Δstrap double ( Inter FRET ) 23E140C/ E407C0 . 07330E140C/ E407C0 . 31 ± 0 . 0240 . 016 ± 0 . 00332E140C/ E407C0 . 45636E140C/ E407C0 . 85342E140C/ E407C1 . 335*CFree WT hTRAP1 ( Inter FRET ) *ATP used30E140C/ E407C0 . 42 ± 0 . 017 . 56 ± 0 . 9925E140C/ E407C0 . 19 ± 0 . 01 ( 0 . 003 ± 0 . 0001 s−1 ) 4 . 3 ± 0 . 2 ( 0 . 071 ± 0 . 004 s−1 ) *CFree Δstrap double ( Inter FRET ) *ATP used30E140C/ E407C1 . 5 ± 0 . 0110 . 6 ± 0 . 425E140C/ E407C0 . 927 . 57*Denotes ATP was used for closure . Relates to Figures 4 , 6 and Figure 6—figure supplement 1 . All other reactions used AMPPNP as the ATP analog . Note that ‘Intra FRET’ in red refers to both probe positions on the same promoter , whereas ‘Inter FRET’ in blue refers to one probe position per promoter . A good way to quantitate the contribution of the strap to the thermal barrier is to measure closure rates as a function of temperature with and without the strap ( Figure 4B , D and Figure 4—figure supplement 1 ) and to calculate the activation energy ( Ea ) towards closure ( i . e . , the temperature dependent barrier height ) . At every temperature sampled removing the strap results in an acceleration of closure compared to WT and an overall loss in temperature dependence ( Figure 4D ) . Comparing the fold changes in closure rates ( Table 3 ) , we see the largest fold change at lower temperatures ( 23°C: 24-fold , 30°C: 16-fold , 32°C: 12-fold , 36°C: sevenfold , and 42°C: threefold ) . This increased impact at lower temperatures is readily evident in an Arrhenius plot calculated from the inter FRET experiments ( Figure 4E ) . The resultant activation energies ( Ea ) taken from the slopes of these curves are 48 . 8 kcal/mol and 29 kcal/mol , for WT and Δstrap respectively . From the difference , the strap appears to be contributing ∼20 kcal/mol towards the Ea of WT hTRAP1 , which we interpret as ∼ 20 kcal/mol of enthalpic stabilization of the open state . Our Ea for WT hTRAP1 is consistent with that measured previously under slightly different conditions ( Leskovar et al . , 2008 ) , but is considerably higher than that calculated for other Hsp90 homologs ( Figure 4—figure supplement 2 ) ( Frey et al . , 2007 ) . As a control , we also measured steady-state ATPase rates on the labeled protein used for the FRET experiments . While these showed differences in absolute ATPase rates between 1 . 5 and fourfold compared with their unlabeled counterparts ( Tables 2 and 3 ) , the relative impact of strap deletion was consistent across experiments . Together , these data support a model in which the N-terminal strap limits closure by inhibiting the rotational movement of the NTD that is necessary to form the catalytically active closed state . To probe the underlying mechanism of the NTD-strap in the closure reaction , we sought to examine the relationship of the strap to the dynamics of the NTD lid ( zTRAP1 residues 191–217 ) that closes over the ATP binding pocket; a mechanism conserved in many ATPases . Previous studies with yHsp90 have suggested a correlation between the ‘β-strand swap’ and dynamics of the NTD lid ( Richter et al . , 2006 ) . In an open conformation and prior to nucleotide binding , the lid makes contacts with helix 1 ( H1 ) ( Richter et al . , 2006; Shiau et al . , 2006; Dollins et al . , 2007; Li et al . , 2012 ) , while in the closed state the lid rotates to secure nucleotide via interactions at conserved sidechains ( Ser193 and Ser195 in zTRAP1 ) inside the nucleotide binding pocket ( Ali et al . , 2006; Lavery et al . , 2014 ) ( Videos 2 and 3 ) . This closed state lid conformation is incompatible with the NTD:MD apo state conformation as it would clash with the MD ( Shiau et al . , 2006; Dollins et al . , 2007 ) . 10 . 7554/eLife . 03487 . 013Video 2 . NTD-strap anti-correlated lid conformational changes . A morph between two conformations of Hsp90 , from the Apo state with cis-protomer interactions between NTD and strap , to the nucleotide bound closed state where the strap makes trans-protomer interactions . This morph demonstrates the significant number of contacts that are lost and then reformed to accommodate movement of the NTD to form the NTD-dimerization interface . ( PDB codes 2IOQ , 4IVG , 4IPE ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 01310 . 7554/eLife . 03487 . 014Video 3 . Movement of the lid to accommodate the NTD-dimerization interface . A morph between two conformations of yHsp90 NTD , either in the APO state or nucleotide bound closed conformation . This morph demonstrates the coordinated movement and changing contacts between both the β-strand ( pink ) and NTD lid ( red ) to facilitate the NTD-dimerization interface of a dimerized Hsp90 molecule . ( PDB codes 4AS9 , 2CG9 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 014 To test whether the strap has a role in lid stabilization , we developed an electron paramagnetic resonance ( EPR ) spectroscopy assay to track lid mobility in the apo and closed states ( ‘Materials and methods’ ) . A cys-free version of zTRAP1 with an Ala201Cys mutation allowed labeling of a fully accessible cysteine residue in the lid with N- ( 1-oxyl- 2 , 2 , 6 , 6-tetramethyl-4-piperidinyl ) maleimide ( MSL ) . We observed a small difference in ATPase activity with the MSL-labeled TRAP1 compared with ATPase rates measured using WT TRAP1 suggesting a minor labeling effect on steady-state catalytic turnover ( ∼1 . 3 fold ) . EPR spectra are sensitive to the rotational mobility of the attached MSL probe making it a useful reporter for changes in local conformational dynamics ( Hubbell et al . , 2000 ) . EPR spectra of full-length zTRAP1 were recorded at 23°C and shown to be more mobile in the nucleotide bound state compared to the apo state ( Figure 5A ) . Mobility of the lid as measured with EPR is consistent with apo structures showing low B-factors in this region due to significant contacts with H1 of the cis protomer ( Richter et al . , 2006; Shiau et al . , 2006 ) . Conversely , crystal structures of TRAP1 and other Hsp90 homologs bound to ATP analogs in the closed and dimerized conformation show that the lid folds over the nucleotide , has increased B-factors and lacks many of the stabilizing contacts with the N-terminal domain found in the apo state ( Ali et al . , 2006; Lavery et al . , 2014 ) . This is consistent with the mobile signature in the EPR observed for the closed conformation . Comparing apo state equilibrium measurements for WT and Δstrap shows little change upon strap deletion ( Figure 5A ) . Fortunately EPR is sufficiently sensitive and the closure kinetics for TRAP1 are sufficiently slow , that it is possible to directly monitor changes in lid state over time . By plotting the change in normalized peak heights over time ( ‘Materials and methods’ , Figure 5B ) , it is apparent that the amplitude changes for both the mobile and immobile peaks are well fit by a single exponential curve for each sample . From this , it is clear that the rate of change between states as monitored by lid mobility is much faster for the Δstrap sample than for WT . The fold difference between rates is on the order of changes in ATPase rates under conditions used in the EPR experiment ( Table 2 ) . Altogether , these data suggest that the local conformational changes of lid closure and NTD-rotation are part of the rate-limiting barrier to the closed state and are regulated by N-terminal residues of the strand swap and extended strap in TRAP1 . 10 . 7554/eLife . 03487 . 015Figure 5 . Lid Closure rate is regulated by the NTD-strap . ( A ) Continuous Wave ( CW ) EPR scans of cysteine Free WT ( top ) and Δstrap zTRAP1 ( bottom ) labeled with a spin-probe on the NTD-lid ( green ) in order to observe changes to the lid in the apo and closed states ( ‘Materials and methods’ ) . In the apo state the lid probe shows signal for both mobile and immobile states , although crystallographic data indicate that even in the mobile state , the majority of the lid is still reasonably well ordered . After addition of AMPPNP , the observed signal shifts indicating a predominantly mobile state of the lid , which corresponds to changes in lid dynamics that accompany NTD rotation and dimerization . Only subtle differences are seen in the mobile:immobile peak ratio upon strap deletion . ( B ) CW-EPR scans at ∼23°C taken for the cysteine-free WT ( red ) and Δstrap zTRAP1 ( blue ) over time after addition of AMPPNP . The percent change in peak height ( final vs start ) over time is plotted for both the immobile ( squares ) and mobile ( circles ) components , showing a clear anti-correlation . The mobile and immobile populations were jointly fit with a single exponential process ( ‘Materials and methods’ ) having a rate constant of 0 . 014 min−1 for WT and 0 . 075 min−1 for Δstrap , demonstrating a strong coupling between the strap and the NTD-lid . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 015 The experiments above collectively suggest a major role for the N-terminal strap as a direct modulator of the kinetic barrier separating the apo and closed states for TRAP1 . Moreover , it appears to be also responsible for the pronounced temperature-sensitivity . Although the experiments above indicate a strong role in modulating the forward closure rate , the TRAP1 crystal structure would suggest that deleting the strap might also compromise the stability of the closed state , thereby enhancing reopening rate and shifting the equilibrium towards the open state . To measure the re-opening rate , inter FRET-labeled hTRAP1 was pre-closed with AMPPNP . After closure was complete a 20-fold excess ADP was added such that upon re-opening of the NTD dimer interface , ADP would exchange resulting in a decreased FRET signal ( Street et al . , 2011 ) . Previous studies found apo state nucleotide on and off-rates to be fast ( Leskovar et al . , 2008 ) , thus the above experiment provides a good approximation of the uni-molecular reopening rate . Monitoring FRET kinetics revealed that strap removal accelerated re-opening of the NTD dimer interface by ∼eightfold ( 0 . 0021 min−1 → 0 . 016 min−1; Figure 6A , Table 3 ) . These data suggest that the strap contacts observed in the closed state ( Lavery et al . , 2014 ) do in fact impact closed-state stability , by about 1 . 2 kcal/mol , however , the larger effect ( ∼16-fold , 0 . 02 min-1 → 0 . 31 min−1 , 1 . 7 kcal/mol ) is on the kinetic barrier corresponding to release of the strap from the apo state . 10 . 7554/eLife . 03487 . 016Figure 6 . The NTD-strap plays a smaller role in additional steps of the ATPase cycle . ( A ) Schematic of dimer closure and re-opening upon addition of AMPPNP ( PNP ) using the dimer closure FRET probe ( left ) . Re-opening of WT hTRAP1 and Δstrap was induced by 20-fold excess ADP after closure with AMPPNP . Re-opening was accelerated by ∼eightfold upon removal of the strap as determined by the ratio of the rates ( table inset ) . ( B ) Steady-state FRET scans of dimer closure FRET in apo and plus ATP in the absence of Mg2+ . Without Mg2+ a closed state accumulates , whereas subsequent addition of Mg2+ ( ‘+ATP & Mg2+’ ) allows hydrolysis to proceed thereby shifting the population to the apo state . ( C ) Schematic of a kinetic experiment using the Mg2+ dependence to separate the rate of hydrolysis from rate of closure . By omitting Mg2+ , the population can be synchronized in a closed state that is unable to hydrolyze ATP . Subsequent rapid addition of Mg2+ leads to ATP hydrolysis , which has now been decoupled from the closure step . ( D ) Kinetic experiments measuring closure and ( E ) ATP hydrolysis . No closed state accumulates if Mg2+ is included in the closure reaction . Again we observe that removal of strap residues leads to an accelerated closure rate , whereas the difference in ATP hydrolysis is small . Kinetic rates for each are listed in the table insets . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 01610 . 7554/eLife . 03487 . 017Figure 6—figure supplement 1 . ATP Hydrolysis at 25°C . ( A–B ) Kinetic experiment designed to decouple ATP hydrolysis from the preceding closure step at 25°C outlined in Figure 6C–E . The measured rates for WT agree well with previous single turnover measurements ( Leskovar et al . , 2008 ) , however , we find that decoupling the closure rate from hydrolysis results in a reassignment of the previous rates with closure being the slowest step . Kinetic rates for each are listed in the table inset . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 017 Because the strap could also play a role in the hydrolysis reaction , we needed a method to decouple closure from ATP hydrolysis . As closure is rate-limiting , even single-turnover experiments would provide an aggregate rate made up of the closure and hydrolysis steps . During the course of our FRET experiments , we discovered that omitting Mg2+ from the reaction buffer results in an accumulation of the closed-state in the presence of ATP without ATP hydrolysis . By contrast , in the presence of ATP and Mg2+ , TRAP1 is predominantly in the apo state as a consequence of hydrolysis ( Figure 6B , D ) . The latter is consistent with previous observations with yHsp90 ( Hessling et al . , 2009 ) and bHsp90 . These observations allowed us to decouple the closure and hydrolysis steps by pre-incubating TRAP1 with excess ATP without Mg2+ , thereby stalling the reaction in the closed state ( illustrated in Figure 6C ) . Upon addition of Mg2+ , ATP is hydrolyzed and the equilibrium shifts predominantly to the apo state as seen by the loss of FRET ( Figure 6B ) . Testing WT and Δstrap in this assay revealed an acceleration of closure with removal of the strap , consistent with experiments using AMPPNP ( Figure 6D ) . Interestingly , the closure rate measured by FRET is significantly faster with ATP than AMPPNP suggesting a significant difference in energetics between the nucleotide analogs ( Table 3 ) . The use of ATP for FRET-based closure measurements better matches our ATPase measurements and points to a correlation between closure rates and ATPase activity ( 0 . 79 min−1 ATPase vs 0 . 42 min−1 FRET Closure , both measurements with Inter FRET probe protein ) , though we do still observe a difference perhaps representing a small Mg2+ contribution . Addition of excess Mg2+ showed the predicted drop in FRET and revealed a minor difference in hydrolysis rate ( ∼1 . 4-fold ) ( Figure 6E , Table 3 ) , suggesting that the strap may also subtly alter lid dynamics in the closed state . The acceleration effects observed for the Δstrap protein are greater at 25°C , where the temperature dependent difference is more pronounced ( Figure 6—figure supplement 1 , Table 3 ) . Notably , our measured closure and hydrolysis rates matched previously reported values for these steps modeled using a global fitting procedure ( Leskovar et al . , 2008 ) . However , the closure and hydrolysis rates measured here ( 0 . 003 s−1 and 0 . 07 s−1 , respectively ) were somewhat arbitrarily assigned to the reverse order in the previously reported model . Since our experiments independently measure both reactions , we can now assign closure to be the slowest and hence rate-limiting step . This model is in good agreement with the other data presented in this study . Our combined data better define the kinetic cycle for TRAP1 and support a model where the strap regulates multiple steps with the largest contribution being to the thermal sensitive rate-limiting kinetic barrier between the apo and nucleotide-bound closed states .
While the crystal structure of the TRAP1 closed state revealed that the strap made stabilizing interactions with the trans protomer , we show here that its dominant role in modulating ATPase activity is to limit the closure kinetics , presumably though analogous cis-protomer interactions in the apo state . Removal of the strap leads to a ∼30-fold increase in ATPase rate and faster closure kinetics that include the smaller conformational steps of NTD-rotation and lid closure , as well as loss of thermal regulation of dimer closure . The strap extension in TRAP1 appears to continue and expand upon the kinetic regulatory affects observed previously for the first eight residues in yHsp90 , which makes contacts on the trans-protomer in the closed state ( Ali et al . , 2006; Lavery et al . , 2014 ) . Deletion of these residues was shown to accelerate the ATPase rate by ∼1 . 5-fold by allowing H1 and the lid to undergo conformational changes necessary to form trans-protomer contacts at the NTD-dimer interface ( Richter et al . , 2002 , 2006 ) . These effects are understood in the light of numerous apo NTD structures showing this strand makes analogous contacts with its own NTD in the apo state ( Shiau et al . , 2006; Dollins et al . , 2007; Li et al . , 2012 ) . In the TRAP1 closed-state structure , the 14 ordered residues wrap around the side of the NTD and add an additional 757 Å2 , as calculated with PISA ( Krissinel and Henrick , 2007 ) , of buried surface area and several new trans-protomer contacts ( Lavery et al . , 2014 ) . We propose that similar additional contacts are made in the apo state ( Videos 2 and 3 ) , which is supported by our own data showing that truncations up the first major contact ( salt bridge ) have no effect on ATPase ( Figure 3 ) . Given the similar accelerating effects on ATPase and dynamics as studied in multiple organisms , it is likely that the β-strand and the strap are acting on the same barrier . Distilling the available information , we outline a model that defines kinetic steps in the Hsp90 ATPase cycle and consequently determine the rates of ATP hydrolysis ( Figure 8A ) . Specifically , after ATP is bound , release of cis contacts of the β-strand/strap is coupled to lid closure and NTD rotation , presenting surfaces that form and stabilize an NTD-dimerized state . Due to closure-induced strain this ultimately results in an asymmetric conformation ( Lavery et al . , 2014 ) . Hydrolysis of one of the two ATPs leads to rearrangement of client binding residues ( red ) between the MD:CTD thus coupling the first ATP to client remodeling when clients are bound to this region . The actual conformational state post hydrolysis of the first ATP is currently unknown , but is here schematized as the symmetric state identified in the yHsp90 crystal structure ( Ali et al . , 2006 ) . After the second ATP is hydrolyzed the chaperone assumes the previously observed compact ADP conformation before resetting the cycle to the apo state . 10 . 7554/eLife . 03487 . 019Figure 8 . Model for the conformational cycle and unique energy landscape of TRAP1 . ( A ) In the absence of nucleotide the chaperone is in equilibrium between various open conformations ( for simplicity we only show the most open ) with the strap folded back onto the cis protomer . Upon binding of ATP , conformational changes necessary for the transition to the closed state are initiated . Here , we propose that the cis contacts of the strap are broken allowing the lid and NTD to undergo conformational changes towards the closed state . After the slow closure step the chaperone assumes the previously reported asymmetric conformation ( Lavery et al . , 2014 ) . Sequential hydrolysis leads to changes in symmetry rearranging the unique MD:CTD interfaces and client binding residues ( red ) before sampling the ADP conformation and resetting the cycle to the apo state equilibrium . ( B ) Model for the unique energy landscape of TRAP1 . Solid lines illustrate the energy landscape of WT TRAP1 , and the dashed lines depict the change in landscape upon the loss of the extended N-terminal strap sequence in TRAP1 . By stabilizing both the apo and closed states , the strap increases the effective height of the energy barrier . This modulates the conformational landscape , and in the case of hTRAP1 provides pronounced temperature sensitivity . DOI: http://dx . doi . org/10 . 7554/eLife . 03487 . 019 Specific regulation of the energetic landscape imparted by the TRAP1 strap is depicted in Figure 8B . Here , we propose the effect of the strap ultimately impacts the kinetic barrier height as the strap stabilizes both the apo and closed states , although the apo state stabilization is dominant . Thus , addition of a structural element that makes analogous interactions in both the apo ( cis ) and closed ( trans ) states provides a novel strategy for kinetic regulation by accentuating the barrier between the apo and closed conformations . While Hsp90 is very highly conserved across species , there are several regions such as the N-terminus , the charge linker and the very C-terminus that have diverged significantly during evolution . As highlighted in Figure 7 , the different classes of Hsp90s segregate quite clearly according to the length of their N-termini , with the bacterial and yeast Hsp90s being the shortest , followed in turn by the metazoan cytosolic Hsp90s , the mitochondrial TRAP1s , and the ER Grp94s . One exception is a recently discovered Hsp90α alternative splice variant that creates a very large N-terminal extension of 122 residues . In keeping with observations here , biochemical analysis revealed that this extension is a negative regulator of ATPase activity ( Tripathi and Obermann , 2013 ) . In the TRAP1 family , conservation of the strap is strong through the known structured region ( His87 in zTRAP1 ) , but decreases towards the N-terminus , and is greatly reduced for TRAP1s from blood fluke , insects , and the sea urchin . Cytosolic Hsp90 has the same drop in conservation and a much shorter strap . By contrast , Grp94 has a very long and very well conserved strap region , with a somewhat variable , but very acidic N-terminus . Despite its long size , deleting the analogous strap region in Grp94 accelerates ATP hydrolysis by only fivefold , although temperature modulation was not investigated ( Dollins et al . , 2007 ) . However , its extreme length , the strong conservation , and the modest effect of deletion on ATPase rates , suggest a possible regulatory role that could couple other phenomena beyond temperature to the rate-limiting conformational changes required for ATP hydrolysis . The observation that the catalytic efficiency of different Hsp90s vary by ∼15-fold ( Richter et al . , 2008 ) suggest that regulation of the rate-limiting step has been highly tuned through evolution for functional importance . In support of this , yHsp90 mutations that accelerate or decelerate ATPase rates result in significant growth defects and loss of client protein folding in vivo ( Nathan and Lindquist , 1995; Prodromou et al . , 2000 ) . The evolution of additional residues at the N-terminus of the Hsp90 gene provides a convenient way to adapt the chaperone's conformational cycle to function with diverse clients encountered by the different homologs or under stressed environmental conditions . Additionally , while the cytosolic Hsp90s are highly regulated by several co-chaperones ( Zuehlke and Johnson , 2010 ) , only one co-chaperone has been identified for the organellar homologs ( Liu et al . , 2010 ) . This brings forth the possibility that the more extended strap in these homologs could directly or indirectly perform some of the regulation that co-chaperones provide to Hsp90 in the cytosol . The marked temperature sensitivity observed with TRAP1 raises the intriguing possibility that it represents a homeostatic response in mitochondria where heat is generated through uncoupling of the electron transport chain ( Rousset et al . , 2004 ) . In keeping with the physiological relevance , we demonstrate that the thermal sensitive kinetic barrier is measurably different between zebrafish and human TRAP1 , which have significantly different physiological temperatures and environments . Additionally , added contacts that the strap provides could be a target for post-translation modifications or even provide a novel binding site for ions , metabolites , or other factors that could modulate the regulatory functions of this element . These observations provide an example of how evolved extensions at the Hsp90 N-terminus can be used to fine-tune chaperone activity to match organism-specific environmental conditions or unique subcellular demands required for optimal function .
Full-length and mutant versions of TNF receptor-associated protein 1 ( TRAP1 ) from Homo sapiens and Danio rerio ( hTRAP1 and zTRAP1 , respectively ) were purified using our previously described protocol ( Lavery et al . , 2014 ) . The coding sequence of proteins used in this study were cloned into the pET151/D-TOPO bacterial expression plasmid ( Life Technologies , Grand Island , NY ) and mutant versions of were generated by standard PCR based methods . Cysteine-free hTRAP1 with encoded cysteine positions ( Glu140Cys or Glu407Cys ) on each or ( Ser133Cys and Glu407Cys ) on a single protomer , allowed for site-specific labeling with maleimide derivative Alexa Fluor 555/647 dyes ( Life Technologies ) for FRET experiments . These constructs were also purified as previously described ( Lavery et al . , 2014 ) , with a final size exclusion chromatography storage buffer of 50 mM Hepes pH 7 . 5 , 100 mM KCl , 500 μM TCEP . Aliquots of stored protein were labeled with fluorescent dyes as described below . WT hTRAP1 was initially diluted to 0 . 1 mg/ml in a buffer containing 20 mM NaH2PO4 pH 7 , 50 mM KCl , and 2 mM MgCl2 , 0 . 02% n-octyl-β-D-glucoside + 2 mM AMPPNP . Reactions were incubated at various temperatures for 1 hr ( or overnight ) , followed by dilution to 0 . 01 mg/ml in the buffer above including 2 mM AMPPNP to maintain nucleotide concentration . 5 µl of the resulting reactions was then incubated for ∼1 min on 400 mesh Cu grids ( Pelco , Redding , CA ) coated with a thin carbon layer ( ∼50–100 Å ) . Following sample incubation , the grid was washed 3× with miliQ water , and lastly stained 3× with uranyl formate pH 6 . The final stain was removed by vacuum until the surface of the grid was dry . Prepared grids were imaged with a TECNAI 12 ( FEI , Hillsboro , OR ) operated at 120 kV . Images were recorded using a 4k × 4k CCD camera ( Gatan , Pleasanton , CA ) at 52 , 000 magnification , at −1 . 5 μm defocus . Representative closed state particles were selected in EMAN ( Ludtke et al . , 1999 ) . TRAP1 homologs and mutant proteins were buffer exchanged into 20 mM Hepes pH 7 . 5 , 50 mM KCl , 2 mM MgCl2 , 1 mM DTT . 75 μM protein ( monomer concentration ) was used as the final concentration for all reactions , and 2 mM AMPPNP was added to initiate closure . Reactions were incubated at various temperatures for 1 hr followed by a spin at max speed in a tabletop centrifuge for 10 min immediately prior to data collection to remove any trace aggregation . Data were collected at the Advanced Light Source ( ALS ) at beamline 12 . 3 . 1 with sequential exposure times of 0 . 5 , 1 , and 0 . 5 s . Each sample collected was subsequently buffer subtracted and time points were averaged using scripts provided at beamline 12 . 3 . 1 and our own in-house software ‘saxs_multiavg . py’ . The scattering data were transformed to P ( r ) vs r using the program GNOM ( Svergun , 1992 ) and Dmax was optimized . The resulting distributions were fit using an in-house least squares fitting program ‘saxs_combine . py’ in the region where non-zero data were present for the target data and closed state model . For the fitting we chose theoretical scattering data for our TRAP1 closed-state model ( Lavery et al . , 2014 ) and the WT apo data for each TRAP1 homolog . The WT apo data were chosen as the best representation of apo for two reasons . ( 1 ) The apo state of Hsp90 proteins consist of a mix of conformations ( Southworth and Agard , 2008 ) of which the various conformations and percent of each remains to be elucidated for TRAP1 , and ( 2 ) removal of the strap ( particularly in hTRAP1 ) induces a shift of the apo distribution towards the closed state as observed for hTRAP1 by SAXS ( data not shown ) , which would result in a value of percent closed for the Δstrap protein that would under represent the true value relative to WT . The theoretical scattering curve for the TRAP1 crystal structure was generated in the program CRYSOL ( Svergun et al . , 1995 ) . The percent of components utilized in the fit and an R factor ( R_merge ) that is similar to a crystallography R factor in nature is output from our least-squares fitting program and values reported in Table 1 . R_merge is defined as the equation belowR_merge=Σ‖Pobs ( r ) |−|Pcalc ( r ) |/|Pobs ( r ) ‖where Pobs ( r ) is the observed probability distribution and Pcalc ( r ) is the calculated modeled fit . Both pieces of in-house software used for SAXS data analysis , ‘saxs_multiavg . py’ and ‘saxs_combine . py’ , have been deposited at GitHub . com ( https://github . com/agardd/saxs_codes ) . Steady-state kinetic measurements for various Hsp90 homolog and mutants were carried out in previously described conditions unless otherwise indicated ( Lavery et al . , 2014 ) . Specific buffer conditions used to measure kinetic rates for cysteine free zTRAP1 proteins used in EPR were 20 mM Hepes pH 7 . 4 , 150 mM NaCl , 2 mM MgCl2 at 23°C with 2 mM ATP ( see EPR method description ) . Buffer conditions used to measure kinetic rates for cysteine free hTRAP1 ( WT and ΔStrap ) used in FRET experiments were 50 mM Hepes pH 7 . 5 , 50 mM KCl , 5 mM MgCl2 with 2 mM ATP ( see FRET method description ) measured at 30°C . Results were plotted using the program R ( R Development Core Team , 2010 ) . Purified protein was labeled with maleimide derivative AlexaFluor 555 ( Donor ) and 647 ( Acceptor ) ( Life Technologies ) at fivefold excess over protein ( pre-mixed at 2 . 5-fold concentration each dye for dual labeled sample ) overnight at 4°C . Labeling reactions were then quenched with twofold β-mercaptoethanol over dye concentration and free dye was removed with desalting columns containing Sephadex G-50 resin ( illustra Nick Columns , GE Healthcare , Pittsburgh , PA ) . For FRET measurements using probes that monitor closure across the dimer ( Glu140Cys , Glu407Cys , ‘Inter FRET’ ) , labeled protein was mixed at a 1:1 ratio with a final concentration of 250 nM . For measurements with the probe that measures NTD rotation ( Ser133Cys and Glu407Cys , ‘Intra FRET’ ) , WT hTRAP1 was mixed in 20-fold excess over labeled protein ( 250 nM labeled protein:5 μM WT ) . Heterodimers for experiments with all FRET probes were formed at 30°C for 30 min in a reaction buffer consisting of 50 mM Hepes pH 7 . 5 , 50 mM KCl , 5 mM MgCl2 . Following heterodimer formation , closure was initiated by addition of 2 mM AMPPNP at various temperatures ( Figure 4 ) . To measure re-opening , 40 mM ADP was rapidly mixed with pre-closed reactions ( closed as in Figure 4 ) . For ATP hydrolysis experiments , closure was initiated with 2 mM ATP in a reaction buffer consisting of 50 mM Hepes pH 7 . 5 , 50 mM KCl . After closure was complete , hydrolysis was initiated by rapid addition of 5 mM MgCl2 . Closure and ATP hydrolysis experiments ( Figures 4 and 6B ) were carried out using a Jobin Yvon fluorometer with excitation and emission monochromator slits set to 2 nm/3 nm ( respectively ) , an integration time of 0 . 3 s , and excitation/emission wavelengths of 532/567 nm ( donor ) and 532/667 nm ( acceptor ) . Re-opening experiments ( Figure 6A ) was measured at 30°C using a SpectraMax5 plate reader with excitation and emission wavelengths as above and with a 540 nm emission cutoff . Kinetic measurements were taken at a time interval to minimize photobleaching . The change in FRET ( ratio of Donor and Acceptor fluorescence—division done to graph positive changes and normalized for visual comparison ) was well fit with a single exponential ( fit in KaleidaGraph , Synergy Software , Reading , PA ) to obtain the rate of closure and NTD rotation ( Fit 1 ) , as well as re-opening and ATP hydrolysis rates ( Fit 2 ) . Fit 1 : m1+m2∗ ( 1−exp∗ ( −m3∗x ) ) Fit 2 : m1+m2∗ ( exp∗ ( −m3∗x ) ) , where m1 is the time zero value , m2 is the amplitude , m3 is the rate constant , and x is time in seconds . For steady-state FRET scans ( taken before and after kinetic measurements ) , reactions were excited at 532 nm and emission was collected from 550–750 nm . FRET scans were normalized such that the area under the curve is 1 . The activation energy was calculated by fitting a plot of the natural log ( ln ) of the observed closure rate ( y-axis ) verses inverse temperature ( x-axis ) using the equation below ( Fit 3 ) Fit 3 : ln ( k ) =−Ea/RT+ln ( A ) , where Ea is the activation energy , T is temperature ( kelvin , K ) , R is the gas constant ( kcal K−1 mol−1 ) , and A is a pre-exponential factor . ATPase rates used to calculate Ea for Hsp90 homologs were taken from reference ( Frey et al . , 2007 ) . Cysteine-free zTRAP1 with a Ala201Cys mutation on the lid was exchanged into non-reducing EPR buffer ( 20 mM Hepes pH 7 . 4 , 150 mM NaCl ) at 100 μM ( monomer concentration ) and labeled by the addition of N- ( 1-oxyl- 2 , 2 , 6 , 6-tetramethyl-4-piperidinyl ) maleimide ( MSL , Sigma , St . Louis , MO ) to 2 . 5× concentration of protein overnight at 4°C . The labeled protein was then run through a Micro Bio-Spin column P-30 ( Bio-Rad , Hercules , CA ) to eliminate free probe . EPR spectra were obtained at ∼100 μM labeled protein ± 2 mM AMPPNP in the buffer above with addition of 2 mM MgCl2 and after heating at 30°C for 30 min to ensure closure has completed ( Figure 5A ) . For the time course ( Figure 5B ) , protein ( apo ) was spiked with 2 mM AMPPNP and EPR scans recorded overtime at room temperature ( ∼23°C ) . EPR measurements were performed with a Bruker EMX EPR spectrometer ( Bruker , Billerica , MA ) in a 50-μl glass capillary . First derivative X-band spectra were recorded in a high-sensitivity microwave cavity using 50-s , 10-mT-wide magnetic field sweeps . The instrument settings were as follows: microwave power , 25 mW; time constant , 164 ms; frequency , 9 . 83 GHz; modulation , 0 . 1 mT at a frequency of 100 kHz . Each spectrum used in the steady-state data analysis was an average of 10–20 sweeps from an individual experimental preparation , with one sweep used for kinetic measurements . Analysis of the raw peak heights indicated that both the mobile and immobile fractions were changing as a concerted single exponential process . As a consequence , to determine the rate constant , it was unnecessary to account for peak overlaps or the starting fraction in each state . To quantify , the raw peak heights at each time point were determined using the Bruker EMX EPR spectrometer software ( Bruker , Billerica , MA ) and converted to a percent change over the time course . The rates of lid closure for WT and Δstrap were estimated by fitting the normalized peak heights for each sample to a single exponential decay process with the same rate constant for the mobile and immobile peaks ( done as a constrained non-linear fit in Prism v6 , GraphPad software , La Jolla , CA ) . | Proteins—which are made of chains of molecules called amino acids—play many important roles in cells . Before a newly made protein can work properly , the amino acid chain has to be folded into the correct three-dimensional shape . Many proteins that have folded incorrectly are harmless , but some can disrupt the cell and cause damage . Although most proteins can fold properly on their own , they are often helped by ‘chaperone’ proteins , which speed up the process and encourage correct folding . Many chaperone proteins belong to a family called the heat shock proteins , which are found in almost all species: from bacteria , to plants and animals . High temperatures can severely impair and destabilize proper protein folding , and the heat shock proteins counteract this by helping to prevent , or correct , protein misfolding . Most animals and plants have at least four genes that make different versions of heat shock protein 90 ( Hsp90 ) . These versions work in different places in the cell and one—called TRAP1—is found in internal compartments called mitochondria . Along with its role in assisting protein folding , TRAP1 also acts as an indicator of the health of the proteins in the mitochondria . One section or ‘domain’ of Hsp90 is able to bind to and break down a molecule called ATP . This releases energy that is used to change the shape of the protein-binding domain—which is responsible for helping other proteins to fold . Recent studies of TRAP1 using a technique called protein crystallography highlighted the presence of a short amino acid tail or ‘strap’ at one end of the protein , but it is not known what role it may play in protein folding . In this study , Partridge et al . reveal that the amino acid strap of TRAP1 controls the breakdown of ATP in a way that depends on the surrounding temperature . Similar straps are also present in the Hsp90 proteins that are found in other parts of the cell . However , the strap is absent from the Hsp90 proteins of yeast and bacteria . These experiments used proteins that had been taken from living cells and placed in an artificial setting , so an important next step will be to study the role of the strap in the folding of proteins inside living cells . Also , future work could investigate the potential role of the protein in maintaining healthy mitochondria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2014 | A novel N-terminal extension in mitochondrial TRAP1 serves as a thermal regulator of chaperone activity |
Hippocampal neurons encode spatial memories by firing at specific locations . As the animal traverses a spatial trajectory , individual locations along the trajectory activate these neurons in a unique firing sequence , which yields a memory code representing the trajectory . How this type of memory code is altered in dementia-producing neurodegenerative disorders is unknown . Here we show that in transgenic rTg4510 mice , a model of tauopathies including Alzheimer's disease , hippocampal neurons did not fire at specific locations , yet displayed robust firing sequences as animals run along familiar or novel trajectories . The sequences seen on the trajectories also appeared during free exploration of open spaces . The spatially dissociated firing sequences suggest that hippocampal neurons in the transgenic mice are not primarily driven by external space but by internally generated brain activities . We propose that tau pathology and/or neurodegeneration renders hippocampal circuits overwhelmed by internal information and therefore prevents them from encoding spatial memories .
Alzheimer's disease ( AD ) is behaviorally characterized by cognitive declines such as memory loss ( Carlesimo and Oscar-Berman , 1992; Salmon and Bondi , 2009 ) , but pathologically defined by β-amyloid plaques and tau pathology including neurofibrillary tangles and tau-mediated neurodegeneration ( Wenk , 2003; Ashe and Zahs , 2010 ) . A current challenge is how to bridge between these two very different aspects of AD . Because memory deficits are necessarily produced by the memory-processing neural circuits in vivo ( Dickerson and Eichenbaum , 2010 ) , a key step toward this challenge is to understand the functional alterations in the memory circuits , including the hippocampus ( HP ) , of the living brain with ongoing pathogenesis . Both the β-amyloid and tau pathologies are likely involved in the memory loss ( Chapman et al . , 1999; Yu et al . , 2001; Cacucci et al . , 2008; Ashe and Zahs , 2010; Palop and Mucke , 2010 ) . In particular , tau pathology and subsequent neurodegeneration play a key role in mediating the memory symptoms seen at a mature stage of AD ( Ashe and Zahs , 2010 ) . Currently , it is unknown how the mnemonic functions of HP are altered by tau pathology in vivo . One well-studied function of HP in vivo is the coding of spatial memories . It is believed that spatial memories in both rodents and humans are encoded by HP place cells that fire at one or a few specific locations of a given space ( place fields ) ( O’Keefe and Dostrovsky , 1971; Wilson and McNaughton , 1993; Burgess and O'Keefe , 2003; Ekstrom et al . , 2003 ) . When the animal runs along a novel spatial trajectory , external sensory input at individual locations along the trajectory drives place cells to fire one after another in a unique sequence , which yields a memory code for the trajectory ( Wilson and McNaughton , 1993; Harris et al . , 2003; Dragoi and Buzsaki , 2006 ) . After the novel experience , the sequence becomes internally established and can be reactivated during later memory recall and/or consolidation with or without partial external input ( Buzsaki , 1989; Wilson and McNaughton , 1994; Skaggs and McNaughton , 1996; Nadasdy et al . , 1999; Lee and Wilson , 2002; Foster and Wilson , 2006; Diba and Buzsaki , 2007; Ji and Wilson , 2007; Gelbard-Sagiv et al . , 2008; Davidson et al . , 2009; Karlsson and Frank , 2009; Carr et al . , 2011 ) . Therefore , in theory , on any given trajectory HP could either retrieve stored internal sequences , form new ones , or combine both , based on the external input that HP receives from the environment . It has been proposed that the interplay between internal and external inputs is important to memory processing in HP and abnormal interaction may lead to memory interference and/or intrusion ( Nakazawa et al . , 2002; McHugh et al . , 2007; Colgin et al . , 2008 ) . The availability of tauopathy animal models provides an opportunity to study how tau pathology alters HP mnemonic functions in vivo . To this end , we set out to examine the spatial memory code in the transgenic rTg4510 ( Tau ) mouse . This model was chosen because its well-characterized pathological and behavioral phenotypes mimic many key features of tau pathology in AD ( Ramsden et al . , 2005; Santacruz et al . , 2005; Berger et al . , 2007 ) . In Tau mice , the overexpression of a mutated version ( P301L ) of the human tau gene restricted in the forebrain leads to age-dependent spatial memory deficits , tau neurofibrillary tangles , and neuronal loss . In this study , we recorded neurons in the CA1 region of HP using the tetrode recording technique ( Wilson and McNaughton , 1993; Buzsaki , 2004; Ji and Wilson , 2007 ) from 7 to 9 month old Tau mice and their control littermates while they performed spatial navigation tasks . We found that HP neurons in Tau mice fail to fire at specific spatial locations and yet maintain robust firing sequences , suggesting that non-spatial , internal brain activities dominate the HP circuits and prevent them encoding external space .
First , we asked whether CA1 place cell activities were altered in Tau mice when they ran along a familiar rectangular track ( Figure 2A ) two sessions a day , 15 min each session . In each session , the animals ran back and forth ( two trajectories ) and repeated 10–40 laps on each trajectory . We analyzed 87 putative pyramidal neurons from WT mice ( WT neurons ) and 93 from Tau mice ( Tau neurons ) that were active on at least one trajectory in at least one session . The median firing rate of Tau neurons during the track running sessions was not significantly different from that of WT neurons ( median and [10% 90%] range rate: WT 1 . 21 [0 . 43 3 . 60] Hz , Tau 0 . 90 [0 . 41 2 . 75] Hz , p=0 . 08 , ranksum-test , same below unless specified otherwise ) . 10 . 7554/eLife . 00647 . 005Figure 2 . Unstable firing locations of Tau neurons led to the loss of overall location-specificity on familiar trajectories . ( A ) The rectangular familiar track . F: food wells . ( B ) Lap-by-lap spike raster of three WT and three Tau neurons , each from a different animal , on one trajectory of the track . The trajectories were linearized and plotted as the x-axis ( arrows: running directions ) . Each tick represents a spike . Bottom curves: firing rates on the trajectories averaged across all laps . Note the unstable firing locations of the Tau neurons . ( C ) – ( E ) Distribution of lap SI ( C ) , trajectory SI ( D ) , and rate-stability ( E ) of WT and Tau neurons ( see text for definitions ) . Plots are histograms normalized by total numbers of samples , each computed for one neuron on one trajectory . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 005 As expected , WT neurons fired spikes at stable , specific locations in every lap that the animal traversed a trajectory . In contrast , spikes of Tau neurons were not confined to specific locations ( Figure 2B ) . Close visual inspections revealed that Tau neurons’ spikes were still specific to one or a few locations in each individual lap , but the locations were changed from lap to lap ( Figure 2B ) . We described a neuron’s firing profile on a trajectory by a rate curve ( the two trajectories were analyzed separately for each neuron ) , which was the neuron’s firing rate as a function of the position along the trajectory . The rate curve was computed for every lap and then averaged across all laps . Whereas the averaged rate curves of typical WT neurons clearly showed one or a few peaks indicative of well-defined place fields , those of typical Tau neurons did not show such prominent peaks ( Figure 2B , ‘bottom curves’ ) . We quantified the location-specificity using spatial information ( SI ) ( Skaggs et al . , 1993 ) . SI was computed for each neuron active on each trajectory at two levels , a lap SI to measure the location-specificity at individual lap level and a trajectory SI to measure the overall location-specificity across all laps based on its averaged rate curve . The median lap SI of WT neurons was only slightly ( 21% ) greater than that of Tau neurons ( Figure 2C; WT 2 . 03 [1 . 00 3 . 09] bits/spike , N = 263; Tau 1 . 68 [0 . 84 2 . 34] bits/spike , N = 262; p=7 . 7 × 10−11 ) , but the median trajectory SI of WT neurons was much greater ( 180% ) than that of Tau neurons ( Figure 2D; WT: 1 . 29 [0 . 55 2 . 61] bits/spike; Tau: 0 . 46 [0 . 15 1 . 29] bits/spike; p=5 . 3 × 10−39 ) . We also quantified the lap-to-lap location change by rate-stability , defined as the average correlation coefficient between any two laps’ rate curves . Tau neurons showed much lower rate-stability than WT neurons ( Figure 2E; WT: 0 . 66 [0 . 30 0 . 86]; Tau: 0 . 075 [0 . 0042 0 . 34]; p=1 . 7 × 10−70 ) . These results reveal that Tau neurons lost their overall location-specificity , mainly due to unstable firing locations . Because WT neurons fired at stable locations , they fired one after another with a stable , position-locked sequence in every lap of a trajectory ( Figure 3A ) . To our surprise , despite their unstable firing locations , Tau neurons still maintained stable firing sequences , that is , they fired with consistent orders across laps ( Figure 3A ) . 10 . 7554/eLife . 00647 . 006Figure 3 . Tau neurons maintained stable firing sequences on familiar trajectories . ( A ) Lap-by-lap spike raster of six WT neurons and that of five Tau neurons on a trajectory ( arrow: running direction ) . Each color-coded row within a lap represents a neuron . Each tick represents a spike . Only five laps are shown . Angled lines: detected firing sequences that matched with the templates derived on the trajectories ( see Figure 3—figure supplements 1 and 2 for details of template generation and sequence detection ) . Vertical lines: landmark positions ( three corners of the track in Figure 2A ) along the trajectory . Note that the Tau neurons shifted their firing locations , but maintained the same firing sequences across the laps . See more data from more animals with more laps in Figure 3—figure supplement 3 . ( B ) Mean number of detected sequences ( in Z-score ) on the track trajectories in WT and Tau mice . The robustness of this sequence detection result is examined in Figure 3—figure supplement 4 . ( C and D ) Mean number of sequences per lap ( C ) and mean location shift ( D ) for the detected sequences in WT and Tau mice . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 00610 . 7554/eLife . 00647 . 007Figure 3—figure supplement 1 . Deriving template sequences based on pair-wise cross-correlations . ( A ) First , we assigned letter identities ( bold letters ) to cells active on a trajectory . Spike raster of six WT and five Tau cells ( same as those in Figure 3A ) during a single running lap are plotted in time . Vertical lines: start and end times of the laps . ( B and C ) Second , we determined whether pairs of cells had stable firing relationships based on the stability of their cross-correlations ( CCs ) . For examples , the CCs of a pair of WT ( cell G and cell I in A ) and a pair of Tau cells ( cell A and cell H in A ) are shown in ( B ) . Each panel shows the color-coded CC for every lap and the bottom curve shows the lap-averaged CC . The color bar is shared by all panels in ( B ) and ( C ) . For every lap , a CC was generated by cross-correlating the two cells' binned firing rates ( bin size: 100 ms ) in the lap . Note the stable peak locations of the lap CCs and the prominent peak of the lap-averaged CC for both the WT and Tau pairs in ( B ) . For each cell pair , we computed a CC-stability , defined as the average correlation-coefficient between any two laps' CCs . To evaluate the significance of the CC-stability , we also computed a shuffled version of CCs ( C ) after the two cells' firing rates were circularly slid with independent , random time intervals ( slide-shuffling ) . Note the inconsistent peaks in the lap CCs of both the WT and Tau pairs and consequently small CC-stability values in ( C ) . For each cell pair , we generated 1000 shuffled CCs by slide-shuffling and therefore obtained 1000 chance level CC-stability values . A cell pair was considered to have a stable CC if ( 1 ) CC-stability is greater than the upper 1% of the chance level ( p<0 . 01 ) and ( 2 ) there was a peak in the lap-averaged CC ( the maximum value of the Z-transformed CC ≥ 1 ) . Third , we ordered the cells in those pairs with a stable CC . For the pairs in ( B ) , because the peak of the lap-averaged CC between cell G and I occurred at a negative time lag ( * ) , we ordered the pair as IG . Because the peak between cell A and H occurred at a positive time lag ( * ) , we ordered the pair as AH . If the peak of a pair occurred at exactly time lag 0 , we randomly ordered the cells in the pair . ( D ) Finally , we constructed template sequences that agreed with all the ordered pairs ( see ‘Materials and methods’ ) . The derived template sequences for the WT ( XIGQOZ ) and Tau ( FABHD ) cells are shown . Spike raster of the cells , ordered according to the templates , are displayed in the time domain during the same five laps as in Figure 3A . Note the repeated occurrence of the template sequences across laps . This also shows that firing sequences of WT and Tau cells in the time domain were very similar to that in the space domain ( Figure 3A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 00710 . 7554/eLife . 00647 . 008Figure 3—figure supplement 2 . Detecting matches with a template sequence . ( A ) An example template sequence ( FABHD ) made of five cells ( A , B , D , F , H ) . ( B ) The spike pattern of the five cells in a time period . Here we show how the spike pattern was searched through to find matches with the template sequence . First , the spike trains of the five cells ( top ) were transformed into instantaneous firing rate curves ( bottom ) using a Gaussian kernel ( σ = 1 s ) ( Ji and Wilson , 2007 ) . The local maxima of the temporal rate curves greater than a threshold ( here was the mean rate of a cell in the entire running session ) were detected ( ticks ) . Boxes: detected matches . ( C ) Second , the detected local maxima of all the cells were ordered into a raw sequence according to their times . Now the spike pattern was translated into a long raw sequence . Boxes: segments of the raw sequence that matched with the template sequence , as determined below . ( D ) Third , we searched through the raw sequence to find the segments that match with the template sequence . A match was defined by a significant rank order correlation . For example , given a segment FFAHBBD , we determined a rank for each letter according to its order in the template ( template rank ) and a rank according to its time ( time rank ) . The correlation coefficient of the two ranks was computed and its significance was assessed by using 1000 random shuffles of the letters in the segment ( shuffle p ) . In pilot computations , we found that the p value determined from Pearson's R ( p=0 . 0103 for this example ) was a close approximation of shuffle p ( 0 . 0095 for this example ) . To save computing time , we later used a threshold ( p<0 . 05 ) on the Pearson's p value to define matches . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 00810 . 7554/eLife . 00647 . 009Figure 3—figure supplement 3 . More examples of WT and Tau firing sequences on familiar trajectories . The first 15 laps are shown . Each panel ( A , B , or C ) was from a separate WT or Tau mouse , which was different from the one in Figure 3A . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 00910 . 7554/eLife . 00647 . 010Figure 3—figure supplement 4 . The robustness of the sequence detection result . We examined the robustness of our sequence detection result to answer two questions . ( A and B ) First , the template sequences were generated by using a threshold to select pairs of cells with consistently high correlations . How did the sequence detection result depend on the particular threshold chosen ? The number of templates ( A ) and mean number of detected sequences ( in Z-score ) ( B ) on familiar trajectories in WT ( black ) and Tau ( red ) mice were computed using a wide range of cell selection threshold ( 0 . 05–0 . 4 ) . Arrows: actual thresholds ( black: WT , red: Tau ) . Different actual thresholds were used for WT and Tau mice because the thresholds were computed from chance-level correlations ( Figure 3—figure supplement 1 ) , which differed between WT and Tau cell pairs . Dashed horizontal lines: actual Z-scores at the actual thresholds . *p<0 . 05 from the actual Z-scores . The Z-scores changed very slowly as the threshold was decreased . Only when the threshold became extremely low ( <0 . 08 ) , did the Z-scores drop significantly . ( C ) Second , Tau cells fired more broadly on the track , which might have increased temporal overlap in their firing patterns and yield more opportunities for sequence detection . Were the Z-scores sensitive to the possible overlapped firing of Tau cells ? To address this , we systematically down-sampled the spike trains during trajectory running and re-computed the Z-scores . We binned spikes of each cell into 1-s long bins and removed spikes within randomly chosen bins . This manipulation reduced the temporal overlap among cells at a time scale of our sequence detection . The result is plotted similarly as in ( B ) . It is clear that the actual Z-scores were insensitive to the down-sampling . Only when more than 60% spikes removed , did Z-scores start to drop significantly below the actual values . Therefore , although the absolute number of detected sequences on the track inevitably depended on the exact parameters chosen , the normalization ( Z-scoring relative to the chance level ) in our sequence detection method resulted in a robust quantification of the sequential structure in the multicell spiking patterns . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 010 To quantitatively demonstrate this , we analyzed the high-order ( ≥4 neurons ) firing sequences ( Lee and Wilson , 2002; Ji and Wilson , 2007 ) of WT/Tau neurons on the familiar track . We first assigned letter identities to those neurons active on a trajectory . We arranged pairs of neurons according to their firing orders identified by their cross-correlations . We then derived a template sequence for the trajectory that agreed with all the ordered pairs . For example , for the five Tau neurons shown in Figure 3A , the cross-correlation between the second ( assigned as neuron A ) and the fourth ( neuron H ) had a positive peak at 0 . 8 s . We ordered them as AH . By ordering all pairs among the five neurons in this manner , we derived a template sequence FABHD for this trajectory . A total of 13 template sequences were derived from WT mice and 8 from Tau mice ( Table 2 ) . For a template sequence on a trajectory , we then detected the firing sequences in the spiking patterns of all laps that matched with the template . To evaluate the significance of the detection , we generated 1000 shuffled copies of the spike patterns by randomly swapping neuron identities and then detected the number of sequences in each copy . The number of detected sequences in the real spike patterns was expressed as a Z-score computed from the mean and standard deviation of the shuffling-generated counts . We found a significant number of detected sequences for 9 out of the 13 WT templates and 6 of the 8 Tau templates ( p<0 . 05 , Z-test; Table 2 ) . The mean ( ±SEM ) Z-scores for all WT and all Tau templates were similarly high ( Figure 3B; WT: 2 . 5 ± 0 . 3; Tau: 2 . 3 ± 0 . 3; p=0 . 67 , one tailed t-test , same below unless specified otherwise ) . Both WT and Tau template sequences were detected in almost every lap ( Figure 3A ) . Although it appeared that Tau sequences occurred more frequently than WT sequences in a given lap ( Figure 3A ) , this difference was not significant ( Figure 3C; WT: 1 . 0 ± 0 . 1 times per lap; Tau: 1 . 4 ± 0 . 3 times per lap; p=0 . 10 ) . These results show that Tau neurons , like WT neurons , formed robust firing sequences . 10 . 7554/eLife . 00647 . 011Table 2 . High-order sequence analysis results for individual template sequences derived on the familiar trajectoriesDOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 011Template sequenceAnimal nameGenotypeSelf traj . Z ( S1T1 ) Z ( S1T2 ) Z ( S2T1 ) Z ( S2T2 ) Z ( O1 ) Z ( O2 ) AGFHJAN4WTS2T20 . 012 . 8−0 . 43 . 2NANAZSBPNAAN19WTS1T12 . 6−0 . 71 . 40 . 30 . 8−1 . 7JHLKZEAN19WTS1T21 . 62 . 7−0 . 41 . 8−0 . 10 . 6IYOBZENAAN19WTS2T11 . 92 . 33 . 51 . 30 . 72 . 7ZSaFbBAN19WTS2T2−0 . 31 . 70 . 91 . 91 . 90 . 3GHKJBDFAN20WTS1T11 . 61 . 22 . 41 . 0−1 . 80 . 9JHAKCAN20WTS1T2−0 . 61 . 3−0 . 32 . 10 . 02−0 . 7GHCKJBDFAN20WTS2T11 . 80 . 31 . 41 . 3−0 . 61 . 4GJAHLCAN20WTS2T20 . 51 . 10 . 40 . 8−1 . 8−0 . 3XIGQOZAN21WTS1T14 . 0−1 . 23 . 6−1 . 0−2 . 0−1 . 0IWQVHPAN21WTS1T2−0 . 83 . 0−1 . 23 . 30 . 20 . 0GOZTHRAN21WTS2T11 . 90 . 53 . 60 . 31 . 0−1 . 5OFINKDVAN21WTS2T2−0 . 93 . 80 . 53 . 3−0 . 5−1 . 4ACHFDGAN5TauS2T13 . 0−0 . 42 . 5−0 . 06−1 . 2−1 . 6FABHDAN13TauS1T14 . 0−0 . 12 . 92 . 02 . 12 . 9BFAHDAN13TauS2T11 . 71 . 12 . 1−0 . 90 . 040 . 5FEBAAN15TauS2T11 . 00 . 93 . 2−0 . 42 . 61 . 9ILFBKAN22TauS1T11 . 40 . 5−0 . 8−0 . 81 . 2−0 . 5FLBDICAN22TauS2T10 . 11 . 42 . 20 . 72 . 22 . 9PKLEDOAN25TauS1T21 . 11 . 90 . 43 . 32 . 42 . 4PKTLEOMAAN25TauS2T20 . 71 . 80 . 01 . 33 . 41 . 9Template sequence: each letter ( case-sensitive ) represents a cell; same letters across different templates but within the same animal represent the same cells . Self traj . : The trajectory where the template was derived . Z: Number of detected matches ( in Z-score ) with a template sequence on a trajectory or in an open box session . S1T1: Trajectory one of session one , S1T2: Trajectory two of session one , so forth; O1: First open box session , so forth; numbers on self-trajectories are in bold; numbers that are statistically significant ( p<0 . 05 ) are in red . NA: not available ( not all experiments included two trajectory running sessions and two open box sessions ) . However , there was a major difference between WT and Tau sequences . Whereas WT template sequences predominantly occurred at similar locations across laps , Tau sequences shifted their locations from lap to lap ( Figure 3A ) . The mean location shift of the detected Tau template sequences was significantly greater than that of WT sequences ( Figure 3D; WT: 12 . 7 ± 1 . 9 cm; Tau: 31 . 3 ± 6 . 2 cm; p=0 . 0035 ) . Therefore , unlike WT sequences , Tau sequences were not anchored to specific spatial locations . We have shown that CA1 neurons maintained robust firing sequences during track running , even though they did not fire at specific locations . We next asked whether this was true during a very different behavior in a different familiar space . After completing the track running sessions , we kept recording the same CA1 neurons while the mice freely explored a familiar open box ( Figure 4A ) for one to two sessions , 15 min each session . 10 . 7554/eLife . 00647 . 012Figure 4 . Tau neurons fired with low location-specificity in a familiar open box . ( A ) The open box with its interior color and cue card ( Cue ) shown . ( B ) Color-coded firing rate maps of three WT and three Tau neurons , each from a different animal , in the open box . Numbers: peak ( red/black ) rates in Hertz . Note the broader firing areas of Tau neurons than those of WT neurons . ( C ) Distribution of open SI of WT and Tau neurons . Plots are histograms normalized by total numbers of samples , each computed for one neuron in one open box session . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 012 We analyzed 74 WT and 68 Tau neurons that were active in at least one open box session . The median firing rate of these Tau neurons was lower than that of WT neurons ( WT 1 . 3 [0 . 5 3 . 8] Hz , Tau 0 . 9 [0 . 5 2 . 7] Hz , p=0 . 0001 ) , which would predict a greater location-specificity of Tau neurons . However , we observed the opposite . As illustrated by their rate maps ( Figure 4B ) , firing rate color-plotted at each position of the two dimensional ( 2D ) open box , firing activities of Tau neurons covered a much larger portion of the box ( meaning less location-specific ) than those of WT neurons . The location-specificity in the box was quantified by SI in the 2D open space ( open SI ) for each neuron active in each open box session . The median open SI of WT neurons was significantly greater than that of Tau neurons ( Figure 4C; WT: 0 . 24 [0 . 04 0 . 76] , N = 131; Tau: 0 . 11 [0 . 04 0 . 34] , N = 119; p=1 . 9 × 10−7 ) . The result indicates that Tau neurons fired with low location-specificity in the familiar open box . We then examined whether there were organized firing sequences of multiple neurons during the open box sessions . Because of the random nature of the free exploration behavior , unlike the repeated trajectory running on the track , we did not expect repetitive high-order firing sequences to appear in the open box . However , when we arranged the neurons according to their firing orders in the template sequences derived during track running , it became apparent that Tau neuron sequences , but not WT sequences , during the track running also appeared frequently during the open box sessions ( Figure 5A ) . We quantified this observation by searching through the spike patterns of open box sessions for sequences that matched with the trajectory template sequences . Here we refer ‘a search’ as to matching one open box session to one trajectory template . We found a significant number of sequences in 10 out of the 16 searches in Tau mice , whereas only in 2 of the 24 searches in WT mice ( p<0 . 05 , Z-test; Table 2 ) . The mean Z-score of all searches in Tau mice was significantly higher than that in WT mice ( Figure 5B; WT: −0 . 1 ± 0 . 3; Tau: 1 . 4 ± 0 . 4; p=0 . 001 ) . Therefore , Tau sequences , but not WT sequences , seen on the track trajectories also appeared robustly during the open box sessions . 10 . 7554/eLife . 00647 . 013Figure 5 . Tau , but not WT , sequences seen on the familiar trajectories also appeared in the familiar open box . ( A ) Spike patterns of the same WT and Tau neurons as in Figure 3A within two time periods of an open box session . Neurons are color-coded . Each tick represents a spike . Angled lines: sequences detected as matches with the Tau sequences shown in Figure 3A . See more data from more animals in Figure 5—figure supplement 1 . ( B ) Mean number ( in Z-score ) of detected WT and Tau sequences in open box sessions . ( C ) Running paths of a Tau mouse for all the detected sequences in one open box session . For every sequence , the animal’s position at its start/end time was marked by a blue/red dot and its running path in between by a purple line . Gray dots: positions of the animal during the entire session . Note that the sequences occurred everywhere in the box . ( D ) Vector representation of the paths in panel ( C ) . For each path , a vector was drawn from its start to end location and was plotted with the start location translated to the origin . Note that there was no preferred direction for the vectors . ( E ) Rayleigh p values for individual searches in open box sessions of Tau mice . Each dot represents a search for one trajectory template sequence in one open box session . Bar: mean values . Red line: p=0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 01310 . 7554/eLife . 00647 . 014Figure 5—figure supplement 1 . More examples of WT and Tau firing patterns in the familiar open box . Neurons in each panel ( A , B , or C ) were the same ones plotted in the corresponding panel ( A , B , or C ) of Figure 3—figure supplement 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 014 Although the animals were allowed to move around freely in the open box , it is possible that some mice ran certain paths repeatedly and resulted in repeatedly occurring firing sequences . To examine this possibility , we plotted the animal’s paths between the start and end times of all detected Tau sequences in a search . These paths appeared to be randomly distributed in the box ( Figure 5C ) , suggesting that the detected sequences were not linked to particular locations or a particular path . To quantitatively demonstrate this , we drew a vector from the start to the end location of each path . We then plotted all the vectors resulted from a search with their starting positions aligned at the origin ( Figure 5D ) . If the paths were not randomly distributed , the vectors should show a preferred direction . However , the vectors of a typical search were randomly orientated ( Figure 5D ) . For each search in Tau mice where a significant number of sequences were found , we computed a Rayleigh p value , which tested a null hypothesis that the vectors had no preferred direction . None of the p values for all the 10 searches was significant ( Figure 5E , p≥0 . 07 ) , indicating that the null hypothesis could not be rejected . Therefore , the detected Tau sequences in open box sessions were not due to repetitive running behavior . Together with the low location-specificity of Tau neurons in the open box sessions ( Figure 4 ) , we concluded that Tau sequences seen in the open box were not primarily driven by spatial locations . Even further , these Tau sequences seen in the open box were the same as those during the track running . Therefore , Tau sequences did not distinguish two completely different spaces during two very different behaviors . This is a striking result given the well-known phenomenon of ‘remapping’ of place cell activities between different environments ( Muller and Kubie , 1987; Leutgeb et al . , 2005; Colgin et al . , 2008 ) . We have demonstrated that Tau neurons displayed robust firing sequences on a familiar track and in a familiar open box despite their apparent lack of spatial specificity . To understand whether this peculiar feature of Tau neurons requires spatial experience , we also recorded CA1 neurons while mice explored a novel open box for one to two sessions and then ran back and forth ( two trajectories ) on a novel track for one to two sessions in a novel room . The findings were similar as those seen in the familiar environments . On the novel track ( Figure 6A ) , 79 WT neurons and 92 tau neurons were active on at least one of its trajectories . Their median firing rates during the novel track running sessions were not significantly different ( WT: 1 . 4 [0 . 4 2 . 7] Hz , Tau: 1 . 1 [0 . 4 3 . 3] Hz , p=0 . 71 ) . Whereas the firing locations of WT neurons became stabilized quickly on the novel trajectories , those of Tau neurons stayed unstable from lap to lap ( Figure 6B ) . As a result , the median trajectory SI of Tau neurons was much lower than that of WT neurons ( WT 0 . 69 [0 . 23 1 . 5] bits/spike , Tau 0 . 28 [0 . 086 0 . 68] bits/spike , p=5 . 9 ×1 0−31 ) , despite a similar median lap SI ( WT 1 . 2 [0 . 58 1 . 8] bits/spike , Tau 1 . 2 [0 . 46 1 . 6] bits/spike , p=0 . 62 ) . The rate-stability of Tau neurons was much lower than that of WT neurons ( WT 0 . 47 [0 . 17 0 . 85] , Tau 0 . 10 [0 . 03 0 . 28] , p=1 . 6 × 10−51 ) . In the novel open box ( Figure 6C ) , 74 WT neurons and 61 Tau neurons were active in at least one session . The median firing rates of these WT and Tau neurons during the novel box sessions were not significantly different ( WT: 1 . 1 [0 . 4 2 . 9] Hz , Tau: 1 . 0 [0 . 5 2 . 6] Hz , p=0 . 38 ) . The rate maps of Tau neurons showed much broader firing areas than those of WT neurons ( Figure 6D ) , suggesting a reduced location-specificity in Tau neurons . This was confirmed by the result that the median open SI of Tau neurons was significantly lower than that of WT neurons ( WT: 0 . 28 [0 . 06 0 . 71] Hz , Tau: 0 . 08 [0 . 02 0 . 32] Hz , p=1 . 3 × 10−12 ) . These results show that Tau neurons fired with little location-specificity either on the novel trajectories or in the novel open box . 10 . 7554/eLife . 00647 . 015Figure 6 . Tau neurons formed robust firing sequences on a novel track and in a novel open box , despite their low location-specificity . ( A ) The L-shaped novel track . F: food wells . ( B ) Lap-by-lap spike raster of a WT and a Tau neuron on one trajectory of the track ( see Figure 2 for details ) . ( C ) The novel open box with its interior color and cue card ( Cue ) shown . ( D ) Firing rate maps of a WT and a Tau neuron on the open box . Numbers: peak ( red/black ) rates in Hertz . ( E ) Examples of WT and Tau sequences ( see Figure 3 for details ) on one trajectory of the L-track . ( F ) Spike patterns of the same WT and Tau neurons within a time period of an open box session . ( G and H ) Mean number of sequences ( in Z-score ) ( G ) and mean location shift ( H ) for the sequences detected on the novel trajectories in WT and Tau mice . ( I ) Mean number of sequences ( in Z-score ) detected during the novel open box sessions in WT and Tau mice . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 015 Nevertheless , Tau neurons , like WT neurons , fired with robust sequences on the novel trajectories . But unlike WT neuron sequences , Tau sequences on the novel trajectories were not anchored to specific locations ( Figure 6E ) and same sequences also appeared in the novel open box sessions ( Figure 6F ) . We derived 15 high-order template sequences on the novel trajectories from WT mice and 8 from Tau mice ( Table 3 ) . The mean Z-scores of WT and Tau template sequences on the novel trajectories were similarly high ( Figure 6G , WT: 2 . 5 ± 0 . 2 , Tau: 2 . 4 ± 0 . 4 , p=0 . 91 ) , but the mean location shift of Tau sequences among laps was greater than that of WT ones ( Figure 6H; WT: 9 . 2 ± 0 . 9 cm , Tau: 18 . 5 ± 3 . 6 cm , p=0 . 0038 ) . The mean location shift values of both WT and Tau sequences shown here were smaller than those on the familiar track , probably due to the shorter length of the novel track . Finally , Tau template sequences were detected in the novel open box sessions significantly more often than WT templates ( Figure 6I , mean Z-scores: WT: 0 . 5 ± 0 . 2 , N = 30 searches; Tau: 1 . 5 ± 0 . 3 , N = 12; p=0 . 0029 ) . Here we emphasize that the novel open box sessions preceded the novel track running . Therefore , Tau sequences seen on the novel trajectories were already preexistent in the prior open box sessions . These findings suggest that the formation of Tau sequences did not primarily depend on spatial experience in an environment . 10 . 7554/eLife . 00647 . 016Table 3 . High-order sequence analysis results for individual template sequences derived on the novel trajectoriesDOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 016Template sequenceAnimal nameGenotypeSelf traj . Z ( S1T1 ) Z ( S1T2 ) Z ( S2T1 ) Z ( S2T2 ) Z ( O1 ) Z ( O2 ) IHFGEAN18WTS1T13 . 40 . 11 . 40 . 21 . 01 . 1HEIFGAN18WTS1T20 . 44 . 3−1 . 42 . 71 . 6−0 . 8HICFBGAN18WTS2T21 . 93 . 5−0 . 33 . 62 . 50 . 8NAOJBDAN19WTS1T12 . 5−1 . 51 . 40 . 10 . 1−0 . 1OKDHFBAJNAN19WTS1T2−0 . 11 . 6−1 . 00 . 71 . 10 . 1LANOBAN19WTS2T13 . 1−0 . 63 . 0−1 . 1−0 . 5−0 . 4OKFCNAAN19WTS2T2−0 . 41 . 3−0 . 41 . 71 . 60 . 1ISRPBGJXKQAYAN20WTS1T11 . 5−1 . 62 . 01 . 20 . 9−1 . 1BIPRSAN20WTS1T2−0 . 11 . 10 . 2−0 . 30 . 30 . 2CTDOISREBXGAN20WTS2T11 . 8−0 . 51 . 8−0 . 51 . 10 . 6SRUPNLAN20WTS2T2−0 . 51 . 81 . 02 . 80 . 20 . 9IXPHNSCAN21WTS1T12 . 5−1 . 53 . 20 . 0−0 . 7−1 . 1PXHYTAN21WTS1T20 . 31 . 8−0 . 4−0 . 11 . 5−0 . 1KIJXSCAN21WTS2T12 . 8−0 . 52 . 70 . 40 . 41 . 1UMNSEOAN21WTS2T2−1 . 6−0 . 41 . 12 . 51 . 70 . 7HGPOLAN14TauS1T11 . 7−0 . 11 . 1−1 . 41 . 4−0 . 3GEDMHCBAN15TauS1T13 . 31 . 72 . 81 . 60 . 53 . 1DMHCBAN15TauS2T14 . 21 . 93 . 8−0 . 40 . 80 . 7GEDMAN15TauS2T23 . 32 . 52 . 92 . 11 . 43 . 1ABDEAN24TauS1T12 . 63 . 0NANA1 . 3NADEABIAN24TauS1T24 . 63 . 5NANA2 . 6NASGHERAN25TauS1T12 . 31 . 4NANA2 . 0NAKMFGHEAN25TauS1T21 . 02 . 3NANA1 . 4NASee Table 2 for details . Does the lack of location-specificity and experience-dependence mean that external space made no contribution to Tau neurons’ firing activities at all ? To answer this , we first examined whether the firing locations of Tau neurons were completely random . Visual inspection of the lap-by-lap spike trains of Tau neurons revealed that the firing locations of Tau neurons were often consistent among a few laps on both the familiar ( Figure 2B ) and novel ( Figure 6B ) trajectories , suggesting that the locations were not completely random . Indeed , randomly sliding rate curves of individual laps further significantly reduced the median rate-stability of Tau neurons on both familiar ( −0 . 0020 [−0 . 034 0 . 042] , p=9 . 3 × 10−49 compared with actual rate-stability ) and novel ( −0 . 0020 [−0 . 032 0 . 049] , p=1 . 8 × 10−60 ) trajectories , indicating that spatial locations still modulated Tau neurons’ firing activities . Second , we examined whether Tau neuron sequences were trajectory-selective . On both the familiar and novel tracks , most animals ran two different trajectories ( back and forth between two food wells ) in each of the two running sessions . For each template sequence derived on a trajectory in a session , we counted the occurrence of the sequence ( in Z-scores ) not only on the trajectory itself ( Self ) , but also on the same trajectory ( ST ) in the other session and on the other trajectory ( OT ) in the same session . On the familiar track , the mean Z-scores of Self and ST were similar for both WT ( Self 2 . 5 ± 0 . 3 , N = 13 templates; ST 2 . 3 ± 0 . 2 , N = 13; p=0 . 53 ) and Tau ( Self 2 . 3 ± 0 . 3 , N = 8; ST 1 . 6 ± 0 . 5 , N = 8; p=0 . 27 ) sequences , but there was a significant difference between Self and OT for both WT ( OT 0 . 29 ± 0 . 3 , N = 13; p=5 . 8 × 10−6 ) and Tau ( OT 0 . 10 ± 0 . 2 , N = 8; p=6 . 1 × 10−5 ) sequences ( Figure 7A ) . The result shows that Tau sequences , like WT sequences , occurred consistently on same trajectories across different sessions and distinguished between different trajectories within same sessions . On the novel track , whereas the mean Z-scores of Self and ST remain comparable for both WT ( Self 2 . 5 ± 0 . 2 , N = 15; ST 1 . 7 ± 0 . 3 , N = 15; p=0 . 06 ) and Tau ( Self 2 . 4 ± 0 . 4 , N = 8; ST 2 . 6 ± 0 . 6 , N = 4; p=0 . 79 ) sequences , there was a significant difference between Self and OT only for WT ( OT −0 . 26 ± 0 . 2 , N = 15; p=3 . 0 × 10−9 ) , but not for Tau ( OT 1 . 9 ± 0 . 5 , N = 8; p=0 . 51 ) sequences ( Figure 7B ) , suggesting that Tau sequences , unlike WT sequences , did not distinguish the two trajectories on the novel track . This analysis shows that Tau sequences were originally not trajectory-selective on the novel track , but became so on the familiar track . Therefore , although Tau neuron’s firing activities were not location-specific and Tau sequences were not primarily driven by space , there was still a modulation of Tau neuron’s firing activities by spatial experience and spatial trajectory . 10 . 7554/eLife . 00647 . 017Figure 7 . Tau sequences were trajectory-selective on the familiar track , but not on the novel track . ( A and B ) Mean number of sequences on three types of trajectories ( Self , ST , OT ) of the familiar ( A ) and novel ( B ) track . For each template sequence derived on a trajectory of a session , we computed the number of sequences ( in Z-score ) that matched with the template on the trajectory itself ( Self ) , on the same trajectory ( ST ) but in the other session , and on the other trajectory ( OT ) in the same session . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 017 Since same Tau sequences sometimes occurred multiple times in a single lap ( Figure 3A ) , we asked whether they tended to appear periodically in space or time . For all the sequences matched with a template in a session , we visualized the spatial/temporal gaps between neighboring sequences by a raster plot and computed the spatial/temporal auto-correlogram of the sequences using their locations/times to examine the periodicity ( Figure 8 ) . Neighboring WT sequences in track sessions displayed a consistent spatial gap and there were periodic peaks in their spatial auto-correlograms ( Figure 8A , left ) . The spatial period was the same as the track length ( ∼2 m for the familiar track , ∼1 m for the novel track ) , which was expected , given the repetitively running of same track trajectories . The periodicity was much less obvious in the temporal domain ( Figure 8A , right ) , because the amount of time that animals spent running and at the food sites varied from lap to lap . In contrast , Tau sequences during track sessions did not show any consistent spatial or temporal gaps in their raster plots and no obvious , periodic peaks were observed in either the spatial or temporal auto-correlograms ( Figure 8B ) . The spatial periodicity at the track length disappeared apparently because Tau sequences were not anchored to track locations . The average auto-correlogram across all sessions confirmed this observation on both the familiar and novel tracks ( Figure 8C , D ) . Using the same analysis , Tau sequences in open box sessions did not show apparent spatial or temporal periodicity either ( Figure 8B , E ) . Therefore , there was no strong evidence for a periodic occurrence of Tau sequences . 10 . 7554/eLife . 00647 . 018Figure 8 . Tau sequences did not display obvious periodicity in space or time . ( A ) The raster and auto-correlogram ( Auto crr ) of all sequences matched with a template during a track session of a WT mouse , plotted in the spatial ( left ) and temporal ( right ) domain . Each tick represents a sequence . For each sequence , we aligned its location/time at 0 and plotted the sequences occurring before and after at corresponding location/time gaps ( top ) . The histogram of this raster plot is equivalent to the auto-correlogram of the sequence locations/times ( bottom ) . The auto correlation value at 0 was truncated . Note the regular intervals in the spatial raster plot and periodic peaks in the spatial auto-correlogram . ( B ) Same as ( A ) , but for the sequences matched with a template during a track session ( Track ) and an open box session ( Open ) of a Tau mouse . Note the inconsistent spatial intervals and the disappearance of the peaks in the track spatial auto-correlogram . ( C ) The average sequence spatial auto-correlograms during familiar ( Fam; WT: N = 9 templates; Tau: N = 6 ) and novel ( Nov; WT: N = 12 templates; Tau: N = 7 ) track sessions of WT and Tau mice . ( D ) Same as ( C ) , but in the temporal domain . ( E ) The average sequence auto-correlograms of Tau mice in familiar ( Fam; N = 10 searches ) and novel ( Nov; N = 4 ) open box sessions in the spatial ( top ) and temporal domain ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 018 We have shown the existence of robust Tau sequences that were not anchored to specific locations and even did not distinguish different environments . Is it possible that these sequences were produced simply by some peculiar behaviors of Tau mice ? Track running and free exploration of open boxes are simple behavioral tasks that both WT and Tau mice were able to perform . Figure 9A , B show the accumulative running paths and linearized individual laps of a WT and Tau mouse on the familiar rectangular track . Figure 9C shows the accumulative running paths of the same mice in the familiar open box . As seen in these raw plots , the overall behavior of the WT and Tau mouse was similar . We then quantified the behavior using the following parameters . The first , most straightforward parameter is the running speed during the running of a track trajectory or during an open box session . In addition , for each trajectory of a track , we measured the animal’s performance by the number of laps on the trajectory and quantified the quality of the running by computing the cumulative travel distance per lap , the number of stops ( defined as speed <4 cm/s lasting ≥2 s ) per lap , and the stopping duration per lap . These parameters were computed for 54 familiar ( WT: 22 , Tau: 32 ) and 46 novel ( WT: 20 , Tau: 26 ) trajectories and 27 familiar ( WT: 11 , Tau: 16 ) and 25 novel ( WT: 12 , Tau: 13 ) open box sessions . The results are shown in Figure 9D–H . On familiar trajectories , Tau mice completed less number of laps and ran with a slower speed than WT mice , but their running quality as measured by stopping and travel distance was similar . On novel trajectories , Tau mice behaved similarly as WT mice or even better on one measure ( less number of stops per lap ) . In open boxes , Tau mice in general ran slightly faster than WT mice , but the difference was significant only in the novel open box . Since robust Tau sequences were found on both the familiar and novel tracks and these sequences appeared in both the familiar and novel open boxes , it is unlikely that the running behavior per se was responsible for the differences between WT and Tau sequences . 10 . 7554/eLife . 00647 . 019Figure 9 . Behavioral quantifications of WT and Tau mice . ( A ) The accumulative running paths of a WT and Tau mouse on the familiar rectangular track . ( B ) The paths in ( A ) were linearized lap-by-lap on the two track trajectories ( ascending and descending lines ) and plotted against time . The positions at the food wells were truncated . ( C ) The accumulative running paths of the WT and Tau mouse during an open box session . ( D–G ) the mean number of laps per session ( D ) , mean travel distance per lap ( E ) , mean number of stops per lap ( F ) , and mean stopping duration per lap ( G ) for WT and Tau mice during familiar and novel track sessions . ( H ) Mean running speed of WT and Tau mice during familiar and novel track ( Track ) and open box ( Open ) sessions . *p<0 . 05; **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 019 Theta ( 6–12 Hz ) oscillation in the CA1 local field potentials ( LFPs ) is associated with normal place cell activities ( Buzsaki , 2002 ) and is known to organize multiple place cells into firing sequences at theta time scale ( O’Keefe and Recce , 1993; Skaggs et al . , 1996; Dragoi and Buzsaki , 2006 ) . We next asked whether theta oscillations were altered in Tau mice . As seen in Figure 10A , the raw and filtered LFP traces of a Tau mouse appeared similar to those of a WT mouse , except that the overall magnitude of the Tau mouse’s LFPs was reduced , which is expected because of the massive neurodegeneration . Power spectral density ( PSD ) analysis of the raw traces showed a clear peak within the theta band in the LFP PSDs of both mice , but with a reduced amplitude in the Tau mouse’s ( Figure 10B , left ) . However , when the PSD values were normalized by the total LFP power , the peak of the Tau mouse’s PSD was comparable with that of the WT mouse ( Figure 10B , right ) , indicating the presence of prominent theta oscillations in the LFPs of the Tau mouse . 10 . 7554/eLife . 00647 . 020Figure 10 . Theta oscillations were present in Tau mice . ( A ) Raw and theta-filtered ( 6–12 Hz ) LFP traces recorded from a WT and Tau mouse . Black bars: time windows during which a WT and a Tau firing sequence were detected . The raw and theta-filtered LFP traces within the time windows are expanded on the bottom . Note the similar theta oscillations between the WT and Tau mouse , but with different scale bars . ( B ) The absolute ( left ) and normalized ( right ) PSD curves of the raw LFP traces shown in ( A ) . Black: WT; Red: Tau . Arrows: peaks in the theta frequency band . ( C–E ) mean peak theta frequency ( C ) , mean absolute theta power ( D ) , and mean normalized theta power ( E ) of raw LFPs recorded from WT and Tau mice and from mice with two other control genotypes ( OC; +/− and −/+ combined ) , during familiar and novel track running ( Track ) and open box sessions ( Open ) . Numbers: number of samples . ( F–H ) same as in ( C ) – ( E ) , but for raw LFPs within the start and end times of detected WT and Tau firing sequences . #p<0 . 05; *p<0 . 01; **p<0 . 001; ***p<0 . 0001 . For ( C ) – ( E ) , the significance threshold was lowered to 0 . 01 because of the multiple ( 3 ) comparisons within a group and therefore only those comparisons with p<0 . 01 were marked . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 020 We quantified the peak frequency and the absolute and normalized theta power for each LFP trace of a recording session . These parameters were compared among WT , Tau , and mice of two other control genotypes ( +/− , −/+ , see ‘Materials and methods’ ) , which showed no obvious tau-mediated neurodegeneration in CA1 ( Figure 1 and Table 1 ) . There was a slight reduction in theta peak frequency in Tau mice ( ∼8 . 5 Hz , compared with ∼9 . 5 Hz in control mice ) during familiar track and open box sessions . This difference became even less obvious in novel sessions ( Figure 10C ) . The absolute theta power was apparently reduced in Tau mice from those of control animals ( Figure 10D ) . However , the normalized theta power in Tau mice was significantly lower than that in control mice only during familiar track sessions , but not during open box sessions or novel track sessions ( Figure 10E ) . We also restricted this analysis on LFPs within the start and end times of detected sequences in WT and Tau mice . The observation was similar ( Figure 10F–H ) . Therefore , although LFPs in Tau mice were overall smaller , prominent theta oscillations were still present and qualitatively similar to those of WT mice , suggesting that abnormal theta oscillation is unlikely the cause of the difference between Tau and WT sequences . We also considered additional factors that may affect Tau neuron firing activities . First , we analyzed whether there was a gender difference in Tau mice . Rather than the sequence-related parameters , we used the rate-stability on track trajectories for this purpose because it allowed the comparison between male and female mice with sufficient samples ( Figure 11A , B ) . Between those neurons recorded from male and those from female WT mice , there was no significant difference in the mean rate-stability either on familiar ( male 0 . 70 [0 . 33 0 . 88] , female 0 . 65 [0 . 29 0 . 86] , p=0 . 21 ) or novel ( male 0 . 54 [0 . 19 0 . 83] , 0 . 46 [0 . 16 0 . 86] , p=0 . 22 ) trajectories . Between those from male and female Tau mice , the mean rate-stability was similar on novel trajectories ( male 0 . 10 [0 . 023 0 . 26] , female 0 . 10 [0 . 037 0 . 31] , p=0 . 51 ) . But on the familiar trajectories , the mean rate-stability of the neurons from female Tau mice was significantly greater than that of male Tau mice ( male 0 . 047 [−0 . 0035 0 . 14] , female 0 . 13 [0 . 031 0 . 42] , p=5 . 5 × 10−16 ) . For both genders and on both the familiar and novel trajectories , the mean rate-stability of WT neurons was always much greater than that of Tau neurons ( p≤1 . 3 × 10−14 ) . These results indicate that the firing stability of CA1 neurons was severely compromised in Tau mice of both genders , but CA1 neurons in female Tau mice did better on familiar trajectories than those in male Tau mice . 10 . 7554/eLife . 00647 . 021Figure 11 . Gender and genotype differences . ( A and B ) Distributions of rate-stability for WT and Tau neurons recorded from male and female mice on familiar ( A ) and novel ( B ) trajectories . ( C and D ) Distributions of rate-stability for neurons recorded from mice with other two control genotype ( +/− , −/+ ) on familiar ( C ) and novel ( D ) trajectories . Plots are histograms normalized by total numbers of samples ( numbers shown ) , each computed for one neuron on one trajectory . DOI: http://dx . doi . org/10 . 7554/eLife . 00647 . 021 Second , we analyzed an additional 163 neurons recorded from four mice with two other control genotypes ( +/− , −/+ ) . We found that they were much more stable than Tau neurons and largely similar to WT neurons . On familiar trajectories ( Figure 11C ) , the median rate-stability values of both +/− and −/+ neurons were not different from that of WT ones ( +/−: 0 . 62 [0 . 35 0 . 80]; −/+: 0 . 63 [0 . 39 0 . 85]; p≥0 . 11 compared with WT ) . Both were significantly greater than that of Tau neurons ( p≤1 . 5 × 10−30 ) . On novel trajectories ( Figure 11D ) , the median rate-stability of −/+ neurons was not different from , but that of +/− was slightly smaller than , that of WT neurons ( +/− 0 . 41 [0 . 090 0 . 69] , −/+ 0 . 49 [0 . 21 0 . 76] , p=0 . 016 between +/− and WT , p=0 . 90 between −/+ and WT ) . Both were significantly greater than the median rate-stability of Tau neurons ( p≥1 . 4 × 10−21 ) .
