Transcriptional elongation machinery controls vulnerability of breast cancer cells to PRC2 inhibitors

CTR9 is the scaffold subunit in Paf1c, a multifunctional complex regulating multiple steps of RNA Pol II-mediated transcription. Using inducible and stable CTR9 knockdown breast cancer cell lines, we discovered that the expression of a subset of KDMs, including KDM6A and Jarid2, is strictly controlled by CTR9. Global analyses of histone modifications revealed a significant increase of H3K27me3 upon loss of CTR9. Loss of CTR9 results in a decrease of H3K4me3 and H3K36me3 in gene bodies, and elevated levels and genome-wide expansion of H3K27me3. Mechanistically, CTR9 depletion triggers a PRC2 subtype switching from PRC2.2 to PRC2.1. As a consequence, CTR9 depletion generates vulnerability that renders breast cancer cells hypersensitive to PRC2 inhibitors. Our findings that CTR9 demarcates PRC2-mediated H3K27me3 levels and genomic distribution, provide a unique mechanism of transition from transcriptionally active to repressive chromatin states and sheds light on the biological functions of CTR9 in development and cancer.

The Polymerase-Associated Factor 1 complex (Paf1c) was originally identified as a Pol II-3 interacting complex in Saccharomyces cerevisiae over 20 years ago. Paf1c has emerged 4 as a highly conserved and multifunctional complex regulating multiple steps of RNA 5 polymerase II (RNAPII)-mediated transcription (Van Oss et al., 2017). In budding yeast, 6 Paf1c consists of five subunits, which includes Paf1, Ctr9, Cdc73, Leo1 and Rtf1. In higher 7 eukaryotic organisms, Rtf1 is loosely attached to Paf1c, while Ski8/Wdr61 is an additional 8 subunit of Paf1c (Zhu et al., 2005). Paf1c regulates multiple phases of transcription, 9 including transcription elongation, transcription termination, and RNA 3'-end 10 polyadenylation (Van Oss et al., 2017). Recent studies show that Paf1 and Ctr9 are 11 essential for Paf1c integrity (Chu et al., 2013;Vos et al., 2018;Yu et al., 2015). Paf1c 12 promotes RNAPII pause release. In addition, it regulates gene expression by controlling  Polycomb repressive complex 2 (PRC2), the sole mammalian multi-subunit 28 complex responsible for H3K27me3, is essential for maintaining cellular identity and 29 development of multicellular organisms (Yu et al., 2019a). PRC2 is comprised of the core 30 activity (Kim and Roberts, 2016). Furthermore, single-cell analysis showed that loss of 23 H3K27me3 was associated with treatment resistant breast cancer (Grosselin et al., 2019), 24 highlighting the need to further understand how chromatin states affect drug sensitivity. tissue-specific expression patterns, and their expression levels and activities are 1 stringently regulated (Lan et al., 2008). Dysregulation of KDMs, such as amplification, 2 mutation, abnormal expression, have been implicated in breast tumorigenesis (Bamodu 3 et al., 2016;Taube et al., 2017). For instance, aberrant expression of KDM5B and KDM6A 4 is associated with aggressive breast cancers (Bamodu et al., 2016;Taube et al., 2017). 5 However, the mechanisms regulating their expression remain largely unknown. 6 Here we identified a subset of KDMs, including PCR2 target genes KDM6A and 7 JARID2, whose expression is precisely controlled by CTR9 levels in breast cancer cells. H3K27me3. This effect is likely attributed to switching from PRC2.2 to PRC2.1, which has 12 high H3K27me3 activity. Moreover, exogenous expression of KDM6A or JARID2 can 13 partially reverse the phenotypic defects in CTR9 knockdown cells. Finally, CTR9 depleted 14 cells become addicted to H3K27me3, and are hypersensitive to PRC2 inhibition. 15 Collectively, our study uncovers a unique mechanism by which a transcriptional elongation 16 factor demarcates the PRC2-mediated H3K27me3 domains in breast cancer cells. The 17 mechanism of regulation of H3K27me3 by CTR9 is likely conserved across cell types, and 18 CTR9-dependent response to EZH2 inhibitors provides therapeutic vulnerability for breast 19 cancer treatment. KDM5B, KDM6A and JARID2. We found that these KDM genes were down-regulated 8 when CTR9 was knocked down by shRNA in MCF7 cells after a 7-day doxycycline (Dox) 9 treatment ( Figure 1B) regardless of whether 17β-estradiol (E2) was present or absent. 