ARID1A maintains transcriptionally repressive H3.3 associated with CHD4-ZMYND8 chromatin interactions

ARID1A is a signature subunit of the mammalian SWI/SNF (BAF) chromatin remodeling complex and is mutated at a high rate in malignancies and benign diseases originating from the uterine endometrium. Through genome-wide analysis of human endometriotic epithelial cells, we show that more than half of ARID1A binding sites are marked by the variant histone H3.3, including active regulatory elements. ARID1A loss leads to H3.3 depletion at ARID1A bound active regulatory elements and a concomitant redistribution of H3.3 towards genic elements. ARID1A interactions with the repressive chromatin remodeler CHD4 (NuRD) are associated with H3.3-containing chromatin regulation. ZMYND8, the CHD4-interacting acetyl-histone H4 reader, specifies ARID1A-CHD4-H3.3 target regulatory activity towards histone H4 lysine 16 acetylation (H4K16ac) to repress super-enhancers. ARID1A, H3.3, CHD4, and ZMYND8 co-repress the expression of genes governing extracellular matrix, motility, adhesion, and epithelial-to-mesenchymal transition. Moreover, these gene expression alterations are observed in human endometriomas. Altogether, these studies demonstrate that cooperation among a histone reader and different types of chromatin remodelers safeguards the endometrium through transcriptionally repressive H3.3.


Introduction 32
The SWI/SNF complex remodels chromatin through ATP-dependent DNA sliding, 33 H2A/H2B dimer eviction, and nucleosome ejection functions (Dechassa et al., 2010;Kassabov, 34 Zhang, Persinger, & Bartholomew, 2003;Whitehouse et al., 1999). SWI/SNF remodeling 35 activities open chromatin and promote accessibility for other DNA-binding factors and chromatin 36 regulators (Clapier, 2021;Clapier & Cairns, 2009). SWI/SNF complex composition is 37 heterogeneous and cell type dependent (Wang et al., 1996). SWI/SNF regulates lineage-specific 38 enhancer activity through multiple mechanisms (Alver et al., 2017). Protein subunit architecture 39 contributes to SWI/SNF complex specificity through specialized cofactor interactions. The in the promoter region were more likely to show differential expression (DE) following ARID1A 178 loss than genes without promoter H3.3 ( Figure 2I, bottom). In addition, locus-scale investigation 179 clearly showed that ARID1A and H3.3 often co-mark active chromatin regulatory elements, which 180 infrequently also includes gene body coating by H3.3, such as at COL1A1, THBS1, and SERPINE1 181 ( Figure 2J). These data collectively suggest H3.3 may be linked to transcriptional regulatory 182 activity by ARID1A at the level of chromatin. 183  regions. Regions are colored based on shARID1A differential H3.3 significance. FDR   We next sought to determine the transcriptional consequences of H3.3 loss in endometrial 295 epithelia. We hypothesized that H3F3B could be knocked down to reduce H3.3 levels for acute 296 transcriptome evaluation without impeding cell health ( Figure 4A). Using siRNA targeting H3F3B 297 (siH3F3B), we observed H3.3 depletion by immunoblotting without affecting the cell cycle 298 Comparing the gene expression changes following H3.3 loss with those following 309 ARID1A loss, we observed significant overlap, with 682 shared dysregulated genes ( Figure 4F). 310 These 682 genes were then grouped by direction of change (upregulated vs. downregulated) to 311 identify genes with the same or different expression patterns following ARID1A vs. H3.3 loss. A 312 significant association was observed between the effects of H3.3 and ARID1A loss indicating 313 shared transcriptional consequences ( Figure 4G). Gene expression changes also positively 314 correlated transcriptome-wide ( Figure 4H). Intriguingly, the 682 genes affected by loss of H3.3 315 and ARID1A were more likely to be repressed by H3.3 ( Figure 4I). 196 genes were identified as 20 mutually repressed by both ARID1A and H3.3, including PLAU, ADAMTS15, C1S, CD82, CCL2, 317 and CLSTN2 ( Figure 4J). In agreement with differential H3.3 patterns, these 196 co-repressed 318 genes were enriched for similar gene sets as observed among the shARID1A direct decreasing 319 promoter H3.3 gene set, including EMT, TNFα signaling, estrogen response, apoptosis, adhesion, 320 migration, extracellular matrix, and collagens ( Figure 4K). Altogether, these data suggest that 321   Color image is presented for molecular weight marker reference. also been shown to suppress super-enhancer hyperactivation (Shen et al., 2016), and, more 379 recently, ZMYND8 and ARID1A were identified in the same screen as key chromatin regulators 380 of EMT (Serresi et al., 2021). Therefore, we investigated the potential roles of ZMYND8 and 381 possible co-factors as mediators of the observed ARID1A-H3.3 co-regulation ( Figure 5B).

