RBBP4 modulates gene activity through acetylation and methylation of histone H3 lysine 27

RBBP4 is a core subunit of polycomb repressive complex 2 (PRC2) and HDAC1/2-containing complexes, which are responsible for histone H3 lysine 27 (H3K27) methylation and deacetylation respectively. However, the mechanisms by which RBBP4 modulates the functions of these complexes remain largely unknown. We generated viable mouse embryonic stem cell lines with RBBP4 mutations that disturbed methylation and acetylation of H3K27 on target chromatin and found that RBBP4 is required for PRC2 assembly and H3K27me3 establishment on target chromatin. Moreover, in the absence of EED and SUZ12, RBBP4 maintained chromatin binding on PRC2 loci, suggesting that the pre-existence of RBBP4 on nucleosomes serves to recruit PRC2 to restore H3K27me3 on newly synthesized histones. As such, disruption of RBBP4 function led to dramatic changes in transcriptional profiles. In spite of the PRC2 association, we found that transcriptional changes were more closely tied to the deregulation of H3K27ac rather than H3K27me3 where increased levels of H3K27ac were found on numerous cis-regulatory elements, especially putative enhancers. These data suggest that RBBP4 controls acetylation levels by adjusting the activity of HDAC complexes. As histone methylation and acetylation have been implicated in cancer and neural disease, RBBP4 could serve as a potential target for disease treatment.

the pre-existence of RBBP4 on nucleosomes serves to recruit PRC2 to restore 23 H3K27me3 on newly synthesized histones. As such, disruption of RBBP4 function led to 24 dramatic changes in transcriptional profiles. In spite of the PRC2 association, we found 25 that transcriptional changes were more closely tied to the deregulation of H3K27ac rather 26 than H3K27me3 where increased levels of H3K27ac were found on numerous cis-27 regulatory elements, especially putative enhancers. These data suggest that RBBP4 28 controls acetylation levels by adjusting the activity of HDAC complexes. As histone 29 methylation and acetylation have been implicated in cancer and neural disease, RBBP4 30 could serve as a potential target for disease treatment. 31

Introduction 37
Histone modifications are key epigenetic regulators that specify the transcriptome during 38 lineage determination and tightly control transcriptional states to preserve cellular identity 39 over cell generations (1). Of the various types of histone modifications, methylation and histone H3 (H3K4me) that regulate similar or differing effects on transcription. These data 49 suggest that crosstalk between histone modifications regulates chromatin function (6). 50 Disruption of H3K27 methylation and acetylation results in numerous cellular defects, 51 including cell proliferation and differentiation anomalies during embryogenesis and tissue 52 specification (7-9). Their dysregulation has also been implicated in disease processes 53 such as tumorigenesis (10). 54 55 PRC2, the methyltransferase with specific activity toward H3K27 is responsible for the 56 mono-, di-and tri-methylation modifications of this residue (H3K27me1/me2/me3) (4).