10 RNAPII binding peaks at the transcription start site (TSS) of several KDM genes also 11 decreased in response to CTR9 depletion ( Figure S1A). We validated the transcriptome 12 array results by RT-qPCR. Figure 1C shows that silencing of CTR9 after a 7-day Dox 13 treatment reduced the total mRNA levels, and reduced the expression of ribosome-14 associated RNAs, which are indicative of actively transcribed mRNAs, of six KDMs. 15 However, the mRNA levels of KDM4B and KDM6B did not significantly change. CTR9's 16 regulation of KDMs was also observed at the protein level. As shown in Figure 1D, the 17 protein levels of six KDMs significantly decreased in the total cell lysates and extracted 18 chromatin fractions. Consistent with their mRNA levels in response to CTR9 depletion,19 KDM4B and KDM6B protein levels also remained unchanged. To exclude the possibility 20 that this observation was specific to MCF7, or due to off-target effects of anti-CTR9 shRNA,  To assess if decreased KDM expression in response to loss of CTR9 is reversible, 28 we treated the MCF7-tet-on-shCtr9 cells with Dox for seven days, removed Dox on Day 29 8, and continued culturing the cells for seven days. CTR9 expression was gradually 30 8 restored after removal of Dox, and returned to its original expression levels after seven 1 days, as we have published previously (Zeng and Xu, 2015). The protein levels of KDMs 2 were also found to be dynamically regulated by the addition and withdrawal of Dox, 3 mirroring the expression pattern of CTR9. As expected, the protein levels of KDM4B and 4 KDM6B remained constant, regardless of changes in CTR9 levels ( Figure 1E). These  Total histones were extracted from nuclear pellets of MCF7-tet-on-shCTR9 cells after 18 treatment with Dox for seven days, and subjected to liquid chromatography, followed by 19 tandem mass spectrometry (LC-MS/MS). The transcription-coupled histone modifications 20 such as H3K4me3 and H3K36me3 decreased when CTR9 was depleted as we reported 21 previously (Zeng and Xu, 2015). Surprisingly, transcriptional repressive histone markers 22 H3K27me2 and H3K27me3 robustly increased ( Figure 2A). The histone modification 23 changes in response to CTR9 KD were validated by western blotting ( Figure 2B) where a 24 significant increase of H3K27me3 was observed after a 7-day Dox treatment. This result 25 suggests that loss of CTR9 results in a global elevation of H3K27me3 levels. Furthermore, 26 the accumulation of H3K27me3 in CTR9 KD cells was not cell type specific, as this was 27 also observed in MCF7, T47D, and BT474 cells stably expressing two distinct CTR9 28 targeting shRNAs (shCtr9, #3 or #5), as measured by ELISA and western blotting of 29 purified histones (Figure S2A-C). In contrast, Dox treatment of MCF7-tet-on parental cells 30 9 did not result in any changes in H3K27me1/2/3 ( Figure S2D). To further interrogate 1 whether elevation of H3K27me3 levels in CTR9 KD cells can be reversed by re-expressing 2 CTR9 by removal of Dox, we performed a time-course Dox addition/removal treatment 3 experiment as described above. Figure 2C shows that H3K27me3 levels increased when 4 CTR9 was depleted, and H3K27me3 returned to its original levels when CTR9 is restored.

5
The dynamic changes of H3K27me3 in response to CTR9 levels were further quantified 6 by flow cytometry using an Alexa Fluor ® 647-labelled H3K27me3 antibody ( Figure 2D) as 7 well as quantitative mass spectrometry analyses ( Figure 2E) . As anticipated, H3K27me3 8 signal intensity increased in response to CTR9 depletion (+Dox, 7 days) as compared with 9 untreated MCF7 cells (+ Dox, 0 days). When CTR9 was re-expressed by removing of Dox 10 for seven days, H3K27me3 levels decreased. To visualize the H3K27me3 changes in 11 individual cells, we performed 3D scanning of immuno-fluorescence staining of 12 H3K27me3 ( Figure 2F). The results showed that, although the response of individual cells 13 varies, possibly due to heterogenous expression levels of shCTR9 (expression indicated 14 by eGFP), the overall intensity of H3K27me3 staining is inversely correlated with CTR9 15 levels in a dynamic manner. Quantification of H3K27me3 intensity is depicted as the ratio 16 of H3K27me3 to nucleus staining intensity from 20 selected cells with the entire nucleus 17 in the 3D volume view ( Figure 2G). Together, these data strongly suggest that CTR9 is a 18 bona fide regulator of cellular H3K27me3 levels. H3K36me3 and H3K27me3 in response to the gradual decrease of CTR9, while the stable 28 CTR9 KD MCF7 cells expressing a scrambled shControl, and two distinct CTR9 shRNAs, 29 allow us to assess permanent changes in histone modifications.

