Meiotic progression in male mice depends on SWI/SNF-regulated gene expression (Menon et al., 2019; Menon et al., 2021). These studies raise the prospect of distinct SWI/SNF subcomplexes governing stage-specific meiotic transcription (Menon et al., 2021). To understand the meiotic functions of the biochemically distinct SWI/SNF BAF subcomplex, we examined the spermatogenic expression profiles of the BAF subunits Arid1a and Arid1b using single-cell RNA-seq data generated from testes (Ernst et al., 2019). Arid1a mRNA expression occurred at various spermatogenic stages with notable expression at pachynema of meiotic prophase I (Fig. S1A-C).

Arid1b mRNA went undetectable until late in spermiogenesis (Fig. S1A, D-E). Consistent with its mRNA expression profile, pachytene spermatocytes featured abundant ARID1A protein, which was uniformly spread (Fig. 1). Late in diplonema, ARID1A appeared preferentially enriched on the sex chromosomes as compared to autosomes (Fig. 1). These data implicate ARID1A in meiotic sex chromosome inactivation (MSCI), a process essential for meiotic progression in males.

To determine whether meiotic spermatocytes require ARID1A, we generated a germ cell-specific knock-out of Arid1a using the Stra8-Cre transgene expressed in spermatogonia (Sadate-Ngatchou et al., 2008). Contrary to our expectations, meiosis appeared unperturbed, as evidenced by seminiferous tubules featuring round and elongated spermatids in Arid1a^cKO testes (Fig. S2A). Furthermore, Arid1a^cKOepididymides featured mature spermatozoa, suggesting that mutants are capable of normal spermatogenesis (Fig. S2B). Although these data indicate that ARID1A is dispensable for spermatogenesis, we wanted to confirm that these results were not a technical artifact arising from inefficient Stra8-Cre mediated excision of the Arid1a^fl allele.

To address this question, we examined Arid1a transcripts isolated from Arid1a^WT and Arid1a^cKO Sta-Put purified populations of pachytene spermatocytes and round spermatids using RT-PCR (Fig. S3A). For cDNA synthesis, we used primers that amplify a 612 bp region spanning the Arid1a floxed exons 5 and 6. Assuming 100% CRE efficiency, we would expect to observe a 281 bp cDNA product associated with mRNA isolated from Arid1a^cKO spermatogenic cells. Instead, we observed cDNAs representative of the floxed (fl) allele and, to a greater extent, the excised (Δ) allele from mRNA isolated from Arid1a^cKO pachytene spermatocytes. These results indicated inefficient CRE activity (Fig. S3A). More importantly, the data suggest that a fraction of pachytene spermatocytes in Arid1a^cKO testes fail to undergo CRE-mediated excision resulting in escapers (internal controls). An examination of ARID1A levels in testes cryosections and meiotic spreads by immunofluorescence (IF) revealed a heterogenous population of pachytene spermatocytes consisting of mutants lacking and internal controls expressing ARID1A (fig. S3B, C).

Furthermore, examining the meiotic profile associated with Arid1a^cKO revealed an accumulation of mutant spermatocytes (ARID1A^-) at mid-pachynema relative to Arid1a^WTspermatocyte spreads (Table 1). This result is indicative of a pachytene arrest. Although a few diplotene spermatocytes displaying a moderate reduction in ARID1A levels occurred in Arid1a^cKO relative to Arid1a^WT meiotic spreads (Table 1), mutant (ARID1A^-) diplotene and metaphase-I spermatocytes were undetectable in Arid1a^cKO testes (fig. S3B). It is possible that the pachytene spermatocytes that escape CRE activity progress to subsequent meiotic stages giving rise to genotypically normal gametes. All round spermatids isolated from Arid1a^cKO testes appeared only to express the normal transcript associated with the floxed allele (fig. S3A). Therefore, the perceived lack of a phenotype is an artifact of inefficient Stra8-Cre inefficiency. But more importantly, the data indicate that ARID1A is essential for progression beyond pachynema.

The meiotic requirement of ARID1A and, more importantly, its association with the sex chromosomes at diplonema prompted us to examine its role in MSCI. We hypothesized that ARID1A might play a role in the transcriptional silencing of the sex-linked genes during meiosis. To address this question, we performed RNA-seq on Sta-Put purified populations of pachytene spermatocytes to profile changes in transcript abundance upon ARID1A deficiency. Differential analysis of gene expression using DeSeq2 (Love et al., 2014) revealed an equal proportion of significantly (FDR < 0.05) misexpressed genes displaying either elevated (up-regulated, n= 5824) or reduced (down-regulated, n=5821) transcript abundance in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 2A). Notably, we detected significant misexpression of 53 % of all the sex-linked coding genes (593/1105) in response to an ARID1A deficiency. Amongst these, 86.4% displayed elevated expression (n=512), whereas only 13.6% displayed reduced (n=81) transcript abundance. This skew was even starker when only considering sex-linked genes misexpressed by a magnitude of 2 LOG-fold or higher. Here, 97% (297/306) of the misexpressed sex-linked genes displayed increased transcript abundance in response to an ARID1A deficiency. Therefore, ARID1A predominantly affects the repression of sex-linked genes during pachynema.

An examination of changes in the average transcript abundance on a chromosome-wide basis showed that expression from the sex chromosomes was significantly higher than that from autosomes in response to the loss of ARID1A at pachynema (Fig. 2B). Therefore, ARID1A regulates the transcriptional repression of the sex chromosomes, implicating it in MSCI. Its predominant influence on sex-linked gene regulation may result from a preferential association of ARID1A with the sex chromosomes during meiosis. This association is undoubtedly the case late in diplonema. However, earlier in pachynema, when MSCI initiates, no sex chromosome-specific pattern of ARID1A enrichment was discernible, at least by IF (Fig. 1).

