The histone variant macroH2A1.1 regulates RNA Polymerase II paused genes within defined chromatin interaction landscapes

The histone variant macroH2A1.1 (mH2A1.1) plays a role in cancer development and metastasis-related processes. To determine the underlying molecular mechanisms, we mapped genome-wide localization of endogenous mH2A1.1 in the human breast cancer cell MDA-MB 231. We demonstrate that mH2A1.1 specifically binds to active promoters and enhancers in addition to facultative heterochromatin. Selective knock-down of mH2A1.1 deregulates expression of hundreds of highly active genes. Depending on the chromatin landscape, mH2A1.1 acts through two distinct molecular mechanisms. The first is to limit excessive transcription in a predefined environment and relies on domain recruitment of mH2A1.1 at the promoter and gene body. The second mechanism is specific to RNA Pol II (Pol II) paused genes. It requires recruitment of mH2A1.1 restricted to the TSS of these genes. Moreover, we show that these processes occur in a predefined local 3D genome organization and are largely independent of enhancer-promoter looping. Among the genes activated by mH2A1.1, genes regulating mammary tumor cell migration are mostly dependent on Pol II release for their expression level, unlike other categories of mH2A1.1-regulated genes. We thus identified an intriguing new mode of transcriptional regulation by mH2A1.1 and propose that mH2A1.1 serves as a transcriptional modulator with a potential role in assisting the conversion of promoter-locked RNA polymerase II into a productive and elongated Pol II.

mapped genome-wide localization of endogenous mH2A1.1 in the human breast cancer cell 23 MDA-MB 231. We demonstrate that mH2A1.1 specifically binds to active promoters and 24 enhancers in addition to facultative heterochromatin. Selective knock-down of mH2A1.1 25 deregulates expression of hundreds of highly active genes. Depending on the chromatin 26 landscape, mH2A1.1 acts through two distinct molecular mechanisms. The first is to limit 27 excessive transcription in a predefined environment and relies on domain recruitment of 28 mH2A1.1 at the promoter and gene body. The second mechanism is specific to RNA Pol II (Pol 29 II) paused genes. It requires recruitment of mH2A1.1 restricted to the TSS of these genes. 30 Moreover, we show that these processes occur in a predefined local 3D genome organization 31 and are largely independent of enhancer-promoter looping. Among the genes activated by 32 mH2A1.1, genes regulating mammary tumor cell migration are mostly dependent on Pol II 33 release for their expression level, unlike other categories of mH2A1.1-regulated genes. We 34 thus identified an intriguing new mode of transcriptional regulation by mH2A1.1 and propose 35 that mH2A1.1 serves as a transcriptional modulator with a potential role in assisting the 36 conversion of promoter-locked RNA polymerase II into a productive and elongated Pol II. 37 38 Introduction 39 40 Histone post-translational modifications, DNA-binding factors and architectural 41 proteins regulate genome organization and dynamics (Luger et al., 2012;Venkatesh & 42 Workman, 2015). In addition, histone variants replace canonical histones in a locus-specific 43 manner, which endows chromatin with properties required to fine-tune DNA accessibility and 44 functions (Buschbeck & Hake, 2017).

