Chromatin opening ability of pioneer factor Pax7 depends on unique isoform and C-terminal domain

Abstract Pioneer factors are transcription factors (TFs) that have the unique ability to recognise their target DNA sequences within closed chromatin. Whereas their interactions with cognate DNA is similar to other TFs, their ability to interact with chromatin remains poorly understood. Having previously defined the modalities of DNA interactions for the pioneer factor Pax7, we have now used natural isoforms of this pioneer as well as deletion and replacement mutants to investigate the Pax7 structural requirements for chromatin interaction and opening. We show that the GL+ natural isoform of Pax7 that has two extra amino acids within the DNA binding paired domain is unable to activate the melanotrope transcriptome and to fully activate a large subset of melanotrope-specific enhancers targeted for Pax7 pioneer action. This enhancer subset remains in the primed state rather than being fully activated, despite the GL+ isoform having similar intrinsic transcriptional activity as the GL– isoform. C-terminal deletions of Pax7 lead to the same loss of pioneer ability, with similar reduced recruitments of the cooperating TF Tpit and of the co-regulators Ash2 and BRG1. This suggests complex interrelations between the DNA binding and C-terminal domains of Pax7 that are crucial for its chromatin opening pioneer ability.


INTRODUCTION
Pioneer transcription factors are unique in their ability to recognise and bind their target DNA sequences in closed chromatin ( 1 ). Their action initiates chromatin opening by a process that has recei v ed recent insight ( 2 ). This unique property has endowed this class of transcription factors with the power to direct new cell fates in de v elopment ( 3 , 4 ). This is achie v ed by mostly targeting ne w enhancer repertoires for chromatin opening and activation of novel transcriptomes in daughter compared to mother cells (5)(6)(7)(8). Hence, the activation of new transcriptomes defines new cell identities. Pioneers also establish an epigenetic memory of these new cell identities through demethylation of target enhancer DNA ( 1 , 8-11 ). It appears as if pioneer ability may not be r equir ed beyond the initial chromatin opening and establishment of epigenetic memory: beyond this epigenetically stable process, pioneer factors often serve as regular transcription factors to maintain differential cell specific transcriptomes ( 12 ). In principle, pioneer factors may elicit these separate functions through unique and different protein domains and co-factors ( 13 ). The identification of unique features within pioneer factors that account for their pioneering compared to transcriptional functions would in itself serve to better define the unique aspects of chromatin opening by pioneer factors.
The structure of a few pioneer factors was probed through structure function studies. For the FOXA factors that specify endoderm deri vati v es, closed chromatin binding was shown to involve the DNA binding domain (DBD) that interacts with its target DNA sequences, as well as the C-terminal domain that interacts with histones H3 and H4 ( 14 ). Further, the FoxA DBD is flanked by wings that have structural similarities to histone H1 and that may contribute to chromatin interactions by mimicry / displacement of histone H1 ( 15 ). The C-terminal domain of the pioneer EBF1 involved in B cell identity was shown to be essential for closed chromatin interaction and pioneering ( 7 ). Mor e r ecentl y, an intrinsicall y disorder ed r egion of the pioneer PU.1 was shown to be important for interaction with linker histone H1-compacted nucleosome arrays ( 16 ). These structure-function studies defined the contributions of different protein domains for pioneer activity but pioneer factor isoforms have not been investigated so far for differential activities. In contr ast, tr anscription factor isoforms with well-defined transcriptional and de v elopmental specificities were described such as the MEF2 transcription factors ( 17 ), and many alternati v ely-spliced transcription factors involved in cancer ( 18 ).
The pioneer factor Pax7 is expressed in a variety of tissues, but its pioneering role is clearly supported only in the pituitary gland where Pax7 opens a new enhancer repertoire to specify the intermediate pituitary tissue and the melanotrope cell fate ( 8 , 19 ). This activity r equir es the corecruitment and cooperation with a nonpioneer factor, the Tbox factor Tpit ( 20 ). Elsewher e, P ax7 is expressed in skeletal muscle and specifies subdomains of the neural tube ( 21 , 22 ). In the myogenic lineage, the related Pax3 and Pax7 transcription factors are critical to implement the myogenic fate and in particular for the transition of progenitors into dif ferentia ted myogenic cells ( 23 ). A clear indication of Pax7 pioneering in these latter systems remains poorly established ( 23 , 24 ), because demonstration of the pioneering ability r equir es access to the tissues or cells of origin prior to chromatin opening by the pioneer factor: this is not always easily accessible as an experimental system.
In the present wor k, we hav e inv estigated the structural features of Pax7 implicated in chromatin opening of the enhancer r epertoir e that specifies the pituitary melanotrope fate. The ability to reprogram a corticotrope cell model, the AtT-20 cells, into melanotropes enabled us to investigate the pioneer ability of naturally occurring isoforms of Pax7, together with a series of deletion and alanine replacement mutants. The Pax7 isoforms differ by just a few amino acids within the DNA binding paired domain but these suffice to create isoforms with or without pioneer ability, but with similar inherent transcriptional activities. The pioneeringdeficient Pax7 isoform is similar to the related Pax3 in their inability to activate the melanotrope transcriptome. A similar loss of pioneering ability was observed for C-terminal deletion mutants of Pax7. It is striking that similar losses of pioneering ability are produced by either introduction of two amino acids within the paired domain or by deletion of C-terminal sequences. The loss of pioneering ability is clearly linked to the loss of chromatin opening rather than to failure of chromatin recruitment. This pattern corresponds to an inability of one isoform and mutants to proceed to the second step of chromatin opening by Pax7 ( 2 ); hence, these forms can only prime but not fully activate a subset of target enhancers that are associated with the melanotrope transcriptome.