25
ARID1A co-immunoprecipitation (co-IP) using an anti-ARID1A antibody was first used 383 to detect physical nuclear interactions with ZMYND8. We previously confirmed the specificity of 384 the anti-ARID1A antibody by co-immunoprecipitation followed by mass spectrometry (Wilson et 385 al., 2019). While ZMYND8 was not detected in the ARID1A pulldown following high salt washes 386 (300 mM KCl), the NuRD catalytic subunit CHD4 was evident ( Figure 5C). We then hypothesized 387 that CHD4-NuRD may serve as an interface between ARID1A and ZMYND8. A reciprocal CHD4 388 co-IP confirmed nuclear interactions with both ARID1A and ZMYND8 ( Figure 5D). To further 389 support that ARID1A, ZMYND8, and CHD4 are found in high molecular weight nuclear 390 complexes of similar size, glycerol gradient sedimentation was performed. Native fractions were 391 observed that included ZMYND8 and members of both SWI/SNF and NuRD ( Figure 5E). These 392 data suggest that physical interactions between ARID1A and CHD4 may regulate H3.3 chromatin 393 with support from ZMYND8. 394 We then examined genome-wide chromatin regulation by CHD4 and ZMYND8 in relation 395 to ARID1A and H3.3. Genome-wide binding profiles of CHD4 and ZMYND8 were measured by 396 ChIP-seq. Roughly 2000 genomic regions were identified with H3.3 and all three chromatin 397 regulators co-localized ( Figure 5F-G). Across all H3.3 peaks genome-wide, ARID1A binding was 398 most strongly enriched at ZMYND8-CHD4 co-bound sites compared to sites occupied by either 399 CHD4 or ZMYND8 alone ( Figure 5H). Notably, when CHD4 was absent, ARID1A binding at 400 H3.3 sites did not correlate with the presence of ZMYND8 ( Figure 5H). These data suggest that 401 CHD4 may be primarily responsible for ARID1A recruitment to H3.3 chromatin. We further 402 investigated ARID1A binding and H3.3 abundance across H3.3 peaks segregated by the presence 403 of CHD4/ZMYND8. ARID1A binding was again strongest at H3.3 peaks co-bound by CHD4 as 404 opposed to those without CHD4 ( Figure 5I). H3.3 abundance was similarly highest at CHD4-26 bound peaks, although CHD4+ZMYND8 peaks showed the overall highest H3.3 levels ( Figure  406 5I). With respect to H3.3 regions dependent on ARID1A chromatin interactions, we observed that 407 baseline H3.3 levels were significantly higher at regions that decreased in H3.3 following ARID1A 408 knockdown if they were co-occupied by CHD4 or ZMYND8, but this was not observed at regions 409 that gained H3.3 following ARID1A loss ( Figure 5J). Intriguingly, we observed that genome-wide 410 H3.3 regions directly maintained by ARID1A chromatin interactions are associated with CHD4 411 but not ZMYND8 ( Figure 5K). Moreover, genome-wide regions that lose H3.3 following ARID1A 412 loss due to disrupted ARID1A chromatin interactions tend to have higher baseline levels of 413 ARID1A, CHD4, and H3.3, but lower levels of ZMYND8 in comparison to stable H3.3 regions 414 ( Figure 5L). Altogether, these results suggest CHD4 and ZMYND8 are associated with ARID1A-415 H3.3 co-regulation. 416

H3.3 may be chromatin context-specific and occur at subsets of regulatory regions bound by 452