We have examined the HP spatial memory code in a tauopathy mouse model , the transgenic rTg4510 mice . Our data show that CA1 neurons in the transgenic mice do not fire at specific locations , but still form robust firing sequences . These sequences are not anchored to precise locations of spatial trajectories and do not even distinguish between two very different spatial environments . The sequences seen on novel trajectories already exist in a prior session of free exploration in an open box . Therefore , CA1 firing sequences in the transgenic mice become so rigid that they no longer primarily respond to spatial environments or spatial experience , and therefore no longer capable of encoding spatial memories . The lack of both spatial specificity and experience-dependence indicates that external spatial input in Tau mice is incapable of taking control of Tau neurons . Nevertheless , Tau neurons do not fall into disorganized firing patterns , but fire with robust sequences . If not external space , what could be the driven force that assembles Tau neurons into these sequences ? It has been recently shown that internal brain activities , those related to memory recall ( Gelbard-Sagiv et al . , 2008; Pastalkova et al . , 2008 ) and those encoding time ( MacDonald et al . , 2011; Eichenbaum , 2013 ) , can produce CA1 firing sequences similar to those shown here . Therefore , here we propose that internal activities underlie the rigid firing sequences in the transgenic mice . On the other hand , although the firing activities of CA1 neurons in the transgenic mice were not location-specific , they are still modulated by external space and spatial experience ( Figure 7 ) . The existence of this external modulation together with the rigid firing sequences paint a picture of ‘cued false activation’ . When the transgenic mice are placed in a space , instead of forming/retrieving the space's memory code , CA1 neurons are cued to activate those internally driven activity patterns irrelevant to the current space . According to this interpretation , the reason why these mice cannot form new spatial memories is because the HP network is dominated by internal brain activities . This interpretation implies that there exists a direct competition between external and internal inputs in HP . Therefore , our data provide evidence that abnormal external-internal interaction may contribute to memory deficits in neurodegenerative diseases . Previous studies have shown that CA1 firing sequences can occur during immobility ( Foster and Wilson , 2006; Diba and Buzsaki , 2007; Davidson et al . , 2009; Karlsson and Frank , 2009; Carr et al . , 2011 ) or when the animal is spatially restricted ( Pastalkova et al . , 2008; MacDonald et al . , 2011 ) . These sequences are believed to be internally generated because external spatial input stays stationary in these situations . In our experiments , the animals’ paths between the start and end times of Tau sequences spanned a distance on the tracks ( Figures 3A and 6E ) and in the open boxes ( Figure 5C ) , clearly showing that the animals were moving when Tau sequences occurred . The sequences also occurred in accompany with prominent theta oscillations ( Figure 10A , H ) . Therefore , the non-spatially driven firing sequences were found even when animals were actively moving through space . In this case , the external spatial input associated with the movement is overtaken by internal input in the control of CA1 neurons’ firing activities . This further extends the recent findings that pre-existing neuronal sequences may influence future memory codes ( Gupta et al . , 2010; Dragoi and Tonegawa , 2011 ) . What changes in the neural circuits of Tau mice could produce the abnormal completion between internal and external input ? Internal activities could originate in CA3 , an area where neurons can drive each other via its extensive recurrent connections ( Lisman , 1999; Nakazawa et al . , 2002 ) . External sensory input reaches CA1 likely through the direct pathway via the entorhinal cortex . In Tau mice , there is evidence that tau pathology starts earlier in the cortex including the entorhinal cortex than in HP , and within HP , neuronal loss is more prominent in CA1 than in CA3 ( Ramsden et al . , 2005 ) . It is likely that the external input to CA1 through the entorhinal cortex is weakened by the pathology , but the internal input from CA3 is relatively spared . In the entorhinal cortex , grid cells fire spikes in a spatially periodic fashion ( Hafting et al . , 2005 ) . In our data , although Tau sequences sometimes occurred repeatedly in a running lap , there was no evidence for either spatial or temporal periodicity in Tau sequences ( Figure 8 ) . Also Tau cells’ firing activities were not tied to specific locations on the tracks ( Figure 2B ) and did not show obvious periodic patterns in the open boxes ( Figure 4B ) . Therefore , Tau sequences are unlikely resulted from periodic firing of entorhinal grid cells . Other potentially relevant changes in Tau mice could include interneurons and theta oscillations , since they greatly shape the firing activities of HP place cells ( O’Keefe and Recce , 1993; Skaggs et al . , 1996; Buzsaki , 2002; Dragoi and Buzsaki , 2006 ) . In Tau mice , the CamKII gene that induces tau pathology and neurodegeneration is not expressed in interneurons and thus should mainly affect pyramidal neurons in HP ( Mayford et al . , 1996; Sik et al . , 1998; Ramsden et al . , 2005 ) . However , interneurons could still die or alter their properties as adaptive changes to the tau pathology and loss of pyramidal neurons . In our datasets , we only recorded 11 putative interneurons from WT mice and 12 from Tau mice . There was a trend that Tau interneurons fired with a lower rate than WT ones ( WT: 24 . 2 [11 . 7 46 . 7] Hz , Tau: 14 . 8 [8 . 8 27 . 2] Hz , p=0 . 05 ) . This rate change could be one of the adaptive changes in Tau mice to keep the overall activity of remaining HP pyramidal neurons at normal level . Regarding theta oscillations , although the medium septum , the structure important for theta generation in HP , could be affected by tau pathology , based on the known expression pattern of CamKII ( Odeh et al . , 2011 ) , our data show that prominent theta oscillations are still present in the CA1 LFPs of Tau mice . In particular , the proportion of theta power in the total LFP power is in many cases similar between WT and Tau mice ( Figure 10E , H ) . Therefore , theta oscillation unlikely plays a major role in the generation of abnormal Tau sequences . Based on these considerations , we propose that the internal input from CA3 , by winning over the external input via the direct pathway from the entorhinal cortex , is the main driving force for producing the rigid Tau sequences . As such , Tau sequences mainly reflect the internal information such as old memories stored in the CA3 recurrent connections . Our data reveal a subtle but significant difference between male and female Tau mice: The firing stability of CA1 neurons on familiar trajectories in female is greater than that in male mice . This is somewhat unexpected , given a previous report that female tau mice show higher level of abnormally phosphorylated tau and worse performance in the Morris water maze task ( MWM ) than male mice ( Yue et al . , 2011 ) . Other previous studies on another tau-related mouse model have found more extensive β-amyloid , but not tau , pathology ( Hirata-Fukae et al . , 2008 ) and more severe impairments on MWM in females than in males ( Clinton et al . , 2007 ) . One possible explanation for the discrepancy is the age of animals used in our study . The sex difference of Tau mice was previously examined at 5 . 5 months old ( Yue et al . , 2011 ) . At much older ages ( 7–9 months ) as in our study , tau pathology becomes much more extensive and possibly saturated and thus the sex difference may become less significant ( Ramsden et al . , 2005 ) . Another possibility is that tau pathology may not only affect spatial memory representation , but also other factors important for MWM . For example , it has been proposed that behavioral testing produces more stress in females than males and thus greater behavioral deficits under pathological conditions ( Clinton et al . , 2007 ) . In our experiments , the animals had been handled for at least 1 month ( for tetrode adjustment and maze training ) before recordings began . The mice in our study might have a reduced level of stress , which exposed the subtle advantage of females in the firing stability of CA1 cells . This advantage in females could be compensated by a stronger stress response to MWM , producing an overall greater impairment on task performance . What pathological features in the transgenic mice are responsible for the observed rigid firing sequences ? Our recordings were conducted on the transgenic mice at 7–9 month old , which is a late , mature stage of tau pathology that displays all typical pathological features including hyperphosphorylated tau , tau tangles , and neurodegeneration ( Ramsden et al . , 2005; Santacruz et al . , 2005 ) . Previous experiments show that stopping the production of the transgenic tau is sufficient to halt the memory deficits at the behavioral level ( Ramsden et al . , 2005; Santacruz et al . , 2005 ) , pointing to a role of the hyperphosphorylated tau protein itself . Another possibility is that neurodegeneration reduces the neuron number in HP and cortex and result in a limited capacity for memory storage . Future studies on tauopathy models at various pathological stages may uncover the exact features of tau pathology and/or neurodegeneration that cause the observed abnormalities . Nevertheless , the current study allows us to identify the functional alterations of CA1 neurons at a mature stage of tau pathology and lay the foundation for probing their pathological causes . Currently , no animal models can fully recapitulate the pathological changes in human AD patients ( Ashe and Zahs , 2010 ) . The limitation in the animal model calls for caution in applying the finding to the human disease . However , tau pathology is a key feature of AD ( Wenk , 2003; Ashe and Zahs , 2010 ) and plays a key role in its behavioral phenotypes ( Ashe and Zahs , 2010 ) . This model does mimic key features of human tauopathy including its age-dependent progression ( Ramsden et al . , 2005; Santacruz et al . , 2005 ) . The functional changes at the cellular level in HP identified in this animal model , such as the low location-specificity and rigid firing sequences , possibly also occur in the human brain with similar tau pathological features . Interestingly , our interpretation that internally generated activities obstruct the formation of new memories is consistent with two typical memory symptoms seen in AD patients: the inability to form new memories ( Carlesimo and Oscar-Berman , 1992; Salmon and Bondi , 2009 ) and the intrusion of old memories ( Butters et al . , 1987; De Anna et al . , 2008 ) . This illustrates that studying the functional changes in memory circuits of animal models can generate novel insight into the memory symptoms of human neurodegenerative diseases including AD .
The mouse colony was maintained by an activator and a responder mouse lines ( Ramsden et al . , 2005; Santacruz et al . , 2005 ) . The activator , bred in a 129S6 background strain , carried the tTA transgene under the CaMKIIa promoter . The responder , bred in the FVB/N strain , carried a transgene encoding the human four-repeat tau with the P301L mutation ( hTauP301L ) . The F1 offspring of the two mouse lines that carries both the hTauP301L and tTA transgenes ( hTau+/tTA+ , Tau mice ) has been shown to overexpress the human tau in the forebrain and display age-dependent progression of tau pathology and neurodegeneration in the hippocampus ( HP ) and the cortex ( Ramsden et al . , 2005; Santacruz et al . , 2005 ) . The animals used in this study were Tau mice , their wildtype ( hTau−/tTA− , WT ) littermates , and littermates with the other two control genotypes , hTau+/tTA− ( +/− ) and hTau−/tTA+ ( −/+ ) , all at 7–9 month old ( Table 1 ) . Both male and female animals were used . A hyperdrive containing eight tetrodes was implanted during a surgery onto the skull of each mouse used in this study . Over the course of 2–4 weeks post surgery , tetrodes were slowly moved down to the CA1 pyramidal cell layer of HP . Starting about one week post surgery , the animal was food-deprived , with weight maintained above 85% of the ab libitum level , and trained in a familiar room to run back and forth ( two trajectories ) on a ∼2 m long rectangular track ( Figure 2A ) for food reward ( liquid condensed milk ) and to freely explore in a 50 × 30 cm open box ( Figure 4A ) . The animal was trained for at least 3 weeks and 15–30 min in each apparatus each day . Recording in the familiar room started when multiple , stable single-units ( neurons ) had been obtained in CA1 and the animal achieved a performance of at least 10 laps per trajectory on the track . The daily recording procedure consisted of two sessions on the track , followed by one to two sessions in the open box . Each session lasted about 15 min and there was a 15-min break between any two sessions , during which the animal rested on an elevated dish ( 25 cm high , 18 cm in diameter ) inside an enclosed box . The recording was repeated 3–10 days . After completing the recordings in the familiar room , recording continued in a novel , different room while the animal freely explored a novel open box ( 30 × 50 cm , Figure 6C ) and ran back and forth on a novel ∼1 m long L-shaped track ( Figure 6A ) . The animal had never been in the novel room or in the novel apparatuses before . The recording schedule was one to two open box sessions , followed by one to two sessions on the L track . Each session lasted about 15 min and there was a 15-min break between any two sessions . After the recording , the animal was sacrificed using pentobarbital overdose ( 200 mg/kg , IP injection ) . Electrical current ( 30 µA , 6–10 s ) was passed through the tetrodes to create small lesions at the recording sites . The animal’s brain was dissected out and fixed in 10% formalin for 24 hr . After its weight and size were measured , the brain was then sectioned for immunohistochemistry . The animal was fixed on a stereotaxic device with body temperature maintained at 36–38 °C and anesthetized with continuous flow of 0 . 5–2% inhalation anesthetic isoflurane . The flow was adjusted to keep the animal’s breathing rate at 40–80 per min . Ten anchor screws were mounted onto the skull . An exposure was made at the coordinates anteroposterior −2 . 0 mm , mediolateral 1 . 5 mm from Bregma . The hyperdrive cannula containing eight tetrodes was lowered to the exposure just above the brain . The drive was fixed in place by dental acrylic . Analgesic ( ketoprofen , 5 mg/kg ) was injected subcutaneously before the animal recovered from the anesthesia . Tetrode recording was made using a DigitaLynx acquisition system ( Neuralynx , Bozeman , MT ) and followed the published procedures ( Ji and Wilson , 2007 ) . For spike recording , voltage signals from four channels of a tetrode were digitally band-pass filtered between 600 Hz and 9 kHz . Spikes were detected with any of the four channels crossing a pre-set triggering threshold ( 50–70 μV ) and were sampled at 32 kHz . Broad band ( 0 . 1 Hz–1 kHz ) local field potentials ( LFPs ) were sampled at 2 kHz sampling rate . Two color diodes ( red , green ) were mounted over the animal's head to track its positions . Positions were sampled at 33 Hz with a resolution approximately 0 . 2 cm . Coronal sections with 20–70 μm thickness were stained using 0 . 2% Cresyl violet and cover-slipped for storage . Tetrodes were identified ( Figure 1 ) , by matching the lesion sites with tetrode depths and their relative positions . Data recorded from a total of 18 mice with four different genotypes were analyzed ( Table 1 ) . For each animal , one day’s data in the familiar room ( the day with most neurons recorded ) and one day’s data in the novel room ( the first day of exposure ) were included in the analysis . Neurons were manually sorted using xclust ( Matthew A Wilson , MIT ) . In total , 333 CA1 neurons in the familiar room and 327 in the novel room were obtained . Among them , 246 in the familiar and 221 in the novel room were classified as putative pyramidal neurons active either on a trajectory of a track or in an open box session ( mean firing rate: ≥0 . 5 Hz and <7 Hz ) . 11 in the familiar and 12 in the novel room were classified as putative interneurons ( mean rate ≥ 7 Hz ) . Further analysis was performed only on the active , putative pyramidal neurons unless otherwise specified . Results are presented as median and [10% 90%] range values with significance determined by the non-parametric ranksum test , or as mean ± SEM values with significance determined by one tailed t-test , unless specified otherwise . | Patients with Alzheimer's disease often forget where they have been or who they have just met . This happens because the neurons in those areas of the brain where memories are processed are dying . Indeed , by the time Alzheimer's disease has been diagnosed , many of the neurons in these regions have already died . The symptoms of Alzheimer's disease are then produced by the remaining neurons . However , the reasons why the remaining neurons cannot make new memories are unknown . In normal mice the neurons in the hippocampus , a part of the brain that is important for memory , are called ‘place cells’ because they are turned on when the mouse is in a specific place . As a consequence , when the mice moves around , different neurons are turned on one by one , and this sequence of activation is believed to be a memory code that represents the places the animal has travelled . Cheng and Ji have explored this phenomenon in mice that have been genetically engineered so that their neurons contain structures called ‘tau tangles’ that are thought to be involved in the death of neurons . More importantly , these transgenic mice suffer age-dependent neuron loss in a way that is similarly to people with Alzheimer's disease . Cheng and Ji implanted tiny sensors into the hippocampus of these mice , and used these sensors to monitor the activity of the remaining hippocampal neurons as the mice moved around while searching for food . They found that the neurons were activated almost everywhere , which indicates that the hippocampal neurons in transgenic mice are no longer working as place cells . However , these neurons were still activated one by one in robust sequences . Moreover , the sequences generated by the transgenic mice were the same in many different surroundings , which suggests that these sequences are not memory codes of the animal's current surroundings . Cheng and Ji propose that the sequences reflect existing memories already stored in the brain , which would suggest that Alzheimer's patients cannot form new memories because the brain is preoccupied by old memories , and thus fails to store the new information that is coming in from the outside world . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Rigid firing sequences undermine spatial memory codes in a neurodegenerative mouse model |
The functions of the TAF subunits of mammalian TFIID in physiological processes remain poorly characterised . In this study , we describe a novel function of TAFs in directing genomic occupancy of a transcriptional activator . Using liver-specific inactivation in mice , we show that the TAF4 subunit of TFIID is required for post-natal hepatocyte maturation . TAF4 promotes pre-initiation complex ( PIC ) formation at post-natal expressed liver function genes and down-regulates a subset of embryonic expressed genes by increased RNA polymerase II pausing . The TAF4–TAF12 heterodimer interacts directly with HNF4A and in vivo TAF4 is necessary to maintain HNF4A-directed embryonic gene expression at post-natal stages and promotes HNF4A occupancy of functional cis-regulatory elements adjacent to the transcription start sites of post-natal expressed genes . Stable HNF4A occupancy of these regulatory elements requires TAF4-dependent PIC formation highlighting that these are mutually dependent events . Local promoter-proximal HNF4A–TFIID interactions therefore act as instructive signals for post-natal hepatocyte differentiation .
TFIID plays a critical role in RNA polymerase II ( Pol II ) pre-initiation complex ( PIC ) formation . The TATA-box binding protein ( TBP ) and 13–14 TBP-associated factors ( TAFs ) assemble to form TFIID from an association between the ‘core complex’ comprising the TAF4–TAF12 and TAF6–TAF9 histone-fold heterodimers together with TAF5 and a second module comprising TBP , TAF1 , TAF2 , and TAF7 ( Cler et al . , 2009; Bieniossek et al . , 2013 ) . The TAF10–TAF8 dimer associates with the core complex , while the role of the TAF10–TAF3 and TAF11–TAF13 dimers is less well described ( Bieniossek et al . , 2013 ) . Extensive genetic analysis of TAFs has been performed in yeast ( Shen et al . , 2003 ) , but their function in mammalian cells particularly in complex physiological processes remains poorly characterised . The best understood are those of the cell-specific TAF paralogues like TAF7L that plays a critical role in male germ cell development and in adipocytes ( Cheng et al . , 2007; Zhou et al . , 2013 , 2014 ) or TAF4B essential for male and female fertility ( Falender et al . , 2005; Voronina et al . , 2007 ) . In contrast , much less is known about the ubiquitously expressed ‘core’ TAFs . For example , mice carrying loss of function alleles for TAF10 , TAF8 , TAF7 , or TBP die between blastocyst and pre-implantation stages upon depletion of the maternal protein ( Voss et al . , 2000; Martianov et al . , 2002; Mohan et al . , 2003; Gegonne et al . , 2012 ) . TAF4 and TAF10 have also been analysed in somatic tissues where they are necessary for normal development of the mouse epidermis ( Indra et al . , 2005; Fadloun et al . , 2007 ) . In contrast , in adult epidermis , loss of TAF10 has no evident phenotype whereas TAF4 plays a critical role in keratinocyte proliferation acting as a cell autonomous and non-cell autonomous tumour suppressor ( Fadloun et al . , 2007 ) . Despite these studies , the role of TAFs in physiological processes in vivo remains largely unknown and the molecular basis of their action is poorly characterised . In this study , we show that TAF4 is essential for activation of the post-natal hepatocyte gene expression programme . The TAF4–TAF12 heterodimer interacts with the nuclear receptor HNF4A to promote its binding to conserved and functional cis-regulatory elements located close to the transcription start sites of liver function genes . The stable binding of HNF4A to these elements is thus dependent on concomitant pre-initiation complex formation . These results reveal that HNF4A interaction with the basal transcription machinery controls not only PIC formation but also the occupation of its cognate regulatory elements .
Immunofluorescence revealed TAF4 expression in P12 wild-type ( WT ) hepatocytes , mesenchymal tissues , endothelial and smooth muscle cells around the hepatic veins and arteries , and cholangiocytes lining the bile ducts ( Figure 1A ) . TAF4 and TBP are expressed in late embryonic hepatocytes and TBP remains strongly expressed in adult hepatocytes , while TAF4 expression is reduced ( Figure 1—figure supplement 1 ) . TAF4 and TBP are therefore expressed from the embryonic to adult stages , with high expression in neonatal liver . Reduced TAF4 expression in adult hepatocytes has been previously reported ( D'Alessio et al . , 2011 ) , but we do not observe the reported strong loss of TBP expression . 10 . 7554/eLife . 03613 . 003Figure 1 . Expression of TAF4 in neonatal liver . ( A ) The first two panels show immunostaining for TAF4 in sections of WT liver at P12 , while the third panel shows immunostaining in sections from TAF4 mutant liver . The boxed region in the upper panel is blown up in the lower panel . Bd; bile duct , ha; hepatic artery , pv; portal vein , and cv; central vein , Hp+; TAF4-expressing hepatocyte , Hp−; TAF4 negative hepatocyte after Cre-mediated inactivation , ch; cholangiocyte , pvm; portal vein mesenchyme , cve; central vein endothelium . ( B ) Immunofluorescence for TAF4 expression in sections through WT and Taf4ahep−/− liver illustrating persistent expression in the extra-hepatic bile duct ( ehbd ) cholangiocytes and associated mesenchyme ( scale bar = 100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 00310 . 7554/eLife . 03613 . 004Figure 1—figure supplement 1 . Comparison of TBP and TAF4 expression in late embryonic to adult stage liver . Immunostaining for TBP or TAF4 in WT livers at the indicated stages ( scale bar = 100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 004 Mice carrying floxed alleles of the Taf4a gene were crossed with animals expressing Cre recombinase under the control of the albumin ( Alb ) promoter to inactivate Taf4 in post-partum murine hepatocytes ( Taf4aaf4hep−/− ) . TAF4 expression is lost from hepatocytes by P12 ( Figure 1A–B ) , although it remains in the portal vein mesenchyme , endothelial cells , cholangiocytes of extra-hepatic bile ducts , and bile ducts close to the liver hilum , but not in cholangiocytes forming peripheral bile ducts ( Figure 1A , B ) . Taf4ahep−/− animals are born at Mendelian ratios and no prenatal abnormalities or lethality were observed ( data not shown ) . Taf4ahep−/− animals are of normal weight and size at birth , show retarded growth by P12 , begin to die between P12 and P14 , and all are dead by P21 ( Figure 2A , B ) . Mutants display severe physiological and morphological alterations caused by perturbed liver function . Cholestatic lesions of mutant livers were evident at P2 with marked jaundice of the animals whose severity rapidly increases over the next days ( Figure 2A , C ) . By P12 , knockout animals display 50% reduced body weight and severe jaundice ( Figure 2A , B ) and their livers become completely yellow with visible patches of bile deposits within the parenchyma ( Figure 2C ) . 10 . 7554/eLife . 03613 . 005Figure 2 . Physiopathology of TAF4 inactivation in post-natal hepatocytes . ( A ) WT and Taf4ahep−/− animals at P4 . ( B ) Body weight of WT and Taf4ahep−/− animals at P12 , N = 20 . ( C ) Livers from animals at P2 and 12 ( scale bar = 2 mm ) . ( D ) Analysis of serum from WT and Taf4ahep−/− animals for the indicated parameters at P10 , N = 6 . ( E ) Hematoxylin–eosin staining of liver sections . Right panel shows a higher resolution of the boxed area of the left panel ( scale bar = 30 µm ) . Pv; portal vein , bd; bile duct , dbd; defective bile duct; ha; hepatic artery , lip; lipid droplets . ( F ) Periodic acid–Schiff and oil red O staining of livers ( scale bar = 100 µm ) . Cv; central vein . ( G ) Electron micrographs of livers ( scale bar = 10 µm ) . Nec; necrosis , n; nucleus , bn; bi-nucleate cell , dcc; defective cell–cell-contacts . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 00510 . 7554/eLife . 03613 . 006Figure 2—figure supplement 1 . Physiopathology of TAF4 inactivation in post-natal hepatocytes . Analysis of serum from WT and Taf4ahep−/− animals for the indicated parameters at P10 , N = 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 006 Blood plasma analysis revealed severe hypoglycemia in Taf4ahep−/− mice with glucose levels below 3 . 0 mmol/l in both males and females that is the most probable cause of early lethality ( Figure 2D ) . In accordance with their jaundiced appearance , we also detected high levels of bile acids and bilirubin in the plasma of the Taf4ahep−/− mice ( Figure 2D ) . In contrast , the levels of free fatty acids ( FFA ) , triglycerides ( TA ) , and total cholesterol do not significantly change , but HDL cholesterol is considerably reduced compared to WT mice ( Figure 2—figure supplement 1 ) . TAF4 knockout in post-natal hepatocytes therefore causes severe liver malfunction , hypoglycemia , and dysfunctional bile and lipid transport and metabolism . TAF4 mutant livers display alterations in parenchyma and lobule organisation with disorganised hepatocyte disposition in lobules and non-aligned hepatic plates due to variations in cell size ( Figure 2E ) . Periportal hepatocytes are markedly vacuolated and often display foamy intra-cytoplasmic inclusions indicating steatosis ( white arrows in Figure 2E , oil red staining in Figure 2F–G ) . PAS staining revealed a lack of glycogen in TAF4 knockouts ( Figure 2F ) , which along with hypoglycemia shows defective carbohydrate metabolism in Tafa4hep−/− mice . We also observed large and bi-nucleate hepatocytes mostly in pericentral areas , many necrotic cells in periportal regions ( Figure 2G ) , and a dramatic reduction in proliferating KI67 positive hepatocytes ( Figure 2H ) . In WT P5 liver , bile ducts were seen in both the hilum and peripheral regions ( Figure 3A ) . In Taf4ahep−/− liver , bile ducts with a clear lumen were formed in the hilum , but their number rapidly decreases towards to periphery where residual cholangiocytes organised as bi-layered ductal structures that fail to undergo normal tubulogenesis ( Figure 3A ) . Staining cholangiocytes at P12 with SOX9 showed normally formed mesenchyme-integrated bile ducts in the periportal regions of the hilum and periphery in WT liver ( Figure 3B ) . Staining for CLDN3 marking the apical pole of the ductal cells revealed properly formed ducts with a clear lumen ( Figure 3—figure supplement 1B ) . In mutant liver , SOX9 positive cholangiocytes were also observed , but even when located next to the large portal vein towards the hilum they did not form tubular ducts ( Figure 3B ) although CLDN3 staining revealed cell polarisation ( Figure 3—figure supplement 1B ) . Staining with ACTA2 showed hepatic arteries adjacent to normal bile ducts in WT liver , which were absent in the peripheral regions of Taf4ahep−/− liver ( Figure 3B ) . Portal vein mesenchyme was also significantly reduced in mutant liver ( Figures 3B and 1A ) . Electron microscopy confirmed proper formation of bile ducts with lumen in WT liver , but disorganised ducts without lumen in Taf4ahep−/− liver ( Figure 3C ) . The presence of ducts in the hilar region and their absence in the peripheral regions suggest that maturation is normally initiated during embryogenesis but is arrested upon TAF4 inactivation . 10 . 7554/eLife . 03613 . 007Figure 3 . Bile duct paucity . ( A ) Hematoxylin–eosin staining of livers at P5 from the hilum towards the periphery as indicated ( scale bar = 100 µm ) . ( B ) Immunofluorescence for SOX9 and ACTA2 in sections through WT and Taf4ahep−/− liver at P12 from the hilum towards the periphery as indicated ( scale bar = 100 µm ) . ( C ) Electron micrographs illustrating bile ducts in WT and Taf4ahep−/− liver showing a normal duct with lumen in WT and defective ducts in the mutant lacking normal lumen ( scale bar = 10 µm ) . pv; portal vein , bd; bile duct , cv; central vein , ha; hepatic artery , tj; tight junction , ch; cholangiocytes , ma; infiltrating macrophage . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 00710 . 7554/eLife . 03613 . 008Figure 3—figure supplement 1 . Defective tight junction formation and loss of bile–blood barrier . ( A ) Electron micrographs of sections from WT and Taf4ahep−/− liver at P12 . Tj; tight junction , mt; mitochondria , bc; bile canaliculus , n; nucleus . Double arrows indicate regions of enhanced intercellular space ( scale bar = 2 µm ) . ( B and C ) Immunofluorescence for tight junction components CLDN3 and TJP1 in sections of WT and Taf4ahep−/− livers at P12 ( scale bar = 100 µm ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 008 The above phenotype is similar to the loss of Notch signalling , associated with Alagille syndrome in humans , essential for normal bile duct numbers and morphogenesis ( Zong et al . , 2009 ) . These similarities suggest that TAF4 may act as a cofactor for transcription factors of the Notch signalling pathway . In addition , the under-development of portal vein mesenchyme and hepatic arteries associated with defective bile duct morphogenesis supports the reported reciprocal nature of biliary , vascular , and mesenchyme maturation ( Hofmann et al . , 2010 ) . Tight junctions formed between neighbouring hepatocytes produce an impermeable blood–bile canaliculi barrier and establish the basolateral–apical axis . The apical membrane is covered with microvilli in bile canaliculi , whereas the basolateral pole faces the blood flow in the sinusoid . Taf4ahep−/− liver displayed abnormal spaces between hepatocyte membranes indicative of impaired cell contacts and reduced number of tight junctions ( Figure 2G and Figure 3—figure supplement 1A ) . CLDN3 and TJP1 staining showed extensive tight junctions all around the membrane of wild-type hepatocytes , whereas , in the Taf4ahep−/− hepatocytes , staining was irregular and patchy , and many cells showed little or no staining ( Figure 3—figure supplement 1B–C ) . Consequently , impaired cell–cell contacts and the absence of tubular bile ducts account for the observed accumulation of bile in the liver parenchyma and in the blood of the mutant animals . We assessed changes in liver gene expression by RNA-seq at P12 . In TAF4-mutant liver , 1408 genes were down-regulated and 776 up-regulated compared to wild-type ( Figure 4A , and Figure 4A—figure supplement 1A and Supplementary file 1 , [Alpern et al . , 2014] ) The changes in expression of a selected set of genes were confirmed by RT-qPCR ( Figure 4B ) . Down-regulated genes are predominately expressed in liver ( endoderm ) , whereas the majority of up-regulated genes are annotated as enriched in brain , nervous system , and testis ( Figure 4A ) . 10 . 7554/eLife . 03613 . 009Figure 4 . Gene expression . ( A ) Ontology of up- and down-regulated genes . ( B ) Comparison of gene expression changes measured by RNA-seq and qPCR . ( C ) Assignment of deregulated transcripts to clusters determined by their expression kinetics during normal liver development at the indicated stages . Gene expression data for last day of embryonic development ( E18 . 5 ) , P7 and P14 , was normalized to have a mean of 0 and a SD of 1 and clustered using fuzzy c-means clustering ( Futschik and Carlisle , 2005 ) implemented in the mfuzz R package . The number of clusters was decided empirically . Each cluster was then overlapped with the down- and up-regulated genes from our RNA-seq . The significance of cluster assignments was assessed by calculation of the hypergeometric p-value . The number of up- and down-regulated genes included in the public data is indicated along with the % of each class present in the cluster ( hypergeometric tests p-values: *p < 1e−06 , **p < 1e−04 , ***p < 0 . 05 ) . ( D ) The Log2 mean expression value during neonatal stages for up- and down-regulated genes relative to their expression at E18 . 5 . The mean expression of down-regulated genes increased after birth , whereas that of the up-regulated genes decreased . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 00910 . 7554/eLife . 03613 . 010Figure 4—figure supplement 1 . Deregulated gene expression in Taf4ahep−/− liver . ( A ) RNA-seq comparison of gene expression in WT and TAF4-mutant liver . ( B ) Ontology analysis of the up- and down-regulated transcripts . ( C ) Schematic integration of deregulated genes in metabolic pathways and cell structure . Down-regulated genes are in blue and up-regulated genes in red . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 010 Down-regulated genes showed significant enrichment for terms associated with liver metabolic functions ( Figure 4A—figure supplement 1B ) and provide insight to the observed phenotype . Genes of the neutral bile acid biosynthetic pathway ( Cyp39a1 , Hsd3b7 , Akr1d1 , Cyp8b1 , Slc27a5 , Acox2 , Scp2 , and Baat ) were repressed as were the bile acid secretion and re-uptake components Abcb1a and Abcg5/8 , respectively ( Figure 4A—figure supplement 1C ) . Defective expression of these genes was reported to be associated with intrahepatic cholestasis and jaundice due to the production of abnormal cytotoxic bile or its impaired secretion ( Lefebvre et al . , 2009 ) , all of which are seen in Taf4ahep−/− animals . Cyp7a1 encoding an initiatory and rate-limiting enzyme responsible for up to 75% of total bile acid production was up-regulated . This may be explained by decreased expression of its transcriptional repressors SHP ( Nr0b2 ) and FXR ( Nr1h4 ) , but also by repression of Cyp8b1 . In Cyp8b1−/− mice , Cyp7a1 expression is up-regulated due to loss of cholic acid , a primary bile acid produced by CYP8B1 ( Lefebvre et al . , 2009 ) . Many glycolysis/gluconeogenesis and fatty acid oxidation genes were repressed such as the glucose transporter GLUT2 ( Slc2a2 ) , Gck , the rate-limiting enzyme of glucose metabolism , Pcx , several glucose inter-conversion enzymes ( Pfkl , Pgk1 , Pdha1 , Eno1 , Fbp1 ) , and Aqp7/Aqp9 , important for glycerol supply and hepatic gluconeogenesis . Reduced expression of the components of all epithelial junction types: tight junctions ( Clnd1/2/3 , Tjp3 ) , adherent junctions ( Cdh1 , Pvrl1/2 ) , desmosomes ( Dsc2 , Pkp1/3 ) , and gap junctions ( Gjb1/2 ) supports the observed disruption of the hepatic epithelium and loss of the blood–bile barrier in the TAF4 mutant animals . Several genes involved in glucose and lipid metabolism were up-regulated in Taf4ahep−/− livers , but surprisingly most encode enzymes that fulfil the complementary function in non-hepatic tissues . For example , Ldhb , Eno2/3 , Bpgm , and Fbp2 are normally expressed in heart , muscle , brain , or erythroid cells , but not in liver . Similarly , G6pdx , an enzyme of the pentose phosphate pathway , is normally highly expressed in adipose tissue but was up-regulated in Taf4ahep−/− livers . We also observed increased expression of fatty acid translocase Cd36 normally expressed in macrophages , muscle , and endothelium . Its up-regulation in liver is associated with insulin resistance , hyperinsulinaemia , and increased steatosis ( Miquilena-Colina et al . , 2011 ) . The brain-specific isoforms of fatty acid-binding protein ( Fabp7 ) and acyl-CoA-synthetase ( Acsl3 ) that activates long chain fatty acids were also up-regulated . Activated FAs are shuttled across the mitochondrial membrane via carnitine palmitoyltransferase-1 that has three isoforms: liver , muscle , and brain . Expression of liver Cpt1a was unchanged , but the muscle Cpt1b isoform was overexpressed . We compared our RNA-seq data with a multi-stage analysis of gene expression in developing liver ( Li et al . , 2009 ) . From this data , we clustered more than 6600 genes with respect to their expression at E18 . 5 , P7 , and P14 comprising 805 of the 1408 down-regulated and 193 of the 776 up-regulated genes . Around 50% of down-regulated genes clustered with those showing low embryonic expression and induction by P7 ( Figure 4C , clusters 1 and 2 ) . In contrast and despite the fact that only a minority of up-regulated transcripts were present in the Li et al . data , they are most represented amongst genes whose expression is high at E18 . 5 and down-regulated between P7 and P14 ( Figure 4C ) . This conclusion was confirmed when considering the mean neonatal expression value for up- and down-regulated genes relative to their expression at E18 . 5 . The mean expression of down-regulated genes increased after birth in WT , whereas that of the up-regulated genes decreased ( Figure 4D ) . Interestingly , several imprinted genes ( Dlk1 , Meg3 , Dio3 , Plagl1 , Peg3 , Peg10 , Igf2as , Airn , Igf2r , Malat1 , Neat1 , and Grb10 ) were up-regulated in mutant liver . These genes regulate fetal growth and are post-natal repressed . Expression of these genes is a hallmark of immature tissue , showing that TAF4 is required for normal post-natal hepatoblast maturation . We investigated expression of TFIID components in Taf4ahep−/− liver . TAF4 expression was strongly decreased in extracts from P12 Taf4ahep−/− liver ( Figure 5A ) but not totally lost since its expression persisted in the non-hepatocyte cells of the liver . Expression of TBP and several TAFs was also diminished while little change was observed for TFIIB , TFIIE , Pol II , and HNF4A ( Figure 5B ) . We performed anti-TBP immunoprecipitation ( IP ) and normalised the amount of IP TBP from WT and Taf4ahep−/− liver . Under these conditions , diminished amounts of all tested TAFs were observed in the TBP IP from Taf4ahep−/− ( Figure 5C ) . Thus , consistent with the critical role of TAF4 in the TFIID core complex ( Wright et al . , 2006 , Bieniossek et al . , 2013 ) , loss of TAF4 led to reduced TBP and TAF accumulation likely due to the disassembly of TFIID witnessed by the reduced co-precipitation of TAFs with TBP . No significant expression of TAF4b that can replace TAF4 in the core complex was seen in extracts from WT and Taf4ahep−/− liver and in the respective TBP IPs ( Figure 5D and data not shown ) . TAF4b therefore cannot substitute for TAF4 to maintain TFIID integrity in this tissue . 10 . 7554/eLife . 03613 . 011Figure 5 . Expression and integrity of TFIID . ( A and B ) Expression of TBP and TAFs in liver nuclear extracts ( 10 , 20 , 30 , 40 µg in A and 20 and 40 µg in B ) . ( C ) Expression of the indicated TAFs in the input , the control anti-HA IP , and the anti-TBP IP . ( D ) Absence of TAF4B in the wild-type and TAF4-mutant liver extracts . An extract from mouse embryonic stem cells ( mESC ) was used as a positive control where TAF4B can be clearly detected . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 011 We performed ChIP-seq from wild-type and Taf4ahep−/− P12 liver using antibodies against TBP , TAF3 , TFIIB , TFIIE , RNA polymerase II ( Pol II ) , trimethylated lysine 4 of histone H3 ( H3K4me3 ) , and CTCF to monitor PIC formation and chromatin organisation . The TBP , TFIIB , and TFIIE ChIP-seq data indicated a global decrease in their genomic occupancy and a 2–2 . 5 fold lower occupancy at the TSS of expressed genes in Taf4ahep−/− liver ( Figure 6A ) . No comparable decrease was seen for CTCF whose distribution is largely unchanged ( Figure 6B ) . A much stronger decrease in PIC formation was seen at the TSS of down-regulated genes . Similarly , an almost two-fold reduction of promoter-proximal paused and elongating Pol II was seen at all expressed genes , with a much stronger reduction at down-regulated genes ( Figure 7A ) . For example , PIC formation , Pol II recruitment , and H3K4me3 were lost at the TSS of Dio1 that is post-natal induced in WT ( Figure 6C ) . Similar results were seen at many other post-natal activated liver function genes indicating that loss of their expression in Taf4ahep−/− liver corresponds to defective PIC formation in the absence of TAF4 . 10 . 7554/eLife . 03613 . 012Figure 6 . Defective pre-initiation complex formation . ( A and B ) Genomic occupancy of the indicated factors and of H3K4me3 in WT and Taf4ahep−/− liver . Left panel shows the global profile and the right panels show occupancy at all expressed genes or up- and down-regulated genes as indicated . ( C ) Integrated read count of the indicated ChIP-seq tracks and the Dio1 or Angptl loci . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 01210 . 7554/eLife . 03613 . 013Figure 6—figure supplement 1 . TAF3 genomic occupancy in liver and ES cells . ( A ) Metaprofiles from ChIP-seq illustrating the different localisations of each factor relative to the TSS ( ±500 nucleotides with respect to the TSS ) in liver and mouse ES cells . ( B ) Density cluster profiles of ChIP-seq against the indicated factors centred on the TAF3 occupied sites ( ±5kb ) . The % of sites in clusters A and B corresponding to proximal , intermediate , and distal elements and for total TAF3 occupancy are indicated . ( C ) Heatmaps showing correlation between the ChIP-seq densities of the indicated factors at regions proximal or distal to the TSS . The % correlation for the more relevant combinations is indicated . ( D ) Comparative density cluster profiles of ChIP-seq against TAF3 in liver and ES cells at either TSS proximal regions ( ±2 kb with respect to the TSS ) or distal regions ( > ± 10 kb with respect to the TSS ) . The ChIP-seq data for ES cells was downloaded from NCBI GEO database with accession GSE30959 ( Liu et al . , 2011 ) . ( E ) Comparison of PIC formation and H3K4me3 with respect to gene expression . For each PIC component a list of associated genes was defined with an occupied site within ±2 kb from the corresponding TSS . Occupancy was normalized by the peak length and calculated as reads per kilobase ( RPK ) . The genes from RNA-seq data ( approx . 23K ) were ranked relative to their expression level , divided into bins of 100 , and attributed to the level of PIC component occupancy . Bins containing less than 20 genes were not included . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 01310 . 7554/eLife . 03613 . 014Figure 7 . Changes in Pol II elongation . ( A ) Pol II genomic occupancy in WT and in Taf4ahep−/− liver . ( B ) ChIP-seq tracks at the Malat or Ehhadh loci . ( C ) Overlays of Pol II ChIP-seq densities at the indicated gene loci . ( D ) ChIP-qPCR of CDK9 at the TSS , gene body ( GB ) , and 3′UTR of the indicated genes . The % input is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 014 More intriguingly , diminished PIC formation was also observed at up-regulated genes with reduced paused Pol II downstream of the TSS ( Figure 7A ) . The Pol II meta-profile shows , however , that levels of elongating Pol II were comparable in WT and mutant indicating reduced Pol II pausing and a relative increase in elongating Pol II in Taf4ahep−/− liver . This is exemplified at the Angptl4 gene where equivalent amounts of PIC and paused Pol II were seen in WT and mutant , whereas a higher density of elongating Pol II was observed in Taf4ahep−/− liver ( Figure 6C ) . Similar results were observed at the Malat1 , Txnip , Tat , and Cyp7a1 genes ( Figure 7B–C ) . At Ehhadh , an increase in both paused and elongating Pol II was observed . Up-regulated Afp expression also reflects an overall increase in Pol II occupancy . ChIP-qPCR indicated enhanced recruitment of the CDK9 subunit of the positive regulator of elongation PTEFb at the TSS , gene body , and 3′UTR of the Angptl4 , Malat , and Ehhadh genes in the mutant , whereas no such increase was seen at down-regulated genes that displayed a generally low level of CDK9 ( Figure 7D ) . These results reveal two distinct mechanisms for gene regulation in Taf4ahep−/− liver . Down-regulation is due to defective PIC formation , while up-regulation of many genes reflects decreased pausing and increased elongating Pol II and CDK9 recruitment . TAF3 was recruited to the TSS as expected for a TFIID component , but TAF3 occupancy and the levels of H3K4me3 at the TSS were much less affected than TBP occupancy and PIC formation ( Figure 6A ) . TAF3 occupancy rather correlated not with TBP and PIC formation but with H3K4me3 in agreement with the idea that it can be recruited independently of TBP and TFIID via interaction of its PHD domain with this mark ( Vermeulen et al . , 2007; Lauberth et al . , 2013 ) . Moreover , TAF3 does not precisely co-localise with TBP at the TSS . TBP and TFIIB localised immediately upstream of TFIIE , whereas TAF3 localised downstream with paused Pol II ( Figure 6—figure supplement 1A ) . Reanalysis of public ES cell data ( Liu et al . , 2011 ) also showed localisation of TAF3 and TAF1 downstream of the TSS , while TBP localised upstream . This differential crosslinking provides insight into how TFIID interacts with promoter DNA in the PIC . TAF3 occupied a large set of distal loci in the absence of TBP and PIC ( Figure 6—figure supplement 1B ) , as previously reported in ES cells ( Liu et al . , 2011 ) . In ES cells , 30% of distal TAF3 sites correlated with CTCF occupancy , whereas in liver only 11% correlation was seen ( Figures 6—figure supplement 1B , C ) . This is not due to differences in analysis procedures , as we independently calculated a 29–31% TAF3–CTCF correlation from the public ES data set ( Figure 6—figure supplement 1C ) . In hepatocytes , the TAF3–CTCF correlation was higher at the TSS ( 0 . 55 ) than at distal regions , the opposite of ES cells . A high TAF3–H3K4me3 correlation was observed at distal sites suggesting that this may be a mechanism of recruitment to these sites . Nevertheless , H3K4me3 levels are low at many of these distal TAF3 sites and other recruitment mechanisms cannot be excluded . Comparison of TAF3 genomic occupancy in liver and ES cells identified common and specifically occupied sites at the TSS , reflecting different sets of active promoters in these two cell types . Moreover , many of the distal regions are also unique to each cell type ( Figure 6—figure supplement 1D ) . Together these data indicate that TAF3 can be recruited to the genome independently of TBP and PIC formation not only at distal regions but also at the TSS . We also noted that TAF3 and H3K4me3 levels did not fully correlate with Pol II occupancy and PIC formation at the most highly expressed genes . TSS occupancy by TBP , TFIIB , TFIIE , and Pol II is positively correlated with gene expression where highest expressed genes showed highest occupancy ( Figure 6—figure supplement 1E ) . In contrast , H3K4me3 and TAF3 displayed reduced occupancy at this small highly expressed genes subclass , exemplified by Afp , Alb , and Ashg , with little Pol II pausing and high levels of elongating Pol II ( Figure 7C ) . Thus TAF3 and H3K4me3 show reduced occupancy at highly expressed genes lacking paused Pol II . Nuclear receptor HNF4A is a major regulator of hepatocyte gene expression ( Hayhurst et al . , 2001; Parviz et al . , 2003 ) . In WT and Taf4ahep−/− liver , HNF4A occupied >68 , 000 binding sites enriched around the TSS and comprising a motif essentially identical to that previously defined ( Figure 8A ) . Nevertheless , TAF4 loss modified HNF4A genomic occupancy as >7100 sites were depleted in the mutant , exemplified by the Dio1 and Slc22a7 loci ( Figure 8—figure supplement 1A ) , and >4100 sites were enriched in the mutant . Globally , sites depleted in the mutant localised close to the TSS ( Figure 8B ) and those associated with 487 down-regulated genes , using a window of ±40 kb with respect to the TSS , define a subpopulation closest to the TSS . In contrast , no TSS enrichment was seen for the small number of depleted sites associated with up-regulated genes ( red in Figure 8B ) . Unlike the sites depleted in the mutant , enriched sites associated with regulated genes showed no strong localisation at the TSS ( Figure 8C ) . A small number of these sites were however associated with up-regulated genes , perhaps corresponding to sites occupied during embryogenesis and lost in wild-type neonatal liver but not in the mutant . 10 . 7554/eLife . 03613 . 015Figure 8 . TAF4 is required to recruit HNF4A to functional CRMs . ( A ) Comparison of HNF4A ChIP-seq in WT and in Taf4ahep−/− liver and location of HNF4A occupied sites relative to the TSS . Comparison of the HNF4A consensus-binding sequence from our data generated by ChIP-MEME with the previously defined sequence . ( B ) Venn diagrams illustrate the number of genes with at least one HNF4A binding site either depleted or enriched in Taf4ahep−/− liver , within ±40 kb with respect to the TSS , intersected with up- or down-regulated genes . Graphs illustrate the locations of either the total , depleted or enriched HNF4A sites associated with the up and down-regulated genes within a window of ±40 kb with respect to the TSS . Total sites are shown in green and the sites associated with up- and down-regulated genes are shown in red and blue , respectively . ( C ) Upper panel shows the % of HNF4A-occupied sites enriched or depleted in Taf4ahep−/− liver that correspond to evolutionarily conserved or non-conserved CRMs . The lower panel shows the % of CRMs in each class that comprise sites for 1 , 2 , or 3 transcription factors in addition to HNF4A . ( D ) Location of HNF4A-occupied sites or CRMs , as indicated , enriched in WT or Taf4ahep−/− livers relative to the TSS corresponding to all peaks , all CRMs , or CRMs associated with down- and up-regulated genes . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 01510 . 7554/eLife . 03613 . 016Figure 8—figure supplement 1 . HNF4A genomic occupancy . ( A ) UCSC view of the TFIIB and HNF4A ChIP-seq tracks at the Dio1 or Slc22a7 loci . Arrows indicate HNF4A-binding sites that are depleted in TAF4-mutant liver . ( B ) Association of HNF4A sites enriched or depleted in the TAF4 mutant liver with active enhancer marks in adult liver . The % of CRMs in each cluster is indicated to the right of the panels . The ChIP-seq data for p300 , H3K4me1 , and H3K27ac in adult liver used from the ENCODE project ( http://genome . ucsc . edu/ENCODE/ ) ( Shen et al . , 2012 ) . ( C ) ChIP-qPCR against the indicated factors that co-occupy four different CRMs . The Dio1 downstream CRM is indicated by * in panel ( A ) . The Slc22a7 CRM is occupied by all 4 factors in WT , whereas the Dio1 and Car5a CRMs are occupied by only 3 of the 4 factors and not by HNF6 . The Acot1-associated CRM that gains HNF4A occupancy in the mutant is occupied by all 4 factors . FOXA1 and CEBPA occupy their sites in the wild-type and mutant , while HNF4A and HNF6 are recruited only in the mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 01610 . 7554/eLife . 03613 . 017Figure 8—figure supplement 2 . HNF4A genomic occupancy during liver development . ( A ) Density cluster profiles of HNF4A ChIP-seq at embryonic day 18 . 5 , P12 and P75 . Peaks in the left panel are centred on those occupied at E18 . 5 , in the middle panel on those occupied at P12 , and in the right panel at P75 . The data for P75 HNF4A occupancy is from Schmidt et al . ( 2010 ) [Arrayexpress E-TABM-722] ) . ( B ) Western blot of HNF4A expression in 30 µg of nuclear extract from WT embryonic , post-natal , and adult liver . ( C ) Density cluster profiles of HNF4A ChIP-seq at embryonic day 18 . 5 , P12 and P75 at the sites depleted or enriched in the TAF4 mutant liver ( left and right panels , respectively ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 017 Not all 68 , 000 detected HNF4A binding sites are functional in terms of gene regulation . At the Dio1 gene for example , two HNF4A binding sites were lost in the mutant , but a third was unchanged ( Figure 8—figure supplement 1A ) . We therefore asked if the HNF4A sites depleted in the mutant correspond to functionally active sites . Comparison with public adult liver data showed that many depleted HNF4A sites associate with p300 , H3K4me1 , and H3K27ac , whereas less of the enriched HNF4A sites show this association ( Figure 8—figure supplement 1B ) . Previous experiments identified evolutionary conserved HNF4A sites , suggesting they are critical for function ( Schmidt et al . , 2010 ) . A subset of HNF4A occupied sites is associated with binding sites for CEBPA , HNF6 , and FOXA2 forming cis-regulatory modules ( CRMs ) many of which are conserved in at least two species ( Ballester et al . , 2014 ) . Other sites , designated here as ‘singletons’ , correspond to HNF4A occupied sites that are not associated with binding sites for these other factors , although some are also conserved between at least two species . 60% of HNF4A sites depleted upon TAF4 inactivation overlap with CRMs , 16% of which are conserved ( Figure 8D ) . These CRMs display a higher proportion of sites for three or four factors and many associate with p300 , H3K4me1 , and H3K27ac ( Figure 8—figure supplement 1B ) . In contrast , only 20% of HNF4A sites enriched in the mutant overlap with CRMs of which only 4% are conserved and they show lesser association with active marks . Furthermore , CRMs depleted in the mutant , and in particular those associated with the 487 down-regulated genes , are strongly enriched at the TSS , whereas those with enhanced HNF4A occupancy appear more random although they are so few that there is no statistical significance ( Figure 8E ) . This allowed us to define a set of 296 down-regulated genes where the depleted HNF4A sites are localised between ±10 kb relative to the TSS ( Supplementary file 1 ) . ChIP-qPCR showed occupancy of CRMs at the Dio1 , Slc22a7 , and Car5a genes by FOXA2 , HNF6 , and CEBPA in WT liver , but not in the TAF4 mutant ( Figure 8—figure supplement 1C ) . Conversely , at one of the rarer CRMs enriched in the mutant , ChIP-qPCR showed increased HNF4A , CEBPA , and HNF6 occupancy , while that of FOXA2 was not significantly modified . TAF4 is therefore required for occupancy of a set of functional CRMs located close to the TSS explaining the failure to activate HNF4A-regulated liver function genes in the TAF4 mutant . We examined HNF4A genomic occupancy at E18 . 5 compared to neonatal and adult liver profiles . 4353 sites were occupied at E18 . 5 compared to >65 , 000 at P12 and >70 , 000 at P75 ( Figure 8—figure supplement 2A ) . The reduced HNF4A occupancy at E18 . 5 is not due to the lack of its expression that is comparable at E18 . 5 and adult stages ( Figure 8—figure supplement 2B ) . The E18 . 5-occupied sites associate with >1800 genes enriched in liver metabolic functions ( Supplementary file 1 ) . While almost all E18 . 5-occupied sites were occupied at P12 and P75 , subsets of sites show stronger occupation at P12 or P75 compared to other stages . We investigated the occupancy of sites enriched or depleted in the TAF4 mutant at the different developmental stages . A significant proportion of depleted sites maps to those with enriched occupancy at the neonatal stage ( Figure 8—figure supplement 2C ) . In contrast , around 30% of sites enriched in the mutant show only very low occupancy at other stages . Thus in the absence of TAF4 , HNF4A occupies a set of sites the vast majority of which are not normally occupied either in embryonic or adult liver . This suggests that there may be a disorganisation of the chromatin landscape in the TAF4-mutant liver that allows HNF4A accessibility to sites that are not accessible in wild type . TAF4 is therefore necessary to promote normal HNF4A genomic occupancy , in particular of promoter-associated functional CRMs . It has previously been shown that HNF4A interacts physically with TFIID via TBP ( Takahashi et al . , 2009 ) . HNF4A correlated with TBP/PIC occupancy at the TSS where HNF4A sites are enriched , but also at distal sites , indicative of functional HNF4A–TFIID interactions in vivo in hepatocytes ( Figure 6—figure supplement 1C ) . Diminished TBP recruitment in TAF4 mutant liver may by itself not explain the loss of PIC formation and HNF4A occupancy of the functional TSS-proximal CRMs . We therefore asked whether HNF4A also interacts with TAF4 , as loss of this interaction would better explain the observed phenotype . HEK cells were transfected with vectors expressing combinations of HNF4A , full-length mouse , or human TAF4 in the presence or absence of TAF12 and the extracts precipitated with anti-HNF4A antibodies . TAF4 alone and the TAF4–TAF12 heterodimer efficiently co-precipitated with HNF4A compared to the negative controls ( Figure 9A , lanes 2 , 3 , and 5 compared to 1 and 4 ) . We also expressed HNF4A along with the C-terminal histone fold-containing region of TAF4 . The TAF4 ( 805–1083 ) –TAF12 heterodimer efficiently co-precipitated with HNF4A compared to the negative control ( lanes 8 and 6 ) , however , TAF4 ( 805–1083 ) alone did not co-precipitate with HNF4A ( lane 7 ) as this short domain is mainly insoluble in the absence of TAF12 . As a further control , no co-precipitation of HNF4A with the TAF11–TAF13 histone fold pair was observed ( lane 9 ) . HNF4A therefore specifically interacts with the TAF4–TAF12 heterodimer . 10 . 7554/eLife . 03613 . 018Figure 9 . HNF4A interacts directly with the TAF4–TAF12 heterodimer via its LBD . ( A ) HEK cells were transfected with vectors expressing the constructs indicated above each lane . The upper panel represents proteins in the transfected cell extracts , the lower the proteins in the anti-HNF4A IP . ( B ) Formation of an HNF4A–TAF4–TAF12 complex from bacterial expressed proteins . SDS-PAGE followed by Coomassie brilliant blue staining of proteins retained on the cobalt-agarose column . Co-expressed proteins are shown above each lane and their locations to the left of the panel . Similar experiments with N-terminal 6HIS-tagged HNF4A were also attempted , but the presence of the tag induced degradation of recombinant HNF4A in bacteria not shown . ( C ) Bacterial co-expression of HNF4A domains with TAF4–TAF12 . Co-expressed proteins are shown above each lane and their locations to the left of the panel . As the HNF4A-DBD co-migrates with TAF4 ( 832–966 ) , it was re-expressed with TAF4 ( 805–1083 ) . ( D ) Model for cooperative HNF4A–TFIID function in hepatocyte gene activation . HNF4A interacts with TBP and TAF4 in TFIID via its DBD and LBD , respectively . These interactions are required for PIC formation on target promoters and occupancy of functional HNF4A sites regulating transcription of liver-specific genes during neonatal hepatic maturation . DOI: http://dx . doi . org/10 . 7554/eLife . 03613 . 018 To confirm this interaction , we co-expressed TAF4 ( amino acids 805–1083 ) –TAF12 ( 29–161 ) , containing their histone fold regions that form a soluble heterodimer in Escherichia coli ( Thuault et al . , 2002 ) , with full length HNF4A . HNF4A was expressed with the 6HIS-tagged TAF4–TAF12 heterodimer and the proteins were purified over a cobalt agarose column and visualised by SDS-PAGE and Coomassie blue staining . HNF4A co-purified in close to stoichiometric amounts with the TAF4–TAF12 heterodimer ( Figure 9B ) , while only trace amounts were non-specifically retained on the column in the absence of tagged TAF4–TAF12 . To better characterise this interaction , we co-expressed TAF4–TAF12 with either the DNA binding domain of HNF4A ( DBD , 55–135 ) , the DBD-hinge-ligand binding domain ( DBD–LBD , 55–377 ) , or the LBD alone ( 148–377 ) . These constructs were co-expressed with 6XHIS-tagged TAF12 ( 25–160 ) and a deletion of TAF4 ( 83–966 ) comprising the HFD . Both the DBD–LBD and the LBD alone were strongly retained on the column only in the presence of TAF4–TAF12 ( Figure 9C , lanes 5 and 6 compared to 1 and 2 ) . As the DBD alone co-migrates with TAF4 ( 832–966 ) and is not visible , we expressed the DBD alone with the longer version of TAF4 ( 805–1083 ) . The DBD was not specifically retained in the presence of TAF4–TAF12 ( lanes 8–10 ) . These data indicate that HNF4A directly forms a complex with the TAF4–TAF12 heterodimer via its LBD .