RBBP4
- 3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3 kb  -3 kb  TSS  TES  3   RBBP7 did not affect genomic patterns of H3K27me3 and SUZ12 ( Figure 1D). Together 152 these data signify that RBBP4 is essential for PRC2 binding on target loci for H3K27me3. 153 154 Differential binding analysis revealed 3201 and 1112 genomic loci with decreased 155 H3K27me3 and SUZ12 respectively ( Figure 1E). Despite these differences, the 156 enrichment of mutated RBBP4 was similar to wildtype RBBP4 on PRC2 loci ( Figure 1D, 157 1F), suggesting that mutant RBBP4 remained nucleosome bound. Although global levels 158 of H3K27me3 in mutant cells were similar to control cells, we did not detect redistribution 159 of H3K27me3 and PRC2 on chromatin to compensate for the reduction across gene 160 bodies. For instance, Fancg and Mybl1, two germ cell specific genes, which are inactive 161 but not marked by PRC2 in ESCs did not show increased H3K27me3 accumulation and 162 SUZ12 binding in Rbbp4 mutants ( Figure S1D, S1E). In mouse ESCs, the major 163 population of loci with depleted H3K27me3 are key developmental genes such as the 164 Hox gene clusters ( Figure 1F), which have bivalent domains that also contain active 165 H3K4me2/3 ( Figure S1F). Fewer genomic regions were identified with increased SUZ12 166 and H3K27me3 occupancy in RBBP4 mutants ( Figure 1E). These sites displayed narrow 167 SUZ12 and H3K27me3 enrichment that failed to meet the threshold of commonly used 168 peak caller, MACS2 (Figure S2 We performed endogenous co-immunoprecipitation assays using whole cell extracts 188 to determine the effect of mutant RBBP4 on PRC2 subunit incorporation. Although the 189 enrichment of mutant RBBP4 on SUZ12 target loci was similar to that for wild-type RBBP4 190 ( Figure 2A), SUZ12 pulled down less mutant RBBP4 compared to wild-type forms of 191 RBBP4 ( Figure 2C). This suggests that RBBP4 mutations interfere with binding of PRC2 192 subunits to target genomic regions. However, the incorporation of SUZ12, EED and EZH2 193 into the PRC2 core is not impacted in Rbbp4 mutant ESCs ( Figure 2C), indicating that 194 PRC2 assembly prior to loading onto target loci is independent of RBBP4. 195 196 F L 1 -1 2 6 1 -2 3 9 1 -3 0 6 F L 1 -1 2 6 1 -2 3 9 1 -3 0 6  Figure 2. The N-terminus of RBBP4 is essential for recruiting SUZ12 to chromatin. (A) The distribution and enrichment of RBBP4, K27me3, K27ac and HDAC1 within -5/+5 kb of the center of SUZ12-enriched loci. 1112 genomic regions which have decreased SUZ12 binding in the mutants were included for analysis. In the heatmap, the genomic regions are ordered as those in SUZ12. (B) Co-immunoprecipitation and Western blot analysis mapping the interaction betwee N-terminus of RBBP4 and SUZ12. (C) Co-immunoprecipitation and Western blot analysis for assessing the interaction between endogenous RBBP4 mutants and SUZ12. (D) Western blot analysis on cellular fractions of RBBP4 and SUZ12. GAPDH and UTX serve as non-chromatin and chromatin-bound protein controls, respectively.

Figure 2. The N-terminus of RBBP4 is essential for recruiting SUZ12 to chromatin
Since dramatic reduction of SUZ12 on PRC2 loci in Rbbp4 mutants did not accompany 197 with decreased global levels of SUZ12, we tested whether RBBP4 mutations led to the 198 redistribution of SUZ12 in cellular compartments. Rbbp4 mutant ESCs did not have more 199 SUZ12 in non-chromatin cellular fractions compared to control cells ( Figure 2D)   into corresponding knockout cell lines, H3K27me3 was restored ( Figure 3B) specifically 220 on PRC2 target genes ( Figure 3C). In contrast, inactive genes that are not marked by 221 H3K27me3 in ESCs, such as Fancg and Mybl1, remain deficient in H3K27me3 (Figure 222 3C). These results suggest that PRC2 target loci are not identified by H3K27me3 223 chromatin modifications alone and that PRC2 subunits rely on another factor which reads 224 chromatin and guides the organization of PRC2 subunits to their target loci. 225

226
We and others have found that loss of EED and SUZ12 destabilizes PRC2-core 227 subunits and most ancillary subunits ( Figure 3A) (19). However, RBBP4 and RBBP7 are 228 stable in the absence of EED and SUZ12 ( Figure 3D) and are also not required for the 229 stability of other PRC2 core subunits (24). These suggests that RBBP4/7 and other PRC2 230 components can function as separate units prior to assembly into PRC2 holo-complexes 231 on target loci. To determine if EED and SUZ12 are indispensable for RBBP4 association 232 with chromatin, we performed ChIP-seq analysis to profile the distribution and enrichment 233 of RBBP4 on chromatin in Suz12KO and EedKO ESCs. We found that RBBP4 was able 234 to bind to its target genomic loci in knockout cells, but at a lower level compared to the 235 controls ( Figure 3E). Next, we separately quantified RBBP4 binding on SUZ12-marked 236 PRC2 loci and HDAC complex loci marked by H3K27ac. Similar to the reduction observed 237 for all RBBP4 target loci, we found slightly reduced RBBP4 binding at SUZ12-, but not 238 H3K27ac-marked loci upon EED depletion, and more striking reduction of RBBP4 binding 239 in the absence of SUZ12 ( Figure 3F) Since RBBP4 exists in different chromatin modifying complexes, we clustered binding loci 246 into three groups based on the enrichment patterns of H3K27me3, H3K27ac, HDAC1, 247 and p300 ( Figure 4A). One cluster (cluster_2) showed strong H3K27me3 signals, 248 indicative of them being PRC2 target loci. This cluster also exhibited enrichment of 249 HDAC1 with concomitant depletion of H3K27ac. However, PRC2 and HDAC1-containing 250 complexes did not interact with each other ( Figure S3). We postulated that in these 251 regions, HDAC1 prevented assembly of activating histone acetylation complexes, while 252 H3K27 residues are methylated to maintain a repressive chromatin state. The other two 253 RBBP4 clusters were marked by high and medium levels of H3K27ac and p300 254 respectively, even though they were also enriched with HDAC1 ( Figure    In probing the ability of the histone deacetylase complex members to interact, we found 273 that the incorporation of mutant RBBP4 into HDAC complexes containing HDAC1 and 274 MTA1 was barely detected ( Figure 4E); however the interaction between complex 275 members MTA1 and HDAC1 was maintained in Rbbp4 mutant cells ( Figure S4A). To test 276 whether the reduction of mutant RBBP4 in histone deacetylase complexes was due to its 277 failure to interact with other subunits, we utilized epitope-tagged RBBP4 and HDAC1 for 278 co-immunoprecipitation assays. Similar to wild type RBBP4, mutants were incorporated 279 into HDAC1-containing complexes ( Figure S4B). This suggests that RBBP4 plays 280 complex roles in regulatory histone deacetylation dynamics, and its presence in HDAC-281 containing complexes is important for regulating the activity of the holo-complexes. 282