Cell culture
AtT-20 / D16v-F2 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and antibiotics (penicillin / streptomycin). To generate stable transgenic AtT-20 Pax7 cell populations, retroviral vectors (pLNCX2) containing the indicated Pax7 constructs (with 3xFlag at both N-and C-termini under control of CMV promoter) were packaged using the Platinium-E Retroviral Packaging Cell Line (Cell Biolabs, catalog #RV-101) and infections performed as described ( 25 ). Selection of retrovirus-infected cell populations was achie v ed with 400 g / ml Geneticin (Gibco, #11811-031). Resistant colonies were pooled to generate populations of hundreds of independent colonies.

Luciferase assays
AtT-20 cells were plated in 12-well dishes (4 × 10 5 cells per well) and transfections performed the next day using PEI 25K (Polysciences #23966-1) in triplicates. Transfected DNA (total 2 g in 150 l DMEM) was added to the PEI 25K solution (4 l per 150 l DMEM) and incubated a t room tempera ture f or 30 min bef ore adding 100 l of the mixture to each well. The media was changed 24 h after transfection and luciferase activity assessed 48 h after transfection. Growth medium was removed from transfected cells and 200 l of lysis buffer (Tris 100 mM, NP40 0.5% and DTT 0.001 M) added to each well. After 10 min on a vigorous shaker, 100 l of cell lysa te superna tant was used for analysis using the GloMax Navigator luminometer (Promega) using the kinetics protocol (100 l injection, speed 100 l / s, integration time 10 s). Statistical significance was assessed by 2-way ANOVA and Tukey`s multiple combinations test.

RNA-seq, ChIP-seq, A T A C-seq
RNA was extracted from 1 000 000 cultured cells for each biological replicate using Qiagen RNeasy plus mini kit. Ribosomal depletion, library preparation and flow cells preparation for sequencing were performed by the IRCM Molecular Biology Core Facility following Illumina recommendations (Illumina, San Diego, CA). ChIP experiments were performed in AtT-20 cells as described earlier ( 8 , 26 ). ATAC-Seq was performed as described in ( 8 Table S1 provides details on the antibodies used for ChIPs. ChIP-Seq and ATAC-Seq average profiles were generated by the average signal function of Easeq software ( http://easeq.net ) and the data exported to Microsoft Excel to create graphs. To view the ChIP-Seq and ATAC-Seq heatmaps, we used the Integrati v e Genome Viewer tool of Easeq ( 27 ) or a custom Python program ( 13 ). All ChIP-Seq and ATAC-Seq data except Flag ChIPs were normalised relati v e to Constituti v e and random sites. Read counts ( ±100 bp around Pax7 summit) were normalized to constituti v e and random sites as follows: normalized value = (raw counts -mean (random sites)) / (mean (constituti v e sites) -mean (random sites)).