TAF4 inactivation in early post-natal hepatocytes led to defective liver organisation and function resulting in death of the animals by P15 . Three major features characterise this phenotype: defective bile duct formation , disorganised hepatocyte epithelium with loss of cell junctions and blood–bile barrier , lack of post-natal gene activation and consequent metabolic abnormalities . Many of these features , indicative of retarded hepatocyte maturation , can be explained by TAF4 acting as a cofactor for HNF4A , an idea that is supported by functional and biochemical data . Several features of the TAF4 knockout are observed when HNF4A is inactivated during hepatoblast development , such as disorganisation of the hepatocyte epithelium and liver architecture , loss of cell junctions , and glycogen accumulation ( Parviz et al . , 2003 ) . While the genes involved in these processes are activated by HNF4A during embryogenesis before TAF4 inactivation , TAF4 is required to maintain their expression as PIC formation and Pol II recruitment at their TSS were diminished at post-natal stages in mutant liver . At most of these genes , however , HNF4A occupancy was not strongly affected by TAF4 inactivation . HNF4A is required for post-natal expression of many liver metabolic function genes ( Hayhurst et al . , 2001; Inoue et al . , 2006; Kyrmizi et al . , 2006 ) . For example , HNF4A is necessary for fatty acid metabolism ( Martinez-Jimenez et al . , 2010 ) and bile acid synthesis ( Inoue et al . , 2006 ) . In these studies , however , HNF4A inactivation occurred between P35 and P40 , while using our mice ( carrying a different Alb-Cre transgene ) TAF4 inactivation is complete by P12 resulting in the loss of HNF4A binding to regulatory elements of at least 296 genes . Consequently , we observe a more severe general loss of liver metabolic functions with many relevant genes amongst these 296 requiring TAF4–HNF4A cooperation . While loss of HNF4A in adults has little effect on the expression of other nuclear receptors such as FXR , PXR , and LXR that play important functions in liver , their expression is strongly diminished in the Taf4ahep−/− liver also contributing to the more severe phenotype . Two different situations can thus be defined . At many , but not all , genes activated by HNF4A pre-natally in the presence of TAF4 , subsequent TAF4 inactivation leads to diminished PIC formation , but has little effect on HNF4A occupancy of their regulatory elements that persists at post-natal stages . In contrast , TAF4 is required to stably recruit HNF4A to regulatory elements and for PIC formation at post-natal activated genes as both processes are disrupted upon TAF4 inactivation . Our data also define for the first time the full scope of HNF4A in maintaining the integrity of the hepatocyte epithelium acquired during embryogenesis and in activation of the post-natal gene expression programme . In previous studies not all of these functions were observed due to the inappropriate timing of HNF4A inactivation that was either too early or too late to reveal them . While loss of HNF4A function accounts for many aspects of the TAF4 phenotype , we do not exclude that TAF4 may act as a cofactor for other transcription factors , for example RPBJ-NCID involved in bile duct morphogenesis . In addition to the above functional data , biochemical data also support the idea that TAF4 acts as a cofactor for HNF4A . Epitope-tagged HNF4A can precipitate TFIID ( Takahashi et al . , 2009 ) . We also show a correlation between HNF4A , TBP/TFIID , and PIC occupancy at distal sites , possibly enhancers , indicative of functional HNF4A–TFIID interactions in vivo . While this may in part be accounted for by interaction of the HNF4A-DBD with TBP ( Takahashi et al . , 2009 ) , we show that HNF4A also interacts with the TAF4–TAF12 heterodimer via its LBD . HNF4A directly forms a complex with the histone fold-containing regions of TAF4–TAF12 when co-expressed in bacteria . In the absence of TAF4 , HNF4A occupancy of TSS-proximal CRMs and/or PIC formation are compromised strongly suggesting that two direct HNF4A–TFIID interactions via the TAF4–TAF12 heterodimer and TBP are required to activate liver function genes ( Figure 9D ) . An analogous mechanism has been described where TAF4 acts as a cofactor for E-proteins ( Chen et al . , 2013 ) . E-proteins interact with the TAFH domain of TAF4 as opposed to the histone-fold region for HNF4A , and this interaction enhances TFIID binding to the core promoter . Our data are consistent with a model where stable binding of HNF4A to TSS proximal CRMs and PIC formation require HNF4A–TFIID interactions . The concomitant loss of PIC formation and CRM occupancy upon TAF4 inactivation indicates that they are mutually dependent events . Given the essential role of TAF4 in TFIID core assembly , the lack of TAF4b to compensate for its absence , and the reduced TBP-TAF co-precipitation , it is likely that TFIID integrity is disrupted in mutant hepatocytes , with the residual TBP-TAF co-precipitation reflecting the presence of intact TFIID in the cell types where TAF4 remains expressed . A similar conclusion was drawn when TAF10 that is also essential for TFIID integrity was inactivated in adult liver ( Tatarakis et al . , 2008 ) . TAF10 inactivation occurred later than TAF4 in this study and despite TFIID disassembly , a milder phenotype is observed as Taf10hep−/− animals die at P35–P38 . Upon TAF10 inactivation , recruitment of TBP and TAFs to many promoters is lost but PIC formation was not affected suggesting that transcription persists in the absence of TBP/TFIID . In contrast , we show that TBP ( and TAF1 , unpublished data ) is recruited to promoters of expressed genes , and its recruitment and PIC formation are diminished to comparable extents . Moreover , TAF3 promoter occupancy at expressed genes is less affected . While loss of TFIID integrity compromises TBP occupancy and PIC formation , interaction with H3K4me3 acts to stabilise TAF3 recruitment independently of TBP . We therefore find no evidence for PIC formation and gene expression in the absence of TBP and TAF3 recruitment in the TAF4 mutant hepatocytes . TAF4 regulates hepatocyte gene expression by controlling HNF4A occupancy , PIC formation , and Pol II pausing . Expression of subset of embryonic genes is up-regulated in the absence of TAF4 due to decreased Pol II pausing , although it is also possible that changes in RNA stability may also contribute to the observed increase in mRNA abundance . Why TAF4 regulates pausing at this set of genes remains to be determined , although we note that they are enriched in short intron-less genes . Also it has been previously described that TAF7 has a negative effect on CDK9 and BRD4 activity ( Devaiah et al . , 2010 ) . Perhaps defective recruitment of TAF7 at these genes due to compromised TFIID integrity would explain the augmented CDK9 recruitment and an increase in its kinase activity would account for the enhanced elongation . Nevertheless , our observations show that down-regulation of many of these genes normally occurs at least initially through increased Pol II-pausing at post-natal stages , while in the absence of TAF4 they continue to be expressed with higher levels of elongating Pol II as in their embryonic mode . We also observed diminished paused Pol II at a small subgroup of highly expressed liver identity genes that may be under the control of ‘super enhancers’ ( Whyte et al . , 2013 ) , where TAF3 and H3K4me3 levels were also reduced . This concomitant reduction suggests that TAF3/H3K4me3 levels may modulate Pol II pausing at this class of genes .
We show that TAF4 is essential for activation of the post-natal hepatocyte gene expression programme acting as a cofactor for HNF4A . Previous models proposed that activators bind their cognate sites and promote PIC formation via interactions with TAFs . We show in vivo that HNF4A stably occupies a set of functional sites only when accompanied by concomitant TAF4-dependent PIC formation showing that these are mutually dependent events ( Figure 9D ) . Many recent studies have highlighted the role of long-range enhancer–promoter interactions and the formation of large-scale chromatin domains ( de Laat and Duboule , 2013 ) . In contrast , we highlight here the importance of local interactions at the TSS . Although HNF4A is also present at distal enhancers , functional HNF4A sites cluster close to the TSS and interactions with TFIID and also possibly the mediator complex ( Malik et al . , 2002 ) play a critical role in gene activation .
The previously described ( Mengus et al . , 2005 ) Taf4alox/lox animals were crossed with Albumin-Cre ( Alb-Cre ) transgenic mice ( Postic and Magnuson , 2000 ) . Experiments were performed in compliance with National Animal Care Guidelines ( European Commission directive 86/609/CEE; French decree no . 87–848 ) . Library preparation and mRNA sequencing were performed as previously described ( Herquel et al . , 2013 ) . The results were confirmed by qRT-PCR for at least 30 independent genes . Gene functional annotation was performed using DAVID ( http://david . abcc . ncifcrf . gov/ ) . For ChIP-seq , mouse livers , freshly isolated or snap frozen in liquid nitrogen , were homogenized by douncing and fixed with 1% PFA for 10 min and fixing was stopped by adding glycine at a final concentration of 0 . 125 M . Alternatively , for TAF3 and for TBP ChIP-seq , chromatin was fragmented by MNase I digestion as follows . Fixed nuclei were resuspended in equal volume of MNase buffer ( 50 mM Tris–HCl , pH 8 . 0 , 15 mM NaCl , 5 mM CaCl2 , 60 mM KCl ) and treated with MNase ( #MO247S; NEB Ipswich , MA ) at 37°C for 10 min . The reaction was stopped by addition of EDTA to a final concentration of 20 mM . Chromatin was incubated with 0 . 3% SDS ( 5 min on ice ) and with 1% Triton X-100 ( 5 min on ice ) and centrifuged at 14 , 000 rpm . ChIP was performed overnight with 50 μg of chromatin and 5 μg of the following antibodies: anti-Pol II ( sc-9001 X ) , anti-TFIIB ( sc-225 X ) , anti-TFIIE ( sc-6935 X ) , anti-HNF4A ( sc-8987 X ) , anti-CDK9 ( sc-8338 X ) , anti-HNF6 ( sc-13050 X ) , anti-CEBPA ( sc-9314 X ) , anti-FOXA2 ( sc-6554 X ) from Santa Cruz ( Santa Cruz , CA ) , anti-TBP ( ab28175 ) from Abcam ( Cambridge UK ) , anti-CTCF ( 07–729 ) and anti-H3K4me3 ( 04–745 ) from Millipore ( Billerica , MA ) , anti-TAF3 ( IGBMC , in house ) . ChIP-seq libraries were prepared as previously described and sequenced on the Illumina Hi-seq2500 as single-end 36-base reads ( Choukrallah et al . , 2012 ) . Peak detection was performed using the MACS software ( Zhang et al . , 2008 ) using the no antibody ChIP as negative control . Data sets were normalised for the number of unique mapped reads for subsequent comparisons . Global clustering analysis and quantitative comparisons were performed using seqMINER ( http://bips . u-strasbg . fr/seqminer/ ) ( Ye et al . , 2011 ) and R ( http://www . r-project . org/ ) and visualized using ggplot2 package . Genomic region annotation was performed with either seqMINER or GREAT ( http://bejerano . stanford . edu/great/public/html/ ) . Overlap between HNF4A-binding sites that are enhanced or depleted in TAF4-mutant liver and CRMs was performed by intersecting the corresponding genomic coordinates . Liver nuclei extracts were made as follows . Livers were homogenized by douncing in cold hypotonic buffer with 25 mM HEPES , pH 7 . 8 , 1 . 5 mM MgCl2 , 10 mM KCl , and 0 . 1% NP-40 , supplemented with Protease Inhibitor Cocktail ( Roche , Basel , Switzerland ) and DTT . Nuclei were washed twice in cold hypotonic buffer followed by centrifugation ( 3000 rpm , 5 min ) and lysed in 50 mM Tris–HCl , pH 7 . 5 , 450 mM NaCl , 0 . 5% NP40 , 5% glycerol for 2 hr on 4°C with constant rotation . Detection was performed using the same antibodies as for ChIP or previously described in house antibodies against TBP and TAFs ( Gangloff et al . , 2001 ) . TAFs , HNF4A , CDK9 , Pol II , and TBP were revealed with the antibodies described elsewhere in the ‘Materials and methods’ . Immunoprecipitation was performed overnight with anti-TBP ( 3G3 ) or anti-HA as control . Livers were fixed at 4°C for 16 hr in 4% paraformaldehyde in phosphate-buffered saline ( PBS ) , washed overnight in PBS , and embedded in paraffin . For immunofluorescence 10 µm sections were treated in a microwave oven in citrate buffer pH 6 . 0 for 15 min at 150 W . Staining was performed with primary anti-TAF4 ( sc-136093 ) and anti-HNF4A ( sc-6556 ) from Santa Cruz , Claudin3 ( ab15102 ) and TBP ( ab51841 ) from Abcam , Sox9 ( AB5535 ) from Chemicon , anti-aSMA ( M0851 ) from DAKO , TJP1 from Invitrogen ( Carlsbad , CA ) ( 339100 ) , KI67 from Novocastra ( Nussloch , Germany ) ( NCL-KI67P ) . Nuclei were counterstained with Hoechst . For histological analysis , sections were stained with Hematoxylin and Eosin , Red Oil O , and Periodic acid–Schiff ( PAS ) following the standard procedures . Full length HNF4A and its indicated deletion mutants , the histone-fold containing fragments of human TAF4 ( 805–1083 ) and ( 832–966 ) and TAF12 ( 29–161 ) , were cloned into the His-Tag vector pEA-tH and/or the native pCS vector ( Diebold et al . , 2011 ) . E . coli BL21 cells were transformed with respective vectors and colonies were used to grow 15 ml cultures at 37°C till the OD600 = 0 . 4 . Protein expression was induced by IPTG and cultures were grown at 25°C overnight . Cells were resuspended in 1 . 5 ml of lysis buffer ( 10 mM Tris–HCl , pH 8 . 0 and 50 mM , 200 mM or 400 mM NaCl ) and sonicated . After the centrifugation , the supernatants were incubated with TALON resin for two hours at 4°C with the permanent agitation . The resin was washed twice with 1 ml of ice-cold lysis buffer and resuspended in 25 µl of Laemmli buffer . Samples were analysed by SDS-PAGE with subsequent Coomassie brilliant blue staining . Total RNA was extracted with the RNeasy Mini Kit ( Qiagen , Venlo , Holland ) following the manufacturer's instructions and treated with RNase-free DNase ( Fermentas ) . RNA ( 3 μg ) was reverse transcribed by using a Maxima Kit ( Fermantas , Pitsburgh , PA ) . The final product was then diluted 10 times and 2 μl were mixed with forward and reverse primers ( 250 nM of each primer at final concentration ) and 5 μl of SYBR Green master mix ( Qiagen ) . The real-time PCR reaction was performed by using the LightCycler 1 . 5 system ( Roche ) . Each cDNA sample was tested at least in triplicate . | To decode the information contained within a gene , a number of processes need to occur . For example , the DNA sequence that makes up the gene needs to be copied to make a molecule of RNA , which is then translated to build the corresponding protein . The first steps in the manufacture of RNA involve a structure called a ‘pre-initiation complex’ moving an enzyme called RNA polymerase II to the start of the gene that needs to be copied . The pre-initiation complex is made up of many types of protein , including a set of proteins called TAFs . However , the way that these proteins work in mammals is not well understood . There are good reasons for this: proteins are often studied by seeing what happens when the protein is removed , but many TAFs are so important that removing them is lethal . Alpern et al . have now studied the function of TAF4 by removing this protein from mouse liver cells . This causes severe hypoglycemia ( that is , a drop in sugar levels in the blood ) . Moreover , it seems as if these cells start dying before they become fully mature . In liver cells lacking TAF4 , some 1408 genes that are normally turned on just after birth are not properly switched on; these genes are necessary for the metabolic functions of the liver . Furthermore , 776 genes that are normally turned off after birth continue to be expressed . It seems that the absence of TAF4 sometimes disrupts the formation of the pre-initiation complex , which would slow down the production of RNA . However , it can also have the opposite effect by increasing the activity of RNA polymerase II , hence making too many copies of RNA from some genes . Alpern et al . also find that TAF4 is needed to allow a protein called HNF4A , which is important in the development of the liver and in controlling metabolism , to interact with over 7000 important DNA sequences . Mutations in HNF4A are responsible for a syndrome known as Maturity Onset of Diabetes in the Young . The next stage in this work will be to explore if these mutations influence the interaction between HNF4A and TAF4 , and if they do , whether these changes contribute to this form of diabetes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Conclusion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | TAF4, a subunit of transcription factor II D, directs promoter occupancy of nuclear receptor HNF4A during post-natal hepatocyte differentiation |
Copulation is the goal of the courtship process , crucial to reproductive success and evolutionary fitness . Identifying the circuitry underlying copulation is a necessary step towards understanding universal principles of circuit operation , and how circuit elements are recruited into the production of ordered action sequences . Here , we identify key sex-specific neurons that mediate copulation in Drosophila , and define a sexually dimorphic motor circuit in the male abdominal ganglion that mediates the action sequence of initiating and terminating copulation . This sexually dimorphic circuit composed of three neuronal classes – motor neurons , interneurons and mechanosensory neurons – controls the mechanics of copulation . By correlating the connectivity , function and activity of these neurons we have determined the logic for how this circuitry is coordinated to generate this male-specific behavior , and sets the stage for a circuit-level dissection of active sensing and modulation of copulatory behavior .
All animals must continuously sequence and coordinate behaviors appropriate to both their environment and internal state if they are to survive and reproduce . Dissecting the neural substrates that initiate , organize , and terminate these behavioral sequences is critical to understanding behavior . Here , we use Drosophila male courtship behavior as a model to address how a compact circuit coordinates these critical action sequences . Drosophila male courtship behavior is a multi-step goal-directed behavior that has evolved to achieve reproductive success ( Villella and Hall , 2008 ) . Typically , the male follows the female , taps her with his forelegs , contacts her genitalia with his mouthparts , sings a species-specific courtship song , and bends his abdomen to copulate ( Villella and Hall , 2008 ) . Successful execution of these discrete sequential motor programs requires continuous integration of multiple sensory cues from the female . Copulation , the direct objective of courtship , is a highly conserved and essential behavioral step for most animals . Copulation itself also consists of an ordered behavioral progression: first , the male engages external genital structures to grasp the female; then he extrudes the intromittent organ , the aedeagus , and initiates copulation ( Kamimura , 2010 ) . The male maintains this posture for 15 min or more while transferring a mixture of sperm and seminal fluid to the female ( Villella and Hall , 2008 ) . Finally , the male terminates copulation by sequential uncoupling of his genitals and detachment from the female . Failure at any point in this complex action sequence , from courtship to termination of copulation , may prevent reproduction . It follows that the neural substrates of male courtship behavior have evolved to carry out the precise action sequences of copulation . Male reproductive behaviors are controlled by neurons expressing the two key sex determination genes , doublesex ( dsx ) and fruitless ( fru ) ( Pavlou and Goodwin , 2013 ) , the male-specific isoforms of which ( FruM and DsxM ) specify male-specific neurons ( Billeter et al . , 2006; Rideout et al . , 2007 , 2010 ) . Manipulating the activity of some , or all , of these neurons profoundly alters male courtship and copulatory behaviors ( Stockinger et al . , 2005; Billeter et al . , 2006; Rideout et al . , 2010; Kohatsu et al . , 2011; Pan et al . , 2011; von Philipsborn et al . , 2011; Tran et al . , 2014 ) . For example , activating fru- or dsx-expressing neurons in males elicits all courtship and copulatory behaviors ( Pan et al . , 2011 ) , while inhibiting all ~650 dsx neurons in males blocks all courtship and copulatory behaviors ( Rideout et al . , 2010 ) . Interestingly , activation of dsx circuitry in males lacking fru-expression ( fruM-null ) elicits robust courtship and copulatory behaviors ( Pan et al . , 2011 ) , suggesting that dsx-specified circuitry encompasses the fundamental neural substrates underlying both courtship and copulation ( Rideout et al . , 2010; Pan et al . , 2011 ) . However , it remains unknown , which of these dsx neurons control male copulation . Gynandromorph studies suggest that copulation is regulated by neurons in both the central brain and the abdominal ganglion ( Abg ) , the most posterior region of the ventral nerve cord ( VNC ) ( Hall , 1979; Ferveur and Greenspan , 1998 ) . Indeed , optogenetic activation of dsx-expressing pC2l cluster of neurons in the male brain has been shown to induce attempts to copulate ( Kohatsu and Yamamoto , 2015 ) . However , activating dsx neurons in solitary headless males elicits abdominal curling ( Pan et al . , 2011 ) and decapitating males in copulo has little effect on the duration of copulation ( Tayler et al . , 2012; Crickmore and Vosshall , 2013 ) , suggesting the Abg can direct many copulatory behaviors independent of any input from the brain . It is therefore likely that descending signals from the brain serve to initiate copulation by triggering a local circuit module in the Abg that in turn coordinates copulation , but the nature of this Abg circuit has remained unclear . While the expression of fru and dsx in sensory- , inter- and motor neurons suggests that they are organized into circuit elements capable of receiving , processing and transferring information that controls sexual behavior ( Pavlou and Goodwin , 2013 ) , surprisingly little is known about the core circuit elements encompassing copulation . To investigate the organizational principles of the sex-specific circuits underlying copulation , we used a Split-GAL4 intersectional approach ( Luan et al . , 2006 ) to identify dsx neurons within the Abg that express the major excitatory and inhibitory neurotransmitters in Drosophila . We found that dsx glutamatergic motor neurons innervate muscles of the genitalia and enable genital attachment and intromission; dsx GABAergic inhibitory neurons mediate genital uncoupling likely by inhibition of key motor neurons; and dsx mechanosensory neurons of the genitalia innervate and activate both dsx GABAergic and dsx glutamatergic neurons in the Abg . These results suggest a model in which dsx configures a sexually dimorphic sensorimotor circuit which allows the male to successfully execute the correct action sequence for both genital attachment and detachment .
To functionally identify sub-populations of dsx neurons with differing neurotransmitter profiles , we generated a novel dsx Split-GAL4 allele ( dsxGAL4-DBD ) by homologous recombination at the dsx locus ( Figure 1a , b ) ( Luan et al . , 2006 ) . We validated the specificity of expression pattern of the dsx Split-GAL4 allele by pairing it with a pan-neuronal matching Split-GAL4 driver ( elavVP16-AD ) ( Luan et al . , 2006 ) ( Figure 1c , d ) . dsx/elav>GFP flies replicated the expression pattern of the previously characterized dsxGAL4 allele in the nervous system ( Figures 1d1–4 , c1–4 , respectively ) without exhibiting GFP expression in non-neural tissues ( Figures 1d5–6 vs . c5–6 ) . We then functionally validated dsxGAL4-DBD by silencing dsx/elav neurons with tetanus toxin light chain ( TNT ) ( Sweeney et al . , 1995 ) , which blocks synaptic vesicle exocytosis , in both males and females ( Figure 1—figure supplement 1 ) . dsx/elav>TNT males and females reproduced the behavioral phenotypes of the previously characterized dsxGAL4>TNT males and females ( Rideout et al . , 2010 ) . Specifically , dsx/elav>TNT males spent very little time courting wild type females ( Figure 1—figure supplement 1a ) , and completely failed to copulate ( Figure 1—figure supplement 1b , c ) and were therefore behaviorally sterile ( Figure 1—figure supplement 1d ) , while dsx/elav>TNT females were infertile ( Figure 1—figure supplement 1e ) , unreceptive ( Figure 1—figure supplement 1f , g ) , and exhibited no post-mating behavioral responses ( Figure 1—figure supplement 1h ) . 10 . 7554/eLife . 20713 . 003Figure 1 . Spatial restriction of GFP expression to dsx neurons using novel dsx Split-GAL4 allele . ( a ) Schematic of doublesex ( dsx ) gene and male and female predicted transcripts . Arrows indicate transcriptional start sites . Colored boxes depict non-sex-specific ( black ) and sex-specific ( red: male and grey: female ) exons . ( b ) Schematic of dsxGAL4 and dsxGAL4-DBD knock-in alleles . ( c ) GFP expression in five day-old males and females driven by dsxGAL4 . ( c1–4 ) dsxGAL4 driving UAS-nuclear GFP ( nGFP ) in ( c1 ) adult male brain and ( c3 ) VNC and ( c2 ) adult female brain and ( c4 ) VNC . ( c5–6 ) Epifluorescence images of dsxGAL4 driving UAS-2XEGFP ( EGFP ) in ( c5 ) adult male and female whole-fly preparations revealing EGFP expression in sub- and peri-cuticular cells and ( c6 ) adult male filleted dorsal abdominal wall revealing EGFP expression in the adult fat body . ( d ) GFP expression in five day-old males and females driven by dsxGAL4-DBD combined with pan-neuronal elavVP16-AD hemidriver . ( d1–4 ) dsxGAL4-DBD/elavVP16-AD ( referred to as dsx/elav in text ) driving UAS-nGFP in ( d1 ) adult male brain and ( d3 ) VNC and ( d2 ) adult female brain and ( d4 ) VNC . Epifluorescence images of dsxGAL4DBD/elavVP16-AD driving UAS-2XEGFP in ( d5 ) adult male and female whole-fly preparations revealing no EGFP expression in sub- and peri-cuticular cells and ( d6 ) adult male filleted dorsal abdominal wall revealing no EGFP expression in the adult fat body . nGFP realized with anti-GFP antibody ( green ) and neuropil counterstained with nc82 ( magenta ) . EGFP realized with anti-GFP antibody ( white ) . ( c1–4 ) and ( d1–d4 ) views are ventral , with anterior up . Scale bar = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 00310 . 7554/eLife . 20713 . 004Figure 1—figure supplement 1 . dsx-expressing neurons specify male and female sexual behaviors . ( a-d ) Effects of silencing dsx/elav neurons on male courtship and copulatory behaviors . ( a ) Courtship index ( mean ± S . E . M . ; n = 30 ) . ( b ) Copulation duration ( mean ± S . E . M . n = 30 ) . ( c ) Percentage of male matings in 1 hr ( n = 30 ) . ( d ) Male fertility ( n = 30 ) . Genotypes indicate males . ( e–h ) Effects of silencing dsx/elav neurons on female courtship behaviors . ( e ) Female fertility ( n = 30 ) . ( f ) Line crossings during copulation ( mean ± S . E . M . ; n = 30 ) . ( g ) Copulation duration ( mean ± S . E . M . ; n = 30 ) . ( h ) Percentage of females that re-mate with the same male in 2 hr ( n = 30 ) . Genotypes indicate females . ( a–h ) **p<0 . 001 , ***p<0 . 0001 by Fisher exact test ( a , b , f , g ) or Kruskal-Wallis and Dunn’s test ( c , d , e , h ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 004 To identify dsx neurons involved in the motor control of the genitalia during copulation , we targeted dsx motor neurons using a Split-GAL4 insertion downstream of the vesicular glutamate transporter ( vGlut ) locus ( Gao et al . , 2008 ) , derived from the OK371-GAL4 enhancer trap ( Mahr and Aberle , 2006 ) . Glutamate is the key excitatory neurotransmitter at the Drosophila neuromuscular junction ( NMJ ) ( Daniels et al . , 2008 ) , and both OK371-GAL4 and vGlutOK371-dVP16-AD ( referred to as vGlutdVP16-AD here onwards ) ( Gao et al . , 2008 ) have been widely used to target motor neurons ( Karuppudurai et al . , 2014; Ting et al . , 2014 ) . We used dsxGAL4-DBD/vGlutdVP16-AD to drive nuclear GFP to count cell numbers and cytoplasmic GFP to visualize neuronal projections ( Figure 2a–g ) . dsx/vGlut neurons were found in three anatomically distinct regions in both males and females: neurons in the Abg , forelegs , and dsx-aDN neurons in the brain ( Figure 2a–e; Table 1 ) . No cells were labeled elsewhere in the PNS , or non-neural tissues . Co-labeling with anti-dvGlut ( Mahr and Aberle , 2006 ) antibody confirmed the glutamatergic identity of dsxGAL4-DBD/vGlutdVP16-AD neurons ( Figure 2—figure supplement 1a–c ) . 10 . 7554/eLife . 20713 . 005Figure 2 . Sexually dimorphic dsx/glutamatergic neurons control genital coupling during copulation . ( a-d ) Sexually dimorphic expression of dsx/vGlut neurons in the brain ( top ) and VNC ( bottom ) of adult males ( a , b ) and females ( c , d ) . ( a , c ) dsx/vGlut cell bodies visualized by vGlutdVP16-AD/dsxGAL4-DBD driving UAS-nGFP in ( a ) male and ( c ) female CNSs . nGFP stained with anti-GFP ( green ) ; neuropil counterstained with anti-nC82 ( magenta ) . ( b , d ) dsx/vGlut projection patterns visualized by vGlutdVP16-AD/dsxGAL4-DBD driving UAS-2XEGFP in ( b ) male and ( d ) female CNSs . EGFP stained with anti-GFP ( black ) ; sexually dimorphic midline crossing ( red arrowhead ) and neurons of the Abg and their descending projections ( red arrow ) are shown . ( e ) dsx/vGlut driven EGFP expression in the T1 tarsi of the male foreleg . Projections from these neurons form the male-specific contralateral commissural bridge in the mesothoracic gangion of the male VNC ( red arrowheads in bottom panels of b , d ) . ( f , g ) dsx/vGlut driven EGFP expression in the ( f ) internal reproductive system and ( g ) abdomen reveals motor neuron arborizations onto ( f ) muscles of the aedeagus and ( g ) dorsal and ventral muscles of the sixth abdominal segment . ( g1-2 ) Higher magnification of ventral ( g1 ) and dorsal ( g2 ) longitudinal muscles of sixth abdominal segment showing dsx/vGlut motor neuron innervations and synaptic termini . EGFP stained with anti-GFP ( green ) . Internal reproductive system and abdominal muscles counterstained with the F-actin specific antibody Phalloidin ( phall; magenta ) . Detail of internal genitalia: testes , ejaculatory bulb ( EjB ) , and aedeagus ( Aed ) indicated . Scale bar = 50 μm . ( h-j ) Effects of silencing dsx/vGlut neurons on male copulatory and courtship behaviors . ( h ) Percentage of successful matings in 1 hr ( n = 24–30 ) . ( i ) Male fertility ( n = 30 ) . ( j ) Courtship index ( mean ± S . E . M . ; n = 24–30 ) . Genotypes indicate males . See also Video 1 . ( k , l ) Effects of thermoactivating dsx/vGlut neurons on male courtship and copulatory behaviors . ( k ) Courtship index ( mean ± S . E . M . ; n = 20–30 ) . ( l ) Percentage of successful matings in 1 hr ( n = 20–30 ) . Statistical comparisons of the experimental genotype at 31°C in ( k-l ) were made against the same genotype at 23°C and all control genotypes at 31°C . Genotypes indicate males . ( m ) Effects of thermoactivating male dsx/vGlut neurons 5 min into copulation . Percentage of copulations terminated over a 2 hr period is graphed ( n = 22–24 ) . ( n ) Video still showing ‘stuck’ dsx/vGlut>TrpA1 male at the end of the 2 hr observation period . See also Video 2 . ( h-l ) **p<0 . 001 , ***p<0 . 0001 by Fisher exact test ( h , i , k ) or Kruskal-Wallis and Dunn’s test ( j , l ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 00510 . 7554/eLife . 20713 . 006Figure 2—figure supplement 1 . Characterisation of dsx/glutamatergic neurons in the adult male CNS . ( a ) Schematic representation of dsx/vGlut neurons in the adult male CNS . Black dotted boxes depict regions shown in b-e . ( b-c ) Co-localization of nGFP and anti-dvGlut in dsx/vGlut neurons in the ( b ) brain and ( c ) VNC of vGlutVP16-AD/ UAS-pStinger; dsxGAL4-DBD/+ males . Scale bar = 10 μm . Single section view of marked regions in b and c ( white dotted box ) showing overlap between nGFP and dvGlut antibody at higher optical magnification shown on right of panels b and c . nGFP stained with anti-GFP ( green ) ; vGlut stained with anti-dvGlut ( magenta ) . ( d-e ) Co-localization between nGFP and anti-FruM in dsx/vGlut neurons in the ( d ) brain and ( e ) VNC of vGlutVP16-AD/ UAS-pStinger; dsxGAL4-DBD/+ males . Solid arrowheads = co-localization; Empty arrowheads = no co-localization . nGFP stained with anti-GFP ( green ) ; FruM stained with anti-FruM ( magenta ) . Scale bar = 10 μm . ( f ) Co-localisation of dsx/vGlut driven EGFP and anti-HRP in male internal reproductive system . All neuronal innervations are revealed with anti-HRP . Expression of dsx/vGlut>EGFP and anti-HRP shown separately in right panels . EGFP stained with anti-GFP ( green ) . Internal reproductive system and abdominal muscles counterstained with the F-actin specific antibody Phalloidin ( phall; blue ) . Detail of internal genitalia: ejaculatory bulb ( EjB ) , and aedeagus ( Aed ) indicated . Scale bar = 50 μm . ( g ) Effects of thermoactivating dsx/vGlut neurons in decapitated males 5 min into copulation . Percentage of copulations terminated over a 2 hr period is graphed ( n = 22–26 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 00610 . 7554/eLife . 20713 . 007Table 1 . Cell counts for dsx-intersected neurons in male and female adult CNS . Male and female dsx/elav , dsx/vGlut and dsx/Gad1 cell counts are listed in black . Subsets of neurons that co-express FruM in males are listed in italics . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 007dsx neuronal clustersdsx/elav dsx/vGlut dsx/Gad1 MaleFemaleMaleFemaleMaleFemaleBrain pC1*52 . 8 ± 4 . 1 ( 12 ) 8 . 3 ± 1 . 6 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) pC2*78 . 3 ± 4 . 8 ( 12 ) 14 . 2 ± 1 . 5 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) 1 . 0 ± 0 ( 12 ) 0 . 9 ± 0 . 3 ( 10 ) 0 ± 0 ( 12 ) pC3*13 . 8 ± 0 . 9 ( 12 ) 8 . 0 ± 1 . 0 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) 3 . 5 ± 0 . 5 ( 12 ) 0 . 5 ± 0 . 5 ( 10 ) 3 . 0 ± 0 ( 12 ) aDN*2 . 0 ± 0 ( 12 ) 2 . 0 ± 0 ( 12 ) 1 . 9 ± 0 . 3 ( 12 ) 0 ± 0 ( 10 ) 2 . 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) 0 ± 0 ( 12 ) SN*1 . 0 ± 0 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . Ventral Nerve Cord TN1*23 . 0 ± 1 . 5 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . TN2*7 . 9 ± 0 . 3 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . 0 ± 0 ( 12 ) n . a . Abg†275 . 0 ± 21 . 7 ( 10 ) 314 . 8 ± 18 . 9 ( 10 ) 79 . 8 ± 2 . 3 ( 10 ) 7 . 4 ± 3 . 1 ( 10 ) 101 . 8 ± 6 . 7 ( 10 ) 151 . 2 ± 3 . 8 ( 10 ) 30 . 0 ± 4 . 8 ( 10 ) 213 . 1 ± 2 . 1 ( 10 ) *Neuronal cluster away from CNS midline . Count represents one cluster per hemisegment of the CNS . †Neuronal cluster spans the CNS midline . Count given is for the entire Abg . Counts represent mean ± S . D . n’s listed in parentheses . The largest sub-population of dsx/vGlut neurons is that of the Abg ( with ~80 in males and ~100 in females; Figure 2a , c; Table 1 ) . These project from the VNC via the abdominal nerve trunk ( Figure 2b , d ) and arborize onto muscles of the genitalia of both sexes . In males this includes the muscles controlling protraction and retraction of the aedaegus ( Figure 2f ) as well as the most distal ( A6 ) longitudinal muscles of the ventral and dorsal abdomen ( Figure 2g ) . We further confirmed that dsx/vGlut neurons encompass all of the motor neurons that innervate all phallic and periphallic musclulature ( Figure 2—figure supplement 1f ) . In females , these neurons innervate muscles of uterus , spermathecal and parovarian ducts ( data not shown ) . No additional muscular innervations were observed . Neurons in the foreleg ( Figure 2e ) project to the prothoracic ganglion and cross the midline in males , but not females ( Figure 2b , d ) , which is typical of foreleg gustatory receptor neurons ( GRNs ) ( Possidente and Murphey , 1989; Mellert et al . , 2010; Rideout et al . , 2010 ) . The two neurons of the dsx-aDN cluster project locally within the dorsal brain and extend to the superior medial protocerebrum ( SEM ) in both sexes ( Figure 2a–d; Table 1 ) . To examine a potential role for fru in specifying the sexual identity of dsx/vGlut neurons , we co-stained samples with FruM antibody ( Figure 2—figure supplement 1a , d , e ) . Interestingly , none of the dsx/vGlut-aDN and only ~10% of dsx/vGlut-Abg neurons co-expressed FruM ( Figure 2—figure supplement 1d , e respectively; Table 1 ) . These fru/dsx/vGlut-Abg neurons likely include the motor neurons that innervate the dorsal and ventral muscles of the sixth abdominal segment ( Nojima et al . , 2010 ) . We tested the role of dsx/vGlut motor neurons in copulation by silencing their activity with TNT . dsx/vGlut>TNT males completely failed to achieve genital coupling ( Figure 2h; Video 1 ) . Even after seven days in the presence of several virgin females , dsx/vGlut>TNT males produced no progeny ( Figure 2i ) . These males also spent less time courting target females ( Figure 2j ) , although they displayed the normal complement of courtship behaviors , including attempting to copulate ( Video 1 ) . We conclude that dsx/vGlut neurons are necessary for successful genital coupling . 10 . 7554/eLife . 20713 . 008Video 1 . Inhibition of dsx/vGlut neurons in males blocks genital coupling and the initiation of copulation . This movie shows a dsx/vGlut>TNT male failing to achieve genital coupling and initiate copulation with a wild-type female . These males do however display the normal complement of courtship behaviors . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 008 Copulation requires motor coordination between the external genitalia and the copulatory organ to facilitate genital attachment . We therefore tested whether simultaneous contraction of these organs prior to or during copulation , by artificial activation of all dsx/vGlut neurons , would prevent genital coupling . We expressed the Drosophila heat-activated cation channel dTrpA1 ( Hamada et al . , 2008 ) in dsx/vGlut neurons and examined the effects of dsx/vGlut thermoactivation prior to copulation . Pairs of dsx/vGlut>dTrpA1 experimental males and wild-type females were heated to 31°C 10 min prior to and throughout a 1 hr observation period ( Figure 2k , l ) . Thermoactivated dsx/vGlut males displayed less overall courtship towards target females ( Figure 2k ) ; as with dsx/vGlut>TNT males , these males displayed all courtship steps , but never successfully copulated ( Figure 2l ) . These results indicate that disrupting the activity of dsx/vGlut neurons pre-copulation , by either complete silencing or activation , perturbs the motor events that are necessary for a male to attach to a female and initiate copulation ( Figure 2h , l ) . Insemination occurs within the first ~8 min of copulation ( Gilchrist and Partridge , 2000 ) . During this critical time , males resist being interrupted by any stressful stimuli , displaying ‘copulation persistence’ ( Crickmore and Vosshall , 2013 ) . To test the role of dsx/vGlut neurons during copulation , we activated dsx/vGlut neurons 5 min into copulation , when fertilization is not complete , and ‘copulation persistence’ is at its peak ( Crickmore and Vosshall , 2013 ) ( Figure 2m , n; Video 2 ) . Compellingly , throughout a 2 hr observation period , thermoactivation of dsx/vGlut neurons in copulo prevented the male from detaching from the female and terminating copulation , which normally occurs after 10–15 min ( Figure 2m , n; Video 2 ) . Extended thermoactivation of dsx/vGlut neurons in copulo did not disrupt sperm transfer , as all matings were fertile . Interestingly , stimulation of dsx/vGlut neurons did not impair the timing of copulation drive; after approximately 15 min , dsx/vGlut>dTrpA1 males attempted to uncouple from the female by dismounting and/or kicking the female genitalia with their hind legs . 10 . 7554/eLife . 20713 . 009Video 2 . Activation of dsx/vGlut neurons in copulating males blocks genital uncoupling and the termination of copulation . This movie shows three dsx/vGlut>TrpA1 males that remain attached to their wild-type female mating partners via their genitals after ~2 hr of in copulo thermal activation ( 31°C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 009 Which of the anatomically distinct dsx/vGlut neurons are responsible for this phenotype ? Amputation of the forelegs results in indiscriminate courtship towards con- and allo-specific males and females ( Fan et al . , 2013 ) , but does not impinge upon copulatory behavior ( s ) , indicating that dsx/vGlut neurons in the foreleg may effectively be ruled out of this particular motor circuit . Furthermore , decapitation of thermoactived dsx/vGlut>dTrpA1 males in copulo did not eliminate the ‘stuck’ phenotype ( Figure 2—figure supplement 1g ) . These results demonstrate that dsx/vGlut-Abg neurons comprise the key motor neurons that control genital coupling during a copulation event . To identify dsx inhibitory neurons we focused on GABA , the major inhibitory neurotransmitter in insects ( Jackson et al . , 1990 ) . As GABA biosynthesis requires glutamic acid decarboxylase 1 ( Gad1 ) ( Erlander and Tobin , 1991 ) , we exploited the intersecting expression domains of dsx and Gad1 using dsxGAL4-DBD and a Gad1 hemi-drivers ( Gad1p65-AD ) ( Diao et al . , 2015 ) . We observed sexually dimorphic groups of dsx/Gad1 neurons in both the Abg and brain ( Figure 3a–d; Table 1 ) . There was no expression in the PNS or non-neural tissues . Co-labeling with anti-GABA antibody confirmed that neurons labeled by dsxGAL4-DBD/Gad1p65-AD are indeed GABAergic ( Figure 3—figure supplement 1a–c ) . 10 . 7554/eLife . 20713 . 010Figure 3 . Sexually dimorphic dsx/GABAergic neurons control genital uncoupling during copulation . ( a-d ) Sexually dimorphic expression of intersected dsx/Gad1 neurons in the brain ( top ) and VNC ( bottom ) of adult males ( a , b ) and females ( c , d ) . ( a , c ) dsx/Gad1 cell bodies visualized by dsxGAL4-DBD/Gad1p65-AD driving UAS-nGFP in ( a ) male and ( c ) female CNSs . nGFP stained with anti-GFP ( green ) ; neuropil counterstained with anti-nC82 ( magenta ) . ( b , d ) dsx/Gad1 projection patterns visualized by dsxGAL4-DBD/Gad1p65-AD driving UAS-2XEGFP in ( b ) male and ( d ) female CNSs . EGFP stained with anti-GFP ( black ) . Scale bar = 50 μm . ( e-i ) Effects of silencing dsx/Gad1 neurons on male copulatory and courtship behaviors . ( e ) Courtship index ( mean ± S . E . M . ; n = 24–30 ) . ( f ) Percentage of successful matings in 1 hr ( n = 24–30 ) . ( g ) Copulation duration ( n = 12–24 ) . ( h ) Percentage of males displaying ‘stuck’ phenotype ( n = 12–24 ) . ( i ) Male fertility ( n = 30 ) . Genotypes indicate males . See also Video 3 . ( j , k ) Effects of thermoactivating dsx/Gad1 neurons on male courtship and copulatory behaviors . ( j ) Percentage of successful mating’s in 1 hr ( n = 20–30 ) . ( k ) Courtship index ( mean ± S . E . M . ; n = 20–30 ) . Statistical comparisons of the experimental genotype at 31°C in ( j , k ) were made against the same genotype at 23°C and all control genotypes at 31°C . Genotypes indicate males . ( l ) Effects of thermoactivating male dsx/Gad1 neurons 5 min into copulation . Percentage of copulations terminated over a 2 hr period is graphed ( n = 20–24 ) . ( m ) Video stills showing that activation of dsx/Gad1 >TrpA1 neurons in copulo results in an almost immediate termination of copulation . Top panel shows dsx/Gad1 >TrpA1 male and wild type female mating 5 min into copulation at the point of shifting the temperature to 31°C . Bottom panel shows the same mating pair 8 s later , at which time the male has terminated copulation . See also Video 4 . ( e-k ) n . s . = not significant , *p<0 . 05 , *p<0 . 05 , **p<0 . 001 , ***p<0 . 0001 by Fisher exact test ( f , h-j ) or Kruskal-Wallis and Dunn’s test ( e , g , k ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 01010 . 7554/eLife . 20713 . 011Figure 3—figure supplement 1 . Characterisation of dsx/GABAergic neurons in the adult male CNS . ( a ) Schematic representation of dsx/Gad1 neurons in the adult male CNS . Black dotted boxes depict regions shown in b-e . ( b-c ) Co-localization of nGFP and anti-GABA in dsx/Gad1 neurons in the ( b ) brain and ( c ) VNC of UAS-pStinger/+; dsxGAL4-DBD/Gad1p65-AD males . Scale bar = 10 μm . Single section view of marked regions in b and c ( numbers and white dotted box ) showing overlap between nGFP and GABA antibody at higher optical magnification shown to the right of panels b and c . nGFP stained with anti-GFP ( green ) ; GABA stained with anti-GABA ( magenta ) . ( d-e ) Co-localization between nGFP and anti-FruM in dsx/Gad1 neurons in the ( d ) brain and ( e ) VNC of UAS-pStinger/+; dsxGAL4-DBD/Gad1p65-AD males . Solid arrowheads = co-localization; Empty arrowheads = no co-localization . nGFP stained with anti-GFP ( green ) ; FruM stained with anti-FruM ( magenta ) . Scale bar = 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 01110 . 7554/eLife . 20713 . 012Figure 3—figure supplement 2 . dsx/GABAergic neurons in the male brain do not specify male copulatory behaviors , but instead mediate male courtship . ( a-b ) Brain-restricted GFP expression in dsx/Gad1 neurons of the adult male CNS . ( a ) Expression of myrGFP in dsx/Gad1 neurons in the brain and not the VNC of OtdFLP/UAS-myrGFP; dsxGAL4-DBD/Gad1p65-AD males . myrGFP stained with anti-GFP ( green ) ; neuropil is counterstained with anti-nC82 ( magenta ) . ( b ) Projections of a 10 μm subset of ( a ) showing skeleton expression of myrGFP in dsx/Gad1-brain neurons . myrGFP shown in black . Scale bar = 50 μm . ( c-e ) Effects of silencing dsx/Gad1-brain neurons on male courtship and copulatory behaviors . ( c ) Courtship index ( mean ± S . E . M . ; n = 18–24 ) . ( d ) Percentage of successful matings in 1 hr ( n = 18–24 ) . ( e ) Copulation duration in seconds ( mean ± S . E . M . ; n = 18–24 ) . Genotypes indicate males . ( f-h ) Effects of thermoactivating dsx/Gad1-brain neurons on male courtship and copulatory behaviors . ( f ) Courtship index ( mean ± S . E . M . ; n = 12–24 ) . ( g ) Percentage of successful mating’s in 1 hr ( n = 12–24 ) . ( h ) Copulation duration in seconds ( mean ± S . E . M . ; n = 12–24 ) . ( i ) Effects of thermoactivating male dsx/Gad1-brain neurons 5 min into copulation . Percentage of copulations terminated over a 2 hr period is graphed ( n = 12–24 ) . ( c ) ***p<0 . 001 , ( d-h ) n . s . = not significant by Fisher exact test ( d and g ) or Kruskal-Wallis and Dunn’s test ( c , e , f , h ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 012 In the Abg , we observed ~150 dsx/Gad1 neurons in males and ~210 in females , or over 50% and 65% of all dsx Abg neurons , respectively ( Table 1 ) . dsx/Gad1-Abg neurons arborize locally within the Abg , except for two neurons that project to the mesothoracic and prothoracic ganglia in males only ( Figure 3b , d ) . In the central brain , dsxGAL4-DBD/Gad1p65-AD labeled a small subset of dsx-pC2 and -pC3 ( Rideout et al . , 2010 ) neurons in males , and an equivalent subset of dsx-pC3 neurons in female that project to the SEM and subesophageal zone ( SEZ; Figure 3a–d , see cell counts in Table 1 ) . To determine whether dsx/Gad1 neurons express fru , we co-labeled male CNSs with FruM antibody ( Figure 3—figure supplement 1d , e ) . We observed fru expression in the dsx/Gad1-pC2 neurons , in one out of the three dsx/Gad1-pC3 neurons per brain hemisphere , and in ~20% of dsx/GABAergic neurons in the Abg ( Figure 3—figure supplement 1d , e; Table 1 ) . We tested the role of dsx/Gad1 neurons in copulation by silencing their activity with TNT ( Figure 3e–i ) . dsx/Gad1>TNT males displayed relatively normal levels of courtship towards target females and displayed all courtship steps , including abdominal bending and attempted copulation ( Figure 3e ) . However , less than 20% of males successfully copulated during a 1 hr observation period ( Figure 3f ) . The males that successfully copulated took significantly longer than controls ( Figure 3g ) . Approximately ~ 80% of these males could not uncouple from the female ( Figure 3h; Video 3 ) , exhibiting a similar ‘stuck’ phenotype to that of dsx/vGlut males thermoactivated in copulo ( Figure 2m , n; Video 2 ) . Sperm transfer appeared normal , as all of these matings produced progeny ( data not shown ) . After one week in the presence of several virgin females , only ~60% of dsx/Gad1>TNT males produced progeny ( Figure 3i ) . The timing of copulatory motivation appeared to be intact , because after approximately 15 min , dsx/Gad1>TNT males dismounted and attempted ( but failed ) to detach . We conclude that dsx/Gad1 neurons functionally oppose dsx/vGlut-Abg neurons: dsx/vGlut-Abg neurons promote genital coupling while dsx/Gad1 neurons promote genital uncoupling . 10 . 7554/eLife . 20713 . 013Video 3 . Inhibition of dsx/Gad1 neurons in males blocks genital uncoupling and the termination of copulation . This movie shows a dsx/Gad1>TNT male displaying the distinctive ‘stuck’ behavior , whereby he has dismounted the wild-type female but remains attached via his genitals for prolonged periods of time . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 013 We hypothesized that if dsx/Gad1 neurons inhibit dsx/vGlut motor neurons , artificial activation of these neurons pre-copulation should prevent males from achieving genital attachment . To test this , we expressed dTrpA1 in dsx/Gad1 neurons and examined the effects of dsx/Gad1 thermoactivation on courtship and copulatory behaviors . Thermoactivating dsx/Gad1 neurons in these males completely blocked genital attachment in the 1 hr observation period ( Figure 3j ) . The males’ inability to achieve genital coupling was not a result of impaired courtship , as they spent normal amounts of time courting ( Figure 3k ) and displayed all courtship behaviors , including vigorous , but failed , attempts to copulate . Specifically , dsx/Gad1>dTrpA1 males attempted to copulate 34 ± 3 times within the first 10 min of courtship , while controls attempted to copulate 7 ± 1 times prior to a successful copulation ( n = 10 ) . As blocking dsx/Gad1 neurons prolongs copulation and activating them prevents it , we conclude that dsx/Gad1 neurons promote genital uncoupling . Silencing dsx/Gad1 neurons and activating dsx/vGlut neurons both disrupt genital coupling , suggesting that dsx/Gad1 neurons may selectively inhibit dsx/vGlut neurons to create the precise pattern of dsx/vGlut activity required for copulation . Artificial activation of dsx/Gad1 neurons in copulo should therefore inhibit genital coupling and result in premature termination of copulation . Compellingly , ~90% of dsx/Gad1>dTrpA1 males dismounted females and terminated copulations within 1 min of thermoactivation ( Figure 3l , m; Video 4 ) ; this was significantly shorter than the termination times of control males ( Figure 3l ) . Given that the majority of dsx/Gad1>dTrpA1 males spent no more ~6 min in copulo , much less than the ~8 min requirement for fertile matings ( Figure 3l ) ( Gilchrist and Partridge , 2000 ) , we checked the fertility of each mating . Matings truncated by thermoactivation never produced any progeny ( Figure 3l ) . These results demonstrate that dsx/Gad1 neurons are sufficient to induce genital uncoupling and terminate copulation . Importantly , the opposing phenotypes that result from thermoactivating dsx/Gad1 and dsx/vGlut neurons in copulo ( premature termination and perpetual copulation , respectively ) demonstrate that dsx/Gad1 neurons functionally oppose dsx/vGlut neurons to inhibit genital coupling for the termination of copulation . 10 . 7554/eLife . 20713 . 014Video 4 . Activation of dsx/Gad1 neurons in copulating males elicits genital uncoupling and the termination of copulation . This movie shows a dsx/vGlut>TrpA1 male and wild-type female copulating pair that have been shifted to 31°C 5 min into copulation . Thermal activation of male dsx/vGlut neurons in this manner results in the near-immediate termination of copulation by dsx/Gad1>TrpA1 males . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 014 To determine whether dsx/Gad1 neurons in the brain or in the Abg caused the genital uncoupling phenotype , we combined the dsxGAL4-DBD/Gad1p65-AD driver with the brain-specific flippase Otd-nls:FLPo ( OtdFLP ) ( Asahina et al . , 2014 ) and UAS-driven reporters or effectors with FRT-flanked stop cassettes ( Figure 3—figure supplement 2 ) . This intersectional combination enabled selective expression of reporters and effectors in dsx/Gad1 neurons only in the brain ( Figure 3—figure supplement 2 ) . Otd/dsx/Gad1>TNT males with silenced dsx/Gad1 brain neurons spent less time courting females than controls ( Figure 3—figure supplement 2c ) but exhibited no copulatory defects ( Figure 3—figure supplement 2d , e ) . In addition , thermoactivation of Otd/dsx/Gad1>dTrpA1 males , both pre-copulation and in copulo , did not impair courtship or copulation ( Figure 3—figure supplement 2f–i ) . Comparing these data to the phenotypes of dsx/Gad1>TNT males ( Figure 3e–i ) demonstrate that the observed copulation defects stem from the dsx/Gad1-Abg neurons . The opposing functions of dsx/Gad1 and dsx/vGlut neurons suggest that dsx/Gad1 neurons may inhibit dsx/vGlut neurons . To test this hypothesis anatomically , we expressed the presynaptic reporter UAS-nSyb-GFP ( Estes et al . , 2000 ) in dsx/Gad1 neurons ( Figures 4a1-a1’ ) , and the dendritic marker ‘DenMark’ ( UAS-DenMark ) ( Nicolaï et al . , 2010 ) in dsx/vGlut neurons ( Figures 4a2-a2’ ) independently , and registered the images to a standardized template VNC ( Jefferis et al . , 2007; Cachero et al . , 2010; Ostrovsky et al . , 2013 ) ( Figure 4a–a’ ) . Computational alignment revealed that dsx/vGlut dendrites overlap with dsx/Gad1 presynaptic boutons ( Figure 4a–a’; Video 5 ) , indicating that dsx/Gad1 neurons are anatomically positioned to inhibit dsx/vGlut neurons . 10 . 7554/eLife . 20713 . 015Figure 4 . dsx/glutamatergic neurons are poised for dsx/GABAergic inhibition . ( a–a2 ) Overlay of expression of ( a1 ) dsx/Gad1 presynaptic boutons and dsx/vGlut dendrites on standardized template VNC . ( a1 ) dsx/Gad1 presynaptic boutons visualized by dsxGAL4-DBD/Gad1p65-AD driving UAS-nSyb::GFP in male VNC stained with anti-GFP ( green ) . ( a2 ) dsx/vGlut dendrites visualized by vGlutdVP16-AD/dsxGAL4-DBD driving UAS-DenMark in male VNC stained with anti-DsRed ( magenta ) ; neuropil counterstained with anti-nC82 ( blue ) . ( a–a2 ) Maximum intensity z-stacks and ( a’–a2’ ) single section images . Solid arrowheads point at regions of close proximity . Scale bar = 50 μm . See also Video 5 . ( b ) Effects of knocking down GABAB-R2 receptor by RNAi in dsx/vGlut neurons on male copulatory behavior . Percentage of males displaying a ‘stuck’ phenotype is graphed ( n = 20–30 ) . *p<0 . 05 by Fisher exact test . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 01510 . 7554/eLife . 20713 . 016Figure 4—figure supplement 1 . Effects of knocking down GABA receptor subunits in dsx/glutamatergic neurons on copulatory behaviors . ( a ) Percentage of successful matings in 1 hr ( n = 24–30 ) . ( b ) Copulation duration ( n = 12–24 ) . ( c ) Male fertility ( n = 30 ) . Genotypes indicate males . Statistical comparisons of the experimental genotype were made against controls of the same receptor . ( d ) dsx/Gad1 neurons have prolific dendrites in regions of the Abg occupied by genital sensory terminals . dsx/Gad1 dendrites visualized by dsxGAL4-DBD/Gad1p65-AD driving UAS-DenMark in male VNC . Maximum intensity z-projection ( d ) and higher magnification image of 10 μm sub-stack ( d’ ) is shown . DenMark stained with anti-dsRed ( green ) ; neuropil counterstained with anti-nC82 ( magenta ) . Scale bar = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 01610 . 7554/eLife . 20713 . 017Video 5 . dsx/Gad1 presynaptic boutons are in close proximity to dsx/vGlut dendrites . This movie shows the 3D reconstruction of a template abdominal ganglion showing an overlay of dsx/vGlut dendrites ( magenta ) and dsx/Gad1 presynaptic boutons ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 017 To test this hypothesis functionally , we asked whether knocking down GABA receptors in dsx/vGlut neurons would recapitulate the dsx/Gad1>TNT phenotype . We behaviorally screened all five GABA receptor subunits by RNAi knockdown in dsx/vGlut neurons ( Figure 4b and Figure 4—figure supplement 1a–c ) . dsx/vGlut>GABAB-R2-RNAi knockdown resulted in a significant defect in copulation termination , with ~15% of males displaying a ‘stuck’ phenotype ( Figure 4b ) . The reduced strength of this phenotype compared to dsx/Gad1>TNT flies may be due to incomplete knockdown of the GABAB receptor by RNAi ( Dietzl et al . , 2007; Lin et al . , 2014 ) . These results suggest that dsx/Gad1 neurons help terminate copulation at least partially through inhibiting dsx/vGlut neurons via metabotropic GABAB receptors . Taken together , our data support the notion that dsx/vGlut neurons are poised for dsx/Gad1 inhibition . Mechanosensory sensilla on the male genital claspers and lateral plates have species-specific roles in establishing correct mating posture and genital coupling during the initial stages of copulation ( Acebes et al . , 2003; Jagadeeshan and Singh , 2006 ) . We postulated that sensory information from the genitalia could provide feedback to neurons controlling copulation during genital attachment and copulation . To identify sensory neurons , we examined the peripheral nervous system ( PNS ) in dsx/elav>GFP flies ( Figure 1d ) . We uncovered novel patterns of dsx-expression in the PNS ( Figure 5 ) . As expected , we identified sexually dimorphic patterns of dsx-expression in several bilateral clusters of mechanosensory neurons of the male and female adult terminalia ( Figure 5b1 , c1 ) , in addition to the foreleg and labellum of both sexes ( Figure 5b2–3 , c2–3 , respectively ) . In males , dsx neurons in the terminalia are associated with bristles of the clasper teeth , lateral plates and anal plates ( Figure 5b1 ) , and are cholinergic ( Yasuyama and Salvaterra , 1999 ) . In females , they are associated with bristles of the anal ( not shown ) and vaginal plates ( Figure 5c1 ) . 10 . 7554/eLife . 20713 . 018Figure 5 . Novel patterns of dsx expression in the male and female peripheral nervous system . ( a ) Cartoon of adult fly depicting regions of dsx-expression in peripheral sense organs in males and females; a1: terminalia , a2: foreleg and a3: labellum of the mouthparts . ( b , c ) Sexually dimorphic dsxGAL4-DBD/elavVP16-AD expression in peripheral sense organs in male and female adult flies . dsx/elav EGFP expression in bristle sensory neurons of the ( b1 ) male clasper teeth , lateral plates , hypandrium and anal plates of the male terminalia and ( c1 ) female vaginal plates of the female terminalia . dsx/elav EGFP expression in sensory neruons of the T1 tarsus of the foreleg in both males ( b2 ) and females ( c2 ) . dsx/elav EGFP expression in sensory neruons of the labellum in both males ( b3 ) and female ( c3 ) . EGFP is shown in green . Scale bar = 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 018 We asked whether these neurons form arborizations that overlap with dsx/Gad1 or dsx/vGlut Abg neurons and therefore might provide input to them . Lacking drivers that specifically label mechanosensory neurons of the genitalia , we used the lipophilic carbocyanide DiD to retrograde label axons of mechanosensory neurons of the claspers , lateral plates , and hypandrium . Consistent with previous findings , we observed sexually dimorphic axonal projections , which show evidence of somatotopic organization in the Abg ( Figure 6a–g; Videos 6–8 ) ( Taylor , 1989 ) . Interestingly however , by incubating the dye for 10 days , we also identified for the first time a single afferent axon ( per hemisegment ) from the claspers that arborizes contralaterally within the Abg and traverses the entire VNC to ultimately terminate in the SEZ of the brain ( Figure 6h ) . We then performed these dye fills in flies expressing the dendritic marker ‘DenMark’ in either dsx/Gad1 or dsx/vGlut neurons ( Figure 6i , i’ ) . Native expression of DenMark in dsx/Gad1 neurons was too weak to visualize dendritic boutons of dsx/Gad1 neurons and establish a definitive relationship with genital neurons . However , dsx/Gad1 neurons were observed to have prolific dendrites in regions of the Abg occupied by genital sensory terminals ( Figure 4—figure supplement 1d ) . Interestingly , projections of clasper , lateral plate and hypandrium neurons clearly interdigitated with the DenMark signal in dsx/vGlut in the Abg ( Figure 6i , i’ ) , suggesting that dsx/vGlut excitatory Abg neurons are poised to receive sensory information from the genitalia . 10 . 7554/eLife . 20713 . 019Figure 6 . Mechanosensory neurons of the genitalia arborize onto the Abg and brain , and interdigitate with glutamatergic dsx motor neurons in the Abg . ( a ) Schematic of male terminalia depicting bristles ( right ) and bristle topography ( dots on left ) . Dye-filled bristle topography in shown with green dots . ( b-d ) Schematic of representative lateral plate ( b ) , clasper ( c ) , and hypandrium ( d ) arborizations in the abdominal ganglion ( Abg ) of male flies , as previously described ( Taylor , 1989 ) . ( e-g ) Representative images showing topographically distinctive patterns of dye-filled genital neuron arborizations in the male Abg . Maximum intensity z-projections of confocal stacks showing unilateral arborizations of ( e ) lateral plate , ( f ) clasper , and ( g ) hypandrium neurons in the male Abg , which reiterate previously described arborizations ( Taylor , 1989 ) . Afferent projections from lateral plate and clasper neurons occupy the same dorso-ventral area but differ in their anterior to posterior positions within the Abg , with clasper neurons ending more posteriorly than lateral plate neurons ( b , c and e , f ) . See also Videos 6 , 7 . Hypandrium neurons typically exhibit a unique contralateral arborization pattern within the Abg ( d , g ) . See also Video 8 . ( h ) A subset of clasper neurons project to the brain . Unilateral dye-fill of clasper neurons together with extended incubation ( 10 days ) reveals single afferent axon ( per hemisphere ) that transverses the VNC and terminates in the subesophageal zone ( SEZ ) of the brain . DiD dye-filled arborizations shown in white . ( e-h ) DiD dye-filled arborizations shown in white . Boundaries of Abg and brain shown with dotted white line . Afferent projections of dye-filled neurons traced with dotted yellow line . D , dorsal , L , lateral , P , posterior , V , ventral . Scale bar = 25 μm . ( i ) Arborisations of clasper , lateral plates and hypandrium neurons interdigitate with dsx/vGlut dendrites in the adult male Abg . Neurons of all three genital structures were unilaterally dye-filled in males expressing dendritic marker ( UAS-DenMark ) in dsx/vGlut neurons . Maximum intensity Z-projection of Abg ( i ) and 10 μm sub-stack ( i’ ) show overt interdigitation ( arrowheads ) between neurons of all three genital structures and dendrites of dsx/vGlut neurons in the Abg . DenMark shown in magenta; DiD shown in green . Scale bar = 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 01910 . 7554/eLife . 20713 . 020Video 6 . Clasper neurons of the male genitalia innervate the abdominal ganglion . This movie shows the 3D reconstruction of an adult male abdominal ganglion with innervations of dye-filled neurons from bristles on the clasper of the male genitalia . White: Lipophilic dye ( DiD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 02010 . 7554/eLife . 20713 . 021Video 7 . Lateral plate neurons of the male genitalia innervate the abdominal ganglion . This movie shows the 3D reconstruction of an adult male abdominal ganglion with innervations of dye-filled neurons from bristles on the lateral plate of the male genitalia . White: Lipophilic dye ( DiD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 02110 . 7554/eLife . 20713 . 022Video 8 . Hypandrium neurons of the male genitalia innervate the abdominal ganglion . This movie shows the 3D reconstruction of an adult male abdominal ganglion with innervations of dye-filled neurons from bristles on the hypandrium of the male genitalia . White: Lipophilic dye ( DiD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 022 To establish direct functional connectivity between dsx/vGlut or dsx/Gad1 neurons and mechanosensory neurons of the genitalia we expressed the calcium indicator GCaMP6m ( UAS-GCaMP ) ( Chen et al . , 2013 ) in dsx/vGlut or dsx/Gad1 neurons and imaged responses evoked by genital stimulation ( Figure 7 ) . We stimulated the genital sensory neurons using a minutien pin attached to a micromanipulator . As a negative control , we stimulated segment A5 of the abdomen . Both dsx/vGlut and dsx/Gad1 Abg neurons responded strongly to genital , but not abdominal , touch ( Figure 7c–j , Videos 9 , 10 ) . Only a subset of neurons responded; activity maps show that not all areas of the neuropil respond significantly ( Figure 7d , g; Videos 9 , 10 ) , which may be due to regional mechanosensory stimulation . Nonetheless , these results show that sensory neurons of the genitalia functionally connect to dsx/vGlut and dsx/Gad1 Abg neurons , and suggest that sensory feedback during copulation alters the activity of an Abg circuit that controls genital coupling and copulation . 10 . 7554/eLife . 20713 . 023Figure 7 . Glutamatergic and GABAergic dsx neurons of the Abg respond to mechanical stimulation of genitalia . ( a , b ) Examples of pin touching a male fly’s abdomen on ( a ) segment A5 and ( b ) genitalia . The fly is illuminated at the VNC by the 910 nm two-photon laser and imaged with an infrared-sensitive camera . ( c , d ) dsx/vGlut>GCaMP6m neuropil in the abdominal ganglion: Pseudocolored activity maps of responses to ( c ) abdominal or ( d ) genital touch , overlaid on grayscale baseline fluorescence See also Video 9 . There is no response to abdominal touch ( c ) . Dotted outline indicates region of interest for panel E . L , lateral , P , posterior . ( e ) ∆F/F of the outlined region in panel D . Bars under traces represent abdominal ( blue ) or genital ( red ) touch . The two traces come from a single movie . ( f , g ) as with ( c , d ) but for dsx/Gad1>GCaMP6m neuropil . See also Video 10 . ( h ) as with ( e ) but referring to the outlined region in ( g ) . ( i ) Average of ∆F/F traces as in ( e ) and ( h ) , aligned to touch onset . dsx/vGlut and dsx/Gad1 neurons in the abdominal ganglion respond strongly to genital touch ( red ) but not abdominal touch ( blue ) . Traces: average ∆F/F ( fluorescence normalized to baseline ) ; shading , S . E . M . ; vertical line , onset of touch . Only touches < 3 s long are included . n = 7 ( vGlut genital ) , 5 ( vGlut abdominal ) , 6 ( Gad1 genital and abdominal ) . ( j ) Average ∆F/F 0–5 s after onset of touch is significantly larger for genital touch than abdominal touch . ** p<0 . 01 , Mann-Whitney test . Scale bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 02310 . 7554/eLife . 20713 . 024Video 9 . dsx/vGlut Abg neurons respond to mechanical stimulation of genitalia . The upper panel shows GCaMP6m signal in dsx/vGlut neuropil in the Abg and the lower panel shows the simultaneous view of the fly’s abdomen , illuminated by the 910 nm laser used for two-photon imaging . The minutien pin touches the genitalia at 0:05 and 0:12 depicted with *** . Movies are 5x actual speed and false colored . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 02410 . 7554/eLife . 20713 . 025Video 10 . dsx/Gad1 Abg neurons respond to mechanical stimulation of genitalia . The upper panel shows GCaMP6m signal in dsx/Gad1 neuropil in the Abg and the lower panel shows the simultaneous view of the fly’s abdomen , illuminated by the 910 nm laser used for two-photon imaging . The minutien pin touches the genitalia at 0:04 and 0:08 depicted with *** . Note that the pin approaches the genitalia but does not quite touch until 0:04 . Movies are 5x actual speed and false colored . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 025
This study defines a sexually dimorphic motor circuit in the Abg that mediates the action sequence of copulation in males . We identified three core dsx-expressing neuronal types – motor neurons , interneurons and mechanosensory neurons – that control the mechanics of copulation . Excitatory motor neurons promote genital coupling and they are opposed by local inhibitory neurons , which prevent it , while sensory neurons of the genitalia provide sensory feedback to the system to ensure a coordinated sequence of motor events that result in successful copulation ( Figure 8 ) . 10 . 7554/eLife . 20713 . 026Figure 8 . Model of circuit organization underlying copulation in males . ( a ) Musculature of male genitalia and terminalia involved in copulation . Protractor muscles shown in orange . Retractor muscles shown in blue . Muscles with no designated colour have unknown functions . D: dorsal , V: ventral . ( b ) dsx/vGlut motor neurons local to the Abg ( blue ) mediate genital coupling by controlling muscles of the phallic and periphallic organs . dsx/Gad1 Abg neurons ( red ) , depicted as a heterogeneous population of neurons , some of which inhibit glutamatergic neurons that control copulatory muscles ( bottom ) , and other which shorten copulation duration ( stopwatch ) by reducing copulation motivation by inhibiting dopaminergic ( DA ) neurons ( top ) . dsx sensory neurons of the genitalia ( grey ) innervate the Abg and brain , and are anatomically and functionally connected to dsx/vGlut motor neurons ( blue ) and dsx/Gad1 inhibitory centres ( red ) of the Abg , likely aiding the male in adopting the correct posture to successfully achieve copulation . Brackets depict control over all encompassing neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 20713 . 026 The tripartite motor circuit that controls the movement of the genitalia during the initiation and termination of copulation is reminiscent of the spinal microcircuits that coordinate limb movements in mammals ( Miri et al . , 2013 ) . These microcircuits exhibit three features , the first of which involves coordinated firing of motor neurons that trigger the contraction of muscles in the appropriate appendages . dsx/vGlut neurons encompass all motor neurons innervating the genital muscles , including both protractor and retractor muscles associated with the phallic and periphallic organs ( Figure 8a ) ( Kamimura , 2010 ) . Triggering the contraction of these muscles in a sequence-specific manner allows the male to clasp to the female’s oviscape prior to protracting the aedeagus and achieving intromission ( Figure 8a , b ) . Prematurely stimulating these contractions likely prevents attachment entirely , much as a human hand cannot grasp an object if it is clenched in a fist . Such a scenario would explain why thermoactivating all dsx/vGlut neurons blocks copulation , rather than stimulating it , while thermoactivating all dsx/vGlut neurons in copulo blocks termination of copulation ( Figure 2 ) . While the dsx/vGlut population may also include interneurons , this possibility is unlikely to affect the direct artificial activation and blockade of dsx/vGlut motor neurons , which are the final output of the circuit . The second critical feature of mammalian motor circuits is inhibition of motor neurons by local interneurons , which facilitates the initiation and coordination of limb movement . Here , dsx/Gad1 neurons likely inhibit dsx/vGlut neurons both to terminate copulation and to ensure successful copulation ( Figure 8b ) . Copulation termination requires both dsx/Gad1 synaptic transmission ( Figure 3 ) and GABAB receptor expression in dsx/vGlut neurons ( Figure 4 ) , suggesting that dsx/Gad1 neurons terminate copulation by metabotropically inhibiting motor neurons . This is supported by the finding that activation of dsx/Gad1 neurons in copulo results in immediate termination of copulation ( Figure 3 ) . dsx/Gad1 interneurons may also prevent males from protracting their copulatory organs indiscriminately . These could be inhibited by descending signals to promote copulation or activated by competing drives upon insemination when continuing copulation is no longer necessary ( Figure 8b ) ( Crickmore and Vosshall , 2013 ) . Aside from termination , blocking dsx/Gad1 neurons also reduces successful copulation ( Figure 3 ) , suggesting that dsx/Gad1 neurons also modulate the timing of motor neuron activity to achieve the correct sequence of muscle contractions required for copulation , perhaps through reciprocal inhibition ( Figure 8b ) . Although we observed a ‘stuck’ phenotype by knocking-down the GABAB-R2 receptor in dsx/vGlut neurons , the role of GABA reception in these neurons is far from resolved ( Figure 4—figure supplement 1 ) . We would be surprised if GABAA receptors were not involved since the coordination of fast motor behavior would normally involve fast-responding GABA receptors . These RNAi results may be due to incomplete knockdown and/or homeostatic adaptation ( Lin et al . , 2014 ) . A direct physiological test of how dsx/Gad1 neurons affect motor neuron activity during copulation awaits new techniques for recording neural activity in the Abg in a copulating male . The final feature of motor circuits is sensory input , which is required to facilitate fine motor coordination . Sensory neurons of the external genitalia likely provide the initial computations about orientation by extracting salient tactile features of the female that lead to appropriate genital attachment , e . g . , during attempted copulation , an objectively purposeful behavior , the male actively acquires sensory inputs ( ‘active sensing’ ) about the female genitalia . These neurons feed forward to dsx/vGlut neurons , dsx/Gad1 interneurons , and to neurons in the brain , suggesting that sensory feedback signals not only local motor circuits in the abdominal ganglion but also higher-level control of male sexual behavior by the brain . Perhaps a function of this input in copulo is to aid the establishment of a dynamic balance of excitation and inhibition that mediates the appropriate positioning of the male throughout an entire copulation event . If dsx/vGlut neurons and dsx/Gad1 neurons oppose each other , then why do subsets of both populations respond to genital touch ( Figure 7 ) ? One explanation might be that the population of dsx/vGlut neurons activated by genital touch might not be inhibited by the population of dsx/Gad1 neurons that are also activated by such stimulation . Alternatively , sensory excitation of dsx/vGlut neurons might outweigh inhibition from dsx/Gad1 neurons , or some dsx/vGlut neurons might indeed be inhibited by genital touch ( via dsx/Gad1 neurons ) , but inhibition of silent neurons would not give a GCaMP response . Addressing this question awaits the discovery of new driver lines that label distinct subsets of dsx/vGlut and dsx/Gad1 neurons and enable simultaneous imaging , and the development of an image registration methodology that is capable of precise identification of individual neurons in the highly dense Abg between imaging preparations . Male copulation is made up of several distinct features: genital coupling , sperm transfer , and copulation duration ( which is thought to be a product of declining ‘persistence’ ) ( Crickmore and Vosshall , 2013 ) . Here , we have dissociated genital coupling from all other features; activating dsx/vGlut neurons in copulo or silencing dsx/Gad1 neurons produces a ‘stuck’ phenotype , where males successfully transfer sperm and attempt to terminate copulation at the appropriate time , but cannot uncouple from the female . A recent study has shown that sperm transfer is mediated by corazonin-expressing neurons activating serotonergic neurons that innervate the accessory glands ( Tayler et al . , 2012 ) . Our results show that these neurons are distinct from dsx/Gad1 and dsx/vGlut neurons and that sperm transfer does not require ( although it may still be modulated by ) dsx/Gad1 neurons or dsx/vGlut neurons . In addition , we demonstrate a dissociation of genital coupling and copulatory ‘persistence’ at the neuronal level . A recent study identified ~8 putative dsx/GABAergic neurons that control copulation ‘persistence’ together with dopaminergic neurons in the VNC ( DA; Figure 8b ) ( Crickmore and Vosshall , 2013 ) . Blocking these eight putative dsx/GABAergic neurons does not affect genital uncoupling but rather results in abnormally long copulation durations , after which males detach without struggle ( Crickmore and Vosshall , 2013 ) . Here , blocking the full complement of 150 dsx/Gad1 neurons does not affect copulation persistence but rather prevents genital uncoupling , evidenced by the ‘stuck’ phenotype . This discrepancy can be explained if the full complement of dsx/Gad1 Abg neurons described here are a heterogeneous population of neurons , some that shorten copulation by reducing copulation motivation and others that inhibit glutamatergic neurons that control copulatory muscles ( Figure 8b ) . Aberrant motor neuron activity patterns caused by blocking the latter might change sensory or internal feedback to the dopaminergic neurons controlling copulation persistence , explaining why blocking all dsx/Gad1 neurons causes a ‘stuck’ phenotype rather than increasing copulation persistence . Such ‘specialist’ inhibitory interneurons are present in many mammalian spinal circuits that drive complex movements ( Graziano , 2006 ) . Our results reveal that distinct populations of dsx/GABAergic neurons likely control different features of copulation ( genital coupling and persistence ) . If the Abg can act autonomously to control copulatory behaviors ( Billeter et al . , 2006; Pan et al . , 2011; Tayler et al . , 2012; Crickmore and Vosshall , 2013; Inagaki et al . , 2014 ) , then what role does the brain play ? In mammals , brain circuits are thought to compute adaptive neural functions that underlie action-selection ( or ‘decision-making’ ) , while spinal circuits communicate bidirectionally with the brain , such that descending pathways activate motor programs and ascending pathways report on their execution ( Arber , 2012 ) . This organizing principle also exists in the fly: male-specific ‘decision-making’ neurons in the brain ( e . g . , P1; pC2I; dopaminergic neurons ) integrate multimodal sensory cues to initiate and modulate courtship and copulatory behaviors ( Kimura et al . , 2008; Yu et al . , 2010; Kohatsu et al . , 2011; von Philipsborn et al . , 2011; Pan et al . , 2012; Inagaki et al . , 2014; Kohatsu and Yamamoto , 2015; Zhang et al . , 2016 ) , while descending ‘command’ neurons ( e . g . , pIP10 ) relay information from the ‘decision-making’ neurons , to activate ‘pattern generator’-like motor circuits in the VNC ( Clyne and Miesenbock , 2008; von Philipsborn et al . , 2011; Inagaki et al . , 2014 ) . We similarly predict that higher-order neurons in the brain control the initiation and modulation of the copulation motor circuit described here . The culmination of the courtship ritual would activate ‘decision-making’ neurons that signal , via descending ‘command’ neurons , to dsx glutamatergic and GABAergic neurons in the Abg , to initiate genital coupling . Meanwhile , descending and ascending signals from these neurons ( including sensory feedback from the genitalia ) would modulate the activity of the motor circuit in response to the male’s internal state and the environment . Indeed , males mate for longer and increase their sperm load when in the presence of rival males ( Bretman et al . , 2009; Garbaczewska et al . , 2013 ) . This behavioral responsiveness most likely relies on physiological modulation of ‘decision-making’ neurons , much as increased courtship behavior in socially isolated male flies correlates with increased excitability of ‘decision-making’ neurons for courtship ( Inagaki et al . , 2014 ) . In addition , while the dsx/vGlut and dsx/Gad1 populations are sexually dimorphic , it is interesting to speculate that similar organizational principles may govern female reproductive behavior , with the analogous female dsx/vGlut and dsx/Gad1 neurons controlling behaviors like ovipositor extrusion and retraction . Our findings provide insight into the circuit logic underlying genital coupling , and serve as an entry point to a circuit-level dissection of copulatory behavior and its modification by social experience .