30
In the CTR9 inducible knockdown system, total H3K27me3 peak numbers 1 increased over two-fold when CTR9 was silenced after a seven-day Dox treatment (Figure   2 3A). The H3K27me3 peaks increased across the genome and were not restricted to 3 specific chromatin regions (i.e., promoter, exon, intron and intergenic regions) ( Figure 3A).

4
Next, we classified H3K27me3 peaks based on peak intersection, and generated four 5 distinct clusters in the vehicle or Dox-treated groups ( Figure 3B- where one peak in the vehicle treated group intersects with one or more peaks in the Dox 10 treated group, or vice versa. Based on this peak classification, Figure 3B (bottom) depicts 11 clustered heatmaps and line plots summarizing the average ChIP-seq signals in both 12 vehicle and Dox-treated groups at the corresponding peak regions using the locus with 13 the highest ChIP-seq signal as the center with ± 2kb expansion. Representative snapshots 14 of the genome browser for each cluster are shown in Figure 3C. Because H3K27me3 15 peaks are too broad to define their regulated genes, we analyzed histone modification 16 changes of 240 previously identified Ctr9 target genes in MCF7 cells. These genes were 17 identified based on two criteria: decreased mRNA expression and reduced RNAPII 18 binding to promoters in response to loss of CTR9 (Zeng et al., 2016;Zeng and Xu, 2015).

19
Among the CTR9 regulated genes, 229 genes elicited either reduced H3K4me3, reduced 20 H3K36me3 signals, or both. Only 10% of CTR9-regulated genes showed increased 21 H3K27me3 signal, all of which simultaneously harbored decreased signal of either 22 H3K4me3, H3K36me3, or both ( Figure 3D, yellow and red bars). Figure 3E showed a 23 significant reduction of H3K4me3 signal, whereas the H3K27me3 signal only modestly 24 increased on CTR9 target genes. The negative correlation between H3K27me3 and 25 H3K4me3/H3K36me3 across the genome was not statistically significant (data not shown), 26 indicating that loss of active histone marks is necessary, but not sufficient, for deposition 27 of H3K27me3 marks.

28
In comparison to the inducible CTR9 KD cells, the global increase of H3K27me3 29 peak numbers was much more pronounced in CTR9 stable KD MCF7 cells (shCtr9#3 and 30 shCtr9#5) when peaks were normalized to control shRNA expressing cells ( Figure S3A).

1
The increased H3K27me3 peaks spread across the genome ( Figure S3A) is indicative of 2 expansion of H3K27me3 domains. In comparison with the CTR9-inducible system, Cluster 3 I H3K27me3 peaks decreased from 7,128 to 3,935 and 2,787 in the shControl cells, 4 respectively ( Figure 4A). In contrast, Cluster II H3K27me3 peaks increased from 23,581 5 to 40,595 and 53,078, respectively, for shCtr9#3 and shCtr9#5 ( Figure 4B). In addition, 6 the peak intensity of H3K27me3 within those TSSs ± 5kb regions with gained H3K27me3 7 peaks after CTR KD also significantly increased in CTR9 stable KD cells ( Figure S3B), as 8 well as the H3K27me3 peak width across the genome in intergenic, exon, intron, and 9 promoter regions ( Figure S3C). The five-fold increase of total H3K27me3 peaks, 10 increased peak intensities, and broadened peak width in response to permanent CTR9 11 knockdown indicate that loss of CTR9 resulted in significant expansion of H3K27me3 12 domains in chromatin. In contrast to what was found in the inducible CTR9 KD cells, a 13 statistically significant, negative correlation between H3K27me3 and H3K4me3 /