We next determined ARID1A genomic localization at a higher resolution by CUT&RUN (Cleavage Under Targets & Release Using Nuclease) in pachytene spermatocytes. We detected ARID1A occupancy at promoters of differentially regulated target genes, irrespective of their chromosomal location, in Arid1a^WT relative to Arid1a^cKO (negative control) pachytene spermatocytes (Fig. S4). This result is consistent with the genome-wide distribution of ARID1A observed in pachytene spermatocyte spreads by IF (Fig. 1). Unnoticeable to IF, pachytene spermatocytes displayed a preferential association of ARID1A with promoters of normally repressed sex-linked genes, when detected by CUT&RUN (Fig. S4B), emphasizing its role in MSCI.

Unlike autosomal homologs that complete pairing during pachynema, the non-homologous regions of the sex chromosomes feature unrepaired DNA double-strand breaks (DSBs). These DNA DSBs recruit γH2Ax, a product of DNA damage response (DDR) signaling pathways essential for establishing and maintaining MSCI (Turner, 2007). Therefore, to determine whether ARID1A regulated MSCI by influencing DDR signaling, we first monitored the association of γH2Ax with the sex chromosomes in response to the loss of ARID1A. By IF, γH2Ax accumulation on the sex body appeared unperturbed in the mutant (ARID1A^-) relative to internal controls (ARID1A⁺) and Arid1a^WT pachytene spermatocytes (Fig. S5A). Consistent with this, the recruitment of ATR (Ataxia Telangiectasia and Rad3 related), a kinase known to phosphorylate γH2Ax and initiate MSCI (Turner et al., 2004; Royo et al., 2013), appeared normal in the absence of ARID1A. These data indicate that sex-linked DDR signaling occurred independently of ARID1A (Fig. S5B). Furthermore, MDC1 (Mediator of DNA damage Checkpoint 1), a known γH2Ax reader that is essential for sex body formation (Ichijima et al., 2011), remains robustly associated with the asynapsed sex chromosomes in the absence of ARID1A during pachynema (Fig. S5C). These data show that the mechanisms underlying ARID1A-mediated repression of the sex-linked transcription are mutually exclusive to DDR pathways regulating sex body formation.

Apart from γH2Ax, the association of the sex chromosomes with RNA polymerase II (RNAPII) distinguishes the sex body from the autosomes during pachynema. Normally, pachytene spermatocytes display reduced levels of RNAPII on the sex body relative to autosomes (Khalil et al., 2004). This sub-nuclear localization of RNAPII coincides with increased transcriptional output, which peaks at diplonema (Ernst et al., 2019), underscoring the importance of targeted mechanisms regulating sex chromosome repression. Therefore, we were curious to test whether ARID1A influenced the nuclear localization of RNAPII during pachynema. We performed IF to monitor the localization of the actively transcribing (elongating) form of RNAPII marked by Serine2 phosphorylation (pSer2) on its carboxy-terminal domain (Noe Gonzalez et al., 2021). We monitored elongating RNAPII (pSer2) because its pausing regulates meiotic transcription (Alexander et al., 2022). By co-staining for RNAPII (pSer2) and SYCP3, we could stage pachynema and identify the sex chromosomes. While RNAPII (pSer2) localization to the autosomes appeared similar, we noticed increased levels of RNAPII (pSer2) association with the sex chromosomes in Arid1a^cKO relative to Arid1a^WT mid and late pachytene spermatocytes (Fig. 3A). By quantifying the RNAPII (pSer2) signal generated by IF, we detected a 1.5-fold increase (p=0.0003) in the average RNAPII (pSer2) fluorescence associated with the sex chromosomes in Arid1a^cKOrelative to Arid1a^WT pachytene spermatocytes (Fig. 3B).

Due to technical difficulties, we could not simultaneously stain for ARID1A alongside SYCP3 and RNAPII (pSer2), making it impossible to distinguish internal controls from mutants in the Arid1a^cKO meiotic spreads. This result suggests that the 1.5-fold increase in RNAPII association with the sex body in Arid1a^cKOrelative to Arid1a^WT pachytene spermatocytes is likely underestimated. Therefore, our results indicate that ARID1A facilitates sex-linked gene repression by limiting the association of elongating RNAPII with the sex chromosomes during pachynema.

Next, we hypothesized that ARID1A-governed chromatin remodeling might underlie its role in limiting RNAPII accessibility on pachytene sex chromosomes. Therefore, we profiled changes in chromatin accessibility in response to the loss of ARID1A using ATAC-seq. Normally, ATAC-seq peaks (accessible chromatin) are mapped comparably across promoters (34%), intergenic (22.8%), and intronic (21.5%) regions. In contrast, the loss of ARID1A resulted in a dramatic shift towards predominantly promoter-associated ATAC-seq peaks with only a minority mapping to intergenic and intronic regions (Fig. 4A).

Consistent with an increase in the proportion of promoter-associated peaks, we detected an increase in accessibility across transcription start sites (TSS) associated with both autosomal and sex-linked DEG’s in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 4B,C). Furthermore, in the case of autosomal targets, chromatin accessibility at their TSSs was enhanced irrespective of their transcriptional status, appearing indistinguishable between down-regulated and up-regulated genes upon the loss of ARID1A (Fig. 4B). In contrast, on the sex chromosomes, the loss of ARID1A seemed to have a prominent effect on TSS’s associated with up-regulated genes, which displayed greater chromatin accessibility relative to their down-regulated counterparts (Fig. 4C). Therefore, the increased promoter accessibility of normally repressed sex-linked genes may underlie their persistent transcription by RNAPII upon the loss of ARID1A. Overall, these data highlight a role for BAF complexes in limiting promoter accessibility, especially on the sex chromosomes during pachynema.

Chromatin remodelers regulate DNA accessibility by altering nucleosome positioning or composition (Clapier et al., 2017). The latter outcome is interesting in the context of MSCI, given that the sex body typically displays a striking enrichment of the variant histone H3.3 while concomitantly appearing depleted of the canonical histones H3.1/3.2 (van der Heijden et al., 2007; Yuen et al., 2014). Given that human ARID1A is known to regulate H3.3 genomic associations (Reske et al., 2022), we tested whether a similar mechanism governed H3.3 localization to the sex chromosomes. To address this, we monitored H3.3 localization in pachytene spermatocytes by IF, simultaneously staining for ARID1A and HORMAD1. The former aided in mutant identification, while the latter labeled the pachytene sex chromosomes. Consistent with previous reports, H3.3 can be seen preferentially coating the sex chromosomes distinguishing it from autosomal chromatin during normal pachynema (Fig. 5A). Interestingly, this sex body association of H3.3 was detected less frequently early in pachynema (17%), becoming more pervasive during the mid (70 %) and late (95%) stages of pachynema. In contrast, in the absence of ARID1A, H3.3 staining on the sex chromosomes was indistinguishable from their autosomal counterparts throughout pachynema (Fig. 5A). Notably, the majority of mutant mid (89%) and late (77%) pachytene spermatocytes lacked the typical sex body enrichment of H3.3 (Fig. 5A), indicating that ARID1A dictates the preferential accumulation of H3.3 on the sex body.

Next, we were curious to know the consequence of reduced H3.3 association on H3.1/3.2 occupancy on the sex body in response to the loss of ARID1A. We performed IF to monitor the association of H3.1/3.2 with the sex chromosomes marked by HORMAD1 in response to the loss of ARID1A. Like H3.3, canonical H3.1/3.2 appeared uniformly distributed genome-wide in most early pachytene spermatocytes (92%). At the onset of pachynema, sex chromosomes mostly displayed comparable H3.3 and H3.1/3.2 levels (Fig. 5A, B). However, this balance changed during the subsequent sub-stages of pachynema. Although robust H3.3 levels remained, H3.1/3.2 appeared excluded from the sex chromosomes in most mid (56%) and all late pachytene spermatocytes (Fig. 5A, B). In contrast, the loss of ARID1A increased the proportion of mid (66%) and late (40%) pachytene spermatocytes retaining H3.1/3.2 on sex-linked chromatin (Fig. 5B). The abnormal retention of canonical H3.1/3.2 coincides with a lack of H3.3 enrichment on sex-linked chromatin in response to the loss of ARID1A.

To determine whether these abnormal kinetics result from a genome-wide deficiency in H3.3, we compared its levels by western blotting to find no difference in H3.3 abundance in Arid1a^cKO relative to Arid1a^WT spermatocyte enriched populations (Fig. S6A). We also considered an alternative possibility that ARID1A might promote the sex-linked enrichment of H3.3 by influencing the expression of cognate chaperones like DAXX and HIRA during murine pachynema (Rogers et al., 2004; van der Heijden et al., 2007). We addressed this by monitoring DAXX and HIRA levels in response to the loss of ARID1A using IF. In the case of DAXX, we examined meiotic spreads to find no changes in either the overall levels or sex chromosome localization of DAXX in response to the loss of ARID1A (Fig. S6B). The same was also true of HIRA, whose levels and nuclear localization in mutants appeared comparable to that seen in internal (from Arid1a^cKO) and normal (Arid1a^WT) spermatocytes (Fig. S6C). Therefore, ARID1A impacts H3.3 accumulation on the sex chromosomes without affecting its expression or incorporation during pachynema. Overall, our data suggest that ARID1A influences MSCI by regulating the composition of sex-linked chromatin.

To define the changes in the sex-linked associations of H3.3 at a higher resolution, we performed CUT&RUN on pachytene spermatocytes isolated from Arid1a^WT and Arid1a^cKOtestes. Despite the appearance of a robust H3.3 IF signal from the sex body during normal pachynema (Fig. 5A), the Macs2 peak caller (Zhang et al., 2008) identified very few sex-linked peaks (n =224). These peaks were overwhelmingly associated with intergenic regions (Fig. S7A). Comparatively, a dramatically higher number of peaks (n=12183) primarily associated with genic (promoter, intron, and exon) regions occurred in Arid1a^cKOpachytene spermatocytes (Fig. S7A). More interestingly, there appeared to be a shift from few but overwhelmingly sex-linked H3.3 peaks (n=224; 98%) to autosomal H3.3 peaks (n=11512; 94.5 %) in response to the loss of ARID1A (Fig. S7A). Additionally, while the proportion of H3.3 sex-linked peaks in Arid1a^cKO pachytene spermatocytes were few (n=671; 5.5%), they outnumbered their Arid1a^WT counterparts by 3-fold. These data suggest that the loss of ARID1A strongly influences H3.3 genomic associations.

Due to the increased representation of promoter-associated peaks in response to the loss of ARID1A, we monitored H3.3 association with TSSs of genes differentially regulated by ARID1A. Consistent with the genomic annotation of H3.3 peaks, we could not detect H3.3 enrichment at TSSs associated with ARID1A-regulated sex-linked genes relative to IgG control. H3.3 signal spread gene-wide in Arid1a^WT pachytene spermatocytes (Fig. S7B, Fig. S8A-top panel). In contrast, we observed a striking accumulation of H3.3 at the TSSs of sex-linked genes misexpressed without ARID1A (Fig. S7B, Fig. S8A). There is a transition from the gene-wide spreading of H3.3 to promoter-proximal enrichment in response to the loss of ARID1A. Concomitantly, X and Y-linked intergenic regions undergo a significant decrease in H3.3 occupancy upon ARID1A loss (Fig. S8A). A similar loss of H3.3 occupancy was also observed at a handful of autosomal intergenic peaks (Fig. S7A, Fig. S8B – top).