GAL4-DBD was targeted to the dsx locus by ends-in homologous recombination as previously described ( Rideout et al . , 2010 ) . In a single cloning step , the GAL4-DBD coding sequence was excised from pCaST-elavGAL4-DBD ( gift from B . White ) with BglII and BamHI restriction enzymes , and ligated into the original dsxGAL4 construct after digestion with BamHI . PCR was used to confirm the predicted recombination event in multiple lines . Two lines were selected for further analysis using multiple UAS reporter transgenes and both exhibited equitable patterns of expression ( data not shown ) , of which a single dsxGAL4-DBD line was then chosen for this study . dsxGAL4-DBD not a dsx null mutant; dsxGAL4-DBD homozygotes are fertile and show normal morphology ( data not shown ) . All flies were raised at 25°C on standard medium in a 12 hr light/12 hr dark cycle at 50% relative humidity . Fly strains used in this study include wild-type Canton-S; dsxGAL4-DBD ( generated in this study ) ; dsxGAL4 ( Rideout et al . , 2010 ) ; elavVP16-AD ( Luan et al . , 2006 ) ; vGlutOK371-dVP16-AD denoted as vGlutdVP16-AD ( Gao et al . , 2008 ) ; Gad1p65-AD ( Diao et al . , 2015 ) ; UAS-TNTG ( Sweeney et al . , 1995 ) ; UAS-dTrpA1 ( Hamada et al . , 2008 ) ; UAS-pStingerII denoted as UAS-nGFP ( Barolo et al . , 2000 ) ; UAS-2XEGFP ( Halfon et al . , 2002 ) ; UAS-DenMark ( Nicolaï et al . , 2010 ) ; UAS-nSyb::GFP ( Estes et al . , 2000 ) ; UAS-GCaMP6m ( Chen et al . , 2013 ) ; UAS>stop>myrGFP ( Yu et al . , 2010 ) ; UAS>stop>TNT and UAS>stop>dTrpA1myc ( von Philipsborn et al . , 2011 ) ; Otd-nls:FLPo ( Asahina et al . , 2014 ) ; UAS-GABAB-R1-RNAi ( VDRC: 101440 ) ; UAS-GABAB-R2-RNAi ( VDRC: 1784 ) ; UAS-GABAB-R3 ( Bloomington: 26729 ) ; UAS-Lcch3-RNAi ( VDRC: 37408 ) and UAS-Rdl-RNAi ( Liu et al . , 2007 ) . All lines and transgenes were in a w+ background for behavioral studies . Crosses with Split-GAL4 and UAS-effectors were raised at 21°C in a 12:12 hr light:dark cycle and at 50% relative humidity . Crosses containing Otd-nls:FLPo , Split-GAL4 and UAS>stop>effectors transgenes or RNAi transgenes were raised at 25°C in a 12:12 hr light:dark cycle and at 50% relative humidity . Flies were reared at 25°C and aged for 4–6 days prior to dissection and staining as previously described ( Rideout et al . , 2010 ) . Samples were dissected in PBS and fixed in 4% ( w/v ) paraformaldehyde ( in PBS ) for 20 min at room temperature ( RT ) . Primary antibody incubation was carried out for 24–48 hr at 4°C . Secondary antibody incubation was carried out for 24 hr at 4°C . Primary antibodies used included: rabbit anti-GFP ( 1:1000 , Invitrogen Molecular Probes , Carlsbad , CA ) , chicken anti-GFP ( 1:1000 , Abcam , UK ) , rabbit anti-DsRed ( 1:1000 , Clontech ) ; mouse mAb nC82 ( 1:10 , DSHB , Univ . of Iowa , IA ) , rabbit anti-βGal ( 1:1000; Cappel , ICN ) , rabbit anti-FruM at 1:3000 ( Billeter et al . , 2006 ) ; rabbit anti-dvGlut at 1:500 ( Mahr and Aberle , 2006 ) and rabbit anti-GABA ( 1:2000 , Sigma-Aldrich ) . Secondary antibodies used included: anti-chicken Alexa Fluor488 , anti-rabbit Alexa Fluor488 , anti-rabbit Alexa Fluor546 , anti-mouse Alexa Fluor546 , anti-rat Alexa Fluor546 , anti-mouse Alexa Fluor633 , anti-rat Alexa Fluor633 , anti-rat Cy5 ( 1:300 , Invitrogen Molecular Probes , Carlsbad , CA ) , HRP-Cy3 conjugate ( 1:100 , Sigma-Aldrich ) and Phalloidin-TRITC and -Alexa Fluor633 conjugates ( 1:100 , Sigma-Aldrich ) . Samples were mounted with Vectashield ( Vector Labs ) and imaged with an Olympus FluoView FV1000 confocal microscope x10 ( air ) , x20 ( air ) , x40 ( oil immersion ) , and x63 ( oil immersion ) objectives . For multi-track ( multiple fluorophore labels ) imaging , each wavelength was sequentially scanned for each optical section through the sample to excite each fluorophore individually and avoid bleed-through . Stacks of optical sections were generated at 1 µm intervals . Images were processed in Imaris ( Bitplane Scientific , AG , Zürich ) and peripheral debris was removed in Adobe Photoshop 7 . 0 . ( Adobe Systems Inc . , San Jose , CA ) . For cell counts , stacks of optical sections obtained by confocal microscopy were transformed into maximum Z-stack projections in Imaris . The fluorophore labeling cells of interest was used as the ‘source channel’ for automatic detection of spherically labeled nuclei with a 3 µm minimum in diameter using the Imaris ‘Spots’ detection module . To ensure that all obvious cells were marked , the threshold was manually shifted to the point at which it automatically detects the maximum number of spherical nuclei , without any observable ectopic detection . Subsequent use of the ‘orthoslicer’ tool allowed for examination of each optical slice . Un-marked nuclei that fell short of automatic-detection were manually marked . A final count of the number of marked nuclei was subsequently calculated . For brain image registration , confocal images of male dsx/vGlut>nSyb and dsx/vGlut>DenMark VNCs were registered onto a D . melanogaster intersex template VNC that was generated by the Jefferis lab ( http://zenodo . org/record/10591# ) , using a Fiji graphical user interface ( GUI ) , and the previously described ( Jefferis et al . , 2007; Cachero et al . , 2010; Ostrovsky et al . , 2013 ) . 2 VNCs per genotype were used for the analysis . The protocol to label the complex axonal branching of mechanosensory neurons of the male genitalia with a lipophilic fluorescent dye is adapted from established protocol ( Kays et al . , 2014 ) . Adult ( 2–3 day old ) male flies were perpendicularly glued onto insect pin heads , decapitated ( or not ) and fixed in 3 . 7% paraformaldehyde in in 0 . 2 M carbonate-bicarbonate buffer at pH 9 . 5 overnight at 4°C . Flies were subsequently washed with ddH2O and gently dried with a Kimwipe tissue . Mechanosensory neurons of the lateral plates , claspers and hypandrium were each or all unilaterally dyed by sub-cuticular injection of DiD dye ( 80 μg/μL in 100% ethanol; Life Technologies , cat . no . D7757 ) using a micropipette and micromanipulator ( settings for a Sutter P-97 Flaming/Brown micropipette puller using standard borosilicate glass of o . d . /i . d . 1 . 00 mm/0 . 78 mm with filament are as follows: heat = 515 , pull = 30 , velocity = 30 , time = 165 ) . Pins with attached dye-filled flies were wedged into clay such that the thorax of the flies was submerged in 0 . 2 M carbonate-bicarbonate buffer while their abdomen and genitals were left above the surface of the buffer . Samples were incubated in the dark for six days to realize innervations in the Abg and 10 days to realize innervations in the brain . CNSs and VNCs were dissected in PBS and imaged with an Olympus FluoView FV1000 confocal microscope within 10 min of mounting in Vectashield ( Vectorlabs ) . 1–3 day old male flies were waxed to tin foil in a perfusion chamber such that the ventral thorax faced the objective through a small hole in the foil . The legs , cuticle and trachea covering the abdominal ganglion were removed and the preparation was superfused with solution ( in mM: TES 5 , NaCl 103 , KCl 3 , CaCl2 1 . 5 , MgCl2 4 , NaHCO3 26 , NaH2PO4 1 , trehalose 8 , and glucose 10 , pH 7 . 3 ) bubbled with carbogen ( 95% O2 , 5% CO2 ) . The genitalia or abdomen were mechanically stimulated using a 0 . 1 mm stainless steel insect pin ( Fine Science Tools 26002–10 ) attached by a thin rod to a manual micromanipulator ( Märzhäuser Wetzlar MM-33 ) . Stimulation of the genitalia or abdomen was recorded at 4 . 07 Hz with a Stingray F-033B camera ( Allied Vision Technologies ) , using the illumination of the fly from the 910 nm laser during two-photon imaging . Two-photon imaging and data analysis was adapted from established protocol ( Lin et al . , 2014 ) . The abdominal ganglion was imaged at 4 . 34 Hz with a pixel dwell time of 3 . 2 µs . The baseline fluorescence ( F0 ) for calculating ∆F/F was defined as the average signal 1–5 s before stimulus onset . For activity maps , we excluded pixels where ∆F ( difference between F0 and mean fluorescence in the first 5 s after stimulus onset ) was less than twice the standard deviation of fluorescence during the F0 period . Regions of interest ( ROIs ) were manually drawn around responsive areas and ∆F/F traces were aligned to the start of genital or abdominal touch . To prevent slow-decaying GCaMP6m signals from the previous stimulus from interfering with this baseline calculation , genital stimuli were only used if they occurred more than 10 s after the start of the previous stimulus . Behavioral means were compared using Kruskal-Wallis ANOVA test and Dunn’s post hoc statistical test where indicated . For Fisher's exact test , two-tail p values were compared with controls . Statistical analyses were performed with the GraphPad Prism software ( version 6 . 0 , GraphPad Software Inc . ) . | Idioms and love songs often euphemistically refer to “the birds and the bees” . Yet for neurobiologists interested in uncovering basic facts about sex and reproduction , the fruit fly has proven much more informative . Male fruit flies court females with a series of “hard-wired” or genetically programmed behaviors . One gene called doublesex generates differences in the anatomy and behavior of males and females in many animal species . In male fruit flies , the doublesex gene is active in roughly 650 neurons , with specific groups of cells controlling distinct steps of the courtship ritual . However , it was not understood how the different steps involved in copulation were coordinated to ensure a successful mating . Pavlou et al . have now identified a circuit of doublesex-expressing neurons that controls copulation itself . The circuit , which is in the fruit fly’s equivalent of the spinal cord , is made up of three types of neurons: motor neurons , inhibitory interneurons and mechanosensory neurons . The motor neurons coordinate the joining of the male’s genitals with those of the female . The inhibitory interneurons promote the release of the male’s genitals by opposing the motor neurons , while the mechanosensory neurons possibly coordinate the activity of the other neurons to generate the correct sequence of events needed for copulation . Pavlou et al . also showed that the mechanism that controls how the male attaches to and detaches from the female is independent of ejaculation , indicating that the mechanics of copulation are separate from those of reproduction . A future challenge will be to understand how command centres in the brain combine these signals with sensory feedback to enable males to execute and modify their copulation-related behaviors . Identifying neural circuits that drive behaviors in fruit flies provide insights into the universal principles by which a nervous system can coordinate complex motor behaviors such as walking and flying . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Neural circuitry coordinating male copulation |
Many microbial populations rapidly adapt to changing environments with multiple variants competing for survival . To quantify such complex evolutionary dynamics in vivo , time resolved and genome wide data including rare variants are essential . We performed whole-genome deep sequencing of HIV-1 populations in 9 untreated patients , with 6-12 longitudinal samples per patient spanning 5-8 years of infection . The data can be accessed and explored via an interactive web application . We show that patterns of minor diversity are reproducible between patients and mirror global HIV-1 diversity , suggesting a universal landscape of fitness costs that control diversity . Reversions towards the ancestral HIV-1 sequence are observed throughout infection and account for almost one third of all sequence changes . Reversion rates depend strongly on conservation . Frequent recombination limits linkage disequilibrium to about 100bp in most of the genome , but strong hitch-hiking due to short range linkage limits diversity .
The human immunodeficiency virus 1 ( HIV-1 ) is a paradigmatic example of a rapidly adapting population characterized by high diversity , strong selection , and recombination . HIV-1 has originated from multiple zoonotic transmissions from apes in the early part of the 20th century ( Sharp and Hahn , 2011 ) , one of which gave rise to the worldwide pandemic . This lineage is called group M and has diversified into different subtypes at a rate of about 1 in 1000 substitutions per year ( Lemey et al . , 2005; Li et al . , 2015 ) . Tens of thousand of HIV-1 group M variants are available in the Los Alamos National Laboratories HIV database ( LANL ) ( Foley et al . , 2013 ) . The evolution of HIV-1 ultimately takes place within infected individuals and can be observed directly in longitudinal samples of virus populations from the same individual . The detailed knowledge of HIV-1 biology paired with historical samples makes HIV-1 an ideal system to study general features of evolution at high mutation rates and strong selection that are otherwise only accessible in evolution experiments ( Elena and Lenski , 2003; Miralles et al . , 1999 ) . During the first few months of an HIV-1 infection , the viral population typically acquires several mutations that mediate escape from cytotoxic T-cells ( CTL ) . The properties and dynamics of these mutations and their effect on epitope presentation and virus control have been extensively studied ( Bernardin et al . , 2005; Goonetilleke et al . , 2009; Jones et al . , 2004; Kearney et al . , 2009; Liu et al . , 2011 ) . Such escape variants often compromise virus replication and have been shown to revert upon transmission of the virus to a human leukocyte antigen ( HLA ) mismatched host ( Brockman et al . , 2010; Leslie et al . , 2004; Li et al . , 2007 ) . However , the quantitative contribution of escape and reversion to HIV-1 evolution and the degree to which costly mutations are compensated by additional mutations is less well understood ( Lythgoe and Fraser , 2012; Schneidewind et al . , 2009 ) . Similar to CTL escape mutations , drug resistance mutations can spread through the viral population within weeks and longitudinal sequence data has been used to study emergence and fitness cost of resistance mutations ( Hedskog et al . , 2010; Little et al . , 2008; Paredes et al . , 2009 ) . Longitudinal sequence data has also been used to track evolution driven by the humoral immune response against HIV-1 which occurs throughout infection and results in high rates of evolution in the variable loops of HIV-1 surface proteins ( Bar et al . , 2012; Richman et al . , 2003; Shankarappa et al . , 1999 ) . In a pioneering study , Shankarappa et al . ( 1999 ) characterized HIV-1 evolution of parts of the gp120 envelope protein over approximately 6–12 years in 11 patients , demonstrating consistent patterns of diversity and divergence . While this early study was limited to about 10 sequences per sample , next-generation sequencing technology today allows deep characterization of intrapatient HIV-1 variability including rare mutations ( Fischer et al . , 2010; Hedskog et al . , 2010; Henn et al . , 2012; Tsibris et al . , 2009 ) . The earlier studies discussed above have either focused on early infection or particular regions of the HIV-1 genome . However , most virus evolution happens during chronic infection simultaneously at many locations in the genome . To develop a comprehensive and quantitative understanding of the evolution and diversification of HIV-1 populations , we generated a whole-genome , deep-sequencing data set covering nine patients over 5–8 years with 6–12 time points per patient . Importantly , the data set covers the entire genome such that no substitution is missed and includes early samples defining the initial population . To our knowledge , this is the only whole-genome deep sequencing data set with long follow-up of multiple patients . We provide interactive and intuitive web-based access to the data set and hope it will become a resource for many future studies , like the data set by Shankarappa et al . ( 1999 ) has been in the last years . Below , we first describe the methodology we developed to sequence the entire HIV-1 genome at great depth . We then analyze the intrapatient evolution of HIV-1 and show that the minor variants in the virus population explore sequence space in predictable fashion at the single site level . At the same time , we observe a strong tendency for reversion towards the global HIV-1 consensus that is not limited to early infection but occurs at an approximately constant rate throughout chronic infection . Reversion is more frequent at sites that are more conserved at the global level , suggesting a direct relationship between intrapatient fitness cost and global conservation . Together with reproducible patterns of intrapatient variation , this link explains why HIV-1 fitness landscapes can be inferred from cross-sectional data ( Dahirel et al . , 2011; Ferguson et al . , 2013; Mann et al . , 2014 ) . We further find frequent recombination , which allows the viral population to evolve independently in different regions of the genome . Nevertheless , recombination is not frequent enough to decouple mutations closer than 100 base pairs and we observe signatures of hitch-hiking at short distances ( Maynard Smith and Haigh , 1974 ) .
The basic steps of our amplification and sequencing pipeline are illustrated in Figure 1 and explained in detail in Materials and methods . Briefly , viral RNA was extracted from 400 μl patient plasma and used for one-step RT-PCR amplification with six overlapping primer sets that span almost the complete HIV-1 genome , similar to the strategy developed by Gall et al . ( 2012 ) . Sequencing libraries were made starting from 0 . 1–1 . 5 ng of DNA with a stringent size selection for long inserts ( >400 bp ) . Sequencing was performed on the Illumina MiSeq platform and sequence reads were quality filtered and assembled using an in-house data processing pipeline . In total , approximately 100 million reads passed the quality filtering . The coverage varied considerably between samples and amplicons , but was mostly of the order of several thousands or more , see Figure 1B . 10 . 7554/eLife . 11282 . 004Figure 1 . Sequencing , coverage , and error rates . ( A ) Schematics of the sample preparation protocol , see text and Materials and methods for details . ( B ) Read coverage for a representative sample . Coverage of separate PCR amplicons is shown in different hues , the black line is the total coverage . The coverage of PCR fragment F5 is lower than the other amplicons , but it is still larger than number of input HIV-1 RNA molecules; this situation is typical in our samples . ( C ) Each blue circle corresponds to a SNP frequency in amplicon F1 of a late sample of patient 11 , while red squares are SNP frequencies in the sequence data generated from 10 , 000 copies of plasmid NL4-3 . The histogram on the right shows the distribution of SNP frequencies in the patient sample and the control . Minor SNPs observed in reads generated from the plasmid , which represent PCR and sequencing errors , did not exceed 0 . 3% . SNP , single nucleotide polymorphism; RT-PCR , reverse transcriptase polychromase chain reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 00410 . 7554/eLife . 11282 . 005Figure 1—figure supplement 1 . Viral load and CD4+ cell counts for all nine patients . Samples that were selected for deep sequencing are indicated as black vertical arrows at the bottom of each panel . Viral load values at 499 or 49 counts/ml actually indicate negative tests , at the respective level of sensitivity ( 500 or 50 counts/ml ) . EDI , estimated date of infection . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 00510 . 7554/eLife . 11282 . 006Figure 1—figure supplement 2 . Tree of minor reads covering p17 from all samples colored by patient ID . No sample cross-contamination is observed . The isolated black tip is the HxB2 reference sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 006 Importantly , sequencing depth is determined not only by coverage but also by template availability and sequencing errors ( Iyer et al . , 2015 ) . We performed a number of control experiments to quantify templates and assess the accuracy of estimates of frequencies of single nucleotide polymorphism ( SNP ) . The results are summarized in the following section and described in detail in Materials and methods . We quantified the number of HIV-1 genomes that contributed to each sequencing library by PCR-limiting dilution ( median: 120 quartiles: 50–500 ) . Hence template availability , rather than coverage , determined the sequencing depth , see Figure 1B . Comparison with routine plasma HIV-1 RNA level measurements performed at time of sampling showed that the median template complementary DNA ( cDNA ) recovery efficiency was 30% . We estimated the error rate of the PCR and sequencing pipeline by amplifying and sequencing a plasmid clone . Figure 1C compares the SNP frequencies observed in a clone to those observed in a patient sample . After quality filtering , PCR and sequencing errors never exceeded 0 . 3% of reads covering a particular position . To detect and control for variation in PCR efficiency among fragments and skewed amplification of different variants , we compared frequencies of variants in overlaps between the six amplicons . A SNP in the overlap is amplified and sequenced twice independently and the concordance of the two measurements of variant frequencies was used to estimate the fragment specific depth , which is limited both by template input and PCR efficiency , see Materials and methods . Frequency estimates were often reproducible to within a few percent . Sometimes , however , variant recovery was poor ( mostly in fragment 5 ) and frequency estimates less accurate . Those fragments are flagged on the website and can be excluded from analyses requiring large depth . We minimized PCR recombination by reducing the number of PCR cycles and optimizing the reaction protocol ( see Materials and methods , Di Giallonardo et al . ( 2013 ) ; Mild et al . ( 2011 ) ) . Control experiments using mixtures of two cultured virus populations showed that less than 10% of reads had experienced RT-PCR recombination . Taken together , our control experiments show that depending on the sample and fragment , we could estimate frequencies of SNPs down to 1% accuracy ( corresponding to several hundred effective templates ) . In some cases , however , the template number was low or template recovery poor such that only presence or absence of a variant could be called . Furthermore , SNPs remained linked through cDNA synthesis , PCR and sequencing . Deep sequencing data sets like the one presented here require substantial filtering , mapping , and processing before they can be used to address specific biological questions . Hence depositing the raw data ( while important ) is of limited use for follow-up analyses . To facilitate reuse of our data , we developed a web application to explore the data set interactively and visualize the patterns of HIV-1 evolution in the study patients . In addition , the website provides easy access to processed data . These include phylogenetic trees , viral loads and CD4+ T cell counts , consensus sequences , major and minor haplotypes in different regions of the genome , frequencies of single nucleotide polymorphisms , and clean sequencing reads . The application allows browsing the data either by patient or by genomic region and provides composite interactive plots to explore how the virus population has changed over time . We hope that this web application , available at hiv . tuebingen . mpg . de , will encourage others to further analyze these data . In most patients , the virus populations were initially homogeneous and diversified over the years , as expected for an infection with a single or small number of similar founder viruses ( Keele et al . , 2008; Salazar-Gonzalez et al . , 2009 ) . In two patients , p3 and p10 , the first sample displayed diversity consistent with the transmission of several variants from the same donor , see Figure 3—figure supplement 1 . For each of the nine patients , we reconstructed the HIV-1 genome sequence of the first sample by an iterative mapping procedure described in Materials and methods . We use this initial consensus sequence to approximate the sequence of the transmitted founder virus ( es ) . Our first sample is estimated to be between 1 and 7 months into infection and a few mutations ( likely CTL escape mutations ) had probably spread through the viral population by that time . Thus , the initial consensus sequence probably will differ slightly from the true founder virus . The number of differences , however , will be small compared to the sequence divergence in the 5 or more years of follow-up such that this initial consensus sequence is a useful approximation of the founder virus ( es ) . Figure 2 shows an example of the dynamics of frequencies of SNP relative to the founder sequence over time , where each dot ( top ) or line ( bottom ) represents the frequency of a nucleotide different from the founder sequence . Interactive versions of this graph are available for the entire genome of all patients at hiv . tuebingen . mpg . de . 10 . 7554/eLife . 11282 . 007Figure 2 . The dynamics of SNP frequencies . The upper panels show single nucleotide polymorphism frequencies along p17 at three time points in patient p1 . The lower panel shows the trajectories of SNPs through time . Color corresponds to position in the sequence . Trajectory that reach high frequencies are shown with thicker and more opaque lines . Analogous data is available for all patients for most of the HIV-1 genome . EDI , estimated date of infection; SNP , single nucleotide polymorphism . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 007 To measure the rate at which the virus population accumulates mutations , we calculated the average distance of each sample from the approximate founder sequence in 300 bp windows . Regressing this distance against time yields the rate of divergence in different regions of the genome , see Figure 3A . As expected , some regions such as the variable loops in gp120 and nef evolve faster , while enzymes – protease ( PR ) , reverse transcriptase ( RT ) – and the rev response element ( RRE ) evolve more slowly . The rate of divergence varies by about a factor of 10 along the genome , but is consistent with typically about 1 . 5-fold differences across patients ( standard deviation of log2 ( fold change ) 0 . 6 ± 0 . 2 ) . The overall pattern of the rate of mutation accumulation agrees with a recent map of HIV genome-wide variation from a population perspective ( Li et al . , 2015 ) and correlates well with entropy in a large HIV-1 group M alignment ( Spearman’s p = 0 . 7 after the same smoothing ) . 10 . 7554/eLife . 11282 . 008Figure 3 . Consistent evolution across the viral genome . The figure shows the rate of sequence divergence averaged in a sliding window of length 300 bp for individual study participants ( in color ) and averaged over all ( black ) . Rapidly evolving ( V loops in gp120 ) and conserved ( RRE ) regions are readily apparent . The divergence rates are determined by linear regression of the distance from the putative founder sequence against time since EDI . This distance includes contributions of minor variants . All positions are given in HxB2 numbering . The corresponding figure for amino acid evolution is provided as Figure 3—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 00810 . 7554/eLife . 11282 . 009Figure 3—figure supplement 1 . Diversity over time for all six PCR fragments and all patients . Early samples of p3 and p10 show diversity at levels suggesting an infection with multiple founder viruses , while the remaining infections appear to be founded by a single variant . EDI , estimated date of infection . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 00910 . 7554/eLife . 11282 . 010Figure 3—figure supplement 2 . Divergence rate like Figure 3 but for amino acids rather than nucleotides . Divergence is estimated in a sliding window of 100 amino acids across the genome ( including tat but excluding rev and the LTRs ) . The resulting picture is similar to what we observe at the nucleotide level . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 010 Having found that coarse patterns of divergence are comparable among patients , we asked whether intrapatient diversity at individual sites in the viral genome follows general and predictable patterns . To this end , we compared diversity at each position to the diversity observed in HIV-1 group M ( see Materials and methods ) . Figure 4A shows the rank correlation between the site-by-site diversity in each patient and a global collection of HIV-1 sequences , both measured by Shannon entropy ( see Materials and methods ) . In all cases , correlation with cross-sectional diversity was initially low , as expected for largely homogeneous populations . As diversity increases within patients , it tends to accumulate at positions that are not conserved , resulting in a rank correlation of about 0 . 4–0 . 5 after about 8 years . These correlations are individually significant and reproducible among different genomic regions ( error bars in Figure 4 ) . 10 . 7554/eLife . 11282 . 011Figure 4 . Within patient variation mirrors global variation . ( A ) Intrapatient variation at individual sites is correlated with diversity at homologous positions in an alignment of sequences representative of HIV-1 group M . This correlation increases reproducibly throughout the infection . Error bars show standard deviations over genomic regions . ( B ) Similarly , the fraction of sites with minor variants above 1% increases over time at the least constrained positions ( quartiles Q3 and Q4 ) , while few sites in the most conserved quartiles ( Q1 and Q2 ) are polymorphic . Figure 4—figure supplements 1 and 2 show the corresponding results for amino acid rather than nucleotide comparisons and patient–patient correlations of diversity , respectively . EDI , estimated date of infection . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01110 . 7554/eLife . 11282 . 012Figure 4—source data 1 . Tab-delimited files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01210 . 7554/eLife . 11282 . 013Figure 4—figure supplement 1 . Within–patient variation mirrors global variation at the amino acid level , similar to what we observed at the nucleotide level . Reference alignments against NL4-3 were used instead of HxB2 to ensure good translation ( HxB2 has frameshift mutations ) . EDI , estimated date of infection; SNP , single nucleotide polymorphism . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01310 . 7554/eLife . 11282 . 014Figure 4—figure supplement 1—source data 1 . Tab-delimited files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01410 . 7554/eLife . 11282 . 015Figure 4—figure supplement 2 . Correlation in nucleotide variation between study participants . Each dot corresponds to one patient pair and reports the mean and standard deviation of diversity correlations over different genomic regions . No clear difference between comparisons within and across subtypes is observed , but the study included only one subtype C ( p6 ) and one recombinant 01_AE ( p1 ) infection . Only late samples were used ( days since EDI ≥ 1500 ) . In general , the correlation between patients is similar to the one to cross-sectional diversity . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 015 Figure 4B offers an alternative perspective on the exploration of sequence space by the HIV-1 populations . We classified nucleotide positions in the genome into four categories ranging from highly conserved positions to less conserved positions within group M ( Q1–Q4 ) using the same alignment as above . Next we asked what fraction of sites within these categories show intrapatient variation at a level of at least 1% . For the least conserved positions , this fraction increased steadily to about 20% after 8 years , while less than 1% of the most conserved sites shows variation above the 1% level . This latter fraction rapidly saturates and does not increase over time . Since variant amplification , sequencing and variant calling does not use any information on cross-sectional conservations , the absence of variation above 1% at conserved sites is further evidence that our amplification and sequencing pipeline does not generate spurious variation . Other thresholds yield similar results . Taken together , the observations in Figure 4 show that variation is not limited by the mutational input , but that HIV-1 populations accumulate diversity wherever mutations do not severely compromise virus replication . At the single nucleotide level , the spectrum of mutational possibilities is explored reproducibly and the level of within-host diversity is predicted by time since infection and cross-sectional diversity . Conserved positions are typically monomorphic even in deeply sequenced samples with high RNA template input . In addition to the reproducible patterns of minor diversity , virus evolution is characterized by adaptations that are specific to the host . Immune selection results in escape mutations that rapidly spread through the population ( Richman et al . , 2003; Walker and McMichael , 2012 ) . Such mutations tend to be nonsynonymous , i . e . , change the viral proteins , while evolution at synonymous sites is expected to be conservative . Nevertheless , synonymous mutations can be affected by ‘selective sweeps’ of linked nonsynonymous mutations ( Maynard Smith and Haigh , 1974 ) . To quantify the degree at which the evolutionary dynamics of HIV-1 is dominated by selective sweeps , we calculated divergence and diversity separately at nonsynonymous and synonymous sites in different parts of the HIV-1 genome . Figure 5A compares nonsynonymous ( solid lines ) and synonymous divergence ( dashed lines ) in different regions of the genome . In agreement with the results presented in Figure 3 , the observed rate of evolution at nonsynonymous sites differed substantially between genomic regions , with env being the fastest and pol the slowest . Divergence at synonymous sites , however , varied very little between different genomic regions indicating random accumulation of synonymous mutations ( rather than positive selection ) . 10 . 7554/eLife . 11282 . 016Figure 5 . Distinct patterns of evolution across mutation types and regions . ( A ) shows divergence at nonsynonymous ( solid ) and synonymous ( dashed ) positions over time for different genomic regions averaged over all patients , measured as average Hamming distance from founder . While synonymous divergence is very similar in different regions , nonsynonyous divergence varies . ( B ) shows diversity through time , measured as average pairwise Hamming distance . Regions with high nonsynonymous diversity ( and divergence ) tend to have low synonymous diversity . Error bars represent standard deviations of patient bootstrap replicates . ( C ) shows the anti-correlation between the rate of nonsynonymous divergence and synonymous diversity in 1kb windows across the genome ( color indicates position on the genome blue→green→yellow→red ) . ( D ) shows the site frequency spectrum of synonymous ( blue ) and nonsynonymous ( green ) SNPs . EDI , estimated date of infection; SNP , single nucleotide polymorphism . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01610 . 7554/eLife . 11282 . 017Figure 5—source data 1 . Tab-delimited files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 017 Figure 5B shows the corresponding plot for diversity , i . e . , the distances between sequences from the same sample . Diversity at nonsynonymous sites ( solid lines ) saturates after about 2–4 years , suggesting that nonsynonymous SNPs either stay at low frequency because they are deleterious or rapidly increase in frequency and fix without contributing much to diversity . Synonymous diversity increases steadily in the 5’ part of the genome ( structural and enzymes ) , while it saturates in the 3’ half of the genome after a few years – the exact opposite of nonsynonymous divergence . Indeed , we observe a strong anti-correlation between synonymous diversity and nonsynonymous divergence , which is further quantified in Figure 5C . This suggests that frequent non-synonymous substitutions limit synonymous diversity because they drive linked synonymous mutations to fixation or to extinction ( Maynard Smith and Haigh , 1974 ) . We will quantify linkage and recombination below , but the differences in diversity accumulation already suggest that linkage is restricted to short distances . The contrasting behavior of synonymous and nonsynonymous SNPs is also seen in the SNP frequency spectrum – the histogram of SNP abundance – shown in Figure 5D . While the spectra agree for frequencies below 20% , synonymous mutations are strongly underrepresented at higher frequencies ( Fisher’s exact test at frequency 0 . 5 , p-value <10-10 ) . This corroborates the interpretation that , due to substantial recombination , sweeping nonsynonymous mutations only occasionally ‘drag’ adjacent synonymous mutations to fixation . Synonymous mutations rarely rise in frequency because of their own effect on fitness , since they usually have small or deleterious phenotypic effects and do not contribute directly in immune evasion ( Zanini and Neher , 2013 ) . The about five-fold excess of nonsynonymous over synonymous SNPs at high frequencies ( see Figure 5D ) shows that the majority of common nonsynonymous mutations spread due to positive selection . Next , we sought to quantify what fraction of nonsynonymous divergence is driven by escape from cytotoxic T-lymphocytes ( CTLs ) . Four-digit HLA types were determined for all patients and a set of putatively targeted HIV-1 epitopes were determined using the epitope binding prediction tool MHCi ( tools . immuneepitope . org/mhci ) . We then asked whether we observed more nonsynonymous substitutions in epitopes predicted to be targeted than expected by chance ( excluding the variable loops of gp120 and the external part of gp41 , see Materials and methods ) . We found a significant enrichment by a factor 1 . 9 in the putatively targeted region ( p-value <3 × 10−6 ) , corresponding to 5 . 5 excess nonsynonymous substitutions , whereby the total number of nonsynonymous substitutions per patient is on average 43 ( median , quartiles 36–64 ) . The set of predicted epitopes will contain false positives and lack true epitopes , hence the actual number of CTL driven substitutions could be higher as for example suggested by Allen et al . ( 2005 ) who report that roughly half of non-envelope mutations are associated with CTL responses . Many CTL escape mutations reduce the replicative capacity of the virus and it is known that such escape mutations often revert upon transmission to a host in which the corresponding epitope is not targeted ( Friedrich et al . , 2004; Herbeck et al . , 2006; Leslie et al . , 2004 ) . The balance between escape and reversion results in association between specific escape mutations and the HLA types of the hosts ( Kawashima et al . , 2009; Palmer et al . , 2013 ) . In a diverse population of hosts , the most common state at a specific site is likely the preferred state , while rare alleles tend to be escape mutations that reduce viral replicative capacity ( Carlson et al . , 2014 ) . To quantify patterns of reversion and fitness cost , we classified sites in the approximate founder sequence of the viral populations in each subject as being identical or different from the HIV-1 group M consensus . Figure 6A shows the fraction of sites where the founder nucleotide is replaced by a mutant during the infection . This fraction is about ten-fold higher if the founder nucleotide differs from the group M consensus than if it is identical to the group M consensus . Reversion towards group M consensus occurs at a roughly constant rate throughout the observation time ( 5–8 years ) . 10 . 7554/eLife . 11282 . 018Figure 6 . Rapid reversion at conserved sites . ( A ) Sites where the founder sequence differed from the subtype or group M consensus ( upper curves ) diverged about tenfold more rapidly than sites that initially agreed with the consensus ( lower curves ) . ( B ) The rate of reversion increased with conservation ( lower variability ) , while divergence away from consensus showed the opposite behavior ( divergence is measured at 5−6 years ) . Error bars report the standard deviation of patient bootstraps . Figure 6—figure supplement 1 shows the corresponding figure for amino acids rather than nucleotides . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01810 . 7554/eLife . 11282 . 019Figure 6—source data 1 . Tab delimited files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 01910 . 7554/eLife . 11282 . 020Figure 6—figure supplement 1 . Patterns of reversion at the amino acid level are similar to those at the nucleotide level . Reference alignments against NL4-3 were used instead of HXB2 to ensure good translation ( HXB2 has frameshift mutations ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 02010 . 7554/eLife . 11282 . 021Figure 6—figure supplement 1—source data 1 . Tab-delimited files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 021 Of all changes accumulated by the viral populations , 30 ± 2 . 5% are reversions towards group M consensus ( mean and standard deviation of patient bootstraps after 4–7 years ) . Similar results are found when comparing with the subtype consensus of each patient virus ( 24 ± 2 . 5% ) . Reversions are between 4 and 5 times more frequent than expected in the absence of a reversion bias ( 7 . 8 and 4 . 5% , respectively ) . These findings agree with results by Allen et al . ( 2005 ) , who report that about 20% of amino acid substitutions are reversions . By focusing on sites where the founder virus differed from the group M consensus , we are predominantly looking at weakly conserved sites . To control for conservation , we carried out the same analysis after stratifying sites by overall level of conservation . Figure 6B shows the result of this analysis , focusing on samples after 5–6 years for the sake of clarity . We find that the rate of reversion is highest at the most conserved sites . Almost 50% of all non-consensus positions at highly conserved sites had reverted to consensus after about 5 years – an almost 1000-fold excess . Even at the least conserved sites , divergence towards group M consensus exceeded divergence away from group M consensus by a factor of 3 . These results suggest that the global HIV-1 group M consensus sequence represents an ‘optimal’ HIV-1 sequence , which acts as an attractor for the evolutionary dynamics within hosts . This attraction is strongest at conserved sites , but extends to the least conserved sites in the genome . To quantify the decoupling of SNPs by recombination , we calculated linkage disequilibrium ( LD ) between SNPs as a function of distance for each of the six fragments , see Figure 7 . For most fragments , we observed a consistent decrease of LD over the first 100–200 bps , with fragment 5 being an exception with linkage of mutations at longer distances . Importantly , our linkage control ( a 50/50 mix of two distinct virus isolates and a total of 1250 RNA molecules per PCR fragment ) shows no decay of LD with distance , suggesting negligible RT-PCR recombination . 10 . 7554/eLife . 11282 . 022Figure 7 . Linkage and recombination . Linkage disequilibrium decays rapidly with distance between SNPs . Colored lines correspond to the different fragments , each averaged over patients . The dashed line shows data from a control experiment for PCR recombination , where two cultured virus populations were mixed . No PCR recombination is observed . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 02210 . 7554/eLife . 11282 . 023Figure 7—source data 1 . Tab-deliminated files with plotted data . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 023 The observed decay of LD in patient samples is consistent with a recombination rate of 10−5/bp/day as estimated in ( Batorsky et al . , 2011; Neher and Leitner , 2010 ) . Our reasoning proceeds as follows . Figure 5B indicates that diversity accumulates over a time frame of 2–4 years , i . e . , about 1000 days . Recombination at a rate of 10−5/bp/day hits a genome on average every 100 bps in 1000 days . Mutations further apart than 100 bps are hence often separated by recombination and retain little linkage consistent with the observed decay length in Figure 7 . The longer linkage in fragment 5 ( env ) might have several reasons that extend beyond our simple argument: ( i ) homologous recombination might be suppressed in the most variable regions , ( ii ) the accuracy of SNP frequency estimates is lower in F5 due to poorer amplification , and ( iii ) the rapid evolution of env due frequent substitutions and sweeps gives less time to break up linkage . In particular , as shown in Figure 5C , frequent and strong selective sweeps affect synonymous diversity in physical proximity along the genome , confirming the presence of linkage at short distances . For phylogenetic analysis , we can extract haplotypes from the sequencing reads up to 500 bp in length . Only in the more diverse regions are 500 bp sufficient for well-resolved phylogenies ( see Figure 8 ) . However , we find that linkage does not extend beyond 100–200 bp . Hence the read length is not a limiting factor . Only during rapid population shifts such as drug resistance evolution , long read technologies such as PacBio would be necessary to capture the evolutionary dynamics ( Nijhuis et al . , 1998 ) . 10 . 7554/eLife . 11282 . 024Figure 8 . Phylogenetic trees of minor genetic variants . In rapidly evolving genomic regions , trees that include minor genetic variants ( haplotypes ) approximate the true phylogeny . Here p17 in gag and the variable loop 3 in env from patient p1 are compared; many more trees are available on the website . Trees are reconstructed using FastTree ( Price et al . , 2009 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 11282 . 024
We have presented a comprehensive portrait of intrapatient evolution of HIV-1 that covers almost the entire genome of the virus , characterizes minor genetic variants , and tracks the fate and dynamics of these variants over a follow-up period of up to 8 years in nine patients . We find that , during the infection , HIV-1 explores the sequence space surrounding the founder virus systematically; similar mutational patterns are observed within different , unrelated patients . Linkage between mutations is limited to approximately 100 bp , so the virus population can accumulate substitutions independently in different regions of the genome as suggested by theoretical models ( Mostowy et al . , 2011; Rouzine and Coffin , 2005 ) . Nonetheless , local dynamics of SNPs is often dominated by hitch-hiking between neighboring mutations , resulting in an anticorrelation between nonsynonymous divergence and synonymous diversity . A large fraction of all substitutions are reversions towards the global HIV-1 consensus sequence , and these reversions steadily accumulate throughout infection . The evolutionary dynamics of HIV-1 populations is the result of stochastic forces like mutation and frequent bottlenecks , deterministic fixation of favorable mutations , and recombination . The relative importance of these forces remains unclear ( Brown , 1997; Frost et al . , 2000; Kouyos et al . , 2006; Maldarelli et al . , 2013; Pennings et al . , 2014; Rouzine and Coffin , 1999 ) . Our observation that intrapatient diversity recapitulates diversity seen across HIV-1 group M and the strong tendency to revert towards consensus suggest that , in chronic infection , selection determines diversity . The reproducible exploration of sequence space can coexist with frequent adaptation only in frequently recombining large populations ( Neher et al . , 2013 ) . We observe that mutations further apart than 100 bp are effectively shuffled by recombination in most parts of genome , consistent with previous estimates of the HIV-1 recombination rate ( Batorsky et al . , 2011; Neher and Leitner , 2010 ) . Linkage and stochastic effects become stronger with increasing frequency of strength of selection , consistent with lower synonymous diversity and more LD in env . While rapid CTL escape at 5–10 sites over the first 2 years of infection has been documented in detail ( Allen et al . , 2005; Goonetilleke et al . , 2009; Herbeck et al . , 2006; Jones et al . , 2004; Liu et al . , 2011 ) and population level associations between specific HLA types and escape variants suggest widespread CTL escape ( Kawashima et al . , 2009 ) , the effect of escape and reversion on long-term evolutionary trends is less clear ( Lythgoe and Fraser , 2012; Roberts et al . , 2015 ) . We find a strong tendency for viral populations to revert towards the global HIV-1 consensus . At sites where the founder sequence differs from the subtype consensus , substitutions are almost five-fold overrepresented: Instead of ≈5% reversions expected based on the fraction of sequence at which the founder virus differs from consensus of the HIV-1 subtype , almost 25% of substitutions are reversions , in agreement with earlier reports on reversion of CTL escapes ( Allen et al . , 2005; Li et al . , 2007 ) . This tendency to revert increases with the level of conservation of the site , suggesting a quantitative relationship between fitness cost and conservation . While reversion is particularly prevalent in acute infection ( Li et al . , 2007 ) , we show that reversion is not limited to early infection but happens throughout chronic infection . The bias towards reversion results in a two- to three-fold reduction of the long-term evolutionary rate of HIV , a trend that is reinforced by selection during transmission ( Carlson et al . , 2014; Sagar et al . , 2009 ) . Inter-individual evolutionary rates of HIV-1 are two to six times lower than intra-individual rates , and a number of possible mechanisms have been suggested to explain this discrepancy ( Lythgoe and Fraser , 2012 ) . Our results strongly indicate that most of this mismatch can be explained by steady reversion during infection; other factors such as retrieval of ‘stored’ latent variants or stage-specific selection might also contribute to the rate mismatch ( Immonen and Leitner , 2014; Lythgoe and Fraser , 2012 ) . The high rate of reversion has implications for phylogenetic dating . Given the five-fold excess of reverting minor variation , reversion would balance divergence once the typical distance from the consensus sequence equals 17% , corresponding to a nucleotide diversity of about 30%; this is remarkably close to the actual divergence between HIV-1 groups M , N and O ( Li et al . , 2015 ) . On longer distances , this simple argument will have to be modified due to compensatory mutations resulting in gradual shift of the preferred state at some positions; nonetheless , it indicates a dramatic slowing down of divergence at a scale of the HIV-1-SIVcpz divergence . This apparent deceleration of evolution could explain the contradictory findings of attempts to date the age of HIV-1 and SIV ( Worobey et al . , 2010 ) . The strong and lasting preference for specific nucleotides needs to be accounted for in phylogenetic analysis , as has recently been shown using experimentally determined fitness landscapes of influenza virus proteins ( Bloom , 2014 ) . Our observations are consistent with results by Ashenberg et al . ( 2013 ) ; Doud et al . ( 2015 ) , who showed that amino acid preferences are mostly conserved between related influenza strains . Similarly , divergence between HIV-1 subtypes is small enough that epistatic interactions have not yet changed the majority of the preferred states . With increasing evolutionary distance between strains , the molecular context and preferred amino acid at a particular position is more and more likely to differ ( Pollock et al . , 2012 ) . Nevertheless , a subset of amino acids preferences are conserved almost universally ( Risso et al . , 2015 ) and enable sequence based homology search . The concordance between intrahost variation and patterns of conservation across HIV-1 group M hints at universal fitness costs of mutations . Recently , cross-sectional conservation has been used as a proxy for fitness costs in models of HIV-1 fitness landscapes ( Dahirel et al . , 2011; Ferguson et al . , 2013; Mann et al . , 2014 ) . Since reproducible intrapatient diversity likely reflects fitness costs of mutations in vivo , our results provide a direct justification for this approach . However , nine patients are insufficient to extend this analysis to fitness interactions between mutations . One limitation of this study was the availability of samples from patients with sufficiently long follow-up without therapy after a well-defined time of infection . The majority of patients were MSM infected with subtype B virus . Thus , we cannot exclude that the aspects of HIV-1 evolution that we have investigated may differ between transmission routes or HIV-1 subtypes . While substitution and recombination errors of our optimized protocol for HIV-1 RNA extraction and an RT-PCR are low , the other main limitation was that the number of available template molecules was small in some samples ( see also Iyer et al . , 2015 ) . In principle , the Primer ID method , which labels and resequences each individual template , allows quantification of templates and almost complete elimination of experimental substitution and recombination errors ( Jabara et al . , 2011 ) . However , we are not aware of a Primer ID protocol for genome-wide sequencing , which was essential to our study . The analysis and means of data sharing of next-generation sequencing data from viral populations is still in its infancy . Raw reads require substantial post-processing before the data can be used to answer biological questions and technology is changing rapidly such that no standardized pipelines exist . To facilitate the exploration and further analysis of our data , we have developed a web server that allows to browse characteristics of individual patients , e . g . , graph the frequencies of single nucleotide polymorphism through time , and download different aspects of the data in convenient format . It is , for example , possible to select an arbitrary region of the genome with length <500 bp , such as the frequently investigated V3 region , and extract an alignment of haplotypes covering this region along with their frequency from all time points belonging to one patient . We hope that the convenient access to processed data will facilitate follow-up studies on other aspects of viral evolution . HIV-1 and other microbial populations evolve in a constant struggle between adaptation to a changing environment and maintenance of functionality . Large mutation rates and population sizes generate standing genetic diversity that is limited by the fitness costs rather than mutation rates . Hence the limiting factor for adaptation is not generating the useful mutations , but combining multiple mutations necessary to survive – e . g . escape mutations and reversions – and pruning deleterious mutations . In HIV-1 , this process is facilitated by frequent recombination . We expect that the systematic exploration of sequence space , the reproducible patterns of minor variation , and frequent reversion will be characteristic of other RNA viruses . Properties of linkage between mutation will differ since mechanisms of recombination are diverse . But even though selective forces , recombination , and time scales will vary among different microbial populations , theoretical models of rapid adaptation population have shown that many features of the evolutionary dynamics are independent of the specific system ( Fisher , 2013; Neher , 2013 ) . Intrapatient evolution of HIV-1 is a unique opportunity to study this evolutionary dynamics directly in vivo .
The inclusion criteria for the study patients were: a ) a relatively well-defined time of infection based on a negative HIV antibody test less than 2 years before a first positive test or a laboratory documented primary HIV infection; b ) no antiretroviral therapy ( ART ) for a minimum of approximately 5 years following diagnosis; and c ) availability of biobank plasma samples covering this time period . The patients were selected from the Venhälsan HIV clinic in Stockholm , Sweden and were diagnosed between 1990 and 2003 . Seven to twelve plasma samples per patient ( 200–1000 μl ) were retrieved from the biobanks of the Karolinska University Hospital and the Public Health Agency of Sweden . These samples had been stored at -70°C following routine HIV RNA quantification . Information about the patients and the samples are summarized in Supplementary file 1 . Results from routine HIV antibody tests , HIV antigen tests , plasma HIV RNA levels and CD4 counts were collected from the patient records . Four-digit HLA typing of HLA class I loci A , B , C and class II loci DR and DQ was performed at the laboratory of immunpathology in Tübingen University Hospital . In the control experiments , we used HIV DNA from the following plasmids; NL4-3 ( subtype B , DNA concentration 110 ng/μl corresponding to 1 . 35 × 1010 copies/μl ) , SF162 ( subtype B , DNA concentration 117 ng/μl corresponding to 1 . 43 × 1010 copies/μl ) , pZM246F_10 ( subtype C , DNA concentration 101 ng/μl corresponding to 1 . 35 × 109 copies/μl ) . We also used HIV RNA from the following virus isolates LAI III ( subtype B , 7500 copies/μl ) , 38540 ( subtype C , 225 , 000 copies/μl ) and 38304 ( subtype B , 45 , 000 copies/μl ) . The patients were classified according to the Fiebig staging system for primary and early HIV infection ( Fiebig et al . , 2003 ) . In addition , we performed BED tests ( Aware BED EIA HIV-1 Incidence Test , Calypte Biomedical Corporation , Portland , OR ) . For each patient , the date of infection was estimated using results from laboratory tests according to the following hierarchical scheme . Comprehensive information about the data used for EDI determinations is provided in Supplementary file 3 . Primers were designed to cover almost the full HIV genome in six overlapping fragments , called fragments F1–F6 as illustrated in Figure 1 . This allowed sequencing of nucleotide positions 571–9567 in the HxB2 reference sequence according to the Sequence Locator Tool available at www . hiv . lanl . gov . Because of the redundancy of the long terminal repeats ( LTRs ) , this means that all genomic regions except positions 482-571 in the R region of the LTRs were sequenced . Primer design was performed using the subtype reference alignment and the PrimerDesign software available at www . hiv . lanl . gov ( Brodin et al . , 2013 ) . PrimerDesign was used to find candidate forward and reverse primers targeting highly conserved regions of the HIV genome , with similar melting temperatures , and with minimal tendency for hairpin and primer-dimer formation . Candidate primers were manually adjusted if needed , tested and sometimes redesigned . For each genome fragment , both outer PCR primers and nested , inner primers were designed; inner primers were only used for template quantification and internal testing purposes . Alternative primer sets were created for genome fragments F3 and F5 because the PCRs with the original primers sometimes were inefficient . For fragment F5 , the amplification problem was not completely alleviated despite trials with several different primer pairs . We believe that this might be be due to the extensive secondary structure in the RRE region . The primers are presented in Supplementary file 2 . All primer positions except the 5’ primer of F1 and the 3’ primer of F6 are contained in neighboring amplicons and hence sequenced . The primer part of the reads itself was trimmed after sequencing ( see below ) . For each sample , 400 μl of plasma ( if available ) was divided into two 200 μl aliquots . Total RNA was extracted using RNeasy Lipid Tissue Mini Kit ( Qiagen Cat . No . 74804 ) . Each aliquot was eluted twice with 50 μl RNase free water to maximize HIV RNA recovery . The four eluates were pooled , giving a total volume of 200 μl of RNA per sample . The RNA was divided into twelve 14 μl aliquots for duplicate one-step RT-PCR with the outer primers for fragments F1 to F6 and Superscript III One-Step RT-PCR with Platinum Taq High Fidelity Enzyme Mix ( Invitrogen , Carlsbad , California , US ) . Remaining RNA was used for template quantification ( see below ) . The one-step RT-PCR was started with cDNA synthesis at 50°C for 30 min and denaturation step at 94°C for 2 min followed by 30 PCR cycles of denaturation at 94°C for 15 s , annealing at 50°C for 30 s and extension at 68°C for 90 s and a final extension step at 68°C for 5 min . A second nested PCR was used for template quantification and in some of the control experiments . For the second PCR , 2 . 5 μl of the product from the first PCR was amplified with Platinum Taq High Fidelity . The second PCR consisted of a denaturation step at 94°C for 2 min , followed by 30 PCR cycles with denaturation at 94°C for 15 s , annealing at 50°C for 20 s and extension at 72°C for 90 s and a final extension at 72°C for 6 min . Other PCR conditions were also tried during assay development . After PCR , the duplicate amplicons from each of the six overlapping PCRs were pooled and purified with Illustra GFX PCR DNA and Gel Band Purification Kit ( 28-9034-70 , VWR ) or AGENCOURT AMPure XP PCR purification kit ( A63881 , Beckman Coulter AB ) . Purified amplicons from each sample were quantified with Qubit assays ( Q32851 , Life Technologies ) and thereafter diluted and pooled in equimolar concentrations . The Illumina Nextera XT library preparation protocol and kit were used to produce DNA libraries . The original protocol was optimized for longer reads and amplicon input in the following fashion: ( i ) input DNA concentration to tagmentation was increased to 0 . 3 ng/μl to reduce overtagmentation; ( ii ) the number of post-tagmentation PCR cycles was raised up to 14 for samples with very low input DNA; ( iii ) post-PCR purification was done using Qiagen Qiaquick columns to maximize large-inserts throughput as compared to magnetic-bead based protocols; ( iv ) sze selection was performed using the SageScience BluePippin system with 1 . 5% agarose gel cassettes and internal marker R2 , selecting sizes of 550–900 bp ( including dual index Nextera adapters , final insert sizes 400–700 bp ) ; ( v ) size-selected eluates were pooled , buffer-exchanged into EB ( 10 mM Tris-HCl ) , and reconcentrated to at least 2 nM . The Illumina MiSeq instrument with 2 × 2 bp or 2 × 2 bp sequencing kits ( MS-102-2003/MS-102-3003 ) was used to sequence the DNA libraries . We performed 26 paired-end sequencing runs . Overall , we obtained around 200 Gbases of output , i . e . , 300 Mbases for each PCR amplicon . The median number of reads per amplicon was 80 , 000 ( quartiles 20 , 000–220 , 000 , max 2 millions ) . All read files have been uploaded to ENA with study accession number PRJEB9618 . Bases were called from the raw images using Casava 1 . 8 . The reads were analyzed from that point on using a custom pipeline written in Python 2 . 7 and C++ . We favored this pipeline over existing programs because HIV-1 is a diverse species and both coverage and genetic diversity typically fluctuate by many orders of magnitude across the genome . The pipeline works as follows: ( i ) reads were mapped onto the HIV-1 reference HxB2 , using the probabilistic mapper Stampy ( Lunter and Goodson , 2011 ) ; ( ii ) mapped reads were classified into one of the six overlapping fragments used for RT-PCR ( ambiguous and chimeric reads were discarded ) , and trimmed for PHRED quality above or equal 30 except for one isolated position per read; RT-PCR primers were also trimmed at this step; ( iii ) a consensus sequence was computed for each fragment in each sample from a subset of the reads , using a chain of overlapping local multiple sequence alignments ( each covering around 150 bp ) ; ( iv ) the reads were re-mapped , this time against their own consensus; ( vii ) genetic distance from the consensus were computed , and reads with a distance higher than a sample- and fragment-specific threshold were discarded . Each threshold , calculated to exclude even traces of cross-contamination that might have happened during RNA extraction , PCR amplification , or library preparation , was established by plotting the distribution of Hamming distances of reads from the sample consensus , and excluding reads that are further away than the tail of the main peak . Contaminations appeared as a second peak at higher distances , recombinants as a fat tail: both were excluded . Reads were also trimmed for mapping errors at the edges ( small indels ) ; ( viii ) filtered reads were mapped a third time against a patient-specific reference that was as similar as possible to the consensus sequence from the earliest time point; ( ix ) reads were re-filtered and checked again for cross-contamination . The tree of minor variants of all patients in Figure 1—figure supplement 1 shows the absence of cross-contamination . The pipeline was equipped with extensive consistency checks for base quality , mapping errors , and contamination , and is based on the open source projects numpy ( van der Walt et al . , 2011 ) , matplotlib ( Hunter , 2007 ) , Biopython ( Cock et al . , 2009 ) , samtools ( Li et al . , 2009 ) , pandas ( McKinney , 2011 ) , and SeqAn ( Döring et al . , 2008 ) . The paired-end reads obtained correspond to inserts of up to 700bp in length and therefore provide information on linkage of mutations up to that distance . However , cDNA synthesis and PCR have both the potential to generate in vitro recombination , and true biological linkage is preserved only if the frequency of in vitro recombination is low . To estimate the in vitro recombination in our experimental setup , virions from two subtype B HIV isolates , LAI III and 38304 ( the latter obtained from a Swedish HIV-1 patient infected in Brazil ) , were mixed in equal concentrations . Aliquots of this mix ( approximately 1250 RNA molecules per PCR fragment ) were amplified with the six overlapping one-step RT-PCRs as described above in both single and nested PCR mode and then sequenced . PCR recombination is known to occur predominantly at high amplicon concentrations due to heteroduplex formation of incompletely extended molecules ( Di Giallonardo et al . , 2013; Mild et al . , 2011 ) . Consistently , we observe no PCR recombination within the first PCR ( which starting at low template input does not saturate ) , while we observe substantial PCR recombination during a second nested PCR . Since our library preparation protocol requires very low input DNA , we do not need nested PCR , avoiding this source of PCR recombination . The fact that PCR recombination occurs during a second PCR shows that the two viruses used for this control ( both subtype B ) are similar enough that divergence does not interfere with heteroduplex formation . Figure 7 includes the linkage disequilibrium observed in the control experiment as a function of distance , showing that linkage is high and not lost with distance during the first PCR . Python scripts that generate each figure shown in the manuscript are available at github . com/neherlab/HIVEVO_figures . In these scripts , all parameters , settings , and calculations are explicitly documented . The website was realized using the Flask web framework , Bootstrap from Twitter as a frontend engine and jQuery and D3 . js as a JavaScript library for interactive plots ( Bostock , 2015; Ronacher , 2015 ) . The webpage is available at hiv . tuebingen . mpg . de . | HIV was transmitted from apes to humans multiple times before the virus began to spread among humans some time in the early 20th century . The virus is now found around the globe and has evolved into several different subtypes . Also , HIV continues to evolve inside each infected individual so that different variations of the virus are often present in the same person at any given time . When a person initially becomes infected , immune cells called killer T-cells seek out and destroy infected cells . However , HIV quickly acquires genetic mutations that allow it to escape the immune system and multiply in the body . There is a cost to this evasion strategy: the mutant forms of the virus often don’t multiply as rapidly as the variant that first infected the individual . This may explain why the virus often mutates back to its previous form in newly infected people . But many questions remain about how these pressures influence the evolution of the virus in individuals and in populations over time . Now , Zanini et al . have sequenced the genomes of HIV samples collected from nine people with HIV in Sweden over the course of five to eight years . These individuals were diagnosed between 1990 and 2003 – when HIV medications were not as widely used as they are now – and did not receive any treatment during the course of the study . Therefore , the study provides a rare opportunity to look at how HIV evolves in the absence of drugs that target the virus . Zanini et al . sequenced the entire genetic code of each form of the virus identified in the samples . The data show that HIV follows predictable patterns of evolution within individuals as well as across human populations . Mutations happen frequently all over the HIV genome . However , the mutant viruses often revert to a common or “optimal” form of the virus throughout the course of infection . This suggests that there is a tradeoff between the benefits of acquiring new mutations and maintaining a set of traits that have enabled the virus to spread so successfully in humans . To make it easier for other researchers to explore the data , Zanini et al . created a web application that allows others to access and create visual representations of viral evolution . Together these findings suggest that it will be possible to achieve a fuller understanding of RNA virus evolution that integrates the molecular biology of the virus and the immune response of the host with the evolutionary changes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"genetics",
"and",
"genomics"
] | 2015 | Population genomics of intrapatient HIV-1 evolution |
The RNA-binding proteins PTBP1 and PTBP2 control programs of alternative splicing during neuronal development . PTBP2 was found to maintain embryonic splicing patterns of many synaptic and cytoskeletal proteins during differentiation of neuronal progenitor cells ( NPCs ) into early neurons . However , the role of the earlier PTBP1 program in embryonic stem cells ( ESCs ) and NPCs was not clear . We show that PTBP1 controls a program of neuronal gene expression that includes the transcription factor Pbx1 . We identify exons specifically regulated by PTBP1 and not PTBP2 as mouse ESCs differentiate into NPCs . We find that PTBP1 represses Pbx1 exon 7 and the expression of the neuronal Pbx1a isoform in ESCs . Using CRISPR-Cas9 to delete regulatory elements for exon 7 , we induce Pbx1a expression in ESCs , finding that this activates transcription of neuronal genes . Thus , PTBP1 controls the activity of Pbx1 to suppress its neuronal transcriptional program prior to induction of NPC development .
Alternative splicing is an important form of gene regulation during tissue development . In the mammalian nervous system , large scale changes in splice site choice produce many new mRNAs encoding protein isoforms with different structures and functions that are specific to neurons ( Li et al . , 2007; Licatalosi and Darnell , 2006; Norris and Calarco , 2012a; Raj and Blencowe , 2015a; Yap and Makeyev , 2013; Zheng and Black , 2013 ) . These splicing patterns are regulated in a temporal and cell-specific manner by the expression of specialized pre-mRNA binding proteins ( RBPs ) ( Black , 2003; Braunschweig et al . , 2013; Fu and Ares , 2014; Lee and Rio , 2015 ) . Several RBPs have been shown to control the expression of isoforms essential for neuronal development and maturation ( Charizanis et al . , 2012; Gehman et al . , 2011; Gehman et al . , 2012; Li et al . , 2014; Licatalosi et al . , 2012; Yano et al . , 2010; Jensen et al . , 2000a; Iijima et al . , 2011a; Ince-Dunn et al . , 2012b; Quesnel-Vallières et al . , 2015b ) . However , the complex programs of splicing affecting neuronal development are only beginning to be characterized , and the cellular functions of the regulated isoforms are often poorly understood . The polypyrimidine tract binding ( PTB ) proteins , PTBP1 and PTBP2 , regulate a large set of splicing events during neuronal differentiation ( Keppetipola et al . , 2012 ) . PTBP1 is expressed in neuronal progenitor cells and many non-neuronal cells . Upon neuronal differentiation , PTBP1 expression is repressed allowing the induction of PTBP2 and the initiation of a neuronal splicing program that is essential for neuronal maturation and survival ( Boutz et al . , 2007; Li et al . , 2014; Licatalosi et al . , 2012; Makeyev et al . , 2007; Tang et al . , 2011; Zheng et al . , 2012; Gueroussov et al . , 2015c ) . In addition to splicing , the PTB proteins affect other aspects of posttranscriptional regulation in neurons , including miRNA targeting ( Xue et al . , 2013; Yap et al . , 2012 ) . It has been shown that depletion of PTBP1 from fibroblasts is sufficient to drive cells towards a neuronal phenotype ( Xue et al . , 2013 ) . As neurons mature , PTBP2 expression is eventually reduced , giving rise to an adult neuronal splicing program ( Li et al . , 2014; Licatalosi et al . , 2012; Tang et al . , 2011; Zheng et al . , 2012 ) . Together , these transitions in PTBP1 and PTBP2 expression define three phases of neuronal maturation each with different splicing programs . PTBP1 and PTBP2 are encoded on separate genes and have very similar sequences and RNA-binding properties ( Markovtsov et al . , 2000; Oberstrass et al . , 2005 ) . These proteins target overlapping but not identical sets of splicing events ( Boutz et al . , 2007; Li et al . , 2014; Llorian et al . , 2010; Tang et al . , 2011; Zheng et al . , 2012 ) . Some exons are responsive to both proteins ( Boutz et al . , 2007; Li et al . , 2014; Spellman et al . , 2007; Zheng et al . , 2012 ) . In these cases , PTBP1 and PTBP2 serve to repress adult splicing patterns during neuronal maturation , yielding isoforms that are expressed only after PTBP2 levels have declined . In other cases , exons are more sensitive to PTBP1 and shift their splicing when PTBP1 is shut off early in neuronal differentiation even though PTBP2 is present ( Boutz et al . , 2007; Makeyev et al . , 2007; Markovtsov et al . , 2000; Tang et al . , 2011 ) . PTBP1 is known to regulate splicing in a wide variety of cell-types , and much of the work on PTBP1 targeting has used non-neuronal cell lines such as HeLa cells ( Llorian et al . , 2010; Spellman et al . , 2007; Xue et al . , 2009; Xue et al . , 2013 ) . The role of the PTBP1-specific splicing program in early progression along the lineage to neuronal progenitor cells has not been examined . Directed differentiation of embryonic stem cells ( ESCs ) provides a versatile model for the study of neuronal commitment , differentiation , and maturation . These systems have illuminated a wide range of genetic contributors to neuronal phenotype including epigenetic modifiers , transcription factors , miRNAs , and signaling molecules ( Temple , 2001; Louvi and Artavanis-Tsakonas , 2006a; Kosik , 2006b; Hirabayashi and Gotoh , 2010; Ronan et al . , 2013a ) . Here we use mouse ESC culture to characterize the PTBP1 splicing program during early neuronal differentiation . We identify a diverse set of splicing events regulated by PTBP1 , including an alternative exon in the homeodomain transcription factor Pbx1 , whose switch in splicing induces a neuronal transcriptional program early in neuronal development .