14
H3K36me3 was detected at the genome-wide level in CTR9 stable KD cells ( Figure S3D).

15
With regard to 240 CTR9-regulated genes, significant number of genes ( ~30 for shCtr9#3 16 and ~80 for shCtr9#5) showed increased H3K27me3 signals in stable CTR9 KD cells 17 ( Figure 4B), and the increase of H3K27me3 is often coupled with decreased H3K36me3 18 but less prominent with decreased H3K4me3 ( Figure 4B). Previous studies have shown 19 that H3K27me3 and H3K36me2 could co-localize in some genomic regions, and depletion 20 of H3K36me2 is accompanied by deposition of H3K27me3 (Brien et al., 2012;Streubel et 21 al., 2018). We hypothesized that upon loss of CTR9, H3K4me3 and H3K36me3 in gene 22 bodies decreases followed by deposition of H3K27me3 marks, and the expansion of 23 H3K27me3 domains. 24 We further investigate the correlation between CTR9 and H3K27me3 level in breast 25 tumors using human breast cancer tissue microarrays (TMAs) containing over 300 breast 26 tumor specimens. After optimizing staining condition with antibodies to CTR9 and 27 H3K27me3 ( Figure S4A), the intensity of immunohistochemistry staining (IHC) was 28 determined by H-score graded optical density in the nucleus of epithelial cells in specimen 29 ( Figure S4B). Representative examples of breast tumor cores with high or low CTR9 IHC 30 12 staining and corresponding inverse H3K27me3 IHC staining were shown in Figure 4C.

1
The H-score based correlation analysis revealed a significant reverse correlation between 2 CTR9 expression and H3K27me3 abundance ( Figure 4D). 3 We next investigated whether CTR9-regulated KDM genes showed dynamic 4 histone modification changes. Indeed, KDM genes showed decreased H3K4me3 peaks 5 near gene promoters ( Figure S5A) when CTR9 was inducibly knocked down by Dox-6 treatment and H3K27me3 peaks increased ( Figure S5B). In contrast, KDM4B, a non-7 CTR9 target gene, did not show significant changes in H3K4me3 or H3K27me3 8 enrichment. H3K27me3 signals at FOXC1 in shControl and shCTR9 stable KD cells illustrated a 20 dramatic increase of H3K27me3 in FOXC1 upon loss of CTR9 ( Figure S6B). These results 21 demonstrate that CTR9 regulates PRC2 target genes in a H3K27me3-dependent manner.