ARID1A-CHD4-ZMYND8. While ARID1A loss causes widespread H3.3 reduction in chromatin, 453 we observed that ARID1A-H3.3 co-regulation is most frequent at active enhancer and super-454 enhancer chromatin states. As such, we next examined ARID1A-CHD4-ZMYND8 co-regulation 455 of H3.3 at enhancers. At ARID1A-bound active enhancers, defined as accessible (ATAC+) 456 H3K27ac peaks located >3 kb from an annotated TSS ( Figure 6A), ZMYND8 binding is associated 457 with presence of CHD4, as expected ( Figure 6B). Moreover, ZMYND8 binding was detected at 458 84.6% of ARID1A+CHD4 super-enhancers (n = 507) as opposed to 63.2% of ARID1A+CHD4 459 typical enhancers (n = 2282) ( Figure 6B). Importantly, ARID1A, CHD4, and ZMYND8 co-460 binding at active enhancers is associated with presence of H3.3, and this association is greater at 461 super-enhancers ( Figure 6C). We previously observed that ARID1A suppresses H3K27-462 hyperacetylation at a subset of active super-enhancers (Wilson et al., 2020). ARID1A and CHD4 463 binding levels are not substantially different at suppressed super-enhancers that become 464 hyperacetylated following ARID1A loss vs. those with stable acetylation ( Figure 6D). Strikingly, 465 ZMYND8 binding and H3.3 abundance are significantly higher at suppressed super-enhancers that 466 become hyperacetylated ( Figure 6D). These data indicate that ZMYND8 is associated with CHD4 467 and ARID1A most frequently at active H3.3-marked super-enhancers that are suppressed by 468 ARID1A. 469 Our data indicate the ZMYND8 module appears to be associated with repressive chromatin 487 targeting by ARID1A at H3.3-marked super-enhancers. Histone tail reader functions of ZMYND8 488 are a plausible mechanism through its BRD, PWWP, and PHD domains, which interact with 489 acetylated H3/H4 residues and methylated H3 residues (Savitsky et al., 2016). Particularly, the 490 ZMYND8 bromodomain was recently described to interact with acetylated H4 tails and recruit 491    Figure 7A). We next determined states enriched for co-binding of 509 ARID1A-CHD4-ZMYND8 and shARID1A direct decreasing H3.3 ( Figure 7B). Both upstream 510 active promoter super-enhancer states (states 2 and 5) showed the strongest enrichment for 511 shARID1A direct decreasing H3.3, while ZMYND8 binding and ARID1A-CHD4-ZMYND8 co-512 binding was most enriched at the H4(K16)ac+ upstream active promoter super-enhancer class 513 (state 2) ( Figure 7B). 514 We further investigated chromatin state identities at reference annotated gene promoter 515 regions (±3 kilobases flanking TSS). As expected, the highly active TSS state (state S1) is the most 516 prevalent promoter chromatin state identity at genes transcriptionally regulated by ARID1A 517 showed stronger enrichment for ARID1A transcriptional repression than those without H4(K16)ac 525 (state 5) ( Figure 7C, right). In agreement, we also observed that chromatin accessibility directly 526 repressed by ARID1A is associated with presence of H4 acetylation (Figure 7-figure supplement  527   4). Promoter super-enhancer H3K27ac peaks were next directly segregated by detection of 528 H4K16ac (H4K16ac+, n = 112; H4K16ac-, n = 115). ARID1A binding and H3.3 abundance were 529 not significantly different between H4K16ac stratified super-enhancers, but ZMYND8 binding 530 was stronger at the H4K16ac+ regions, while CHD4 binding was lower at H4K16ac+ regions 35 ( Figure 7D). Further, a correlation of measured chromatin features across all 227 promoter-532 proximal super-enhancer constituent enhancers supported that ZMYND8 binding is associated 533 with acetylated H4 marks (Figure 7-figure supplement 5). These analyses collectively suggest 534 that H4(K16)ac marked promoter super-enhancers may recruit repressive ARID1A-CHD4-535 ZMYND8 complexes to regulate H3.3. 