To examine PTBP1 and PTBP2 expression in anin vitro model of neuronal development , we differentiated mouse embryonic stem cells ( ESCs ) into motor neurons ( MNs ) . Mouse ESCs ( Day -2 ) were grown in aggregate culture for two days to form embryoid bodies ( EBs; Day 0 ) , which were then treated with retinoic acid ( RA ) and a Sonic hedgehog ( Shh ) pathway agonist for 5 days to induce MN formation ( Wichterle et al . , 2002; Adams et al . , 2015d ) . To facilitate MN identification and isolation , we used a mouse ESC line that expresses eGFP under the control of the MN specific HB9 promoter ( Wichterle et al . , 2002 ) . At Day 5 , the GFP+ embryoid bodies were dissociated and MNs plated onto Matrigel in the presence of neurotrophic factors to permit further MN maturation ( Figure 1A ) . 10 . 7554/eLife . 09268 . 003Figure 1 . The transition in PTB protein expression occurs during in vitro neuronal differentiation . ( A ) HB9-GFP ESCs express eGFP under the MN-specific promoter HB9 . The addition of retinoic acid ( RA ) and a Sonic hedgehog ( Shh ) agonist drives MN differentiation . ( B ) Western blot shows the loss of PTBP1 protein and gain of PTBP2 and GFP proteins as ESCs differentiate into HB9+ MNs . U170K served as a loading control . ( C ) Quantification of relative PTBP1 and PTBP2 protein expression across MN differentiation . Error bars represent standard error of the mean ( SEM , n=2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00310 . 7554/eLife . 09268 . 004Figure 1—source data 1 . Splicing changes identified by RNA-seq during ESC neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00410 . 7554/eLife . 09268 . 005Figure 1—figure supplement 1 . Transitions in PTB protein expression as ESCs differentiate into MNs , NPCs , and neurons . ( A ) Day 2 MN cultures express Nestin . ( B ) Day 5 MN cultures co-express GFP and HB9 . ( C ) Day 8 MN cultures co-express GFP and PTBP2 . ( D ) Western blot of PTBP1 and PTBP2 protein expression in ESCs , sorted Day 5 GFP+ MNs , and sorted Day 8 GFP+ MNs . GAPDH served as a loading control . ( E ) NPC cultures homogenously express the progenitor marker Nestin , ( F ) while very few cells express the neuronal marker TuJ1 . ( G ) Western blot and ( H ) quantification of PTBP1 and PTBP2 as ESCs differentiate into NPCs , Day 4 RA-derived neurons , and Day 6 MN cultures . U170K served as a loading control . Borders of western blots indicate different lanes within the same gel . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00510 . 7554/eLife . 09268 . 006Figure 1—figure supplement 2 . Transitions in alternative splicing occur during neuronal differentiation . ( A ) Alternative splicing events in ESCs , NPCs , and GFP+ MNs were quantified using SpliceTrap ( Wu et al . , 2011 ) . Bar graph and ( B ) heat map of cassette exons , 5’ and 3’ splice sites , and retained introns whose splicing differs between each sample pair ( ΔPSI ≥ 15% ) ( Figure 1—source data 1 ) . ( C , D ) Gene ontology analyses of genes showing the differential splice events compared to all expressed genes were performed using DAVID ( FDR<0 . 05 ) ( Huang et al . , 2009a ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 006 Similar to observations in the brain , as ESCs differentiate into MNs , PTBP1 expression declines , while PTBP2 expression is induced ( Figure 1B and C ) ( Boutz et al . , 2007; Tang et al . , 2011; Zheng et al . , 2012 ) . By immunofluorescence , Day 2 MNs express the progenitor marker Nestin , whereas Day 5 GFP+ cells express the postmitotic MN marker HB9 and the neuronal marker TuJ1 ( Figure 1—figure supplement 1A and B , and data not shown ) . The presence of PTBP2 in GFP+ cells was also seen by immunofluorescence ( Figure 1—figure supplement 1C ) . To monitor PTBP1 and PTBP2 protein specifically in the differentiated neurons , we used fluorescence-activated cell sorting ( FACS ) to isolate GFP+ cells at Days 5 and 8 . By western blot , these GFP positive cells express low amounts of PTBP1 and high levels of PTBP2 at Day 5 . At Day 8 only trace amounts of PTBP1 are found , while PTBP2 expression remains high ( Figure 1—figure supplement 1D ) . We did not observe the late decline in PTBP2 expression that occurs during neuronal maturationin vivo . This decline in PTBP2 expression may take place with longer culture times , or cells may not fully mature in this system . These results show that the switch from PTBP1 to PTBP2 expression occurs during MN differentiation , similar to earlier observations in the developing cortex . We previously observed high levels of PTBP1 and low levels of PTBP2 in neuronal progenitor cells ( NPCs ) ( Boutz et al . , 2007; Li et al . , 2014; Licatalosi et al . , 2012 ) . Nestin+ NPCs are observed early in EB cultures , but these cells could not be isolated in pure populations ( Figure 1—figure supplement 1A ) . To examine NPCs in more detail , ESCs were grown in N2B27 media and expanded in the presence of EGF and FGF , to yield cultures where greater than 80% of the cells were Nestin+ and with very few TuJ1+ neurons ( Conti et al . , 2005; Ying et al . , 2003 ) ( Figure 1—figure supplement 1E and F ) . We examined both the Hb9-GFP line and the 46C ESC line used in the original monolayer NPC protocol , finding minimal differences in marker staining ( Ying et al . , 2003 ) . By western blot , we found that both PTBP1 and PTBP2 proteins were expressed in these NPC cultures . As these NPCs were differentiated into neurons with retinoic acid , PTBP1 protein expression was lost while PTBP2 protein was induced ( Figure 1—figure supplement 1G and H ) . The EB and monolayer culture systems likely give rise to different neuronal populations with differences in gene expression attributable to different neuronal types ( Gaspard et al . , 2009; Wichterle et al . , 2002; Lee et al . , 2000b; Murashov et al . , 2005a; Su et al . , 2006c; Salero and Hatten , 2007 ) . Nevertheless , the transition in PTBP1-PTBP2 protein expression is common to both models of in vitro neuronal development , as well as embryonic brain . We next examined alternative splicing patterns under different states of PTBP1 and PTBP2 expression . Poly-A plus RNA was isolated from undifferentiated ESCs ( high PTBP1 and low PTBP2 ) , Nestin+ progenitor cells ( intermediate PTBP1 and PTBP2 ) , and post-mitotic HB9-GFP+ MNs ( low PTBP1 and high PTBP2 ) . Strand specific libraries for each cell type were subjected to 100 nt paired-end sequencing to generate 200 to 250 million mapped reads per sample . Alternative splicing events were quantified using SpliceTrap to identify cassette exons , 5’ and 3’ splice sites , and retained introns that differ between all sample pairs ( Wu et al . , 2011 ) . PSI values for the HB9 and 46C cell lines were highly correlated ( R2 = 0 . 92 and 0 . 90 in ESCs and NPCs respectively ) , indicating that the two cell lines are similar in splicing phenotype ( data not shown ) . Differentially spliced events were filtered for p-value below 0 . 05 and ranked by the change in percent spliced in value ( ΔPSI ) . Using a ΔPSI cutoff of 15% , we identified a large set of alternative splicing events that change across neuronal differentiation ( Figure 1—figure supplement 2A and B ) . With these filters 1201 differential splicing events were identified between ESCs and MNs , 1218 differential events between NPCs and MNs , and 750 differential events between ESC and NPC ( Figure 1—source data 1 ) . To identify enriched functional categories , transcripts exhibiting differential splicing were compared to the total expressed transcript set by Gene Ontology analyses ( Huang et al . , 2009a ) . Differential splicing events between NPCs and MNs were enriched ( FDR<0 . 05 ) in biological processes and cellular compartments relevant to neuronal differentiation such as cell projection organization , axonogenesis , and cell morphogenesis ( Figure 1—figure supplement 1C and D ) . It is expected that a subset of these splicing changes are the result of changes in PTBP1 and PTBP2 expression . To identify neuronal splicing events controlled by the PTB proteins and in particular PTBP1 , RNAi knockdowns were performed with two independent sets of control , PTBP1 , and PTBP2 siRNAs . PTBP1 siRNA treatment depleted the protein by 80% ( n=3 ) and 85% ( n=2 ) in ESCs and NPCs respectively ( Figure 2A and B ) . As seen previously , PTBP1 depletion induced PTBP2 protein expression by four-fold ( Boutz et al . , 2007; Makeyev et al . , 2007; Spellman et al . , 2007 ) ( Figure 2B ) . This change in PTBP2 expression is comparable to the 3 and 13-fold increases in PTBP2 protein seen when ESCs differentiate into NPCs and Day 4 RA-derived neurons ( Figure 1—figure supplement 2G ) . To assess the effect of PTBP2 , we performed PTBP1/2 double knockdowns to remove 84% and 93% PTBP2 in ESCs and NPCs respectively ( Figure 2A and B ) . 50 nt paired-end RNA sequencing was performed on the control and knockdown samples , and the splicing was analyzed with SpliceTrap . PSI values between the siRNA pairs showed a strong correlation ( R2 = 0 . 97 for each pair , data not shown ) , and the samples were pooled to increase read depth . 10 . 7554/eLife . 09268 . 007Figure 2 . PTBP1 regulates a large set of neuronal exons in ESCs and NPCs . ( A , B ) Left: Western blots of ESCs and NPCs treated with siControl , siPtbp1 , or both siPtbp1 and siPtbp2 . Similar results were obtained with an independent set of siRNAs . Right: Bar graphs showing the relative PTBP1 and PTBP2 protein expression ± SEM following siRNA treatment in ESCs ( n=3 ) and NPCs ( n=2 ) . Cells with the highest level of either protein are normalized to 1 . ( C ) Scatter plots compare the splicing changes in individual exons resulting from PTB protein depletion in ESCs ( X-axis ) with splicing changes between ESCs and MNs ( Y-axis ) ( Figure 2—source data 2 ) . ( D ) Scatter plots compare the splicing changes resulting from PTB protein depletion in NPCs ( X-axis ) with splicing changes between NPCs and MNs ( Y-axis ) ( Figure 2—source data 4 ) . ( C , D ) Data points in red correspond to neuronally spliced exons that are PTB protein repressed ( ΔPSI ≥ 15% ) . Data points in blue correspond to neuronally skipped exons that are PTB protein activated ( ΔPSI ≤ -15% ) . Data points in grey correspond to the other cassette exons measured by SpliceTrap . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00710 . 7554/eLife . 09268 . 008Figure 2—source data 1 . Splicing changes identified by RNA-seq following PTB protein depletion in ESCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00810 . 7554/eLife . 09268 . 009Figure 2—source data 2 . Neuronal cassette exons regulated by the PTB proteins in ESCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 00910 . 7554/eLife . 09268 . 010Figure 2—source data 3 . Splicing changes identified by RNA-seq following PTB protein depletion in NPCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01010 . 7554/eLife . 09268 . 011Figure 2—source data 4 . Neuronal cassette exons regulated by the PTB proteins in NPCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01110 . 7554/eLife . 09268 . 012Figure 2—source data 5 . Cassette exons co-regulated by PTBP1 and PTBP2 in NPCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01210 . 7554/eLife . 09268 . 013Figure 2—figure supplement 1 . The PTB proteins regulate a large set of exons in ESCs and NPCs . ( A ) PTBP2 protein is depleted from NPCs using two independent siRNAs . ( B ) Bar chart and ( C , D ) heat maps of differentially spliced events between Control samples and PTB protein knockdown samples in ESCs and NPCs ( ΔPSI ≥ 15% ) ( Figure 2—source data 1 and 3 ) . ( E , F ) Scatter plots compare the splicing changes from PTBP1 depletion with splicing changes from PTBP1 and PTBP2 dual depletion in ( C ) ESCs ( D ) and NPCs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 013 Using the P-value and PSI cutoffs previously described , we identified 418 splicing events altered by PTBP1 knockdown in ESCs , including 362 cassette exons , 50 alternative 5’ or 3’ splice sites , and 6 retained introns ( Figure 2—figure supplement 1B and C , Figure 2—source data 1 ) . This number was only slightly increased by double PTBP1/2 depletion , with similar changes in PSI when compared with the single knockdown ( Figure 2—figure supplement 1C ) . The splicing changes between the single and double knockdowns showed a strong correlation ( R2 = 0 . 91 , Figure 2—figure supplement 1E ) , indicating that these events are more strongly affected by PTBP1 . A majority of the exons altered by PTB protein depletion also change in the same direction during ESC to MN differentiation ( exons in the upper right and lower left quadrants of the plots in Figure 2C ) . Applying a ΔPSI cutoff of 15% for both data sets , 30% of neuronal exons are regulated by PTBP1 , including 97 neuronally activated exons that are PTBP1 repressed , and 56 neuronally skipped exons that are PTBP1 activated ( indicated by red and blue data points respectively in Figure 2C , left panel; Figure 2—source data 2 ) . 31% of neuronal exons were PTBP1/2 repressed ( 101 ) and activated ( 55 ) following PTBP1/2 double knockdown ( Figure 2C , right panel ) . As expected , many exons that change during neuronal differentiation are not affected by PTB protein depletion , indicating likely regulation by other proteins ( grey data points along the y-axes of the plots in Figure 2C ) . PTBP1 depletion from NPCs altered 325 splicing events ( 264 cassette exons , 44 alternative 5’ or 3’ splice sites , and 17 retained introns , Figure 2—figure supplement 1B , Figure 2—source data 3 ) . Comparing this list to the splicing changes between NPCs and Day 5 GFP+ MNs identified 16% of neuronal exons regulated by PTBP1 ( 66 PTBP1-repressed exons increased with differentiation; 23 PTBP1-activated exons decreased with differentiation; Figure 2D , left panel; Figure 2—source data 4 ) . Interestingly , PTBP1/2 double knockdown increased the number of regulated neuronal exons to 30% ( 116 PTBP1/2-repressed exons increased with differentiation; 45 PTBP1/2-activated exons decreased with differentiation; Figure 2D , middle panel ) . These changes in splicing after double PTBP1/2 knockdown in NPCs also tended to be larger than PTBP1 knockdown alone ( Figure 2—figure supplement 1D ) , yielding a lower correlation than seen in ESC ( R2 = 0 . 69 , Figure 2—figure supplement 1F ) . Knockdown of PTBP2 alone in NPCs had only a small effect on the splicing of neuronal exons , indicating that PTBP2 largely regulates a subset of PTBP1 targets , with relatively few targets of its own ( Figure 2—figure supplement 1A and D ) ( Figure 2D , right panel ) . This is in agreement with previous results that a large set of exons in the developing mouse brain is regulated by both PTBP1 and PTBP2 . These include Dlg4 ( Psd-95 ) exon 18 ( Boutz et al . , 2007; Li et al . , 2014; Zheng et al . , 2012 ) . In these new data , PTBP1 and PTBP2 single knockdown each only moderately increased Dlg4 exon 18 splicing ( ΔPSI = 8% and 13% respectively ) , whereas PTBP1/2 double knockdown increased exon 18 splicing by 42% . Other exons where PTBP2 can compensate for the loss of PTBP1 are found in the Magi1 , Ap2a1 , and Tnrc6a transcripts ( Figure 2—source data 5 ) . Thus , PTBP1 and PTBP1/2 depletion from NPCs identifies exons that are coregulated by both proteins . Whereas , depletion from ESCs identifies exons affected largely by PTBP1 . The PTBP target transcripts in ESCs and NPCs include both unique and overlapping splicing events ( 80 shared , 73 unique to ESCs , 81 unique to NPCs ) . To examine possible common functions of the PTBP1/2 regulatory programs , Gene Ontology analyses were performed relative to the transcripts expressed in each cell type ( Huang et al . , 2009 ) . These yielded fewer enriched terms compared to total set of neuronally regulated exons ( Figure 1—figure supplement 2C and D ) . Neuronal exons regulated by PTBP1 in ESCs showed significant enrichment ( FDR = 0 . 04 ) in the cellular compartment ontology term , cell junction . Neuronal exons regulated by PTBP1/2 in NPCs were enriched ( FDR = 0 . 01 ) in the cellular compartment ontology term , cytoskeleton . The total set of neuronal exons identified during ESC and NPC differentiation were also enriched for these terms among others ( Figure 1—figure supplement 2C and D ) . The relative lack of enriched terms might reflect a broad number of functions for PTBP targets , but this will require further analysis to assess . Enrichment for cytoskeletal functions is in agreement with previous analyses of PTBP2 function in mice , and with the dramatic cytoskeletal changes during the transition from PTBP1 to PTBP2 during early neuron differentiation ( Li et al . , 2014; Licatalosi et al . , 2012 ) . To identify direct binding of PTBP1 to transcripts in these cells , we performed crosslinking-immunoprecipitation followed by sequencing ( iCLIP-seq ) ( König et al . , 2010; König et al . , 2011 ) . UV irradiated ESC and NPC cultures were immunoprecipitated with PTBP1 antibody or with Flag antibody as a negative control ( Figure 3—figure supplement 1A ) . Cross-linked RNA fragments were converted into cDNA using a modification of the iCLIP protocol and subjected to high-density sequencing ( König et al . , 2010 ) . After removing PCR duplicates , unique sequencing reads were mapped to the UCSC known Genes table ( Hsu et al . , 2006d ) . To identify significant crosslinking sites , the FDR for each position was calculated ( König et al . , 2010 ) . Applying an FDR cutoff of 0 . 01 , we obtained 918 , 725 and 914 , 084 significant reads of RNA crosslinked to PTBP1 in ESCs and NPCs respectively . The PTBP1-bound sequences were analyzed for base content and location . The iCLIP protocol generally yields fragments that terminate one nucleotide downstream of the crosslink site ( König et al . , 2010 ) . We compiled a set of binding regions extending 20 nucleotides upstream and downstream of a crosslink site . We measured the frequency of all pentamer motifs within these binding sites , relative to randomly chosen intervals from the same introns . As expected , PTBP1 binding sites in both ESC and NPC were highly enriched for CU-rich pentamers . The top 20 motifs included pentamers comprised of only C and U nucleotides , as well as some with G nucleotides , in agreement with the described binding specificity of PTBP1 ( Figure 3—figure supplement 1B ) ( Amir-Ahmady et al . , 2005; Han et al . , 2014; Licatalosi et al . , 2012; Llorian et al . , 2010; Oberstrass et al . , 2005 ) . The enriched motifs were very similar between ESC and NPC samples ( R2 = 0 . 93 , Figure 3—figure supplement 1B ) . The locations of the binding sites were also consistent with previous observations , with most being found in introns and some in 3’ UTRs ( Figure 3—figure supplement 1C and D ) ( Xue et al . , 2009 ) . We next examined the alternative exons sensitive to PTBP1 knockdown ( ΔPSI ≥ 15% ) for PTBP1 iCLIP clusters within 500 nt upstream or downstream ( ≥ four significant reads per cluster ) . In ESCs , we identified 170 PTBP1-dependent exons that are bound by PTBP1 ( Figure 3—figure supplement 1E , Figure 3—source data 1 ) . In NPCs , we identified 108 PTBP1-dependent exons with PTBP1 binding ( Figure 3—figure supplement 1F , Figure 3—source data 1 ) . These include many known PTBP1 target transcripts such as Ganab , Rod1 , Tpm1 , Rbm27 , Fam38a , Mink1 , and Ptbp2 ( Boutz et al . , 2007; Makeyev et al . , 2007; Spellman et al . , 2007; Xue et al . , 2009 ) . Some known target exons with PTBP1 binding did not exhibit splicing changes above our cutoff , including Pkm2 , Eif4g2 , Dlg4 , and Ptbp1 ( Xue et al . , 2009; Zheng et al . , 2013 ) . These exons may be controlled by additional factors in ESCs and NPCs . Overlapping all of the datasets , we defined a minimal set of 104 direct PTBP1 targets in ESCs and NPCs , whose splicing changes during neuronal differentiation ( 70 Ptbp1 repressed , 34 Ptbp1 activated; Figure 3—source data 2 ) . Many other exons are expected to be direct targets but were excluded due to failure to fulfill a cutoff value in one dataset . The exons in this minimal set affect a wide range of cellular processes including GTPase signaling ( Dock7 ) , synaptic function ( Gabbr1 ) , and transcriptional regulation ( Tcf20 ) ( Figure 3A–D ) . For example , Gabbr1 is a subunit of a metabotropic gamma-aminobutyric acid ( GABA ) receptor . PTBP1 induced skipping of Gabbr1 exon 15 leads to premature translation termination in exon 16 and predicted nonsense mediated decay of the Gabbr1 mRNA ( Makeyev et al . , 2007 ) . We observe PTBP1 binding within exon 15 and upstream , consistent with the increased exon 15 splicing observed during neuronal differentiation and after PTBP1 depletion ( Figure 3A ) . This gene is similar to Dlg4 , where PTBP1 regulates its overall expression during development rather than the production of a neuronal isoform . However unlike Dlg4 , Gabbr1 is more sensitive to PTBP1 than PTBP2 , allowing its expression to be induced earlier . 10 . 7554/eLife . 09268 . 014Figure 3 . Examples of neuronally spliced cassette exons that are PTBP1 repressed . ( A-D ) Genome Browser tracks show aligned RNA-seq reads from ESCs ( red ) , NPCs ( blue ) , and GFP+ MNs ( yellow ) . The scale indicates the number of mapped reads in the highest peak . PSI values for the cassette exons were calculated with SpliceTrap ( Wu et al . , 2011 ) . Significant PTBP1 iCLIP sequencing reads from ESCs ( magenta ) and NPCs ( green ) are overlaid on the Genome Browser tracks . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01410 . 7554/eLife . 09268 . 015Figure 3—source data 1 . PTBP1-regulated cassette exons with PTBP1 binding . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01510 . 7554/eLife . 09268 . 016Figure 3—source data 2 . Direct PTBP1 target exons during neuronal differentiation . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01610 . 7554/eLife . 09268 . 017Figure 3—figure supplement 1 . The base content and location of PTBP1 iCLIP-sequencing reads are consistent with known PTBP1 binding properties . ( A ) Autoradiograph of the nitrocellulose membrane containing protein-RNA complexes isolated from ESCs and NPCs using either Flag or PTBP1 antibodies . Boxes indicate the region of the membrane used for iCLIP library preparation . ( B ) Scatter plot compares the calculated pentamer Z-scores from ESCs with scores from NPCs . The top 23 pentamer motifs , which had Z scores >200 , are listed . ( C , D ) Pie charts showing the percent of the significant PTBP1 iCLIP reads that were found in 5’ UTR ( blue ) , coding sequence ( CDS ) ( red ) , intron ( green ) , and 3’ UTR ( purple ) regions . ( E , F ) All expressed cassette exons with nearby PTBP1 iCLIP clusters ( within 500 nt , blue circles ) are overlapped with PTBP1-responsive cassette exons ( green circles , ΔPSI ≥ 15% ) ( Figure 3—source data 1 ) . P-values were calculated using the hypergeometric test . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 017 The above data define the PTBP1 regulatory program in a new level of detail , and show that for early cells along the neuronal lineage , the PTBP1 program is distinct from that regulated by PTBP2 . PTBP1 target transcripts influence an array of important cellular functions , and understanding how splicing alters each target’s activity in differentiating cells will be a challenge . We were particularly interested in how the early PTBP1 program affects gene regulation . We identified a number of transcriptional regulators whose splicing is controlled by PTBP1 during ESC differentiation into NPCs , including Tcf20 , Med23 , Gatad2a , and Pbx1 . These early changes in the structure of transcriptional regulators have the potential to broadly alter gene expression programs during neuronal development ( Gabut et al . , 2011; Raj et al . , 2011; Demir and Dickson , 2005b ) . We focused on the homeodomain transcription factor Pbx1 because of its known interactions with the Hox family of developmental regulators ( Piper et al . , 1999a; LaRonde-LeBlanc , 2003a ) . Multiple studies implicate Pbx1 and its family members as transcriptional regulators during neuronal development ( Maeda et al . , 2002a; Vitobello et al . , 2011b; Sgadò et al . , 2012c; Schulte et al . , 2014 ) . Exon 7 in the Pbx1 transcript is conserved across mammals and its inclusion leads to production of the Pbx1a isoform . Skipping exon 7 shifts the translational reading frame to introduce a new termination codon ( PTC ) in exon 8 ( Figure 4A ) . This translation stop does not result in NMD but instead generates the shorter isoform , Pbx1b , which retains the DNA binding homeodomain but lacks the 83 amino acids at the C-terminus of Pbx1a . During development , early embryonic tissues predominantly express Pbx1b , whereas Pbx1a is found in neural tissues ( Redmond et al . , 1996; Schnabel et al . , 2001b ) . We found a similar isoform switch in cultured cells , where ESCs express low levels of Pbx1b . As ESCs differentiate into NPCs and HB9+ MNs , both overall Pbx1 expression and exon 7 splicing are induced , such that MNs predominantly express the Pbx1a isoform ( Figure 4B and C ) . Previous work indicates that the Pbx1a and Pbx1b isoforms have different transcriptional activities and cellular functions . The proteins differ in their ability to activate or repress reporter gene expression , their interactions with the transcriptional corepressors NCoR1 and NCoR2 , and their activity for cell transformation when fused to E2A ( Kamps et al . , 1991; Di Rocco et al . , 1997; Asahara et al . , 1999b ) . However , the role of these functional differences during neuronal differentiation is not understood . 10 . 7554/eLife . 09268 . 018Figure 4 . PTBP1 regulates a switch in Pbx1 isoform expression . ( A ) Exons are represented as boxes , while horizontal lines represent introns . Inclusion of exon 7 results in the longer Pbx1a protein ( top ) . Skipping of exon 7 results in a frameshift and translation termination at a stop codon ( * ) in exon 8 to yield the shorter Pbx1b protein isoform ( bottom ) . The DNA-binding homeodomain ( HD ) is indicated by a blue box . ( B ) Genome Browser tracks of aligned RNA-seq reads show exon 7 inclusion as ESCs ( red ) differentiate into NPCs ( blue ) and GFP+ MNs ( yellow ) . A significant PTBP1 iCLIP cluster is present upstream of exon 7 in NPCs ( green ) . ( C ) Both Pbx1 exon 7 ( top panel , solid line bottom panel , n=3 ) and Pbx1 mRNA expression ( dashed line bottom panel , n=2 ) are induced with MN differentiation . ( D , E ) ESCs and NPCs were treated with siControl , siPtbp1 , or both siPtbp1 and siPtbp2 . ( D ) RT-PCR of Pbx1 exon 7 splicing following siRNA treatment in ESCs ( 24 PCR cycles ) and NPCs ( 19 PCR cycles ) . ( E ) Bar chart of Pbx1 exon 7 splicing ( Mean ± SEM , n=3 ) . Statistical analyses were performed using paired one-tailed Student’s t-test ( P-value<0 . 01** , 0 . 05* ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 018 We find that Pbx1 exon 7 is regulated by PTBP1 . Exon 7 is derepressed as ESCs differentiate into NPCs and further into motor neurons ( Figure 4B and C ) . Depletion of PTBP1 from ESCs or NPCs strongly induces exon 7 splicing ( Figure 4D and E ) . In NPCs , a PTBP1 iCLIP cluster is found in the intron upstream of exon 7 ( Figure 4B ) . These iCLIP tags are not seen in ESCs presumably due to lower Pbx1 transcript levels in these cells ( Figure 4B and C ) . These data indicate that one consequence of PTBP1 depletion during neuronal differentiation is to switch Pbx1 expression from the Pbx1b isoform to Pbx1a . PTBP1 represses the production of Pbx1a in ESCs and during early neuronal differentiation . To examine the role of Pbx1a repression , we modified the Pbx1 locus using the CRISPR-Cas9 technology to force exon 7 splicing in cells containing PTBP1 ( Doudna and Charpentier , 2014; Cong and Zhang , 2015e ) . Intron 6 upstream of exon 7 contains a PTBP1 iCLIP cluster as well as regions of mammalian sequence conservation , possibly indicative of additional splicing regulatory elements . Guide RNA pairs were designed to target the Cas9 endonuclease to sites within intron 6 ( Figure 5A ) . After Cas9 catalyzes double-strand breaks at each site , non-homologous end joining ( NHEJ ) of the free ends will potentially generate genomic deletions within intron 6 ( Figure 5—figure supplement 1A ) . The largest deletion , between guides 1 and 4 , would remove 5 . 5 kb of intron 6 containing all of the conserved regions , while preserving the splice sites ( Figure 5A ) . HB9-GFP ESCs were transiently transfected with expression plasmids for Cas9 and two guide RNAs , and clones were isolated from single transfected cells and genotyped by PCR , with the deletion event confirmed by sequencing . Using guides g1 and g4 , we identified no homozygous deletions among 60 ESC clones , although heterozygotes were readily isolated ( Figure 5—figure supplement 1B and C ) . We expanded five heterozygous deletion clones with preserved splice sites . Notably , we were still unable to isolate homozygous null cells after reintroducing guides 1 and 4 into a heterozygous cell line . Shorter deletions were also tested . Guides 3 and 4 generated a 1 . 4 kb deletion that includes the PTBP1 binding site . Again this deletion was only isolated as a heterozygote ( 1/6 clones ) . In contrast , guides 2 and 3 generated a 1 . 5 kb deletion from the middle of the intron that was isolated as a homozygous allele ( 3/10 clones ) . 10 . 7554/eLife . 09268 . 019Figure 5 . Intron 6 deletions in Pbx1 induce exon 7 splicing . ( A ) Guide RNAs were designed to target regions along Pbx1 intron 6 . The location of the PTBP1 iCLIP cluster is indicated by the black box . ( B ) RT-PCR of Pbx1 exon 7 splicing in ESC clones carrying Cas9-mediated deletions . The asterisk ( * ) indicates a sporadic non-specific band . ( C ) Bar chart of Pbx1 exon 7 splicing in ESC clones with intron 6 deletions . Error bars indicate the SEM for three independent clones , except for g3-4 where only one clone was isolated . ( D ) RT-PCR of Pbx1 exon 7 splicing and ( E ) western blot of Pbx1 in D2 MN cultures show that Pbx1a is the predominant isoform in the I6 +/- cell line at day 2 . Solid borders of RT-PCR gels indicate non-adjacent gel lanes . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 01910 . 7554/eLife . 09268 . 020Figure 5—figure supplement 1 . Cas9-targeted deletions of Pbx1 intron 6 switch Pbx1 isoform expression . ( A ) Schematic of Cas9-targeted deletion of Pbx1 intron 6 . ( B ) Location of primers used for intron 6 genotyping PCRs . A list of primer sequences is provided in Supplementary file 1 ( C ) Genotyping PCR of the wild type and the intron 6 deletion cell lines . Table showing the number of clones screened and number of edited cell lines isolated . ( D ) Bar graph of Pbx1 exon 7 splicing in D2 MN wild type ( n=3 clones ) and I6 +/- ( n=5 clones ) cultures . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 020 Interestingly , these deletions had different effects on Pbx1 splicing . Clones carrying the heterozygous Intron 6 g1-g4 deletion ( I6 +/- ) exhibited dramatically increased exon 7 splicing ( Figure 5B and C ) . Exon 7 was included in only 16% of the Pbx1 mRNA in clones of the parental ESC line , whereas exon 7 was included in 51% of the mRNA in the I6 +/- clones . This indicates that RNA from the mutant allele is almost entirely splicing in exon 7 and confirms that intron 6 contains splicing silencer elements for this exon . We also tested the other deletions within intron 6 for their effects on exon 7 splicing . The deletion derived from guides 3 and 4 was similar to the long deletion using guides 1 and 4 , with the heterozygous deletion clone exhibiting strongly induced exon 7 splicing ( g3-4 in Figure 5B and C ) . In contrast , the homozygous intronic deletion derived from guides 2 and 3 had minimal effect , with exon 7 splicing similar to wild type in 3 separate clones ( g2-3 in Figure 5B and C ) . The inability to isolate cells carrying homozygous mutations that increase exon 7 splicing suggests that either the gain of Pbx1a or the loss of Pbx1b is deleterious for ESC growth . To characterize how early Pbx1a expression alters transcriptional programs , we measured steady-state transcript levels in cells with different Pbx1 genotypes by RNA-seq . Wild type ( n=3 ) and I6 +/- ( n=5 ) ESC clones were differentiated into MNs . At Day 2 , the I6 +/- cells predominantly expressed the Pbx1a mRNA isoform , leading to dramatically increased Pbx1a protein compared to wildtype cells ( Figure 5D and E ) . These Day 2 samples were subjected to 50 nt single-end RNA sequencing , with the data analyzed by the Cufflinks pipeline . We performed hierarchical clustering on the replicates , and as expected the gene expression profiles were similar across replicates , with the profiles of the wild type and Pbx1 mutant cell replicates segregating into two groups ( data not shown ) . Using a twofold change in FPKM as the cutoff , we identified 33 differentially expressed genes in the I6 +/- cultures ( 31 induced , 2 repressed ) . Examining these genes with the largest changes in expression confirmed that a Pbx1a neuronal regulatory program was activated in the mutant cells . Of the 20 transcripts that are most altered in I6 +/- , 14 are also induced as wildtype ESCs differentiate in NPCs and MNs ( Figure 6—source data 1 and 2 ) . Multiple genes of this group are known to play important roles in neuronal differentiation including Phox2b , Cntn2 , Ntng2 , Olig1 , Isl1 , Nrp2 , Ngfr , Nav2 , and Slit1 ( Figure 6B ) , and many were previously shown to be affected by Pbx1 ( Igf2 , Isl1 , Dlk1 , and Meox1 ) ( Kim et al . , 2002; Jürgens et al . , 2009b; Thiaville et al . , 2012d ) . Notably , the gene exhibiting the largest change in expression was Phox2b , a marker of hindbrain visceral motor neurons and a known Pbx1a target ( Pattyn et al . , 2000c; Samad , 2004 ) . 10 . 7554/eLife . 09268 . 021Figure 6 . The early expression of Pbx1a promotes neuronal gene expression . ( A ) Genome Browser tracks of aligned RNA-seq reads from wild type ( top ) and I6 +/- ( bottom ) Day 2 MN cultures . The asterisk ( * ) indicates Pbx1 exon 7 , which is induced in the mutant . ( B ) Bar charts of the top 20 induced genes in the I6 +/- cell lines compared to expression in wild type cells ( Figure 6—source data 1 ) . Genes highlighted in bold have known roles in neuronal differentiation . Error bars indicate SEM . ( C ) Gene ontology analysis of the 196 genes induced by 1 . 5 fold in the I6 +/- cultures . ( D ) Hoxc5 expression increases during neuronal differentiation ( left panel , Figure 6—source data 2 ) and the induction of Pbx1a . ( E ) Relative Pbx1 ( left panel ) and Meis1 ( right panel ) binding in D2 MN cultures ( n=3 ) . Statistical analyses were performed using Welch’s t-test ( P-value<0 . 05* , <0 . 01** ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 02110 . 7554/eLife . 09268 . 022Figure 6—source data 1 . Expression changes in I6 +/- Day 2 MN cultures . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 02210 . 7554/eLife . 09268 . 023Figure 6—source data 2 . Expression values in ESCs , NPCs , and GFP+ MNs . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 02310 . 7554/eLife . 09268 . 024Figure 6—figure supplement 1 . Pbx1 binds target neuronal genes . ( A-E ) Gene expression of several Pbx1 targets increases during neuronal differentiation ( left panels , Figure 6—source data 2 ) . Inducing Pbx1a expression increases the expression of these genes ( middle left panel , purple bars ) . ChIP-qPCR demonstrates that Pbx1 ( middle panel , blue bars ) , Meis1 ( middle right panel , red bars ) , and Prep1 ( right panel , orange bars ) bind neuronal target genes in D2 MN cultures . Statistical analyses were performed using Welch’s t-test ( P-value<0 . 05* , <0 . 01** ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09268 . 024 To increase the number of differentially expressed genes , we relaxed the cutoff to a 1 . 5-fold change in expression and identified 196 induced and 25 repressed genes in the mutant relative to the wild type cells . Of the 196 induced genes , 139 increase expression with ESC neuronal differentiation . Gene ontology ( GO ) analyses were performed on the induced gene set to assess functional enrichments relative to the total transcripts present in the D2 MN cultures . Within the biological processes ontology , several terms were significantly enriched ( FDR<0 . 05 ) including axonogenesis , pattern specification , regulation of transcription , cell adhesion , cell motion , cell fate commitment , and heart development ( Figure 6C ) . Notably , the induced genes include the MN markers Olig2 , Lhx3 , and Mnx1 ( Figure 6—source data 1 ) , again indicating that Pbx1a is activating a neuronal transcriptional program . Overlapping the 196 induced events with a published Pbx1 ChIP data set identified 52 genes with nearby Pbx1 binding ( Penkov et al . , 2013b ) ( Figure 6—source data 1 ) . Even though these binding events were identified in E11 . 5 embryos , 26 . 5% of induced transcripts in D2 MNs exhibit Pbx1 binding . This is compared to 14 . 8% of all expressed genes , constituting a significant enrichment of Pbx1 binding sites adjacent to induced transcripts ( p=1 . 06e-5 by hypergeometric test ) . We were particularly interested in the up-regulation of several transcription factors with known roles in neuronal differentiation , including the homeobox transcription factor Hoxc5 ( Figure 6D ) . Hoxc5 affects several aspects of MN development and function , including the regulation of the phrenic motor neuron column ( Liu et al . , 2001c; Dasen et al . , 2003b; Philippidou et al . , 2012e ) . To confirm Pbx1 binding in D2 MN cultures , we performed ChIP-qPCR . All the genes tested , including Hoxc5 , Hoxa3 , Midn , Tshz3 , Zfp503 , and Igf1r showed higher Pbx1 binding than an IgG control , although the significance relative to the control varied ( n=3 , Figure 6E , Figure 6—figure supplement 1A–E ) . Differences in Pbx1 binding were not observed between the wildtype and I6+/- cell lines , as might be expected since both splice variants maintain the DNA binding homeodomain of Pbx1 . Pbx1 is known to cooperate with the transcriptional cofactors Meis1 and Prep1 ( Pknox1 ) in regulating gene expression . To assess whether Pbx1 isoforms differentially affect Meis1 and Prep1 recruitment , we performed ChIP-qPCR in D2 MN cultures at shared binding sites , identified in previous ChIP-seq studies ( Penkov et al . , 2013b ) . Most genes showed no change in Prep1 or Meis1 binding between wildtype and I6+/- , indicating that the Pbx1b and Pbx1a isoforms likely differ in other interactions . Notably , there was a significant increase in Meis1 binding at the Hoxc5 locus when Pbx1a was induced in I6+/- ( Figure 6E ) . Thus , Pbx1a may increase Hoxc5 expression through enhanced recruitment of Meis1 . Together the data indicate that switching to Pbx1a increases the expression of genes affecting neuronal fate , including Hoxc5 .
Neuronal development is driven by a complex set of genetic events affecting every step in gene expression from transcriptional regulation by specialized DNA binding proteins to translational repression by miRNAs . These levels of regulation are intricately connected , where one type of regulator will target effectors at different steps in the gene expression pathway . In particular , alternative splicing plays a key role in defining the neuronal proteome with thousands of spliced isoforms found specifically in neurons or particular neuronal subtypes ( Barbosa-Morais et al . , 2012; Grabowski and Black , 2001; Merkin et al . , 2012 ) . A variety of RNA binding proteins alter splicing patterns during development and in the adult brain , including members of the PTBP , Rbfox , Nova , CELF , SRRM and other protein families ( Zheng and Black , 2013; Norris and Calarco , 2012a; Raj and Blencowe , 2015a ) . The changes in isoform expression controlled by these factors determine the activity of many essential neuronal genes . However , these regulatory programs are incompletely characterized and the cellular roles of their alternatively spliced products are often not understood . The polypyrimidine tract binding proteins , PTBP1 and PTBP2 , influence neuronal differentiation through programs of posttranscriptional gene regulation . In previous work , we found that the transition from PTBP1 to PTBP2 expression during neuronal differentiation causes a large scale reprogramming of splicing patterns ( Boutz et al . , 2007 ) . More recently , we identified a second transition late in neuronal maturation , where PTBP2 levels decline allowing for the expression of many adult-specific spliced isoforms ( Li et al . , 2014; Licatalosi et al . , 2012; Zheng et al . , 2012 ) . Like many splicing regulators , the PTB proteins also affect other aspects of mRNA metabolism . The depletion of PTBP1 from mouse embryonic fibroblasts ( MEFs ) was shown to induce the trans-differentiation of these cells into neurons ( Xue et al . , 2013 ) . This was linked to PTBP1 bound in 3’ UTRs altering miRNA targeting of transcripts driving neuronal differentiation , including the REST transcriptional repressor . In contrast , alterations in neuronal differentiation were not observed in mice carrying a conditional Ptbp1 knockout mutation in nestin-positive NPCs ( Shibasaki et al . , 2013 ) . Similarly , we have not found that PTBP1 depletion is sufficient to induce neuronal differentiation of ESCs . This could reflect a difference between ESCs and NPCs compared to MEFs , or differences in culture conditions . Notably , we find here that , like its 3’ UTR targets , the PTBP1 splicing program also includes targets that enhance neuronal development , such as Pbx1 . Earlier studies demonstrated that while PTBP1 and PTBP2 share many targets , exons can have different sensitivities to the two proteins ( Keppetipola et al . , 2012 ) . This raised questions about the mechanisms of their differential targeting and the biological roles of the early and late neuronal splicing programs . The late program was illuminated in the PTBP2 knockout mouse where many exons that are repressed by both proteins were identified ( Li et al . , 2014; Licatalosi et al . , 2012 ) . However , the early program composed of exons primarily targeted by PTBP1 was harder to define in the developing brain . By examining the PTBP1 splicing program in ESC culture , we now identify a large set of targets primarily responsive to PTBP1 regulation . Interestingly in ESCs , exons that are sensitive PTBP1 depletion are largely not responsive to the concomitant PTBP2 induction . This contrasts with NPCs where exons whose splicing shifts with PTBP1 depletion are often more strongly affected by PTBP1/2 co-depletion . We previously found that in mice heterozygous for the PTBP2 knockout mutation , certain exons were spliced at 50% the level seen in the homozygous null , indicating a strong effect of PTBP2 concentration . Similarly , we find here that certain exons , such as Gabbr1 exon 15 , are very sensitive to moderate changes in PTBP1 expression , such as the reduction seen when ESCs differentiate into NPCs . Like PTBP2 , the PTBP1 targets display a range of responsiveness to PTBP1 concentration , and the ESC system gives us a new tool for examining this earlier PTBP1 dependent program . Alternative exons are usually regulated by ensembles of splicing factors acting to repress or activate their splicing ( Fu and Ares , 2014; Lee and Rio , 2015 ) . A question of interest is how targeting by the PTB proteins is affected by other neuronal splicing factors . Two known PTBP1 cofactors that may also affect PTBP2 are Matrin3 and Raver1 ( Coelho et al . , 2015; Rideau et al . , 2006; Huttelmaier , et al . , 2001d ) . Both these proteins are likely to affect the activity of PTBP1 and PTBP2 on certain targets . We find that both proteins are well expressed in ES cells , and Matrin3 is strongly upregulated with differentiation into NPCs and neurons . We recently identified several potential cofactors that alter the splicing of the PTBP1/2 target exon in Dlg4 ( Zheng et al . , 2013 ) . These will also be interesting to examine in relation to additional PTBP targets , and whether they more strongly affect exons controlled by PTBP1 , PTBP2 or both proteins . A protein that can counteract PTBP repression is nSR100/SRRM4 , which is induced with neuronal differentiation and whose targets include some PTBP1/2 targets ( Calarco et al . , 2009; Raj et al . , 2014 ) . SRRM4 expression coincides with PTBP2 , and its role may be to specifically antagonize the effects of PTBP2 on certain exons in immature neurons . It will be also interesting to identify the SRRM4 target exons during NPC differentiation and to assess their sensitivity to PTBP2 compared to PTBP1 . The intent of this study was to define a set of targets that are primarily responsive to PTBP1 , and thus may affect early neuronal lineage commitment and differentiation . We identify a diverse group of transcripts that are sensitive to PTBP1 depletion from ESCs and which change their splicing when ESCs differentiate into NPCs and then into early neurons , as PTBP1 is reduced . Among these targets , we focused on Pbx1 , which contains an alternative exon 7 that is highly responsive to PTBP1 concentration and is implicated in neuronal transcriptional regulation . Using Cas9 genome editing , we created Pbx1 alleles that eliminate PTBP1 repression of exon 7 , causing its premature splicing during development ( Doudna and Charpentier , 2014; Cong and Zhang , 2015e ) . This early switch in Pbx1 isoforms was sufficient to induce a variety of transcripts , including multiple mRNAs implicated in neuronal lineage commitment . Thus , one function of PTBP1 is to repress the neuronal form of Pbx1 and prevent its action at target genes early in development . It will be interesting to further examine how different interactions of the two Pbx1 isoforms might lead to their different activities . Pbx1b lacks 83 amino acids at the C-terminus of Pbx1a but retains domains that facilitate interactions with the Hox , Prep , and Meis transcription factors ( Longobardi et al . , 2013 ) . An earlier study found that Pbx1a and Pbx1b differ in their interactions with the transcriptional corepressors NCoR1 and NCoR2 and their ability to repress reporter gene expression ( Asahara et al . , 1999b ) . Our data indicates that Pbx1a may enhance Meis1 binding at the Hoxc5 locus to activate its expression . Thus , the changes in Pbx1 structure likely influence enhancer complex assembly at specific genes . A splicing factor will often regulate a large network of targets that may overlap with the targets of other factors . The complexity of these regulatory programs is a challenge for understanding their biological roles . Mouse mutations in splicing regulators often exhibit lethal or highly pleiotropic phenotypes . Among the many affected transcripts , few isoforms are usually sufficiently analyzed to understand their altered activity . Cas9 genome editing offers an tool for examining differential isoform activity in the developing nervous system ( Gueroussov et al . , 2015c ) . Combining genome editing with stem cell differentiation allows the characterization of isoform activity in specific cells at specific times in development , and provides means for investigating the function of other PTB targets within the larger programs of neurogenesis .
HB9-GFP and 46C mouse ESCs were gifts of H . Wichterle and T . Jessell , Columbia University , and of J . Sanford , UCSC , respectively . Cells were confirmed for expression of appropriate developmental and transgenic markers by immunostaining , and to be mycoplasma-free by PCR-based testing . 46C and HB9-GFP mouse ESCs were grown on 0 . 1% gelatin-coated dish with CF1 mouse embryonic fibroblasts ( Applied StemCell , Inc . , Menlo Park , CA ) in ESC media . ESC media consisted of DMEM ( Fisher Scientific , Hampton , NH ) supplemented with 15% ESC-qualified fetal bovine serum ( Life Technologies , Carlsbad , CA ) , 1X non-essential amino acids ( Life Technologies ) , 1X GlutaMAX ( Life Technologies ) , 1X ESC-qualified nucleosides ( EMD Millipore , Billerica , MA ) , 0 . 1 mM 2-Mercaptoethanol ( Sigma-Aldrich , St . Louis , MO ) , and 1000U/ml ESGRO leukemia inhibitor factor ( EMD Millipore ) . 46C and HB9-GFP ESCs were differentiated into NPCs according to ( Conti et al . , 2005 ) , with minor modifications . Briefly , feeder-free ESCs were differentiated on 0 . 1% gelatin-coated dish in N2B27 media . N2B27 media consisted of 1:1 mixture of Neurobasal media ( Life Technologies ) and DMEM/F12 ( Fisher Scientific ) supplemented with 0 . 5X B27 without Vitamin A ( Life Technologies ) , 0 . 5X N2 Supplement ( Life Technologies ) , 1X GlutaMAX ( Life Technologies ) , and 0 . 1 mM 2-Mercaptoethanol ( Sigma-Aldrich ) . After seven days of differentiation , cells were trypsinized and grown in aggregate culture for two days in N2B27 supplemented with 10 ng/ml recombinant human EGF ( PeproTech , Rocky Hill , NJ ) and 10 ng/ml recombinant human FGF-basic ( PeproTech ) . Aggregates were plated on poly-ornithine-coated ( 15 ug/ml , Sigma ) and fibronectin-coated ( 1 . 5 ug/ml , Sigma-Aldrich ) dishes in NPC media . NPC media consisted of DMEM/F12 ( Fisher Scientific ) supplemented 1X B27 without Vitamin A ( Life Technologies ) , 1X GlutaMAX ( Life Technologies ) , 0 . 1 mM 2-Mercaptoethanol ( Sigma-Aldrich ) , 10 ng/ml recombinant human EGF ( PeproTech ) , and 10 ng/ml recombinant human FGF-basic ( PeproTech ) . NPCs were maintained on poly-ornithine-coated ( 15 ug/ml , Sigma-Aldrich ) and fibronectin-coated ( 1 . 5 ug/ml , Sigma-Aldrich ) dishes in NPC media for one to two passages . NPCs were differentiated with the removal of EGF and FGF , along with the addition of 1 uM all-trans retinoic acid ( RA , Sigma-Aldrich ) . HB9-GFP ESCs were differentiated to MNs according to ( Wichterle et al . , 2002 ) , with minor modifications . Briefly , feeder-free ESCs were grown as aggregate culture for two days in MN media . MN media consisted of 1:1 mixture of Neurobasal media ( Life Technologies ) and DMEM/F12 ( Fisher Scientific ) supplemented with 10% Knockout Serum Replacement ( Life Technologies ) , 1X GlutaMAX ( Life Technologies ) , and 0 . 1 mM 2-Mercaptoethanol ( Sigma-Aldrich ) . After two days , the MN media was supplemented with 1X N2 Supplement ( Life Technologies ) , 1 uM all-trans retinoic acid ( RA , Sigma-Aldrich ) , and 1 uM smoothened agonist ( EMD Millipore ) . After five days of RA and SAG addition , EBs were plated on BD Matrigel ( BD Biosciences ) , with the addition of 10 ng/ml BDNF ( R&D Systems , Minneapolis , MN ) , 10 ng/ml GDNF ( R&D Systems ) , and 10 ng/ml CNTF ( R&D Systems ) . EBs were collected 5 or 8 days after RA and SAG addition . EBs were dissociated in Acutase ( Innovative Cell Technologies , San Diego , CA ) for 10 min and sorted using a BD FACSAria cell sorter at the UCLA Broad Stem Cell center core facility . 46C ESCs and NPCs were transfected with Silencer Select Ptbp1 ( Life Technologies; s72335 , s72336 ) and Ptbp2 ( Life Technologies; s80149 , s80149 ) siRNAs using RNAiMax ( Life Technologies ) according to the manufacturer’s recommendations . Silencer Negative Control siRNA #1 ( Life Technologies ) and siGENOME Non-Targeting siRNA Pool #1 ( GE Healthcare , Pittsburg , PA ) were used as negative controls . ESCs were reverse transfected with 20 nM of siRNAs , treated again 24 hr later , and collected 72 hr post-transfections . NPCs were reverse transfected with 10 nM of siRNAs and collected 48 hr post-transfections . Western blots were performed on total protein from cell cultures lysed in RIPA buffer supplemented with protease inhibitors ( Roche , Basel , Switzerland ) and Benzonase ( Sigma-Aldrich ) . Lysates were diluted in 4X SDS loading buffer , heated for at 95°C for 10 min , and loaded onto 10% polyacrylamide Laemmli SDS PAGE gels . Gels were run under standard electrophoresis conditions . Transfers were performed on a Novex X-Cell mini-cell transfer apparatus ( Life Technologies ) onto Immobilon-FL PVDF membranes ( EMD Millipore ) . The membranes were probed under standard conditions with primary antibodies overnight at 4°C . The following primary antibodies were used: rabbit anti-PTBP1 antibody PTB-NT ( 1:3000 ) ( Markovtsov et al . , 2000 ) , rabbit anti-PTBP2 antibody nPTB-IS2 ( 1:1000 ) ( Sharma et al . , 2005 ) , rabbit anti-Pbx1 antibody ( 1:500 , Cell Signaling , Danvers , MA ) , rabbit anti-U170K antibody ( 1:1000 ) ( Sharma et al . , 2005 ) , and mouse anti-GAPDH 65C antibody ( 1:10000 , Life Technologies ) . For fluorescent detection , the membranes were probed with ECL Plex Cy3 and Cy5-conjugated goat anti-mouse and goat anti-rabbit secondary antibodies ( 1:2000; GE Healthcare ) . The blots were then scanned on a Typhoon Phosphorimager ( GE Healthcare ) and quantified using ImageQuant software ( GE Healthcare ) . For chemiluminescent detection , the membranes were probed with Amersham ECL HRP Conjugated Antibodies ( 1:4000 , GE Healthcare ) and developed using SuperSignal West Femto Maximum Sensitivity Substrate ( Life Technologies ) and Kodak BioMax XAR film ( Sigma-Aldrich ) . Cell cultures were fixed in 4% paraformaldehyde ( Electron Microscopy Sciences , Hatfield , PA ) . Adherent fixed cells were incubated in permeabilization buffer ( PBS , 0 . 25% Triton X-100 ) , washed with block solution ( PBS , 0 . 1% Triton X-100 , 1% bovine serum albumin ) , and probed with primary antibodies overnight at 4°C . Fixed EB cell cultures were cryoprotected in 30% sucrose-PBS , frozen in Tissue-Tek OCT ( Electron Microscopy Sciences ) , and stored at −80°C until use . 10 μM sections were prepared on a cryostat . Sections were washed with block solution , and probed with primary antibodies overnight at 4°C . The following primary antibodies were used: chicken anti-GFP antibody ( 1:1000 , Abcam , Cambridge , United Kingdom ) , rabbit anti-HB9 antibody ( 1:8000 ) ( Arber et al . , 1999c ) , rabbit anti-PTBP2 antibody nPTB-IS2 ( 1:1000 ) , mouse anti-Nestin antibody ( 1:250 , Developmental Studies Hybridoma Bank , Iowa City , IA ) , and mouse-anti TuJ1 antibody ( 1:1000 , Covance , Princeton , NJ ) . Samples were washed in PBS before incubation with Alexa-conjugated secondary antibodies ( Life Technologies ) for 2 hr at room temperature . Samples were rinsed in PBS and mounted with Prolong Gold AntiFade containing nuclear stain DAPI ( Life Technologies ) . Images were acquired using the Carl Zeiss Laser Scanning System LSM 510 META confocal microscope . Total RNA was collected from cell cultures using Trizol ( Life Technologies ) according to the manufacturer’s instructions . RNA was quantified ( A260 ) using a Nanodrop-1000 spectrophotometer ( Nanodrop Technologies/Thermo Fisher , San Jose , CA ) . Total RNA ( 0 . 5–1 μg ) was used for each sample in a 10 μL reaction with 0 . 25 μL of SuperScript III RT ( Life Technologies ) . PCR reactions contained 200 , 000–500 , 000 counts per minute of 32P-labeled or 0 . 2 μM FAM-labeled reverse primer . PCR reactions were run for 24 cycles for ESCs , 19 cycles for NPCs , and 20 cycles for MN cultures at an annealing temperature of 60°C . PCR reactions were mixed 1:2 with 95% formamide and loaded onto 8% polyacrylamide , 7 . 5 M urea gels . Gels were run under standard electrophoresis conditions , dried , and imaged on a Typhoon Imager ( GE Healthcare ) . Bands were quantified with ImageQuant software ( GE Healthcare ) . Real-time PCR was performed using SensiFAST SYBER Lo-ROX Kit ( Bioline , London , United Kingdom ) on a QuantStudio 6 Real-Time PCR System ( Life Technologies ) according to the manufacturers’ instructions . Relative mRNA levels were determined using a standard curve with beta-actin as a normalizing control . A list of primer sequences is available in Supplementary file 1 . Guide RNAs targeting Pbx1 intron 6 were designed using http://crispr . mit . edu/ and cloned into a modified pX330-U6-Chimeric_BB-CBh-hSpCas9 plasmid ( gift of B . Stahl and J . Doudna , University of California , Berkley ) according to the published protocol ( Ran et al . , 2013 ) . To generate Cas9-targeted deletions , HB9-GFP ESCs were transfected with two modified pX330 constructs and pBABE-puro ( Addgene , Cambridge , MA; #42230 ) using BioT ( Bioland Scientific , Paramount , CA ) according to the manufacturer’s recommendations . Transfected cells were then treated with 0 . 5 ug/ml puromycin ( InVivoGen , San Diego , CA ) . Following puromycin selection , clonal ESC lines were isolated and genotyped for intron 6 deletions . Some intron 6 deletions using guides 1 and 4 resulted in the loss of splice sites needed for exon 7 splicing . After genotyping to identify intron 6 deletions , ESC clones were subjected to secondary tests to confirm exon 7 splicing . A list of guide RNA and genotyping primer sequences is in Supplementary file 1 . 46C ESCs and NPCs were cross-linked at 100 mJ/cm2 . Cell pellets were collected and flash-frozen for iCLIP library preparation ( König et al . , 2010 ) . Modifications to the protocol are described in Supplementary file 2 . Flag and PTBP1 libraries were subjected to 100 bp single-end sequencing at the UCLA Broad Stem Cell center core facility ( Illumina HiSeq2000 ) . Data analyses were performed according to ( König et al . , 2010 ) , with minor modifications . Briefly , PCR duplicates were removed by comparing random portions of the sequenced barcodes . Unique reads were mapped to the mouse genome ( mm9/NCBI37 ) using Bowtie , allowing up to two nucleotide mismatches ( Langmead et al . , 2009c ) . Mapped reads were further mapped to the longest transcripts in Known Gene table ( Hsu et al . , 2006d ) . Crosslink sites were defined as the nucleotide upstream of each iCLIP read , and the False Discovery Rate ( FDR ) calculated according to ( König et al . , 2010 ) , with significant crosslink sites ( FDR<0 . 01 ) used for clustering and downstream analyses . Significant crosslinking sites were extended 20 nucleotides on either side , and overlapping sites compiled into clusters . Genomic sequences 30 nucleotides upstream and downstream from each crosslink site were used for motif enrichment analyses . Z-scores for all pentamer motifs were calculated by comparing motif frequencies relative to randomly chosen intervals from the same introns . Cassette exons were evaluated for PTBP1 clusters containing ≥ four significant reads , and defined as likely targets if a cluster was present within the cassette exon or within the intron sequence 500 nt upstream or downstream . To identify splicing changes during neuronal differentiation , polyA-plus RNA was isolated from HB9-GFP ESCs , NPCs , and Day 5 GFP+ MNs . Paired-end libraries were constructed using the TruSeq mRNA Library Prep Kit ( Illumina , San Diego , CA ) with modifications to generate strand-specific libraries ( Li et al . , 2014 ) . The libraries were subjected to 100 nt paired-end sequencing at the UCLA Broad Stem Cell center core facility ( Illumina HiSeq2000 ) . Data were analyzed using the Cufflinks pipeline to generate 200–250 million mapped reads per sample ( Trapnell et al . , 2012 ) . Alternative exon inclusion levels were determined by mapping to exon duo and trio databases using SpliceTrap ( Wu et al . , 2011 ) . Gene Ontology analyses were performed using DAVID ( Huang et al . , 2009a ) . To identify splicing changes following PTB protein depletion , polyA-plus RNA was isolated from 46C ESCs treated with siControl , siPtbp1 , or siPtbp1/2 ( n=3 ) ; and 46C NPCs treated with siControl , siPtbp1 , siPtbp1/2 , or siPtbp2 ( n=2 ) . Paired-end , strand-specific libraries were constructed using the TruSeq Stranded mRNA Library Prep Kit ( Illumina ) . The libraries were subjected to 50 nt paired-end sequencing ( Illumina HiSeq2000 ) to generate 30 to 50 million mapped reads per replicate . The replicates were then pooled for SpliceTrap analyses ( Wu et al . , 2011 ) . To identify gene expression changes with Pbx1a induction , polyA-plus RNA was isolated from HB9-GFP Day 2 MNs differentiated from wild type ( n=3 ) and I6 +/- ( n=5 ) ESC clones . Sequencing libraries were constructed using the TruSeq Stranded mRNA Library Prep Kit ( Illumina ) and subjected to 50 nt single-end sequencing ( Illumina HiSeq2000 ) to generate 25 to 40 million mapped reads per replicate . Differential gene expression was calculated using the Cufflinks pipeline ( q-value < 0 . 05 ) . Normalized FPKM values were calculated by Cuffnorm and were used to perform a Welch’s t-test ( P-value <0 . 05 ) . Gene Ontology analyses were performed using DAVID ( Huang et al . , 2009a ) . Differentially expressed genes were evaluated for Pbx1 ChIP clusters ( Penkov et al . , 2013b ) . Pbx1 binding was determined if a ChIP cluster was present within a window 1 kb upstream and downstream of the gene locus . HB9-GFP EBs were collected 2 days after RA and SAG addition , and dissociated in 0 . 25% Trypsin ( Life Technologies ) for 10 min . Cell pellets were collected and flash-frozen for ChIP-qPCR . Cells were cross-linked using 1% methanol-free formaldehyde , and chromatin was isolated using a previously published protocol ( Schjerven et al . , 2013c ) . ChIP was performed using antibodies against Pbx1 ( Cell Signaling ) , Meis1 ( AbCam ) , and Prep1/Pknox1 ( Thermo Fisher ) . Primers for qPCR are provided in Supplementary file 3 . Datasets are submitted to GEO with Accession Number GSE71179 , downloadable at: http://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? token=stmpwqeyppetvkt&acc=GSE71179 A list of Genome Browser sessions is provided in Supplementary file 4 . | The neurons that transmit information around the nervous system develop in several stages . Embryonic stem cells specialize to form neuronal progenitor cells , which then develop into neurons . These cell types have different characteristics , in part because they make different proteins or different versions of the same proteins . To make a protein , the DNA sequence of a gene is used to build a molecule of ribonucleic acid ( RNA ) that acts as a template for the protein . However , not all of this sequence codes for the protein . The non-coding regions must be removed from the RNA , and the remaining “exons” joined together to form the final “mRNA” template . Not all of the exons are necessarily included in the final mRNA molecule . By joining together different combinations of exons , several different versions of a protein can be produced from a single gene . This process is known as alternative splicing . One way that alternative splicing is controlled is through proteins that bind to RNA and determine which exons are included or excluded from the final mRNA molecule . PTBP1 is an RNA-binding protein that controls alternative splicing in embryonic stem cells and neuronal progenitor cells . Embryonic stem cells have the ability to develop into all the cells of the body . In contrast , neuronal progenitor cells are restricted in their development and only give rise to specialized cells of the nervous system . The role of PTBP1 in these properties was not clear . Linares et al . have now used a range of techniques to study the RNA molecules produced in these two cell types and how these RNAs change when PTBP1 is removed . This identified many RNAs whose splicing is regulated by PTBP1 , including mRNAs of the gene that produces a protein called Pbx1 , which is an important regulator of neuronal development . Further investigation revealed that PTBP1 prevents a particular exon being included in the mRNA template for Pbx1 . This creates an embryonic stem cell form of Pbx1 that does not affect neuronal genes . Removal of PTBP1 allows splicing of the Pbx1 exon and produces a version of Pbx1 that is found in neuronal progenitor cells and which turns on neuronal genes . Thus , through its action on Pbx1 , one role of PTBP1 is to enable stem cells to maintain their non-neuronal properties and prevent their premature development into neuronal progenitor cells . The gene for Pbx1 is only one of many genes controlled by PTBP1 at the level of splicing . One challenge for the future will be to understand how these genes work together in a common program that determines the properties of stem cells . Another question regards how the different Pbx1 proteins in stem cells and in neuronal progenitors can exert different effects in the cells where they are made . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"developmental",
"biology"
] | 2015 | The splicing regulator PTBP1 controls the activity of the transcription factor Pbx1 during neuronal differentiation |
Intraflagellar transport ( IFT ) is a highly conserved mechanism for motor-driven transport of cargo within cilia , but how this cargo is selectively transported to cilia is unclear . WDR35/IFT121 is a component of the IFT-A complex best known for its role in ciliary retrograde transport . In the absence of WDR35 , small mutant cilia form but fail to enrich in diverse classes of ciliary membrane proteins . In Wdr35 mouse mutants , the non-core IFT-A components are degraded and core components accumulate at the ciliary base . We reveal deep sequence homology of WDR35 and other IFT-A subunits to α and ß′ COPI coatomer subunits and demonstrate an accumulation of ‘coat-less’ vesicles that fail to fuse with Wdr35 mutant cilia . We determine that recombinant non-core IFT-As can bind directly to lipids and provide the first in situ evidence of a novel coat function for WDR35 , likely with other IFT-A proteins , in delivering ciliary membrane cargo necessary for cilia elongation .
The primary cilium is a highly specialized sensory organelle and signaling hub compartmentalized from the rest of the cell and positioned with a unique interface towards the extracellular environment . Analogous to a cell’s antenna , many roles for cilia have emerged in development , disease , and homeostasis ( Reiter and Leroux , 2017 ) . Enrichment of signaling receptors and effectors in ciliary membranes is critical for cilia function , yet all biosynthesis of cilia-localized membrane proteins occurs in the endoplasmic reticulum and is sorted by the Golgi and vesicular membrane traffic system to efficiently route cargo-laden vesicles for incorporation into the elongating ciliary membrane . The details of this highly efficient , directed transport process for the delivery of diverse cargos to cilia remain unclear . In mammalian cells , electron microscopy ( EM ) studies reveal the Golgi stacks closely apposed to the mother centriole ( Sorokin , 1962; Wheatley , 1969 ) . During intracellular ciliogenesis , small vesicles are recruited , most likely from the Golgi , to the mother centriole , where they fuse to form a large preciliary vesicle ( PCV ) attached at the distal appendages ( Yee and Reiter , 2015 ) . More secondary vesicles later fuse with the PCV , allowing elongation of cilia . Interestingly , the Golgi remains close to mature cilia , suggesting a continuous exchange of materials , enabling cilia maintenance ( Sorokin , 1962; Wheatley , 1969 ) . Several proteins essential for ciliogenesis localize to both the Golgi and the mother centriole and are implicated in this early stage of ciliogenesis including CCDC41 ( CEP83 ) , IFT20 , HOOK2 , and CEP164 ( Baron Gaillard et al . , 2011; Follit et al . , 2006; Graser et al . , 2007; Joo et al . , 2013; Schmidt et al . , 2012; Tanos et al . , 2013 ) . In some cases , including HOOK2 and CEP164 , these components recruit Rab8a and Rabin-8 , which facilitate membrane transport to cilia ( Baron Gaillard et al . , 2011; Moritz et al . , 2001; Nachury et al . , 2007 ) . For some specific ciliary cargos , including rhodopsin ( Wang and Deretic , 2014 ) and PKD2 ( Follit et al . , 2008; Follit et al . , 2006; Hoffmeister et al . , 2011; Kim et al . , 2014; Noda et al . , 2016 ) , Golgi-to-cilia transport mechanisms have been described . However , these processes seem to involve cargo-specific traffic modules . A more universal Golgi-to-cilia transport machinery , if one exists , has yet to be identified . In contrast to traffic to cilia , movement of cargos within cilia requires highly conserved motor-driven macromolecular cargo binding complexes that traffic along axonemal microtubules closely apposed against the ciliary membrane , in a process known as intraflagellar transport ( IFT ) ( Cole , 2009; Kozminski et al . , 1993; Pazour et al . , 1998; Pigino et al . , 2009; Rogowski et al . , 2013; Rosenbaum and Witman , 2002 ) . Bidirectional movement of IFT complexes regulates cilia content; the IFT-B complex aids in kinesin-dependent anterograde transport of cargo , whilst the IFT-A complex is required for retrograde transport driven by dynein motors ( Blacque et al . , 2006; Efimenko et al . , 2006; Jonassen et al . , 2012; Lee et al . , 2008; Piperno et al . , 1998; Tran et al . , 2008; Tsao and Gorovsky , 2008 ) . The IFT-A complex is composed of three core ( IFT144/WDR19 , IFT140/WDTC2 , IFT122/WDR10 ) and three non-core proteins ( IFT139/TTC21B/THM1 , IFT121/WDR35 , and IFT43 ) ( Behal et al . , 2012; Hirano et al . , 2017; Piperno et al . , 1998 ) . However , beyond classical retrograde ciliary traffic defects ( an inappropriate accumulation of cargos within the cilium ) mutations in IFT144 , IFT140 , IFT122 , IFT121/WDR35 , and IFT43 result in either severe reduction in cilia length or complete loss of cilia , implying they also have critical roles in transport of cargo to cilia ( Avidor-Reiss et al . , 2004; Caparrós-Martín et al . , 2015; Duran et al . , 2017; Hirano et al . , 2017; Liem et al . , 2012; Mill et al . , 2011; Takahara et al . , 2018; Zhu et al . , 2017 ) . Indeed , several IFT-A mutants fail to localize a range of ciliary membrane proteins including EVC1/2 , SMO , ARL13B , INPP5E , and SSTR3 to cilia ( Brear et al . , 2014; Caparrós-Martín et al . , 2015; Fu et al . , 2016; Hirano et al . , 2017; Jensen et al . , 2010; Lee et al . , 2008; Liem et al . , 2012; Mukhopadhyay et al . , 2010; Takahara et al . , 2018 ) . However , the mechanism of transport and the location of any IFT-A extra-ciliary function remain unclear . The movement of cargos between membranes of spatially separated organelles in the cytoplasm involves vesicular traffic . Indeed , IFT proteins have been observed to localize to various endomembranes and vesicular compartments outside cilia . For example , the IFT-B protein IFT20 localizes to the Golgi ( Follit et al . , 2006; Noda et al . , 2016 ) , whereas both IFT-B ( IFT20 , IFT27 , IFT46 , IFT52 , IFT57 , IFT88 , and IFT172 ) and IFT-A proteins ( IFT139 , IFT140 ) cluster around periciliary vesicles , shown by immuno-EM and light microscopy ( Sedmak and Wolfrum , 2010; Wood et al . , 2012; Wood and Rosenbaum , 2014 ) . Direct interaction of IFTs with membranes in vitro has also been described where the adaptor TULP3 and phosphoinositides mediate the membrane association of IFT-As ( Mukhopadhyay et al . , 2010 ) . More recently , purified IFT172 was shown to bind to lipids and pinch off smaller vesicles , similar in size to classic COPI vesicles ( Wang et al . , 2018 ) . It has been postulated that IFT proteins have evolved from membrane traffic coat complexes: soluble multimeric protein complexes that ‘coat’ donor membranes , facilitating cargo enrichment and membrane remodeling prior to traffic and fusion with target membranes ( Jékely and Arendt , 2006; van Dam et al . , 2013 ) . Nonetheless , a functional requirement for an IFT-dependent vesicle-to-cilia traffic module and what its dynamic architecture may resemble is currently unknown . To dissect how traffic of newly synthesized ciliary membrane proteins to the cilium occurs , we undertook a series of biochemical and imaging experiments in Wdr35 null mouse embryonic fibroblasts ( MEFs ) ( Caparrós-Martín et al . , 2015; Mill et al . , 2011 ) . To distinguish extra-ciliary functions from canonical retrograde traffic defects , we compared Wdr35-/- phenotypes with those of the retrograde IFT dynein Dync2h1-/- ( Criswell et al . , 1996; Huangfu and Anderson , 2005; Porter et al . , 1999; Signor et al . , 1999 ) . Whilst accumulations of intact IFT-B proteins were observed inside cilia in both mutants , only in the absence of WDR35 does the IFT-A holocomplex fragment and fail to enter Wdr35-/- cilia . Without intact IFT-A , we observe broad defects in the ciliary import of diverse membrane and lipidated proteins , as well as an accumulation of ‘coat-less’ vesicles around the base of Wdr35 mutants , which fail to fuse with the ciliary sheath . We demonstrate that together recombinant non-core IFT-A proteins ( WDR35 , IFT43 , and IFT139 ) are sufficient to specifically bind lipids in vitro . Together with our localization data , our results provide the first in situ evidence of a WDR35-dependent coat required to deliver essential cargo from vesicles to cilia .