46
Among the histone variants, macroH2A1 (mH2A1), a vertebrate-specific (Pehrson & 47 Fuji, 1998; Rivera-Casas et al., 2016) histone H2A variant, is composed of an N-terminal "H2A-48 like" domain (64 % identical to H2A) and a C-terminal 25 kDa "macro" domain. These two 49 domains are joined by an unstructured 41 amino acid long "linker" domain that positions the 50 macro domain outside of the nucleosome . Expression of the highly 51 conserved H2AFY gene produces two splicing isoforms, mH2A1.1 and mH2A1.2, whose 52 sequences differ in a 30 amino-acid region within the macro domain . 53 54 mH2A1 was originally found to be enriched on the transcriptionally silent X 55 chromosome (Costanzi & Pehrson, 1998). mH2A1 is also present at autosomes, forming large 56 domains in association with histone marks associated with heterochromatin, such as 57 H3K27me3 and H3K9me3 (Douet et  cells could be dependent on cellular context and remains to be clarified. 89 In this work, we identified and characterized the role of mH2A1.1 in the regulation of 90 gene expression in TNBC cells. We found that mH2A1.1 modulates the expression of hundreds 91 of highly expressed genes, while mH2A1.1 deficiency does not affect the expression of silent 92 or low expressed genes. Many of these mH2A1.1-regulated genes are involved in cytoskeletal 93 organization, certainly giving mH2A1.1 its role in controlling the migratory properties of these 94 tumor cells. This transcriptional function of mH2A1.1 is however bifunctional, with inhibitory 95 or stimulatory of target gene. Although not requiring ad-hoc rewiring of promoter-enhancer 96 contacts, this functional dichotomy clearly depends on the chromatin landscape in which 97 these genes are located and relies on differential recruitment of mH2A1.1. The activating 98 effect of mH2A1.1 requires tight recruitment of mH2A1.1 to the TSS of related genes. 99 Conversely, genes inhibited by mH2A1.1 recruit this histone variant over larger domains, 100 present further upstream and downstream of the TSS. Mechanistically, we determined that 101 the expression level of mH2A1.1-activated genes is dependent on the Pol II pause process. Among the 945 genes whose expression was 115 significantly modified in the mH2A1.1 KD cells, 533 genes (56.3%) were down-regulated 116 (called mH2A1.1-activated genes or AG) and 412 genes (43.7%) were up-regulated (called 117 mH2A1.1-repressed genes or RG) (Fig 1A, S1 Table). In general, gene expression changes 118 induced by mH2A1.1 depletion are moderate (Fig 1A). Altered gene expression was confirmed 119 by RT-qPCR on a subset of genes using two different siRNAs directed against mH2A1.1 (S1F-H 120 Fig). All mH2A1.1-regulated genes, both RG and AG, were found among the moderately to 121 highly expressed genes in WT MBA-MB231 cells (Fig 1B, C). Silenced genes in MDA-MB231 122 cells were not activated upon mH2A1.1 depletion (Fig 1B, C). We concluded that mH2A1.1 123 participates in fine-tuning actively transcribed genes expression. We next asked if the role of 124 mH2A1.1 variant in controlling expression of active genes depends on its association with 125 certain genomic regions including gene regulatory regions and specific epigenetic contexts. 126 mH2A1.1 associates with gene regulatory regions. We developed a ChIP-grade polyclonal 127 antibody that exclusively recognizes mH2A1.1 (Ab αmH2A1.1) (S2 Table and  generated ChIP-seq data of mH2A1.1 (S3 Table). The obtained dataset was compared to the 129 one obtained using a commercially available ChIP-grade antibody (Ab37264, Ab αmH2A1) 130 directed against total mH2A1 (S2, S3 Tables and S2E-G Fig). The two datasets were highly 131 similar with a Pearson coefficient correlation (PCC) of 0.92 (Fig 1D). Therefore, we decided to 132 conserved common peaks between the two ChIP-seq data for further analysis (Materials and 133 Methods). We identified 11.849 mH2A1.1 peaks, covering ≈ 7 % of the genome. Analysis of 134 the genomic distribution of mH2A1.1 showed that the vast majority of mH2A1.1 peaks 135 correspond to annotated promoters (TSS +/-1 kb), while 22% of mH2A1.1 peaks were 136 associated with distal intergenic regions (Fig 1E). We confirmed the enrichment of mH2A1.1 137 in a subset of regions corresponding to ChIP-seq peaks by ChIP-qPCR in WT cells, as well as its 138 decrease in mH2A1.1 KD cells (two mH2A1.1-specific RNAi), using either mH2A1.1 or mH2A1-139 specific antibodies (S3 Fig). 140 mH2A1.1 binds promoters of its target genes. We next asked if mH2A1.1 binding occurred in 141 specific chromatin environments. Genome-wide, we analyzed the correlation between 142 mH2A1.1 and a subset of heterochromatin marks (H3K9me3, H3K27me3, H2AK119ub), 143 chromatin-bound components (Pol II, BRD4, RING1B, PARP1, PCGF2) as well as euchromatin 144 marks (H3K4me1, H3K4me3, H3K36me3, H3K27ac, H3.3) (Fig 1D). mH2A1.1 positively 145 correlated with its well documented partner PARP1 in the MDA-MB231 cell line (Fig 1D) (Fig 1D, S4A histone marks tended to be inversely proportional (S4B- C Fig). Moreover, we found that high 153 H3K27me3-H3K9me3 difference is mainly associated with genomic regions whereas low 154 H3K27me3-H3K9me3 difference is more associated with intergenomic regions (S4D Fig). We 155 propose that in this cell line, in which H3K9me3 is over-present, difference in the intensity 156 signal between H3K27me3 and H3K9me3 could be used to distinguish "facultative-like" 157 heterochromatin and "constitutive-like" heterochromatin. To decipher whether mH2A1.1 158 preferentially overlaps with H3K9me3-or H3K27me3-marked heterochromatin domains, we 159 analyzed its association depending on the difference in the signal intensity of both marks. We 160 found that mH2A1.1 (and PARP1) binding were proportional to the abundance of H3K27me3 161 minus H3K9me3 (S4B, S4E, S4F Fig), indicating that these two proteins predominantly 162 associate with "facultative-like heterochromatin" in this TNBC breast cancer cell.

