Selective Inhibition of Ezh2 by a Small Molecule Inhibitor Blocks Tumor Cells Proliferation

Selective inhibition of Ezh2 by a small molecule inhibitor blocks tumor cells proliferation

Wei Qia,1, HoMan Chana,1,2, Lin Tenga, Ling Lia, Shannon Chuaia, Ruipeng Zhanga, Jue Zenga, Min Lia, Hong Fana, Ying Lina, Justin Gua, Ophelia Ardayﬁob, Ji-Hu Zhangb, Xiaoxia Yana, Jialuo Fanga, Yuan Mia, Man Zhanga, Tao Zhoua, Grace Fenga, Zijun Chena, Guobin Lia, Teddy Yanga, Kehao Zhaoa, Xianghui Liua, Zhengtian Yua, Chris X. Lua, Peter Atadjaa, and En Lia,3

aChina Novartis Institutes for BioMedical Research, Shanghai 201203, China; and bCenter for Proteomic Chemistry, Novartis Institutes for BioMedical Research, Cambridge, MA 02139

Edited* by Rudolf Jaenisch, Whitehead Institute for Biomedical Research, Cambridge, MA, and approved November 12, 2012 (received for review June 18, 2012)

Ezh2 (Enhancer of zeste homolog 2) protein is the enzymatic mutations and their enhanced activity toward H3K27me2 pro- component of the Polycomb repressive complex 2 (PRC2), which motes DLBCL cell proliferation. represses gene expression by methylating lysine 27 of histone H3 In this study, we developed an S-Adenosyl methionine (SAM) (H3K27) and regulates cell proliferation and differentiation during competitive inhibitor of Ezh2, EI1, which inhibits the methyl- embryonic development. Recently, hot-spot mutations of Ezh2 transferase activity of the Ezh2/PRC2 with high selectivity across were identiﬁed in diffused large B-cell lymphomas and follicular an HMT panel. EI1 treated cells exhibited decreased H3K27 lymphomas. To investigate if tumor growth is dependent on the methylation without changes of other histone H3 methylation enzymatic activity of Ezh2, we developed a potent and selective marks. Using this tool inhibitor, we showed that inhibition of Ezh2 in DLBCL cells carrying hot-spot mutations and some Ezh2 small molecule inhibitor, EI1, which inhibits the enzymatic activity overexpressed tumor cell lines results in decreased proliferation, of Ezh2 through direct binding to the enzyme and competing with cell cycle arrest, and apoptosis. Thus, inhibition of Ezh2 enzy- the methyl group donor S-Adenosyl methionine. EI1-treated cells matic activity may provide a therapeutic option for the treatment exhibit genome-wide loss of H3K27 methylation and activation of

of DLBCL and other cancers. CELL BIOLOGY PRC2 target genes. Furthermore, inhibition of Ezh2 by EI1 in diffused large B-cell lymphomas cells carrying the Y641 mutations Results results in decreased proliferation, cell cycle arrest, and apoptosis. Identification and Biochemical Characterization of EI1 as a Potent and These results provide strong validation of Ezh2 as a potential Selective Inhibitor of PRC2. To identify inhibitors of Ezh2/PRC2, therapeutic target for the treatment of cancer. we performed a high-throughput screen using recombinant PRC2 protein complex containing Ezh2, Suz12, EED, RbAP48, zh2 is the mammalian homolog of Enhancer of Zeste, the key and AEBP2 (24). EI1 (Fig. 1A) was designed based on one Ecomponent of Polycomb repressive complex 2 (PRC2), which chemical scaffold identified from the high-throughput screen. represses gene expression in development (1). Ezh2 and the This compound demonstrated potent, concentration-dependent other two proteins, Suz12 and EED, form the core PRC2 complex, inhibition of the enzymatic activity against both Ezh2 wild-type which possesses histone methyltransferase (HMT) activity and and Y641F mutant enzymes with IC50 of 15 ± 2 nM and 13 ± 3 methylates H3K27 (2, 3). H3K27 methylation has been correlated nM, respectively (Fig. 1B). To understand the mode of inhibition with transcriptional repression and heterochromatin formation (4). for EI1, SAM competition experiments were carried out under the The Drosophila ortholog, E(z), was initially identified as a regulator conditions of saturated substrate peptide. In accordance with of homeotic gene expression and body segmentation (5). In the Cheng–Prusoff relationship for a competitive binding mode, mammals, Ezh2 is highly expressed in stem cells and actively pro- the IC50 value of EI1 increased linearly with increasing concen- liferating cells, and down-regulated in differentiated cells (6). Ezh2 tration of SAM. The fitting of the data into the Cheng–Prusoff knockout in mice leads to developmental abnormality and embry- equation for competitive inhibition using linear regression analysis ± onic lethality (7, 8). Ezh2-null ES cells are viable (8) and Ezh2 is gave a Ki value of 13 3nM(Fig.1C). involved in maintaining pluripotency and repression of lineage- Although SAM is the common cofactor for all HMTs, EI1 differentiation genes (8–10). showed remarkable selectivity against Ezh2 over other HMTs Ezh2 is overexpressed in a broad spectrum of tumors, in- (Table 1). All biochemical reactions were carefully characterized cluding prostate cancer, breast cancer, myeloma, hepatocellular with enzymology studies and the SAM and substrate concen- carcinoma, gastric cancer, and so forth (1, 6). Overexpression of trations were kept at their respective Km for most of the HMTs. Ezh2 in mouse mammary gland leads to epithelial hyperplasia Strikingly, EI1 displayed ∼90-fold selectivity for Ezh2 over Ezh1, (11). Multiple studies using si/shRNA show that reduction of and >10,000-fold selectivity over other HMTs (Table 1). In Ezh2 expression in tumor cell lines inhibits cell proliferation (12), migration, and invasion (13) or angiogenesis (14), and leads to apoptosis (15). Ezh2 contains the characteristic SET domain Author contributions: W.Q., H.C., C.X.L., P.A., and E.L. designed research; W.Q., L.T., L.L., [Suppressor of variegation 3–9 (Suv (3–9), Enhancer of zeste R.Z., and Y.L. performed research; J.Z., M.L., H.F., J.G., O.A., J.-H.Z., X.Y., J.F., Y.M., M.Z., (E(z)) and Trithorax] (16) present in most HMTs. Recurrent T.Z., G.F., Z.C., G.L., T.Y., and K.Z. contributed new reagents/analytic tools; W.Q., L.T., S.C., somatic mutations in the SET domain of Ezh2 was identified X.L., and Z.Y. analyzed data; and W.Q., H.C., and E.L. wrote the paper. in diffused large B-cell lymphomas (DLBCL) patients (17–22). The authors declare no conflict of interest. These mutations lead to the change of tyrosine (Y) 641 to *This Direct Submission article had a prearranged editor. phenylalanine (F), serine (S), asparagine (N), histidine (H), Data deposition: The data reported in this paper have been deposited in the Gene Ex- cysteine (C), or alanine (A) 677 to glycine (G). The Y641 mutant pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE41315). and A677G proteins show enhanced activity on H3K27me2 1W.Q. and H.C. contributed equally to this work. peptide (20, 23). Cancer cell lines heterozygous for these Ezh2 2Present address: Oncology, Novartis Institutes for BioMedical Research, Cambridge, mutants show increased H3K27me3 and decreased H3K27me2 MA 02139. level (17, 20, 21, 23). Wild-type and mutant Ezh2 proteins may 3To whom correspondence should be addressed. E-mail: [email protected]. work cooperatively in cells to maintain a high level of H3K27me3 This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. (23). It is therefore hypothesized that the Ezh2 hot-spot 1073/pnas.1210371110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1210371110 PNAS Early Edition | 1of6 Downloaded by guest on September 25, 2021 A time-course study was performed in WSU-DLCL2 cells to understand the kinetics of H3K27 methylation inhibition. The H3K27me3 level was decreased 24 h after EI1 treatment, and reached the lowest level after 4–5 d (Fig. 2B). To test the speci- ficity of EI1, we examined other histone H3 lysine methylation marks. As shown in Fig. 2C, the other marks were not changed. p16 is a well characterized Ezh2 target gene (26–28) and its promoter is directly bound by Ezh2 and enriched for H3K27me3 (8, 27). Depletion of Ezh2 in G401 cells leads to up-regulation of p16 expression and senescence (26). We therefore examined the effect of EI1 on p16 expression in G401 and found EI1 could activate p16 expression in a dose-dependent manner (Fig. 2D, Left). Consistently, EI1 also caused growth inhibition of G401 (Fig. S3). In a time-course study, p16 expression was activated 2 d after EI1 treatment and the expression level increased by 20- fold at day 5 (Fig. 2D, Right). The decrease of H3K27me3 was readily observable 24 h after EI treatment (Fig. S1D). These results indicate that the decrease of H3K27me3 precedes the Fig. 1. Biochemical characterization of EI1 inhibition of PRC2 complex. (A) activation of p16 expression. Chemical structure of EI1. (B) Inhibition curves of EI1 for wild-type enzyme ChIP experiments were performed to examine the change of with unmethylated H3K27 as the substrate and Y641F enzyme with di- H3K27me3 and Ezh2 status on p16 promoter (Fig. 2E). Both methylated H3K27 as the substrate. Each data point represents the mean of Ezh2 and H3K27me3 were enriched at the p16 promoter (Fig. two replicates at each concentration of the compounds. The data are fitto 2F), as previously reported (8, 27). Specific and significant en- a dose–response equation using PRISM. (C) Plot of IC50 values of EI1 as richment of H3K27me3 and Ezh2 were observed at region A and a function of SAM concentration relative to the Michaelis–Menten constant – B, and the enrichment decreased toward region C (Fig. 2F). (Km) of SAM. Using the Cheng Prusoff equation for competitive inhibitors Following EI1 treatment, H3K27me3 signals were significantly [IC50 = Ki (1+[S]/Km)], the Ki value of EI1 was derived from the linear regression fitting. reduced across the p16 promoter (Fig. 2F). Interestingly, Ezh2 remained bound at the p16 promoter (Fig. 2F). Therefore, EI1 activated p16 expression by suppressing H3K27me3, but not contrast, another well-characterized SAM competitive HMT Ezh2 occupancy at the promoter. inhibitor, sinefungin, showed no selectivity toward Ezh2 (Table 1). Taken together, our results indicate that EI1 is a potent and EI1 Caused Similar Phenotypes as Ezh2 Knockout in Mouse Embryonic selective inhibitor against Ezh2. Fibroblasts. To assess the selectivity of EI1 in cells, we established SV40-T immortalized mouse embryonic fibroblasts (MEFs) carry- fl EI1 Inhibits Cellular H3K27 Methylation and Activates Ezh2 Target ing the Cre-ERT2 transgene and two oxed Ezh2 loci (29). Treat- ment with 4-OH-tamoxifen (4-OH-T) led to significant depletion of Gene Expression. Previous studies using si/shRNA to knock down Ezh2 and H3K27me3 (Fig. 3A). EI1 treatment did not affect Ezh2 Ezh2 or other components of PRC2 showed that DLBCL and protein level but caused similar loss of H3K27me3 (Fig. 3A). The rhabdoid tumor cells were highly dependent on PRC2 for pro- immortalized MEFs showed decreased proliferation (∼40–50%) liferation (25, 26). Therefore, we tested the effect of EI1 in these following Ezh2 depletion by 4-OH-T (Fig. 3B). Consistently, cell lines, including the DLBCL cells carrying Ezh2 mutations Y641F Y641N a similar decrease in proliferation was observed in EI1-treated [WSU-DLCL2 (Ezh2 ), SH-DHL6 (Ezh2 )] or wild-type MEF. Importantly, there was no further reduction of proliferation Ezh2 [(OCI-LY19 (Ezh2WT), and GA10 (Ezh2WT)] and a rhab- WT when the MEF was treated with EI1 in combination with 4-OH-T doid tumor line G401 (Ezh2 ). EI1 dramatically inhibited the (Fig. 3B). Therefore, EI1 treatment mimics Ezh2 knockout in the H3K27me3 and H3K27me2 levels in these cells in a dose-de- MEFs to inhibit H3K27me3 and cell proliferation. pendent manner, but H3K27me1 was largely unchanged (Fig. 2A and Fig. S1 A and B). The effect was similar in these cell lines, EI1 Selectively Inhibited the Growth of DLBCL Cells Carrying Ezh2 although they have different basal H3K27me3 and H3K27me2 Mutation. To test if inhibition of Ezh2 might be efficacious to levels. For example, in WSU-DLCL2, SU-DHL6, and Karpas422 inhibit cancer cell growth, siRNAs specific to Ezh2 was designed with Ezh2 mutations, the H3K27me3 level was much higher than and transfected into OCI-LY19 (Ezh2WT) and WSU-DLCL2 that in DLBCL cells with wild-type Ezh2 (Fig. S1C). (Ezh2Y641F). Two independent siRNAs, named siEzh2-1 and

Table 1. HMT Proﬁling of EI1 with a panel of HMT enzymes Sinefungin Enzyme EI1 IC50 (nM) EI1 fold-selectivity* Sinefungin IC50 (nM) fold-selectivity* Assay with enzyme complex Substrate

Ezh2 9.4 1 20 1 Ezh2/SUZ12/EED/AEBP2/RbAP48 H3[21–44] Ezh1 1,340 142 33 1.65 Ezh1/SUZ12/EED/AEBP2/RbAP48 H3[21–44] G9a >100,000 >10,000 18 0.9 H3[1–21] Suv39H2 >100,000 >10,000 25 1.25 H3[1–21] Set7/9 >100,000 >10,000 1.5 0.075 H3[1–21] CARM 1 >100,000 >10,000 0.5 0.025 H3[1–21] SmyD2 >100,000 >10,000 0.15 0.0075 p53[363–393] SETD8 >100,000 >10,000 >60 >3H4[11–33] NSD3 >100,000 >10,000 45 2.25 H3 [1–45] SETD2 >100,000 >10,000 16 0.8 H3[21–44] MLL >100,000 >10,000 12 0.6 MLL/WDR5/RBBP5/ASH2L H3[1–21] Dot1L >100,000 >10,000 >60 >3 Nucleosome

*Fold selectivity is calculated as the ratio of the IC50 of the enzyme tested over the IC50 of Ezh2.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1210371110 Qi et al. Downloaded by guest on September 25, 2021 number of WSU-DLCL2 colonies were decreased after EI1 treatment. In contrast, the colony formation of OCI-LY19 was not affected. Taken together, these results indicate that EI1 is not generally toxic to cells, but selectively inhibits the neoplastic properties of DLBCL cells with Ezh2 mutation.

EI1 Causes Cell Cycle Arrest and Apoptosis of DLBCL Cells with Y641 Mutations. To understand the role of Ezh2 in the regulation of the proliferation of DLBCL lines carrying Ezh2 Y641 mutations, we examined the DNA synthesis and cell cycle progression by BrdU incorporation and FACS analysis. In WSU-DLCL2 cells + treated with EI1 for 7 d, significant decrease of BrdU cells (reduced from 41–8.72%) was detected (Fig. 5A). The percentage of cells in G2/M also decreased, whereas the percentage of cells in G1 was increased (Fig. 5 A and B). Similar results were observed in two other Ezh2 mutant DLBCL cell lines (Fig. 5B). In contrast, the Ezh2 wild-type cells OCI-LY19, GA10, and Toledo did not show significant changes of DNA synthesis or cell cycle distribution (Fig. 5 A and B). Cell cycle progression is regulated by the activity of cyclin- dependent kinases and levels of various cyclins (31). We therefore surveyed the expression of cyclins in EI1-treated WSU- DLCL2 cells. Cyclin A and B1 protein levels were diminished after 6 d of EI1 treatment (Fig. 5C). Similar effects were observed in other DLBCL cells with Ezh2 mutations (Fig. S4). This finding is consistent with the reduction of the G2/M population from FACS analysis. A mild increase of sub-G1 population was observed in EI1-treated SU-DHL6 and Karpas422, suggesting an

increase of apoptosis. To further examine this phenomenon, we CELL BIOLOGY carried out the Western blotting of procaspase 3 and cleaved (c)- caspase 3 in EI1-treated SU-DHL6 cells. Indeed, c-caspase 3 level D Fig. 2. EI1 inhibits cellular H3K27 methylation and activates p16 expression. was gradually elevated with increasing treatment time (Fig. 5 ). Taking these data together, we show that inhibition of the enzy- (A) Inhibition of H3K27 methylation by different concentrations of EI1. SU- fi DHL6 (Ezh2Y641N) and OCI-LY19 (Ezh2WT) cells were treated with EI1 for 4 d, matic activity of Ezh2 by EI1 can signi cantly block cell cycle and G401 cells were treated with EI1 for 2 d at the indicated concen- progression and induce apoptosis in Ezh2 mutant DLBCL cells. trations. H3K27me3, H3K27me2, H3K27me1, and H3 were detected by immunoblot. (B) Time course of H3K27 methylation inhibition by EI1. WSU- Inhibition of Ezh2 Causes Down-Regulation of a Proliferation Gene DLCL2 (Ezh2Y641F) cells were treated with EI1 at 10 μM for indicated time Signature and Up-Regulation of a Memory B-Cell Signature in periods. (C) Immunoblot against histone H3 methylation modifications using DLBCL. We examined the gene-expression changes in Karpas422 WSU-DLCL2 (Ezh2Y641F) cell lysates treated with indicated concentrations of (Ezh2Y641N) after EI1 treatment in a time-course experiment. EI1- EI1. (D) qPCR analysis of p16 mRNA level in G401 (Ezh2WT) cells. p16 mRNA is treated cells were compared with the corresponding controls to normalized to GAPDH level and plotted as fold to control base line samples. obtain the significantly differentially expressed gene lists using (Left) The expression of p16 gene in G401 cells treated by different concen- a two-tailed Student t test at each time point. A cutoff of P value < trations of EI1 for 2 d. (Right) Time course of p16 expression in G401 cells > μ 0.05 and fold-change 1.5 was used to select the EI1 regulated treated by EI1 (10 M) for indicated time. (E) Graphic presentation of p16 genes. The number of genes that showed statistically significant genomic locus. The regions A, B, and C indicate the primer location for ChIP- change increased with increasing treatment time (Fig. 6A). More PCR. (F) ChIP-PCR using anit-Ezh2 and H3K27me3 antibodies at the indicated regions on p16 promoter before and after 48 h of EI1 treatment. Rabbit and genes were up-regulated than down-regulated between days 1 and mouse IgG are used as controls. Relative enrichment is calculated as percent 3 after EI1 treatment (Fig. 6A), suggesting the early response to of input. Ezh2 inhibition is primarily gene up-regulation, which is consistent with Ezh2’s role in transcriptional repression. By day 6, the number

siEzh2-2, knocked down Ezh2 mRNA by ∼80% (Fig. S2). Both siRNAs inhibited the proliferation of WSU-DLCL2 but had little effect on OCI-LY19 (Fig. 4A), suggesting the proliferation of DLBCL cells with Ezh2 mutations is dependent on Ezh2. To test whether EI1 treatment has a similar effect as Ezh2 knockdown, we measured the proliferation after EI1 treatment and IC50s of EI1 on both wild-type and mutant Ezh2 DLBCL lines. The proliferation of Ezh2 mutant DLBCL cells, including WSU-DLCL2, SU-DHL6, Karpas422, DB, and SU-DHL4, was strongly inhibited by EI1 in a dose-dependent manner (Fig. 4 B and C and Fig. S3). In contrast, Ezh2 wild-type DLBCL cells, including OCI-LY19 and GA10, were not affected by EI1, and To- ledo was only weakly affected by EI1 (Fig. 4 B and C and Fig. S3). Fig. 3. EI1 inhibits H3K27me3 similarly as Ezh2 knockout. (A) Inhibition of Thus, speciﬁc inhibition of Ezh2 enzymatic activity by EI1 selec- H3K27me3 by EI1 and Ezh2 knockout induced by 4-OH-T. Immortalized MEF (Ezh2f/f) cells were treated with 3.3 μM EI1, 4-OH-T, or the combination of EI1 tively inhibited the growth of DLBCL cells carrying Ezh2 mutations. and 4-OH-T. After cell harvesting, indicated proteins were detected by im- Soft agar colony formation evaluates if cancer cells can pro- munoblot. (B) Proliferation of the immortalized MEFs after treatment by liferate in an anchorage-independent manner (30). We tested 3.3 μM EI1, Ezh2 knockout induced by 4-OH-T, or the combination of EI1 and the effect of EI1 on the colony formation ability of WSU-DLCL2 4-OH-T. Viable cells were counted at days 3, 6, and 11. The results were and OCI-LY19 cells. As shown in Fig. 4D, both the size and the plotted on a linear scale.

Qi et al. PNAS Early Edition | 3of6 Downloaded by guest on September 25, 2021 Fig. 4. EI1 selectively inhibits the proliferation and colony formation of DLBCL cell lines carrying Ezh2 Y641 mutations. (A) Proliferation of WSU-DLCL2 (Ezh2Y641F) and OCI-LY19 (Ezh2WT) cells. Viable cells were counted every 3 or 4 d in the presence of siRNAs against Ezh2 and results were plotted on a log- arithmic scale. (B) Proliferation of WSU-DLCL2 (Ezh2Y641F), SH-DHL-6 (Ezh2Y641N), OCI-LY19 (Ezh2WT), and GA10 (Ezh2WT) cells in the presence of EI1 at the indicated concentrations. Viable cells were similarly counted as in A.(C) Effect of EI1 on the proliferation of DLBCL cell lines with wild-type Ezh2, or cells carrying Ezh2 Y641 mutations. Cells were similarly cultured and counted as in B and IC50 values were calculated at 14 or 15 d of treatment. The level bars indicate median for each group (see also Fig. S3). (D) Colony formation of WSU-DLCL2 (Ezh2Y641F) and OCI-LY19 (Ezh2WT) cells in the presence of EI1 at the indicated concentrations. (Upper) On day14, photomicroscopy of cell colonies was taken at 20× magniﬁcation. (Lower) Quantiﬁcation of colony formation by AlamarBlue. The data are plotted as percentage of DMSO.

of down-regulated genes dramatically increased, which could be consistent with our data on cell proliferation and cell cycle (Figs. because of the secondary response (Fig. 6A). 4 and 5) (34, 35). Interestingly, genes up-regulated by EI1 showed To investigate if genes with expression changes at early time significant enrichment of a gene set distinguishing splenic mar- points are more likely direct PRC2 targets, two differential ex- ginal zone memory B cell from the germinal center B cell pressed gene lists were generated at days 3 and 9 after EI1 treat- (Splenic_marginal_zone_B cell_gt_GC_B cell signature from ment, respectively. These lists were compared with the PRC2 and Lymphoma/Leukemia Molecular Profiling Project), indicating that H3K27me3 target genes in human embryonic fibroblasts and Ezh2 inhibition may promote differentiation to memory B-cell centroblasts, respectively (10, 25). The gene list at day 3 showed (Fig. 6C) (35, 36). We compared our data with a geneset that better enrichment to the known PRC2 target genes than the day negatively correlated with Ezh2 mRNA level in DLBCL (25), and 9 gene list (Fig. S5). Almost all these genes are up-regulated, found that this gene signature showed significant enrichment in the suggesting early response genes are direct Ezh2 targets. We further genes up-regulated by EI1 treatment (Fig. 6D). Intriguingly, our selected 13 genes that were up-regulated by EI1 from microarray data also displayed significant enrichment of a geneset regulated in and revalidated their expression by quantitative PCR (qPCR) (Fig. prostate cancer cell after siRNA-mediated knockdown of Ezh2 S6A). ChIP experiments were performed to determine if these (Fig. 6 E and F and Fig. S7) (37). genes were labeled by H3K27me3 at the promoters. EI1 treatment significantly decreased the H3K27me3 enrichment in these 13 Discussion gene promoters, but not the negative control GAPDH promoter In this study, we developed a potent small molecule inhibitor, (Fig. S6B). These data further strengthen the notion that the de- EI1, which inhibits Ezh2 activity through competition with the pletion of H3K27me3 at target gene promoters correlates to EI1- cofactor SAM. EI1 is highly selective against Ezh2 over its close dependent transcriptional up-regulation. homolog Ezh1 and other HMTs. The SET domains of Ezh1 and To further delineate the pathways that are responsive to Ezh2 Ezh2 are 90% identical (Fig. S8A). Structural modeling shows inhibition, we performed gene set enrichment analysis (GSEA) the divergent residues I626 and C663 in Ezh2 (T664 and S664 in using the microarray data (P value < 0.05 and fold-change > 1.5) Ezh1) are around the EI1 binding pocket, even though no direct after 6 d of EI1 treatment (32, 33). The analysis revealed that the interaction with EI1 (Fig. S8 B and C), suggesting these residues genes down-regulated by EI1 showed significant enrichment of a may contribute to the selectivity. Future structural studies of proliferation signature in DLBCL (Fig. 6B) (Proliferation_DLBCL EI1-Ezh2 cocrystal structure may help to understand how such signature from Lymphoma/Leukemia Molecular Profiling Project), high selectivity is achieved.

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1210371110 Qi et al. Downloaded by guest on September 25, 2021 Fig. 5. EI1 causes cell cycle G0/1 arrest and apoptosis in DLBCL cell lines carrying Ezh2 Y641 mutations. (A) Cell cycle analysis of WSU-DLCL2 (Ezh2Y641F)andOCI- LY19 (Ezh2WT) cells after 7-d treatment with DMSO or EI1 (10 μM). The analysis was performed using BrdU incorporation with 7-AAD staining followed by ﬂow cytometry. Annotations indicate distribution of cells to G0/1, S, and G2/M phases of cell cycle. One repre- sentative result of multiple experiments is shown. (B) Cell cycle changes analyzed by FACS after treatment as in A. The cell lines are: WSU-DLCL2 (Ezh2Y641F), SU- DHL6 (Ezh2Y641N), Karpas422 (Ezh2Y641N), OCI-LY19 (Ezh2WT), GA10 (Ezh2WT), and Toledo (Ezh2WT). (C) Immunoblot analysis. Cells were similarly cultured and treated as in A (see also Fig. S4). (D) Immunoblot using SU-DHL6 cell lysates treated with 10 μMofEI1 for indicated time periods.

EI1 treated cells show specific reduction of H3K27me2 and Ezh1. Although it was not shown whether the gene up-regulation CELL BIOLOGY H3K27me3 (Fig. 2C). Using Ezh2-floxed MEF cells, we further by these compounds correlated with the decrease of H3K27me3 demonstrated that EI1 had similar effect as Ezh2 knockout on mark, both compounds could inhibit the proliferation of several H3K27 methylation and proliferation, indicating that EI1 is Ezh2 mutant lymphoma cell lines. This finding is consistent with a highly specific Ezh2 inhibitor suitable for functional studies. our findings and independently validates Ezh2 as a potential tar- siRNA knockdown of PRC2 complex, such as Ezh2, Eed, and get for cancer treatment. Suz12, could disrupt the complex and block cancer cell pro- All Ezh2 mutations are heterozygous in DLBCL and the liferation (1, 12–14, 37). Here, we demonstrated that inhibition mutant cells contain much higher levels of H3K27me3 than of HMT activity of Ezh2 by EI1 without disrupting the chromatin wild-type cells (Fig. S1C) (17, 20, 21). Ezh2-Y641 mutants have binding of PRC2 is sufficient to blocks cancer cell proliferation weak biochemical activity toward unmethylated H3K27 and and colony formation, particularly in Ezh2 mutant containing H3K27me1 peptide substrates, whereas they have much higher DLBCL cells. Thus, a small molecule inhibitor of Ezh2 can be catalytic activity converting H3K27me2 to H3K27me3 (20, 23, 40). efficacious in treating cancers. Wild-type and mutant Ezh2 complexes may collaborate to convert During the review of our manuscript, Knutson et al. (38) and unmethylated H3K27 toward high levels of H3K27me3 (40). In- McCabe et al. (39) published two Ezh2 inhibitors, which also act terestingly, although H3K27me2 and H3K27me3 are similarly in a SAM-competitive manner and are highly selective over depleted in DLBCL cells with or without Ezh2 mutation, Ezh2

Fig. 6. Ezh2 inhibition by EI1 causes down-regulation of DLBCL proliferation signature and up-regulation of memory B-cell signature. (A) Gene-expression changes of Karpas422 cell at different time points after EI1 (5 μM) treatment. Changes in gene expression were ﬁl- tered by fold-change >1.5 or ≤1.5 and by P < 0.05. (B and F). GSEA showing enrichment of “Proliferation signature in DLBCL” (B)(34)and“Nuytten Ezh2 targets down” (F) (37) in genes down-regulated by 6 d of EI1 treatment. (C–E). GSEA showing enrichment of “Splenic marginal zone memory B cell high vs. GC B cell” (C)(36),“genes negative correlated with Ezh2 mRNA in DLBCL” (D) (25), and “Nuytten Ezh2 targets up” (E) (37) in genes up-regulated by 6 d of EI1 treatment. See also Table S1.

Qi et al. PNAS Early Edition | 5of6 Downloaded by guest on September 25, 2021 mutant cells are specifically growth-inhibited by EI1 (Figs. 2A and Materials and Methods 4 B and C,andFig. S3). This finding suggests that the Ezh2 mu- Cell Culture. Cells were maintained in a humidified incubator at 37 °C, 5%

tation in DLBCL may be an oncogenic driver. Because EI1 is (vol/vol) CO2. OCI-LY19, GA10, Toledo, WSU-DLCL2, DB, SU-DHL4, Karpas422 equally active against both the wild-type and Y641 mutant Ezh2, it and SU-DHL6 were cultured in RPMI-1640 (Invitrogen, 11875) with 15% (vol/vol) will be interesting to develop a mutant-selective inhibitor in the FBS (Invitrogen, 10099-141). G401 was cultured in DMEM (Invitrogen, 11995) with future to evaluate if it will be equally efficacious as EI1. 10% (vol/vol) FBS and 0.055 mM 2-mercaptoethanol (Sigma, M7522). MEFs were In this study, we conducted a detailed cell cycle analysis of EI1 from the Antoine Peters laboratory (41) andculturedinDMEMwith10%(vol/vol) treated DLBCL cells. Ezh2 mutant cells are growth-arrested by FBS. The immortalized MEF was generated by retroviral transduction of pBABE- SV40 T and selected with hygromycin (Invitrogen). Single clones were selected EI1 with G1 arrest (Fig. 5). Expression of cyclin A and B1 is based on the inducible knockout of Ezh2 in response to 4-OH-T (100 nM). down-regulated (Fig. 5). GSEA analysis showed EI1 can reac- tivate the expression of PRC2 target genes (Fig. S5) (37). After Biochemical Assay. PRC2 biochemical reaction was carried out in reaction 6 d of treatment, the proliferation signature was significantly buffer (20 mM Tris pH 8.0, 0.1% BSA, 0.01% Triton, 0.5 mM DTT). The 20 nM down-regulated (Fig. 6B) (34) and accompanied by the activation wild-type PRC2 complex, 1 μM SAM, and 1.5 μM H3K27me0 (21–44); or 10 nM of a set of memory B-cell–specific genes (Fig. 6C) (36). These Y641F mutant complex, 1 μM SAM, and 3 μM H3K27me2 (21-44) were used, results suggest inhibition of Ezh2 may induce growth arrest and respectively, because they have differential substrate preference (38). differentiation into memory B cells. We found multiple genes up- regulated by EI1 have H3K27me3 enrichment at their promoters. Microarray Analysis. RNAwaspurifiedfromKarpas422after1,2, 3, 6, and9d after μ EI1 effectively reduced the H3K27me3 level on these promoters DMSO or EI1 (5 M) treatment. Triplicate samples were collected and hybridized and activated their expression. At the p16 promoter, EI1 sup- to U133_Plus 2 Chips (Affymetrix). The gene-expression data were normalized using the Robust Multiarray Averaging method (42). To generate a differentially pressed H3K27me3 level but Ezh2 remained bound to the pro- expressed gene list, EI1-treated samples were compared with the corresponding moter. To our knowledge, this demonstration that inhibition of DMSO controls using a cutoff of fold-change > 1.5 and P value < 0.05 (Student histone H3K27me3 can occur without dissociation of the Ezh2 t test). GSEA was performed as previously described (33). complex from its target gene is unique. Ezh2 can interact with Additional details of biochemical and cellular reagents and protocols are other chromatin modifying enzymes, such as histone deacetylases described in SI Materials and Methods.SeeTables S2 and S3 for primers used. and DNA methyltransferases (4) to repress transcription. Future studies should investigate if all Ezh2 target genes can be suffi- ACKNOWLEDGMENTS. We thank Youzhen Wang, Jeff Zhang, Jingquan Dai, Wei Yi, Shaolian Zhou, Yaochang Xu, Martin Pfeifer, Daniela Gabriel, Clemens ciently reactivated by reversing promoter H3K27me3 level alone. Schuffler, Jan Weiler, Chandra Vargeese, Keith Bowman, Julien Papillon, and Such knowledge might help to understand how to use Ezh2 Nicholas Keen for support and discussions; and Antoine Peters (Friedrich inhibitors as single agent or in combination for cancer treatment. Miescher Institute) for discussion and reagents.