We utilized primary MEFs carrying null mutations in two components of the retrograde IFT machinery ( Figure 1E ) , one part of the motor complex that moves cargos ( the retrograde dynein heavy chain Dync2h1 ) , and the non-core IFT-A component Wdr35 , in order to dissect the stage at which ciliogenesis defects occurred ( Caparrós-Martín et al . , 2015; Mill et al . , 2011 ) . Cilia length measured by acetylated α-tubulin staining was drastically reduced in both Wdr35-/- ( 0 . 48 µm mean ± 0 . 35 SD ) and Dync2h1-/- ( 0 . 76 µm mean ± 0 . 35 SD ) mutants compared to wild type ( WT ) ( 2 µm mean ± 0 . 45 SD ) MEFs ( Figure 1A and B ) . Given there was no reduction in cilia number ( Figure 1C ) , as previously shown ( Fu et al . , 2016; Liem et al . , 2012; Mukhopadhyay et al . , 2010 ) , our results suggest that DYNC2H1 and WDR35 are needed for cilia elongation at later stages of ciliogenesis . Defects in centriolar satellite traffic , implicated in ciliogenesis , were previously reported for WDR35 mutant RPE-1 cells ( Fu et al . , 2016 ) ; however , we saw no difference in levels or localization of endogenously tagged PCM1 protein ( PCM1-SNAP ) , which marks centriolar satellites in MEFs ( Figure 1—figure supplement 1 , Figure 1—video 1 ) . In Caenorhabditis elegans non-core IFT-A mutants , extension of the MKS module into the axoneme from the transition zone due to failure of cargo retrieval had been reported ( Scheidel and Blacque , 2018 ) . However , we observed intact transition zone modules as shown by NPHP1 and MKS1 localization in both mammalian mutants ( Figure 1D and E ) . We noted that Wdr35-/- axonemes , while acetylated , were not polyglutamylated , suggesting differences in tubulin post-translational modifications ( PTMs ) and stability ( Figure 1D ) . Together , these data suggest that the initial steps of ciliogenesis occur in both Dync2h1-/- and Wdr35-/- mutants; however , subsequent axoneme elongation may be differentially affected . Axoneme elongation during cilia assembly requires the import of key building blocks from their place of synthesis in the cell body into the cilium across the transition zone via IFT . We focused first on the anterograde , IFT-B machinery , monitoring two subunits IFT81 and IFT88 . We found that IFT-B complex proteins have similar retrograde traffic defects in both Wdr35-/- and Dync2h1-/- cells ( Figure 2A and B ) , accumulating beyond the length of the acetylated axoneme . We next looked to see if IFT-B complex assembly is disturbed in the absence of WDR35 by immunoprecipitation ( IP ) of endogenous IFT88 , followed by mass spectrometry ( MS ) to identify co-purifying subunits . IFT88 is the link between IFT-B1 and IFT-B2 complexes ( Figure 1E ) , interacting with IFT38 on the IFT-B2 side and IFT52 on the IFT-B1 side ( Katoh et al . , 2016; Mourão et al . , 2016; Taschner et al . , 2016 ) . MS analysis of immunoprecipitates from E11 . 5 Wdr35+/+ and Wdr35-/- embryo lysates revealed no statistically significant differences in stoichiometric composition of IFT-B complexes ( Figure 2C and D ) . We were able to isolate almost the entire IFT-B complex ( 14 out of 16 IFT-B components ) aside from IFT70 , which is not yet reported in mouse as well as IFT25 , which binds IFT27 to form a heterodimer ( Bhogaraju et al . , 2011; Funabashi et al . , 2017; Katoh et al . , 2016; Wang et al . , 2009 ) and is necessary for Hh signaling ( Keady et al . , 2012 ) . Because the composition of the IFT-B complex and its ability to enter cilia each appear unaltered , we conclude that exit from cilia is impaired in the absence of WDR35 . We next examined the composition of the IFT-A holocomplex in WT vs . Wdr35-/- embryos by IP of endogenous IFT-A core protein IFT-140 and its interactors . Whilst IFT140 immunoprecipitated all six components of the IFT-A complex from Wdr35+/+ embryo lysates , in Wdr35-/- samples both non-core components IFT139 and IFT43 were missing from our MS datasets ( Figure 3A ) . Their absence was confirmed by immunoblotting ( Figure 3B ) . Moreover , core components IFT122 and IFT144 were also significantly reduced in the purified mutant complex ( Figure 3A ) , suggesting that WDR35 is critical for the stability of the IFT-A complex and its components . We also compared total cellular levels of IFT-A component proteins and found that IFT139 and IFT43 levels were also undetectable on blots with lysates from both Wdr35-/- embryos ( Figure 3C and D ) and MEFs ( Figure 3—figure supplement 1A and B ) . This suggests that WDR35 is not only critical for the formation of stable IFT-A holocomplex but is also required for stability of its non-core components . In contrast , the individual core components of the IFT-A complex were nearly equally expressed in WT and Wdr35-/- lysates , except for IFT122 , which had higher expression levels in Wdr35-/- MEFs ( Figure 3C and D , Figure 3—figure supplement 1A and B ) . Other core components have been shown to have higher levels in the absence of WDR35 in human fibroblasts ( Duran et al . , 2017 ) . Thus , our work also supports previous studies demonstrating a level of interdependence in the levels of IFT-A subunits , which might be required for their coordinated function ( Behal and Cole , 2013; Duran et al . , 2017; Fu et al . , 2016; Picariello et al . , 2019; Zhu et al . , 2017 ) . These results suggest that WDR35 might be a link between IFT-A core and non-core proteins , which when absent results in the decreased abundance of IFT-A non-core subunits . To further distinguish between increased protein degradation from transcriptional changes , control and mutant MEFs were treated with the proteasome inhibitor MG-132 ( 20 µM ) ( Figure 3E ) . Treated cells displayed increased levels of IFT43 , which suggests that in the absence of WDR35 , non-core proteins may be targeted by the proteasomal degradation pathway . Interestingly , IFT139 and IFT121 are degraded in IFT43 null cells and both are rescued similarly by MG-132 treatment ( Zhu et al . , 2017 ) , confirming that the stability of IFT-A complex proteins is interdependent . We next looked at the localization and levels of the IFT-A components by immunofluorescence . IFT-A components were present in Dync2h1-/- cilia , suggesting that entry of IFT-A holocomplexes is not affected , but return from the distal tip is compromised in the absence of the dynein motor ( Figure 3F , Figure 3—figure supplement 1C ) . In contrast , in Wdr35-/- MEFs , IFT-A core components fail to enter cilia and remain restricted at the ciliary base ( Figure 3F ) , as shown by the difference in length covered by IFT-A components relative to cilia length measured by acetylated tubulin staining ( Figure 3—figure supplement 1C ) , whereas non-core proteins were undetectable , consistent with degradation ( Figure 3C–F ) . These results are consistent with previous reports of the interdependence of IFT-A components for ciliary localization . IFT140 is decreased in cilia of IFT122 mutants in mouse and fly ( Lee et al . , 2008; Qin et al . , 2011 ) , IFT139 is reduced in the flagella of Chlamydomonas with IFT144 mutation ( Iomini et al . , 2009 ) , and IFT144 fails to get recruited into cilia in WDR35-/- RPE cells ( Fu et al . , 2016 ) . IFT-A proteins require CPLANE chaperones for holocomplex assembly and cilia entry ( Toriyama et al . , 2016 ) . In all cases , failure of IFT-A holocomplex integrity impairs its recruitment into the cilia axoneme . Recent cryo-EM work had suggested that IFT-A is being carried by IFT-B trains inside the Chlamydomonas flagella in WT cells and these structures are missing in the IFT139 mutant ( Jordan et al . , 2018 ) . Our work in the mammalian system in the absence of WDR35 has a similar effect with IFT-B proteins accumulating inside the cilium whilst IFT-A core proteins accumulate proximal to the cilia base , and the non-core components are degraded . Cilia membrane protein cargos are synthesized in the cell body ( rough ER ) and traffic into cilia through a variety of direct and indirect routes . These include lateral diffusion from the plasma membrane ( Leaf and Von Zastrow , 2015; Milenkovic et al . , 2009 ) , recycling of plasma membrane proteins via the endocytic pathway ( Boehlke et al . , 2010 ) , as well as more directly from Golgi-derived vesicles ( Follit et al . , 2008; Follit et al . , 2006; Kim et al . , 2014; Witzgall , 2018 ) . Moreover , ciliary membrane content is dynamically regulated in response to external signals . First , we tested appropriate dynamic localization of the GPCR Smoothened ( SMO ) , which is recruited to the cilia in response to Hh ligand ( Figure 4A ) . SMO is already present in Dync2h1-/- mutant cilia , even in the absence of Hh . In contrast , even in the presence of Hh activation , SMO fails to enter Wdr35-/- cilia . We investigated endogenous levels and localizations of membrane-associated GTPases ARL13B and ARL3 , which are enriched in cilia in control cells ( Figure 4A ) . We saw that while they accumulate in excess in Dync2h1-/- mutants as per a retrograde defect , strikingly they fail to be recruited into Wdr35-/- cilia . Detecting low levels of endogenous protein localization and their mislocalization in Wdr35 mutants by immunofluorescence can be challenging . To overcome this , we transiently expressed membrane cargos , including fluorescently tagged SMO and ARL13B ( Figure 4B , Figure 4—video 1 ) , which effectively traffic into the cilia of WT cells . However , they fail to localize to Wdr35-/- cilia , with some accumulation at the cilia base . Interestingly , in our Wdr35-/- cells , SMO was predominantly localized to vesicles in the cytoplasm of mutant cells , whereas overexpressed ARL13B when not transported into cilia , is concentrated on other membranes , particularly the plasma membrane and pericentrosomal vesicles ( Figure 4B , Figure 4—video 1 ) . In trypanosomes , localization of flagellar membrane proteins was shown to be dependent on lipid rafts highly enriched in axonemes ( Tyler et al . , 2009 ) . Here , dual acylation was shown to direct potential association with lipid rafts , membrane microdomains that function as specialized platforms for protein/lipid transport and signaling . Indeed , ARL13B requires palmitoylation for its cilia membrane targeting and ciliogenesis in worms and mammals ( Cevik et al . , 2010; Li et al . , 2010; Roy et al . , 2017 ) , where it acts as the cilia-localized GEF for ARL3 , driving it to release lipid-modified cargos from carriers UNC119 and PDE6δ within cilia membranes ( Gotthardt et al . , 2015; Kapoor et al . , 2015 ) . As ARL13B and ARL3 fail to localize to mutant cilia , we next asked about the ability to recruit general lipidated cargo in Wdr35-/- MEFs . We examined the localization of lipidated motifs tagged to EGFP ( Williams et al . , 2014 ) to look at specialized membrane microdomains . In WT MEFs , untagged EGFP is present in the cell , but not in the cilium . When tagged with either myristoylation and palmitoylation ( MyrPalm ) or dual palmitoylation ( PalmPalm ) motifs , EGFP robustly enriches within cilia ( Figure 4C ) . We observed no enrichment of dual geranylation ( GerGer ) modified EGFP within control fibroblast primary cilia ( data not shown ) in contrast to the low-level expression previously reported in the most proximal portions of highly specialized olfactory sensory cilia ( Williams et al . , 2014 ) . This suggests that cell-type and cilia-specific differences exist . In marked contrast to WT cells , in Wdr35-/- MEFs , both the myristoylation and palmitoylation ( MyrPalm ) or dual palmitoylation ( PalmPalm ) EGFP failed to concentrate in mutant cilia ( Figure 4C ) . This failure to recruit lipidated cargos into Wdr35 mutant cilia is consistent with a more general traffic disruption of ciliary-destined membrane microdomains , containing broad categories of the membrane and membrane-associated cargos . It has previously been suggested that IFTs evolved from a protocoatomer ( Jékely and Arendt , 2006; Taschner et al . , 2012; van Dam et al . , 2013 ) . Three classic coat complexes ( COPI , COPII , and clathrin ) exist and perform similar functions but on different membranes and follow different routes through the cell . They are made of different protein components , which share a similar division of labor , characterized functionally as either adaptors or cage-forming proteins . Although components like the cage proteins share significant structural homology in organization of protein domains , they do not share detectable sequence homology ( Field et al . , 2011 ) . Given the defects in ciliary membrane content observed in the Wdr35 mutant cilia , we hypothesized that WDR35 , in collaboration with other IFT-A complex proteins , may be required for moving ciliary membrane cargos between donor membranes , such as the Golgi or endosomes , to their destination ciliary membrane , in a manner comparable to coat complexes . WD40 repeat ( WDR ) and tetratricopeptide repeat ( TPR ) motifs are common throughout cellular proteomes and are involved in a wide range of biological processes . Agnostic of structure , we used deep sequence analysis of the whole human proteome and homology modeling to ask which proteins were most similar to IFT-A components . Simple alignment strategies with proteins such as IFT subunits , which contain tandem repeat motifs , could erroneously align with other repeat proteins to suggest a close evolutionary relationship where none exists ( i . e . , false positives ) . To address this , we used four IFT-A subunits ( IFT144 , IFT140 , IFT122 , and IFT121 ) and two of IFT-B ( IFT80 and IFT172 ) as seed sequences for multiple iterative rounds of homology searches via profile-HMM alignment ( Remmert et al . , 2011 ) . We then clustered the resulting proteins based on sequence similarity , as previously described ( Wells et al . , 2017; Wells and Marsh , 2019 ) . This was repeated using the COP protein subunits as seeds for reverse analysis . Together , these reciprocal analyses revealed that out of the entire proteome COPI α and β′ cluster most closely with six IFT proteins ( two IFT-B and four IFT-A components ) , both having TPR and WD40 repeats ( Figure 5A ) . In contrast , homology searches with COPI β and COPI ϒ1/2 , which have HEAT/ARM repeats , did not yield any hits with IFT components , as was the case with COPI ε , which has TPRs but no WD40 domains . COPI δ and COPI ε1/2 , which have no identifiable repeat domains , are most closely related to adaptor protein complex subunits AP2 and AP3 . In summary , using multiple rounds of sequence homology searches , we generated a broad set of putatively related repeat proteins , clustering of which reveals clear relationships between coatomers and IFT-A/B complex components . Next , we used SWISS-MODEL ( Waterhouse et al . , 2018 ) to predict the structures of IFT-A proteins . COPI α ( COPA ) and β′ ( COPB2 ) structures were top hits with 12–15% sequence identity and 26–27% sequence similarity to four IFT-A complex proteins ( IFT144 , IFT140 , IFT122 , and WDR35 ) . Based on the target-template alignment models , built using ProMod3 , ribbon diagrams of all four IFT-A subunits modeled structures with two N-terminal seven-bladed WD40 β propellers and C-terminal extended TPRs , also found in α and β′ COPI proteins ( Figure 5—figure supplement 1A ) , as previously modeled for WDR35 ( Mill et al . , 2011 ) . The remaining two IFT-A subunits were not possible to model accurately . IFT139 contains only TPRs with limited sequence similarity to the ε subunit of COPI coatomer ( van Dam et al . , 2013 ) . IFT43 is the smallest and unstructured protein and could not be modeled and is presumed to be made of α-helices ( Taschner et al . , 2012 ) . While undertaking this work , a crystal structure for IFT80 was published , highlighting that despite the same domain organization IFT80 adopted a distinctive 3D conformation of the second β-propeller domain from β′-COP and also formed a dimer unlike the triskelion COPI cage ( Taschner et al . , 2018 ) . However , purified IFT172 adopted two configurations by negative stain EM when incubated with and without lipids ( Wang et al . , 2018 ) . Thus , respecting the limitations of homology modeling without solved structures , we found four IFT-A proteins ( IFT144 , IFT140 , IFT122 , and IFT121 ) to have very high sequence and structural similarities to COPI α and β′ subunits with N-terminal WD40 repeats and C-terminal TPR region ( Figure 5—figure supplement 1A ) . Given the structural homology of WDR35 and IFT-As to COPI proteins , which derive vesicles from the Golgi , we asked whether WDR35 and IFT-As were sufficient to directly bind membranes . To test if the IFT-A complex directly associates with lipids in vitro , we purified recombinantly expressed IFT-A non-core complex ( IFT139/121/43 ) as well as the dimeric IFT121/43 and the isolated IFT43 subunit of the unicellular organisms Chlamydomonas reinhardtii using eukaryotic expression systems ( Figure 5B and C , Figure 5—figure supplement 1B–D ) . All three samples were soluble , eluted as stoichiometric proteins from size-exclusion chromatography ( SEC ) , and were positively identified by MS . The heterotrimeric IFT-A complex purified from mammalian cells was assessed for lipid binding using membrane lipid strips , detecting any bound protein complex using antibodies against the His-Tag on IFT43 . From the protein-lipid overlay results in Figure 5C , the His-GFP-tagged IFT-A trimeric complex displays strong binding to phosphatidic acid ( PA ) as well as weaker binding to phosphatidylserine ( PS ) . Thus the IFT-A trimeric complex binds to these negatively charged ( anionic ) phosphoglycerates exclusively , without binding to neutral or inositol-based lipids as had been reported for the IFT-A adaptor TULP3 ( Mukhopadhyay et al . , 2010 ) . Although there are no reports that PA is a constitutive phospholipid of Golgi apparatus in C . reinhardtii , it was shown to be the third most abundant phospholipid in cilia ( Lechtreck et al . , 2013 ) . As a low abundance phospholipid , PA is known to play both important structural roles facilitating membrane curvature during vesicle fusion and fission events ( Arisz and Munnik , 2011; Zhukovsky et al . , 2019 ) as well as signaling functions such as flagellar excision in response to environmental stresses ( Goedhart and Gadella , 2004; Lechtreck et al . , 2013; Quarmby et al . , 1992 ) . To further investigate which subunit of the IFT-A non-core complex is responsible for lipid binding , His-tagged IFT121/43 and IFT43 were also tested in the lipid-strip assay ( Figure 5—figure supplement 1B-D ) . Neither the IFT121/43 dimer nor IFT43 alone showed detectable lipid binding , demonstrating that the IFT139 subunit is essential for lipid interaction by the non-core IFT-A complex . In order to further test whether the trimeric IFT-A complex was capable of specifically binding to PA-containing liposomes , we performed negative stain EM of purified proteins incubated with liposomes composed of PE/PG/PA or POPC ( control ) ( Figure 5D ) . The IFT-A trimer was observed to associate with PE/PG/PA liposomes but not to control liposomes ( Figure 5D ) . Consistent with the lipid overlay assay ( Figure 5—figure supplement 1B-D ) , the IFT121/IFT43 dimer displays only weak association with PE/PG/PA liposomes ( Figure 5D ) . The structural homology of IFT-As to COPI proteins and the ability of the non-core IFT-A complex to bind directly to lipids in vitro led us to ask whether IFT-A complex may function similarly to COPI vesicle coats assisting vesicular transport between the Golgi and cilia in vivo . We undertook ultrastructural studies to examine traffic phenotypes with higher resolution around cilia in MEFs . In all genotypes , ciliation was observed to start very close to the nucleus and remain close to the Golgi stacks throughout cilia elongation ( Figure 6A , Figure 6—video 1 , Figure 6B , Figure 6—video 2 , Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 2 , Figure 6—video 4 ) . In control MEFs , even after 24 hr of serum starvation , very few ( ~1% ) cilia were observed to emerge from the cell , highlighting the deep-seated ciliary pocket in MEFs ( Figure 6B , Figure 6—video 2 , Figure 6—figure supplements 1 and 4A ) , and as described for RPE-1 cells ( Molla-Herman et al . , 2010 ) . In control MEFs , polymerized microtubules formed a well-structured axoneme ( Figure 6B , Figure 6—video 2 , Figure 6—figure supplements 1 and 4A ) as previously described in MEFs ( Rogowski et al . , 2013 ) and reported in other primary cilia ( Kiesel et al . , 2020; Molla-Herman et al . , 2010 ) . Additionally , microtubules can be seen attached at the cilia base and radiating in different directions in the cell ( Figure 6—figure supplement 1 ) . In contrast to the well-defined ciliary membrane and well-polymerized microtubules of the control axoneme , Wdr35-/- cilia have ‘wavy’ membranes and disorganized microtubules ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 4B ) . Mammalian Dync2h1-/- mutants retained a well-defined ciliary membrane and an apparently well-structured axoneme present throughout ( Figure 6—figure supplement 4C , Figure 6—video 5 ) , similar to previous reports of the fla14 dynein mutant in Chlamydomonas ( Pigino et al . , 2009 ) . Stacked standing trains with a periodicity of 40 nm were reported in fla-14 mutants ( Pigino et al . , 2009; Stepanek and Pigino , 2016 ) , and in our Dync2h1-/- mutant axonemes , we observed similar stacking of stalled IFT trains with a periodicity of 40 nm , irrespective of the length of mutant cilia ( Figure 6—figure supplement 4C , Figure 6—video 5 , and Liem et al . , 2012 ) . Although IFT-Bs also accumulated in Wdr35-/- cilia ( Figure 2A and B ) , these stripes were not observed ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 4B ) , suggesting that both IFT-B and IFT-A are required to form the higher ordered IFT trains that stall in Dync2h1 mutants . We further tested our hypothesis that IFT-A acts as a coat-like complex for vesicles targeted to cilia by transmission electron microscopy ( TEM ) analysis of ciliated MEFs . We observed electron-dense-coated vesicles between the Golgi and cilia in WT MEFs ( Figure 6A , Figure 6—video 1 ) . We also observed these coated vesicles clustering at the cilia base ( Figure 6B , Figure 6—video 2 ) and bulging from ciliary pockets and ciliary sheaths in WT MEFs ( Figure 6—figure supplement 1 ) . These electron-dense vesicles around control cilia were more prominent at the early stage of ciliogenesis in EM ( Figure 6A , Figure 6—video 1 ) . In contrast , in Wdr35-/- mutant cells , there is a 10-fold increase in the average number of vesicles around the ciliary base ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 2 , Figure 6—video 4 , Figure 6—figure supplement 4B , Figure 7—figure supplement 1A; quantified in Figure 7B ) . Importantly , virtually all of these mutant vesicles lack the electron-dense coats observed in control cells ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 2 , Figure 6—video 4 , Figure 6—figure supplement 4B , Figure 7E , Figure 7—figure supplement 1B; quantified in Figure 7D ) . Notably , we did observe other electron-dense coats , likely clathrin , on budding vesicles at the plasma membrane in these same Wdr35 mutant cells , emphasizing that other coats are preserved in these conditions ( Figure 6—figure supplement 2 , Figure 6—video 4 , Figure 7E ) . Moreover , no difference in the density or distribution of periciliary clathrin-positive vesicles is observed around the base of Wdr35-/- mutant cilia ( Figure 7—figure supplement 1C and D ) . In contrast , the accumulation of coatless vesicles spreads in a volume ~2 µm3 around the Wdr35-/- ciliary base ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 2 , Figure 6—video 4; quantified in Figure 7B , Figure 7—figure supplement 1A ) . In spite of their proximity to the ciliary sheath and their abundance , fusion events were not observed in Wdr35-/- mutants ( Figure 6C , Figure 6—video 3 , Figure 6—figure supplement 2 , Figure 6—video 4; quantified in Figure 8D ) . We do not believe that this periciliary vesicle accumulation phenotype is a general defect in global membrane traffic as the accumulation of vesicles lacking electron densities occurs specifically around mutant cilia , and not at other regions of Wdr35 mutant cells ( Figure 6—figure supplement 3 , Figure 6—video 5 ) . Clathrin-mediated endocytosis at the ciliary pocket is proposed to regulate internalization of ligand/receptor complexes or membrane content at the base of cilia ( Molla-Herman et al . , 2010 ) . To test whether these vesicles might be important for the import or export of cargo directed to cilia , we analyzed Dync2h1-/- cilia , which we showed to contain increases in IFTs ( Figures 2 and 3 ) and membrane protein cargo ( Figure 4 ) in the absence of retrograde transport . Consistent with the redistribution of IFT pools from the base into the ciliary compartment ( Figure 2A and Figure 3F ) , we observed no vesicles at the base of Dync2h1-/- cilia ( Figure 6—video 6 , Figure 7; quantified in Figure 8D ) . Interestingly , ectosomes , which are previously reported to regulate the content of cilia in a variety of systems ( Cao et al . , 2015; Nager et al . , 2017; Wood and Rosenbaum , 2014 ) , budding from the tip were much more prevalent in Dync2h1-/- cilia than in WT cells ( Figure 6—figure supplement 4C , Figure 6—video 6 ) . We interpret these data as evidence that the coated vesicles around the WT cilia function to transport cargo possibly from the Golgi or via an endosomal intermediate to the cilia . In the absence of WDR35 , non-coated vesicles accumulate around the ciliary base , marking a failure in this process in either the formation and/or maintenance of this coat and subsequent fusion at the target ciliary pocket . To further confirm our hypothesis that these electron-dense vesicles directed to cilia contain WDR35 and IFT-A proteins , we performed correlative light and electron microscopy ( CLEM ) imaging in Wdr35-/- MEFs expressing WDR35-EmGFP , which we had previously shown to completely rescue cilia phenotypes ( Figure 1A and B , Figure 8A ) . Expressing WDR35-EmGFP in Wdr35-/- ensures that every WDR35 particle was labeled with EmGFP , minimizing competition with non-labeled species . Using Airyscan confocal imaging of WDR35-EmGFP MEFs grown on grids for subsequent TEM , we saw WDR35-EmGFP enriched at the ciliary base of rescued mutant cilia . Moreover , we observed that this signal coincided with the reappearance of electron-dense vesicles in the TEM images ( Figure 8A and B ) . We also observed recovery of fusion events of coated vesicles at the cilia base in cells expressing WDR35-EmGFP as well as rescue of the periciliary vesicle accumulation phenotype ( Figure 8B and C; quantified in Figure 8D , Figure 8—figure supplement 1 , Figure 8—video 1 ) . Next , we performed immunogold labeling directly on 70 nm sections and observed sparse but specific labeling of GFP-positive particles at the cilia base , within the axoneme and around putative vesicles at the cilia base and ciliary sheath ( Figure 8—figure supplements 2 and 3 ) . Together , these results demonstrate that WDR35 is required for the formation of these coated vesicles and that these coated vesicles coincided with WDR35-EmGFP signal , confirming that WDR35 supports the assembly of a novel coat on vesicles destined to deliver membrane cargos to cilia .
Vesicle coat proteins , with the archetypal members clathrin and the coat protein complexes I and II ( COPI and COPII , respectively ) , are macromolecular machines that play two central roles in the homeostasis of the cell’s endomembrane system . They enable vesicle formation and select protein and lipid cargo packaged for delivery from a specific donor to functionally segregated compartments . Given the deep sequence structural similarities between IFT-A and COPI subunits and the ability of the non-core IFT-A to bind directly to lipids in vitro , coupled to the phenotypic defects in Wdr35-/- cells ( including lack of ciliary enrichment of a broad range of membrane cargos and the absence of electron-densities on accumulated periciliary vesicles ) , we propose a novel function for WDR35 and other IFT-A proteins to act as a coat-like complex that is critical for the transport of ciliary membrane cargo into cilia . Two other macromolecular complexes have been proposed to form vesicle-associated coats involved in ciliary traffic: clathrin ( Kaplan et al . , 2010; Molla-Herman et al . , 2010 ) and the BBSome complex ( Jin et al . , 2010 ) . Clathrin is a classical vesicle coating protein with some documented activity at the ciliary pocket ( Clement et al . , 2013; Pedersen et al . , 2016 ) . From static images , the directionality of events is difficult to resolve: fission ( endocytosis ) or fusion ( exocytosis ) . Clathrin vesicles can be both endocytic , where they concentrate cargos and curve off donor membranes for selective transport into the cytoplasm , or exocytic , where they can use fuse to release their contents . For example , a subset of AP-1 clathrin vesicles were shown to traffic between the trans Golgi and basolateral membranes of polarized epithelial cells ( Fölsch et al . , 1999 ) via the recycling endosome compartment ( Futter et al . , 1998 ) . Indeed , in both C . elegans ( Bae et al . , 2006; Dwyer et al . , 1998; Kaplan et al . , 2010; Ou et al . , 2007 ) and trypanosomes ( Vince et al . , 2008 ) , deletion or depletion of AP-1 leads to defects in cilia assembly and protein traffic into cilia . However , in mammalian cells , depletion of clathrin and clathrin-associated proteins results in a normal number of cilia with normal lengths ( Kaplan et al . , 2010; Molla-Herman et al . , 2010 ) , as opposed to the drastically reduced size of Wdr35-/- cilia ( Caparrós-Martín et al . , 2015; Fu et al . , 2016; Mill et al . , 2011 ) . This suggests that clathrin is dispensable for vesicular transport into mammalian cilia . Although electron-dense vesicles were observed invaginating from the mammalian ciliary pocket , the electron density on these vesicular invaginations was unchanged in the absence of clathrin ( Molla-Herman et al . , 2010 ) . Using live cell imaging , the directionality of clathrin-mediated traffic was reported to be largely away from cilia ( Molla-Herman et al . , 2010 ) . Importantly , we still observe clathrin-coated endocytic structures on the plasma membrane of Wdr35-/- cells ( Figure 6—figure supplement 2 , Figure 6—video 4 ) , and we found no difference in the distribution of clathrin intensity in a volume of ~2 µm3 around the ciliary base in Wdr35-/- cilia compared to controls ( Figure 7—figure supplement 1C and D ) . Moreover , studies on clathrin-mediated exocytosis demonstrated that depletion of human clathrin heavy or light chains results in increased total fusion events with complete release of membrane cargos from vesicles in fibrosarcoma cells ( Jaiswal et al . , 2009 ) , the opposite to what is observed in Wdr35 mutants where vesicles stack up adjacent to the ciliary sheath but do not fuse . The BBSome is a macromolecular machine of Bardet–Biedl syndrome ( BBS ) proteins , which is also postulated to have evolved from an early ancestral coat complex ( Jékely and Arendt , 2006; van Dam et al . , 2013 ) . The BBSome shares similar structural elements to the archetypal coats and plays a role in cilia function ( Nachury , 2018 ) . In contrast to IFT , mutations in BBSome components , including ARL6/BBS3 , do not affect cilia assembly and length regulation ( Domire et al . , 2011; Eguether et al . , 2014; Lechtreck et al . , 2013; Lechtreck et al . , 2009; Liew et al . , 2014; Nager et al . , 2017; Shinde et al . , 2020; Xu et al . , 2015; Ye et al . , 2018 ) . Instead , they generally are required for regulating cilia content , mostly for the export of ciliary membrane proteins . Although this suggests that BBSomes regulate movement of ciliary components between compartments , endogenous localization of the BBSome remains unclear , without evidence supporting endomembrane or plasma membrane localization . In contrast , IFT20 localizes to the Golgi ( Follit et al . , 2006; Noda et al . , 2016 ) . Moreover , whilst there is in vitro evidence that BBSomes can cluster on liposomes , they do not deform membranes , a key step in vesicle formation by coatomers ( Jin et al . , 2010 ) . In contrast , purified IFT172 , an IFT-B component that is also homologous to COPI α and ß′ like WDR35 , can not only assemble on liposomes with high affinity but can also bud 50 nm vesicles consistent with coatomer-sized products ( Wang et al . , 2018 ) . We report here that the purified trimer of non-core IFT-A ( WDR35 , IFT43 , and IFT139 ) can also directly and specifically bind to lipids , notably PA , which is involved in membrane deformation in COPI maturation and exocytosis ( Yang et al . , 2008; Zeniou-Meyer et al . , 2007 ) . We are currently testing whether non-core IFT-A can also pinch off vesicles . Together , the evidence , including its evolutionary conservation of the BBSome with more classical coat proteins ( Jékely and Arendt , 2006; van Dam et al . , 2013 ) , interaction with in vitro membranes in the presence of the ARF-like GTPase ARL-6 , interaction with phospholipids ( Jin et al . , 2010; Nachury et al . , 2007 ) , and recent cryo-EM structures of the complex ( Chou et al . , 2019; Klink et al . , 2020; Singh et al . , 2020; Yang et al . , 2020 ) , suggests that the BBSome may be working as an adaptor for IFT-A-mediated cage formation , similar to other coat adaptors for clathrin ( i . e . , AP1/AP2 ) or COP ( i . e . , β- , γ- , δ- , and ζ-COP for COPI ) . Our data suggest that the electron density observed on vesicles around the ciliary base in control cells is neither clathrin nor BBSome in nature , and is likely composed of WDR35/IFT-A . Our study demonstrates a requirement for IFT-A to deliver ciliary membrane cargo into cilia , potentially by acting as a vesicle coat operating between the Golgi and the ciliary base . Archetypal coatomer protein complexes , including COPII , COPI , and clathrin , concentrate cargo within donor membranes and pinch off vesicles ( fission ) , which then travel to their target organelle membranes , where SNARE and Rab GTPases assist their fusion ( Bonifacino and Glick , 2004 ) . In these cases , the electron-dense coats are progressively dismantled such that uncoated vesicles can fuse with acceptor membranes , presumably to facilitate access to the fusion machinery , such as SNAREs , on the surface of the vesicle . As a result of interactions with cargo and lipids with the vesicles , there is evidence that the COPI coat can remain stable on membranes after fission . Moreover , this suggests that COPI vesicle uncoating may be incomplete , such that residual COPI on the vesicle surface enables vesicle recognition and tethering necessary for fusion to the correct acceptor membrane ( Orci et al . , 1998 ) . In contrast to the trail of electron-dense vesicles between the Golgi and the base of cilia in control cells , we observed 10 times more vesicles stalled around the cilia base of Wdr35-/- MEFs . These all lack an electron-dense coat , suggesting that these transport vesicles are formed but fail to fuse at the ciliary target membrane in the absence of WDR35 . This raises a question as to why a protein like WDR35 , which shares structural homology to fission-inducing proteins , gives phenotypes consistent with a fusion-facilitating protein . One possibility is that while Wdr35-/- MEFs are missing one COPI α/β′- homolog , the other three core IFT-As ( IFT144 , IFT140 , and IFT122 ) may be sufficient to compensate by providing interaction motifs necessary for the fission of vesicles from donor membranes such as the Golgi . Indeed , we show IFT122 to be upregulated in Wdr35-/- mutant cells , similar to previous reports in WDR35 patient cells ( Duran et al . , 2017 ) . However , we and others have demonstrated that in the absence of WDR35 the IFT-A complex is unstable ( Zhu et al . , 2017 ) such that any core IFT-A coat on the vesicles from donor membranes such as the Golgi may be easily disassembled . It is interesting to note that non-core IFT139 and IFT43 are helical ( Taschner et al . , 2012 ) similar to SNARE proteins that mediate vesicle fusion with target membranes . Importantly , we show here that these components , which are degraded in the absence of WDR35 , could help mediate the fusion of vesicles with the ciliary pocket or base to transfer membrane cargos into the growing cilia sheath . Indeed , we show that purified non-core IFT-A complex is sufficient to specifically bind PA , which is present in ciliary membranes , as well as the Golgi and the recycling endosome compartment ( Farmer et al . , 2021; Lechtreck et al . , 2013; Yang et al . , 2008 ) . The lipid composition of membranes is known to determine their curvature ( McMahon and Boucrot , 2015 ) ; PA being conical in shape concentrates on more curved regions of membranes , resulting in nanoscopic-negative curvature such as found in the ciliary pocket ( Zhukovsky et al . , 2019 ) . Moreover , with a small head group , negative charge , and a phosphomonoester group , PA interacts with proteins and lipids in several subcellular compartments that facilitate fission and fusion of membranes ( Zhukovsky et al . , 2019 ) . From our liposome assay , we speculate that IFT139 binding to the IFT121/43 dimer increases the binding affinity to lipids . Indeed , on its own , the IFT-A dimer signal is below the threshold of detection in the protein-lipid overlay assays but observed to weakly associate to PA-containing liposomes . In our purification of the non-core IFT-A complex with the affinity tag on IFT43 , only IFT43/121 , rather than IFT43/139 , was co-purified together with the trimeric complex , indicating that IFT121 interacts with both IFT43 and IFT139 , and is responsible for mediating the interactions between IFT43 and IFT139 , which is consistent with what has been previously reported ( Behal et al . , 2012; Zhu et al . , 2017 ) . In the Wdr35-/- mutant , and likely IFT139 or IFT43 KO strains , the non-core IFT-A complex will not form , which leads to non-coated vesicles ( Figure 9 ) . Important next steps will be to systematically investigate vesicular traffic defects in other IFT-A mutants , as well as identify the GTPase that acts to drive formation , uncoating , and fusion of these vesicles . Recruitment , remodeling , and regulation of protein coats involve cycles of GTP hydrolysis , for example , ARF1 regulates recruitment to membranes of the COPI coatomer ( Dodonova et al . , 2017 ) . It is interesting to note that we and others have been unable to purify IFT-A complex with any GTPases ( Mukhopadhyay et al . , 2010 ) , suggesting that any interaction is transient . This is even in conditions where we can purify endogenous IFT-B complexes with its associated GTPases IFT22/RABL5 and IFT27/RABL4 . In COPI , recruitment of coat components to donor membranes starts with the insertion of small GTPase ARF1 into membranes ( Dodonova et al . , 2017 ) . So far only one ARF , ARF4 acting at the TGN ( Mazelova et al . , 2009; Wang et al . , 2017 ) , has been implicated in ciliary traffic . However , it plays non-ciliary roles and shows early lethality in mouse knockouts without affecting cilia assembly ( Follit et al . , 2014 ) . Mutations in several related ARLs have defects in cilia structure and/or content , including ARL3 , ARL6 , and ARL13B ( Alkanderi et al . , 2018; Cantagrel et al . , 2008; Fan et al . , 2004 ) . At least in the case of ARL13B and ARL3 , they fail to accumulate and/or enter mutant cilia , even when overexpressed in the absence of WDR35 , although periciliary vesicular staining can be observed . Rab GTPases have been implicated in the ciliary targeting of vesicular cargos ( Blacque et al . , 2018 ) . Notably , expression of dominant negative RAB8 in Xenopus photoreceptors ( Moritz et al . , 2001 ) results in a strikingly similar accumulation of vesicles to our Wdr35 mutants , which fail to fuse with the ciliary base . Similarly , in RPE-1 cells , dominant negative RAB8 impairs traffic of ciliary membrane cargos ( Nachury et al . , 2007 ) . However , functional redundancy between RABs may exist as neither single nor Rab8a;Rab8b double mutant mice have defects in cilia formation . On the other hand , defects in ciliation were observed when Rab10 was additionally knocked down in Rab8a;Rab8b double mutant cells ( Sato et al . , 2014 ) . Excitingly , our work demonstrates IFT-As to be important for the later stage of ciliogenesis , similar to GTPases like RAB23 ( Gerondopoulos et al . , 2019 ) or RSG-1 ( Agbu et al . , 2018; Toriyama et al . , 2016 ) . Given that these GTPases have also been shown to sequentially interact with CPLANE subunits INTU and FUZ , which are also required for IFT-A holocomplex assembly ( Gerondopoulos et al . , 2019; Toriyama et al . , 2016 ) , they will be priorities for future investigations . We have demonstrated that an IFT-A-dependent coat for membrane vesicles exists and is necessary for their fusion with the ciliary sheath and ciliary pocket , which is continuous with the ciliary membrane . We also showed that this coat is necessary to efficiently deliver cilia-destined signaling molecules into the elongating axoneme of the cilium . This raises the possibility that some of this IFT-A-dependent coat may remain upon vesicle fusion as a now linear ‘train’ carrying membrane cargos to be picked up by cytosolically assembling IFT-B particles allowing import across the transition zone and then anterograde IFT within the cilium ( Figure 9 , inset B ) . Excitingly , we show that purified non-core IFT-A including WDR35 is sufficient to bind selectively to PA . This low abundance signaling lipid has well-described roles in vesicle traffic where it promotes COPI vesicle fission in the Golgi ( Yang et al . , 2008 ) , maintenance of the endosome recycling compartment ( ERC ) ( Farmer et al . , 2021 ) , as well as promoting exocytosis through formation of fusion-competent granules ( Zeniou-Meyer et al . , 2007 ) . Defining at which points in vesicular traffic IFT-A-dependent coats act , both fission and fusion , within cells as well as the biochemical nature of lipids and cargos these vesicles carry will be required . Given its efficacy , this IFT-dependent ‘targeted delivery’ module may also be repurposed for other non-ciliary membrane targeting events via polarized exocytosis . Notably in the immune synapse of T cells , where IFT20 is required for rapid clustering of TCRs necessary for T cell activation ( Finetti et al . , 2009 ) , as well as photoreceptor dendrites in which IFT localization to vesicles tracking towards the postsynaptic membranes was observed ( Sedmak and Wolfrum , 2010 ) , where dendritic exocytosis is implicated in synaptic plasticity and neuronal morphology ( Kennedy and Ehlers , 2011 ) . Future studies into this IFT-dependent coat complex and the membrane traffic processes it controls may expand our phenotypic understanding of the ciliopathies beyond the cilium .
Primary MEFs were harvested from E11 . 5 embryos and cultured in complete media ( Opti MEM-I [Gibco , 31985-047] supplemented with 10% fetal calf serum ( FCS ) and 1% penicillin-streptomycin [P/S] and 0 . 026 µl β-mercaptoethanol ) and incubated at 37℃ in a hypoxic incubator ( 3% O2 and 5% CO2 ) . To induce ciliogenesis , 70–80% confluent cells were serum-starved for 24 hr . Genotyping was done as described before for the Wdr35 line ( Mill et al . , 2011 ) and Dync2h1 line ( Caparrós-Martín et al . , 2015 ) . Pcm1-SNAP mouse line was made by Dr . Emma Hall ( Hall E . et al . , unpublished ) by endogenous tagging of PCM1 by CRISPR . Pcm1SNAP mouse line was crossed with Wdr35-/+ and genotyped to screen E11 . 5 embryos homozygous for both Wdr35-/- and Pcm1SNAP/SNAP . MEFs prepared from these embryos were used to image PCM1 localization in WT and Wdr35-/- using antibodies and other reagents listed in Table 1 and Table 2 . Cells were trypsinized to a single-cell suspension and resuspended in 10 µl Resuspension Buffer R per 0 . 5 × 105 cells/transfection reaction , mixed with plasmid DNA ( 0 . 75 µg/transfection ) ( Table 3 ) and electroporated ( voltage 1350 V , width 30 ms , one pulse ) using a Neon Nucleofection kit ( Thermo Fisher Scientific MPK-1096 ) , according to the manufacturer’s protocol . Transfected cells are harvested or visualized 24–48 hr post electroporation . Primary MEFs ( 0 . 5 × 105 cells/transfection ) were electroporated with ARL13B-EGFP or Smoothened-GFP using the Neon Transfection System , 10 µl kit ( Thermo Fisher Scientific , MPK-1096 ) and seeded in 24-well glass-bottomed plates ( Greiner Sensoplates , 662892 ) with prewarmed media ( Opti-MEMI [1×] [Gibco , 31985-047] , 10% FCS and 0 . 026 µl β-mercaptoethanol ) . Samples were incubated in antibiotic-free media 37℃/5% CO2/3% O2 overnight and then serum-starved for 24 hr . SiR-tubulin kit ( Spirochrome , SC002 ) , a 1 mM stock solution , was prepared in anhydrous DMSO and stored at –20℃ , without aliquoting . For staining , 1:5000 ( 200 nmol ) of SiR-tubulin stock was diluted in serum-free media and added to cells for 1 hr in the hypoxic incubator , then live imaged without washing . For live-cell PCM1 imaging , MEFs electroporated with ARL13B-EGFP were incubated with 1:1500 TMR-SNAP ( New England Biolabs , S9105S , stock 30 nmol ) in low serum media in the hypoxic incubator for 30 min . Cells were washed twice with low serum media for 1 hr each in the incubator . Samples were then incubated for 1 hr in 1:5000 SiR-tubulin ( 200 nmol ) . Hoechst 344442 ( Thermo Fisher Scientific , H1399 ) was added 10 min before imaging . Plates were allowed to equilibrate in the Okolabs stage top incubator before confocal imaging on the Leica SP5 using the LAS-AF software , 405 nm diode , argon and 561 and 648 nm laser lines , three Photomultiplier tubes , and one HyD GaSP detector , as per the requirement of the experiment . Images were scanned using a 63× 1 . 4 NA oil immersion objective and processed using ImageJ and Imaris software . Embryos were lysed and homogenized in IP lysis buffer ( 10 µl/mg ) at 4℃ on a rotator for 30 min . Composition of IP lysis buffer is ( 50 mM Tris-HCl [pH 7 . 5] ) , 100 mM NaCl , 10% glycerol , 0 . 5 mM EDTA , 0 . 5% IGEPAL , and 1/100 Halt protease and phosphatase inhibitor ( Thermo Fisher Scientific , 78443 ) and a tablet of Protease Inhibitor Tablet – one tablet per 10 ml ( cOmplete Mini , Roche , 11836170001 ) . The lysate was cleared by spinning at 4℃ , 14 , 000 rpm , for 20 min . The protein concentration was determined using the BCA Protein Assay Kit as per the manufacturer’s instruction ( Thermo Fisher Scientific , 23225 ) . For each IP , 500 µg of protein was incubated with 3 µg of each antibody overnight at 4℃ ( Table 1 ) with mild agitation ( side-to-side ) . IP of immunocomplexes was done using PureProteome Protein G magnetic beads ( Millipore LSKMAGG10 ) . 30 µl beads/IP were equilibrated with 500 µl IP lysis buffer by gentle agitation for 5 min at 4℃ . Tubes were placed on a magnet for 2 min , and the buffer was aspirated off with the fine pipette . 200 µl antibody-lysate mix was added to each tube of 30 µl equilibrated beads and incubated for 45 min with agitation to concentrate immunoglobulin complexes on beads at 4℃ . Washes ( eight times ) were performed , each lasting 5 min . Washes were as follows: 2× washes in Wash Buffer-1 ( same as IP lysis buffer ) , followed by 2× washes with Wash Buffer-2 ( IP lysis buffer with reduced 0 . 2% IGEPAL ) , finally 4× washes with Wash Buffer-3 ( IP lysis buffer without any IGEPAL detergent ) . All wash buffers were aspirated , and dry beads were stored at –80℃ , or samples were sent immediately for MS . All MS experiments were done at the IGMM Mass Spectrometry facility as per their published protocol ( Turriziani et al . , 2014 ) . Briefly , the immunocomplexes collected on magnetic beads were processed to generate tryptic peptides . Proteins were eluted from beads by incubating at 27℃ for 30 min in elution buffer ( 2 M urea , 50 mM Tris-HCl pH 7 . 5 , and 5 µg/ml trypsin ) . The sample was centrifuged , bead pellets washed twice , and the supernatant from samples digested overnight at room temperature ( RT ) . Iodoacetamide was added to the samples to inhibit disulfide bond formation and incubated for 30 min in the dark . Following this , trifluoroacetic acid ( TFA ) was added to stop tryptic digestion . Desalting and pre-fractionation of the digested peptides were done by manually using C18 pipette stage-tips filled with 3 M Empore disc activated with 50% acetonitrile and 0 . 1% TFA and then washed once with 0 . 1% TFA . The peptide mixtures were passed manually along to the column with a syringe to concentrate and purify the analytes . Peptides were subsequently eluted twice in 50% acetonitrile and 0 . 1% TFA and both eluates were combined . Samples were concentrated and resuspended in 0 . 1% TFA . This was followed by chromatographic separation on a Reprosil column along a 3–32% acetonitrile gradient . The LC setup was attached to a Q-Exactive mass spectrometer , and ion mass spectra were obtained following HPLC during a tandem MS run . Mass spectra were analyzed using MaxQuant software . Label-free quantification intensity ( LFQ ) values were obtained for analysis by identifying mass/charge ratio , and their intensities at a specific elution time for individual peptides . The data were collected for both control ( GFP ) and specific proteins IPs ( i . e . , IFT88 , IFT140; Table 1 ) . LFQ values for the proteins were obtained by summing the ion intensities corresponding to peptides after assigning the unique peptides to proteins . The ratio of LFQ intensities of test:control was taken , where higher the ratio better corresponds to a better enrichment of protein in complex . Complete MS data are available on ProteomeXchange ( PXD022652 ) . The relative concentration of IFTs was calculated after normalizing the individual test values with respective GFP-LFQs , as shown in the figures . Cells or tissues were lysed in 1× Cell Lysis Buffer with the addition of 1/100 Halt protease and phosphatase inhibitor ( Thermo Fisher Scientific , 78443 ) and a cOmplete Protease Inhibitor Tablet , one tablet per 10 ml ( cOmplete Mini , Roche , 11836170001 ) . Prepare 1× Cell Lysis Buffer by diluting 10× stock in ddH20 ( Cell Signaling Technology [10x #9803]: 20 mM Tris-HCl [pH 7 . 5] , 150 mM NaCl , 1 mM Na2 EDTA , 1 mM EGTA , 1% Triton-X100 , 2 . 5 mM sodium pyrophosphate , 1 mM β-glycerophosphate , 1 mM Na3VO4 , 1 µg/ml leupeptin ) . The lysate from embryos was homogenized at 4℃ for 30 min and from cells was sonicated briefly ( 5× , 10 s pulses , Bioruptor Diagenode ) to lyse the tissue or cells . The lysate was centrifuged at 14 , 000 g at 4℃ for 30 min and the supernatant transferred to a fresh tube . Ready-to-use SDS-PAGE gels ( NuPage Novex precast gels , Thermo Fisher Scientific ) were used to separate proteins . The resolved proteins on the gel were transferred to PVDF ( Hybond P , GE HealthCare ) using the XCell II Blot module as per the manufacturer’s instruction . The membrane was then blocked with a 10% solution of dried skimmed milk ( Marvel Premier Foods ) made in 1× TBST ( 0 . 05% Tween-20 in TBS ) for 1 hr RT , washed with PBS and incubated with primary antibody ( Table 1 ) diluted in 1% skimmed milk solution in 1× TBST overnight at 4℃ on shaker/roller . Membranes were washed in 1× TBST 3 , 10 min followed by a 1× wash with PBS , and incubated in HRP-conjugated secondary antibody from appropriate species ( Table 2 ) for 1 hr at RT , diluted in a solution of 1× TBST and 1% milk . Blot was then washed with 1× TBST , three times and with PBS twice . After the washes , signals were detected by the Super Signal ELISA Femto kit ( Thermo Fisher Scientific , 37074 ) or Super Signal ELISA Pico kit ( Thermo Fisher Scientific , 37069 ) . Protein bands were visualized digitally by transmission light imaging on ImageQuant LAS 4000 ( GE HealthCare ) and analyzed using ImageQuant TL software . Protein bands on blots were quantified with ImageJ/Fiji software by measuring individual bands intensity and normalizing intensities with loading control bands on the same blot . Cells were washed two times with warm PBS , then fixed in either 4% PFA in 1× PHEM/PBS 15 min at RT , 2% fresh glutaraldehyde in 1× PHEM for 15 min , or pre-extracted for 30 s on ice in PEM ( 0 . 1 M PIPES pH 6 . 8 , 2 mM EGTA , 1 mM MgSO4 ) prior to fixing in ice-cold methanol on ice for 10 min according to Table 1 , then washed twice with PBS . 1× PHEM ( pH 6 . 9 ) contains 60 mM PIPES , 25 mM HEPES , 10 mM EGTA , and 4 mM MgSO4·7 H20 . The cells were treated twice with 50 mM NH3Cl for 15 min each for PFA-fixed cells , or 0 . 01 mg of NaBH4 in 1× PBS for 7 min for glutaraldehyde-fixed cells to quench autofluorescence . Cells were then washed twice with PBS . Cells were permeabilized with 0 . 25% Triton-X 100/TBS for 10 min at RT . Cells were rinsed twice in 1× TBS for 5 min . Blocking for non-specific binding was done by incubating samples in 10% donkey serum in 0 . 2% Tween-20/TBS for 60 min at RT . Samples were washed twice with PBS . Primary antibodies ( Table 1 ) were added to samples and incubated for 60 min at RT or 4℃ overnight in dilutant made of 1% donkey serum in 0 . 025% Triton X-100/TBS . Samples were washed in 0 . 25% Triton-X 100/TBS 4–6 times , 10 min each . Secondary antibodies diluted in 1% donkey serum and 0 . 025% Triton X-100/TBS were incubated on samples for 60 min at RT . Samples were washed with 0 . 25% Triton-X 100/TBS 4–6 times 10 min , stained with DAPI ( 1:1000 ) in PBS for 5 min at RT , again washed with PBS and directly imaged or coverslips were added on slides using ProLong Gold antifade ( Thermo Fisher Scientific ) , according to the manufacturer’s instructions . Confocal imaging was done on a Leica SP5 using the LAS-AF software , 405 nm diode , argon and 561 and 648 nm laser lines , three Photomultiplier tubes , and one HyD GaSP detector , as per the requirement of the experiment . Images were scanned using a 63× 1 . 4 NA oil immersion objective and later processed using ImageJ and Imaris software . The sequence match of IFT-A proteins was found by iterative rounds of homology searches via alignment for sequence proximity-based clustering as described before ( Wells et al . , 2017; Wells and Marsh , 2019 ) . Further SWISS-MODEL server was used to model IFT-A complex protein structures as described on the server ( Waterhouse et al . , 2018 ) . Briefly , a template search with BLAST and HHblits was performed against the SWISS-MODEL template library . The target sequence was searched with BLAST against the primary amino acid sequence contained in the SMTL . An initial HHblits profile , followed by one iteration of HHblits against NR20 , was run and the obtained profile then searched against all profiles of the SMTL . The top hit in all of IFTA searches was 3mkqA ( Lee and Goldberg , 2010 ) , a coatomer β′ subunit 2 . 5 Å X-ray structure with 14–20% sequence identity and 25–30% sequence similarity with different IFT-A proteins . A coatomer α subunit was also found within these top matches . Models were built on the target-template alignment using ProMod3 . Coordinates that are conserved between the target and the template were copied from the template to the model . Insertions and deletions were remodeled using a fragment library . Side chains were then rebuilt . Finally , the geometry of the resulting model was regularized by using a force field . In case loop modeling with ProMod3 fails , an alternative model was built with PROMOD-II . The global and per-residue model quality has been assessed using the QMEAN scoring function . The obtained model was processed later in Pymol software for structural analysis . The codon-optimized sequences for C . reinhardtii IFT-A trimeric complex ( UniProt accession codes: IFT43_A8HYP5 , IFT121_A8JFR3 , and IFT139_A9XPA6 ) were assembled into a single construct for expression in mammalian cells . The IFT43 gene was fused to TEV cleavable His-GFP-tag at the N-terminus for affinity purification and inserted into pAceCMV vector while the IFT139 and IFT121 subunits were untagged . The pAceCMV_His-GFP-IFT43 , pIDC_IFT121 , and pIDK_IFT139 were fused using an in vitro Cre recombinase ( New England Biolabs ) by the LoxP sites in the vectors to form the IFT-A trimer construction . Large-scale transient expression of the IFT-A trimeric complex in mammalian HEK293S cells was carried out by transfection of the IFT-A trimer construct using PEI ( 40 kDa linear polyethylenimine , 1 mg/ml stock in water ) . Before transfection , sterile and high-quality DNA was prepared using a NucleoBond Maxiprep Kit ( MACHEREY-NAGEL ) with 200 ml overnight culture of DH5α cells containing the construct . HEK293S cells were cultivated 1 day before the transfection in medium ( FreeStyle 293 Expression Medium , Thermo Fisher ) with 1% FBS and 1% penicillin/streptomycin . Cultures were incubated in a humidified incubator with 5% CO2 at 37°C with 130 rpm shaking to let the cells grow . The cells were diluted to 1 . 1 × 106 cells/ml before transfection with fresh and warm medium . The transfection mixtures were prepared using a sterile flow bench . For expression in 1 l of HEK293S cells , 1000 μg IFT-A trimer DNA were diluted into 25 ml medium without antibiotics or FBS . In another tube , 3000 μg PEI were diluted in 20 ml medium and added to the diluted DNA dropwise . The mixture was incubated at RT for 5 min to let the PEI-DNA transfection complex form . The mixture was added dropwise to cells and mixed gently by swirling the flask . Cells were incubated at 37°C in a CO2 incubator for 48 hr . The cells were harvested by centrifugation at 800 × g for 10 min at 4°C , and the cell pellet was flash-frozen in liquid nitrogen and stored at –80°C until use . The IFT-A trimeric complex was purified using the His-tag on the IFT43 for affinity chromatography . Briefly , a frozen pellet from 1 . 5 l of HEK293S cell culture was thawed on ice and resuspended in lysis buffer ( 50 mM HEPES pH 7 . 4 , 250 mM NaCl , 2 mM MgCl2 , 10% [v/v] glycerol , and 5 mM β-mercaptoethanol ) supplemented with 1 μl DNase and one cOmplete Protease Inhibitor Tablet ( cOmplete-EDTA Free Protease Inhibitor Tablet , Roche Applied Science ) to a final volume of 20 ml . Cells were lysed in a dounce-type tissue grinder ( Wheaton ) using 30 strokes . The cell lysate was cleared by centrifugation at 48 , 000 × g for 45 min at 4°C . The clarified supernatant was loaded onto a 1 ml TALON column ( HiTrap , Cytiva ) pre-equilibrated with lysis buffer at 4°C . The bound protein was washed with 15 mM imidazole in QA buffer ( 20 mM Tris–HCl pH 7 . 5 , 10% glycerol , 50 mM NaCl , and 5 mM β-mercaptoethanol ) , followed by elution with 150 mM imidazole in QA buffer . The elution containing the IFT-A proteins was loaded onto a 5 ml Q column ( HiTrap Q FF , Merck-Millipore ) , and the bound IFT-A proteins were eluted in QA buffer with a 50–500 mM gradient of NaCl . The elution fractions containing the IFT-A proteins were concentrated to 500 μl in a 100 kDa molecular weight cutoff concentrator ( Amicon Ultracel , Merck-Millipore ) for subsequent SEC on a pre-equilibrated Superose 6 Increase column ( 10/300 GL , Merck-Millipore ) in SEC buffer ( 10 mM HEPES pH 7 . 5 , 150 mM NaCl , 2 mM MgCl2 , 1 mM DTT , 5% glycerol ) . The SEC peak fractions were analyzed by SDS-PAGE and resulted in the sample used in Figure 5B and C . DNA sequences encoding for the IFT43 with an N-terminal TEV cleavable hexa-histidine tag and untagged IFT121 were cloned into the two expression cassettes of the pFL vector . The gene encoding for IFT139 was cloned in another pFL vector . The expression and purification of His-IFT43 , His-IFT43/121 complex , and His-IFT43/121/139 complex was performed as previously described for the CrODA16 protein ( Taschner et al . , 2017 ) with the following modifications . Two recombinant baculoviruses for IFT139 and IFT43/121 were generated from separate constructs . The expression was carried out in sf21 suspension cells by co-infection with these two recombinant baculoviruses . After 3 days of incubation at 27°C , the cells were harvested by centrifugation . The His-43 , His-IFT43/121 , and His-IFT43/121/139 were purified using a similar purification procedure to that for the His-GFP-tagged IFT-A trimer describe above and were purified by Ni-NTA affinity , ion-exchange , and SEC . The SEC fractions containing His-tagged IFT-A proteins were used for the lipid overlay assays in Figure 5—figure supplement 1B–D . The purified His-IFT43/121 and His-IFT43/121/139 were digested overnight using TEV protease for removal of His-tag . The resulting IFT-A proteins were loaded onto SEC , and fractions containing untagged IFT-A proteins were further used for the binding assay with liposomes in Figure 5D . To detect the direct binding between non-core IFT-A complexes and lipids , the His-GFP-tagged IFT-A trimeric complex or His-tagged proteins purified from insect cells and Membrane Lipid Strips ( Echelon Biosciences , P-6002 ) with 100 pmol of 15 different lipids were used following the manufacturer’s protocol . The strips were blocked in 3% ( w/v ) BSA in TBS-T buffer ( 50 mM Tris [pH 7 . 4] , 150 mM NaCl , and 0 . 1% [v/v] Tween 20 ) at 4°C overnight in dark with gentle agitation . After blocking , they were washed in TBS-T buffer three times and 5 min each , followed by incubation at RT for 1 hr with IFT-A proteins in SEC buffer supplemented with 3% ( w/v ) BSA . The strips were washed three times in TBS-T as before and soaked in 3% ( w/v ) BSA in TBS-T with primary antibody against His-tag ( THE His Tag Antibody , Mouse , GenScript ) at a 1:2500 dilution for 1 hr at RT . Strips were washed three times and incubated with horseradish peroxidase ( HRP ) -conjugated polyclonal rabbit anti-mouse immunoglobulins ( 1:1000 dilution , Dako ) for 1 hr followed by three TBS-T washes . An ECL Prime Western Blotting reagent ( Amersham ) was used as the substrate for the HRP , and the binding of IFT-A proteins onto spotted lipids was recorded with the ChemiDoc imaging system ( Bio-Rad ) . The POPC-liposomes and PA-containing liposomes ( PE/PG/PA ) were purchased from T&T Scientific Corp . The liposomes ( PE/PG/PA ) have a similar phospholipids composition to that of Chlamydomonas ciliary membrane as reported previously ( Lechtreck et al . , 2013 ) . The percentage of PA was 11 . 36% while the ratios of PE and PG as the framework of liposomes were requantified to 63 . 18 and 25 . 46% , respectively . To observe the binding between IFT-A complexes with liposomes , the liposomes ( PE/PG/PA , 0 . 20 mM ) were applied to homemade carbon grids directly or after incubation with IFT-A complexes ( untagged IFT139/121/43 trimer or IFT121/43 dimer , 0 . 25 µM ) at 25°C for 10 min . 3 μl of the sample were applied to the plasma-cleaned grids for 30 s before it was blotted , and the sample was stained with 2% ( w/v ) uranyl-format staining by applying 3 μl of stain three times on the grids . The negative stain grids were imaged on an FEI Tecnai G2 Spirit TEM operated at 120 kV with a 67 , 000× nominal magnification corresponding to the digital pixel size of 1 . 59 Å/pixel . The electron micrographs were recorded on a water-cooling 4k CMOS CaMeRa ( TemCam-F416 ) . The mixture of IFT-A trimer ( 0 . 10 µM ) with POPC liposomes ( 0 . 20 mM ) , as a negative control , was checked using negative staining EM by following the same procedure . WDR35-EmGFP and ARL13B-mKate expressing Wdr35-/- MEFs were serum-starved for 24 hr , stained with Hoechst 33342 ( R37605 ) for 10 min in culture condition , fixed with 4% PFA and 0 . 1% GA in 1× PHEM and imaged on Zeiss LSM 880 upright single-photon point scanning confocal system with Quasar detector ( 32 spectral detection channels in the GaAsP detector plus 2PMTs ) and transmitted light detector , Airyscan detector for high-resolution imaging . Cells were grown on 35 mm glass-bottom dishes with grids ( cat . no . P35G-1 . 5-1 . 4C-GRID ) and firstly brightfield images were made with Plan-Apochromat 10×/0 . 45 M27 objective to save the coordinates of cells needed for the correlation with the respective TEM data . Confocal and Airyscan imaging was done using Plan-Apochromat 63×/1 . 4 oil DMC M27 objective , 405 nm laser diode , 458 , 477 , 488 , 514 nm multiline integrated argon laser and 594 nm integrated HeNe laser . Z-stack was acquired sequentially to get the whole 3D volume of the cell , and the image was further deconvolved using the in-built software . After Airyscan imaging , the sample was processed for TEM as described above . 70 nm sections were made for the regions of saved coordinates from brightfield imaging , mounted on grids and imaged on FEI Morgagni TEM ( 100 kV ) microscope . Wdr35-/- MEFs expressing WDR35-EmGFP and ARL13B-mKate2 ( Table 2 ) were serum-starved for 4 hr . MEF cells were grown on 6 mm sapphire disks ( Wohlwend GmbH , Switzerland , 1292 ) and high-pressure frozen ( EM ICE , Leica Microsystems ) . The frozen samples were processed by freeze substitution in a Leica AFS2 temperature-controlling machine ( Leica Microsystems ) using 0 . 01% uranyl acetate ( Polyscience Europe GmbH , 21446 ) and 4% water in glass distilled acetone ( EMS , E10015 ) as freeze substitution medium and then embedded in Lowicryl HM-20 ( Polysciences , 15924-1 ) . 70-nm-thick serial sections were sectioned on a Leica Ultracut UCT ultramicrotome ( Leica Microsystems ) . Sections were labeled with anti GFP antibody , 1:20 ( Abcam , ab6556; Table 1 ) followed by secondary goat anti-rabbit antibody coupled to 10 nm gold , 1:30 ( BBI Solutions , batch 008721; Table 2 ) . Before antibody staining , grids were incubated twice section side for 10 min each on blocking buffer PBG ( 0 . 5% BSA/0 . 1% fish skin gelatin in PBS ) . Following blocking , grids were incubated for 1 hr in primary-Ab/PBG in a wet chamber , given five 2 min washes with PBG and incubated for 1 hr in secondary-Ab/PBG . Grids were washed five times for 2 min with PBG , followed by five 2 min washes with PBS . Antibodies were subsequently fixed for 1 min 0 . 1% glutaraldehyde/PBS , followed by five 2 min washes with PBS and five 2 min washes with H2O . After immunogold labeling , the sections were stained with 1% uranyl acetate ( Polyscience Europe GmbH , 21446 ) in water for 8 min and 0 . 04% lead citrate ( EMS , 17800 ) for 5 min . The sections were imaged using Tecnai 12 ( Thermo Fisher Scientific , formerly FEI/Philips ) at 100 kV with TVIPS F214 and F416 cameras ( TVIPS , Gauting , Germany ) . All image processing were performed using Fiji ( Schindelin et al . , 2012 ) . Macros for quantification of PCM1 ( RadialIntensityFromCentrosomes . ijm ) and clathrin ( 3DMeanIntensityfromUserDirectedPoints . ijm ) can be found on GitHub ( https://github . com/IGC-Advanced-Imaging-Resource/Quidwai2020_WDR35paper; Quidwai et al . , 2021; copy archived at swh:1:rev:96b375ac31f1451dea93943fac7f563ad348ee69 ) . To measure PCM1 intensity radially from the centrosomes , an average intensity projection of the z-stack was obtained , and the γ-tubulin signal was segmented using RenyiEntropy threshold and the Analyze Particles tool to obtain masks of the centrosomes . The selections obtained from the masks were enlarged using the ‘Make Band’ function to create a band region of interest ( ROI ) . This was done by increasing in 1 µm increments until there were five bands . The centrosome masks and the surrounding bands were measured on the PCM1 channel of the average intensity projection image . To quantify clathrin intensity around the cilia base , a point was manually selected as the center of the basal point . The user was blinded to file name and condition while quantification took place . This point was expanded 1 µm in each direction to create a shall of 2 µm diameter in x , y , and z . This shell was then measured using the 3D image suite in ImageJ ( Ollion et al . , 2013 ) . Etomo and IMOD ( Kremer et al . , 1996 ) were used to reconstruct tomograms and manually segment tomograms , respectively . These segmentations were used to create objects using the 3D Image suite in Fiji . The 3D centroids were obtained and the manually segmented ROI on the 2D slice that the 3D centroid was on was selected to move forward with . A 20 nm width band around this ROI was measured using the ‘Make Band’ function . The integrated density of this band ROI was quantified as an indication of how electron dense the region around the user segmented vesicle is . 3D objects were measured using the 3D Image Suite . Statistical analyses were carried out in GraphPad Prism 8 . | Most human cells have at least one small hair-like structure on their surface called a cilium . These structures can act as antennae and allow the cell to sense signals from the rest of the body . To do this , they contain proteins that differ from the rest of the cell . The content of cilia depends on regulated delivery of these proteins in and out of cilia by a process called the intraflagellar transport or IFT , which involves a large complex made of several proteins . This complex shuttles the cargo proteins back and forth between the base and the tip of the cilia . However , ciliary proteins are not produced in the cilia; instead , they are made in a different part of the cell and then they are transported to the ciliary base . At the point where they enter the cilia , they were thought to bind to the assembling IFT ‘trains’ and be transported across the ciliary gate to the positions where they are needed in cilia . One of the components of the IFT machinery is a protein called WDR35 , also known as IFT121 . If the gene that codes for this protein is faulty or missing , it results in severe disorders in both humans and mice including a range of potentially lethal skeletal dysplasias . Interestingly , without WDR35 , cells cannot build functional cilia . The absence of this protein not only disrupts IFT , stopping certain ciliary proteins and their associated membranes from entering cilia; it also causes a ‘traffic jam’ with a pile-up of transport intermediates from the place in cell where they are made to the cilia . It is unclear why a mutation in one of the components of the IFT would have this effect , raising the question of whether WDR35 , or IFTs a whole , has another role in bringing the cargo proteins into the cilia . To understand this phenomenon , Quidwai et al . analysed the structure of WDR35 and other IFT proteins and found that they are very similar to a protein complex called COPI , which is involved in transporting membrane proteins around the cell . When certain proteins are newly made , they are stored in small lipid bubbles – called vesicles – that then selectively move to where the proteins are needed . COPI coats these vesicles , helping them get to where they need to go in a process called vesicular transport . Quidwai et al . found that WDR35 and other IFT proteins are able to bind to specific types of lipid molecules , suggesting that they might be assisting in a form of vesicle transport too . Indeed , when mouse cells grown in the lab were genetically engineered so they could not produce WDR35 , coatless vesicles accumulated around the base of the cilia . Adding back WDR35 to these mutant cells rescued these defects in vesicle transport to cilia as well as allowed functional cilia to be formed . These results provide evidence that WDR35 , likely with other IFT proteins , acts as a COPI-like complex to deliver proteins to growing cilia . Further research will investigate the composition of these vesicles that transport proteins to cilia , and help pinpoint where they originate . Quidwai et al . ’s findings not only shed light on how different genetic mutations found in patients with cilia dysfunction affect different steps of transporting proteins to and within cilia . They also increase our understanding of the cellular roadmap by which cells shuttle building blocks around in order to assemble these important ‘antennae’ . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2021 | A WDR35-dependent coat protein complex transports ciliary membrane cargo vesicles to cilia |
TRPV1 channels support the detection of noxious and nociceptive input . Currently available functional and structural data suggest that TRPV1 channels have two gates within their permeation pathway: one formed by a ′bundle-crossing′ at the intracellular entrance and a second constriction at the selectivity filter . To describe conformational changes associated with channel gating , the fluorescent non-canonical amino acid coumarin-tyrosine was genetically encoded at Y671 , a residue proximal to the selectivity filter . Total internal reflection fluorescence microscopy was performed to image the conformational dynamics of the channels in live cells . Photon counts and optical fluctuations from coumarin encoded within TRPV1 tetramers correlates with channel activation by capsaicin , providing an optical marker of conformational dynamics at the selectivity filter . In agreement with the fluorescence data , molecular dynamics simulations display alternating solvent exposure of Y671 in the closed and open states . Overall , the data point to a dynamic selectivity filter that may serve as a gate for permeation .