Nuclear extracts
To pr epar e nuclear extracts, AtT-20 cells from a 100 mm Petri dish (15-20 ×

Emsa
Double-stranded DNA probes were described and labeled as in Pelletier et al. ( 13 ) .Gel shift assays were performed using 1 g of nuclear protein extracts from different AtT-20 cell lines pre-incubated 10 min on ice with 1 g of nonspecific carrier DN A pol y(dI-dC) and pol y(dA-dT) in binding buffer (25 mM HEPES pH 7.9, 84 mM KCl, 10% glycerol, 5 mM DTT). For purified polypeptides, 300 fmoles were used per binding reactions. Equal amounts of radiolabeled DNA probes (50 000 cpm) were then added for a total volume of 20 l and incubated 60 min on ice. If applicable, 1 g of Flag M2 antibody (F3165, SIGMA) was added to the mixture during the last 10 min of incubation.
Nondenaturing polyacrylamide gels (40 mM Tris, 195 mM glycine, 0.08% APS, 0.5 l / ml TEMED, 5% Acrylamide:Bisacrylamide in 19:1 ratio), were prerun in Trisglycine buffer (40 mM Tris and 195 mM glycine) for 60 min a t 300 V a t 4 • C . The gel shift reaction mixture was then loaded, and the gel run for 3 h at 300 V at 4 • C. Gels were dryed on 3MM W ha tman paper in a gel dryer at 80 • C during 60 min under vacuum. Autor adiogr aphy films (HyBlot CL, catalog # E3018) were then exposed with the dried gels for 10-48 h using an intensifier screen.

RT-qPCR
RNA was extracted from AtT-20 cells using the RNeasy Plus Mini kit (Qiagen #74134), and cDNA synthesized using 5 g RNA, oligo-dT and SuperScript III re v erse transcriptase (Invitrogen, #18080-044) following manufactur er's r ecommendations. The r esulting cDNAs (5 l of 1 / 50 dilution) were analyzed by qPCR using SYBR Green reagent (ThermoFisher Scientific #A25741) supplemented with 500 nM of each gene-specific primer pair (chosen in exons separated by intron of at least 1Kb and provided in Supplementary Table S2) in a total volume of 10 l. At least two biological replicates were analyzed in duplicates for each condition using a ViiA ™ 7 Real-Time PCR device (ThermoFisher Scientific), and results were analyzed using the accompanying software. Gene expression values were normalized relati v e to the TBP transcript and statistical significance assessed by 2-way ANOVA and Tukey`s multiple combinations test.

ChIP-qPCR
ChIP experiments were performed in AtT-20 cells as described ( 8 ). Imm unoprecipitated DN As were diluted in 100 ul and were analyzed by qPCR using the SYBRgreen reagent (ThermoFisher Scientific #A25741) completed by 500 nM of each enhancer-specific primer pair (provided in Supplementary Table S3) in a total volume of 10 l. At least two biological replicates were assessed in duplicates for each condition using a real-time ViiA ™ 7 (ThermoFisher scientific) PCR device, and the r esults wer e analyzed using the associated softwar e. ChIP r ecruitment values wer e normalized to those of two random loci (H3K9me3poor, H3K9me3rich as listed in Supplementary Table S3).

NGS data analysis
ChIP-seq and ATAC-seq paired-end sequenced reads were trimmed using Trimmomatic / 0.36 ( 28 ) and aligned to the mouse mm10 r efer ence genome using Bowtie / 2.3.5 (-v ery-sensiti v e -no-mixed -no-unal) ( 29 ). Bam files were created using view from SAMtools / 1.9 ( 30 ) and duplicated reads were removed using MarkDuplicates from Picar d / 2.17.3 (Broad Institute). Cov er age tr acks wer e cr eated using bamCoverage (-normalizeUsing RPKM -bs 10 -e) from deepTools / 3.3.0 ( 31 ). Peaks were called against sequenced input DNA or control Flag ChIP using callpeaks (-f BAMPE -p (1e-3) -g mm) from MACS / 2.1.1 ( 32 ). Pax7 recruitment site categories are based on previously published lists ( 13 ). In brief, Pax7 ChIP-seq peaks were categorized as 'Pioneered' if they had no signal before, and gains (p-value < 1e-5 for ChIP and < 1e-3 for ATAC) of H3K4me1, p300 and ATAC-seq, after Pax7 expression. 'Activated' sites gained p300 after Pax7 and Constituti v e sites wher e alr eady positi v e for all marks before Pax7. To determine H3K4me1 peak sub-categories, H3K4me1 signals were quantitated within a 200 pb window by easeq ( 33 ) around the summit of the Pax7 peaks. We next compared these with regions at 500 bp on either side of the Pax7 summits to determine whether they are equal to, lower than, or greater than, the central region of the Pax7 peaks.