283
We found protein levels of p300, a histone acetylase specialized for H3K27ac, was 284 strikingly decreased in the mutants ( Figure S4C), which may explain the extensive 285 reduction of H3K27ac levels in mutant cells. However, p300 mRNA levels were only 286 decreased by about 10% in Rbbp4 mutants compared to the control ( Figure S4D), 287 suggesting possible involvement of RBBP4 in stabilizing p300 protein. However, no 288 physical interaction between p300 and RBBP4/HDAC1/MTA1 was detected ( Figure S4C). 289 One future question is based on how the crosstalk between H3K27ac writer and eraser 290 complexes coordinate to regulate acetylation levels.   Unexpectedly, 115 deregulated genes in the Eed knockout were inversely regulated with 323 dramatic changes in Rbbp4 mutants (Table S1) to the acquisition of H3K27ac rather than a loss of H3K27me3. We also observed that 340 Thbs1 was silenced with loss of H3K27ac upstream of its promoter, but not via gaining 341 repressive H3K27me3 ( Figure 5D). Together, these data shows that RBBP4-mediated 342 metabolism of H3K27ac on chromatin plays a key role for controlling gene activity. 343 344 6. RBBP4 regulates gene activity by controlling H3K27 acetylation on enhancers and 345

super-enhancers 346
To explore the mechanisms by which RBBP4 disruption alters H3K27ac to perturb gene 347 expression, we characterized the genomic features of those H3K27ac marked regions. 348 Using the ENCODE mouse cis-regulatory elements database as a reference, we found 349 that both up-and down-regulated H3K27ac loci are significantly enriched with enhancer-350 like signatures (5836 and 7165 respectively) (Table1). Selected loci also cover many 351 promoters and other cis-regulatory elements as defined by DNase hypersensitivity, 352 H3K4me3 and CTCF binding (Table 1). We found RBBP4 associated with sites normally 353 devoid of H3K27me3 and SUZ12 but marked with H3K27ac thereby resembling putative 354 enhancers ( Figure 6A). These sites displayed both increased and decreased H3K27ac   (Table 1). RBBP4 and HDAC1 bind to super-enhancer 372 regions and exhibit similar distribution patterns as H3K27ac ( Figure 6C). Compared to 373 E14 cells, less RBBP4 was bound to super-enhancers in RBBP4 mutants, but this did not 374 affect the binding of HDAC1 at these regions ( Figure 6D). has been shown to be involved in the maintenance of mouse ESC pluripotency (24). We 378 24 focused on Klf5 since it is a regulator of ESC pluripotency (37) and prevents differentiation 379 toward mesoderm (38). Given that Klf5 is important for controlling and defining ESC 380 identity, we predicted that the Klf5 locus and its upstream region, which are broadly 381 marked by H3K27ac (Figure 6E), is a super-enhancer. Here, we observed H3K27ac 382 levels across this region were elevated along with decreased binding of RBBP4 in mutant 383 cell lines ( Figure 6E). Correspondingly, the transcription of Klf5 was enhanced (Figure 384  Regarding PRC2, we found that RBBP4 protein stability is maintained independent of 415 other core subunits, along with its ability to bind PRC2 target chromatin in the absence of 416 EED and SUZ12. We also found that loss of RBBP4 did not affect other PRC2 subunits, 417 confirming that these complex members can independently exist in separate complexes 418 prior to assembly on target chromatin. However, disruption of RBBP4 impairs the 419 recruitment of SUZ12 and EZH2 to PRC2 target loci, leading to a decrease in H3K27me3. 420 Compared to a developmental arrest at the gastrulation stage due to depletion of EED 442 fine-tunes gene expression rather than acting as categorical "on-off" switches (51). As 461 Rbbp4 mutations did not arrest ESC proliferation, it is possible that the mutations did not 462 disrupt the functions of other RBBP4-containing complexes, such as Sin3, which are 463 essential for ESC self-renewal (52). However, it is known that under self-renewal 464 conditions, NuRD binds to a subset of pluripotency genes (Klf4, Klf5, and Tbx3) where it 465 confines their expression (53). Consistent with this, RBBP4 mutant ESCs had numerous 466 upregulated genes, including pluripotency genes, which are surrounded by increased 467 levels of H3K27ac. Altogether, this suggests that RBBP4 helps to maintain pluripotency 468 regulatory networks through regulating NuRD's deacetylation activity on H3K27. 469 28 Regarding crosstalk between histone modifications, it has been established that 470 RBBP4 is involved in the regulation of acetylation and methylation on the same lysine 471 residues of histone H3 for transcriptional repression. The genomic distribution of these 472 two modifications is mutually exclusive and we did not observe interactions between 473 PRC2 and NuRD, indicating that methylation and deacetylation of H3K27 are 474 independent events in gene repression. In support of this, we found only a few 475 overlapping dysregulated genes between Rbbp4 disruption and Eed knockout in ESCs. 476 Compared to ablation of Eed, Suz12, and Ezh2, RBBP4 disruption caused more 477 widespread transcriptional deregulation (24, 32, 33). We also found a subset of genes 478 that were discordantly regulated between Eed and Rbbp4 mutants with respect to loss or 479 gain of H3K27ac. Together, these data suggest that RBBP4 controls gene activity 480 primarily through regulating H3K27ac levels. 481