Cation channels from the TRPV family are functional tetramers with a central pore domain that is shaped by TM5 and TM6 transmembrane segments from each of the four subunits ( Ramsey et al . , 2006 ) ( Figure 1a ) . As in Shaker-related tetrameric ion channels , the pore domain houses the ion selectivity filter and an intracellular ‘bundle crossing’ formed by the α-helical S6 segments . Functional and structural data suggest that the selectivity filter and inner bundle crossing have the potential to control the flux of ions according to the electrochemical gradient ( Cao et al . , 2013; Gao et al . , 2016; Salazar et al . , 2009 ) . The lower constriction is highly conserved ( Palovcak et al . , 2015 ) and associated to the canonical gate ( Cao et al . , 2013; Steinberg et al . , 2014 ) , but less is known about the putative role of the upper constriction at the ionic selectivity filter , possibly due to the functionally sensitive nature of this region to side-chain mutagenesis . Nevertheless , high-resolution structures of TRPV1 suggest that the upper constriction undergoes structural changes in the presence of DkTx ( Gao et al . , 2016 ) , a spider toxin that increases the open probability of the channel ( Siemens et al . , 2006 ) . Moreover , the inhibition of the unitary conductance of TRPV1 channels by protons suggests that the selectivity filter region is flexible and is capable of adopting multiple , functionally distinct conformational states ( Liu et al . , 2009; Wen et al . , 2016 ) . However , it is not known if such conformational dynamics at the selectivity filter might be coupled to channel activation by agonists , such as capsaicin , and if filter gating is sufficient or acts in concert with the lower bundle crossing to control the ionic conductance . Organic dyes with environmentally sensitive fluorescence have been used for nearly 20 years to describe membrane protein conformational changes and ion channel motions ( Mannuzzu et al . , 1996 ) . The voltage-clamp fluorometry ( VCF ) approach relies on the covalent attachment of such fluorophores , usually via introduced cysteine residues , and allows for the simultaneous recording of fluorescence signals and voltage-clamped ionic currents from expressed ion channels , often in the Xenopus oocyte ( Cha and Bezanilla , 1997 ) . This technique has proven to be useful for the real-time description of subtle , and in some cases electrophysiologically silent , conformational changes in voltage- and ligand-gated ion channels ( Cha et al . , 1999; Islas and Zagotta , 2006; Pless and Lynch , 2009 ) . However , excessive background labeling of intracellular sites , at endogenous cysteine residues for instance , and the prerequisite for solvent accessibility of the labeling site serve to limit the utility of the technique substantially for the study of intracellular or solvent-restricted residues . A possible solution to these limitations is the use of genetically encoded fluorophores in the form of non-canonical amino acids ( f-ncAA ) ( Drabkin et al . , 1996 ) . Such strategies employ an orthogonal suppressor tRNA and an evolved tRNA synthetase ( RS ) , which can be used to encode the ncAA at any site in the reading frame of the target gene ( Drabkin et al . , 1996; Sakamoto et al . , 2002 ) . This approach has been used for site-specific incorporation of fluorescent ncAAs on soluble proteins in prokaryotic systems ( Summerer et al . , 2006; Wang et al . , 2006 ) , membrane proteins expressed in Xenopus oocytes ( Kalstrup and Blunck , 2013; ) , and proteins expressed in mammalian cells ( Chatterjee et al . , 2013; Luo et al . , 2014; Shen et al . , 2011; Steinberg et al . , 2016; Zagotta et al . , 2016 ) . The f-ncAA coumarin has potential for the study of expressed protein conformational dynamics because of its small size and its exquisite sensitivity to environmental polarity ( Liu et al . , 2015; Wagner , 2009; Wang et al . , 2006 ) . This sensitivity can be exploited to report on variations ( i . e . dielectric changes ) in the surroundings of the incorporated f-ncAA and , in turn , it can be interpreted as a signature of a protein conformational change . Notable examples include the use of coumarin and the amber codon suppression technique to report on peptide binding to viral proteins ( Ugwumba et al . , 2011 ) , on the photoregulation of firefly luciferase and on the subcellular localization of proteins in living mammalian cells ( Luo et al . , 2014 ) , as well as onstructural rearrangements in isolated NaK channels ( Liu et al . , 2015 ) . Here , the spectral dynamics of an encoded hydroxy-coumarin were determined in individual live cells with total internal reflection ( TIRF ) microscopy and rapid imaging of single emitters to study the conformational changes occurring within the pore of the capsaicin receptor TRPV1 during activation .
Previous cysteine accessibility experiments showed that Y671C residues define the boundary for both silver ( Ag+ ) state-dependent accessibility and disulphide bond formation in the closed state ( Salazar et al . , 2009 ) . Thus , we reasoned that a genetically encoded coumarin residue at position Y671 ( Y671Coum ) would be functionally tolerated and if so , could potentially be well positioned to report on local structural rearrangements occurring within the pore during gating ( Figure 1a ) . As expected for cells containing an encoded fluorescent probe , HEK-293T cells heterologously expressing both TRPV1-YFP and amber codon-containing TRPV1 mutants ( TRPV1YFP/TRPV1TAG ) exhibit coumarin–YFP colocalization ( Figure 1b; Figure 1—figure supplement 1 ) consistent with the presence of coumarin and YFP co-fluorescence . Whole-cell patch-clamp electrophysiological recordings performed at room temperature ( 22°C ) demonstrate the functional expression of the coumarin-containing TRPV1 channels ( TRPV1TAG ) when expressed alone or at 1:7 molar ratio ( TRPV1TAG:TRPV1–YFP ) ( Figure 1c ) . The later condition was chosen to simplify the imaging of single emitters ( i . e . one coumarin molecule per channel tetramer or cluster of tetramers ) . Both macroscopic channel kinetics and the conductance-voltage ( G-V ) relationship were similar between wild type ( WT ) and coumarin-containing channels ( Figure 1c ) . The G-V relationship from wild type and coumarin channels were well fit by a Boltzmann function with a half-maximal activation voltage ( V0 . 5 ) near +100 mV ( Figure 1d ) . Thus , the data suggest that position Y671 in TRPV1 is amenable to nonsense suppression and encoding of f-ncAA coumarin . In order to demonstrate the ability of the genetically encoded hydroxy-coumarin side chain to act as a reporter for protein activity in live cells , HEK Y671-coumarin TRPV1 expressing cells were imaged within the evanescent field in total internal reflection fluorescence ( TIRF ) microscopy , with a single-photon counting avalanche diode ( SPAD ) camera ( 64 × 32 pixels array with in-pixel counters ) ( Bronzi et al . , 2014; Michalet et al . , 2013 ) . Specifically , photon counts of the emission signal were examined to determine the effect of capsaicin , the canonical agonist of the TRPV1 channel ( Figure 1e and f; Figure 1—figure supplement 2 ) . The photon count distribution showed clear differences between pixels representing background and pixels from fluorescent spots , which presumably arise from fluorescent channel subunits ( Figure 1f ) . Further , when the cell is exposed to a maximal concentration of capsaicin ( 3 μM ) , the photon counting rate further shifted to higher values ( Figure 1f ) . The modes observed for these populations showed significant differences ( n = 4; p<0 . 05 ) . One interpretation of the capsaicin-induced increase in photon emission counts is that it stems from local changes in the immediate environment of the dye . To simplify the analysis and interpretation of the optical data , we set out to measure the fluctuations of the fluorescence signal emitted from a single diffraction-limited emitter directly . Given the large pixel size and the low fill-factor of the SPAD array used ( Bronzi et al . , 2014; Michalet et al . , 2013 ) , the precise localization of a diffraction-limited signal is not possible under the condition of overexpression of membrane proteins in living cells . Therefore , steady state fluctuations of fluorescence were recoded by an electron-multiplying charge coupled device ( EM-CCD ) camera to resolve changes in the optical properties of the dye ( Blunck et al . , 2008 ) . As in the electrophysiological experiments , cDNAs encoding for both wild-type channels ( TRPV1 ) and amber-codon-containing mutants were mixed in 7:1 ratio to promote the formation of uneven channel tetramers . This strategy results in a small yet discrete number of diffraction-limited coumarin-emitting spots as observed by TIRF microscopy ( Figure 2a ) . The coumarin signal colocalizes well with the YFP signal corresponding to wild-type YFP-tagged channel subunits ( Figure 2a ) . Close examination of these fluorescent spots by photobleaching showed a stepwise behavior allowing for ex-post identification of spots containing a single emitter ( Figure 2b ) . A clear change in fluorescence was evidenced by the ensemble average of several diffraction-limited spots once the cell is exposed to capsaicin ( 3 μM; Figure 2c ) . The ensemble recording showed not only an increase of fluorescence after agonist treatment but also increases in the standard deviation of the signal’s mean value ( Figure 2c ) . Given the Y671 position within the TRPV1 selectivity filter , we reasoned that the environmental sensitivity of coumarin may well provide kinetic information on channel gating . Stationary noise analysis was performed on the basis of the autocorrelation ( C ( ∆t ) ) of the fluorescence signal ( Figure 3a ) . The autocovariance function can be used to calculate the correlation between values of a noisy signal taken at increasing time intervals , ∆t . In this regard , C ( ∆t ) will be higher for small values of ∆t ( i . e . for data points close to each other ) and this function rapidly decays as ∆t become larger . In the present example , this analysis allows for the identification of a characteristic time constant ( τ ) of the exponential decay , which describes the autocorrelation of the fluorescent signal originating from single emitters . The obtained τ has an inverse relationship to the rates defining the equilibrium of the reaction between open and closed states ( Anderson and Stevens , 1973; Zingsheim and Neher , 1974 ) . The calculated autocorrelation function discriminates well between the three conditions tested ( i . e background , no capsaicin , and capsaicin present in the media ) . The background signal is obtained from cells that have the coumarin amino acid but lack the expressed TRPV1TAG cDNA . This condition generated noise with a single , fast time constant ( τ = 8 . 5 ms ) that was easily separated from the double exponential decay time constants observed in the signal originating from the encoded coumarin conditions . In the absence of capsaicin , the coumarin signal autocorrelation had two time constants with corresponding values of 12 ± 6 ms and 78 ± 22 ms . Incubations with saturating capsaicin concentrations ( 3 µM ) changed these time constants to 13 ± 4 ms and 148 ± 38 ms ( n = 7; Figure 3a ) , consistent with the possibility that the encoded coumarin is reporting on an agonist-induced increase in τ . Such a result can be interpreted as an stabilization of the bright state after ligand binding ( i . e . longer bursts of activity ) either alone or in combination with the destabilization of the dark or non-emitting state , a possibility consistent with the previously determined distribution of closed and open states in TRPV1 channels when activated by capsaicin ( Hui et al . , 2003 ) . In order to begin to distinguish between these possibilities , we examined unitary transitions in the optical data ( Figure 3b; see Materials and methods ) . To do so , we inferred that the signal minima ( i . e . L0 ) would correspond to the magnitude of background signal and would corroborate that fluorescence transitions drop from a given maxima ( L1 ) to the observed minima ( L0 ) as expected ( Figure 3b , inset ) . We analyzed the signal in the presence of capsaicin using a threshold-crossing criterion , where the threshold was chosen as 2 . 5 times the standard deviation of the mean amplitude of background fluctuations ( see Materials and methods ) . This strategy produced an idealized recording , similar to the analysis of single ion channel current idealization trace ( Figure 3b ) . An analysis of this record shows that the agonist does not affect the amplitude of the fluorescence fluctuations , but rather it promotes a shortening of the mean dwell-time of the bright state , which we defined as L1 , going from 398 ± 133 ms to 156 ± 52 ms ( n = 5; Figure 3c and d ) . On the other hand , capsaicin increases the probability of being in the bright state ( Figure 3 ) . Taken together , these two observations support the notion that capsaicin increases the burst length of transitions to the bright state . This is in agreement with the autocorrelation data , which show that capsaicin increases the time constant , and consistent with an increased burst length of bright events . The data from single emitters also supports previous data showing that the main gating effect of capsaicin is increasing the burst-length of openings by shortening the closed state , instead of prolonging the duration of the open state ( Hui et al . , 2003 ) . Overall , the fluorescence data is consistent with the possibility that coumarin at position Y671 reports on the transition between open and closed states of the channel . Further , given the physicochemical properties of coumarin ( i . e . that it is quenched in an aqueous environment ) , the data suggest that position Y671 is more water-exposed and thus has decreased in fluorescent output , in the closed state compared to the open state ( Wagner , 2009 ) . By comparison , the available TRPV1 structural data place W426 in an apparently invariant solvent-exposed region ( Figure 3—figure supplement 1a ) , thus , the open-closed transition would not be expected to induce evident changes in local solvation ( Gao et al . , 2016 ) . In order to investigate the specificity of the spectral effects measured at Y671 with coumarin , a second site , W426 , was examined in parallel . Electrophysiological analysis confirmed that W426coum is a functional channel ( Figure 3—figure supplement 1b and c ) , and the photon-count analysis suggests that the fluorescence signal from this position is insensitive to capsaicin administration ( Figure 3—figure supplement 1d and e ) . Further , the fluctuations of coumarin fluorescence at 426coum are unresponsive to capsaicin treatment , as determined by the calculated bright state probability , PL1 ( Figure 3e; Figure 3—figure supplement 1f and g ) . Conversely , the PL1 of position 671coum reported capsaicin dose-dependent fluorescent changes which are well fitted by the capsaicin dose-response curve obtained from capsaicin-activated currents at −70 mV ( Figure 3e; Figure 3—figure supplement 2 ) . Lastly , when the fluorescence data from Y671coum was normalized and compared with the electrophysiological data taken from whole-cell macroscopic recordings , we observed a close match between the calculated Ec50 for optical and electrophysiological data , 0 . 33 µM versus 0 . 43 µM , respectively ( Figure 3f ) . Taken together , the data and biophysical analysis suggest a continuum of coordinated motions within the pore , with the Y671coum optical data reporting on a transition to the open conducting conformation in the selectivity filter . To relate the changes in coumarin fluorescence to possible structural transitions experienced by the channel during gating , we analyzed the molecular dynamics ( MD ) trajectories of wild-type and 671coum TRPV1 channels in the closed and open states ( about 750 ns each ) . First , we compared the conformations that Y671 adopts in the closed and open states of TRPV1 ( Figure 4a and b ) . In the closed state , the representative conformation shows that three out of four phenyl moieties are approximately perpendicular to the membrane and do not contact each other ( Figure 4a ) . In this conformation , phenyl moieties create transient hydrogen bonds with the main-chain of F640 of the selectivity filter . In the open state , the planar groups of the four tyrosine residues instead adopt a parallel orientation to the membrane , forming a ring around the central pore ( Figure 4b ) . These opposing orientations in the closed and open conformations predicts an increase in the hydration of the central pore , with the four Y671 moieties being more hydrated in the closed state ( Figure 4c ) . Surprisingly , the open state shows an intermittent occupancy by water molecules of the region surrounded by the four Y671 side chains . Despite being somewhat counterintuitive , this observation is not in contradiction with the conductive nature of the open state: even when devoid of water molecules , the environment provided by the ring of hydroxyl groups from Y671 is highly polar and does not impede ion permeation . In spite of these water density fluctuations , the pore is , on average , continuously hydrated only in the open state ( Figure 5a ) . Na+-permeation events in our simulations ( in the absence of any applied external field ) were observed in the open state , but not in the closed state . An additional insight provided by the simulations is that the hydration of Y671 is different in the two conformations: in the open channel , one face of the phenyl ring is exposed and the adjacent ring is partially buried by F640 and M644 , whereas in the closed state , both faces of the Y671 side chain are exposed to water . To describe hydration of Y671 quantitatively , we calculated the corresponding solvent accessible surface area ( SASA ) in the open and closed conformations . The evolution of the Y671 SASA reveals that the two MD trajectories start to diverge quite early: in particular , in both conformations , the Y671 SASA drops down from ~35% to ~20% at the beginning of the equilibration; then , after ~30 ns , in the closed and open states , respectively , the SASA either increases up to ~30% or continues to decrease until it reaches an equilibrium value of ~15% ( Figure 4d ) . Further , it can be observed that although the open state displays the side chains of Y671 residues engaged in mutual hydrogen bonding interactions , they establish vdW interactions with nearby residues F640 and M644 , whose side chains are in between the selectivity filter backbone groups of two adjacent subunits ( Figure 4e ) . In the closed state , F640 and M644 residues still interact with Y671 as in the open state; however , the different orientation of the side chain phenyl groups allows for a rigid displacement of the pore helix toward the extracellular side of the pore , narrowing the diameter of the selectivity filter ( Figure 4a and e and Figure 5a and b ) . To explore the plausibility of this hypothetical gating mechanism , we performed additional simulations in whichfour coumarin moieties were modeled in the place of the side chains of Y671 . The rationale for these additional calculations was to ascertain whether the bulkier coumarin moiety ( as compared to tyrosine phenyl group ) can accommodate configurations similar to those accomodated by the wild type . In particular , would the ‘ring configuration’ in the open state allow ion permeation in the case of coumarin ? Is coumarin in the ‘vertical configuration’ able to establish H-bonds with and thus promote the displacement of the pore helix in the closed state ? Although the length of the simulations performed on the coumarin mutant would need to be extended in order to establish a more quantitative comparison with the experimental data , these cursory calculations suggest that the proposed gating mechanism is not incompatible with the molecular size of the coumarin moiety ( Figure 5c; Figure 5—figure supplement 1 ) .
Ion channel gating occurs in the range of hundreds of µs and it is one of the fastest protein-dependent processes occurring within cells ( Hille , 2001 ) . This rapid time-frame imposes intrinsic technical barriers in the detection of a limited number of photons coming from few emitting molecules and a consequent low signal-to-noise ratio under these conditions ( Ha , 2014 ) . Given the design of our imaging setups and our experimental conditions ( i . e . transient overexpression ) , we are bound to either the low ( 64 × 32 pixel ) spatial resolution of the ns-time-scale sampling of the SPAD imager ( which does not allow us to discriminate single emitters ) or to a low signal-to-noise ratio restricting us to image at a maximum of 500 Hz ( ms-time scale ) for the individual diffraction-limited spots observed by an EM-CCD camera . As reported previously ( Blunck et al . , 2008 ) , we observed that the latter approach is useful for conformational transitions in the order of 2–5 ms , thus barring optical access to the fast intra-burst channel activity that occurs in tens of µs . Therefore , the unitary transition analysis presented here was limited to the description of each burst as a single opening at the expense of missing activity during short openings . None the less , such an analysis , although lacking detail on unitary transitions , is a valid approach for estimating the duration of the mean open and closed state for a ligand-gated channel such as TRPV1 , which has an open transition lasting for several hundreds of ms ( Hui et al . , 2003 ) . The ideal experimental system would allow for sparse distribution ( every ~1 micrometer ) of the emitter-expressing receptors at the plasma membrane , so that the photon-count profile of individual molecules can be obtained with the SPAD imager with nanosecond resolution . However , in the absence of such technical advances , standard biophysical approaches are in place to provide insights on unitary gating events from an expressed population of channels . Interestingly , the similar rate constants obtained by noise analysis of the optical signal and from the subsequent modeling of the unitary transitions , suggest a strong internal consistency of data . Another limitation to consider in the analysis is the intrinsic blinking of the dye . In the present case , this represents roughly 12% of the PL1 transitions and therefore obfuscates efforts to obtain absolute values for the open probability . Our solution to bypass this issue was to linearly subtract this ‘background open probability’ post-hoc and to normalize the data to its maximal theoretical response . Still , even with this coarse approximation , the optical data is in surprising agreement with whole-cell electrophysiological recording of channel gating . Further , this approach allowed for the direct measurement of the steady-state activity of membrane receptors undergoing fast molecular rearrangements . Specifically , from our fluorescence recordings , we reasoned that Y671coum experiences a change of environment as part of the process of pore rearrangement associated to channel’s opening . This possibility is supported by the dynamic range for the optical measurements presented here , and closely overlaps with the dynamic range of burst activity observed in single-channel recordings of TRPV1 in response to capsaicin ( i . e . 0 . 1 to 1 μM ) ( Hui et al . , 2003 ) . The molecular dynamics simulations showed that Y671 could undergo three significant conformational changes during the channel’s transition from the conducting to the non-conducting conformation: ( i ) a side chain re-orientation with respect to the pore and pore helix , ( ii ) increased water exposure , and ( iii ) a reshaping of the H-bond interaction network . The spectral properties of coumarin are highly sensitive to changes in solvation and/or to changes in the coordination of the hydroxyl group of the dye ( Wagner , 2009 ) . Therefore , any of the modifications detected on MD simulations performed for the wild-type channel can potentially explain the observed changes in fluorescence . As a consequence , a detailed structural interpretation at the level of side-chain dynamics for the fluorescence fluctuations remains out of reach . Moreover , the data presented here do not prove a causal relationship between the motion of the aromatic side chain and the state of the pore helix .
HEK293T cells ( ATCC Cat# CRL-3216 , RRID:CVCL_0063 ) were cultured in DMEM ( Gibco Inc . ) supplied with 10% FBS ( Gibco Inc . ) . Plasmocin ( Sigma ) was used following the provider’s instructions to maintain the cell culture free of mycoplasma contamination . Cells were tested for mycoplasma by standard PCR . Cells were prepared at ( 60–70 ) % confluence and transfected with lipofectamine 2000 , ( Life technologies ) . The target gene was transfected 2–3 hr after the initial transfection of the pair tRNAtag/CoumRS . The ncAA was added to the cell culture to produce a final concentration of between 0 . 5 and 1 μM at the time of the second transfection procedure . Under these transfection conditions , the efficiency of incorporation of non-natural amino acid was as low as 12% , with readthrough calculated to be about 4% ( Steinberg et al . , 2016 ) . Imaging: to simplify the analysis of the optical data , we transfected a mixture of TRPV1YFP/TRPV1TAG in a 7:1 ratio in an attempt to drive the system to incorporate as few coumarin molecules per tetramer as possible . The ratio of expression was not controlled as we concentrated efforts on single emitters identified ex-post . 24 hr after transfection of the target gene , cells were disaggregated and plated on poly-L-lysine treated glass covers . The f-ncAA-containing media was removed at least 12 hr prior to the experiment , allowing cells to clear the soluble ncAA . Generally , the cells were recorded 36–48 hr after the second transfection . Electrophysiology: cells were plated 2–3 hr before patching , and incubated with normal media lacking the supplemented coumarin . L- ( 7-hydroxycoumarin-4-yl ) ethylglycine was obtained as described before ( Wang et al . , 2006 ) . Ringer solution for imaging experiments contained: 140 mM NaCl , 8 mM KCl , 8 mM HEPES , and 1 mM MgCl2 at pH 7 . 4 . External electrophysiology solutions contained 145 mM NaCl , 2 mM CaCl2 , 5 mM KCl , 10 mM HEPES and 10 mM glucose , pH 7 . 4; internal electrophysiology solutions contained 135 mM CsF , 5 mM KCl , 2 mM MgCl2 , 1 mM CaCl2 , 4 mM EGTA and 20 mM HEPES , pH 7 . 4 . To analyze Y671 hydration at the atomic level , we used the molecular dynamics trajectories obtained in our previous work ( Kasimova et al . , 2017 ) . These trajectories were generated starting from the cryo-EM structure of the TRPV1 capsaicin-bound state ( Cao et al . , 2013 ) . In one simulation , we initialized the system by inserting several water molecules ( from 4 to 6 ) inside the channel peripheral cavities ( located between the S4–S5 linker and the S6 C-terminus ) ; in the second simulation , we left these cavities empty . During the equilibration ( about 750 ns ) , the molecular structures converged , respectively , to the closed and open states . To estimate water density along the central pore , we implemented the following strategy . First , using HOLE software ( Smart et al . , 1996 ) , we calculated the pore radius profile for each instantaneous configuration along the trajectory with a stride of 1 ns . Then , for the same set of frames , we calculated the three-dimensional histograms of water occupancy using the Volmap tool of VMD ( RRID:SCR_001820 ) ( Humphrey et al . , 1996 ) . Finally , we integrated the water occupancy in the XY plane ( perpendicular to the channel central pore ) using the pore radius profile as a boundary of the integration domain . The solvent accessible surface area ( SASA ) of Y671 in the closed and open states was estimated using the following procedure . From the MD trajectories , we extracted a set of sub-trajectories , each containing 10 frames taken with a stride of 0 . 2 ns . For each sub-trajectory , we computed the three-dimensional histogram of atomic occupancy ( for all the atoms except water ) using the Volmap tool of VMD ( Humphrey et al . , 1996 ) . We then used this map to define a molecular surface . To this end , we first discretized the map by assigning a value of 1 or 0 , depending on whether or not the local occupancy is larger than a preset threshold . We considered all the bins with a value of 1 and located 1 . 5 Å away from a bin with a value of 0 . Finally , the Y671 SASA was calculated as the overlap between the solvent-accessible surface and the Y671 residue . Note that the SASA was normalized to the Y671 maximal surface area . Molecular dynamics simulations were performed in the open and closed states of the TRPV1 coumarin mutant . The force-field parameters for coumarin were obtained from the CHARMM general force field ( Vanommeslaeghe and MacKerell , 2012 , Vanommeslaeghe et al . , 2012 ) . Simulations were performed using NAMD2 . 10 ( RRID:SCR_014894 ) ( Phillips et al . , 2005 ) at constant temperature and pressure ( one atm ) using the Langevin piston approach . For the vdW interactions , we used a cutoff of 11 Å with a switching function between 8 Å and 11 Å . The long-range component of electrostatic interactions was calculated using the Particle Mesh Ewald approach using a cutoff for the short-range component of 11 Å . The equations of motion were integrated using a multiple time-step algorithm , with a time step of 2 fs and long-range interactions calculated every other step . Trajectories were collected for 100 ns . Cells were imaged using an inverted Olympus IX71 microscope main body and through-the-objective TIRF mode . Both 405 nm and 473 nm solid-state lasers ( Coherent , Santa Clara , CA ) were used to excite coumarin and YFP , respectively . Laser beams were focused to the backplane of a high-numerical aperture objective ( Olympus 60X , N . A . 1 . 49 , oil ) by a combination of focusing lens . Fluorescence emission was collected by an Andor iXonEM + 860 EM-CCD camera ( Andor/Oxford Instruments , Belfast , UK ) , after passing through an emission filter for each acquisition wavelength band of interest ( coumarin: 450/70 nm; YFP:540/40 nm; Semrock , US ) . The light coming to the sample was controlled by a 12 mm mechanical shutter ( Vincent associates , Rochester , NY ) , all measurements were performed under continuous light . All imaging experiments were done at room temperature ( 20–22°C ) . For localization and co-localization , images were recorded at 10–100 ms intervals ( 100–10 Hz ) . For autocorrelation and unitary fluctuation analysis , images were acquired at 2 ms intervals ( 500 Hz ) under constant illumination at 405 nm ( 75% laser power equivalent to 8 mW s2 mm2 ) . Laser and focus control was performed using micromanager . Acquisition and digitalization was done with Andor Solis software ( Andor/Oxford Instruments , Ireland ) . Photon counts were made using a single-photon avalanche diode ( SPAD ) photon-counting camera ( 64 × 32 pixels array ) , which is space correlated with the EM-CCD camera , thus allowing for the location of limiting diffraction puncta . A continuous light stimulation of 1 µs duration was used to excite coumarin emitters and photons were collected during that period . Laser TTL triggering and piezoelectric focus control ( PIFOC-721 , PI , Germany ) was performed using Micro-Manager software ( Vale Lab , UCSD , US ) . Acquisition and digitalization was done through a custom code written in C++ ( SPADlab at POLIMI; http://www . everyphotoncounts . com/ ) . The code used to readout the SPAD array is provided as source code in Dryad repository ( https://doi . org/10 . 5061/dryad . 1kc2c ) . For localization and co-localization , the set of frames was averaged to increase the signal-to-noise ratio ( SNR=μ/σ ) , which is usually less than two for the single frame at 75% laser illumination . Noise analysis of the steady-state signal by autocorrelation was performed as employed in the past to investigate the kinetic properties of ion channel unitary events from recordings with low signal-to-noise ratios ( Anderson and Stevens , 1973; Zingsheim and Neher , 1974 ) . For autocorrelation , a post-acquisition Gaussian digital filter was used at one fourth of the acquisition frequency . The background signal was averaged and the standard deviation ( SD ) calculated . Unitary fluctuation analysis: when analyzing the fluctuations of fluorescence , we aim to avoid false-positive signals and therefore defined 2 . 5 SD ( μ ±2 . 5 σ = 0 . 987 ) of the background signal ( or the equivalent base level ) as a threshold to discriminate the two regimes ( Peterson and Harris , 2010 ) , basal level ( L0 ) and higher signal level ( L1 ) . We reasoned that fluorescence fluctuations , intrinsic to the probe , might interfere with our calculations; Soluble coumarin deposited on a glass coverslip showed12% of positive transitions , a value we linearly subtracted from all the individual points taken from the recorded mutants ( Y671coum or W426coum ) , so that a new corrected dataset having the apparent PL1 was obtained after subtraction . The fluorescence of the single emitter bleaches in less than 1 min of continuous illumination at 75% laser power . Normalization of optical data was performed by dividing signal amplitude by the maximal signal obtained ( F/Fmax ) . Autocorrelation , idealization , and the analysis of unitary transitions was performed with pClamp 8 . 0 software ( RRID:SCR_011323 ) . Whole-cell currents were obtained from transiently transfected HEK-293T cells . Gigaseals were formed by using 2–4 MΩ borosilicate pipettes ( Warner Instruments , Hamden , CT ) . Junction potentials ( 6–8 mV ) were corrected for CsF/NaCl solutions . Seal resistance was 2–3 GΩ in all cases . Series resistance and cell capacitance were analogically compensated for directly on the amplifier . The mean maximum voltage drop was about 4 mV . Whole-cell voltage clamp was performed , and macroscopic currents from voltage steps were acquired at 20 kHz and filtered at 10 kHz . Ramp protocols were acquired at 10 kHz and filtered at 5 kHz . Currents are presented in terms of densities . Data were acquired using an Axopatch 200B and a digidata 1320 ( Molecular Devices Inc . , Sunnyvale , CA ) . The acquisition and basic analysis of the data were performed with pClamp 8 . 0 software . Single photon counting . Individual cells on one day of transfection contribute with several puncta , which are considered as replicates . Therefore the histogram presented in Figure 1g corresponds to n = 1 but comes from the average of several individual histograms ( in this case , five on each curve ) . The final statistical procedure was done by comparing the mode of the different populations , n = 4 for each condition . A p-value of 0 . 05 was considered significant when comparing the modes of the photon count ( Mann-Whitney test ) . Idealization and unitary transitions . The idealization was performed first by checking that the base level was similar to the background noise , then setting a defined threshold on two standard deviations of background noise average . The comparison between the dwell times was performed in paired experiments ( 0 versus 3 uM , n = 11 ) . The full data set for the dose-response curve was analyzed using one-way ANOVA . In general , when averaged , data are shown with the correspondent SEM . The statistical analysis was computed in Microcal OriginPro ver9 ( OriginLab ) ( RRID:SCR_002815 ) . Figures were prepared using Microcal OriginPro ver9 . VMD ( structural data ) , and ImageJ ( RRID:SCR_003070 ) . Source data is available through the Dryad repository ( https://doi . org/10 . 5061/dryad . 1kc2c ) | Cells use proteins on their surface as sensors to help them to assess and explore their environments and adapt to external conditions . The transient receptor potential ( TRP ) ion channels form one such family of proteins . Sodium , potassium and calcium ions can move through TRP channels to enter and exit cells , and by doing so trigger changes in the cell that help it respond to an external stimulus . The channels have physical “gates” that can open and close to control the flow of the ions . When the TRP channel detects a stimulus – which could take the form of specific chemicals , or a change in temperature , pressure or voltage – it changes shape , causing the gate to open . Researchers have a number of unanswered questions about how TRP channels work . Where in the channels are gates located ? How do the channels control the flow of ions , and how do they communicate with each other ? And which regions of the protein sense environmental cues ? As a result , new technologies are being developed to make it easier to study the types of rearrangements that TRP channels experience when they activate . Steinberg , Kasimova et al . have used total internal reflection microscopy – a method that images fluorescent molecules – to measure the conformational change of a single TRP channel in a living cell . This channel , called TRPV1 , senses external heat and plays an important role in pain perception . Its gate can also be opened by the pungent compound of chili pepper , capsaicin . The results of the experiments suggest that a selectivity filter region in TRPV1 channels changes its shape when the channel opens in response to capsaicin . Then , this selectivity filter appears to do double duty – it controls which types of ions pass through the channels as well as controlling their flow rate . Because of its role in pain perception , having a better understanding of how TRPV1 works will be valuable for researchers who are trying to develop new pain relief treatments . The so-called ‘seeing is believing’ method used by Steinberg , Kasimova et al . could also be used to study other membrane proteins , both to guide drug development and to improve our understanding of how cells interact with their environment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"computational",
"and",
"systems",
"biology"
] | 2017 | Conformational dynamics in TRPV1 channels reported by an encoded coumarin amino acid |
In systemic light chain amyloidosis , an overexpressed antibody light chain ( LC ) forms fibrils which deposit in organs and cause their failure . While it is well-established that mutations in the LC’s VL domain are important prerequisites , the mechanisms which render a patient LC amyloidogenic are ill-defined . In this study , we performed an in-depth analysis of the factors and mutations responsible for the pathogenic transformation of a patient-derived λ LC , by recombinantly expressing variants in E . coli . We show that proteolytic cleavage of the patient LC resulting in an isolated VL domain is essential for fibril formation . Out of 11 mutations in the patient VL , only one , a leucine to valine mutation , is responsible for fibril formation . It disrupts a hydrophobic network rendering the C-terminal segment of VL more dynamic and decreasing domain stability . Thus , the combination of proteolytic cleavage and the destabilizing mutation trigger conformational changes that turn the LC pathogenic .
Amyloid light chain ( AL ) amyloidosis is the most common form of systemic amyloidosis . It is the result of an anomalous monoclonal plasma B cell proliferation in the bone marrow which leads to a subsequent overproduction of immunoglobulin light chains ( LC ) ( Baden et al . , 2009; Gertz , 2016 ) . Normally , immunoglobulins ( Igs ) are secreted as disulfide-bonded complexes comprising two heavy chains ( HCs ) and two LCs ( Baden et al . , 2008; Feige et al . , 2010 ) . In AL patients , free LCs excessively escape cellular protein quality control and are secreted into the blood stream . From there , the amyloid precursor LCs can be taken up by cells or deposit extracellularly in organs or tissues ( Baden et al . , 2008; Falk et al . , 1997; Feige et al . , 2010; Gertz , 2016; Kastritis and Dimopoulos , 2016; Marin-Argany et al . , 2016 ) . The LCs are composed of an N-terminal variable ( VL ) and a constant domain ( CL ) . The VL sequence is the result of DNA rearrangements involving a variable ( V ) , a diversity ( D ) and a joining ( J ) gene segment ( Bernard et al . , 1978; Hozumi and Tonegawa , 1976; Maki et al . , 1980; Tonegawa , 1983 ) . At later stages of differentiation , directed hypermutation of the V region exon further increases the natural diversity of VL domains ( Feige et al . , 2010; Jung and Alt , 2004; Li et al . , 2004; Wilson et al . , 1998 ) . As a result , each AL amyloidosis patient possesses a different set of somatic mutations and hence a unique disease-responsible LC sequence ( Blancas-Mejía et al . , 2015 ) . These mutations often destabilize patient LCs compared to non-amyloidogenic LCs . The destabilization was linked to amyloid fibril formation propensity ( Baden et al . , 2008; Blancas-Mejía et al . , 2015; Hurle et al . , 1994; Raffen et al . , 1999; Ramirez-Alvarado , 2012; Wall et al . , 1999 ) . Even though full length LCs and other proteins have recently been found as part of fibrils , patient biopsies predominantly revealed VL domains that result from proteolytic cleavage of the LC as the main constituent of amyloid deposits ( Blancas-Mejía et al . , 2015; Buxbaum , 1986; Enqvist et al . , 2009; Gallo et al . , 1996; Glenner et al . , 1970; Hurle et al . , 1994; Nokwe et al . , 2014; Pepys , 2006; Simpson et al . , 2009 ) . Partial unfolding of the pathogenic VLs leads to exposed residue segments that entail the stacking capability of the domain , leading to the tightly packed , dehydrated cross-β amyloid fibril structure ( Annamalai et al . , 2016; Brumshtein et al . , 2018; Kelly , 1998; Merlini and Bellotti , 2003 ) . However , changes in stability alone cannot explain the differences in fibril formation propensity ( Nokwe et al . , 2014 ) . Since every patient suffers from a unique LC , it is generally not clear which of the mutated residues trigger the disease . Thus , the underlying reason for VL variants in AL patients is still largely enigmatic and the mechanistic principles need to be defined . In this study , we analyzed a patient-derived amyloidogenic LC truncation of the λ subtype with a view to determine the mechanism that renders it pathogenic . The patient variant ( referred to as Pat-1 ) contains 11 mutations in the VL domain compared to the corresponding germline sequence ( referred to as WT-1 ) . A comparison of mutant and germline VL regarding their structure , stability and amyloid formation propensity allowed us to identify the key mutation responsible for amyloidogenicity and to elucidate its effects on conformational dynamics that result in fibril formation .
The LC variant studied here was identified in a female patient who was diagnosed with a smoldering myeloma and AL amyloidosis at the age of 50 years at the Amyloidosis Center of the University of Heidelberg . The LC responsible for the disease was of λ subtype . The levels of the free λ LCs were highly elevated with 7572 mg/l in serum , compared to κ which was normal with 8 mg/l . The clonal plasma cell infiltration of the bone marrow was also high reaching 80% . Heart , kidney , lung , and soft tissue were positive for amyloid deposits and also clinically involved . More specifically , the cardiac stage was IV ( Kumar et al . , 2012 ) or III b ( Wechalekar et al . , 2013 ) and the renal stage was 3 ( Palladini et al . , 2015 ) indicating that the patient was suffering from advanced AL amyloidosis with poor prognosis . Indeed , 4 weeks after initiation of chemotherapy , the patient died of cardiac arrest . The primary structure of the pathogenic LC , named Pat-1 , was deduced from the sequence of the complementary DNA ( cDNA ) of the plasma cell clone that caused AL amyloidosis . The corresponding germline LC sequence , named WT-1 , was identified by database-assisted ( abYsis , IgBLAST ) primary structure alignments searching for the germline LC with the lowest number of residue changes with respect to the Pat-1 sequence ( Figure 1A ) . The two sequences differ in 11 point mutations , all located in the VL domain . Five of these are located in the constant framework regions ( FR ) and six are in the hypervariable complementarity determining regions ( CDR ) . We were interested in how frequently the mutant amino acids occur at the respective positions in the antibody repertoire . To this end , we used the Kabat classification system which provides quantitative information on the relative abundance of each amino acid at each position within antibodies ( Johnson and Wu , 2000; Wu and Kabat , 1970 ) . For 9 of the 11 positions , the mutated residues in the Pat-1 VL are less abundant than the ones present in the WT-1 sequence ( Figure 1—figure supplement 1 ) indicating potential negative effects on antibody structure and stability . The two mutated residues that show higher general frequency ( S26 and D53 ) are located in the hypervariable CDRs . In AL , either the LC , truncated LCs or the VL domain have been shown to form fibrils ( Blancas-Mejía et al . , 2015; Buxbaum , 1986; Enqvist et al . , 2009; Hurle et al . , 1994; Nokwe et al . , 2014; Pepys , 2006; Simpson et al . , 2009 ) . To determine which form is present in the case of Pat-1 , we analyzed patient-derived abdominal fat tissue containing amyloid . These deposits were previously shown to be identical to organ-specific fibrils ( Annamalai et al . , 2017 ) . Extraction of the fibrils followed by SDS-PAGE , MALDI fingerprint analysis of the specific band ( data not shown ) and its mass revealed that in this patient predominantly the VL domain was deposited ( Figure 1B , Figure 1—figure supplement 1 . As the fibril load in organs correlates to the severity of the disease in AL patients , we studied fibril formation in vitro . To test whether proteolytic processing of the LC is a prerequisite for fibril formation , we produced the respective constructs recombinantly in E . coli and monitored amyloid formation of the purified proteins . Typically , fibril formation takes a long period of time . It involves unfolding , formation of partially folded intermediates and oligomers which then assemble into fibrils . It was shown earlier that small amounts of SDS presented during the incubation accelerate the process ( Kihara et al . , 2005; Nokwe et al . , 2015; Yamamoto et al . , 2004 ) . In this study , we use SDS in concentrations that do not affect the native secondary structure ( Figure 1—figure supplement 2 ) . In these assays , monitored by thioflavin T ( ThT ) fluorescence , full length Pat-1 and WT-1 LCs did not form fibrils ( Figure 1C ) . However , incubation of the VL domain of Pat-1 revealed the presence of fibrils after a lag phase , while the WT-1 VL stayed soluble . We thus conclude that the Pat-1 VL but not Pat-1 LC is the disease-relevant amyloidogenic species . Transmission electron microscopy ( TEM ) confirmed the presence of fibrils for Pat-1 VL samples , whereas no fibrils were detected in the WT-1 VL and both full length LCs samples ( Figure 1—figure supplement 3 ) . In line with these findings , all further experiments were performed only with the VL domains . To determine potential structural differences between the two VL variants Pat-1 and WT-1 , the X-ray structure of both proteins were solved to 2 . 5 Å ( Pat-1 ) and to 1 . 55 Å ( WT-1 ) resolutions . Despite forming crystals of different space groups , both domains display highly similar structural properties ( Figure 2—figure supplement 1 ) . They exhibit the typical Ig fold consisting of nine β-sheets forming a greek-key β−barrel topology with the three hypervariable CDR loops in spatial proximity ( Figure 2A ) . At first sight , none of the mutations is located at a position that could explain the difference in fibril formation tendency . Six out of the 11 variations are found in the CDRs ( T26S , S28N , V30F , G32D , E53D and S55D ) . The five substitutions in the FRs are located in two regions . Two of them which flank the CDR3 loop are in close proximity to the VL-VH interface ( Y90F and T105S ) , without disrupting the contact surface in comparison to WT-1 . The remaining three mutations ( P15L , L81V and Q82L ) are opposite to the antigen-binding site , located close to the C-terminal part of the VL domain . Thus , judging from the positions of the mutations in the domain structure and the superposition of both crystal structures ( Figure 2—figure supplement 1 ) , no reasonable conclusions concerning their contributions to fibril formation could be drawn . To gain further insight into the consequences of the mutations on the conformation of the VL domain , we analyzed whether the variants forms dimers in solution as reported for some LCs and their fragments ( Brumshtein et al . , 2014; Rennella et al . , 2019 ) . Analytical ultracentrifugation showed that the VLs studied are monomers in solution with sedimentation coefficients of approximately ~1 . 6 s ( Figure 2B ) . To test whether the mutations affect the thermal stability of the Pat-1 and WT-1 VL , the loss of secondary structure upon heating was monitored by far-UV CD-spectroscopy ( Figure 2C ) . The melting temperatures ( Tm ) at which 50% of the protein is unfolded were determined to be 48 . 9°C and 57 . 0°C for Pat-1 and WT-1 , respectively . Furthermore , the stabilities against chemical unfolding were followed by changes in the tryptophan fluorescence intensity in the presence of increasing GdmCl concentrations . Both VL domains showed sigmoidal unfolding curves and the cm-value ( the GdmCl concentrations at which 50% of the protein is unfolded ) was found to be increased by 0 . 46 M GdmCl for WT-1 compared to Pat-1 ( Figure 2D ) . These results revealed that the patient-derived variant is significantly less stable than the germline protein . To identify the mutation ( s ) decisive for stability and amyloidogenic properties of the Pat-1 VL domain , we substituted all mutations of Pat-1 individually with the respective residue present in the WT-1 VL . When we analyzed their thermal and chemical stabilities , we found that 10 out of the 11 Pat-1 VL point mutants either ( i ) only had minor effects on the stability ( S26T , N28S and S105T ) , ( ii ) were even destabilizing the domain ( L15P , F30V and D53E ) , ( iii ) showed a negative effect on thermal stability ( D32G and D55S ) , or , iv ) showed a slight increase in chemical stability ( L82Q and F90Y ) ( Figure 3A , B ) . In contrast , the mutation V81L resulted in a substantial shift of the Pat-1 VL towards the thermal and chemical stability of WT-1 with a TM of 54 . 6°C and a cM of 1 . 36 M ( WT-1: TM = 57 . 0°C , cM = 1 . 43 M ) ( Figure 3A , B ) . Remarkably , when we performed fibril formation assays , we found that all Pat-1 point mutants except V81L formed amyloids ( Figure 3C ) , suggesting that valine at position 81 is the main cause for the pathogenicity of Pat-1 . All fibril forming variants showed an even lower halftime ( t1/2 ) of fibril formation than Pat-1 . The maximal ThT intensity , which seems to correlate to the amount of fibrils present , was in most cases equal to Pat-1 . The only exception is the F30V variant which showed an increased maximum ThT fluorescence ( Figure 3—figure supplement 2 ) . Valine and leucine are closely related amino acids; both exhibit hydrophobic characteristics and differ in only a methylene moiety . Since isoleucine is identical to leucine concerning the constituent atoms , we substituted V81 by isoleucine and analyzed this mutant regarding stability and amyloid formation propensity . Intriguingly , V81I did not exhibit a significant increase in thermal stability compared to Pat-1 ( Figure 3—figure supplement 1A ) . Furthermore , we found that V81I , unlike V81L , readily forms fibrils ( Figure 3—figure supplement 1B ) . Thus , we conclude that the protective effect observed for the V81L variant is highly specific to leucine . While the experiments so far suggest that the V81L mutation can rescue the Pat-1 VL from amyloid formation , it was an open question whether vice versa replacement of leucine by valine in the context of WT-1 would be sufficient to induce fibril formation . We therefore cloned WT-1 L81V and analyzed its properties . The thermal stability of WT-1 L81V decreased by 5 . 5°C compared to WT-1 ( Figure 3D ) and the cM value for chemical unfolding , is reduced by 0 . 42 M GdmCl ( Figure 3E ) . Hence , the introduction of valine at position 81 shifts WT-1 toward the stability of Pat-1 . In the same line , upon incubation the WT-1 L81V mutant showed an increase of ThT fluorescence indicating fibril formation ( Figure 3F ) . Together , these results show that it is possible to reconstitute the properties of the AL patient mutant by exchanging only one amino acid in the wild-type protein . The crystal structure of Pat-1 showed that V81 is in close proximity to L15 and L82 opposite of the antigen-binding site ( Figure 2A ) . Focusing on the surface properties in this region , we found that these three amino acids form a hydrophobic surface patch which is not present in the WT-1 protein ( Figure 4A , B ) . To determine the combined effect of the mutations at position 15 , 81 and 82 ( L15P , V81L and L82Q ) , we designed additional double and triple mutants carrying all possible combinations of these substitutions in Pat-1 and determined their stabilities against thermal and chemical denaturation . The combination of the substitutions at position 81 and 82 enhanced the increase in stability slightly , whereas mutating leucine at position 15 to proline in combination with the other point mutations led to a decrease in stability ( Figure 4C , D ) . Fibril formation kinetics monitored by ThT fluorescence showed that the stability changes of the analyzed VL variants correlate with the amyloid propensity . The anti-amyloidogenic effect of V81L could be preserved in combination with the L82Q mutation . The presence of L15P in the mutants increased fibril formation ( Figure 4E , Figure 4—figure supplement 1 ) . From these results , we conclude that the hydrophobic surface spot in Pat-1 plays a key role in pathogenesis . Minimizing the area by the back-mutation of the hydrophobic residue L82 to glutamine supports the anti-amyloidogenic effect of V81L . L15P alone and in all combinations tested had negative effects on stability and fibril formation . To gain insight into conformational dynamics of the Pat-1 and WT-1 VL domains , we monitored the kinetics of hydrogen/deuterium exchange in the protein backbone by mass spectrometry ( H/DX-MS ) ( Bai and Englander , 1996; Rand et al . , 2014 ) . We obtained complete sequence coverage in the H/DX-MS for both variants . Our analysis indicates that Pat-1 exhibits higher conformational dynamics as indicated by an overall higher HD/X rate ( Figure 5A ) . The most affected segment was the C-terminal framework region 4 ( FR4 , residues 101–113 ) . Additionally , the region from amino acid 11 to 20 ( FR1 ) , comprising the residue at position 15 which is part of the surface-exposed hydrophobic spot , is striking since the fractional uptake of deuterium in Pat-1 increases in this area . When the differences in fractional deuterium uptake between Pat-1 and WT-1 are mapped onto the X-ray structure of Pat-1 , it becomes obvious that the Pat-1 VL shows a higher flexibility in most parts of the domain compared to WT-1 ( Figure 5B ) . To determine the underlying molecular reasons for the differences in dynamics and stability , we analyzed side chain interactions in the structures of the VL domains using the BIOVA Discovery Studio software . We found that the mutations at position 15 , 81 and 82 result in a change in hydrophobic interactions between Pat-1 and WT-1 ( Figure 6 ) . P15 is located in FR1 , in a region that shows a conspicuous change in deuterium exchange . In WT-1 , this proline is not involved in a hydrophobic interaction network . In contrast , in the patient mutant L15 interacts with three amino acids located in FR3 , including the two other point mutations in the hydrophobic surface spot , and two residues in the C-terminal region ( V111 and L112 ) . Thus , the contact of L15 to the very dynamic C-terminal segment , as indicated by the HD/X rate , induces the enhanced conformational dynamics of the respective loop region in FR1 . V81 in Pat-1 forms six hydrophobic interactions with residues in FR2 and FR3 , two of them , the L15 and L82 point mutations , are located within this hydrophobic surface area . However , V81 contacts only one amino acid in the C-terminal segment ( V111 ) . The hydrophobic interaction pattern of the WT-1 L81 misses one interaction each in FR2 and in FR3 but two more interactions to the C-terminal region appear . Overall , this leads to three hydrophobic interactions of L81 to the C-terminal region ( V109 , V111 and L112 ) . For the mutation at position 82 , the biggest effect comes from the change of the nature of interaction from hydrophobic to hydrophilic and the resulting decrease of hydrophobic surface area . L82 can form five hydrophobic interactions with FR2-4 . Q82 however forms hydrophilic interactions to four residues , all located in FR3: R64 , E84 , D85 and E86 . We conclude that the network of interactions in the C-terminal region is markedly changed in Pat-1 by the mutations at position 15 , 81 and 82 leading to a highly dynamic C-terminal region . To further analyze the effect of the protective back mutation of V81L in Pat-1 and the destructive substitution of L15P on the C-terminal network , we performed molecular dynamics ( MD ) free energy simulations . By Umbrella Sampling ( US ) simulations , the dissociation of the C-terminal segment ( residues beyond 102 ) was induced during MD-simulations for WT-1 , Pat-1 , Pat-1 V81L and the double mutant Pat-1 V81L L15P and the associated change in free energy was calculated . The free energy of dissociating this part of the structure from the otherwise folded protein can be taken as a relative estimate for the influence of the mutation on the stability of the protein fold . The calculated free energy profiles of the dissociation of the C-terminal region are shown in Figure 7 , RMSD values are shown in Figure 7—figure supplement 1 . The free energy simulations exhibit similar dissociation free energies for the WT-1 and the single back mutation V81L , which were higher ( by ~1 kcal·mol−1 ) in comparison to Pat-1 and the double back mutation V81L L15P . This indicates an increased protein stability for WT-1 and the back mutation of Pat-1 V81L . The effect of the V81L mutation is mainly due to a small cavity that is present in the case of V81 and that is filled if V81 is replaced by a slightly larger residue ( V81L ) ( Figure 7 ) . However , a destabilizing effect due to the additional back mutation L15P could be observed . Here , the double back mutation exhibited an opposite effect compared to the single back mutation , by producing a significant reduction of the dissociation free energy ( by ~1 kcal·mol−1 ) relative to the patient mutant . This effect can be explained by a change in the local protein backbone structure ( due to the L15P ) that changes the local packing geometry and offsets the effect of the V81L mutation .
Patients suffering from systemic AL amyloidosis often show different patterns of affected organs as well as widely varying symptoms . This impedes both diagnosis and treatment ( Blancas-Mejía et al . , 2015; Gertz , 2016; Gertz and Kyle , 1997; Hurle et al . , 1994; Li et al . , 2004; Merlini and Bellotti , 2003; Palladini and Merlini , 2009; Ramirez-Alvarado , 2012; Schönland et al . , 2012 ) . Even though the importance of mutations in the VL domain of overexpressed LCs is established ( Baden et al . , 2009; Hurle et al . , 1994; Martin and Ramirez-Alvarado , 2010; Nokwe et al . , 2014 ) , the mechanisms of pathogenicity are still unclear . The molecular analysis is hampered by the presence of various mutations in VL domains from different patients and the uniqueness of each patient’s LC . Also , the importance of the proteolytic cleavage for amyloidogenicity is in most cases unclear . Our comprehensive characterization of a pathogenic LC variant allows us to answer these questions and provides a comprehensive picture of the requirements to initiate fibril formation . For Pat-1 , the full length LC is resistant to fibril formation . Proteolytic cleavage in the linker region between the VL and CL domain is a prerequisite to unleash the amyloidogenic potential of the mutations in VL . The deposition of truncations of very similar sizes in the patient’s tissue suggests a complex picture concerning the proteases involved in the generation of pathogenic LC truncations . Recent cryo-EM structures of AL fibrils showed a highly diffuse density for the C-terminal residues of the VL domain indicating a flexible orientation . Thus , the exact pathogenic truncation site between VL and CL does not seem to be crucial ( Radamaker et al . , 2019; Swuec et al . , 2019 ) . NMR studies further suggest the existence of different fibril topologies in VL amyloid fibrils . They may contain well-ordered and rigid C-terminal ends or a highly ordered hydrophobic core domain ( Hora et al . , 2017; Piehl et al . , 2017 ) . For a better understanding of the disease , we conclude that the analysis of biopsies will provide important information about the presence of full length or truncated LCs in the patient’s organs and tissues and thus about the disease-causing LC species ( Annamalai et al . , 2016; Ramirez-Alvarado , 2012; Weber et al . , 2018 ) . It is obvious that in AL , the large amount of insoluble aggregates deposited in organs interferes with their function . However , the mechanism of toxicity remains unknown . In general , there are different hypotheses on the nature of the toxic species in amyloid diseases . Besides fibrils , Oligomers formed on the pathway to highly organized amyloid structures are considered toxic ( Merlini , 2017; Riek and Eisenberg , 2016 ) . In this study , we focused on analyzing fibril formation as their presence is directly correlated to pathogenicity in AL . The results of the in vitro fibril formation assays are in line with the predominant deposition of VL species in the patient: neither the Pat-1 LC nor the WT-1 LC formed fibrils; also the WT-1 VL was resistant to amyloid formation . Only the Pat-1 VL readily formed fibrils . When viewed together , these findings suggest a strong effect of the LC context on the Pat-1 VL domain; the point mutations present in the VL domain only become effective when the CL domain has been removed ( Weber et al . , 2018 ) . The current view holds that the overall destabilization of the VL domain is a major indicator for fibril formation propensity ( Blancas-Mejía et al . , 2015; Morgan and Kelly , 2016; Nokwe et al . , 2014; Ramirez-Alvarado , 2012 ) . The patient-derived mutant Pat-1 fits this general picture . However , a global destabilization alone is not sufficient to provide hints on the underlying processes leading to fibril formation . Furthermore , also VL domains with wildtype-like stability were found to form fibrils ( Nokwe et al . , 2016 ) . For a mechanistic understanding of the disease , it is important to identify which of the sequence differences between germline and patient mutant are causative . The VL domains Pat-1 and WT-1 differ in 11 point mutations resulting in a general sequence identity of ~90% . By serially substituting each of the point mutations in Pat-1 with the corresponding amino acid present in WT-1 , we identified one specific substitution , valine to leucine mutation at position 81 , as the key modification responsible for reversing the pathogenic properties of Pat-1 and enhancing the overall stability of the VL domain . Amino acid frequencies in antibody sequences hint toward a negative effect of valine at position 81 ( 13 % compared to 77% for leucine ) ( Johnson and Wu , 2000; Wu and Kabat , 1970 ) . That this substitution at position 81 is indeed sufficient to also render the germline VL domain amyloidogenic was demonstrated by introducing the V81 residue into the sequence of WT-1 . The effect on the stability and fibril formation was very specific for the leucine residue , for example a substitution with the highly similar isoleucine residue did not yield comparable effects . Residue 81 was found to be part of a surface-exposed hydrophobic area in Pat-1 together with the two other mutated residues L15 and L82 . Since surface-exposed hydrophobic areas are known to be energetically unfavorable ( Eisenhaber and Argos , 1996; Moelbert et al . , 2004; Young et al . , 1994 ) , we investigated these amino acids regarding their stability and amyloidogenic properties . Especially the L82Q substitution seemed to be a promising candidate since this mutation changed the nature of the amino acid side chain and of possible interactions . Besides , the amino acid frequency at this position decreases from 68% for glutamine to 1% for leucine in Pat-1 , supporting the idea that this substitution could have a severe effect on domain architecture ( Wu and Kabat , 1970 ) . Against our expectations , the single L82Q substitution did not affect the stability of Pat-1 and the mutant still formed fibrils . However , a combination of L81 and Q82 increased the stabilizing effect observed for L81 and the variant remained soluble . This stresses the important role of the expanded hydrophobic surface area for amyloidogenesis that is reduced by the L82Q back mutation . Surprisingly , substituting L15 with the germline residue P15 resulted in lower thermal and chemical stabilities than the ones observed for the patient mutant . The L15P mutation was also disadvantageous in combinations with L81 and/or Q82 . All the P15-containing VL variants readily formed fibrils with an even earlier onset than the patient VL itself , stressing the negative effect of P15 . This does not fit our expectations since according to the amino acid frequency a leucine at this position occurs in only 12% of the cases , while proline is present in the germline with a frequency of 61% . Even though these results clearly show that stability changes are a major element affecting fibril formation , for a mechanistic explanation the structural consequences of the mutations need to be considered . In the crystal structure of Pat-1 , V81 exhibits six hydrophobic interactions spread over FR2-4 . These include interactions with the other two residues of the surface-exposed hydrophobic area , L15 and L82 . Likewise , L81 , present in WT-1 , exhibits six hydrophobic interactions . However , the difference is that L81 reaches three amino acids in the C-terminal region ( V109 , V111 and L112 ) , whereas V81 only extends its interactions to one residue ( V111 ) . This fits the analysis of the conformational dynamics of the two VLs by HD/X which demonstrates a highly increased flexibility of the C-terminal segment in the patient-derived mutant as the underlying cause for destabilization and amyloid formation . A leucine at this position leads to a tighter packing of the domain than a valine residue . Therefore , neighboring residues are in closer proximity and form more hydrophobic interactions leading to a stabilization of the C-terminal region . Changed hydrophobic interactions and thereby changed conformational dynamics can also explain the negative effect of P15 . The leucine of Pat-1 at position 15 is tightly imbedded into a network of hydrophobic interactions . It facilitates five hydrophobic interactions to V81 , L82 and A83 in FR3 and to V111 and L112 in the C-Terminal region . Even though these interactions increase the dynamics of the loop region around residue 15 , the additional interactions to the C-terminal segment positively affect the overall stabilization of the domain . A proline residue , on the other hand , induces a local change of the protein backbone structure and packing geometry abrogating hydrophobic interactions . This in turn affects a large segment of the protein although the VL variant still exhibits secondary , tertiary and quaternary structure comparable to the WT-1 , indicating increased local dynamics . Thus , our results demonstrate a complex interplay of interactions that is responsible for the patient variant’s pathogenicity . Our results further show that small and seemingly insignificant changes of a side chain can lead to a fatal change in intramolecular interactions , which are in the case of Pat-1 of hydrophobic nature . The assays in this study were performed under simplified conditions in vitro compared to the environment the LCs experience in the human body . In the human body , many parameter including proteolysis , interactions with plasma factors and shear forces contribute to fibril formation and complicate the understanding of the pathogenicity . Furthermore , it is well-known that amyloid deposits consist not only of the disease-causing LC but also contain many other factors like glycosaminoglycans , lipids , apolipoprotein E or other proteins . Many of these factors have already been shown to influence fibril formation in the context of amyloid diseases ( Gellermann et al . , 2005; Wyatt et al . , 2012 ) . In the future their mechanistic influences on amyloid formation need to be determined to obtain a molecular understanding of disease progression , This may also offer new perspectives for potential treatment options . In summary , our study presents a general strategy how to investigate LCs associated with AL amyloidosis . An important starting point is the determination of the patient LC sequence and the nature of the LC truncation present in the fibrils in vivo . The identification and classification of the mutated residues requires a comparison with the most homologous germline sequence and with the Kabat frequency database . This is the basis for a mutational analysis required to distinguish between active and silent mutations . The combination of structural and dynamic analyses of mutants together with fibril assays allows identifying the disease-causing residues and their contributions to stability and propensity for fibril formation . For defining the underlying molecular mechanism , analyses of the structural changes and dynamic consequences induced by the identified mutations are required . This information will contribute to better diagnosis and even treatment options .
The cDNA sequence of the light chain Pat-1 was obtained from CD138-enriched bone marrow plasma cells . The sequencing was performed as described elsewhere ( Annamalai et al . , 2016 ) . The DNA sequence was deposited in the GenBank ( https://www . ncbi . nlm . nih . gov/genbank/ ) under the accession number MK962887 . To isolate fibrils from abdominal fat tissue , 25 mg patient tissue was diced with a scalpel and washed five times with 0 . 5 mL Tris calcium buffer ( 20 mM Tris , 138 mM NaCl , 2 mM CaCl2 , 0 . 1% ( wt/vol ) NaN3 , pH 8 . 0 ) . Before each washing step , the sample was vortexed and centrifuged at 3 , 100 g for 1 min at 4°C . The supernatant was discarded . The pellet was dissolved in a solution of 5 mg/mL Clostridium histolyticum collagenase ( Sigma ) in Tris calcium buffer . The mixture was incubated overnight at 37°C and 750 rpm in a horizontal orbital shaker and thencentrifuged at 3100 g for 30 min at 4°C . The pellet was resuspended in 0 . 25 mL Tris ethylenediaminetetraacetic acid ( EDTA ) buffer ( 20 mM Tris , 140 mM NaCl , 10 mM EDTA , 0 . 1% ( wt/vol ) NaN3 , pH 8 . 0 ) and homogenized using a Kontes pellet pestle . The homogenate was centrifuged for 5 min at 3100 g at 4°C . The supernatant was removed and the homogenization step was repeated twice resulting in washing fractions 1–3 . The remaining pellet was again homogenized with the pestle in 0 . 1 mL of ice-cold water and centrifuged for 5 min at 3100 g at 4°C . The supernatant was stored as water extract one and the step was repeated four more times resulting in extracts 2–5 . DNA synthesis for the LCs and VLs of Pat-1 and WT-1 was performed by Invitrogen ( Carlsbad ) . Single mutations were introduced into the Pat-1 sequence by site-directed mutagenesis . Primers carrying the mutations were designed with NEBaseChanger and the PCR was performed according to the manufacturer’s protocol . All variants were expressed and purified as previously described ( Nokwe et al . , 2014; Simpson et al . , 2009 ) . In brief , the plasmids were transformed in E . coli BL21 ( DE3 ) -star cells and protein expression took place at 37°C overnight . Cells were harvested and inclusion bodies were prepared as previously described ( Thies and Pirkl , 2000 ) . The pellet was solubilized and unfolded in 25 mM Tris-HCl ( pH 8 ) , 5 mM EDTA , 8 M urea and 2 mM β-mercaptoethanol at room temperature for a minimum of 2 hr . The solubilized protein was loaded onto a Q-Sepharose anion exchange column equilibrated in 25 mM Tris-HCl ( pH 8 . 0 ) , 5 mM EDTA and 5 M urea . The LCs and VLs were eluted in the flow-through fractions and refolded by dialysis against 250 mM Tris-HCl ( pH 8 . 0 ) , 100 mM L-Arg , 5 mM EDTA , 1 mM oxidized glutathione and 0 . 5 mM reduced glutathione at 4°C overnight . To remove aggregates and impurities , the refolded proteins were purified using a Superdex 75 16/60 gel-filtration column ( GE Healthcare , Uppsala , Sweden ) equilibrated in PBS buffer . Recovery and purity of intact proteins were analyzed by SDS-PAGE . Initial crystallization hits for both WT-1 and Pat-1 were obtained using the vapor diffusion method at 20°C . Equal amounts of protein sample ( about 30 mg\mL , solved in 20 mM Tris-HCl , pH 7 . 5 , 40 mM NaCl ) and reservoir solution were mixed for setting up 0 . 4 µL sitting drops . WT-1 protein crystallized in presence of 0 . 2 M CaCl2 , 0 . 1 M HEPES , pH 7 . 5 , 28% PEG400 . Requiring no further cryo-protection , crystals were directly vitrified in liquid nitrogen at 100 K . A high-resolution data set was measured at beam line ID30 at the European Synchrotron Radiation Facility ( ESRF , Grenoble , France ) using radiation of λ = 0 . 98 Å . For Pat-1 , suitable crystals were grown in 2 µL hanging drops with reservoir solutions optimized around the initial crystallization condition of 0 . 5 M ( NH4 ) 2SO4 , 0 . 1 M tri-sodium-citrate , pH 5 . 6 , 1 . 0 M Li2SO4 . For all crystals , mother liquor supplemented with 40% glycerin was used for cryo-protection . To obtain experimental phasing information , crystals were soaked for 4 hr in drops containing potassium tetrachloroaureate- ( III ) -hydrate dissolved in mother liquor . An anomalous data set was measured at the peak wavelength of the Au ( L-III ) edge ( λ = 1 . 039 Å , f'=−17 . 0 , f''=10 ) at beam line X06SA at the Paul Scherrer Institute , Swiss Light Source ( Villingen , Switzerland ) . In addition , a native data set was recorded at beam line ID30 at the ESFR . For all data sets , initial analysis , data processing , scaling and reduction were performed using the XDS software package ( Kabsch , 1993 ) . Further structure determination made use of different programs , distributed together by the ccp4i program suite ( Winn et al . , 2011 ) . The WT-1 structure was solved to 1 . 55 Å resolution by Patterson search calculation techniques using PHASER ( McCoy et al . , 2007 ) and the atomic coordinates of PDB entry 4NKI ( Hao et al . , 2015 ) as a starting model . Applying the experimental phase information provided by the Au-SAD data set measured from the Pat-1 crystal , the automated structure solution pipeline Crank2 ( Skubák and Pannu , 2013 ) generated an initial model to 3 . 5 Å resolution . This , in turn , proved to be suitable for solving the native Pat-1 data set , thereby expanding the phase information to 2 . 5 Å . Iterative rounds of model building and refinement with Coot ( Emsley et al . , 2010 ) and Refmac5 ( Vagin et al . , 2004 ) followed by addition of water molecules with applying the ARP/wARP software package ( Perrakis et al . , 1997 ) further improved the structure models for both WT-1 and Pat-1 . This resulted in final Rvalues of Rwork = 14 . 2% and Rfree = 16 . 5% or Rwork = 19 . 8% and Rfree = 24 . 5% , respectively . Besides , both models were found to have good stereochemistry as analyzed by Molprobity ( Chen et al . , 2010 ) . Further details regarding data collection and refinement are listed in Figure 2—source data 1 . Atomic coordinates and structure factors for WT-1 and Pat-1 have been deposited in the RCSB Protein Data Bank under the PDB IDs 6SM1 and 6SM2 , respectively . Thermal transitions were recorded using a Jasco J-715 spectropolarimeter ( Jasco , Grossumstadt , Germany ) equipped with a Peltier element . Protein unfolding was followed by monitoring the signal change at 205 nm at a heating rate of 30°/h . All measurements were performed using a 10 µM protein solution in a quartz cuvette with 1 mm pathlength . Tryptophan fluorescence measurements were carried out in a 10 × 2 mm quartz cuvette using a FluoroMax-4 spectrofluorometer ( Horiba Jobin Yvon , Bensheim , Germany ) . The measurements were performed with slit widths of 3 nm for excitation and 4 nm for emission , respectively . The protein concentration was 1 µM and the temperature 20°C . Unfolding transitions were carried out by denaturing the samples overnight in GdmCl concentrations from 0 to 4 M . The fluorescence intensity was measured at 349 nm every second for 50 s , and the average was taken . Analysis of the data was carried out assuming a two-state unfolding as described previously ( Pace , 1986; Santoro and Bolen , 1988 ) . AUC measurements were carried out using a ProteomLab XL-I ( Beckman , Krefeld , Germany ) equipped with absorbance optics . The protein concentration for the measurements was 20 µM in PBS buffer . A total of 350 µL per sample was loaded into assembled cells with quartz windows and 12-mm-path-length charcoal-filled epon double-sector centerpieces . The measurements took place at 42 , 000 rpm in an eight-hole Beckman-Coulter AN50-ti rotor at 20°C . Sedimentation was continuously scanned with a radial resolution of 30 µm and monitored at 280 nm . Data analysis was carried out with SEDFIT using the continuous c ( S ) distribution mode ( Brown and Schuck , 2006; Schuck , 2000 ) . ThT assays were performed in black 96 well microplates ( #437112 , Nunc , ThermoFisher Scientific , Roskilde , Denmark ) . The fibril formation kinetics were followed by measuring every plate well at 440 nm excitation and 480 nm emission wavelengths every 30 min with a Tecan Genios plate reader ( Tecan Group Ltd . , Männedorf , Switzerland ) ( Gade Malmos et al . , 2017 ) . To remove aggregates and oligomers and prevent seed formation during the assay , monomer isolation was performed prior to the experiment by ultracentrifugation in an Optima MAX-E ultracentrifuge , ( Beckman , Krefeld , Germany ) . Assays were performed in a final volume of 250 µL per well with 20 µM protein 10 µM ThT in PBS buffer ( pH 7 . 4 ) containing 0 . 5 mM SDS to support fibril formation ( Kihara et al . , 2005; Nokwe et al . , 2015; Yamamoto et al . , 2004 ) and 0 . 05% NaN3 . Microplates were covered with a Crystal Clear PP sealing foil ( HJ-Bioanalytik GmbH , Erkelenz , Germany ) and kept in the plate reader at 37°C under continuous orbital shaking of 180 rpm . To obtain TEM micrographs , 10 µL samples were taken from the completed ThT-assay wells , applied onto a 200-mesh activated copper grid and incubated for 1 min . The samples were washed with 20 µL H2O and negatively stained with 8 µl of a 1 . 5% ( w/v ) uranyl acetate solution for 1 min . Excess solutions were removed using a filter paper . Micrographs were recorded on a JEOL JEM-1400 Plus transmission electron microscope ( JEOL Germany GmbH , Freising , Germany ) at 120 kV . H/DX-MS experiments were performed on a fully automated system equipped with a Leap robot ( HTS PAL; Leap Technologies , NC ) , a Waters ACQUITY M-Class UPLC , a H/DX manager ( Waters Corp . , Milford , MA ) and a Synapt G2-S mass spectrometer ( Waters Corp . , Milford , MA ) , as described elsewhere ( Zhang et al . , 2014 ) . The protein samples were diluted in a ratio of 1:20 with deuterium oxide containing PBS buffer ( pH 7 . 4 ) and incubated for 0 s , 10 s , 1 min , 10 min , 30 min or 2 hr . The exchange was stopped by diluting the labeled protein 1:1 in quenching buffer ( 200 mM Na2HPO4 × 2 H2O , 200 mM NaH2PO4 × 2H2O , 250 mM Tris ( 2-carboxyethyl ) phosphine , 3 M GdmCl , pH 2 . 2 ) at 1°C . Digestion was performed on-line using an immobilized Waters Enzymate BEH Pepsin Column ( 2 . 1 × 30 mm ) at 20°C . Peptides were trapped and separated at 0°C on a Waters AQUITY UPLC BEH C18 column ( 1 . 7 µm , 1 . 0 × 100 mm ) by a H2O to acetonitrile gradient with both eluents containing 0 . 1% formic acid ( v/v ) . Eluting peptides were subjected to the Synapt TOF mass spectrometer by electrospray ionization . Samples were pipetted by a LEAP autosampler ( HTS PAL; Leap Technologies , NC ) . Data analysis was conducted with the Waters Protein Lynx Global Server PLGs ( version 3 . 0 . 3 ) and DynamX ( Version 3 . 0 ) software package . Molecular dynamics ( MD ) simulations were performed employing the Amber16 simulation package ( Case et al . , 2016 ) . In order to analyse the stability of the VL domain under specific mutations , Umbrella sampling ( US ) simulations were carried out with the pmemd . cuda module of the Amber16 package . The simulated constructs were the wild type WT-1 , the patient mutant Pat-1 , the single back mutation of the patient mutant Pat-1 V81L and the double back mutation of the patient mutant Pat-1 V81L L15P . For this purpose , the Amber ff14SB force field and the TIP3P solvent model ( Jorgensen et al . , 1983 ) were used . Each construct was solvated in water in a periodic solvent octahedron box with a minimum distance of 10 . 0 Å between each atom of the protein and the edge of the periodic box . Apart from that , a neutralization of the solution was achieved by adding Na+ and Cl- ions . Relaxation of the structure was carried out with an energy minimization of maximum 1500 minimization cycles . Each system was then heated up in steps of 100K , for 10ps each , until a temperature of 300 K was reached , whereby the system was harmonically restraint to the start structure with a restraint force constant of 25 . 0 kcal/molŲ . Afterwards , the system was equilibrated by gradually reducing the restraint force , in five steps of 10ps each , to a restraint force constant of 0 . 5 kcal/molŲ . For the heating and equilibration , MD-steps of 2 ps were used . Subsequently , hydrogen mass repartitioning ( HMR ) was performed in order to enable an increment in the simulation time step from 2 ps to 4 ps for the production simulations ( Hopkins et al . , 2015 ) . The US method allows the system to overcome an energy barrier by implementing an additional quadratic restraining potential to guide the system along a selected reaction coordinate . It possible to extract the free energy along the reaction coordinate of interest in which the configurations vary energetically and the system overcomes possible energy barrier . For the purpose of obtaining information about the folding stability of the protein , we simulated the system along a path in which the C-terminus dissociates from the protein . For this , we divided the dissociation path in 19 umbrella windows , which vary in the used harmonic restraining potentials , that is the restraining force constant ( K ) and the reference value around which the system is forced to stay close to ( dref ) . These penalty potentials were then selected in such a manner that the distribution of states was shifted along the reaction coordinate , the distribution of states converged around the desired reference value for each umbrella window and allowing for sufficient overlap between neighboring distributions . The reaction coordinate was in this case the distance between the center of mass of the C-terminal ( residues 103 to 109 ) and the remaining residues of the protein ( 1 to 102 ) . The sets of umbrella windows generated were the following: i ) 11 consecutive simulations with dref varying between 12 . 0 Å and 22 . 0 Å with a step of 1 . 0 Å and a force constant of K = 2 . 5 kcal / Å2 mol , ii ) eight consecutive simulations with dref varying between 14 . 0 Å and 17 . 5 Å with a step of 0 . 5 Å and a force constant of K = 4 . 0 kcal / Å2 mol . Here , positional restraints with a force constant of Kpos = 0 . 05 kcal / Å2 were applied to the alpha carbons ( Cα ) of residues 1 to 102 . Each umbrella window was simulated for 100ns , but only for the last 50 ns of each window a trajectory was written and used to calculate the free energy profile ( or potential-of-mean-force: PMF ) by means of the WHAM algorithm ( Kumar et al . , 1992 ) along the reaction coordinate . | Amyloid light chain amyloidosis , shortened to AL amyloidosis , is a rare and often fatal disease . It is caused by a disorder of the bone marrow . Usually , cells in the bone marrow produce Y-shaped proteins called antibodies to fight infections . In AL amyloidosis , these cells release too much of the short arm of the antibody , known as its light chain , and the light chains also carry mutations . The antibodies are no longer able to assemble properly , and instead misfold and form structures , known as amyloid fibrils . The fibrils build up outside the cells , gradually causing damage to tissues and organs that can lead to life-threatening organ failure . Due to the rareness of the disease , diagnosis is often overlooked and delayed . People experience widely varying symptoms , depending on the organs affected . Also , given the diversity of antibodies people make , every person with AL amyloidosis has a variety of mutations implicated in their disease . It is thought that mutations in the antibody light chain make it unstable and prone to misfolding , but it remains unclear which specific mutations trigger a cascade of amyloid fibril formation . Now , Kazman et al . have pinpointed the exact mechanism in one case of the disease . First , tissue biopsies from a woman with advanced AL amyloidosis were analyzed , and the defunct antibody light chain was isolated . Eleven mutations were identified in the antibody light chain , only one of which was found to be responsible for the formation of the harmful fibrils . The next step was to determine how this one small change was so damaging . The experiments showed that after the antibody light chain was cut in two , a process that happens naturally in the body , this single mutation transforms it into a protein capable of causing disease . In this ‘bedside to lab bench’ study , Kazman et al . have succeeded in determining the molecular origin of one case of AL amyloidosis . The results have also shown that the instability of antibodies due to mutation does not alone explain the formation of amyloid fibrils in this disease and that the cutting of this protein in two is also important . It is hoped that , in the long run , this work will lead to new diagnostics and treatment options for people with AL amyloidosis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2020 | Fatal amyloid formation in a patient’s antibody light chain is caused by a single point mutation |
Axonal branching allows a neuron to connect to several targets , increasing neuronal circuit complexity . While axonal branching is well described , the mechanisms that control it remain largely unknown . We find that in the Drosophila CNS branches develop through a process of excessive growth followed by pruning . In vivo high-resolution live imaging of developing brains as well as loss and gain of function experiments show that activation of Epidermal Growth Factor Receptor ( EGFR ) is necessary for branch dynamics and the final branching pattern . Live imaging also reveals that intrinsic asymmetry in EGFR localization regulates the balance between dynamic and static filopodia . Elimination of signaling asymmetry by either loss or gain of EGFR function results in reduced dynamics leading to excessive branch formation . In summary , we propose that the dynamic process of axon branch development is mediated by differential local distribution of signaling receptors .