164
When specifically examining the chromatin landscape at promoters (TSS +/-1kb), we 165 found that mH2A1.1 enrichment correlated with H3K4me3 and H3K27ac, as well as with BRD4, 166 H3.3 and Pol II binding (Fig 2A, 2B & S5A, S5B Fig), suggesting a role for mH2A1.1 in 167 transcription initiation-regulated processes. At promoters, mH2A1.1 distribution inversely 168 coincided with heterochromatin marks (Fig 2A & S5A, S5B Fig). We further determined that 169 enrichment of mH2A1.1 centered at the TSS was proportional to the level of transcription ( and similar to the one of Pol II at the NFR, bordered by H3K27ac marked nucleosome regions, 172 but larger than that of Pol II (Fig 2E, S5F, S5G Fig).

174
We next questioned whether this profile was linked to the mechanism by which 175 mH2A1.1 regulates gene expression. To this end, we separately determined mH2A1.1 binding 176 at genes repressed or activated (RG and AG) by mH2A1.1. At both gene categories, mH2A1.1 177 was highly enriched at the TSS (+/-2 kb) (Fig 3A-B). However, mH2A1.1 binding was restricted 178 to the TSS of mH2A1.1 AGs, while it associated both with promoter regions and the gene 179 bodies of mH2A1.1 RGs (Fig 3). mH2A1.1 association correlated with the level of binding of 180 Pol II, H3.3 and BRD4 (Fig 3C). Interestingly at RGs, we detected Pol II at the promoter and 181 over the elongation-characteristic H3K36me-marked gene body (Fig 3B; S6 Fig) We plotted the ChIP-seq signal of Pol II and H3K36me3 around TSS +/-10 kbp for each gene 193 ranked according to their pausing index (Fig 4A). In agreement with the literature (Elrod et al.,194 2019), the level of H3K36me3 was greater over the body of genes with low PI compared to 195 genes with high PI (Fig 4A, and S8A). We further observed that confinement of mH2A1.1 to 196 the TSS and its absence from the gene body was characteristic of genes with a high PI (Fig 4A,  197 and S8A). In agreement, the width of mH2A1.1 peaks overlapping with TSSs, as well as that of 198 Pol II peaks, correlated negatively with the PI (S8B-C Fig). H3.3 follows the same binding profile 199 as mH2A1.1. BRD4, RING1B and PARP1 were mainly enriched at the TSS, slightly more at high 200 PI-genes (Fig 4A, S8A and S9).

202
The majority of mH2A1.1-AGs (85%) have a pausing index greater than 2 (Fig S9C), a PI 203 value that can be used as a threshold to distinguish paused from not paused genes (Day et al.,204 2016). In agreement, the average PI of these genes was significantly higher than any other 205 gene category tested (Fig 4B-D). At the difference, RGs are enriched in low PI-marked genes 206 (Fig 4B-C). Furthermore, ChIPqPCR analysis of Pol II at three mH2A1.1-activated genes (RBL1, 207 GTF2H3 and E2F3) showed that the amount of Pol II at promoter regions was multiplied by a 208 ~3-fold factor upon siRNA reduction of mH2A1.1 (Fig 4E and S10A). On average, we observe 209 that depletion of mH2A1.1 induces an increase in the PI of RBL1, GTF2H3, and E2F3 genes by 210 a factor of 1.4, 1.6, and 1.7, respectively. These results suggest that mH2A1.1 may enhance 211 gene expression by promoting Pol II pause release. 212 213 mH2A1.1 binds enhancers. In addition to promoters, mH2A1.1 also associates with intergenic 214 regions (Fig 1E). Genome-wide, mH2A1.1 binding highly correlated with H3K4me1 (PCC of 215 0.55) and to a lesser extent with H3K27ac (PCC of 0.18), two chromatin marks which 216 characterize enhancer regions (Creyghton et al., 2010) (Fig 1D). In agreement, we found that 217 mH2A1.1 binding was significantly enriched at enhancers (Fisher exact test: p-value < 2.2x10 -218 16 and odd ratio = 5.23) (Fig 5A-B (Fig 5E and S11C). Overall, these results 224 show that mH2A1.1 binds enhancers and super-enhancers in association with BRD4 and 225 RING1B. 226 227 mH2A1.1-target genes regulation does not require changes in enhancer-promoter looping.