22
The increased total H3K27me3 levels, and broadened H3K27me3 peak-width  Figure 5D). To test if the inverted levels of PRC2 13 facultative subunits in response to changes in CTR9 levels resulted in switching of PRC2 14 subtypes, we performed co-immunoprecipitation using antibodies against the core 15 subunits of the PRC2 complex EZH2 and SUZ12 to pull down other core and auxiliary 16 proteins from nuclear extract. Figure 5E shows that, in the presence of CTR9, PRC2 core 17 subunits tended to precipitate PRC2.2-specific JARID2 and AEBP2. When CTR9 was 18 depleted, PRC2.1-specific subunits PCL2/3 and EPOP were enriched instead. We reason 19 that this switch is likely due to the relative abundance of facultative subunits corresponding 20 to CTR9 levels. In addition, the inversely correlated expression levels of CTR9 and 21 H3K27me3 levels can also be detected in breast cancer cell lines ( Figure 5F). CTR9 levels   All data support that CTR9 has a profound impact on H3K27me3 by regulating both 'writer'  for survival. If this were true, we expect that CTR9 knockdown cells would elicit higher 25 sensitivity to EZH2 inhibitors than parental cells. Cell viability was measured after 26 treatment with UNC1999, a chemical inhibitor targeting both EZH2 and EZH1, and 27 UNC2400, a structurally similar but inactive analog compound, to exclude the possibility 28 of off-target effects (Konze et al., 2013). As expected, CTR9 KD MCF7 cells were more 29 15 sensitive to UNC1999 than parental MCF7 cells, as shown by MTT assays, whereas EZH2 1 KD MCF7 cells (shEZH2) were insensitive to UNC1999, serving as a negative control 2 ( Figure 7A). A similar result was observed by cell counting ( Figure 7B) and colony 3 formation assays ( Figure 7C). UNC1999 inhibited the growth of CTR9 KD MCF7 cells in 4 a dose-dependent manner, whereas the negative analog UNC2400 showed no cytotoxic 5 effects. To exclude a drug-specific effect, we tested two additional mechanistically distinct 6 PRC2 inhibitors, GSK343 and EED 226 (Qi et al., 2017;Verma et al., 2012). The results 7 ( Figure S8A-B) were similar to those of UNC1999. Collectively, our data indicate that 8 depletion of CTR9 leads to increased sensitivity towards the PRC2 inhibitors. To 9 determine whether the EZH2 inhibitor causes elevated apoptosis or necrosis upon CTR9 10 depletion, we performed flow cytometry analyses after labeling cells with PI and annexin 11 V-FITC ( Figure 7D). The dose-responsive increase of apoptosis by UNC1999 was 12 quantified in Figure 7E. Both apoptotic and necrotic cells were detected in CTR9 KD cells,  The results shown that the CTR9 depleted spheroids (shCtr9#3/shCtr9#5) disassemble in 24 a much rapid manner in contrast to the control group (shControl), whereas 50 M 25 UNC2400 treatment does not affect the integrity of 3D spheroids for both CTR9 KD and 26 control KD groups. Collectively, EZH2 inhibitor UNC1999 can inhibit growth and induce 27 apoptosis in MCF7 cells and increase their sensitivity to EZH2 inhibitors depends on the 28 levels of CTR9. CTR9-depleted cells gain H3K27me3 and elicit stronger sensitivity to 29 EZH2 inhibitors than parental cells in both 2D monolayer and 3D spheroid models. Here we report that CTR9 governs the establishment of H3K27me3 repressive domains, 3 beyond its well characterized functions in transcriptional regulation and transcription-4 coupled histone modifications (i.e., H2Bub, H3K4me3, H3K36me3). This discovery 5 provides an explanation for the discrepancy between the phenotypes observed in lower 6 eukaryotes (i.e. yeast) and multicellular organisms when CTR9 is depleted. While Ctr9 is  Third, we envision that the PRC2 subtype switching caused by decreased JARID2 21 in response to CTR9 depletion is the major mechanism for the global H3K27me3 increase  therapeutic target for TNBC. Our data that TNBC cell lines display higher levels of EZH2 28 and H3K27me3 ( Figure 5F) support the application of PRC2 inhibitors in treating TNBC.

29
Moreover, depletion of CTR9 in ER + cells increases sensitivity of cells to PRC2 inhibition 30 20 by >10 fold, suggesting that EZH2 inhibitors may also be applicable to ER-positive breast 1 cancer with lower levels of CTR9. We speculate that CTR9 levels, rather than ER status, 2 is a predictive biomarker for PRC2 dependency in breast cancer cells. Since Ctr9 3 depletion generates therapeutic vulnerability to pharmacological inhibition of PRC2, the 4 CTR9 expression levels may be used as a guideline for predicting PRC2 dependency and 5 EZH2 inhibitor sensitivity in broad cancer types.

5
Nuclei pellets were separated by centrifugation at 14,000 rpm at 4℃ for 10 mins.

5
Co-Immunoprecipitation using nuclear extract 6 Co-IP was performed as previously described (Malte, B., et.al., 2016). Briefly, 7 immunoprecipitations were performed in IP buffer [50mM Tris-HCl pH7.5,150mM NaCl, 8 2mM MgCl2, 0.5% NP-40 and 10% Glycerol] supplemented with protease and 9 phosphatase inhibitors before use. Approximately 1.5 -2 mg nuclear protein extract, as 10 quantified by a Bradford assay, was mixed with 5 g of antibody and 50 l of protein A 11 magnetic Dynabeads (Invitrogen, washed previously 3X in IP buffer) per IP reaction in 12 750 l total volume. Beads were washed three times with IP buffer, and once with PBST 13 the following day. Proteins were eluted in 75 l 2X SDS loading buffer with 50nM DTT and 14 heated at 95℃ for 15 mins before loading on an SDS-PAGE gel.

16
Peptide pulldown using nuclear extract 17 Peptide pulldown was adapted from protocol described in (Malte, B., et.al., 2016). Lysine (provided in kit) and the wells will turn into yellow. The absorbance of each well was read 11 on a microplate reader at 450 nm with an optional reference wavelength of 655 nm.

12
Absorbance output among different samples will first eliminate the background noise 13 measured by the blank wells and subsequently normalized by corresponding absorbance 14 in control wells coated with H3-total antibody.

