H3 K27M and EZHIP Impede H3K27-Methylation Spreading by Inhibiting Allosterically Stimulated PRC2 Siddhant U

bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

H3 K27M and EZHIP impede H3K27-methylation spreading by inhibiting allosterically stimulated PRC2 Siddhant U. Jain1, Andrew Q. Rashoff1,#, Samuel D. Krabbenhoft1,#, Dominik Hoelper1, Truman J. Do1, Tyler J. Gibson1, Stefan M Lundgren1, Eliana R. Bondra1, Shriya Deshmukh2, Ashot S. Harutyunyan3, Nikoleta Juretic3, Nada Jabado3, Melissa M. Harrison1, Peter W. Lewis1,*

1Department of Biomolecular Chemistry, School of Medicine and Public Health, University of Wisconsin, Madison, WI 53715, USA 2Department of experimental Medicine, McGill University, Montreal, QC, Canada 3Department of Human Genetics, McGill University, Montreal, QC H3A 1B1, Canada. Department of Pediatrics, 3McGill University, and The Research Institute of the McGill University Health Center, Montreal, QC H4A 3J1, Canada # These authors contributed equally * Corresponding Author

Abstract Diffuse midline gliomas and posterior fossa type-A ependymomas contain the highly recurrent histone H3 K27M mutation and the H3 K27M-mimic EZHIP, respectively. In vitro, H3 K27M and EZHIP are competitive inhibitors of Polycomb Repressive Complex 2 (PRC2) lysine methyltransferase activity. In vivo, these proteins reduce overall H3K27me3 levels, however residual peaks of H3K27me3 remain at CpG islands through an unknown mechanism. Here, we report that EZHIP and H3 K27M preferentially interact with an allosterically activated form of PRC2 in vivo. The formation of H3 K27M- and EZHIP-PRC2 complexes occurs at CpG islands containing H3K27me3 and impedes PRC2 and H3K27me3 spreading. While EZHIP is not found outside of placental mammals, we find that expression of human EZHIP reduces H3K27me3 in Drosophila melanogaster through a conserved molecular mechanism. Our results highlight the mechanistic similarities between EZHIP and H3 K27M in vivo and provide mechanistic insight for the retention of residual H3K27me3 in tumors driven by these oncogenes.

Diffuse midline gliomas (DMGs) are highly aggressive pediatric tumors with a poor prognosis. About ~80% of DMGs contain the recurrent histone H3 lysine 27 (H3 K27M) mutation 1,2. Despite representing a small fraction of total histone H3 (3-17%), H3 K27M causes a global reduction of H3 lysine 27 trimethylation (H3K27me3) in DMGs 3-8. The K27M mutation occurs in genes encoding histone H3 proteins that are assembled into nucleosomes through replication-coupled (H3.1/2) and replication-independent mechanisms (H3.3). Posterior fossa type-A ependymomas (PFA ependymomas) display similar gene expression, DNA methylation, and low H3K27me3 profiles as DMGs 9. However, instead of recurrent H3 K27M mutations, PFA ependymomas aberrantly express a newly discovered gene, EZHIP (also known as CXorf67, CATACOMB) 10. EZHIP-expressing ependymomas are generally found in young children (median age 3 years) and display poorer prognosis as compared to other posterior fossa ependymomas 11. While numerous studies have described the biochemical activities of H3 K27M oncohistones, their precise role and function in modulating gene expression remains unclear. Moreover, the molecular mechanisms of EZHIP in cells remains largely uncharacterized. Polycomb Repressive Complex 2 (PRC2) catalyzes H3K27me3, which is involved in transcriptional silencing. The oncoproteins EZHIP and H3 K27M are competitive inhibitors of PRC2 12-14. A C-terminal peptide within EZHIP mimics the H3 K27M sequence and is necessary and sufficient to inhibit PRC2 activity in vitro. Expression of EZHIP or H3 K27M in cells leads to global reduction of H3K27me2/3 12-15. However, hundreds of CpG islands (CGIs), that represent high-affinity PRC2-binding sites retain residual H3K27me3 12,16-18. Residual PRC2 activity is necessary for survival of H3 K27M-containing gliomas 16-18. It has been proposed that retention of H3K27me3 at CGIs is necessary to silence tumor suppressor genes and maintain cell proliferation. Nonetheless, the molecular mechanism by which EZHIP and H3 K27M reduce H3K27me3 specifically from intergenic regions but disproportionately retain H3K27me3 at CGIs remains elusive. PRC2 interacts with unmethylated CGIs through its auxiliary subunits, Polycomb-like proteins (PCLs) or JARID2, where it catalyzes high levels of H3K27me3 19-25. H3K27me3, initially catalyzed at CGIs, interacts with the EED subunit and allosterically stimulates PRC2 catalytic activity by ~8-fold 26,27. In addition to H3K27me3, trimethylation of K116 in the JARID2 subunit can also stimulate EZH2 activity in an EED-dependent manner 28. This ‘read-write’ mechanism triggers PRC2 spreading into the intergenic regions and formation of broad H3K27me3 domains. Remarkably, the inhibitory potential of H3 K27M and EZHIP oncoproteins is substantially enhanced by allosteric stimulation of PRC2 in vitro 12,29-31. It is unclear whether the preferential inhibition of allosterically stimulated PRC2 by EZHIP and H3 K27M plays a functional role in vivo. bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Here, we demonstrate that EZHIP preferentially interacts with allosterically stimulated PRC2 in vitro and in vivo. EZHIP impedes PRC2 spreading by stabilizing a high-affinity complex between H3K27me3-PRC2-EZHIP at CGIs containing residual H3K27me3. Using reChIP experiments, we demonstrate that H3 K27M oncohistones interact with and stall PRC2 spreading at CGIs. Finally, despite its absence in non-placental mammals, we demonstrate that EZHIP reduces H3K27me3 in Drosophila melanogaster through a conserved molecular mechanism. In summary, we provide evidence that EZHIP and H3 K27M bind PRC2 in vivo and reduce H3K27me2/3 in trans by blocking PRC2 spreading.

Results: EZHIP preferentially interacts with allosterically stimulated PRC2 in vivo Expression of EZHIP in cells leads to an overall reduction of H3K27me3, however residual H3K27me3 is retained at CGIs (Jain et al., 2019). The mechanism by which these narrow H3K27me3 peaks persists in tumors expressing EZHIP remains elusive. Therefore, we used ChIP-Sequencing (ChIP-seq) to identify the genomic regions where EZHIP inhibits PRC2 in mouse embryonic fibroblasts expressing EZHIP. Surprisingly, we found that EZHIP and EZH2 colocalized with residual H3K27me3 at CGIs (Figure 1A-D). Depletion of EED through transient expression of Cre-recombinase abolished EZHIP enrichment suggesting that EZHIP binds to chromatin indirectly through its interaction with PRC2 (Figure 1A, B; S1A, B). A K27M-like peptide (KLP) within the C-terminus of EZHIP interacts with EZH2 catalytic site residues (where M406 is equivalent to H3 K27M) 12. Consequently, EZHIP M406E mutant within the KLP failed to inhibit PRC2 in vitro and did not immunoprecipitate PRC2 subunits from nuclear extract (Figure 1E, F; S1A). Moreover, we did not identify enrichment of EZHIP M406E at CGIs with PRC2 and H3K27me3 in vivo (Figure 1G; S1C). These data linked in vitro PRC2 inhibition and in vivo reduction of H3K27me3 to interaction between EZHIP KLP and EZH2 at CGIs. Previously, we showed that the PRC2 inhibitory potential of EZHIP is significantly enhanced in the presence of H3K27me3 peptide 12. The interaction between EED and H3K27me3 is proposed to induce or stabilize an EZH2 conformation that has increased affinity towards the peptide substrates and inhibitors, such as EZHIP and H3 K27M (Figure 1H). We assessed EZHIP- PRC2 interactions using FLAG affinity purification in the presence or absence of a H3K27me3- competitive inhibitor of EED, A-395 32. Cells treated with A-395 exhibit low H3K27me3 levels due to reduced allosteric activation of PRC2 32. We found that EZHIP immunoprecipitated substantially lower amounts of PRC2 subunits from MEF nuclear extracts in the presence of A-395 (Figure 1I). These data suggest that EZHIP preferentially interacts with allosterically stimulated PRC2 and bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

help explain why the inhibitory protein EZHIP interacts with PRC2 at regions containing H3K27me3 in vivo.