228
Because mH2A1.1 binds enhancer and promoter regions (Fig 3 and Fig 5) We aggregated the total number of detected interactions per gene for mH2A1.1-activated, -234 repressed or -independent genes. For each category, the average number of interactions 235 detected per gene was identical in control and mH2A1.1 KD cells (Fig 6A). The average 236 intensity of these interactions with adjacent genomic regions (+/-1.5 Mb around the gene) 237 (Fig 6B-D do not appear to require mH2A1.1. Yet, quantification of the PCHiC interactions showed that 240 mH2A1.1-AGs have on average a greater number of interactions than mH2A1.1-RGs (Fig 6A,  241 6E). However, interactions at mH2A1.1-AGs showed weaker signal intensities (Fig 6B, S12A), 242 suggesting that AG and RG reside within two types of interaction landscapes. 243 244 mH2A1.1 is enriched at the enhancers associated with RGs than those associated with 245 AGs (S12B-C). Active chromatin marks and co-activators are also more abundant to RGs 246 related enhancers than AGs related ones (H3.3, Pol II, BRD4) (S12B-C, S13 Fig) in agreement 247 with the fact that these genes are in average more transcribed than AGs. Although, loss of 248 mH2A1.1 did not induce any global changes in promoter contact numbers or frequencies, 249 closer inspection of the interaction landscape of a few mH2A1.1-regulated genes revealed 250 reproducible changes in the intensity of interactions at certain enhancers in mH2A1.1 KD vs 251 wt cells (Fig 6F, 6G and S14B-C). For example, we observed an increase in the intensity of some 252 interactions at the RG FRAS1 upon mH2A1.1 depletion (Fig 6F), and a decrease in the intensity 253 of interactions at the AG ARDDC3 (Fig 6G). But this cannot be generalized, having also 254 observed a decrease in the intensity of some interactions at RGs and an increase in the 255 intensity of interactions at AGs (data not shown). Thus, the gain or loss of interactions does 256 not appear to be related to transcriptional changes mediated by the loss of mH2A1.1. In 257 addition, ChIP-qPCR of Pol II at the TSS of these AGs showed an increase in Pol II association 258 at TSS upon loss of mH2A1.1 (Fig 6H and S10B). This observation is similar to that one 259 observed for AGs (Fig 4E) whose 3D organization is not affected by mH2A1.1 depletion (S14D MDA-MB231 cells became more elongated after 2-3 days (Fig 7A and S16A). Using 274 immunofluorescence against cytoskeleton proteins (actin, tubulin-α, vimentin), we observed 275 that the cytoskeleton organization was modified by the loss of mH2A1.1 (Fig 7A). Moreover, of MDA-MB231 cells was significantly increased compared to control cells (Fig 7A-B). The 283 effect of knocking down the other isoform, mH2A1.2, was opposite to that of mH2A1.1 (S16 284 Fig).

286
Strikingly, mH2A1.1-AGs involved in cytoskeleton organization and cell adhesion were 287 also amongst genes with a high Pol II pausing index (Fig 7C, D), compared to cell cycle and 288 DNA repair mH2A1.1-AGs (S15C Fig). Overall, we conclude that mH2A1.1 impedes the 289 migration capacity of MDA-MB231 breast cancer cells in part by promoting expression of 290 genes modulating cell migration capacity. 291 292

293
Regulating gene expression in a particular cell type requires fine-tuning transcriptional 294 response. The concentration and the relative ratio between factors required for these 295 regulatory mechanisms ensure rapid adjustments to maintain homeostasis or to respond to 296 stimuli and stress. In this study, we identify the histone variant mH2A1.1 as a means to operate 297 these adjustments in TNBC cells.