The establishment of functional neuronal networks relies on the correct incorporation of a neuron into a developing circuit . An extended neurite network enables a single neuron to process information from multiple input cells and to relay that information to a wide range of targets . Neurite formation during development is a dynamic process and therefore tight regulation seems necessary to achieve connection specificity . At earlier steps of circuit formation , axon guidance , an intensively investigated process , combines intrinsic factors and extracellular cues to form a trajectory towards the general target area ( Williams et al . , 2003; Schnorrer and Dickson , 2004; Kolodkin and Tessier-Lavigne , 2011; Pappu et al . , 2011 ) . Subsequently , the formation of precise axonal connections within the target area relies on the development of the correct number of axonal branches . Currently , the mechanisms regulating axonal branch number and accuracy are largely unknown and subject to much debate . In mammals , a common mechanism to regulate axon branch number is excessive axonal outgrowth and exuberant branch formation during development followed by a refinement process called pruning ( Low and Cheng , 2006 ) . Pruning encompasses the removal of relatively short axon terminals and branch arbors innervating a common target area as seen in the mouse peripheral and central nervous systems ( Sanes and Lichtman , 1999; Hashimoto et al . , 2009 ) . In addition , long axon collaterals innervating distant target areas occurring for example in corticospinal tract ( CST ) axons of layer V neurons can be eliminated ( Weimann et al . , 1999 ) . Removal of short redundant or inappropriate branches occurs typically via retraction of short branches whereas longer tracts are eliminated primarily by degeneration ( Luo and O’Leary , 2005 ) . A process involving features of both pruning mechanisms , termed axosome shedding , has been observed in mammals ( Bishop et al . , 2004 ) . An important question is how branch refinement is regulated . For a long time activity-dependent mechanisms were thought to be the major factor underlying regulation of pruning in the mammalian system ( McLaughlin et al . , 2003; Yu et al . , 2004; Huberman et al . , 2006; Hashimoto et al . , 2009 ) . However , several studies in various vertebrate systems suggest that this may not be universally true ( Crowley and Katz , 2000; Bagri et al . , 2003; Pfeiffenberger et al . , 2006; Cang et al . , 2008; Sun et al . , 2011; Wei et al . , 2011 ) . Thus , although there is ample description of axonal branch refinement in vertebrate systems , much remains to be elucidated about the mechanisms underlying them . In Drosophila deterministic genetic programs are thought to account for the stereotypic development of the vast majority of neuronal connections ( Jefferis et al . , 2001; Hiesinger et al . , 2006 ) . Nevertheless , a specialized form of pruning also occurs in Drosophila , namely the widely studied remodeling of insect networks during metamorphosis . In holometabolous insects , like the fruit fly , many cells need to accommodate two distinct morphological and behavioral states within a lifetime . In the nervous system neuronal arbors have to remodel extensively to allow the reiterative use of larval neuronal populations to form adult circuits . Interestingly , the molting hormone Ecdysone is not only necessary for body transformation but also for the regulation of remodeling events in the nervous system ( Truman , 1990 ) . This system resembles partially the emergence of an adult network from initial projections as seen in vertebrates in the visual and motor cortex ( O’Leary and Koester , 1993 ) . In this study , we focus on axonal branch refinement of the dorsal cluster neurons ( DCNs ) in the central nervous system ( CNS ) of Drosophila ( Hassan et al . , 2000 ) . DCNs form only adult-specific neuronal projections and therefore unlike sensory neurons ( Williams et al . , 2006 ) and mushroom body neurons ( Boulanger et al . , 2011 ) , DCN axons are not remodeled during metamorphosis . DCN axons innervate the optic lobes via an initial phase of long-range axonal growth and retraction steps , followed by the establishment of a stereotypic number of axonal branches by an unknown mechanism . In this work , we first describe that this wiring pattern is achieved through initially excessive axonal branch growth followed by refinement during brain development . Next , we show that the refinement process is regulated through local activation of EGFR signaling in part by EGF-secreting sensory axons . We find that EGFR shows intrinsic differential distribution between individual developing DCN axonal branches and that the appropriate level of signaling is required for proper axonal branching . Mechanistically , we find that , in this context , the EGFR acts via regulating actin cytoskeleton dynamics , and not the canonical mitogen activated kinase ( MAPK ) pathway . Finally , high-resolution 4D live imaging of pupal brain explants shows that inhibition of EGFR signaling causes a dramatic reduction in axonal branch dynamics leading to the failure of axonal branch pruning .
The dorsal cluster neurons ( DCNs ) establish a complex neurite network in the Drosophila adult optic lobes . A small subset of neurons from this cluster extend their axons in the outer part of the optic lobe , the medulla ( Me ) ( Srahna et al . , 2006; Langen et al . , 2013 ) , where they form a stereotypic pattern of axonal branches ( Figure 1A , B ) . This pattern can be readily visualized using the ato-Gal4 driver in combination with a UAS-driven marker of choice such as CD8-GFP . Flip-out single cell clones ( Wong et al . , 2002 ) reveal the branch pattern of an individual axon derived from a single neuron of the 12 medulla innervating DCNs ( Figure 1C ) . False color labeling and tracing ( Longair et al . , 2011 ) of single DCN Me axons and their branches ( Figure 1D ) reveals that each axon generates 6–8 primary branches , with a mean of 7 branches . This stereotypic pattern is achieved by hot spots of branches extending in dorsal and ventral direction from each main axon shaft . The first main branch point is located at the border between lobula and Me with one or two branches . The next major branch point with often two branches is situated in Me layers M7–M8 and in this location branches from distinct neighboring axons are often in close contact forming a grid-like pattern . The terminal set of up to four branches is distributed over the M1–M3 layers and is more often intermingled with neighboring axon branches . In between the two most distal branch points intermediate branches occur occasionally . DCN branches never extend beyond the Me neuropil . 10 . 7554/eLife . 01699 . 003Figure 1 . The axonal network of medulla dorsal cluster neurons ( DCNs ) in the adult central nervous system of Drosophila . ( A ) Dorsal cluster neurons , labeled with lacZ ( red ) using the atoGal4-14a driver , with its dendritic and axonal projections in the optic lobes of the CNS . Using the FLP-out system , an individual neuron is labeled with mCD8-GFP ( green ) within the background of the entire cluster . AtoGal4-14a is used in all the following experiments except when stated otherwise . ( B ) DCN axons , labeled with mCD8-GFP , form a stereotypic pattern of axonal branches within the medulla ( Me ) of the adult optic lobe . ( C ) Using the FLP-out system the axon and branches of an individual neuron are labeled with mCD8-GFP ( green ) within the background of the entire cluster labeled with lacZ ( red ) . ( D ) False color labeling of one Me DCN axon with its main shaft ( green ) and branches ( magenta ) using a tracer tool . The scale bars represent 100 µm in ( A ) and 20 µm in ( B–D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 003 We carried out a targeted screen using loss and gain of function transgenes for signal transduction and axon guidance receptors to identify pathways that might regulate axon branch development . We noted excessive branching in the adult DCNs using a dominant-negative construct of the EGFR . To validate these findings we first analyzed flies carrying a viable hypomorphic loss of function mutation for the receptor ( EGFRT1 ) . In this genetic background DCN axons show short ectopic branches ( Figure 2A ) highlighted using the tracing tool ( Figure 2A′ ) . Since the proper development of the optic lobes depends on EGFR signaling ( Huang et al . , 1998 ) , reduced EGFR signaling might indirectly influence DCN axon formation and morphology . To investigate whether the EGFR is required in the DCNs for axonal branch refinement , we sought to generate DCN MARCM EGFR-null clones ( Lee and Luo , 1999 ) , whereby EGFR function is removed at the time of neuronal birth . We obtained very few clones , suggesting that the EGFR may be required early during development for cell viability . The clones we did obtain showed ectopic branching defects , but also severe axon targeting phenotypes , suggesting that the EGFR is required early in DCN development and precluding further analysis of these clones ( Figure 2—figure supplement 1 ) . To avoid these early defects , we used the ato-Gal4 driver , which is expressed in postmitotic DCNs after the initiation of axonal outgrowth ( Srahna et al . , 2006; Zheng et al . , 2006; Langen et al . , 2013 ) , to express two different dominant negative alleles of the EGFR ( uas-EGFRDN-A , Freeman , 1996; uas-EGFRDN-B , Buff et al . , 1998 ) and EGFRRNAi ( uas-EGFRRNAi , VDRC107130 ) . In all three cases the DCNs show a significant increase of axon branches in the adult CNS . Compared to an average of 7 primary branches under wild type conditions , we observed a significant increase to 10 . 5 primary branches per axon in EGFRDN-A expressing DCNs ( Figure 2B , B′ , E ) . Single cell clones in wild type ( Figure 2F ) and EGFRDN−A background ( Figure 2G ) show the branch increase on single cell level . Expression of the second , weaker , EGFRDN allele ( Urban et al . , 2004 ) ( EGFRDN-B ) resulted in an increase to an average of 8 . 3 branches per axon ( Figure 2C , C′ , E ) , and EGFR knock-down with RNAi leads to a similar increase to 8 . 5 branches per axon ( Figure 2D , D′ , E ) . In the case of the EGFRDN-A axonal branches appear thin and spike-like suggesting that they are immature . Interestingly , inhibition of EGFR results not only in an increase of the average branch number , but also increases the variability in the branch numbers between individual axons ( Figure 2—figure supplement 3 ) , even within the same individual brain , suggesting that EGFR signaling may regulate the accuracy and robustness of the branching process . 10 . 7554/eLife . 01699 . 004Figure 2 . EGF-receptor downregulation in the DCNs results in excessive axonal branches in the adult . ( A ) The homozygous hypomorphic allele EGFRT1 shows additional , short branches . ( B–D ) Downregulation of the EGFR specifically in the DCNs results in an increase of adult branches via overexpression of ( B ) a dominant-negative form A ( UAS-EGFRDN-A ) , ( C ) a dominant-negative form B ( UAS-EGFRDN-B ) and ( D ) a RNAi against EGFR ( UAS-EGFRRNAi ) . ( A′–D′ ) Visualization of branches ( purple ) along a single main axon shaft ( green ) using the tracing tool reveals excessive branches of the aforementioned genotypes in ( A–D ) . ( E ) Quantification of adult primary branch numbers per axon for the genotypes shown in ( B–D ) shows significant increase of branches . Control 6 . 96 ± 1 . 34 ( n = 60 ) , EGFRDN-A 10 . 5 ± 2 . 5 ( n = 45 , p<0 . 001 ) EGFRDN-B 8 . 3 ± 1 . 46 ( n = 40 , p<0 . 001 ) , EGFRRNAi 8 . 5 ± 1 . 09 ( n = 40 , p<0 . 001 ) . ( F–G ) Adult Drosophila brain in which the neuropil is marked with DN-Cad ( red ) . Flip out DCN clones are generated in control ( F ) and EGFRDN−A ( G ) background . Error bars represent SEM . Non-parametric ANOVA Kruskal–Wallis test . ***p<0 . 001 . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 00410 . 7554/eLife . 01699 . 005Figure 2—figure supplement 1 . EGFRnull MARCM clones show early branch growth defects . ( A ) Example of an adult Drosophila brain in which a single GFP-positive EGFR mutant DCN is generated using the MARCM technique . ( A′ ) High magnification of the axonal projection of the clone in ( A ) in the contralateral brain half . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 00510 . 7554/eLife . 01699 . 006Figure 2—figure supplement 2 . Spi release from photoreceptor axons regulates DCN axon branch pruning . ( A ) The homozygous , hypomorphic allele Spiscp2 shows ectopic short branches . ( A′ ) Visualization of branches ( purple ) along a single main axon shaft ( green ) , using the tracing tool . ( B–D ) Co-innervation of DCN axon branches ( green ) and photoreceptor axons ( anti-chaoptin , red ) in the medulla at 36 hr APF ( B ) , at 48 hr APF ( C ) and adult ( D ) . ( E ) Vein-LacZ expression in unknown cells , probably glia , very close to the more proximal branches of the DCNs , suggesting that Vein may be the other ligand activating EGFR in the DCNs . ( F–G ) No difference in branch number was observed after downregulation of Spi using UAS-SpiRNAi in the DCNs using the ato-Gal4-14a . ( H–I ) Spi downregulation in photoreceptors using the GMR-Gal4 driver causes a significant increase in DCN branches whereby DCN axons are labeled with LexAop-myr-GFP driven by an IMAGO LexA knock-in into the atonal locus ( atolexA ) . ( J ) Analysis of the RNAi experiment represented in ( F–I ) . Significant increase of DCN branch numbers occurs only after photoreceptor specific Spi downregulation . Control ( atoGal4-14a>UAS-CD8-GFP ) 5 . 6 ± 1 . 1 ( n = 12 ) , SpiRNAi ( atoGal4-14a>UAS-SpiRNAi ) 5 . 9 ± 1 . 1 ( n = 12 ) , control ( GMR-Gal4>UAS-CD8-GFP ) 5 . 42 ± 0 . 9 ( n = 12 ) , SpiRNAi ( GMR-Gal4>UAS-SpiRNAi ) 7 . 17 ± 2 . 17 ( n = 12 ) . Error bars represent SEM . t test . *p<0 . 05 . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 00610 . 7554/eLife . 01699 . 007Figure 2—figure supplement 3 . Distribution of axon branch numbers in control and EGFR-DN flies . Inhibition of EGFR signalling in the DCNs increases both the average number and the variability of axonal branch numbers . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 007 Activation of the EGFR requires binding to its EGF ligands . To confirm that EGFR signaling regulates DCN axon branching , we first tested adult hypomorphic mutants for the EGFR ligand Spitz ( Spi ) whose role in optic lobe development is well described ( reviewed in Salecker et al . , 1998 ) . Reduction of Spi activity results in ectopic short branches indistinguishable from those seen in EGFR hypomorphic mutants ( Figure 2—figure supplement 2A , A′; compare to Figure 2A , A′ ) . To determine the source of the EGF signal that regulates DCN branch refinement , we considered two possibilities . First , DCN axons themselves might release an activating ligand to initiate an autocrine signaling mechanism , as seen in the p75-TNR axo-axonal competition of mouse and rat sympathetic axons innervating the eye ( Singh et al . , 2008 ) . Second , neurons in the target neuropil might release EGF to regulate branch refinement . A subset of retinal photoreceptors known as R8 and R7 have axon terminals that innervate the medulla . Photoreceptors are known to secrete Spi to initiate a number of EGFR-dependent events in the developing optic lobes ( Huang and Kunes , 1998; Huang et al . , 1998; Yogev et al . , 2010 ) . We analyzed the coincidence of innervation of the medulla by R7 and R8 photoreceptor axons using the photoreceptor specific marker mAb 24B10 ( Fujita et al . , 1982 ) . Overlap between DCN and R7/8 axons can be seen at different times during brain development and in the adult brain ( Figure 2—figure supplement 2B–D ) . To distinguish between the two models , we used SpiRNAi driven by either ato-Gal4 ( DCNs ) or GMR-Gal4 ( photoreceptors ) to down regulate Spi expression in the DCNs or photoreceptors , respectively . To visualize DCN branch formation while down regulating Spi specifically in the photoreceptors , we used the Gal4-independent LexA-based binary expression system ( Lai and Lee , 2006 ) . Specifically , we took advantage of the atoLexA IMAGO ( Choi et al . , 2009 ) knock-in allele we recently generated ( Langen et al . , 2013 ) and used it to drive LexAop-GFP expression in DCNs . Whereas we find no significant difference in branch number upon knock-down of Spi in the DCNs ( Figure 2—figure supplement 2F , G , J ) , Spi knock-down in photoreceptors causes a significant increase in DCN branches ( Figure 2—figure supplement 2H , I , J ) . In addition , to Spi release from photoreceptors , we observed cells expressing a reporter for the EGFR ligand Vein in close proximity to DCN axons ( Figure 2—figure supplement 2E ) , suggesting a second source of EGF within the brain . Taken together , these results show that EGFR signaling regulates DCN axonal branch development . In theory , the adult branching pattern of DCN medulla axons can be established via one of at least two distinct mechanisms during development . On the one hand , accurate target innervation might proceed via the direct formation of the correct number of branches . Alternatively , the specificity of axonal branching might be the result of initial excessive outgrowth and exuberant branch formation during development followed by a refinement process to eliminate the majority of branches , as in refinement observed in mammalian visual map formation ( Feldheim and O’Leary , 2010 ) , for example . To distinguish between these two models , we characterized branching of wild type DCN axons at different time points after puparium formation ( APF ) during brain development . Between 36 hr and 54 hr APF DCN axons form extensive branches at multiple positions along the growing axon ( Figure 3A–C ) . Between 60 hr and 72 hr APF pruning begins to be evident ( Figure 3D–F′ ) . At 84 hr APF , the eventual adult branch pattern of 6–8 branches is apparent ( Figure 3G ) and little or no further pruning appears to occur beyond that point ( Figure 3H , I ) . This developmental pattern is not an artifact of the expression of the membrane bound marker CD8-GFP , as two other intracellular axonal markers ( nSyb-GFP and Syt-GFP ) yield the same results ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 01699 . 008Figure 3 . Loss of EGFR function impairs developmental axon branch pruning . ( A–I ) Axonal branch pattern at different pupal stages shows excessive branching at early to mid-pupal development . Successive refinement of exuberant branches can be observed between 60 hr and 96 hr ( arrowhead , compare D–H ) . Branch morphology at ( A ) 36 hr APF , ( B and B′ ) 48 hr APF , ( C ) 54 hr APF , ( D ) 60 hr APF , ( E ) 64 hr APF , ( F and F′ ) 72 hr APF , ( G ) 84 hr APF , ( H ) 96 hr APF and ( I ) adult stage . High magnification of branches is shown in B′ and F′ . ( J–O ) Axonal branch pattern at different pupal stages of EGFRDN expressing DCNs shows excessive branching at early to mid-pupal time points similar to wild type . Impaired refinement of exuberant branches can be observed between 60 hr and 96 hr ( arrow , compare L–O ) . Branch morphology at ( J ) 36 hr APF , ( K and K′ ) 48 hr APF , ( L ) 60 hr APF , ( M and M′ ) 72 hr APF , ( N ) 84 hr APF , ( O ) 96 hr APF . High magnification of branches is shown in K′ and M′ . ( P ) Quantification of branches at the second branch point at 48 hr and 72 hr APF comparing control and EGFRDN using the Skeleton Analysis tool of ImageJ ( ‘Materials and methods’ ) . EGFR downregulation does not result in increased branches at 48 hr APF compared to control . Significant decrease of developmental branch numbers at 72 hr APF occurs due to refinement in control . No significant decrease in branch number was observed after EGFR downregulation between 48 hr and 72 hr APF . Compared to control more branches persist after EGFR downregulation at 72 hr APF . Control ( 48 hr APF ) 49 . 33 ± 9 . 87 ( n = 18 ) , control ( 72 hr APF ) 22 . 75 ± 9 . 1 ( n = 18 , p<0 . 01 ) , EGFRDN ( 48 hr APF ) 45 . 77 ± 10 . 96 ( n = 16 ) , EGFRDN ( 72 hr APF ) 37 . 3 ± 3 . 83 ( n = 14 ) ( to control 72 hr APF , p<0 . 05 ) . Error bars represent SEM . t test . *p<0 . 05; **p<0 . 01 . The scale bars represent 20 µm except in B′ , K′ and M′ with 10 µm . ( Q ) Schematic representation of the role of EGFR signaling in DCN axonal branch formation . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 00810 . 7554/eLife . 01699 . 009Figure 3—figure supplement 1 . DCN axon branches . ( A–F ) DCN axon branches labeled with nSyb-GFP during pupal development at ( A ) 36 hr APF , ( B ) 48 hr APF , ( C ) 60 hr APF , ( D ) 72 hr APF , ( E ) 84 hr APF , ( F ) 96 hr APF . ( G–L ) DCN axon branches labeled with nSyb-GFP during pupal development at ( G ) 36 hr APF , ( H ) 48 hr APF , ( I ) 60 hr APF , ( J ) 72 hr APF , ( K ) 84 hr APF , ( L ) 96 hr APF . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 00910 . 7554/eLife . 01699 . 010Figure 3—figure supplement 2 . Branch growth is not enhanced in aged EGFRDN flies . ( A ) DCN axon branching ( green ) of 2-days-old EGFRDN flies does not exceed the neuropil marked with DN-cadherin ( red ) . ( B ) DCN axon branching ( green ) of aged EGFRDN flies at 18 days does not exceed the neuropil marked with DN-cadherin ( red ) . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 010 Excessive axonal branches in EGFR mutant adults may be the result either of increased branch growth or of failure of branch pruning . To distinguish between these two possibilities , we analyzed DCNs expressing EGFRDN during pupal development . Between 36 hr and 48 hr APF axon branching at the second branch point is similar to wild type ( Figure 3J , K ) . An initial difference in branch phenotype can be observed at 60 hr APF and subsequently at 72 hr APF , the typical refinement seen in wild type is largely absent in the EGFRDN background ( Figure 3L–M′ ) . The failure to prune is evident at 84 hr and 96 hr APF ( Figure 3N , O ) where DCN axons show excessive axonal branches . To rule out developmental delay as a cause we examined 2-day vs 18-day-old EGFRDN flies . These flies are indistinguishable from 96 hr APF EGFRDN flies indicating no further branch refinement ( Figure 3—figure supplement 2 ) . Finally , we quantified axonal branch pruning at 48 hr and 72 hr APF by counting the number of branch end-points at these two time points in wildtype and EGFRDN flies , respectively . While there is no significant difference between the two genotypes at 48 hr APF , quantification at 72 hr APF confirms the increased amount of branches in the EGFRDN background compared to wild type ( Figure 3P ) . In addition , the significant decrease in branch number seen in wild type axons between 48 hr and 72 hr is not observed in the EGFRDN axons ( Figure 3P ) . In summary , these data show that EGFR signaling is required to generate the correct number of axonal branches through the reduction of branch precursors formed during development ( Figure 3Q ) . To gain insight into the role of EGFR during axonal branching , we turned to primary embryonic Drosophila neuronal culture ( Prokop et al . , 2011; Sanchez-Soriano et al . , 2010 ) . After 2 days in culture wildtype Drosophila primary neurons sprout on average ∼2 . 5 primary axonal branches , whereas neurons expressing EGFRDN show a significant increase in branch number ( Figure 4A–C ) , suggesting that regulation of axonal branching by the EGFR is a process intrinsic to neurons and common to different neuronal subtypes . Axonal branches develop from dynamic filopodia that gets stabilized during the axonal branching process . We quantified the dynamics of filopodia under WT and EGFR loss of function conditions . We find that in growing wildtype neurons less than 10% of filopodia are static during the imaging time window of 3 min . In contrast , EGFRDN neurons have a significant increase in the percentage of static filopodia to ∼30% ( Figure 4D–F ) . An increase in static filopodia may suggest that more of the transient protrusions are stabilized into branches . An indication of the maturation of filopodia into branches is the invasion of microtubules into axonal filopodia ( Gallo , 2011 ) . Accordingly , we find that EGFRDN induces an increase in microtubules invading axonal filopodia ( Figure 4—figure supplement 1 ) . These data suggest that EGFR signaling regulates branch formation by controlling the dynamics of immature protrusions . To examine the localization of the EGFR in primary neurons , we cultured neurons from animals expressing C-terminally GFP-tagged EGFR ( UAS-EGFRGFP ) and performed live imaging experiments . We find that the EGFR is dynamically transported into and out of axonal branches and their filopodia ( Figure 5A; Video 1 , Video 2 ) , with slightly , but significantly , higher levels in dynamic filopodia compared to static filopodia ( Figure 5A′ , A″ , Figure 5—figure supplement 1 ) . In summary , our data indicate that EGFR is differentially localized to static vs dynamic filopodia and that its activity promotes dynamic filopodial behavior and consequent adjustment of branch number . 10 . 7554/eLife . 01699 . 011Figure 4 . EGFR regulates filopodia dynamics in primary Drosophila neuronal cultures . ( A and B ) Branch formation in cultured primary Drosophila neurons ( 2 days ) . ( B–B′ ) Overexpression of UAS-EGFRDN using the sca-Gal4 driver results in an increase of branches when compared to ( A–A′ ) wild type ( control ) . For the visualization of branches , neurons were stained with anti-tubulin ( green ) and phalloidin ( magenta ) . ( C ) Quantification of primary branch numbers per axon shows significant increase of branches in UAS-EGFRDN neurons ( control: 2 . 48 ± 0 . 2 ( n = 83 ) ; EGFRDN: 3 . 57 ± 0 . 24; n = 74 , p<0 . 001 ) . ( D–E ) Still images from videos of ( D ) wild type and ( E ) UAS-EGFRDN-expressing neurons . Overexpression of UAS-EGFRDN using the sca-Gal4 driver results in a decrease of filopodia dynamics in primary Drosophila neurons cultured for 6–8 hr . Different filopodia are marked by colored arrows and can be followed over time . ( F ) Quantification of static vs dynamic ( extensions and retractions ) behaviors shows a significant distribution change between wild type vs EGFRDN-expressing filopodia ( control: static = 10 , dynamic = 110; EGFRDN: static = 41 , dynamic = 86 , p<0 . 001 ) . Error bars represent SEM . Mann–Whitney test . ***p<0 . 001 . The scale bars in ( A–B ) represent 10 µm and in ( D–E ) represent 3 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01110 . 7554/eLife . 01699 . 012Figure 4—figure supplement 1 . Increase in filopodia containing microtubules by expression of UAS-EGFRDN . ( A and B ) Filopodia containing microtubules in cultured primary Drosophila neurons ( 6 hr ) . ( B ) Overexpression of UAS-EGFRDN using the sca-Gal4 driver results in an increase of filopodia containing microtubules ( arrowhead ) when compared to ( A ) wild type ( control ) . Neurons were stained with anti-tubulin ( green ) and phalloidin ( magenta ) . ( C ) Quantification of filopodia containing microtubules , the percentage of filopodia with microtubules is increase in UAS-EGFRDN neurons ( in percentages , control: 13 ± 0 . 8 ( n = 253 ) ; EGFRDN: 17 ± 0 . 7; n = 239 , p<0 . 001 ) . Error bars represent SEM . Mann–Whitney test . ***p<0 . 001 . The scale bars in ( A and B ) represent 3 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01210 . 7554/eLife . 01699 . 013Figure 5 . EGFR shows differential localization in filopodia of primary Drosophila neurons . ( A–B ) UAS-EGFRGFP expressed with elav-Gal4 in wild type ( A ) and EGFRDN ( B ) primary Drosophila neurons . False color image displaying a heat map of an EGFRGFP-expressing growth cone . EGFRGFP expression in dynamic ( A′ and B′ ) and static filopodia ( A″ and B″ ) is followed over time in wild type ( A′ and A″ ) and EGFRDN ( B′ and B″ ) . A′ and B′ each shows one filopodia growing and one retracting ( C ) . To quantify EGFRGFP intensity in static vs dynamic filopodia in the absence ( control ) or presence of EGFRDN , we calculated the ratio of EGFRGFP in dynamic minus static filopodia ( GFP maximal intensity of each dynamic phase minus the mean of GFP maximal intensity in static filopodia ) . The difference in EGFRGFP levels between dynamic filopodia and static filopodia are significantly reduced in the presence of EGFRDN ( control dynamic-static: 0 . 1046 ± 0 . 009 , n=216; EGFRDN dynamic-static: 0 . 0349 ± 0 . 0121 , n=124 , p<0 . 001 ) . Error bars represent SEM . Mann–-Whitney test . ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01310 . 7554/eLife . 01699 . 014Figure 5—figure supplement 1 . Localization of EGFR in cultured neurons . UAS-EGFRGFP under control of sca-Gal4 is significantly higher expressed in dynamic filopodia compared to static filopodia in cultured wild-type neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01410 . 7554/eLife . 01699 . 015Figure 5—figure supplement 2 . Colocalization of EGFR with Rab11 and Rab5 in the growth cone . ( A–B ) Growth cones from primary Drosophila neurons expressing UAS-EGFRGFP with elav-Gal4 . Neurons were inmunostained for the recycling endosomal marker Rab11 ( A , B , A‴ , B‴ in red ) , GFP ( A′ , B′ , A‴ , B‴ in green ) and the early endosomal marker Rab5 ( A″ , B″ , A‴ , B‴ in blue ) . Note that a fraction of EGFR granules colocalises with Rab11 ( arrows ) or Rab5 ( arrowheads ) . The scale bars in ( A ) represent 1 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01510 . 7554/eLife . 01699 . 016Video 1 . EGFR-GFP cell culture filopodia . This video is related to Figure 5 . Live imaging time-lapse video of axons from different primary neurons grown in culture for 4 days . UAS-EGFRGFP is expressed with elav-Gal4 driver . Images were collected every 4 s . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01610 . 7554/eLife . 01699 . 017Video 2 . EGFR-GFP cell culture filopodia . This video is related to Figure 5 . Live imaging time-lapse video of axons from different primary neurons grown in culture for 4 days . UAS-EGFRGFP is expressed with elav-Gal4 driver . Images were collected every 4 s . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 017 We wondered whether differential EGFR localization is itself dependent on EGFR signaling activity . To this end , we compared levels of EGFRGFP in filopodia of control vs EGFRDN neurons . We find that the difference in EGFRGFP levels between dynamic and static filopodia drops dramatically upon inhibition of EGFR signaling ( Figure 5B , B′ , B″ , C ) . EGFR signaling depends on receptor endocytosis upon ligand binding ( Haigler et al . , 1979 ) . Interestingly , EGFRGFP traffics actively all along the axonal shafts , branches and filopodia in cultured neurons ( Video 3 ) , and we find EGFR-GFP puncta partially co-localize with both Rab5 and Rab11 , suggesting that EGFR is present on early and recycling endosomes ( Figure 5—figure supplement 2 ) . These data indicate that recycling might lead to differential EGFR localization in filopodia . Because the EGFRDN used here can still bind ligand but fails to signal , one interesting possibility is that these dominant negative receptors may titrate ligand away from the functional receptor and thus inhibit not only signaling , but also internalization . To test the putative role of endocytosis in receptor dynamics , we live-imaged EGFRGFP localization in filopodia before and after inhibition of endocytosis using the Dynamin inhibitor Dyngo ( Harper et al . , 2011 ) . In untreated wild type neurons , EGFRGFP levels vary between individual filopodia and within each filopodium over time ( Figure 6A–C ) . Upon inhibition of endocytosis , the overall levels of EGFRGFP in filopodia decrease and the fluctuation of EGFRGFP between and within filopodia is significantly reduced ( Figure 6A′–C ) . This is accompanied by a dramatic reduction in filopodial dynamics ( Figure 6D; Video 4 ) , suggesting that receptor endocytosis and recycling regulates EGFR localization and dynamics in filopodia . 10 . 7554/eLife . 01699 . 018Video 3 . EGFR-GFP cell culture filopodia . This video is related to Figure 5 . Live imaging time-lapse video of axons from different primary neurons grown in culture for 4 days . UAS-EGFRGFP is expressed with elav-Gal4 driver . Images were collected every 4 s . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01810 . 7554/eLife . 01699 . 019Figure 6 . Differential EGFR localization in filopodial of primary Drosophila neurons requires endocytosis . ( A–A′ ) UAS-EGFRGFP expressed with elav-Gal4 driver in primary Drosophila neurons . False color image displaying a heat map of an EGFRGFP-expressing growth cone before ( A ) and after ( A′ ) treatment with Dyngo . ( B ) Maximal intensity of EGFRGFP in filopodia within one neuron over time ( 2 min ) , before and after treatment ( indicated by dotted line ) with Dyngo . ( C ) Scatter plot from EGFRGFP maximal intensities from filopodia from 6 neurons , showing a significant decrease in levels after treatment with Dyngo ( EGFRGFP maximal intensities in DMSO: 1 . 229 ± 0 . 0074 , n = 1403; EGFRGFP maximal intensities in Dyngo: 0 . 768 ± 0 . 005 , n = 1401 , p<0 . 001 ) . ( D ) Effect of Dyngo on filopodia dynamics . Quantification of static vs dynamic ( extensions and retractions ) behaviors of filopodia shows a significant distribution change between controls ( EGFRGFP-expressing neurons in DMSO ) and Dyngo-treated EGFRGFP-expressing neurons ( control: static = 28 , dynamic = 72; Dyngo treated: static = 66 , dynamic = 32 , p<0 . 001 ) . Mann–Whitney test . ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 01910 . 7554/eLife . 01699 . 020Video 4 . Comparison egfr vs egfr + dingo . This video is related to Figure 6 . Video shows the side-by-side comparison of dynamic behavior of filopodia with and without Dyngo-4a ( a dynamin inhibitor ) treatment . The intensity of the EGFR-GFP signal is displayed in yellow . The outlines of the filopodia have been segmented by subsequently thresholding and outline detection of the fluorescent signal . To show the dynamic behavior of the filopodia in the time-lapse video , the outlines of the following two frames ( 4 and 8 s ahead of the current frames ) are displayed in red and in blue respectively . The untreated filopodia move more than the Dyngo-4a treated ones . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 020 Next , we asked if the EGFR is differentially localized and regulates branching dynamics in vivo . EGFR transcription ( Schejter et al . , 1986 ) and function in the Drosophila developing and adult brain have been documented , where it plays a role in neuronal survival ( Botella et al . , 2003 ) and sleep regulation ( Foltenyi et al . , 2007 ) . However , attempts to detect the EGFR protein using immunohistochemistry have thus far failed , most likely due to very low expression levels . We attempted to circumvent this problem by generating a genomic rescue construct tagged at the C-terminal end with GFP , identical to the UAS-EGFRGFP used in our cell culture experiments . This construct rescues the embryonic lethality of EGFR null mutants to full adult viability with no visible defects . We examined the expression of genomic EGFRGFP during brain development and find that it is broadly expressed in the developing neuropil , especially the distal medulla ( Figure 7—figure supplement 1A ) , suggesting that EGFR signaling may be generally involved in the regulation of CNS connectivity . Indeed , inhibition of EGFR activity in the lateral neurons ventral ( LNv ) also causes excessive axonal branching ( Figure 7—figure supplement 1B ) . Unfortunately , expression levels of the genomic EGFRGFP transgene were too low to allow analysis at sub-cellular , single axon branch resolution . To examine subcellular EGFR distribution , we expressed UAS-EGFRGFP in the DCNs . In DCNs , EGFRGFP is detected in a punctate pattern in the cell bodies ( Figure 7A , B , insets ) , along the axons and in axonal branches ( Figure 7—figure supplement 1C–E ) . At ∼56 hr APF , when extensive growth and pruning occur , EGFRGFP is unevenly distributed across different branches of the same axon ( Figure 7A–A‴ ) and we find no stereotypic pattern across different individual axons or individual brains ( Figure 7—figure supplement 1F ) . At ∼72 hr APF , after significant pruning has occurred , EGFRGFP is distributed more uniformly across the remaining unrefined branches ( Figure 7B–B‴ ) . Note that both wildtype-untagged EGFR and EGFRGFP do not change the DCN branching pattern ( Figure 7—figure supplement 4 ) , hinting that asymmetric receptor signaling is governed by differential receptor distribution , rather than total receptor levels per se . In the LNv , EGFRGFP is expressed in cell bodies and low levels are present along the growing axons during development ( Figure 7—figure supplement 1G ) . In contrast , in adult LNv UAS-EGFRGFP becomes restricted to neuronal soma ( Figure 7—figure supplement 1G′ ) . Thus , remarkably , even overexpressed EGFRGFP is present at relatively low levels and shows regulated developmental localization in different neuronal populations in vivo . 10 . 7554/eLife . 01699 . 021Figure 7 . EGFR mediates a probabilistic branch refinement process . ( A–B ) EGFR localization examined by expressing UAS-EGFRGFP ( green ) in the DCNs ( red , UAS-cherryRFP ) during pupal development at ( A ) 56 hr APF and ( B ) 72 hr APF . EGFRGFP expression was observed in a punctate pattern in the cell bodies ( insets in A and B ) and along the axonal branches ( A and B ) . Images A/B and A′/B′ were subjected to thresholding and merged ( A‴/B‴ ) . Differential localization results in branches with ( A‴ , arrowheads ) and without ( A‴ , arrows ) EGFRGFP at 56 hr APF , whereas most if not all branches contain EGFRGFP at 72 hr APF ( B‴ , arrowheads ) . High magnification shows EGFR localization at branches at 56 hr APF ( A2 ) and 72 hr APF ( B2 ) . ( C ) Z-stack projections from live imaging time-lapse videos of control axons at around 40 hr APF between t0 = 0 min ( C1 ) and t2 = 10 min ( C3 ) with 5-min intervals . ( D ) Z-stack projections from live imaging time-lapse videos of EGFRDN axons at around 40 hr APF between t0 = 0 min ( D1 ) and t2 = 10 min ( D3 ) with 5 min intervals . Arrows indicate branches being pruned while arrowheads point to growing branches . ( E ) Visualization of growth ( green ) and retraction ( purple ) events between t0 = 0 min ( C1 ) and t1 = 5 min ( C2 ) in control . ( F ) Visualization of growth ( green ) and retraction ( purple ) events between t1 = 5 min ( D2 ) and t2 = 10 min ( D3 ) in EGFRDN . ( G ) Quantification of growth and retraction dynamics at branches using the tracer tool shows significant decrease in branch lengths in EGFRDN compared to control . Control ( growth ) 7 . 75 ± 2 . 65 ( n = 8 ) , EGFRDN ( growth ) 2 . 97 ± 0 . 56 ( n = 9 , p<0 . 001 ) . Control ( retraction ) 7 . 4 ± 2 . 28 ( n = 8 ) , EGFRDN ( retraction ) 3 ± 1 . 08 ( n = 8 , p<0 . 001 ) . Horizontal lines represent the mean for each data set . t test . ***p<0 . 001 . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02110 . 7554/eLife . 01699 . 022Figure 7—figure supplement 1 . Localization of EGFR . ( A ) Expression of genomic EGFRGFP ( green ) during brain development at 48 hr APF and the neuropil marked with DN-cadherin ( blue ) . ( B ) Axonal arbor of lateral neurons ventral ( LNv ) ( green ) in wild type and EGFRDN . ( C–E ) EGFR localization by expressing UAS-EGFRGFP ( green ) in the adult DCNs ( red , UAS-cherryRFP ) in one brain hemisphere ( C ) , in the cell bodies ( D ) and in the axonal branches ( E ) . The scale bars represent 20 µm except in C with 100 µm . ( F ) EGFR localization by expressing UAS-EGFRGFP ( green ) in the DCNs ( red , UAS-cherryRFP ) of two different individual flies during pupal development at 56 hr APF . Differential localization results in branches with ( arrowheads ) and without ( arrows ) EGFR localization at 56 hr APF . The scale bars represent 20 µm . ( G ) EGFR localization examined by expressing UAS-EGFRGFP ( green ) in the LNvs ( red , UAS-lacZ ) at ( G ) larval stage ( L3 ) and in ( G′ ) adult . The scale bar represents 60 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02210 . 7554/eLife . 01699 . 023Figure 7—figure supplement 2 . DCN branch pattern in cultured pupal brains . ( A ) DCN branch morphology of a brain dissected from pupae at 48 hr APF and then cultured under standard conditions ( ‘Materials and methods’ ) for 48 hr . Morphology of the neuropils has been visualized by nc82 staining ( magenta ) . ( B–C ) Axon branch morphology in the optic lobe of pupal brain dissected at 48 hr APF and cultured for ( B ) 24 hr and ( C ) 48 hr . ( D ) EGFR shows differential and dynamic localization in developing dorsal cluster neurons in vivo . UAS-CD8-RFP and UAS-EGFR-GFP were expressed with ato-Gal4 in wild-type Drosophila brains . Intact eye–brain complexes were imaged live at 45% APF . Maximum projection images demonstrating a single DCN axon terminal from live imaging time-lapse videos at t = 0 min ( left ) t = 18 min ( middle ) and t = 46 min ( right ) , for both channels ( upper ) and only EGFR channel ( lower ) . Two directly opposing branches of the same DCN axon were followed over time . At t = 0 both branches have significant levels of EGFR signal ( arrow and arrowhead ) . 18 min later upper ( arrow ) branch retains its EGFR signal while lower ( arrowhead ) branch demonstrates a significant decrease . 28 min later the lower branch demonstrates a slight increase in the signal while the upper branch almost completely loses it . Scale bars correspond to 5 µm in all images . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02310 . 7554/eLife . 01699 . 024Figure 7—figure supplement 3 . UAS-EGFRGFP localizes and functions similar to endogenous EGFR . ( A ) Expression of UAS-EGFR-GFP in photoreceptor neurons at developing L3 eye disc using GMR-Gal4 . ( A′ ) Expression of endogenous EGFR revealed by immunohistochemistry on developing Canton-S L3 eye disc . The scale bars represent 20 µm . ( B–D′ ) EGFR-GFP overexpression in the wing produces increase in vein tissue . ( B ) Control bearing Dpp-Gal4 , showing the wild-type vein pattern , ( B′ ) Zoom in of a ROI in ( B ) . ( C ) Flies over-expressing wild-type-untagged EGFR in the wing using Dpp-Gal4 ( along vein L3 , black arrow ) show a vein-specific increase in vein thickness ( yellow arrow ) and formation of ectopic veins ( yellow arrowheads ) , ( C′ ) zoom in of a ROI in ( C ) . ( D ) Flies expressing EGFR-GFP using Dpp-Gal4 produces similar phenotypes to wild type EGFR over-expression ( yellow arrow and arrowhead ) , ( D′ ) zoom in of a ROI in ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02410 . 7554/eLife . 01699 . 025Figure 7—figure supplement 4 . Overexpression of wild-type EGFR does not cause a significant increase in axonal branching . Quantification of the number of axonal branches in DCNs overexpressing untagged or GFP-tagged EGFR . Neither causes a significant increase in the average number of axonal branches . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 025 DCN branches develop and prune during pupal development when the brain is not easily accessible to live imaging . To overcome this limitation , we modified the protocol for long-term adult brain explant culture ( Ayaz et al . , 2008 ) to support long-term pupal brain culture . This protocol supports the morphologically normal development of Drosophila pupal brains ( Figure 7—figure supplement 2A–C ) . We sought to probe the basis of the regulation of developmental branch pruning by the EGFR . To this end , we performed high-resolution 4D live imaging to analyze real-time DCN axon branch formation by pairing the brain explant culture technique with resonant confocal microscopy of the cultured brains in a closed perfusion chamber ( Williamson and Hiesinger , 2010 ) . Imaging of developing wild type pupal brains ( 40 hr–60 hr APF ) shows that wildtype DCN axon branches are dynamic ( Video 5 ) . Branch growth and removal occurs within minutes and can span up to 11 . 5 μm within 5 min with an average of 7 . 5 μm during this period ( Figure 7C1–C3 , E ) . Furthermore , wildtype branches behave differently from each other , as indicated by the spread of growth and retraction speeds of different branches ( Figure 7G ) . In contrast , the growth dynamics in the EGFRDN expressing neurons are reduced in speed ( Video 6 ) . Growth and retraction processes of single branches are decreased to an average of 3 μm within 5 min and the dynamics show strikingly reduced variability between individual branches ( Figure 7D1–D3 , F , G ) . We wondered whether we could exploit the new developing brain culture system to ask whether EGFR is dynamically trafficked within DCN branches in vivo as these branches grow and retract . To this end , we generated flies expressing both EGFR-GFP and a red fluorescent protein ( td-Tomato ) in the DCNs . Live imaging ( Video 7 ) of pupal brains from these animals and analysis of still images form these videos ( Figure 7—figure supplement 2D ) confirms that , like in primary neurons in culture , EGFR is trafficked dynamically as axons grows and retracts their branches in vivo; finally , it should be noted that EGFRGFP shows similar localization and activity to its wild type counterpart ( Figure 7—figure supplement 3 ) , in agreement with the fact that the identically tagged genomic construct rescues the null mutant to full viability . 10 . 7554/eLife . 01699 . 026Video 5 . brain culture WT 40 hr . This video is related to Figure 7 . Live imaging time-lapse videos of control axons at around 40hr APF . Corresponds to images presented in Figure 7C and quantified in Figure 7G . Images were collected every 5 min for 45 min . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02610 . 7554/eLife . 01699 . 027Video 6 . brain culture EGFR-DN 40 hr . This video is related to Figure 7 . Live imaging time-lapse videos of EGFRDN axons at around 40hr APF . Corresponds to images presented in Figure 7D and quantified in Figure 7G . Images were collected every 5 min for 40 min . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02710 . 7554/eLife . 01699 . 028Video 7 . brain culture EGFR-GFP . This video is related to Figure 7 . EGFR shows differential and dynamic localization in developing dorsal cluster neurons in vivo . UAS-CD8-RFP and UAS-EGFR-GFP were expressed with ato-Gal4 in wildtype Drosophila brains . Intact eye–brain complexes were imaged live at 45% APF . Maximum projection images demonstrating three DCN axon terminals in a time-lapse video of 3 hr with 2 min time intervals . Extension of the axons over time could be observed especially in the upper axon . All axons demonstrate rapid filopodial dynamics as well as changes in EGFR localization over time . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 028 To analyze downstream players of EGFR-dependent refinement we first focused on the canonical EGFR pathway . Activation of the MAPK cascade and transcriptional changes in the nucleus are main features of this pathway ( Vivekanand and Rebay , 2006 ) . A nuclear marker for active MAPK signaling is double phosphorylated ERK ( dpERK ) . Despite the fact that we verify expression in developing L3 eye disc ( Figure 8—figure supplement 1A ) , we were not able to detect dpERK in developing DCN ( Figure 8—figure supplement 1B-B″ ) . One caveat is that activation of ERK might be difficult to detect due to low expression levels and timing issues . To further investigate if the canonical pathway is involved , we analyzed the effect of MAPK pathway genes ( Vivekanand and Rebay , 2006 ) on DCN axon refinement . Expression of Ras1RNAi , Drk RNAi , a constitutively active form of ERK or a constitutively active form of Ras1 did not change the DCN branching pattern ( Figure 8—figure supplement 1C–G ) . These results indicate that refinement occurs independently of the canonical EGFR pathway . The fast growth and retraction rates of axonal branches in wildtype brains , altered growth dynamics upon EGFR inhibition and the well-established role for cytoskeletal proteins in branch formation ( Gallo , 2011 ) together suggest that EGFR activation may act via cytoskeleton regulation in this case . We used actin-GFP ( Verkhusha et al . , 1999 ) and Utrophin-GFP ( Rauzi and Lenne , 2011 ) expression in the DCNs to examine the distribution of total actin and filamentous actin ( F-actin ) , respectively , in wild type vs EGFRDN backgrounds . The F-actin binding protein Utrophin ( Galkin et al . , 2002 ) was utilized to analyze the distribution of actin filaments in wildtype and EGFRDN axons . Utrophin-GFP reveals that F-actin is largely confined to the branches ( Figure 8A–A‴ , arrowheads ) with low levels of F-actin in the axon shafts ( Figure 8A‴ , arrow ) of wildtype brains . In contrast , in DCNs expressing EGFRDN F-actin distribution appears weaker and more diffused over the axon shaft and axon branches ( Figure 8B–B‴ , arrowheads ) . Similar to F-actin , total actin-GFP concentrates at the branch tips ( Figure 8—figure supplement 2A–A‴ , arrowheads ) and little actin is present within the main axon shaft . In contrast , in EGFRDN DCNs total actin also accumulates in blebs along the entire length of the axons and their branches ( Figure 8—figure supplement 2B–B‴ , arrows ) . Axonal swellings have considerably lower F-actin accumulation ( Figure 8B‴ , asterisk ) compared to total actin ( Figure 8—figure supplement 2B‴ , asterisk ) suggesting that in EGFRDN axons , monomeric actin is retained in axonal swellings along the axons , thus potentially inhibiting efficient actin polymerization dynamics at the branch tips . 10 . 7554/eLife . 01699 . 029Figure 8 . EGFR regulates actin polymerization in DCN axonal branches . ( A–B ) Utrophin ( F-actin ) localization in adult DCN ( red , UAS-cherryRFP ) by expressing UAS-utrophin-GFP ( green ) in ( A ) control and ( B ) EGFRDN . ( A‴–B‴ ) High magnification of the branch tips . Arrowheads show localization of Utrophin at the branch tips . Arrows show Utrophin localization along axon shafts . Asterisk in ( B‴ ) shows weak Utrophin accumulation in axonal swellings . ( C ) Overexpression of a constitutively active form of EGFR ( UAS-EGFRCA ) results in increased branching in the adult DCN . ( C′ ) Visualization of branches ( purple ) along a single main axon shaft ( green ) , using the tracing tool . ( D ) Quantification of adult primary branch numbers per axons shows significant increase of branches in EGFRCA compared to control . Control 6 . 96 ± 1 . 34 ( n = 60 ) , EGFRCA 8 . 22 ± 1 . 47 ( n = 55 , p<0 . 001 ) . Error bars represent SEM . Mann–Whitney test . ***p<0 . 001 . ( E–H ) Axonal branch pattern at different pupal stages shows excessive branching during mid-pupal development . Branch morphology at ( E ) 36 hr APF , ( F ) 48 hr APF , ( G ) 60 hr APF , and ( H ) 72 hr APF . ( I–I′ ) Utrophin ( F-actin ) localization in adult DCN ( red , UAS-cherryRFP , I′ ) by expressing UAS-utrophin-GFP ( green , I ) in an EGFRCA background . ( I″ ) Merge of DCNs ( red ) and Utrophin ( green ) . ( I‴ ) High magnification of the branch tips . Arrowheads show localization of Utrophin at the branch tips . Arrowheads show localization of Utrophin at the branch tips . Arrows show Utrophin localization along axon shafts . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 02910 . 7554/eLife . 01699 . 030Figure 8—figure supplement 1 . The canonical MAPK pathway is not involved in DCN refinement . ( A ) Localization of phosphorylated ERK ( dpERK , red ) in developing eye discs as readout for activated EGFR pathway . ( B ) Absence of dpERK ( red , B′ ) at the DCN cell bodies ( green , B ) during pupal development . ( B″ ) Merge of DCNs ( green ) and dpERK ( red ) . ( C–F ) DCN branch pattern in neurons with genetically altered MAPK pathway activity by expression of ( C ) a constitutively active form of ERK ( ERKCA ) , ( D ) Ras1RNAi , ( E ) a constitutively active form of Ras1 ( Ras1CA ) and ( F ) Drk RNAi . ( G ) Quantifications of the genotypes in ( C–F ) do not reveal significant changes in DCN branch numbers compared to control . Control 6 . 96 ± 1 . 34 ( n = 60 ) , ERKCA 7 . 43 ± 1 . 07 ( n = 30 ) , Ras1RNAi 7 . 2 ± 1 . 71 ( n = 59 ) , Ras1CA 7 . 07 ± 1 . 44 ( n = 14 ) , DRKRNAi 7 . 36 ± 1 . 22 ( n = 14 ) . Non-parametric ANOVA Kruskal–Wallis test was not significant . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 03010 . 7554/eLife . 01699 . 031Figure 8—figure supplement 2 . EGFR regulates actin polymerization in DCN axonal branches . ( A–C ) Actin localization in adult DCN ( red , UAS-cherryRFP ) by expressing UAS-Actin-GFP ( green ) in ( A ) control , ( B ) EGFRDN and ( C ) EGFRCA . ( A‴–B‴ ) High magnification of the branch tips . Arrowheads show localization of Actin at the branch tips . Arrows show actin localization along axon shafts . Asterisk in ( B ) shows total actin accumulation in axonal swellings . The scale bars represent 20 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 03110 . 7554/eLife . 01699 . 032Figure 8—figure supplement 3 . Branch increase in cultured neurons by expression of EGFRCA . ( A–A′ ) Overexpression of UAS-EGFRCA using the sca-Gal4 driver in neurons cultured for 2 days results in an increase of branches when compared to wild type ( compare to Figure 4A–A′ ) . For the visualization of branches , neurons were stained with anti-tubulin ( green ) and phalloidin ( magenta ) . ( B ) Quantification of primary branch numbers per axon shows significant increase of branches in UAS-EGFRCA neurons ( control: 1 ± 0 . 05 ( n = 235 ) ; EGFRCA: 1 . 37 ± 0 . 09; n = 120 , p<0 . 01 ) . Error bars represent SEM . Mann–Whitney test . **p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 032 Our data thus far suggest a model whereby dynamic localization of the EGFR results in differential signaling between developing filopodia and axonal branches . This enhances actin dynamics and results in the proper balance of branch growth and pruning . However , an alternative possibility is that EGFR signaling simply instructs branch retraction . Both models predict increased branch numbers when EGFR signaling is inhibited . However , if EGFR signaling instructs branch pruning , activated EGFR would result in reduced axonal branching . In contrast , if EGFR signaling asymmetry is indeed required for the correct number of DCN axonal branches , then constitutive activation of the EGFR should also result in increased axonal branching . To distinguish between the two models , we analyzed the effect of a constitutively active form ( UAS-EGFRCA ) . In agreement with a differential local signaling model , EGFRCA induces a significant increase of DCN branches both in vitro ( Figure 8—figure supplement 3 ) and in vivo ( Figure 8C–D ) similar to down regulation of EGFR signaling . Importantly , similar to loss of EGFR function , the increase in branch number induced by gain of EGFR function is also due to reduced pruning during development ( Figure 8E–H ) . Furthermore , in EGFRCA axons total actin and F-actin also distribute more uniformly across the axonal projection ( Figure 8I–I‴ , Figure 8—figure supplement 2C–C‴ , arrows ) , again suggesting reduction of efficient polymerization dynamics . In summary , EGFR signaling affects branch growth and retraction likely through the regulation of actin polymerization .
The refinement of exuberant branches is a crucial step during the development of a neuronal network . In this work , we exploit an adult-specific model circuit , the dorsal cluster neurons , to study developmental neurite pruning processes in the CNS of Drosophila . DCN axons form a stereotyped number of branches innervating the medulla through initial excessive axon branch formation followed by a refinement process . Our data suggest a model ( Figure 9 ) whereby uneven distribution of EGFR to developing DCN axonal branches is required to eliminate exuberant branches and help generate the correct adult connectivity pattern . 10 . 7554/eLife . 01699 . 033Figure 9 . A model for EGFR function in axonal branching . Local asymmetries in tyrosine kinase receptor activity in axonal branch , driven by differential distribution of active receptor molecules in filopodia , generate dynamical behavior and drive branch pruning . Gray dots represent EGFR puncta trafficked along the axon shaft ( red ) while yellow dots represent active EGFR puncta within branches ( green ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01699 . 033 During mammalian development neurites are generally formed in excessive numbers and subsequently refined to form the mature circuit ( Low and Cheng , 2006 ) . This mechanism ensures that all targets are properly innervated , it enables further specification of connections by the target environment like neighboring neurons and glia ( Stevens et al . , 2007 ) and permits the removal of exuberant or mistargeted branches . Studying real-time events in the mammalian system involving CNS refinement is challenging . The Drosophila developing brain culture system used in this work combined with live imaging allows examination and manipulation of neuronal growth dynamics . Our data suggest that EGFR signaling , in part triggered by the co-innervation of the target neuropil by sensory neurons from the retina , is a crucial determinant of axonal branch refinement by the regulation of filopodial growth and retraction dynamics . Finally , we find that EGFR activity regulates actin polymerization dynamics at the branch tips . Consistent with this notion , we find that interfering with actin dynamics in vivo by inhibition of the small GTPase RhoA or constitutive activation of the actin filament severing protein Cofilin , is sufficient to cause ectopic axon branch formation in the DCNs ( data not shown ) . EGFR expression has been observed in neurites of mammalian neurons ( Gerecke et al . , 2004; Chen et al . , 2005 ) and knock-out of the EGFR in the mouse results in increased neurite branching in the skin ( Maklad et al . , 2009 ) , suggesting that the mechanism we identify in the fly CNS may be more generally utilized . In summary , we report evidence for the notion that differential branch signaling is a determinant of connection specificity . We show that intrinsically asymmetric EGFR localization and signaling is required for efficient branch pruning . Several lines of evidence support this conclusion . First , EGFR is asymmetrically localized in branches and filopodia both in vivo and in cultured primary neurons . Second , both inhibition and constitutive activation result in failure of axonal branch refinement . Third , overexpression of the wildtype receptor , which is differentially localized and trafficked , is not sufficient to produce a phenotype . This argues that receptor localization dynamics—possibly mediated by endocytosis—rather than total EGFR levels , is the cue for filopodial collapse and subsequent axonal branch pruning . What explains the link between regulation of dynamic behavior and the generation of a specific number of axonal branches ? A hint to this comes from three observations . First , both loss and gain of EGFR function increase proportion of static filopodia from less than 10% to more than 30% , subsequently increasing the number of axonal branches . Second , this filopodial behavior correlates with small , but significant and highly dynamic differences in EGFR localization . Third , loss of EGFR signaling increases the variability in axon branch number . Based on these observations we propose that in wildtype neurons most dynamic filopodia collapse over time , resulting in continuous redistribution of EGFR among fewer and fewer remaining filopodia . This process stops usually when only one filopodium remains at a given branching point , and occasionally when EGFR happens to distribute equally between the last two filopodia . This probabilistic process does not require an additional mechanism of branch ‘tagging and selection’ and can explain both EGFR loss of function phenotypes: increased branch number and increased variability . What remains to be determined is the interaction between EGFR-dependent branch dynamics and the specificity of the spatial pattern of branches .
Fly stocks were cultured on standard fly food . All experiments were performed in temperature-controlled incubators at 25°C or 28°C . The GAL4 driver lines used in this study are: ato-Gal4-14a ( Hassan et al . , 2000 ) , sca-Gal4 , elav-Gal4 . The UAS-reporter stocks were the following: UAS-CD8-GFP , UAS-CD8-cherryRFP , UAS-LacZ , UAS-EGFRDN-A ( gift from M Freeman ) , UAS-EGFRDN-B , UAS-EGFRRNAi ( VDRC107130 ) , UAS-SpiRNAi ( TRiP , JF03322 ) , UAS-EGFRCA , UAS-Utrophin-GFP ( gift from T Lecuit ) , UAS-Moesin-GFP ( Dutta et al . , 2002 ) , lexAop-myr-GFP , atolexA . Additional fly stocks and mutants used were: Canton-S , EGFRT1 . For FLP-out system experiments yw , hs-FLP; UAS-FRT CD2 , y FRT mCD8::GFP; atoGal4-14a , UAS-LacZ was crossed out to Canton-S or UAS-EGFRDN-A . atoLexA was created by knocking LexA into the ato locus to drive LexAop-myr-GFP expression . Drosophila primary neuron cultures were generated as described previously ( Sanchez-Soriano et al . , 2010; Prokop et al . , 2011 ) . In brief , stage 11 embryos ( 6–7 hr AEL at 25°C ) were homogenised , treated for 5 min at 37°C with dispersion medium , washed and dissolved in Schneider’s medium . Then , the aliquots were transferred to coverslips , kept as hanging drop cultures in airtight special culture chambers ( Deak et al . , 1980 ) for 6 hr or 4 days at 26°C . Live imaging of primary neurons was performed on a Delta Vision ( RT ) ( Applied Precision , Issaquah , WA ) restoration microscope using a ( 100 × 3 phase ) objective and the ( Sedat ) filter set ( Chroma Technology , Germany ) . The images were collected using a Coolsnap HQ ( Photometrics , Tuscon , AZ ) camera , image acquisition was through Softworx . For immunocytochemistry , cells were fixed ( 30′ in 4% paraformaldehyde in 0 . 05 M phosphate buffer , pH 7 . 2 ) , washed in PBS 0 . 1% Triton X-100 ( PBT ) , then incubated with antisera diluted in PBT . To inhibit endocytosis , cells were incubated for 6 min with 0 . 14 mM dynamin inhibitor Dyngo-4a ( Abcam ) , diluted in Schneider’s medium from stock solution in DMSO . For controls , equivalent concentrations of DMSO were diluted in Schneider’s medium . The effect of the dynamin inhibitor Dyngo-4a on levels of EGFRGFP was quantified in FIJI , by measuring the maximal intensities at the distal ends of filopodia during 2 min before and after drug treatment . Previous to quantification , the background of acquired images was subtracted ( atrous wavelet transform , scales 1–8 minus low pass image ) . The GFP intensity of each filopodia was normalized to the mean of maximal intensities of all filopodia within a cell before and after treatment . The following primary antibodies were used in the in vivo experiments: rabbit anti-GFP ( 1:1000; Invitrogen ) , mouse anti-GFP ( 1:500; Invitrogen ) , mouse anti β-galactosidase ( 1:1000; Promega ) , rabbit anti β-galactosidase ( 1:1000; Cappel ) , mouse MAb 24B10 anti-Chaoptin ( 1:200; DSHB ) , rabbit anti-DsRed ( 1:500; Clontech ) , mouse anti-NC82 ( 1:100; DSHB ) , rat anti-DN cadherin ( 1:20; DHSB DN-EX#8 ) . The following primary antibodies were used in the in vitro experiments: mouse anti-tubulin ( 1:1000; Sigma ) , goat anti-GFP ( 1:1000; Abcam ) . The incubation with the primary antibodies was followed by several washes in PBT ( 1 hr ) and a final incubation with the appropriate fluorescent secondary antibodies ( in vivo: Alexa 488 , 555 or 647 , Molecular Probes , 1:500 , in vitro: FITC- or Cy3-conjugated affinity-purified secondary antibodies ( donkey , 1:200 [Jackson ImmunoResearch] ) ) . In vitro filamentous actin was detected with TRITC-conjugated phalloidin ( Sigma ) . After several washes in PBT the samples were mounted in vectashield . Confocal stacks of fixed brains were made using Zeiss LSM 510 or Leica SP6 confocal microscopes . Neuronal cell culture imaging was conducted with an AxioCam camera mounted on an Olympus BX50WI microscope . DCN live imaging was conducted with a Leica SP6 resonance scanning confocal microsocope . In general , a confocal stack comprising the axonal projection of Dorsal Cluster Neurons ( 30–40 single projections ) was recorded every 5 min . Resonance scanning allowed high scan speed with lower laser intensities and therefore ensures preservation of living tissue due to decreased photo-toxicity . Projection images were generated and further processed with ImageJ . For tracking of axon branches we have used the ‘simple neurite tracer’ a plugin for ImageJ from Mark Longair ( Fiji , http://pacific . mpi-cbg . de ) . Images of medulla axons were skeletonized and subsequently automatically analyzed using the ‘Skeletonize3D’ and ‘AnalyzeSkeleton’ free plugins for ImageJ/FIJI ( freely downloadable from the FIJI website: URL: http://pacific . mpi-cbg . de/wiki/index . php/Fiji ) . Number of developmental branches is the number of end points from the skeleton . Staged pupal brains were dissected in cold Schneider’s Drosophila Medium ( GIBCO ) and transferred to the culture plate inserts and cultured according to the whole brain explant system described previously ( Ayaz et al . , 2008 ) . After allowing the pupal brains to attach to the membrane of the culture plate insert for a minimum of 8 hr , the membrane was cut out of the plastic insert and carefully transferred to a closed confocal imaging perfusion chamber ( Harvard IC30 confocal imaging chamber ) connected to a peristaltic pump that slowly perfuses culture solution over the live tissue . A fast resonant scanning confocal microscope ( Leica TCS SP5 ) with special high-aperture immersion lenses was used to allow three-dimensional recordings over time at faster frame rates which reduces phototoxicity . Live imaging was performed as previously described ( Williamson and Hiesinger , 2010 ) . For tracking of growth and retraction dyamics we have used the ‘simple neurite tracer’ a plugin for ImageJ from Mark Longair ( Fiji , http://pacific . mpi-cbg . de ) . We have traced dynamic axon branches by using the tip of an axon at time point t0 as starting point and the tip of the same axon at time point t1 . The length of the resulting fragment represents the length of the growing or retracting axon . UAS-EGFRGFP was created by fusing the Drosophila egfr cDNA from the pUC13-DERII construct ( Schejter et al . , 1986 ) and eGFP cDNA ( Clontech ) from pStinger into pUAST-Attb vector ( Genbank EF362409 . 1 ) . Two Gly-Gly-Ser bridges ( GGSGGS ) have been introduced between the two open reading frames . Transgenic flies were created at GenetiVision Inc . ( Houston , USA ) using PhiC31-mediated transgenesis in the VK37 docking site ( 2L , 22A3 ) and in the VK31 docking site ( 3L , 62E1 ) . Levels of EGFRGFP in static vs dynamic filopodia were quantified in FIJI , by drawing a box at the distal ends of filopodia and measuring the maximal intensities within . Only neurons with both static and dynamic filopodia were used for the analysis , and the GFP intensity of each filopodia was normalized to the mean of maximal intensities of all filopodia within a cell . For dynamic filopodia , measurements were taken during the first 8 s of the retraction or extension . For static filopodia , measurements were taken during 8 s at the middle of the recording period . These measurements were used to calculate the ratio of EGFRGFP in dynamic minus static filopodia ( GFP maximal intensity of each dynamic phase minus the mean of GFP maximal intensity in static filopodia ) . For non-normally distributed samples the nonparametric ANOVA Kruskal–Wallis test with Dunn’s multiple comparisons for Figure 2E and the Mann–Whitney test for Figure 8D was performed . Student’s t test was used for Figures 3P and 7G . For neuronal culture experiments , the Mann–Whitney test was used for Figure 4C , F , 5C and 6D . | In the human brain , 100 billion neurons form 100 trillion connections . Each neuron consists of a cell body with numerous small branch-like projections known as dendrites ( from the Greek word for ‘tree’ ) , plus a long cable-like structure called the axon . Neurons receive electrical inputs from neighboring cells via their dendrites , and then relay these signals onto other cells in their network via their axons . The development of the brain relies on new neurons integrating successfully into existing networks . Axon branching helps with this by enabling a single neuron to establish connections with several cells , but it is unclear how individual neurons decide when and where to form branches . Now , Zschätzsch et al . have revealed the mechanism behind this process in the fruit fly , Drosophila . Mutant flies that lack a protein called EGFR produce abnormal numbers of axon branches , suggesting that this molecule regulates branch formation . Indeed in fruit flies , just as in mammals , the developing brain initially produces excessive numbers of branches , which are subsequently pruned to leave only those that have formed appropriate connections . In Drosophila , an uneven distribution of EGFR between branches belonging to the same axon acts as a signal to regulate this pruning process . To examine this mechanism in more detail , high-resolution four-dimensional imaging was used to study brains that had been removed from Drosophila pupae and kept alive in special culture chambers . Axon branching and loss could now be followed in real time , and were found to occur more slowly in brains that lacked EGFR . The receptor controlled the branching of axons by influencing the distribution of another protein called actin , which is a key component of the internal skeleton that gives cells their structure . In addition to providing new insights into a fundamental aspect of brain development , the work of Zschätzsch et al . also highlights the importance of stochastic events in shaping the network of connections within the developing brain . These findings may well be relevant to ongoing efforts to map the human brain ‘connectome’ . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Regulation of branching dynamics by axon-intrinsic asymmetries in Tyrosine Kinase Receptor signaling |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.