EZHIP reduces PRC2 spreading by stalling it at CpG islands We hypothesize that the formation of a catalytically inactive ternary complex, H3K27me3-PRC2- EZHIP, restrains PRC2 from spreading into the intergenic regions and hence, increases PRC2 residency at CGIs containing H3K27me3 (Figure 2A; S2A). Using quantitative ChIP-Seq, we found that expression of wildtype EZHIP, but not the PRC2-binding deficient M406E mutant, led to a substantial increase of EZH2 occupancy at residual H3K27me3 sites containing high CpG density (Figure 2B, C; S2B-E). Simultaneously, EZH2 enrichment was significantly reduced from CpG-poor domains of H3K27me3 (Figure S2C, D), suggesting that EZHIP stalls PRC2 at CGIs that serve as PRC2 recruitment sites. To test our model in human cancers, we used SUZ12 CUT&RUN data from U2OS osteosarcoma cells that express endogenous EZHIP 15. Mirroring our results in MEFs expressing EZHIP, wildtype U2OS cells exhibit sharp SUZ12 peaks, but these peaks redistributed to broader intergenic domains upon loss of EZHIP (Figure 2D, E; S2F, G). These results indicate that EZHIP impedes PRC2 spreading by stalling it at recruitment sites.

H3 K27M mutations interact with and stall PRC2 at CpG island EZHIP inhibits PRC2 through a mechanism that is similar to H3 K27M oncohistones and leads to PRC2 redistribution from intergenic regions to CGIs by interacting with allosterically stimulated PRC2. Nevertheless, whether H3 K27M directly interacts with PRC2 in vivo at steady state has remained controversial because H3 K27M and PRC2 occupancies do not correlate positively in cells 17,18,30. The amount of H3 K27M far exceeds that of PRC2 in the nucleus 30. H3 K27M, being histones, are incorporated into nucleosomes and are therefore present throughout the genome. In contrast, PRC2 is localized in relatively narrow peaks at CGIs, and this may explain poor correlation between the enrichment of H3 K27M and PRC2. Previously, we and others have found that H3 K27M preferentially binds and inhibits PRC2 in the presence of H3K27me3, similar to EZHIP 12,29-31. Therefore, we hypothesize that H3 K27M also stalls PRC2 at CGIs containing residual H3K27me3 by binding allosterically stimulated PRC2. To directly map genomic regions where H3 K27M interacts with PRC2 in cells, we used H3 K27M ChIP followed by EZH2 reChIP-Seq (Figure 3A; S3A, B). We selected reads with fragment size smaller than 400 bp for our analyses to only capture PRC2 bound to mono- or di- nucleosomes. As would be predicted from our model, H3 K27M-bound EZH2 was enriched at CGIs containing residual H3K27me3 (Figure 3B-E; S3C, D). We did not detect EZH2 reChIP enrichment in control cells that did not express a FLAG-tagged H3 K27M transgene, confirming bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

our detection of only H3 K27M-bound EZH2 instead of a background signal (Figure S3C). These data suggest that H3.1 and H3.3 K27M directly interact with PRC2 at CGIs, similar to EZHIP. Next, we examined changes in PRC2 distribution in cells expressing H3 K27M by mapping overall EZH2 binding profile. Similar to cells expressing EZHIP, expression of H3.1 or H3.3 K27M led to a significant increase in EZH2 occupancy at CGIs containing residual H3K27me3 and a concurrent reduction of EZH2 from non-CpG-rich regions (Figure 3F; S3F-K). To corroborate our model in gliomas, we profiled genomic distribution of SUZ12 in patient-derived DMG cell lines with wildtype H3 or H3 K27M mutations 18. DMGs lines containing the H3 K27M mutation had significantly higher SUZ12 enrichment at CGIs relative to H3 wildtype glioma lines (Figure 3G; S4A-D). Importantly, Cas9-mediated genetic ablation of H3.3 K27M substantially reduced SUZ12 occupancy at these sites (Figure 3H, S4E-K). Taken together, our results suggest that H3 K27M directly interacts with PRC2 in vivo and stalls it at CGIs containing residual H3K27me3.

Mammalian EZHIP inhibits Drosophila PRC2 through a conserved mechanism Our studies demonstrate remarkable similarities in the mechanism through which EZHIP and H3 K27M inhibit PRC2 through directly interacting with it at its recruitment sites. While histone H3 is highly conserved among eukaryotes, EZHIP is only present in placental mammals. Previous studies have found that expression of H3 K27M in fruit flies largely phenocopies loss of PRC2 activity 33. Since EZHIP mimics the molecular function of H3 K27M oncohistone, we hypothesized that human EZHIP may inhibit Drosophila PRC2 despite its evolutionary absence in flies. Indeed, we found that expression of human EZHIP or H3.3 K27M in imaginal wing discs led to substantial reduction of H3K27me3, relative to EZHIP M406E or H3 K27R controls (Figure 4A). These data further highlight the similarity between the molecular functions of the two oncogenes. The cis-regulatory elements involved in PRC2 recruitment that account for the global H3K27 methylation profile have not been identified in mammals. However, PRC2 is recruited to Polycomb Response Elements (PREs) through its Pho subunit in Drosophila melanogaster 34,35. Therefore, Drosophila presents an excellent model to study and validate PRC2 recruitment versus spreading defects mediated by EZHIP. Having validated the ability of EZHIP to inhibit H3K27me3 in vivo, we established a copper-inducible system to express EZHIP in Drosophila S2 cells and showed that expression of EZHIP led to a dose-dependent reduction of H3K27me3, consistent with a competitive mode of inhibition with respect to histone substrates (Figure 4B, C). Similar to EZHIP in mammalian cells, ChIP-seq for H3K27me3 identified regions that retained residual levels of H3K27me3 upon EZHIP induction in S2 cells (Figure 4C; S5A, B). Furthermore, majority of the PREs retained residual H3K27me3 in cells expressing EZHIP (Figure 4C, S5B). bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

To systematically distinguish between PRC2 recruitment versus spreading defects, we identified genomic regions that disproportionately retained H3K27me3 relative to global H3K27me3 (Figure 4D). Because PREs are characterized by being nucleosome-free regions, we used ATAC-Seq to identify potential PRC2 recruitment sites within the residual H3K27me3 regions. We found that 178/326 (55%) of the regions that retained H3K27me3 contained previously annotated PREs and displayed polyhomeotic (PH) enrichment (Figure 4E-G, S5C, D). Moreover, an additional 111 (34%) sites also displayed PH occupancy, accounting for ~90% of regions that retained H3K27me3 in cells expressing EZHIP (Figure 4E-F). Notably, the amplitude of PH enrichment at unannotated PH binding sites was lower as compared to annotated PREs (Figure 4G), likely representing cell-type specific, weaker PRC2-recruitment sites 36. These results further support the model that EZHIP preferentially inhibits PRC2 spreading. Moreover, this model provides a mechanistic explanation for the presence of sharp H3K27me3 peaks in DMGs and PFA ependymomas at genes containing CGIs including the tumor suppressor, CDKN2A.