299
We present the first genome-wide map of endogenous histone variant mH2A1.1 in 300 human breast cancer cells. We discovered that the mH2A1.1 variant specifically associates 301 with transcription regulatory elements, promoters and enhancers, in addition to large 302 domains of facultative heterochromatin. Binding to promoters occurred in sharp, narrow 303 peaks as opposed to the larger signals detected in heterochromatin seen previously ( . Moreover, we found that selective depletion of the mH2A1.1 isoform was sufficient to 306 modify expression of hundreds of actively transcribed genes in the MDA-MB231 TNBC cell line 307 (Fig 1A-C). All of these genes are highly expressed in this cell line. We uncovered two distinct 308 mechanisms through which mH2A1.1 regulates their transcription and link them to the 309 chromatin landscape in which the affected genes reside. 310 311 The first mechanism consists in dampening transcription of highly expressed genes. 312 Indeed, in the absence of mH2A1.1, these mH2A1.1-RGs are overexpressed. mH2A1.1 binds 313 the gene bodies alongside RNA pol II, as well as their associated promoters (Fig 3 and S12B). 314 Importantly, these domains are also characterized by the presence of RING1B and the 315 Polycomb-induced histone modification, H2AK119ub (Chan et al., 2018) on their enhancers 316 and promoters (Fig S7, S12 found for mH2A1.1-targets genes (Fig 1B,C). We can postulate that the presence of mH2A1. suggesting that mH2A1.1 may limit transcriptional noise and serve as a brake (Fig 7E).

327
The second mechanism is specific to genes in which RNA pol II is paused. Here, in 328 contrast to the RGs, mH2A1.1 recruitment is restricted to the TSS of these genes. The deletion 329 of mH2A1.1 leads to a reduction of the transcriptional level of these genes (mH2A1.1-AGs) as 330 well as an accumulation of Pol II at their TSS (Fig 1A-C, Fig 4E, Fig 6H). It is indeed tempting to 331 combine these two observations to propose that mH2A1.1 may assist in the conversion of 332 promoter-locked RNA polymerase II into a productive and elongated Pol II. process, resulting in the maintenance of torsional stress, accumulation of Pol II and inhibition 341 of transcription. Finally, the chromatin organization at mH2A1.1-AGs, which are mostly 342 characterized by Pol II pausing, seems more dynamic than at mH2A1.1-RGs, with more 343 frequent but weaker contacts detected by PCHiC (Fig 6). In these domains, as Pol II is retained 344 in pause at the TSS, its release could be facilitated by transient TSS-enhancer contacts in 345 search for co-occupancy of coactivators and Pol II (Fig 7E). This observation can be generalized 346 to all paused genes in the MDA-MB231 cell line (data not shown) identifying a new 347 characteristic of paused genes with respect to their 3D organization of the genome. 348 Conversely, the 3D organization of the mH2A1.1-RG loci appears to be relatively stable, 349 reminiscent of a productive and well-organized environment for transcription (Fig 6, Fig 7E).

351
Currently, only one study has analyzed transcriptional activities of mH2A1.1 related to 352 its genomic localization, and that is in murine muscle cell line C2C12 (Hurtado-Bagès et al., Here, in addition to that silencing of 379 mH2A1.1 enhances cell migration in the MDA-MB231 cell line (Fig 7A-B), we begin to identify 380 the underlying molecular mechanism with direct transcriptional stimulation by mH2A1.1 of 381 genes highly dependent on the Pol II pausing (Fig 7C-D).