Motifs analysis
The motifs recognized by Pax7 were described by Pelletier et al. ( 13 ). Presence or absence of a specific Pax7 motif within 50 bp on each side of the Pax7 ChIP summit at pioneer sites was assessed using the command Homer findMotifsGenome.pl ( http://homer.ucsd. edu/homer/ngs/peakMotifs.html ). The number of Pax7 motifs present within a window of 200 bp around the summit of each pioneer site was assessed using a custom program based on Python ( 13 ).

RESULTS
The pioneer factor Pax7 is expressed as different isoforms that result from differential exon splicing ( 22 , 37 ). Prior work defined the nature of isoforms expressed in muscle ( 21 , 37 ) and brain ( 38 , 39 ) tissue and we assessed the relati v e e xpr ession of P ax7 isoforms in the pituitary gland where we previously characterized the pioneer ability of P ax7 ( 8 , 19 , 20 ). The P ax7 isoforms ar e characterized by the differential insertion of either a single or a pair of amino acids at two positions within the DNA binding paired (PD) domain. The first differential splicing involves a glutamine (Q) r esidue inserted befor e helix 4 of the helix-turn-helix (HTH) RED domain (Q+) whereas the second differential splicing e v ent involv es a gly cine-leucine pair inserted (GL+) at the beginning of helix 6 of the same RED HTH domain ( Figure 1 A and Supplementary Figure S1a). In the pituitary intermediate lobe, the predominant isoform as assessed by RT-qPCR is the Q + GL-isoform that r epr esents about 75% of the mRN A w hereas the remainder is the Q + GL+ isoform, with no expression of the Q-isoforms (Figure 1 A). The Pax7 Q + GL-isoform was used in our prior work ( 8 , 13 , 19 , 20 ) and constitutes the r efer ence isoform in the present work.

GL+ isoforms fail to activate the melanotrope transcriptome
In order to assess the properties of the Pax7 isoforms, we used the gain-of-function system previously developed (Figure 1 B) in mouse AtT-20 cells that r epr esent a model of pituitary corticotrope cells and that undergo reprogramming into melanotropes following Pax7 expression ( 19 ). Pools of AtT-20 cells each expressing a different Pax7 isoform under control of retroviral sequences were obtained and shown to express similar levels of PAX7 proteins and le v els that are similar to those of the mouse pituitary intermediate lobe (Supplementary Figure S1b). Complete transcriptomes were obtained for these cells by RNA-Seq and anal yzed for differentiall y expr essed genes (Figur e 1 C and Supplementary Figure S1c). Comparison of the transcriptomes re v eals that GL+ isoforms affect expression of far fewer genes than the GL-isoforms; further, the presence of Q (Q + isoforms) has a marginal effect (Figure 1 C and Supplementary Figure S1d). Significantly, volcano plot representa-tion of differentially expressed genes between the Q + GLand Q + GL + isoforms re v eals that most (91%) previously identified ( 8 ) melanotrope genes are preferentially activated in Q + GL-compared to Q + GL+ cells (Figure 1 D and Supplementary Figure S1e, f). The ability of Pax7 to implement this melanotrope program of gene expression thus appear to be mostly dependent on the GL-isoforms and to be pre v ented by the insertion of GL in the PD domain of the GL+ isoform.
This differential gene expression could be due to differences in transcriptional activation capability, and this was assessed directly by co-transfection of the various isoforms with reporter plasmids containing either the natural enhancer of the Pcsk2 gene (Figure 1 E) or reporters containing multimers of various Pax7 DNA binding sites ( 13 ) (Supplementary Figure S1g). The different Pax7 isoforms have very similar transcriptional ability in all these systems. Hence, it is not transcriptional ability that accounts for the unique properties of the isoforms. In striking contrast, the le v els of the melanotrope-specific Pcsk2 mRNA ar e very differ ent in cells expr essing differ ent isoforms (Figure 1 F). Indeed, the GL-isoforms express significantly higher le v els of Pcsk2 mRNA than the GL+ isoforms with the greatest difference being between the two isoforms expressed in pituitary, namely the predominant Q + GLcompared to the Q + GL+ isoform. Expression of Pcsk2 was previously shown to depend on the Pax7 pioneer ability ( 8 , 19 ).