Discussion In this study, we used genomic approaches to demonstrate that glioma-driving oncoproteins EZHIP and H3 K27M directly interact with PRC2 in vivo and reduce H3K27me3 in trans. Due to the broad chromatin deposition profile of histone H3, it had been challenging to assess PRC2-H3 K27M interaction in vivo using such approaches. We show that the non-histone, H3 K27M-mimic EZHIP occupies the same sites as PRC2 in vivo. Importantly, we successfully detected H3 K27M- PRC2 interaction in vivo using a sequential ChIP strategy. These data support the model that H3 K27M inhibits PRC2 activity in trans and link the numerous studies that characterized the PRC2- K27M interactions in vitro to the in vivo loss of H3K27me3. Previous studies revealed that H3 K27M and EZHIP have substantially higher affinity for allosterically stimulated PRC2. In the complete absence of H3K27me3, EZHIP and H3 K27M appear to be poor inhibitors of PRC2 and hence, allow initial H3K27me3 at high-affinity recruitment sites 37,38. However, these oncoproteins form high-affinity complexes with H3K27me3- bound PRC2 at these recruitment sites to prevent further catalysis of H3K27me3 at distal sites. As a result, PRC2-recruitment sites disproportionately retain H3K27me3 at equilibrium in these tumors. We propose that tumors expressing EZHIP or H3 K27M retain residual H3K27me3 at PRC2 recruitment sites because of their preferential interaction with allosterically stimulated PRC2. We found that EZHIP and H3 K27M bind to allosterically stimulated PRC2 at CGIs containing residual H3K27me3. While we focused on H3K27me3-stimulated PRC2 in this study, it is known that Jarid2 K116me3 also stimulates PRC2 by interacting with EED. Additionally, bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

Jarid2 serves to target PRC2 at its recruitment sites through an independent mechanism 19,20,22,25,28,39. Therefore, enhanced binding of EZHIP and H3 K27M to PRC2 stimulated by Jarid2- K116me3 may also contribute to increased PRC2 residence at CGIs and stalling its spread. Recurrent H3 K27M mutations and aberrant EZHIP expression are frequently found in DMGs and PFA ependymomas, respectively. Our study suggests that EZHIP and H3 K27M inhibit PRC2 activity in trans, through a similar molecular mechanism in vitro and in vivo. Remarkably, two recent studies discovered aberrant expression of EZHIP in a subset of DMGs lacking H3 K27M mutations 40,41. Similarly, a small fraction of PFA ependymomas contain H3 K27M mutations that are mutually exclusively with EZHIP expression 10. Our finding that EZHIP and H3 K27M have similar underlying biochemical mechanisms are consistent with the clinical observations that both oncoproteins appear to drive the same subtype of gliomas. Moreover, pharmacological interventions that have been proposed for H3 K27M-positive gliomas might be promising candidates in gliomas expressing EZHIP 8,16,17,42-44. We validated our findings that EZHIP disproportionately blocks PRC2 spreading while sparing residual H3K27me3 at recruitment sites using Drosophila melanogaster S2 cells. Half of the sites that disproportionately retained H3K27me3 were previously characterized cis-acting PREs. However, we also found more than one hundred new, weak PRC2-binding sites that likely represent tissue-specific PREs. Previous studies used a combination of H3K27me3 ChIP- and ATAC- Seq to identify PREs in Drosophila melanogaster. Expression of EZHIP might provide a tool to filter out majority of the genomic regions containing H3K27me3 and a more sensitive method to detect tissue specific PREs in future studies.

Methods Transgenic cell lines and culture Mouse embryonic fibroblasts used in this study containing loxP sites flanking exon 3-6 of EED were described previously 12. Cells were cultured in DMEM supplemented with 10% FBS, 1x glutamax and 1x penicillin-streptomycin. Lentiviruses were produces by co-transfecting packaging vectors (psPAX2 and pMD2.G) and transfer vector (pCDH-EF1a-MCS-PuroR) in HEK- 293T cells. Cells were transduced with lentiviruses for 2 days and selected using 1.5 µg/µl puromycin for 4 days. Mouse and human EZHIP DNA sequences were used in mouse and human cell lines respectively; however, only human amino acid numbers were used to avoid confusion. S2 cells in this study were cultured in Schneider’s Media (Thermo Fisher) containing 10% FBS (Omega Scientific), and 1% antibiotic/antimycotic (Thermo Fisher). Human EZHIP and H3.3 genes was cloned into the pMT-puro vector for copper-inducible expression with the bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

metallothionein (MT) promoter. Transfections were performed with 2 µg plasmid DNA, using Effectene Transfection Reagent for random genome integration, and selected using 2 µg/ml puromycin for approximately three weeks.

Fly Stocks All stocks were grown on molasses food at 22°C (room temperature). N-terminally FLAG-tagged EZHIP, EZHP(M406E), H3K27M or H3K27R were cloned into pUASt-attB (DGRC#1419) and integrated into ZH-86Fb on the third chromosome using PhiC31 integrase-mediated recombination into BDSC#24749 with fluorescence marker removed (Best Gene). en-Gal4, UAS- RFP/CyO (II) (BDSC#30557) was used to drive expression in the larval wing disc.

Immunohistochemistry Crawling third instar larvae were harvested and dissected in pre-chilled (4ºC) 1X PBS. Wing imaginal discs were removed from larvae and placed into 1X PBS on ice and fixed in 4% formaldehyde for 30 minutes. Fixed wing imaginal discs were washed in 1X PBS + 0.1% Triton X-100, (PBST) and blocked in PBST + 1% BSA (PAT). After removal of PAT, wing discs were resuspended in PAT + anti-H3K27me3 (1:1600) (Cell Signaling Technology #9733S) and incubated overnight at 4ºC, washed in PBST, incubated in PBST + 2% normal goat serum for 10 minutes, followed by incubation with PBST + 2% normal goat serum and goat anti-rabbit DyLight 488 conjugated secondary antibody (1:2000) (Fisher Scientific #35552). Larvae were imaged at 10X using a Nikon Ti2-E epifluorescent microscope.

Production of stable S2 cell lines for inducible EZHIP expression S2 cells in this study were cultured at 25°C in Schneider’s Media (Thermo Fisher) containing 10% FBS (Omega Scientific), and 1% antibiotic/antimycotic (Thermo Fisher). FLAG-tagged human EZHIP was cloned into the pMT-puro vector (Addgene #17923). Transfections were performed with 2 µg plasmid DNA, using Effectene Transfection Reagent (Qiagen), and cells were selected using 2 µg/ml puromycin for approximately three weeks. For induction, 5 µM or 10 µM copper sulfate was added to cells at one million cells/ ml density. Cells were incubated for 72 hours and harvested for immunoblot or ChIP.

FLAG affinity Purification ~80 million cells were homogenized in hypotonic lysis buffer (15 mM HEPES pH 7.9, 4 mM MgCl2, 10 mM KCl, 1 mM EDTA, 8 mM PMSF) to isolate nuclei. Nuclei were resuspended in Buffer-M bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

(15 mM HEPES pH 7.9, 1 mM CaCl2, 30 mM KCl, 1X protease inhibitor cocktail, 8 mM PMSF, 1 mM beta-mercaptoethanol) and treated with 750 units of MNase for 20 min at 37 ºC. MNase digestion was quenched and nuclear extract was prepared by adding 10 mM EDTA, 5 mM EGTA, 270 mM KCl, 0.05% Triton X-100). Nuclear extract was incubated with 75 µl of M2 anti-FLAG affinity gel (Sigma A2220) for 2 hours. Beads were washed 5-times with wash buffer (15 mM HEPES pH 7.9, 500 mM KCl, 1 mM EDTA, 0.05% Triton X-100, 8 mM PMSF) and captured proteins were eluted using 300 µg/ml of 3x FLAG peptides. For FLAG affinity purification in the presence of A-395, 1 µM A-395 (or DMSO control) was added to cultured cells for 6 hours before cells were harvested and nuclear extract was prepared. 1 µM A-395 (or DMSO) was added to all buffers throughout the protocol.

Immunoprecipitation of pre-deposition complexes Lysate from 40 × 106 HEK-293T cells transduced with H3.3-FLAG-HA transgenes was prepared by resuspension in 3.0 ml lysis buffer [20 mM HEPES pH 7.9, 200 mM KCl, 0.5 mM EDTA, 2 mM MgCl2, 0.2 % Triton X-100, 2× Protease Inhibitor Cocktail (Roche), 2 mM 2-mercaptoethanol, 1 mM benzamidine, 0.4 mM PMSF, 300 μM S-adenosyl methionine), followed by douncing and separation of the insoluble fraction by centrifugation. Per sample, 30 μl of packed anti-FLAG M2 beads (Sigma) were added to the lysate and incubated rotating at 4 °C for 2 h. Beads were transferred onto microspin columns (Enzymax) and washed three times with wash buffer (20 mM HEPES pH 7.9, 300 mM KCl, 1 mM EDTA, 0.12 % Triton X-100, 0.4 mM PMSF, 1 mM benzamidine, 150 μM S-adenosyl methionine) for 5 min. Finally, samples were eluted with 2 × 25 μl elution buffer [wash buffer supplemented with 500 ng μl-1 3×FLAG peptide (Tufts University Peptide Core Facility)] via incubation for 5 min on ice and centrifugation at 300 g.

Histone Methyltransferase Assays 200 nM oligonucleosome arrays or 25 µM H3 peptide (18-32) substrates were incubated with 20 nM recombinant PRC2 complex, 4 µM S-adenosyl Methionine (1 µM 3H-SAM; 3 µM cold SAM) and 20 µM H3K27me3 stimulatory peptide in 25 mM Tris pH 8.0, 2 mM MgSO4, 5 mM DTT, 0.4 mM PMSF for 90 min. Reaction was spotted on phosphocellulose membrane (Whatman p81) and

dried for 10 min. Filters were washed 3-times with 100 mM NaHCO3 for 5 min each, rinsed in acetone and dried for 10 min. Scintillation counting was performed using Tri-Carb 2910 TR liquid Scintillation analyzer (Perkin Elmer). For fluorography, reaction was resolved on 15% SDS-PAGE gel, stained with Coomassie, incubated in Amersham Amplify Fluorography reagent (GE bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

healthcare) for 10 min and dried under vacuum. Films capturing fluorography signal were developed after 24-48 hours. Experiment specific details are in figure legend.

Chromatin Immunoprecipitation ~20 million mammalian cells or ~80 million S2 cells were cross-linked with 0.8% paraformaldehyde for 8 min at room temperature and quenched with 0.2 M glycine. Cells were lysed by resuspending in lysis buffer (50 mM HEPES pH 7.9, 140 mL NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP40, 0.25% Triton X-100, 0.8 mM PMSF). Nuclei were washed once and resuspended in digestion buffer (50 mM HEPES pH 7.9, 1 mM CaCl2, 20 mM NaCl, 1x protease inhibitor cocktail, and 0.8 mM PMSF), and treated with 200 units of MNase for 10 min. Reaction was quenched by adding 10 mM EDTA, 5 mM EGTA, 80 mM NaCl, 0.1% sodium deoxycholate, 0.5% N-lauroyl sarcosine. Mono-nucleosomes were solubilized by sonication using covaries S220 (160 peak incidental power, 5% duty factor, 200 cycles/burst, 45” ON-30” OFF) 3-times. 1% Triton X-100 was added to the chromatin and insoluble chromatin was removed using centrifugation. Chromatin was dialyzed against RIPA buffer (10 mM Tris pH 7.6, 1 mM EDTA, 0.1% SDS, 0.1% sodium deoxycholate, 1% Triton X-100) for 2 hours. Chromatin concentration was measured using qubit and spike-in chromatin was added at 1:40 ratio. Chromatin was incubated with primary antibodies overnight. Antibodies was captured using Dynabeads for 4 hours and washed 3x using RIPA buffer, 2x using RIPA-NaCl and 2x with LiCl buffer. Chromatin was eluted in 10 mM Tris, 1 mM EDTA, and 1% SDS, incubated with proteinaseK, and RNaseA and DNA was purified using PCR purification columns. For reChIP, chromatin was eluted using a competing 3x-FLAG peptide for 30 min at 10ºC. Eluate was diluted 10-times with RIPA buffer and incubated with EZH2 antibody overnight. Final washes and elution were carried out as described for conventional one- step ChIPs. Eluted DNA was diluted 1:50 for qPCR analysis. Sequencing libraries were prepared using NEB Next Ultra kit. ChIPs were performed in at least two independent replicates with similar results, at least one replicate was sequenced using NGS, both replicates were used for qPCR but data in the figures only represent technical replicates; ChIP-qPCR results are displayed as barchart (mean ± SE), p-values were determined by paired, non-parametric t-test.

ChIP-Sequencing analysis Reads that passed quality score were aligned to mouse (mm9) or human (hg19) or drosophila (dm6) genomes using bowtie 1 with default parameters. Sample normalization factor was determined as Rx = 10^6 / (total reads aligned to exogenous reference genome). Sam files were converted to bam files using samtools. Bigwig files were generated using deeptools and Rx scale- bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

factor was used for sample normalization. Residual H3K27me3 sites in cells expressing EZHIP or H3 K27M were determined as peaks found in two independent ChIP-Seq experiments. Spreading sites were determined by subtracting EZH2 peaks from broad H3K27me3 peaks present in control samples (EZHIP M406E or H3 K27R). For EZH2 reChIP-Seq analyses, aligned paired-end sequencing reads were filtered by fragment length < 400 bp. Samples were normalized by RPKM and corresponding FLAG ChIP (H3 K27M ChIP). EZH2 reChIP in MEFs not expressing H3 K27M transgene was used as negative control. No conclusion about the amplitude of reChIP signals were drawn in the manuscript. Deeptools was used for data visualization. Peaks were called using mosaics-HMM (typically using FDR= 0.01, maxgap= 2-10K, minsize= 1K). Statistical analysis was performed using R. For identification of potential PRC2 binding sites in S2 cells, regions containing disproportionate retention of H3K27me3 were determined as the difference in H3K27me3 RPKM enrichment in control and cells expressing EZHIP. Bins with change in H3K27me3 < 10 within 5KB were merged and regions with delH3K27me3 < 500 were removed. Finally, ATAC-Seq peaks within these regions with residual H3K27me3 were defined as potential PRC2 recruitment sites. Annotations of PREs in dm6 genome were obtained from a recent report 45.

ATAC-Sequencing 2 × 105 Drosophila S2 cells were washed once with 1X PBS and then resuspended in 100 µL ATAC lysis buffer (10mM Tris 7.5, 10mM NaCl, 3mM MgCl2, 0.1% NP-40). Cells were centrifuged at 600 × g for 10 minutes at 4°C. The resulting pellet was resuspended in 47.5 µL buffer TD (Illumina 15027866) before adding 2.5 µL Tn5 transposase (Tagment DNA Enzyme, Illumina 15027865) and incubating in 37°C water bath for 30 minutes. The tagmented DNA was immediately purified using MinElute Cleanup Kit (Qiagen 28204) and eluted in 10 µL buffer EB. Tagmented DNA was amplified with 12 cycles of PCR using the NEBNext Hi-Fi 2X PCR Master Mix (NEB M0541) and unique dual index primers. Libraries were purified using a 1.2X ratio of Axygen magnetic beads. 150bp, paire-end sequencing was performed at the University of Wisconsin-Madison Biotechnology Center on the Illumina Nova Seq 6000 platform.

ATAC-Sequencing analysis Raw reads were trimmed to remove adapter sequences using NGmerge (Gaspar, 2018, BMC Bioinformatics). Trimmed reads were aligned to the Drosophila (dm6) genome using bowtie2 with the following parameters: --very-sensitive, --no-mixed, --no-discordant, -X 5000, -k 2. Only reads with a mapping quality score > 30 that aligned to major chromosomes (2, 3, 4, X, Y) were retained for downstream analysis. In order to enrich for fragments originating from nucleosome-free bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

regions, only fragments < 100 bp were retained. Peak calling was performed on accessible fragments using MACS2 with the following parameters: -f BAMPE --keep-dup all -g 1.2e8 --call- summits.

Acknowledgments This research was supported by funding from P01CA196539 (to P.W.L., T.W.M., and N.J.); the Greater Milwaukee Foundation (to P.W.L.), the Sidney Kimmel Foundation (Kimmel Scholar Award to P.W.L.), a startup provided by the Wisconsin Institute for Discovery (to P.W.L.). A.B. is supported by Fonds de Recherche du Québec-Santé. This work was performed within the context of the I-CHANGE consortium and supported by funding from Genome Canada, Genome Quebec, The Institute for Cancer Research of the Canadian Institutes for Health Research (CIHR), McGill University and the Montreal Children’s Hospital Foundation. N.J. is a member of the Penny Cole lab and the recipient of a Chercheur Clinician Senior Award.

Author Contribution S.U.J and P.W.L conceptualized the study and wrote the manuscript. S.U.J, A.Q.R., S.D.K., D.H., T.J.D., T.J.G., S.M.L. and E.R.B performed the experiments. S.D., A.S.H. and N.J. helped with NGS samples. N.J., M.M.H, and P.W.L. supervised the experiments. All authors read and edited the manuscript.

Declaration of Interests Authors declare no conflict of interest.

References 1. Schwartzentruber, J. et al. Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature 482, 226–231 (2012). 2. Wu, G. et al. Somatic histone H3 alterations in pediatric diffuse intrinsic pontine gliomas and non-brainstem glioblastomas. Nat Genet 44, 251–253 (2012). 3. Lewis, P. W. et al. Inhibition of PRC2 activity by a gain-of-function H3 mutation found in pediatric glioblastoma. Science 340, 857–861 (2013). 4. Chan, K.-M. et al. The histone H3.3K27M mutation in pediatric glioma reprograms H3K27 methylation and gene expression. Genes & Development 27, 985–990 (2013). 5. Bender, S. et al. Reduced H3K27me3 and DNA hypomethylation are major drivers of gene expression in K27M mutant pediatric high-grade gliomas. Cancer Cell 24, 660–672 (2013). 6. Larson, J. D. et al. Histone H3.3 K27M Accelerates Spontaneous Brainstem Glioma and Drives Restricted Changes in Bivalent Gene Expression. Cancer Cell 35, 140–155.e7 (2019). 7. Silveira, A. B. et al. H3.3 K27M depletion increases differentiation and extends latency of diffuse intrinsic pontine glioma growth in vivo. Acta Neuropathol 137, 637–655 (2019). bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

8. Anastas, J. N. et al. Re-programing Chromatin with a Bifunctional LSD1/HDAC Inhibitor Induces Therapeutic Differentiation in DIPG. Cancer Cell 36, 528–544.e10 (2019). 9. Bayliss, J. et al. Lowered H3K27me3 and DNA hypomethylation define poorly prognostic pediatric posterior fossa ependymomas. Sci Transl Med 8, 366ra161–366ra161 (2016). 10. Pajtler, K. W. et al. Molecular heterogeneity and CXorf67 alterations in posterior fossa group A (PFA) ependymomas. Acta Neuropathol 136, 211–226 (2018). 11. Pajtler, K. W. et al. Molecular Classification of Ependymal Tumors across All CNS Compartments, Histopathological Grades, and Age Groups. Cancer Cell 27, 728–743 (2015). 12. Jain, S. U. et al. PFA ependymoma-associated protein EZHIP inhibits PRC2 activity through a H3 K27M-like mechanism. Nat Commun 10, 2146 (2019). 13. Hübner, J.-M. et al. EZHIP / CXorf67 mimics K27M mutated oncohistones and functions as an intrinsic inhibitor of PRC2 function in aggressive posterior fossa ependymoma. Neuro-Oncology 21, 878–889 (2019). 14. Piunti, A. et al. CATACOMB: An endogenous inducible gene that antagonizes H3K27 methylation activity of Polycomb repressive complex 2 via an H3K27M-like mechanism. Sci Adv 5, eaax2887 (2019). 15. Ragazzini, R. et al. EZHIP constrains Polycomb Repressive Complex 2 activity in germ cells. 42, 531–52 (2019). 16. Mohammad, F. et al. EZH2 is a potential therapeutic target for H3K27M-mutant pediatric gliomas. Nat. Med. 23, 483–492 (2017). 17. Piunti, A. et al. Therapeutic targeting of polycomb and BET bromodomain proteins in diffuse intrinsic pontine gliomas. Nat. Med. 23, 493–500 (2017). 18. Harutyunyan, A. S. et al. H3K27M induces defective chromatin spread of PRC2-mediated repressive H3K27me2/me3 and is essential for glioma tumorigenesis. Nat Commun 1–13 (2019). doi:10.1038/s41467-019-09140-x 19. Højfeldt, J. W. et al. Non-core Subunits of the PRC2 Complex Are Collectively Required for Its Target-Site Specificity. Molecular Cell 76, 423–436.e3 (2019). 20. Healy, E. et al. PRC2.1 and PRC2.2 Synergize to Coordinate H3K27 Trimethylation. Molecular Cell 76, 437–452.e6 (2019). 21. Li, H. et al. Polycomb-like proteins link the PRC2 complex to CpG islands. Nature 549, 287–291 (2017). 22. Pasini, D. et al. JARID2 regulates binding of the Polycomb repressive complex 2 to target genes in ES cells. Nature 464, 306–310 (2010). 23. Peng, J. C. et al. Jarid2/Jumonji coordinates control of PRC2 enzymatic activity and target gene occupancy in pluripotent cells. Cell 139, 1290–1302 (2009). 24. Li, G. et al. Jarid2 and PRC2, partners in regulating gene expression. Genes & Development 24, 368–380 (2010). 25. Son, J., Shen, S. S., Margueron, R. & Reinberg, D. Nucleosome-binding activities within JARID2 and EZH1 regulate the function of PRC2 on chromatin. Genes & Development 27, 2663–2677 (2013). 26. Hansen, K. H. et al. A model for transmission of the H3K27me3 epigenetic mark. Nat. Cell Biol. 10, 1291–1300 (2008). 27. Margueron, R. et al. Role of the polycomb protein EED in the propagation of repressive histone marks. Nature 461, 762–767 (2009). 28. Sanulli, S. et al. Jarid2 Methylation via the PRC2 Complex Regulates H3K27me3 Deposition during Cell Differentiation. Molecular Cell 57, 769–783 (2015). 29. Justin, N. et al. Structural basis of oncogenic histone H3K27M inhibition of human polycomb repressive complex 2. Nat Commun 7, 11316 (2016). 30. Stafford, J. M. et al. Multiple modes of PRC2 inhibition elicit global chromatin alterations in H3K27M pediatric glioma. Sci Adv 4, eaau5935 (2018). bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

31. Diehl, K. L. et al. PRC2 engages a bivalent H3K27M-H3K27me3 dinucleosome inhibitor. Proc. Natl. Acad. Sci. U.S.A. 116, 22152–22157 (2019). 32. He, Y. et al. The EED protein-protein interaction inhibitor A-395 inactivates the PRC2 complex. Nat. Chem. Biol. 13, 389–395 (2017). 33. Herz, H.-M. et al. Histone H3 lysine-to-methionine mutants as a paradigm to study chromatin signaling. Science 345, 1065–1070 (2014). 34. Orsi, G. A. et al. High-resolution mapping defines the cooperative architecture of Polycomb response elements. Genome Res. 24, 809–820 (2014). 35. Müller, J. & Kassis, J. A. Polycomb response elements and targeting of Polycomb group proteins in Drosophila. Curr. Opin. Genet. Dev. 16, 476–484 (2006). 36. De, S., Mitra, A., Cheng, Y., Pfeifer, K. & Kassis, J. A. Formation of a Polycomb-Domain in the Absence of Strong Polycomb Response Elements. PLoS Genet 12, e1006200 (2016). 37. Højfeldt, J. W. et al. Accurate H3K27 methylation can be established de novo by SUZ12- directed PRC2. Nat. Struct. Mol. Biol. 25, 225–232 (2018). 38. Oksuz, O. et al. Capturing the Onset of PRC2-Mediated Repressive Domain Formation. Molecular Cell 70, 1149–1162.e5 (2018). 39. Shen, X. et al. Jumonji modulates polycomb activity and self-renewal versus differentiation of stem cells. Cell 139, 1303–1314 (2009). 40. Castel, D. et al. Histone H3 wild-type DIPG/DMG overexpressing EZHIP extend the spectrum diffuse midline gliomas with PRC2 inhibition beyond H3-K27M mutation. Acta Neuropathol 1–5 (2020). doi:10.1007/s00401-020-02142-w 41. Pratt, D. et al. Diffuse intrinsic pontine glioma-like tumor with EZHIP expression and molecular features of PFA ependymoma. Acta Neuropathol Commun 8, 37–5 (2020). 42. Grasso, C. S. et al. Functionally defined therapeutic targets in diffuse intrinsic pontine glioma. Nat. Med. 21, 555–559 (2015). 43. Nagaraja, S. et al. Histone Variant and Cell Context Determine H3K27M Reprogramming of the Enhancer Landscape and Oncogenic State. Molecular Cell 76, 965–980.e12 (2019). 44. Krug, B. et al. Pervasive H3K27 Acetylation Leads to ERV Expression and a Therapeutic Vulnerability in H3K27M Gliomas. Cancer Cell 35, 782–797.e8 (2019). 45. Alecki, C. et al. RNA-DNA strand exchange by the Drosophila Polycomb complex PRC2. Nat Commun 11, 1781–14 (2020).

bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint Figure 1 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. K27me3 ChIP EZHIP ChIP Cells expressing EZHIP 9 A EEDf/f Cre EED-/- C K27me3 D G EZHIP EZHIP EZHIP EZHIP 10 KB 10 KB 400 WT M406E WT M406E 0-900 0-900 313 EEDf/f 429 0-900 0-900 403

ChIP 98% EZH2 -/- EED 426 6 200

0-1.5 0-2 CpG Island Number of peaks f/f 43 EED 6 EZHIP 0 0-1.5 0-2 EZH2 EZHIP ChIP

K27me3 EED-/- Peaks

0-50 0-50 -20kb Peaks (n= 438) EZHIP 20kb -20kb 20kb -20kb 20kb-20kb 20kb f/f EED E 0 16 0 16 0 84 0 84 EZHIP WT EZHIP M406E ChIP-Rx RPKM 0-50 0-50 [EZHIP] ChIP

EZHIP (nM): 250 75 150 25 75 150 EED-/- H Inefficient High Affinity CGIs: EZHIP-PRC2 Binding Complex Onecut3 Lhx5 Sdsl B Fluorography EZH2 EZH2 KLP EZH2 ChIP K27me3 ChIP EZHIP ChIP EED EED EEDf/f EED-/- EEDf/f EED-/- EEDf/f EED-/- EZHIP KLP K27me3 Coommassie EZHIP

Input FLAG-IP Input FLAG-IP F I

ControlM406EWT ControlM406EWT ControlA-395 DMSOControlA-395 DMSO FLAG- FLAG- EZHIP EZHIP EZHIP Peaks (n= 438) EZHIP Ezh2 Ezh2 -10kb 10kb -10kb 10kb-10kb 10kb-10kb 10kb-10kb 10kb-10kb 10kb

0 200 0 200 0 1.6 0 1.6 0 84 0 84 Suz12 Suz12 ChIP-Rx ChIP-Rx RPKM

EZHIP preferentially intereacts with allosterically stimulated PRC2 (A) EZH2, H3K27me3 and EZHIP ChIP-Seq enrichments in EEFf/f or EED-/- MEFs expressing FLAG-tagged EZHIP. (B) EZH2, H3K27me3 and EZHIP ChIP enrichments in EEFf/f or EED-/- MEFs expressing EZHIP at EZHIP peaks. (C) Over- lap between EZHIP, H3K27me3 and EZH2 peaks in MEFs expressing EZHIP. (D) Fraction of EZHIP peaks that contain CpG islands (blue). (E) HMT assays: 20 nM PRC2 was incubated with 200 nM nucleosomes, 4 µM S-adenosyl Methi- onine and 1 µM 3H-S-adenosyl Methionine. EZHIP WT or M406E was titrated into the reaction mixture. PRC2 activity on histone H3 was detected using fluorography following SDS-PAGE. (F) Immunoblots of inputs and elutions from FLAG-affinity purification of EZHIP WT or M406E from MEFs. (G) Normalized enrichment of H3K27me3 and FLAG-EZHIP ChIPs from MEFs expressing EZHIP WT or M406E mutant. (H) Binding of H3K27me3 to EED leads to a conformational change in PRC2, which increases its affinity for EZHIP. (I) Immunoblots of inputs and elutions from FLAG-affinity purification of EZHIP WT from MEFs in the presence or absence of EED inhibitor, A-395. Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is 10the KB author/funder.Ezh2 ChIP All rights reserved.0-950 No reuseEZH2 allowed ChIP without permission. SUZ12 A PRC2 PRC2 PRC2 B EZHIP C EZHIP EZHIP E U2OS EZHIP-/- M406E PCLs M406E WT U2OS Jarid2 0-950 12 EZHIP WT 10

Recruitment Spreading Hit-&-Run Neurod2 Ppp1r1b 8 U2-OS human Osteosarcoma cells PRC2 D 10 KB SUZ12 Cut-&-Tag PRC2PRC2 0-30 RPKM

+/+ ChIP-Rx EZHIP EZHIP 6 PRC2

0-30 42 SUZ12 Peaks (n= 570) EZHIP-/- Residual H3K27me3 (n= 1126)

Recruitment Spreading Hit-&-Run 20 60 100 140 180

TMPRSS5 ZBTB16 NXPE1 -10Kb 10Kb -10Kb 10Kb -10Kb 10Kb -10Kb 10Kb EZHIP inhibits PRC2 spreading by stalling it at CGIs containing residual H3K27me3 (A) EZHIP-mediated re-distribution of PRC2. Normally, PRC2 catalyzes H3K27 methylation on upto 90% of total Histone H3. There- fore, only a small fraction of PRC2 is present at recruitment sites (CGIs). In cells expressing EZHIP, EZHIP stalls PRC2 spreading by forming a high affinity complex with allosterically stimulated PRC2 at recruitment sites containing residual H3K27me3 and thereby, increasing its steady state occupancy at CGIs. (B) EZH2 ChIP-Rx profiles in MEFs expressing EZHIP WT or M406E. (C) EZH2 ChIP-Rx enrichment centered at residual H3K27me3 peaks. (D) SUZ12 CUT&RUN occupancy (RPKM) in U2OS cells expressing endogenous EZHIP or EZHIP-/- cells. (E) SUZ12 enrichment (RPKM) at SUZ12 peaks in U2OS cells. Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737H3.3 K27M; this version posted August 5, 2020. The copyright holder forH3.1 this preprintK27M ABMEFs H3.3 K27M Cells C D H3.1 K27M Cells 5 Kb E H3K27M-FLAG (which was0-20 not certified by peer review)5 Kb is theK27me3 author/funder.K27M-bound All rights reserved.0-20 No reuse allowed without permission.K27me3 K27M-bound ChIP EZH2 reChIP ChIP EZH2 reChIP FLAG FLAG ChIP FLAG ChIP ChIP 0-10 0-10 K27me3 Ezh2 reChIP K27me3 ChIP ChIP

0.5-10 Illumina Paired- 0.5-10 End Seq EZH2 EZH2 reChIP reChIP Residual K27me3 (n= 1053) frag. ≤ 400 bp Residual K27me3 (n= 1192) -10kb 10kb -10kb 10kb -10kb 10kb -10kb 10kb Wnt9b Wnt3 Wnt9b Wnt3 0 1 2 0 1 2 3 0 1 2 0 1 2 3 ChIP Enrichment ChIP Enrichment EZH2 ChIP 50 KB 0-200 F G H SUZ12 ChIPs H3.3 K27R SUZ12 ChIPs BT245 DIPG-XIII H3 Wildtype H3 K27M K27M K27M 0-200 K27M Knockout K27M Knockout pcGBM G477 DIPG-VI DIPG-XIII 18 H3.3 20 K27M 16 18 Uqcr10 Nf2 Nefh Ap1b1 14 0-4000-300 25 KB 16

H3.1 14 12 K27R 0-400 Normalized Enrichment 0-300 Normalized Enrichment 12 10 H3.1 Residual K27me3 Peaks K27M 8 H3K27me3 Peaks (n= 2497) 10 -10kb 10kb -10kb 10kb -10kb 10kb-10kb 10kb -10kb 10kb -10kb 10kb -10kb 10kb-10kb 10kb Rprml Gosr2 Wnt9b Wnt3 Nsf n= 4529 n= 6423 H3 K27M reduces H3K27me3 in trans by stalling PRC2 at CpG islands (A) Sequential ChIP-Seq used to identify genomic localtions of H3K27M-bound Ezh2. (B) H3.3 K27M-FLAG ChIP, H3K27me3 ChIP and Ezh2 reChIP in cells expressing FLAG-tagged H3.3 K27M. (C) H3K27me3 and EZH2 reChIP profiles in cells expressing H3.3K27M at residual H3K27me3 peaks. (D and E) Same as B and C respectively but for cells expressing FLAG-tagged H3.1K27M. (F) Ezh2 ChIP-Rx enrichment in cells expressing H3.3 (top) or H3.1 (bottom) K27M or K27R. (G) SUZ12 ChIP enrichment in DIPGs containing H3K27M mutations or H3WT at combined H3K27me3 peaks (Harutyunyan et al 2019). (H) Heat- map displaying SUZ12 enrichment in DIPGs cell lines containing H3K27M mutation or corrosponding H3K27M knockout cells. FigurebioRxiv 4 preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A B C [Cu2+] H3K27me3 ChIP 100 Kb EZHIP M406E EZHIP WT H3.3 K27R H3.3 K27M 0-15 Control EZHIP [Cu2+]: 10 5 10 10 µM

FLAG Control 0-15

K27me3 5 µM 0-15 EZHIP H3K27me3 RFP K27me2 10 µM

Tubulin PREs: Merge Ubx Glut3 iab-8 Actn3 Dad m-cup beat-llb Dscam3 Sgs5

89% Known PRE D E Annotated PREs Ph peaks F unannotated PH binding site 0-400 Control EZHIP 178 (55%) 111(34%) 37 Control K27me3 K27me3 K27me3 0 100 200 300 0-200 Number of sites Disproportionately ∆K27me3 retained K27me3 G ATAC-Seq PH ChIP Annotated 0-400 PREs ATAC-Seq ATAC-Seq PH Peaks 0-400

100 200 PH ChIP

100 200 300

Read Density Polycomb Read Density Recruitment Sites

0 -2KB 2KB -2KB 2KB grh wbl CR44432 rib

Human EZHIP reduces H3K27me3 in Drosophila through a conserved mechanism. (A) H3K27me3 staining (green) of wing imaginal discs from third instar larvae expressing either EZHIP WT, EZHIP M406E, H3 K27M, or H3 K27R (as indicated) driven by en-GAL4. RFP (red) indicates the region of en-GAL4 expression. (B) Immunoblots of whole cell extracts from S2 cells expressing FLAG-tagged EZHIP induced with 5 µM or 10 µM copper for 72 hours. Empty vector was used as control. (C) Reference-normalized H3K27me3 profile in S2 cells as described in B. (D) Strategy to identify putative PRC2-recruitment sites within residual H3K27me3 peaks in S2 cells. Genomic regions that disproportionately retained H3K27me3 in cells expressing EZHIP were determined as a difference of internally-normalized H3K27me3 (EZHIP-Control). (E) Fraction of putative PRC2 recruitment sites within residual H3K27me3 that contained annotated Polycomb Response Elements (PREs) or Ph peaks. (F) Baseline H3K27me3 ChIP, ∆H3K27me3 (EZHIP-Control), ATAC-Seq and PH ChIP-Seq profiles (related to figure S7C). Previously annotated PREs and other PRC2 recruitment sites are highlighted in red and purple respectively. (G) ATAC-Seq and Ph ChIP-Seq densities at putative PRC2 recruitment sites. bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Supplementary Figure 1:

0-2 M406E f/f -/- A EED EED B EZH2 ChIP-qPCR ** C 0-2

0.6 EEDf/f ** ** K27me3 WT - EED-/- 0-35 WT M406E WT 0.4 M406E FLAG- 0-35 * FLAG- EZHIP WT EZHIP 0.2 Percent Input 0 Ass1 Fubp3 Qrfp Ppapdc3 EZH2 ** 0.8 f/f H3K27me3 ChIP-qPCR EED D 2.0 H3K27me3 ChIP-qPCR 0.6 EED-/- ** M406E K27me2 ** 1.5 EZHIP WT 0.4 1.0 K27me3 ** 0.2

Percent Input 0.5 0 H3 0 ** FLAG-EZHIP ChIP-qPCR 0.8 FLAG-EZHIP ChIP-qPCR 0.8 ** f/f EED ** M406E ** 0.6 -/- 0.6 EED EZHIP WT ** ** 0.4 0.4 0.2 0.2 Percent Input Percent Input Percent Input 0 0 Rpl3 Wnt3 RFX4 Rpl3 Gapdh Crhr1Wnt9b Crhr1Wnt9bWnt3 RFX4 Hoxb13 Onecut3 Gapdh Hoxb13 Onecut3 Negative Intergenic Residual H3K27me3 Negative Intergenic Residual H3K27me3

EZHIP colocalizes with PRC2 and residual H3K27me3 in vivo: (A) Immunoblots of whole cell extracts from cells expressing EZHIP wild-type, EZHIP M406E, and EED knockout cells expressing EZHIP WT. (B) ChIP-qPCR of EZH2, H3K27me3 and EZHIP-FLAG ChIPs in cells expressing EZHIP before and after Cre-recombinase mediated EED knockout. (C) Genomic profile on H3K27me3 and EZHIP-FLAG ChIPs in cells expressing EZHIP wildtype or M406E. (D) ChIP-qP- CR analyses of H3K27me3 and EZHIP-FLAG ChIPs in cells expressing EZHIP wildtype or M406E. (**p<0.05) Supplementary Figure 2: EZH2 ChIP-Rx bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint A (which was not certified by peer review) is theB author/funder. All rightsC reserved.Residual No reuse allowedSpreading without permission.D 100 PRC2 PRC2 K27me3 Sites PRC2 p<2.2e-16 PCLs p<2.2e-16 n=5818 Jarid2 n=1126 80

60 Spreading

40 Recruitment Spreading Hit-&-Run ChIP-Rx EZHIP CpG Density 20 K27me3 Residual Fraction of EZH2 reads 0

PRC2 0.00 0.02 0.04 0.06 0 0.04 0.08

EZHIP 0 0.5 1.0 1.5 EZHIP EZHIP PCLs Residual Spread M406E WT Jarid2 K27me3 EZHIP EZHIP EZHIP EZHIP M406E WT M406E WT SUZ12 Cut-&-Run E 15 EZH2 ChIP-qPCR F Recruitment Spreading Sites Sites Recruitment Spreading Hit-&-Run EZHIP WT ** 10 M406E p<2.2e-16 p< 2.2e-16 n=12032 ** ** n=570 PRC2 PRC2 PRC2PRC2 Redistribution 5 EZHIP PRC2 Percent Input ** n.s. ** 0

Crhr1 Wnt9b Wnt3 RFX4 Gapdh Onecut3 Recruitment Spreading Hit-&-Run Negative Residual H3K27me3 Intergenic SUZ12 density (RPKM) 25.9% 0 2 4 6 8 0.0 0.2 0.4 0.6 0.8 1.0 EZHIP+/+ EZHIP-/- EZHIP+/+ EZHIP-/-

EZHIP sequesters PRC2 at CpG islands containing residual H3K27me3 and prevents PRC2 spreading (A) Schematic depicting the mechanism of EZHIP-mediated PRC2 redistribution. In normal cells, PRC2 is recruited to its high affinity sites at CGIs through polycomb like proteins (PCLs) or Jarid2, where it initiates H3K27me3. Binding of initial H3K27me3 to EED insitigates allosteric-stimulation of EZH2 and PRC2 spreading in cis. Therefore, concentrations of PRC2 at CGIs and spreading sites reaches an equili- birium. In cells expressing EZHIP, PRC2 is able to catalyze initial H3K27me3 since EZHIP has poorer affinity for unstimulated PRC2. However, allosteric stimulation of PRC2 by initial H3K27me3 leads to the formation of a high-affinity EZHIP-PRC2 complex at CGIs. Therefore, PRC2 occupancy is shifted towards CGI away from spreading sites at equilibrium, which results in a disproportionate reduction of intergenic H3K27me3. (B) Boxplot displaying CpG densities (total number of CpGs/length of the region) at residual H3K27me3 (red) and spreading (green) sites. (C) Boxplot displaying the reference-normalized EZH2 enrichment at recruitment and spreading sites in cells expressing EZHIP WT or M406E mutant. (D) Barchart displaying the fraction of EZH2 read densities within the recruitment or spreading sites in cells expressing EZHIP WT or M406E. (E) EZH2 ChIP enrichment in cells expressing EZHIP WT or M406E as measured by qPCR (p<0.05**). (F) Boxplot displaying the reference-normalized EZH2 enrichment at recruitment and spreading sites in EZHIP+/+ and EZHIP-/- U2OS osteosarcoma cells. SupplementarybioRxiv preprint doi: Figure https://doi.org/10.1101/2020.08.05.236737 3: ; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 0-12 A B C H3.1 100 Kb H3.1 H3.3 K27M 0-12 H3.3 K27MK27R K27MK27R - H3.1K27MH3.3K27M K27M 0-12

FLAG FLAG ChIP Control 500 EZH2 1-10 400 H3.1 300 K27M K27me3 1-10 200 H3.3 K27me2 100 K27M 1-10 EZH2 reChIP H3 EtBr stained Control agarose gel

Prdm13 Usp45 Fbxl4 H3.1 K27M H3.3 K27M D E F K27me3 EZH2 G K27me3 EZH2 H3.1 K27M H3.1 K27R H3.1 K27M H3.3 K27M H3.3 K27R H3.3 K27M Percent Input Percent Input 0 0.1 0.2 0 0.1 0.2

Rpl3 Rpl3 Hoxb8Crhr1 RFX4 Hoxb8Crhr1 RFX4 Gapdh Onecut3 Gapdh Onecut3 Residual H3K27me3 (n= 1249) Residual H3K27me3 (n= 1079) -10kb 10kb -10kb 10kb -10kb 10kb -10kb 10kb -10kb 10kb -10kb 10kb

0 1 2 20 40 60 80 0 1 2 20 40 60 80 ChIP-Rx ChIP-Rx ChIP-Rx ChIP-Rx

H3.1 K27M H3.3 K27M H EZH2 ChIP-qPCR I J n= 1079 9856 K n= 1249 14273 20 20 ** H3.1 K27M ** H3.3 K27M H3.3 K27R ** H3.1 K27R ** ** 10 10 p < 2.2e-16 p < 2.2e-16 ** ** Percent Input Percent Input

0 0 CpG Density CpG Density Wnt3 RFX4 Crhr1 Wnt9b Wnt3 RFX4 Gapdh Crhr1 Wnt9b Gapdh Onecut3 Onecut3 0.00 0.02 0.04 Neg. Control Residual K27me3 Neg. Control Residual K27me3 0.00 0.02 0.04 0.06 Retained Lost Retained Lost

H3K27M sequester PRC2 at CpG islands containing residual H3K27me3. (A) Immunoblots from whole cell extracts of MEFs expressing H3.1 or H3.3 K27M and K27R. (B) Ethidium bromide stained agarose gel displaying the fragment size distribution of nucleosomes used for FLAG ChIP followed by Ezh2 reChIP experiments. (C) Genome browser view of FLAG-EZH2 ChIP-reChIP densities in cells expressing H3.1K27M or H3.3K27M. EZH2 reChIP enrichment was not found in control cells lacking FLAG-tagged H3 transgene. (D and E) EZH2 reChIP enrichment at a subset of promoters in cells expressing H3.1K27M (D) or H3.3K27M (E) as measured by qPCR. (F and G) Heatmap displaying EZH2 enrichment at residual H3K27me3 in cells expressing H3.1 K27M and K27R (F), or H3.3 K27M and K27R (G) . (H and I) EZH2 ChIP enrichment in cells expressing H3.1K27M (H) or H3.3 K27M (I) at a subset of target loci measured by qPCR (** p<0.05). (J and K) Boxplot displaying CpG densities at residual H3K27me3 peaks and regions that displayed loss of H3K27me3. bioRxiv preprint doi: https://doi.org/10.1101/2020.08.05.236737; this version posted August 5, 2020. The copyright holder for this preprint Supplementary(which was Figure not certified 4 by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. H3K27me3 width CpG Density Suz12 Enrichment n= 9863 11767 6423 4101 n= 9479 2497 n= 9863 11767 6423 4101 A SUZ12 ChIP B C D 0.5, 0-200 10 Kb n.s.

p<2.2e-16 p<2.2e-16 DIPG-VI p<2.2e-16 0-200 0.8 1.0 DIPG-XIII

0-200 0.07, n.s. CpG Density pcGBM

0-200 0.2 SUZ12 Enrichment (RPKM) H3K27me3 Peak Width (KB) H3K27me3 Peak Width H3 Wildtype H3 K27M 0 0.4 0.6 0 50 150 250 G477 0.00 0.04 0.08 H3 WT H3 K27M G477 G477 pcGBM2 DIPG-XIIIDIPG-VI Spread residual peaks pcGBM2 DIPG-XIIIDIPG-VI NR146771 PRDM6 H3 WT H3 K27M H3 WT H3 K27M

n= 6423 10401 n= 6423 7354 RNA Expression E FG K (Median Normalized TPM) 4 0-300 0 5 10 15 p<2.2e-16 p<2.2e-16 K27M EZH1 EZH2 Cas9 0-300 EED DIPG-XIII K27M SUZ12 Knockout CpG Density RBBP4 0 1 2 3 0.00 0.02 0.04 0.06 RBBP7 FOXD1 SUZ12 Enrichment (RPKM) K27M KO Residual Spread K27me3 AEBP2 H I n= 4529 6945J n= 4529 4864 JARID2 G477 WT SUZ12 ChIP 10 Kb 0-250 pcGBM WT 2.2e-16 p<2.2e-16 MTF2 K27M 0.08 DIPG-VI K27M PHF1 DIPG-XIII K27M

BT245 Cas9 0-250 PHF19 DIPG-XIII K27M KO EPOP K27M CpG Density BT245 K27M Knockout PALI1 BT245 K27M KO 0 0.04 0.0 0.5 1.0 1.5 SUZ12 Enrichment (RPKM) K27M KO Residual Spread FOXD1 K27me3 H3 K27M sequesters PRC2 at CpG islands containing residual H3K27me3 in Diffuse Midline Glio- mas (A) Genome browser view of SUZ12 occupancies in DMG cell lines containing H3 K27M mutations (DIPG-XIII and DIPG-VI) or H3 WT (G477 and pcGMB2). (B) Boxplots displaying the H3K27me3 peak width in DMG cell lines as describes in A. (C) Boxplot displaying CpG densities within residual H3K27me3 peaks (red) or spreading sites that lost H3K27me3 (blue). (D) Boxplots displaying RPKM normalized SUZ12 enrichment at residual H3K27me3 peaks. (E and H) Genome browser view of SUZ12 occupancies in DIPG-XIII (E) or BT245 (H) cell line containing H3.3 K27M mutation or after cas9-mediated genetic depletion of H3.3 K27M. (F and I) Boxplot displaying SUZ12 RPKM enrichment at residual H3K27me3 peaks in DIPG-XIII cells (F) or BT245 cells (I) containing H3.3 K27M or H3.3 K27M Knockout cells. (G and J) Boxplot displaying CpG densities wihin residual H3K27me3 peaks or at spreading sites that gained H3K27me3 upon H3.3 K27M knockout in DIPG-XIII (G) or BT245 cells (J). (K) Plot displaying the RNA expression of PRC2 subunits in DMGs containing H3 WT (blue) or H2 K27M mutations (red) or H3.3 K27M knockout cells (green). None of the PRC2 subunits were differentially expressed in H3 K27M mutant tumors or upon H3 K27M knockout. bioRxivSupplementary preprint doi: https://doi.org/10.1101/2020.08.05.236737 Figure 5 ; this version posted August 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. ATAC-Seq PH ChIP 50

A B D 200 Lost K27me3

41, 40

3 High EZHIP 10.4% 150 Control 1 Low EZHIP 1003 960 6 30

(68, 17.2%) Annotated PREs High EZHIP 1 100 637 (68, 72.3%)

Low EZHIP 20 Normalized Read Density Normalized Read Density 50 10 0 C PH Binding Sites

Control K27me3

K27me3 (EZHIP-Control) mbl Dcr-2 Elk Hsf IM23 edl sano hppy Rep Obp56h Efhc1.2 Ims insc ktub Sdc Tbp

Control K27me3

K27me3 (EZHIP-Control)

ATAC-Seq

PH ChIP

grh wbl CR44432 rib CR43160 CG9235

Human EZHIP reduces global H3K27me3 but retains residual H3K27me3 at Polycomb -recruitment sites in Drosophila S2 cells (A) Venn diagram displaying the overlap between H3K27me3 peaks found in control S2 cells (blue) or cells induced to express EZHIP at 5 µM (yellow) and 10 µM (red) concentrations. (B) Density of annotated PREs within the residual peaks of H3K27me3 in S2 cells expressing EZHIP at 10 µM Cu2+. (C) Genome browser view of H3K27me3 ChIP, H3K27me3 (EZHIP-Control), ATAC-Seq and PH ChIP-Seq profiles (related to figure 4G). Previously annotated PREs and other polycomb recruitment sites are highlighted in red and purple respectively. (D) Heatmap displaying ATAC-Seq and PH ChIP-Seq read enrichments at polycomb recruitment sites identified in Figure 4D.