383
We further demonstrate that mH2A1.1-bound chromatin co-localizes with the 384 H3K9me3 histone mark (Fig S4A). A fraction of these sites is devoid of H3K27me3 and could 385 correspond to the identified mH2A localization at constitutive heterochromatin (Douet et al.,386 2017). However, the vast majority of mH2A1.1-bound H3K9me3-decorated chromatin 387 contained also tri-methylated H3K27 (Fig S4). This difference may be a feature of the MDA- Despite the association of mH2A1.1 with heterochromatin, its phenotypic knockdown 396 was not sufficient to reactivate silenced genes present in these domains (Fig 1B, 1C and S4). 397 Different hypothesis could explain this result. The first hypothesis could be that mH2A1 398 isoforms (mH2A1.1 and mH2A1.2) have redundant actions at heterochromatin. Here, we 399 specifically depleted mH2A1.1 without affecting the expression of mH2A1.2 ( Fig S1D). participates to this process especially since we have identified a preferential association of 427 mH2A1.1 with SEs (Fig 5D-E). SEs are known to play an important part in many diseases, 428 including several cancers in which they drive expression of oncogenes (   genes are considered as paused genes with PI of 3,28, 2,9 and 3,2, respectively. Right panel: 504 ChIP-qPCR of Pol II on WT and mH2A1.1-depleted cells. Hetero corresponds to a negative 505 position. For each gene, Pol II enrichment was evaluated on the TSS and a gene body region. 506 Results from additional biological replicates are given S10A. (from + 50bp to TES). ns, not significant. "****" = p-value < 2.2x10 -16 . (C) Overlap of mH2A1.1-689 regulated genes with paused genes. Enrichment of mH2A1.1-target genes with paused genes 690 are measured using fisher exact tests. Of note, only mH2A1.1-target genes characterized by a 691 PI were used to generate this Venn diagram. 692 693 S10 Fig. mH2A1.1 Tables   768   769  S1 Table:  with rotation (or 1h30 at RT). Primary antibodies are described in the S2 Table. Rabbit anti-801 mH2A1.1 antibody was generated according to immunization protocol from Agro-Bio -La 802 fierté Saint-Aubin -France. Membranes were next incubated with secondary antibody in PBS-803 Tween 0.4% -Milk 5% 1h at RT with rotation and the signal was detected using 804 chemiluminescence. Secondary antibodies are described in the S2 22.10 6 cells) and 20 µg of the corresponding antibody were used. Each ChIP were sequenced 875 by the MGX genomic platform (Montpellier) using the Hi-seq2500 Illumina sequencer. 876 ChIPqPCR of Pol II were done following the same protocol as for Pol II ChIP-seq. qPCR were 877 done on ChIP samples and Input (1% of DNA used for ChIP). qPCR results are normalized using 878 the signal obtained with the input (% of input). Primers used are given in S6 Table.  879  880 Strand-specific total RNA library preparation. Total RNA was isolated using the RNAeasy midi 881 kit (Qiagen). RNA-seq quality and quantity control were performed using a Nanodrop 882 (NanoDrop2000, Thermo) and BioAnalyser. Library preparation and sequencing was done by 883 GeT core facility, Toulouse, France (http://get.genotoul.fr) with the kit TruSeq Stranded total 884 RNA according to manufacturer's institutions. Sequencing was done HiSeq3000-HWI-J00115 885 according to the manufacturer's protocol.  Corresponding ChIP-seq data generated from genomic DNA (Input) were used as control for 907 every bigWig files normalization (options: --normalizeUsing RPKM --operation subtract --908 binSize 50 bp --smoothLength 150 bp). Peaks were determined with the enrichR function of 909 NormR package (Helmuth J, 2018). NormR parameters were adjusted depending on the bigwig 910 profiles for each ChIPseq data. mH2A1.1-specific peaks were used for all analysis and 911 correspond to the commun peaks between mH2A1.1 and mH2A1 ChIPseq. Number of peaks 912 for each ChIP-seq data are listed in S3 Table. All downstream analyses were mainly performed 913 with R studio. ChIP-seq signal and peaks positions visualization were obtained with IGV 914 (Thorvaldsdóttir et al., 2013). Boxplots were done with ggplot2 (H. Wickham., 2016). 915 Distributions of mH2A1 isoforms and H3K27me3/H3K9me3 common peaks identified at 916 specific genomic features were calculated using ChIPseeker package with default parameters 917 (Figs 1E and S4D) (Yu et al., 2015). Statistical analyses are presented in Statistics and 918 Reproducibility paragraph. 919 920 Identification of "putative" enhancers and super-enhancers. All putative enhancers were 921 determined with ROSE utility tools based on H3K27ac signal outside TSS (+/-2 kb) to avoid TSS 922 bias (Fig 5A-B) (Blinka et al., 2017). TSS annotation is based on 923 TxDb.Hsapiens.UCSC.hg38.knownGene release (n=25,668 annotated genes). Super-enhancers 924 were determined with ROSE utility tools based on H3K27ac signal (options : stitching_distance 925 = 12.5 kb and TSS_exclusion_zone_size : 2500 bp) (Fig 5D-E)  smaller than 1 kb were excluded from the analysis. Moreover, pausing index were not 935 calculated for the genes having a Pol II density lower than 1.2 in the promoter-proximal region 936 and a Pol II density in the transcribed regions lower than 0. Using this threshold, we only 937 calculated pausing index for transcribed genes having a Pol II binding (n=10,564 genes).