482
A previous study showed sparse increases of H3K27ac along the genome as a 483 consequence of PRC2 loss, and suggested chromatin hyperacetylation rather than 484 specific loss of repressive control at target genes leads to the early developmental failure 485 induced by PRC2 disruption (54). In mouse ESCs, Hox cluster genes are highly enriched 486 with H3K27me3, but nearly all these genes did not experience transcriptional activation 487 due to loss of H3K27me3. H3K27me3 marks poised chromatin and helps to maintain a 488 transcriptionally silenced state, which is critical for spatiotemporal activation of cell-type 489 specific genes during development. In contrast, H3K27ac is a more potent gene 490 expression activator that actively and dynamically regulates transcriptional levels. This leads us to propose a model for RBBP4's function in regulating methylation and deacetylation/acetylation of H3K27. On one hand, RBBP4 consistently binds to parental histones during DNA replication and guides the assembly of other PRC2 subunits to adjacent newly synthesized histones for maintaining the genomic landscape of H3K27me3 during cell proliferation ( Figure 7A). One the other hand, RBBP4 facilitates the deacetylase activity of HDAC complexes for efficient removal of acetyl group on H3K27. Meanwhile, RBBP4 also promotes H3K27ac through maintaining p300 levels.
Altogether therefore it is possible that RBBP4 assists in controlling H3K27ac levels on cis-regulatory elements for exquisite programming of the transcriptome with regards to cellular context. Better understanding of the molecular mechanisms of RBBP4 will help to elucidate its functional specificities in chromatin regulation and gene expression. In addition to our findings in this study that are relevant to developmental biology, pharmacological interventions to manipulate H3K27me3 and H3K27ac are increasingly explored for cancer therapy. As RBBP4 sits at the intersection of the H3K27 epigenome, it warrants this interest, as it may hold broad implications for chemotherapy applications.

Sample-size estimation
The experiments in this study were performed with cell lines, and, because there was no biological variation to account for, no power analysis was performed in advance.
Also, since all experiments were done using cell-lines, all replication was technical in nature. That said, Western blots used 1 or more technical replicates per cell line. These The cells were treated with 2 μg/ml puromycin for 2 days and recovered in normal culture 32 medium until ESC colonies grew. Targeted colonies were genotyped by PCR and verified by DNA sequencing and Western blot analysis.

ChIP-seq analysis
ChIP-seq was performed as described (57)  Differential binding analysis were performed using CSAW and significant differences in counts were called at a false discovery rate (FDR)≤0.05 (60). The ChIP-seq experiment for each antibody and cell line was performed once.