All pax7 isoforms exhibit similar genomic occupancy
Since the intrinsic transcriptional activity of the Pax7 isoforms does not explain their differential effect on mRNA expression, we assessed the ability of each isoform to be recruited to chromatin and trigger chromatin opening. We first performed ChIP-Seq for P ax7 r ecruitment in cells expr essing the differ ent isoforms and globally all isoforms showed similar binding (Figure 2 A and Supplementary Figure S2). Subtle differences appeared howe v er when Pax7 recruitment is compared at previously described subsets of enhancers ( 8 ) (Figure 2 B), namely Constituti v e (open and acti v e before Pax7), Activated (gain of ATAC, H3K4me1 and p300 signals after Pax7 binding with prior weak signals) and in particular at Pioneer enhancers (that had no acti v e mar ks befor e P ax7 but gained marks after). Wher eas binding was similar for all isoforms at Constituti v e and Activated enhancers, recruitment of the Q + GL+ isoforms is impaired, but not lost, at Pioneer enhancers compared to the r efer ence Q + GL-isoform (Figur e 2 C and D), except for a small group of 115 Pioneer sites (13%, labelled as GLsensiti v e binding) where binding is lost (Figure 2 C).
The small decrease in genomic recruitment of the GL+ isoforms could be due to intrinsic DNA binding ability, and this was assessed by in vitro binding in electrophoretic mobility shift assay (EMSA) experiments (Figure 2 E). We previously ( 13 )  The 'common' subsets r epr esent genes that ar e differ entially expr essed f or all isof orms compar ed to P ax7 GL-wher eas the 'differ ent' subsets r epr esent genes that are differ entially expr essed in at least one isoform. The scale +7 / -7 r epr esents the log 2 fold change with P values < 0.05. The T test P values of gene expression changes are sho wn belo w the heatmaps. ( D ) Volcano plot comparison of differentially expressed genes in AtT-20 cells expressing the Q+ GL+ isoform compared to the reference Q + GL-isoform. Highlighted melanotrope (red dots) and corticotrope (black dots) genes reflects the ability of the r efer ence P ax7 isoform to activate the melanotrope transcriptome and to r epr ess genes of the corticotrope transcriptome as pr eviously defined ( 8 ).
( E ) Transcriptional activity of differ ent P ax7 isoforms was assessed using a luciferase reporter dri v en by the melanotrope-specific Pcsk2 gene enhancer ( 19 ). ability for the various isoforms with a slightly weaker binding of the GL+ compared to the GL-isoforms: this was true for all probes studied (Figure 2 E). It is thus possible that the weaker recruitment observed for the GL+ isoforms in ChIP experiments may in part be due to lower intrinsic DNA binding affinity. Howe v er, the decr ease in Q + GL+ r ecruitment compared to the Q + GL-isoforms (Figure 2 C) appears greater than expected from differences in in vitro binding.

Melanotrope enhancers are primed rather than activated by GL+ isoforms
In order to assess the ability of Pax7 isoforms to initiate chromatin opening, we performed ATAC-  Figure S3a-e). It is ∼45% of the Pioneer sites that fail to show significant ATAC-Seq signal in presence of Q + GL+ compared to Q + GL-isoform. We assessed whether the underlying Pax7 DNA binding motifs present in each subset of enhancers could account for the sensitivity to insertion of the GL residues in the PD domain. Neither frequency of individual motifs (Supplementary Figure S3f) nor total number of Pax7 binding motifs (Supplementary Figure S3g) can explain the difference between the properties of the two Pioneered subsets.

Activa ted subca tegories (Supplementary
To further assess Pax7-dependent deposition of acti v e enhancer marks, we performed ChIP-Seq for H3K4me1, a mark of potentially acti v e enhancers. Interestingly, the subset of enhancers that did not show significant ATAC-Seq signals, do show a small gain in H3K4me1 that mostly presents as a single weak peak compared to the bimodal H3K4me1 distribution typically observed at fully acti v e enhancers (Supplementary Figure S3c). This single peak H3K4me1 profile is typical of Primed enhancers and hence the subset of enhancers that are impaired by the presence of GL+ fails to be fully activated but is nonetheless remodelled from the inacti v e to the Primed state. Since this change in H3K4me1 profile is the most significant GL-dependent altera tion, we quantita ted the H3K4me1 profiles as single or bimodal profiles, the latter subgroup being further subdivided as symmetric or asymmetric bimodal profiles (Figure 3 A, B). Examples of these profile changes are shown for two Pax7-dependent enhancers (Figure 3 D). The H3K4me1 profiles thus provide the best illustration of the impact of GL insertion for pioneer action: whereas a subset of sites is insensiti v e (GL-insensiti v e), thr ee subsets that r etain P ax7 binding with decreased ATAC signal (Figure 3 C) exhibit a shift from bimodal to single peak (either symmetrical or asymmetric) H3K4me1, indicati v e of a failure of nucleosome displacement (Figure 3 B, E).
We then assessed how stable Pax7 binding may differ between the subsets of pioneered enhancers (Figure 3 F) and whether this may account for the difference in chromatin opening re v ealed by ATAC-Seq (Figur e 3 G). While P ax7 binding is more affected for the GL-sensiti v e subsets compared to the GL-insensiti v e subset, significant binding remained for most sites within the subsets of GL-sensiti v e enhancers (Figure 3 F). Howe v er, the loss of chromatin opening ability re v ealed by ATAC-Seq is significantly different between the subsets (Figure 3 G). The analysis of transcription factor DNA binding motifs did not provide clues to explain the differences between GL subsets (Supplementary Figure S4).
We further compared the status of the GL-sensiti v e and GL-insensiti v e enhancers by performing ChIP-Seq for Tpit, a Pax7 cooperating nonpioneer ( 20 ), and chromatin remodelling proteins associated with Pax7 pioneering ( 8 ). Wher eas r ecruitment of the Pax7-cooper ating tr anscription factor Tpit and of the remodelling complex proteins Ash2 and Brg1 are similar for Q + GL+ and Q + GL-at the GLinsensiti v e subset of enhancers, their recruitment is about half for the GL-sensiti v e subset ( n = 259) of enhancers that exhibit a single peak of H3K4me1 (Figure 3 H-J and Supplementary Figure S5). While these recruitment decreases ar e corr ela ted with weaker chroma tin opening (ATAC), they ar e mor e strikingly associated with a failur e of nucleosome eviction (single peak rather than bimodal H3K4me1 profiles) and enhancers that appear stalled in the Primed state (Figure 3 E). This could be due to inadequate recruitment of Tpit and / or of the SWI / SNF complex containing Brg1.
Reduced chromatin recruitment of the Pax7 GL + isoform (Figure 3 A) could account, at least to some extent, for its decreased pioneering ability as re v ealed by weaker ATAC-Seq (Figure 3 C) and H3K4me1 deposition (Figure  3 B). Howe v er, P ax7 ChIP-Seq r ecruitment str ength does not correlate well with either H3K4me1 or ATAC signals for both Pax7 GL-and Pax7 GL+ isoforms (Supplementary Figure S6). Further, each subset of GL-sensiti v e Pioneer sites (Figure 3 A) have strong Pax7 ChIP-Seq signals that overlap the signal strengths of the GL-insensiti v e subset, arguing against a strict relation between chromatin recruitment strength and pioneering outcome. This point is best exemplified by extracting the top and bottom quartiles (25%) of the Pioneered sites (based on Pax7 recruitment signals of all 755 sites, excluding the 'No binding' subset) and assessing their average profiles of Pax7 GL+, H3K4me1 ChIP-Seq and ATAC-Seq signals at the GLsensiti v e compared to GL-insensiti v e subsets (Figure 3 K). This analysis clearly highlights that the strong Pax7 recruitment sites (top quartile -25%) of both GL-sensiti v e and GL-insensiti v e subsets have different outcomes with single peaks and bimodal H3K4me1 distributions, respecti v ely. In contrast and in support of the conclusion that recruitment strength does not correlate with outcome, the weaker sites (bottom quartile) also show either single peak (GLsensiti v e) and bimodal (GL-insensiti v e) H3K4me1 profiles despite quite weak and similar Pax7 GL+ recruitment (Figure 3 K). Thus in spite of a global decrease in chromatin recruitment (ChIP-Seq) for the Pax7 GL+ isoform that may contribute to the loss of pioneer ability, the outcome of its action appears to be qualitati v ely different for the GLsensiti v e and GL-insensiti v e targets at both strong and weak Pax7 GL+ recruitment sites.

Pax3 (GL-) fails to open the GL-sensitive melanotrope specific enhancers
The Pax transcription factor that is most similar to Pax7 is Pax3 and indeed, these two factors cooperate for implementation of the skeletal muscle program ( 40 ). Even though Pax3 is not expressed in the pituitary, it is interesting to compare the pioneering ability of those two factors in view of their co-expression in other tissues. Pax3 was thus introduced into AtT-20 cells and expressed at similar le v els as Pax7 (Supplementary Figur e S7a, b). P ax3 and Pax7 have very similar sequence in their DNA binding domains and are relati v ely conserv ed in their C-terminal regions ( Figure 4 A and Supplementary Figure S7a), but Pax3 does not have the OAR sequence present in the C-terminus of P ax7. P ax7 and P ax3 have similar in vitro binding ability when assessed by EMSA (Figure 4 B). Howe v er, AtT-20 cells expressing Pax3 showed altered expression of far fewer genes compared to Pax7 Q + GL-and strikingly, a similar subset of genes that are unaffected by the GL+ isoform of Pax7 also fail to be upregulated by Pax3 that is

Pax7 C-terminus is required for recruitment to melanotrope specific enhancers
To define Pax7 domains involved in pioneer activity, we conducted C-terminal deletions of Pax7 (Figure 5 A). The choice of C-terminal deletions endpoints was guided by alignment of the mouse Pax7 and Pax3 sequences in order to delete blocks of homology where possible (Supple-mentary Figure S7a). Pools of AtT-20 cells expressing these deletions mutants at similar le v els (Supplementary Figure  S7b) were first assessed for in vitro DNA binding ability. These gel-shift experiments used the different probes recognized by either or both PD and HD domains and show similar in vitro binding activity for all deletion mutants, including Pax7-291 that is deleted of the entire C-terminus (Figure  5 B). The smallest C-deletion mutant, Pax7-465, is deleted of C-terminal sequences that are unique to Pax7 compared to Pax3 and that contain the OAR motif. Pax7-465 exhibits similar recruitment by ChIP-Seq as the reference Pax7-503 at pioneered sites ( Constituti v e sites. Finally, almost complete deletion of the C-terminus to position 291 abrogates recruitment to Pioneer sites and reduces recruitment to Constituti v e sites (Figure 5 C) despite intact in vitro (EMSA) binding (Figure 5 B). The critical interval between positions 431 and 465 is interesting because it is quite conserved compared to Pax3 and because it contains multiple putati v e sites of phosphorylation (Supplementary Figure S7a). To identify critical residues within this interval for Pax7 recruitment to pioneer sites, we performed a series of alanine replacement mutations throughout the interval and assessed those Pax7 mutants (Supplementary Figure S7b-d). Mutagenesis of either residues 459-461 or 463-464 to alanines had similar deleterious effects on recruitment to the Pcsk2 pioneer site (Supplementary Figure S7c) but not to a transcriptionally activated site ( Pde2a ), and corresponding effects on Pcsk2 mRNA le v els (Supplementary Figure S7d). These mutations have marginal effects on a gene, Lmcd1, that is transcriptionally activated (ie not requiring pioneer action) by Pax7 (Supplementary Figure S7d This behavior is reminiscent of sites that are affected by the insertion of GL in the Pax7 Q + GL+ isoform ( Figure  3 ). Comparison of the GL-sensiti v e subsets of Pioneer sites with those affected by the Pax7-465Ala 459-461 showed that the former subset is overlapping with sites that show the loss of Pax7 recruitment or a shift from H3K4me1 bimodal to single peak with loss of ATAC signal following C-terminal deletion (Figure 6 A-E). Hence, enhancers that exhibit impaired chromatin opening together with Primed rather than Acti v e state because of GL+ insertion in the PD domain are also dependent on C-terminal sequences between 459-465 for chromatin remodelling. Overall, the Pax7-465Ala 459-461 mutant shows impairment at a greater number of sites compared to the GL+ isoform but the latter are all included in the former group. The dependence of the same subset of enhancers on these two structural features of Pax7 for pioneer action may suggest that these enhancer's contexts are similar. Further, these enhancer subsets exhibit parallel losses of Tpit recruitment (Figure 6 F). In view of the importance of Tpit recruitment for complete enhancer opening ( 2 ), the reduced recruitment of Tpit at GL-sensiti v e enhancers can account for the observed blockade at the Primed state.

DISCUSSION
We showed in the present work that the GL+ isoform of Pax7 produced by differential splicing of an extra two amino acids into the DNA binding PD domain loses significant pioneer ability without any effect on its transcription activation potential. The loss of pioneer ability is not complete as it affects a subset of its Pioneer sites in AtT-20 cells. The affected subsets include one (about 13% of Pioneer sites) that loses Pax7 recruitment in ChIP-Seq experiments and other subsets (three GL-sensiti v e subsets alltogether r epr esenting 59% of Pioneer sites) that fail to be completely opened and remain in the Primed state ( Figure 3 ). Nonetheless, these subsets are significant as they lead to failure to activate most of the melanotrope transcriptome. The similar transcriptional activity of the various Pax7 isoforms (Figure 1 E) suggests that they may be redundant for transcriptional maintenance. Howe v er, the isoform-specific pioneering ability implies tha t implementa tion of new enhancer r epertoir es by P ax7 during de v elopment may be linked to production of the appropriate isoform, the GL-isoform in the pituitary. While the different isoforms of Pax7 are expressed in adult tissues, there is no documentation of these differential splicing e v ents at critical times during either pituitary, muscle or brain de v elopment.
Another differential splicing of the Pax7 gene appears crucial for function: indeed, a splicing mutant resulting in loss of the OAR-containing isoform of human PAX7 is associated with a global neurode v elopmental delay ( 39 ). This mutation results in production of PAX7 isoforms that are similar to Pax7-465 terminating with coding sequence of the Pax7 exon 8. Since our analyses ( Figure 5 ) indicate that Pax7-465 is sufficient for pioneer ability and function in pituitary cells, this suggests that some PAX7 functions in neural tissues might be uniquely dependent on the OAR domain.
Could relati v e af finity for genomic / chroma tin target sites be the limiting factor that accounts for loss of pioneering ability by the GL + isoform and / or the C-terminal deletion m utants? A slightl y lower a pparent affinity of the GL + isof orm f or target DNA in EMSA (Figure 2  to C-terminal deletions ( Figure 5 ) show overlapping distributions of Pax7 ChIP-seq recruitment strength and ATAC signals. So, these distributions argue against DNA binding strength as a primary determinant of pioneer ability (Figure 6 A). Also, the apparent Pax7 ChIP recruitment signal intensity in itself is not well correlated with chromatin opening ability (Supplementary Figure S6) and whether considering strong or weak Pax7 GL + sites of the GLsensiti v e or GL-insensiti v e subsets, their response is not strictly correlated with Pax7 recruitment strength (Figure 3 K). A similar conclusion was reached by analysis of initial recruitments at differ ent P ax7 pioneer sites wher e initial r ecruitment str ength is inversely correlated with the le v els of linker histone H1 ( 2 ). But it could be a combination of DNA and chromatin interaction strengths, together with ability to recruit the cooperating TF Tpit (as discussed below), that ultimately determine chromatin modification and opening. In this respect, it is noteworthy that Pax7-291 which is deleted of most C-terminal sequences with an intact DNA binding domain, binds DNA in vitro ( Within the Pax7 C-terminus, the progressi v e deletions suggest that sequences within both the aa291-431 and aa431-465 regions contribute to chromatin interaction at the closed Pioneer sites. Similarly, chromatin and linker histone H1 interaction of the pioneer PU.1 was mapped outside its DNA binding domain ( 16 ). In summary, despite some clear relationship between recruitment strength and pioneering ability, it is likely that other parameters are critical to explain the loss of pioneering ability. It is noteworthy that the GL+ isoform or the deletion mutants that are impaired in pioneering ability exhibit subsets of targets that appear blocked in the primed state rather than being fully activated by the r efer ence P ax7 (Figur e 7 ). Blockade at the primed state is expected of a pioneering pr ocess going thr ough the first step of chr omatin opening but failing to undergo the second cell-cycle dependent step that involves nucleosome displacement and recruitment of the SWI / SNF complex ( 2 ). We also showed that this second step r equir es the cooperating nonpioneer factor Tpit in addition to cell replication ( 2 ). It is thus possible that pioneer ability of Pax7 mutants may result in the impaired ability of these mutants to recruit / interact with Tpit. The loss of Tpit recruitment at the GL-sensiti v e and Pax465Ala 459-461 subsets of impaired pioneer sites is similar (Figure 6 F) and hence this is a reasonable explanation. Further investigation of the Pax7-Tpit interaction would be needed to support this model behond the present correlation. It is howe v er noteworthy that the subset of sites that are only primed by Pax7 in AtT-20 cells does not recruit Tpit ( 2 , 8 ) and knockout of the Tpit gene results in failure of Pax7 pioneer capacity in the mouse pituitary ( 20 ). Further, Pax3 behaves very similarly to the GL+ Pax7 isoform in pituitary cells despite Pax3 being GL-and it also shows reduced Tpit recruitment ( Figure 6 F). This may be taken to suggest that the impact of GL insertion in Pax7 is at a tridimensional le v el, and this would be consistent with prior work suggesting that a unique conformation of Pax7 (that r equir es both intact paired and homeo domains) is involved in pioneering ability ( 13 ).
Consider ed globally, the r esults pr esent a pictur e of gradual loss of pioneering ability by the GL + and deletion mutants that affect the same subsets of pioneer sites suggesting that it is the intrinsic properties of those sites that underlie their sensitivity. The nature of the critical properties that are involved remains to be better defined.

DA T A A V AILABILITY
Sequencing data are available under GSE87185 (previous data) and GSE225231 (data from present paper).