536 To identify genes targeted by the ARID1A-CHD4-ZMYND8 regulation of repressive 537 H3.3, we also used siRNA to deplete CHD4 (siCHD4) and ZMYND8 (siZMYND8) followed by   Figure 7G). These mechanistic co-repressed genes were enriched for EMT, 551 adhesion, development, locomotion, collagens, and extracellular matrix gene sets ( Figure 7H). 552 Further, 68% of these genes were marked by gene body H4K16ac, an enrichment compared to less 553 than half of all expressed genes (Figure 7-figure supplement 7). Two physiologically relevant 36 target genes revealed through integrative epigenomic analysis are PLAU and TRIO, both of which 555 are located within broad H4K16ac+ domains and near active H3.3+ super-enhancers co-bound by 556 ARID1A, CHD4, and ZMYND8 ( Figure 7I). ARID1A loss leads to decreased promoter H3.3 557 abundance and transcriptional hyperactivation of PLAU and TRIO ( Figure 7I). We also observed 558 that co-knockdown of ARID1A and CHD4 led to increased induction of PLAU compared to either 559 knockdown separately (Figure 7-figure supplement 8).       lead to destabilization of repressive NuRD. One hypothesized cooperative mechanism that could 800 explain how ARID1A and CHD4 maintain H3.3 in active chromatin is through different aspects 801 of remodeler activity conferred by each complex. ARID1A could remodel or remove canonical 802 H3 nucleosomes-a typical role of SWI/SNF-while CHD4 could function to assemble or 803 stabilize H3.3. In this scenario, failure to remove canonical H3 following ARID1A loss may impair 804 the ability of CHD4 to assemble or stabilize H3.3, and thus loss of either ARID1A or CHD4 may 805 break this cycle. The related CHD1 remodeler is known to be required for H3.3 deposition into 806 chromatin in vivo (Konev et al., 2007), further suggesting a necessary role for remodeler activity 807 in H3.3 nucleosome assembly. It is important to note that H3.3 has a defined chaperone associated 808 with transcriptional activity, HIRA (Shi et al., 2017). P400 is another SWI/SNF-like remodeler 58 recently shown to exchange H3.3 nucleosomes that could also possibly collaborate with SWI/SNF 810 (Pradhan et al., 2016). 811 In silico analyses from the ReMap 2020 transcriptional regulator peak database (Cheneby 812 et al., 2020) (Adhikary et al., 2016), and TIP60-mediated H4 820 acetylation can functionally recruit ZMYND8 through this mechanism to repress transcription with 821 CHD4 in response to DNA damage (Gong et al., 2015). H4K16ac was also previously reported to 822 be required for the NoRC repressor complex to bind and silence chromatin (Zhou & Grummt, 823 2005). Our data also indicate that ARID1A directly suppresses chromatin accessibility at sites 824 marked by H4 acetylation, suggesting that SWI/SNF chromatin remodeler activity may be 825 may orchestrate this regulatory activity. ZMYND8 has been reported to be a super-enhancer factor 59 that suppresses hyperactivation (Shen et al., 2016). Corroborating our results, ZMYND8 was 833 previously shown to associate with NuRD at super-enhancers (Spruijt et al., 2016). We found that 834 super-enhancers that become hyperacetylated following ARID1A loss are normally associated 835 with the highest levels of H3.3 and ZMYND8 binding. In our proposed model, ZMYND8 836 bromodomain interactions with H4 acetylated tails recruit CHD4 and ARID1A for transcriptional 837 repression at active chromatin. This most notably occurs at H3.3+ super-enhancers, where all three 838 factors co-localize most frequently. Further work will seek to explain how ZMYND8 specifies this 839 repressive activity, as CHD4 and ARID1A can also function toward transcriptional activation.