1. Simon JA, Lange CA (2008) Roles of the EZH2 histone methyltransferase in cancer 23. Yap DB, et al. (2011) Somatic mutations at EZH2 Y641 act dominantly through epigenetics. Mutat Res 647(1–2):21–29. a mechanism of selectively altered PRC2 catalytic activity, to increase H3K27 trime- 2. Cao R, et al. (2002) Role of histone H3 lysine 27 methylation in Polycomb-group si- thylation. Blood 117(8):2451–2459. lencing. Science 298(5595):1039–1043. 24. Cao R, Zhang Y (2004) SUZ12 is required for both the histone methyltransferase ac- 3. Morey L, Helin K (2010) Polycomb group protein-mediated repression of transcrip- tivity and the silencing function of the EED-EZH2 complex. Mol Cell 15(1):57–67. tion. Trends Biochem Sci 35(6):323–332. 25. Velichutina I, et al. (2010) EZH2-mediated epigenetic silencing in germinal center B 4. Margueron R, Reinberg D (2011) The Polycomb complex PRC2 and its mark in life. cells contributes to proliferation and lymphomagenesis. Blood 116(24):5247–5255. Nature 469(7330):343–349. 26. Wilson BG, et al. (2010) Epigenetic antagonism between polycomb and SWI/SNF 5. Lewis EB (1978) A gene complex controlling segmentation in Drosophila. Nature complexes during oncogenic transformation. Cancer Cell 18(4):316–328. 276(5688):565–570. 27. Agherbi H, et al. (2009) Polycomb mediated epigenetic silencing and replication 6. Sparmann A, van Lohuizen M (2006) Polycomb silencers control cell fate, de- timing at the INK4a/ARF locus during senescence. PLoS ONE 4(5):e5622. velopment and cancer. Nat Rev Cancer 6(11):846–856. 28. Kotake Y, et al. (2007) pRB family proteins are required for H3K27 trimethylation and 7. O’Carroll D, et al. (2001) The polycomb-group gene Ezh2 is required for early mouse Polycomb repression complexes binding to and silencing p16INK4alpha tumor sup- – development. Mol Cell Biol 21(13):4330–4336. pressor gene. Genes Dev 21(1):49 54. 8. Shen X, et al. (2008) EZH1 mediates methylation on histone H3 lysine 27 and complements 29. Conzen SD, Cole CN (1995) The three transforming regions of SV40 T antigen are required fi – EZH2 in maintaining stem cell identity and executing pluripotency. Mol Cell 32(4):491–502. for immortalization of primary mouse embryo broblasts. Oncogene 11(11):2295 2302. 9. Boyer LA, et al. (2006) Polycomb complexes repress developmental regulators in 30. Shoemaker RH, et al. (1985) Application of a human tumor colony-forming assay to new drug screening. Cancer Res 45(5):2145–2153. murine embryonic stem cells. Nature 441(7091):349–353. 31. Malumbres M, Barbacid M (2009) Cell cycle, CDKs and cancer: A changing paradigm. 10. Lee TI, et al. (2006) Control of developmental regulators by Polycomb in human Nat Rev Cancer 9(3):153–166. embryonic stem cells. Cell 125(2):301–313. 32. Mootha VK, et al. (2003) PGC-1alpha-responsive genes involved in oxidative phosphor- 11. Li X, et al. (2009) Targeted overexpression of EZH2 in the mammary gland disrupts ylation are coordinately downregulated in human diabetes. Nat Genet 34(3):267–273. ductal morphogenesis and causes epithelial hyperplasia. Am J Pathol 175(3):1246–1254. 33. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based ap- 12. Gonzalez ME, et al. (2009) Downregulation of EZH2 decreases growth of estrogen re- proach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA ceptor-negative invasive breast carcinoma and requires BRCA1. Oncogene 28(6):843–853. 102(43):15545–15550. 13. Au SL, et al. (2012) Enhancer of zeste homolog 2 epigenetically silences multiple tumor 34. Rosenwald A, et al.; Lymphoma/Leukemia Molecular Profiling Project (2002) The use suppressor microRNAs to promote liver cancer metastasis. Hepatology 56(2):622–631. of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell 14. Lu C, et al. (2010) Regulation of tumor angiogenesis by EZH2. Cancer Cell 18(2): lymphoma. N Engl J Med 346(25):1937–1947. 185–197. 35. Shaffer AL, et al. (2006) A library of gene expression signatures to illuminate normal 15. Wu ZL, et al. (2010) Polycomb protein EZH2 regulates E2F1-dependent apoptosis and pathological lymphoid biology. Immunol Rev 210:67–85. – through epigenetically modulating Bim expression. Cell Death Differ 17(5):801 810. 36. Weller S, et al. (2004) Human blood IgM “memory” B cells are circulating splenic 16. Schapira M (2011) Structural chemistry of human SET domain protein methyl- marginal zone B cells harboring a prediversified immunoglobulin repertoire. Blood – transferases. Curr Chem Genomics 5(Suppl 1):85 94. 104(12):3647–3654. 17. Bödör C, et al. (2011) EZH2 Y641 mutations in follicular lymphoma. Leukemia 25(4): 37. Nuytten M, et al. (2008) The transcriptional repressor NIPP1 is an essential player in – 726 729. EZH2-mediated gene silencing. Oncogene 27(10):1449–1460. 18. Capello D, et al. (2011) Mutations of CD79A, CD79B and EZH2 genes in immunode- 38. Knutson SK, et al. (2012) A selective inhibitor of EZH2 blocks H3K27 methylation and fi – ciency-related non-Hodgkin lymphomas. Br J Haematol 152(6):777 780. kills mutant lymphoma cells. Nat Chem Biol 8(11):890–896. 19. Lohr JG, et al. (2012) Discovery and prioritization of somatic mutations in diffuse large B-cell 39. McCabe MT, et al. (2012) EZH2 inhibition as a therapeutic strategy for lymphoma with lymphoma (DLBCL) by whole-exome sequencing. Proc Natl Acad Sci USA 109(10):3879–3884. EZH2-activating mutations. Nature, 10.1038/nature11606. 20. McCabe MT, et al. (2012) Mutation of A677 in histone methyltransferase EZH2 in 40. Sneeringer CJ, et al. (2010) Coordinated activities of wild-type plus mutant EZH2 drive human B-cell lymphoma promotes hypertrimethylation of histone H3 on lysine 27 tumor-associated hypertrimethylation of lysine 27 on histone H3 (H3K27) in human B- (H3K27). Proc Natl Acad Sci USA 109(8):2989–2994. cell lymphomas. Proc Natl Acad Sci USA 107(49):20980–20985. 21. Morin RD, et al. (2010) Somatic mutations altering EZH2 (Tyr641) in follicular and 41. Terranova R, et al. (2008) Polycomb group proteins Ezh2 and Rnf2 direct genomic diffuse large B-cell lymphomas of germinal-center origin. Nat Genet 42(2):181–185. contraction and imprinted repression in early mouse embryos. Dev Cell 15(5):668–679. 22. Ryan RJ, et al. (2011) EZH2 codon 641 mutations are common in BCL2-rearranged 42. Irizarry RA, et al. (2003) Exploration, normalization, and summaries of high density germinal center B cell lymphomas. PLoS ONE 6(12):e28585. oligonucleotide array probe level data. Biostatistics 4(2):249–264.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1210371110 Qi et al. Downloaded by guest on September 25, 2021