Unlike the sex-linked targets, TSSs of ARID1A regulated autosomal genes displayed H3.3 enrichment relative to the IgG control (Fig. S7C, Fig. S8B – bottom). However, this TSS occupancy appeared enhanced in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. S7C, Fig. S8B). Therefore, TSSs of both sex-linked and autosomal gene targets display concordant changes in H3.3 occupancy. Overall, ARID1A occupancy at TSSs (Fig.S4) of differentially regulated genes prevents the accumulation of H3.3 at target promoters.

Our analysis of the genome-wide localization of H3.3 in response to the loss of ARID1A would argue that the abnormal increase in promoter-associated H3.3 results from its redistribution from high-affinity sites (peaks) that are predominantly sex-linked and intergenic during normal pachynema (Fig. S7A, Fig. S8). We focused our attention on these H3.3-occupied intergenic sex-linked sites. Although the Macs2 algorithm identified only 224 sex-linked intergenic peaks, we speculated that this approach filters out regions enriched for H3.3 that fail to meet the peak calling threshold compounded by the pervasive spreading of H3.3 on the sex chromosomes (Fig. 5A, Fig. S7B, S8A). We adopted an alternative strategy that involved identifying potential H3.3 bound sites by monitoring their occupancy at ARID1A-governed accessible chromatin (Fig. 4A). We compared Arid1a^WT and Arid1a^cKO pachytene spermatocyte associated ATAC-seq peaks to identify mutually exclusive and overlapping peaks using BEDTools (Quinlan and Hall, 2010). The resulting peaks could be categorized into lost (n=9), common (n=7194), or gained (n=2246) based on their absence, persistence, or appearance, respectively, in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 6A). Next, we examined the ATAC-seq signal at these regions to find an increase in accessibility not only at gained peaks as expected but also at common peaks in Arid1a^cKO relative to Arid1a^WTpachytene spermatocytes.

Despite being labeled as lost, these regions did not display the expected loss in ATAC-seq signal in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 6A, left panel). The ATAC signal appears unchanged, highlighting the possibility that lost regions were labeled as such because they failed to satisfy Macs2 peak thresholds in Arid1a^cKOpachytene spermatocytes, especially given the comparatively higher magnitude of ATAC signal surrounding the common and gained peaks (Fig. 6A, left panel). We restricted our analysis to common and gained sites constituting 99 % of sex-linked regions subject to ARID1A-regulated chromatin accessibility. By plotting the average H3.3 coverage at sites associated with common and gained peaks, we detected striking differences in H3.3 occupancy. Briefly, both common and gained open chromatin regions consisted of loci differentiated by either a loss or robust increase in H3.3 occupancy in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 6A, left panel). To gain further insight, we performed k-means clustering, based on H3.3 occupancy, to identify 4 clusters associated with either common (C1-C4) or gained (G1-G4) peaks (Fig. 6B). The clusters associated with common ATAC-seq peaks displayed contrasting changes in H3.3 occupancy in response to the loss of ARID1A. C1 and C2 represented regions deficient for H3.3, while C3 and C4 represent regions that gained H3.3 binding in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes (Fig. 6B, left panel). Similarly, clusters associated with gained ATAC-seq peaks, namely, G1 and G3, represented sites that gained H3.3 occupancy, while G2 identified sites that lost H3.3 occupancy. We also noticed a gain in H3.3 binding at regions associated with G4 in response to the loss of ARID1A, albeit at significantly reduced coverage relative to G1 and G2 (Fig. 6B, right panel). Genomic annotations of common and gained k-means clusters revealed that the sites that displayed a loss of H3.3 binding (represented by C1 + C2, G2) in response to the loss of ARID1A were intergenic (Fig. S9A).

In contrast, those that gained H3.3 localization (represented by C3, C4, G1+G3, and G4) distributed differentially between intergenic and genic regions (Fig. S9A). Therefore, these data support our idea that ARID1A restricts H3.3 occupancy primarily to intergenic sites on the sex chromosomes. Furthermore, the loss of ARID1A triggers an abnormal redistribution of H3.3 to genic regions. Therefore, the increased promoter occupancy of H3.3 is an indirect consequence of ARID1A loss.

To gain further insight into the mechanisms regulating H3.3 associations on the sex chromosomes, we examined the regions associated with various k-means clusters to investigate the enrichment of relevant motifs. To our surprise, we detected an enrichment of the motif associated with the DSB hotspot specifier, PRDM9 (PR domain-containing protein 9), in clusters related to gained peaks, G1 and G3 (Fig . S9B, top). Additionally, we observed an increased frequency of PRDM9 motif centered at gained ATAC-seq peaks associated with G1 and G3 (Fig . S9B, middle).

Interestingly, regions associated with G1 and G3 usually are devoid of H3.3 (Fig. 6B), implying a potential antagonism between H3.3 occupancy and the occurrence of meiotic DNA DSBs. Gained ATAC-seq peaks associated with G2, which usually displays H3.3 occupancy, appeared devoid of PRDM9 motif occurrences relative to G1 and G3 (Fig. S9B, middle). Next, we determined the frequency of the PRDM9 motif at k-means clusters associated with common ATAC-seq peaks. These regions also displayed a dichotomous association with H3.3 (Fig. 6B). C3 and C4, usually deficient in H3.3, showed an increased frequency of PRDM9 motifs relative to C2, which typically displays H3.3 binding (Fig. S9B, bottom). These data confirm an antagonistic relationship between H3.3 and meiotic DNA DSBs.

Since the PRDM9 motifs identified from SSDS (Single-Stranded DNA Seq) data were generated to map the genomic associations of DNA repair factor DMC1(Brick et al., 2012), we monitored its enrichment at the various k-means clusters. We also monitored the association of markers of meiotic DNA DSBs, such as the PRDM9 catalyzed H3K4me3 at common and gained k-means clusters using published ChIP-seq data from pachytene spermatocytes (Maezawa et al., 2018a). Consistent with the motif analyses, DMC1 occupancy and H3K4me3 enrichment occurred at gained and common ATAC-seq peaks associated with k-means clusters, normally devoid of H3.3, namely, G1, G3, G4, C3, and C4 (Fig. S9C, D).

These data suggest that ARID1A might indirectly influence DNA DSB repair on the sex chromosomes by regulating the localization of H3.3. To test this conclusion, we monitored the association of DMC1 on the sex chromosomes, in response to the loss of ARID1A, by IF. Normally, DMC1 foci appear distributed along the non-homologous arms of the X and Y chromosomes from early to mid-pachynema, only disappearing by late pachynema (Fig. 7A, left, panel insets) (Moens et al., 2002). In contrast, in the absence of ARID1A, we observed a marked reduction in the number of DMC1 foci (Fig. 7A, right, panel insets). Quantification of DMC1 foci in Arid1a^cKO relative to Arid1a^WT pachytene spermatocytes revealed a significant decrease in the sex-linked association of DMC1 in the absence of ARID1A (Fig. 7B). Along with DMC1, the mitotic DNA recombinase, RAD51, is also known to localize to the non-homologous arms of the sex chromosomes (Moens et al., 1997; Moens et al., 2002). Therefore, we determined whether ARID1A also influenced the sex-linked association of RAD51. First, we analyzed previously generated RAD51-SSDS data (Hinch et al., 2020) to monitor its enrichment across the various k-means clusters associated with common and gained ATAC-seq peaks. Unlike DMC1, we found no detectable levels of RAD51 at any k-means clusters (Fig. S10A). Furthermore, the formation of RAD51 foci along the asynapsed axes of the sex chromosomes appeared unchanged in Arid1a^cKO relative to Arid1a^WTpachytene spermatocytes (Fig. S10B). Therefore, ARID1A is necessary for the normal sex-linked association of DMC1 but not RAD51. More importantly, our data highlight a dual role for ARID1A in regulating sex chromosome repression and DNA repair (Fig. 8).

Generation of the Arid1a^tm1Mag/Mmnc mutant allele was previously described (Chandler et al., 2015). The mice are available through the regional mutant mouse resource and research center (MMRRC; Stock number: 041418-UNC; www.mmrrc.org). Arid1a floxed mice (Arid1a^tm1Mag/Mmnc) exist on a CD1 genetic background. Crosses with testes-specific Stra8-Cre mice (expressed in P3 spermatogonia)(Sadate-Ngatchou et al., 2008) resulted in conditional knockouts. Arid1a^fl/fl; Stra8-Cre^Tg/0 females were crossed to either Arid1a^fl+or Arid1a^flfl males to obtain Arid1a^flfl;Stra8-Cre^Tg/0(Arid1a^cKO) and either Arid1a^flfl or Arid1a^fl+ (Arid1a^WT) males. Haploinsufficiency associated with Arid1a negated the possibility of transmitting the Cre through the paternal germline. Genotyping primers: Arid1a^fl+ alleles – (F)-5^’ – CTAGGTGGAAGGTAGCTGACTGA -3^’; (R) 5^’ – TACACGGAGTCAGGCTGAGC -3^’ (PCR product sizes-fl: 300 bp; +: 200 bp), and Stra8-Cre – (F) 5^’ -GTGCAAGCTGAACAACAGGA-3^’, (R) 5^’ -AGGGACACAGCATTGGAGTC-3^’ (PCR product size-Cre: 150 bp). Mice were housed with a 12hr light cycle (temperature – 20-24° C, humidity – 30-70%). All animal work followed approved UNC Chapel Hill IACUC protocols.

Arid1a^WT and Arid1a^cKO t2estes and cauda epididymites were fixed overnight in Bouins solution (Fisher scientific Ricca chemical; 11-201) at 4° C, followed by dehydration in ethanol series (50%, 70%, and 100%) before embedding in paraffin. The animal histopathology core prepared stained tissue sections (hematoxylin and eosin for cauda epididymites; Periodic acid-Schiff for testes). Staining and morphology determined seminiferous tubule staging (Ahmed and Rooij, 2009; Meistrich and Hess, 2013). Summary of staging is provided (Supplementary table 1).

Testes cryosections and spermatocyte spreads from juvenile (P18, P19, and P26) and adult Arid1a^WT and Arid1a^cKO mice were examined with indirect immunofluorescence (IF). 8-10 μm thick cryosections of testes were prepared using published methods (Menon et al., 2021). Whole testes were fixed in 10% neutral buffered formalin (NBF) at 4° C for 20 minutes. They were cut in half, followed by 40 minutes of incubation in 10% NBF. Fixed tissues were washed three times (10 minutes per wash) in room temperature (RT) phosphate-buffered saline (PBS) pH7.4. The tissues were then treated with a series of sucrose washes -10% (30 min), 20% (30 min), and 30% (1 hr). Tissues were then incubated at 4° C overnight in a 1:1 ratio of a solution of 30% sucrose/optimum cutting temperature (OCT, Tissue-Tek; 62550-01), followed by embedding in OCT, sectioning, and storage at -80⁰ C. Before immunostaining, cryosections were thawed on a heating block at 42⁰ C, rehydrated for 10 minutes in PBS, and incubated in 10mM citric acid buffer (pH6.0) for 10 minutes at 80-90⁰ C (antigen retrieval), during which addition of fresh citrate buffer (80-90 ⁰C) occurred every 2 minutes, and then allowed to cool for 20 minutes. For immunostaining, cryosections were incubated overnight in a humidified chamber with primary antibodies at 4⁰ C and then followed by a secondary antibody incubation at RT for 1 hr. Sera were diluted with antibody dilution buffer in PBS (ADB: 3% bovine serum albumin; 10% goat serum; 0.5% Triton-X 100). Two identical blocking steps preceded antibody incubations. These steps involved sequential incubations in (i) PBS, (ii) PBS/0.1% Triton-X 100, and (iii) blocking solution (1:10 dilution of ADB in PBS) for 10 minutes each at RT. Immunostained slides were washed twice in Kodak Photo-Flo 200 (PBS/0.32%), counterstained with DAPI, washed again in PBS/0.32% Photo-Flo 200, and then mounted in Prolong Gold antifade medium (P-36931; Life Technologies).

Spermatocyte spreads were prepared as previously described (Gray et al., 2020) with minor modifications. Briefly, seminiferous tubules were incubated in hypotonic buffer (30mM Tris pH8.2, 50mM Sucrose, 17 mM Trisodium dihydrate, 5mM EDTA, 0.5mM DTT, 0.1 mM PMSF) for 10-20 minutes on ice. Tubules were minced in 100 mM sucrose (200 µL/testis) until the solution turned turbid. The resulting cell suspension was isolated, avoiding tissue debris, and held on ice. Cells were fixed and spread by dropping 20 µL of cell suspension directly onto glass slides coated with paraformaldehyde solution (1 % Paraformaldehyde; 0.15 % Triton X-100; 625 nM Sodium Borate, pH 9.2). Uniform spreading was ensured by tilting the slides along the near and opposite edges. Next, slides were incubated in a humidified chamber for 1 hr, following which they quickly air-dried and then washed two times in PBS/0.32% Photo-Flo 200 and once in H₂O/0.32% Photo-Flo 200. Air-dried slides were immediately processed for immunostaining as described above for cryosections or stored at – 80⁰ C. Immunostained spreads were washed 2x with PBS/0.32% Photo-Flo 200, counterstained with DAPI, and washed again with Photo-Flo 200. Stained spreads were mounted in Prolong Gold antifade medium. The list of antibodies for IF is provided in supplementary table 2. The Leica Dmi8 fluorescent microscope was used to capture images. Z-stacks were deconvoluted using Leica Dmi8 image software (Huygens essentials version 20.04).

Fluorescence intensity measurements and object (foci) counting were performed with Fiji (Schindelin et al., 2012). RNAPII (pSer2) IF signal intensities were measured using the method described previously (Menon et al., 2021). Briefly, pixel intensities were recorded from regions of interest (ROI) which include the sex chromosomes (sex body) and an area lacking cells (background fluorescence). Corrected total RNAPII-pSer2 fluorescence (CTRF) was calculated using method previously described (https://theolb.readthedocs.io/en/latest/imaging/measuring-cell-fluorescence-using-imagej.html), where CTRF= Integrated density associated with sex body – (Area of sex body x Mean fluorescence of background).

For counting axial DMC1 foci, images were cropped to an area encompassing the sex chromosomes. Cropped images were converted to 8-bit grayscale, and the DMC1 fluorescent channel selected for further processing. An ROI was set, manual thresholding was performed to select DMC foci, and noise (speckles) removed using Fiji median filter tool with a radius set to 2.0-4.0. Finally, DMC1 was enumerated with the count particles tool. Statistical significance was calculated using an unpaired student’s t-test. Metadata associated with quantification of RNAPII (pSer2) fluorescence and DMC1 foci are provided (Supplementary table 3,4)

Histones from P19 Arid1a^WT and Arid1a^cKOspermatogenic cells were acid-extracted using published methods (Shechter et al., 2007). Seminiferous tubules were digested in 1 mL of Enriched Krebs-Ringer Bicarbonate (EKRB) Medium (EKRB salts – 1.2 mM KH₂PO₄, 1.3 mM CaCl₂, 119.4 mM NaCl, 4.8 mM KCl, 1.2 mM MgSO₄, 11.1 mM Dextrose; 25.2 mM NaHCO₃; 1X essential amino acids -Invitrogen, 11130051; 1X non-essential amino acids-Gibco,11140050) supplemented with 0.5 mg/mL collagenase (10 minutes, rotated at 32 °C). Digested tubules sedimented by gravity for 5 minutes at RT. The supernatant was removed, and tubules were resuspended in 1 mL EKRB medium and then settled for 5 minutes at RT. The sedimented tubules were transferred to 1 mL EKRB medium supplemented with 0.025% trypsin and 4 µg/mL DNase-I. The suspension was incubated for 10 minutes on a rotator at 32° C. Digestion was halted with 10% Fetal bovine serum (FBS), and single-cell suspensions were prepared by pipetting the slurry and then filtering it sequentially through 70 µm and 40 µm filters. Cells were pelleted at 600g for 5 minutes, washed once in PBS, and processed for histone extraction. Briefly, cell pellets were resuspended in 1 mL hypotonic lysis buffer (10mM Tris-Cl pH8.0, 1mM KCl, 1.5mM MgCl₂, 1mM DTT) and incubated on a rotator at 4° C for 30 minutes. Nuclei were pelleted at 10,000 g for 10 minutes at 4° C. Nuclear pellets were resuspended in 400 µL 0.2N HCl and incubated for 1 hr at 4 °C, followed by centrifugation at 14,000 g for 10 minutes at 4° C to isolate supernatants containing histones that were precipitated with Trichloroacetic acid (TCA), and then added to the nuclear suspension at a final concentration of 33% and incubated on ice for 30 minutes. Precipitated histones were pelleted at 14,000g for 10 minutes at 4° C and washed two times in 1 mL ice-cold acetone. Finally, washed histones were air-dried at RT for 20 minutes and resuspended in 100 µL ddH₂O. For western blots, histones were separated on a 15 % SDS-polyacrylamide gel and transferred to PVDF (Polyvinylidene Difluoride) membrane using a semi-dry transfer apparatus (Bio-Rad). Sample loading was assessed by REVERT^TM total protein stain (LI-COR, 926-11010). Blots were scanned on a LI-COR Odyssey CLx imager and viewed using Image Studio Version 5.2.5. Antibodies and their corresponding dilutions are listed (Supplementary table 2).

Pachytene spermatocytes were purified from Arid1a^WTand Arid1a^cKO adult testis by the unit gravity sedimentation procedure using a Sta-Put apparatus (Bellvé, 1993; O’Brien, 1993). Testis from at least 12 adult males (>P60) were dissected into EKRB, decapsulated, and digested in 0.5mg/ml collagenase for 15 min at 32° C. The resulting tubules were gently washed with EKRB and digested with 0.25mg/ml trypsin supplemented with 4µg/ml DNase for 15 min at 32° C. Following trypsin incubation, additional DNase (4µg/mL) was added to reduce viscosity further. Samples were triturated for 3 minutes with a plastic transfer pipet to fragment flagella, break intercellular bridges between germ cells and further disperse cells into suspension. Trypsin digestion was stopped by adding 0.25mg/mL trypsin inhibitor (Sigma/EPRO, T9003), and cell aggregates were removed by filtration using 40 µm cell strainers. Cells from the flow-through were pelleted at 500g for 5 min at 4° C, resuspended in EKRB + 0.5% BSA, and adjusted to a 3.33x10⁸ cells/mL density. 10⁹ cells from this single-cell suspension were loaded onto a 2-4% linear BSA gradient and allowed to sediment in the Sta-Put apparatus at 4° C. After 2hr and 40min of undisturbed sedimentation, 10ml fractions were collected at 45 sec per tube. Fractions were pelleted at 500g for 5 min at 4°C and resuspended in 1ml EKRB + 0.5%BSA. Fractions containing >80% pachytene spermatocytes were identified by light microscopy, pooled, and stored in cryoprotective media (DMEM supplemented with 10%FBS and 20%DMSO) at -80° C pending further analysis.

Total RNA was isolated in TRIzol^TM reagent (Invitrogen, 15596026) from Sta-Put purified fractions enriched for pachytene spermatocytes and round spermatids. The extracted RNA was purified using the Direct-zol RNA kit (Zymo, R2050). cDNA was synthesized using a random primer mix (New England Biolabs, S1330S) and ProtoScript^® II reverse transcriptase (New England Biolabs, M0368L). RT-PCR was performed on 1:10 cDNA dilutions using Sso Fast EvaGreen supermix (Bio-Rad, 172-5280) on a thermocycler (BioRad, C1000). Amplicons associated with either floxed or excised Arid1a cDNA’s were detected using primers: (F)-5^’ – TCCAGTAAGGGAGGGCAAGAAGAT -3^’; (R) 5^’ – GTAGTTGGCGTTGGGCAAGGCATTA -3^’ (PCR product sizes-fl: 612 bp; Δ: 281 bp)

RNA was extracted and purified in quadruplicate from frozen pellets of Arid1a^WT and Arid1a^cKO pachytene spermatocytes obtained bSalmony Sta-Put gravity sedimentation by methods described above. All samples displayed RNA integrity number (RIN) > 7.0 as determined by Agilent Tapestation (High sensitivity RNA screen tape). Sequencing libraries were prepared from 1 µg of total RNA using a Kapa mRNA HyperPrep kit (Roche, KK8580) per the manufacturer’s instruction. Libraries were quantified using Qubit^TM dsDNA Hs assay kit (Invitrogen, Q32854) pooled at equimolar concentrations, and sequenced on a single lane of the Illumina NovaSeq 6000 S Prime flow cell (50 bp reads, paired-end).

Quantification of gene expression was done using Salmon (Patro et al., 2017). Transcript counts at the gene level were summarized using tximport (Soneson et al., 2016) and then imported to perform a differential gene expression analysis using DESeq2 (Love et al., 2014). The mouse (mm10) gene/transcript annotations were retrieved using AnotationDbi (Hervé Pagès et al., 2021) and TxDb.Mus musculus.UCSC.mm10.known gene (Bioconductor Core Team, 2019) R packages from Bioconductor. Low-count genes (< 10 reads) were pre-filtered and significant differences in counts were called at a false discovery rate (FDR)≤ 0.05. Supplementary Table 5 lists differentially expressed genes.