939
"Not paused" genes were defined as genes that have a PI lower to 2 (n=3,356 Each scare on the heatmap shows the results of a fisher exact test between the two groups 955 tested. The positive or negative association between the two groups tested is established by 956 the odd ratio, represented by the "score" (log 2 (Odd Ratio) = LOR) and the color scale, that is 957 proportional to the score. Significatively of the overlap is assessed by the p-value, represented 958 by the stars (* ≤ 0.05 and highly significant when ** ≤ 0.01; *** ≤ 0.001; **** ≤0.0001). Groups 959 used for the analysis were divided in equal size according to the ChIP-seq signal. were removed and not used to generate the profiles. Heatmaps profiles were also performed 966 with R Seqplot package using bigwig files (function getPlotSetArray and plotHeatmap) 967 (Stempor & Ahringer, 2016). Some heatmaps profiles were also ranked according to ChIPseq 968 signal, PI index or gene expression log2Fold change. On all heatmaps, colour intensity reflects 969 level of ChIP-seq enrichment. Colour intensity autoscale were always used excepted for 970 heatmaps S12B and S13A to compared to relative enrichment between mH2A1.1-target genes 971 and their associated enhancers. On profiles and heatmaps, gene directionality was not 972 ignored, meaning that all gene bodies are artificially placed on the right place of the plots. with EnhancedVolcano package (Kevin Blighe, Sharmila Rana, 2018) (Fig 1A). The mH2A1.1 KD 995 de-regulated genes are listed in S1 Table.  996  997 Promoter Capture HiC and library preparation. PCHIC data were generated on MDA-MB 231 998 cells in control and mH2A1.1 KD cell using the siRNA#1 (see Methods part "Transfection of 999 siRNA and plasmids and S5 Table). PCHi-C was essentially performed as in (Schoenfelder et  Intensity of interactions were estimated based on CHi-C read counts for each biological 1008 replicate. Only reads > 5 for each biological replicate were kept and interactions between a 1009 bait and the other-end closer than 1.5 Mbp. Finally, for each interaction, read counts were 1010 quantile normalized using the function "normalizeBetweenArrays" from the LIMMA package 1011 (Ritchie et al., 2015). Means between biological replicates were used. ChICMaxima Browser 1012 were used to generate PCHiC profiles (https://github.com/yousra291987/ChiCMaxima) (Ben 1013 Zouari et al., 2019). ChiCMaxima-called and merged interactions wereoverlapped with 1014 enhancers using the findOverlaps function from the R GenomicRanges package (Lawrence et 1015(Lawrence et al., 2013.More than one enhancer can significantly be in interaction with mH2A1.1-regulated 1016 genes. To simplify, only one enhancer per gene was conserved to generate the heatmaps in 1017 S12B and S13A. Some mH2A1.1-target genes are not present in those heatmaps because they 1018 do not have PCHIC-called interactions with an enhancer or did not have PCHi-C capture 1019 oligonucleotides. Overlaps of those genes with mH2A1.1-regulated genes were done (mH2A1.1-activated 1030 (n=64/533), mH2A1.1-repressed genes (n=18/412). We finally took genes related to DNA 1031 repair (GO:0006281), cellular response to DNA damage stimulus (GO:0006974) (n=533). 1032 Overlaps of those genes with mH2A1.1-regulated genes were done (mH2A1.1-activated 1033 (n=37/533), mH2A1.1-repressed genes (n=4/412). For fisher test heatmap with PI, only genes 1034 having a PI were used. N indicates the number of genes used for the analysis in Fig 4B, Fig 7D  1035 and Fig 15C. 10% FBS medium in the lower chamber. Following incubation for 16h at 37°C, the cells were 1043 fixed with 3.7% formaldehyde for 2 min at RT. Cells permeabilization was carried out with 1044 methanol incubation for 20 min at RT. Cells were then stained with Giesma for 15 min at RT.