Cell proliferation and two-dimensional (2D) colony formation assays 17
For cell counting-based proliferation assays, 1 × 10 5 cells were seeded into six 3.5 cm 18 petri dishes for compound treatment. MCF7-tet-on-shCtr9 cells were pretreated with 19 vehicle or 500 ng/mL Dox for 5 days before seeding to 3.5 cm petri dishes. Media were 20 changed every 48 hours. Cells were trypsinized and counted after Trypan blue exclusion 21 using an automated cell counter (Bio-Rad).

29
For 2D colony formation assays, 1000 cells were seeded into 3.5 cm petri dishes for 30 compound treatment, and media was refreshed every 4 days. After treatment with 31 compounds for 12-15 days, cells were washed with DPBS, fixed with 4% formaldehyde 32 for 10 min at room temperature, and stained with 0.05% crystal violet for 30 min at room 33 temperature. Images were taken on a Leica inverted microscope using the Leica 34 Application Suite. Colony numbers were counted using ImagePro software.

Chromatin immunoprecipitation (ChIP) 31
Cells in 15-cm dishes were washed once with PBS before cross-linking with PBS 32 containing 1% formaldehyde for 15mins at room temperature. Crosslinking was quenched 33 with 0.125 M glycine for 5 minutes at room temperature before two washes with ice-cold 34 PBS. Cells were scraped, harvested by centrifugation, and subjected to ChIP assays.

35
Crosslinked cells were lysed with lysis buffer 1 (10 mM HEPES pH 7.0, 10 mM EDTA, 0.5 36 mM EGTA, 0.25% Triton X-100, supplemented with 0.5 mM PMSF before use) with 37 rotation at 4 °C for 10 minutes. The crude nuclear pellets were collected by centrifugation 38 at 1500 rpm for 4 minutes at 4 °C. The supernatant was discarded, and the chromatin was fragments by sonication in ice-water bath at 4 °C using a Branson Sonifier 450 with a 1 microtip (40% amplitude, 3 seconds on, 10 seconds off, 3 minutes total pulse time).
2 Sonicated chromatin was centrifuged at 15,000 rpm for 15 minutes at 10 °C, and 3 concentration of nuclear proteins was determined using the BioRad Protein Assay 4 (BioRad). Equal amounts of total nuclear proteins were used for ChIP. Nuclear proteins 5 were supplemented with nuclear lysis buffer to achieve same final volumes between 6 different samples, and then diluted 1:10 with dilution buffer (20 mM Tris-HCl pH 8.1, 150 7 mM NaCl, 2 mM EDTA, 1% Triton X-100, supplemented with 1 x protease inhibitor cocktail 8 before use). Five percent of the chromatin fraction was removed and saved as input, and 9 the rest was pre-cleared with a normal IgG control before incubating with the antibody of 10 interest overnight at 4 °C.

11
On the following day, the immune complexes were incubated with Dynabeads TM  Table II. Fold change was determined by normalizing the 28 ChIPed DNA signal to the input DNA signal.

ChIP-seq library preparation 31
Prior to ChIP-seq library preparation, the concentration and size distribution of the ChIPed 32 DNA samples was determined using a Qubit Fluorometer (Thermo Fisher Scientific) and  ChIP-seq data analysis 6 ChIP-seq reads were aligned to human genome (hg38) by Bowtie (version 1.1.2) with 7 command line options '--quiet -q -v 2 -a --best --strata -m 1 --phred33-quals -S'. We 8 removed reads that were unmapped or PCR/optical duplicates and those that did not pass 9 platform/vendor quality controls. ChIP-seq peaks were called by MACS (version does not belong any of the other three peak categories. Peaks from the first three 42 categories were also considered as 'genic peak'.

43
To study H3K27me3 signals around TSS, we selected protein-coding transcripts that with 'exon region' and do not overlapped with any 'promoter region' or 'intron region'; (iv)

17
'intergenic peak': peaks that does not belong any of the other three peak categories.

18
Peaks from the first three categories were also considered as 'genic peak'.

19
For Figure S4D, 240 Ctr9-regulated genes were identified in previous published work.  The accession number for the ChIP-seq data reported in this paper is GEO: GSE133318.

13
Source data have been provided separately with supplementary files.

17
Statistical comparisons between two groups for RT-qPCR data and proliferation analyses 18 were performed with Graphpad Prism software 7.0 using a paired two tails t-test. The

19
sample size (n) is indicated in the figure legends and represents biological replicates.