CUT&RUN assays were performed on cryopreserved Arid1a^WTand Arid1a^cKO pachytene spermatocytes (500,000 cells/genotype) obtained by Sta-Put, using a slightly modified version of previously described methods (Meers et al., 2019; Skene and Henikoff, 2017). For ARID1A, CUT&RUN was performed in triplicates per genotype with two different antibodies. H3.3 CUT&RUN was performed in duplicates on Arid1a^WT and triplicates on Arid1a^cKO samples. Aliquots of cryopreserved Arid1a^WT and Arid1a^cKO pachytene spermatocytes were quickly thawed in a water bath at room temperature (RT), pelleted at 600g for 5 minutes, and washed [wash buffer: 20mM HEPES(Na) pH7.5, 150 mM NaCl, 0.5 mM spermidine, EDTA-free protease inhibitor] three times. Cells were gently resuspended in wash buffer and bound to Concanavalin-A coated magnetic beads (20 µL beads/500,000 cells) by mixing on a rotator for 15 minutes. The bead slurry was separated using a magnet followed by permeabilization in 50 µL antibody buffer (wash buffer + 0.05% Digitonin + 2mM EDTA) per sample in 0.2 mL 8-strip PCR tubes. Primary antibody was added to the samples and incubated on a nutator at 4° C overnight. Cells were then separated using a magnet, washed two times in 200 µL Dig-wash buffer (wash buffer + 0.05% Digitonin), and incubated with a secondary antibody prepared in 50 µL of Dig-wash buffer/sample for 30 minutes on a nutator at 4° C. After antibody incubations, samples were washed twice in 200 µL Dig-wash buffer and then incubated on a nutator at 4° C with 50 µL Dig-wash buffer containing 1000ng/mL Protein A/G Micrococcal Nuclease (pA/G-MNase) for 1 hr. Following this, samples were washed twice in 200 µL Dig-wash buffer, resuspended in 50 µL Dig-wash buffer, cooled down to 0° C, and supplemented with CaCl₂ at a final concentration of 2 mM to activate MNase. Samples were digested at 0 °C for 40 minutes. The resulting chromatin fragments were extracted from the bead slurry by incubating in 2X STOP buffer (340 mM NaCl, 20mM EDTA, 4mM EGTA, 0.05% Digitonin, 100 µg/mL RNase A) at 37° C for 30 minutes. DNA was purified using ChIP DNA clean and concentrator kit (Zymo, D5205). Libraries were prepared using the Kapa Hyperprep kit (Roche, KK8504), examined for size, and quantified on an Agilent 2100 bioanalyzer using the high-sensitivity DNA kit (Agilent, 5067-4626). The libraries were pooled in equimolar amounts and sequenced on a single lane of the Illumina NovaSeq 6000 S Prime flow cell (50 bp reads, paired-end). Antibody details are listed in Supplementary table 2. The pA/G-MNase (Addgene ID: 123461) used for the CUT&RUN experiments was purified at the Protein Expression and Purification core at UNC Chapel Hill.

CUT&RUN data analyses were performed as previously described (Menon et al., 2021). Briefly, fastq files were analyzed with fastqc (version 0.11.9), then processed with TrimGalore (version 0.6.2, https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) keeping the --trim-n option. Trimmed reads were aligned to mm10 (mouse) reference genome using bowtie2 (Langmead and Salzberg, 2012) parameters: bowtie2 --very-sensitive-local --no-mixed --no-unal --dovetail --no-discordant. Alignments were outputted in BAM format using samtools (Li et al., 2009) and filtered for PCR duplicates using Picard tools, MarkDuplicates (https://broadinstitute.github.io/picard/). Deduplicated BAM files were used to generate bigWig files filtered for ENCODE mm10 blocked regions (Amemiya et al., 2019) with DeepTools, bamCoverage (Ramírez et al., 2016) with the following options: -bs 30 --smoothLength 60 -bl -- scaleFactor. The option --scaleFactor was used to normalize for composition bias between CUT&RUN libraries generated from Arid1a^WT and Arid1a^cKO pachytene spermatocytes. csaw (Lun and Smyth, 2015) derived normalization factors (nf) were used to calculate the effective library size (library size X nf). The inverse of the effective library size per million was supplied to --scaleFactor to generate normalized bigWig files. Peak calling was performed with MACS2 (version 2.1.2) with options: -f BAMPE --broad --broad-cutoff 0.05 --keep-dup all, on filtered BAM files associated with ARID1A and H3.3 relative to an IgG control previously generated from pachytene enriched populations obtained from P18 testes (Chakraborty and Magnuson, 2022). Peak overlaps across replicates were determined using bedtools intersect (Quinlan and Hall, 2010). Peak calls are provided in supplementary table 6. Peaks were annotated using HOMER (version 4.9.1), annotatePeaks.pl. Replicate bigWig files were averaged using WiggleTools (Zerbino et al., 2014). Averaged bigWig files generated heatmaps with DeepTools, computeMatrix, and plotHeatmap. Data was visualized on Integrative Genomics Viewer (Robinson et al., 2011).

Cryopreserved aliquots of Arid1a^WT and Arid1a^cKOpachytene spermatocytes were processed in quadruplicate for ATAC-seq using the standard Omni-ATAC protocol (Corces et al., 2017) with minor changes. These involved increasing the cell number to 100,000 per replicate, using Diagenode tagmentase (loaded Tn5 transposase, C01070012) and Diagenode 2X tagmentation buffer (C01019043). Libraries were quantified using NEBNext® kit for Illumina® (New England Biolabs, E7630S), pooled at equimolar amounts, and sequenced on a single lane of the Illumina NovaSeq 6000 S Prime flow cell (50 bp reads, paired-end).

After analyzing the quality of fastq files with fastqc (version 0.11.9), they were processed with TrimGalore (version 0.6.2), keeping the --trim-n option. Alignments to mm10 reference genome were performed with bowtie2 (Langmead and Salzberg, 2012) parameters: bowtie2 --very-sensitive -X 2000, and outputted in BAM format using samtools. Resultant BAM files were filtered for PCR duplicates using Picard tools, MarkDuplicates, and mitochondrial reads using samtools view (code: samtools view -h Dedup.bam | grep -v ’chrM’ | samtools view -b -h -f 0x2 – | samtools sort > Dedup_noChrM.bam). Filtered BAM files were used to call peaks with MACS2 (parameters: -f BAMPE -n de -q 0.05 --nomodel --nolambda --keep-dup all) and generate normalized bigWig files with DeepTools, bamCoverage (options: -bs 30 --smoothLength 60 -bl -- normalized using RPKM). Peak overlaps across replicates were determined using bedtools intersect (Quinlan and Hall, 2010). Peak calls are provided in supplementary table 7. Peak annotation (annotatePeaks.pl) and motif analyses (findMotifsGenome.pl) were performed with HOMER (version 4.9.1). PRDM9 motif frequency near ATAC-seq peaks was determined using HOMER annotatePeaks.pl with options: -size 20000 -hist 10. Replicate bigWig files were averaged using WiggleTools (Zerbino et al., 2014). Averaged bigWig files generated heatmaps with DeepTools, computeMatrix, and plotHeatmap. Data was visualized on Integrative Genomics Viewer (Robinson et al., 2011).