Data availability 855
All new data generated in this study have been deposited to the Gene Expression Omnibus (GEO) 856 at SuperSeries accession GSE190557, and access is available with reviewer token: 857 ijmtkuuivtwzbwz. Previously generated and re-analyzed data were retrieved from GEO at 858 accessions GSE121198 and GSE148474 and analyzed as previously described ( were centrifuged and resuspended in TEB buffer (PBS supplemented with 0.5% Triton X-100, 5 891 mM sodium butyrate, 2 mM phenylmethylsulfonyl fluoride, 1X protease inhibitor cocktail) and 892 incubated on a 3D spindle nutator at 4 °C for 10 minutes. Cells were centrifuged at 3000 RPM for 893 10 minutes at 4 °C. TEB wash step was repeated once. Following second wash, pellet was 894 resuspended in 0.2 N HCl, and incubated on 3D spindle nutator at 4 °C overnight. The following 895 day, samples were neutralized with 1:10 volume 1M Tris-HCl pH 8.3. Sample was centrifuged at 896 3000 RPM for 10 minutes at 4 °C, and supernatant containing histone proteins was collected. 897 898

Co-immunoprecipitation (co-IP). 900
Nuclear extracts were prepared as previously described (Chandler et  Nuclear extracts were prepared, dialyzed, and quantified as described in the co-IP methods section. 916 Density sedimentation by glycerol gradient was performed and probed similar to published reports 917 (Mashtalir et al., 2018). Briefly, 4.5 mL 10-30% linear glycerol gradients were prepared using an 918 ÄKTA start (Cytiva) from density sedimentation buffer (25 mM HEPES, 0.1 mM EDTA, 12.5 919 mM MgCl2, 100 mM KCl, 1 mM DTT) additionally containing 30% and 10% glycerol for initial 920 and target concentrations, respectively. 200 µg nuclear lyase was overlaid on the glycerol gradient 921 followed by ultracentrifugation at 40,000 rpm in an AH-650 swinging bucket rotor (ThermoFisher 63 Scientific) for 16 hours at 4 °C. 225 µL gradient fractions were collected and concentrated using 923 StrataClean resin (Agilent). Concentrated fractions were eluted in 1.5X Laemmli buffer + 37.5 924 mM DTT and run on SDS-PAGE for immunoblotting. 925 926

Immunoblotting. 927
Whole-cell protein lysates were prepared as previously described (Wilson et al., 2020). Proteins 928 were quantified with the BCA Protein Assay Kit (Pierce, ThermoFisher Scientific). Protein 929 samples in Laemmli buffer + DTT were denatured at 94 °C for 3 minutes prior to running on SDS-930 PAGE gels (6% gels for co-IP and glycerol gradients, 15% gels for histone extracts, and 4-15% 931 gradient gels for whole-cell protein lysates). Gels containing histone extracts were wet transferred 932 to nitrocellulose membranes at 4 °C for 3 hours at 400 mA current, then dried at room temperature 933 followed by re-hydration in TBS + 0.1% Tween-20 (TBS-T) and blocking with Odyssey blocking 934 buffer (LI-COR). All other gels were semi-dry transferred to PVDF using a Trans-Blot Turbo (Bio-935 Rad) according to the manufacturer's protocol designed for high-molecular weight proteins, and 936 blocked with either 5% BSA or 5% milk in TBS. The following primary antibodies were used: