Discovery of a first-in-class EZH2 selective degrader

Abstract

The enhancer of zeste homolog 2 (EZH2) is the main enzymatic subunit of the PRC2 complex, which catalyzes trimethylation of histone H3 lysine 27 (H3K27me3) to promote transcriptional silencing. EZH2 is overexpressed in multiple types of cancer including triple-negative breast cancer (TNBC), and high expression levels correlate with poor prognosis. Several EZH2 inhibitors, which inhibit the methyltransferase activity of EZH2, have shown promise in treating sarcoma and follicular lymphoma in clinics. However, EZH2 inhibitors are ineffective at blocking proliferation of TNBC cells, even though they effectively reduce the H3K27me3 mark. Using a hydrophobic tagging approach, we generated MS1943, a first-in-class EZH2 selective degrader that effectively reduces EZH2 levels in cells. Importantly, MS1943 has a profound cytotoxic effect in multiple TNBC cells, while sparing normal cells, and is efficacious in vivo, suggesting that pharmacologic degradation of EZH2 can be advantageous for treating the cancers that are dependent on EZH2.

EZH2 (enhancer of zeste homolog 2) is one of the most important histone methyltransferases (HMTs) and is the main catalytic subunit of the polycomb repressive complex 2 (PRC2) that catalyzes methylation of histone 3 lysine 27 (H3K27)^1,2. To be catalytically active, EZH2 minimally requires two other PRC2 components, EED (embryonic ectoderm development) and SUZ12 (suppressor of zeste 12 protein homolog). The trimethylation of H3K27 (H3K27me3) is a transcriptionally repressive epigenetic mark that regulates gene expression, differentiation and development³, and hypertrimethylation of H3K27 drives tumorigenesis and progression of several types of tumors including diffuse large B-cell lymphoma and malignant rhabdoid tumor (MRT)⁴. Numerous EZH2 inhibitors, which inhibit the methyltransferase activity of EZH2/PRC2 (that is, reducing H3K27me3) have been developed⁵, including UNC1999 and C24, the EZH2 inhibitors previously discovered by us^6,7. Among them, EPZ6438^8,9, GSK126¹⁰, CPI-1205¹¹ and PF-06821497¹² have entered clinical development for the treatment of several types of tumor including sarcoma, lymphoma and MRT, where inhibition of the enzymatic activity of EZH2/PRC2 can effectively block the growth of tumor cells^4,5. It has also been reported that the roles of EZH2 in cancers can be independent of the canonical role of PRC2 or the catalytic function of EZH2⁴. For example, in hormone-refractory prostate cancer, phosphorylation of EZH2 switched its function from a polycomb repressor to a transcriptional coactivator by catalyzing the methylation of androgen receptor (AR)¹³. The catalytically independent functions of EZH2 have also been discovered^14,15. For example, EZH2 controls the protein translation of p53 gain-of-function (GOF) mutants by binding to p53 mRNA, and knocking down EZH2 was shown to be efficacious in p53 GOF prostate cancer in vivo models¹⁴.

Triple-negative breast cancer (TNBC) represents 12–20% of all breast cancers. TNBC has poor prognosis, high recurrence, a low survival rate and has higher incidence in African-American and Hispanic women^16,17. Currently, there are no effective therapies for treating a substantial portion of TNBC patients¹⁸. EZH2 is overexpressed in many cancers, including breast and prostate cancers^4,19–21. In breast cancer, EZH2 has been identified as a major driver for disease development and progression, and high expression level of EZH2 correlates with poor prognosis^19,22–27. Importantly, however, EZH2 inhibitors that do not affect EZH2 protein levels in cells are ineffective at blocking proliferation of TNBC and other breast cancer cell lines^6,28 even though knockdown of EZH2 via RNA interference is sufficient to block tumor growth²⁵. Taken together, these results suggest that expression of EZH2, but not the methyltransferase activity of EZH2, is critical for TNBC and other breast cancer progression. We therefore hypothesized that EZH2 selective degraders—compounds that selectively reduce EZH2 protein levels—could provide an effective therapeutic approach for treating TNBC and other types of cancer that are dependent on EZH2.

PROTACs (proteolysis targeting chimeras) and hydrophobic tagging are successful technologies/strategies for selective degradation of the target protein^29,30. Although PROTAC technology has been rapidly gaining momentum in the drug discovery field, the hydrophobic tagging approach has received considerably less attention from the biomedical community. The hydrophobic tagging approach utilizes a bulky and hydrophobic group attaching to a small-molecule binder of the target protein. The binding of this bivalent compound to the target protein leads to misfolding of the target protein and its subsequent degradation by the proteasome^29,31. This approach has been successfully applied to the selective degradation of Her3, using a covalent inhibitor of Her3 as an irreversible binder to Her3³². So far, there is no report on the selective degradation of EZH2 using the PROTAC or hydrophobic tagging technology. Furthermore, it is unprecedented that attaching a hydrophobic tag to a non-covalent small-molecule binder can result in effective degradation of the target protein. Here, we report the discovery of a first-in-class EZH2 selective degrader (MS1943, 1), which was designed by linking a non-covalent inhibitor of EZH2 to a bulky adamantyl group, and describe characterization of this EZH2 degrader in vitro and in vivo. We have demonstrated that MS1943 effectively reduces EZH2 protein levels and selectively kills EZH2-dependent TNBC cells over normal cells while EZH2 inhibitors do not reduce EZH2 protein levels and are ineffective in blocking the proliferation of TNBC cells. We have also shown that treatment with MS1943 phenocopies the effect of knockout (KO) or knockdown (KD) of EZH2 in MS1943-sensitive and -insensitive cell lines. Importantly, MS1943 is efficacious in vivo, and we have established a pharmacokinetics (PK)/pharmacodynamics (PD) relationship in a mouse xenograft model. Our RNA expression analysis revealed that treatment with MS1943 was associated with activation of the unfolded protein response (UPR) in a TNBC cell line. We further confirmed that MS1943 treatment de-repressed several UPR-related genes in MS1943-sensitive TNBC cell lines but not in a MS1943-insensitive cell line. Taken together, our results suggest that pharmacologic degradation of EZH2 is a promising therapeutic approach for TNBC.

Results

Discovery and biochemical characterization of MS1943.

We previously discovered the EZH2 selective inhibitor C24 (2) (structure shown in Supplementary Fig. 1), which displays high potency for EZH2 (half-maximum inhibitory concentration, IC₅₀ = 12 nM) and was found to be >200-fold selective for EZH2 over the highly homologous H3K27 methyltransferase EZH1 (IC₅₀ > 2.5 μM)⁷ and UNC1999 (structure shown in Supplementary Fig. 1), which is a potent and selective inhibitor for both EZH2 and EZH1 over other methyltransferases⁶. Using the previously published crystal structures of PRC2, including co-crystal structures of PRC2 in complex with an EZH2 inhibitor^11,12,33,34, we performed docking of C24 into human PRC2 crystal structures and identified that the piperazine moiety of C24 is solvent-exposed, thus presenting a suitable handle to introduce a hydrophobic tag without interfering with EZH2 binding⁷. This potential linker attachment point was validated by the biotinylated UNC1999 (UNC2399; structure shown in Supplementary Fig. 1) we synthesized previously, which maintained high potency for EZH2 (IC₅₀ = 17 nM) and can effectively pull down EZH2 from cell lysates⁶. Based on these insights, we designed and synthesized a series of bivalent compounds by connecting the piperazine group of C24, via a linker, to various hydrophobic groups such as an adamantyl group. We selected C24 instead of UNC1999 as the binder to EZH2 because C24 was highly selective for EZH2 over EZH1. We then assessed the effects of these bivalent compounds on reducing EZH2 protein levels and on growth inhibition of TNBC cells. From these studies, we identified MS1943 (Fig. 1a) as a promising lead compound. MS1943 maintained high potency (IC₅₀ = 120 nM) for inhibiting EZH2 methyltransferase activity (Fig. 1b) and was highly selective for EZH2 over a wide range of methyltransferases including EZH1 (Fig. 1b and Supplementary Fig. 2). It was also inactive against 45 kinases (<10% inhibition at 10 μM, Supplementary Table 1), displayed <50% inhibition at 10 μM for 43 out of 45 G protein-coupled receptors (GPCRs), ion channels and transporters and showed 50% and 59% of inhibition, respectively, at 10 μM for Alpha1A and Sigma 2 receptors (Supplementary Table 2). Overall, MS1943 exhibited excellent selectivity for EZH2 over a broad range of methyltransferases and common drug targets.

Fig. 1 |. MS1943 is an EZH2 selective degrader.

a, Chemical structure of MS1943 (1). b, MS1943 retained high potency for EZH2 and was selective for EZH2 over EZH1 in radioactive methyltransferase inhibition assays. The EZH2 assay results are shown as the mean ± s.d. from two independent experiments in three technical replicates. c,d, MS1943, but not the EZH2 inhibitor C24, reduced EZH2 protein levels in MDA-MB-468 cells in a concentration- (c) and time-dependent manner (d): western blot analysis of EZH2, H3K27me3, SUZ12, EED and EZH1 in MDA-MB-468 cells treated with the aforementioned compounds (see Supplementary Figs. 16 and 17 for source blot images). Results are representative of at least two independent experiments.

Characterization of MS1943 in cells.

We first tested MS1943 alongside C24 for its ability to reduce EZH2 protein levels in MDA-MB-468 cells, a TNBC cell line. MS1943 was able to reduce EZH2 and SUZ12 protein levels in a concentration- and time-dependent manner, but interestingly without affecting EED protein levels, whereas the H3K27me3 mark was also suppressed (Fig. 1c,d and Supplementary Fig. 3a,c). MS1943 had minimum effect in reducing EZH1 protein levels (Fig. 1d). In contrast, C24 reduced the H3K27me3 mark more effectively than MS1943 but had no effect on protein levels of any of the PRC2 components tested, including EZH2 (Fig. 1c,d and Supplementary Fig. 3a,c,d). Moreover, MS1943 modestly reduced H3K27me2 and had no significant effect on H3K27me and H3K4me3 (Supplementary Fig. 3d). Compared with C24, MS1943 was also less effective at reducing the H3K27me2 mark, consistent with the H3K27me3 results. Furthermore, EZH2 and H3K27me3 levels were not changed upon treatment with AM41–44A (3), which contains only the adamantyl moiety, serving as a negative control (Supplementary Fig. 3a,b).

We next tested MS1943 and C24 in other cancer and normal cell lines. We found that MS1943, but not C24, effectively reduced EZH2 levels in BT549, HCC70 and MDA-MB-231 TNBC cells, as well as KARPAS-422 and SUDHL8 lymphoma cells and PNT2 non-cancerous prostate cells (Supplementary Figs. 4 and 5a–d), while C24 still reduced the H3K27me3 mark more effectively than MS1943 (Supplementary Figs. 4 and 5a–d). In addition, MS1943 concentration-dependently reduced EZH2 protein levels in MCF7 cells, an ER+ luminal breast cancer cell line (Supplementary Fig. 5e). Similar to MDA-MB-468 cells, MS1943 did not reduce EZH1 protein levels in BT549 cells (Supplementary Fig. 4a). Interestingly, MS1943 reduced SUZ12 levels in BT549 cells but not in HCC70 cells (Supplementary Fig. 4a,b), suggesting that the effects of MS1943 on other components of the PRC2 complex are context-dependent. We also found a variability in the kinetics of EZH2 degradation among the different cells. For example, MS1943 reduced EZH2 protein levels at 24 h and more profoundly at 48 h, but did not significantly reduce EZH2 protein levels at 6 and 12 h in MDA-MB-468 cells (Supplementary Fig. 3c), while in HCC70 cells, it significantly reduced EZH2 protein levels starting at 6 h and the reduction of EZH2 protein levels was maintained up to 48 h (Supplementary Fig. 4b,c). It was also found that MS1943 reduced EZH2 protein levels more effectively at 48 h than at 24 h in BT549, MDA-MB-231, PNT2, KARPAS-422 and SUDHL8 cells (Supplementary Figs. 4a and 5a–d). Collectively, these data show that MS1943 was able to effectively reduce EZH2 protein levels in various types of cell. To further characterize MS1943, we assessed its effects on EZH2 protein levels in MDA-MB-468 and HCC70 cells with or without a functional proteasome. Of note, co-treatment with the proteasome inhibitor MG132 was able to rescue EZH2 levels, at least in part, in a concentration-dependent manner in both MDA-MB-468 and HCC70 cells (Supplementary Fig. 6a,b), suggesting that the reduction of EZH2 protein levels resulted from the MS1943 treatment is mediated by a proteasome-related pathway.

We next asked whether MS1943 has any superior cytotoxicity when compared to EZH2 inhibitiors. We thus compared the EZH2 inhibitors C24, GSK126, CPI-1205 and EPZ6438 (structures shown in Supplementary Fig. 1) side by side with MS1943, for their ability to inhibit MDA-MB-468 cell growth. MS1943 (concentration for 50% of maximal inhibition of cell growth (GI₅₀) = 2.2 μM) but not EZH2 inhibitors effectively inhibited cell proliferation (Fig. 2a,b). Given that MS1943 was effective at reducing EZH2 levels in cells of different contexts, we asked whether that would result in cytotoxicity in other TNBC, cancer and non-cancerous cells. Importantly, MS1943 was also effective at inhibiting growth in several other TNBC cells that express EZH2²¹, such as HCC1187, HCC70, BT549 and HCC1954, as well as in MCF7, KARPAS-422 and SUDHL8 cells, with GI₅₀ values ranging from 1.1 to 4.9 μM (Supplementary Figs. 7 and 8), while it displayed little or no cytotoxicity in MDA-MB-231 (TNBC), K562 (leukemia), A549 (non-small-cell lung cancer), HCC827 (small-cell lung cancer) and non-cancerous MCF10A (breast) and PNT2 (prostate) cells (Supplementary Fig. 9). On the other hand, C24 failed to inhibit cell growth at up to 8 μM in most of the aforementioned cells except the germinal center B cells KARPAS-422, which are known to be relatively senstitive to EZH2 inhibition^9,10 (Supplementary Fig. 8b). Similar results were obtained using a variety of different assays including crystal violet and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays, highlighting the superior efficacy of MS1943 in inhibiting growth in TNBC cells (Fig. 2c and Supplementary Fig. 10a,b). Moreover, time-lapse images from our cell growth assays using a live imaging system indicated that MS1943, but not C24, was able to elicit a cytotoxic effect in MDA-MB-468, HCC1187 and HCC70 cells at 4 μM (Supplementary Fig. 11 and Supplementary Videos 1–9). Indeed, MS1943, but not C24, was able to induce cell death in MDA-MB-468 cells beyond background levels (cells treated with DMSO), as measured by inclusion of DRAQ7 dye, which stains dead and permeabilized cells (Fig. 2d).

Fig. 2 |. MS1943 (but not EZH2 inhibitors) inhibits cell growth and induces cell death in TNBC cells.

a, MS1943, but not EZH2 inhibitors GSK126, CPI-1205, EPZ6438 and C24, displayed concentration-dependent toxicity in MDA-MB-468 cells in MTT viability assays. The results are shown as individual dots per independent experiment, and different colors indicate compound concentrations. Results shown as black dots represent the mean and error bars indicate ±s.d. for three independent experiments per treatment. P values calculated from two-tailed Student’s t-tests with unequal variance between DMSO and different treatments are provided. b, MS1943, but not C24, inhibited growth of MDA-MB-468 cells with a GI₅₀ value of 2.2 μM in a proliferation assay. The results are shown as mean ± s.d. for three independent experiments. c, MS1943, but not C24, inhibited growth of MDA-MB-468 cells in a crystal violet assay. The results are representative of three independent experiments. d, MS1943 induced cell death in MDA-MB-468 cells, as measured by increased inclusion of DRAQ7. The results are shown as individual dots per independent experiment and colors represent the different treatment groups. Results in black dots represent the mean and error bars indicate ±s.d. for three independent experiments per treatment. P values from two-tailed Student’s t-tests with equal variance between DMSO and different treatments are indicated.

Our data, so far, suggest that although MS1943 is able to reduce EZH2 protein levels in a number of cell types, it displays cytotoxicity that may be dependent on whether EZH2 is driving cancer growth. To test this hypothesis and provide further evidence that the antiproliferative activity of MS1943 is mainly due to EZH2 degradation, we performed EZH2 KO experiments using the CRISPR/Cas9 technology in three TNBC cell lines (BT549 and MDA-MB-468 cells, which are sensitive to MS1943, and MDA-MB-231 cells, which are insensitive to MS1943) (Supplementary Fig. 12a and Fig. 3a–c). We found that deletion of EZH2 in the two MS1943-sensitive cell lines (BT549 and MDA-MB-468) resulted in significant cell growth inhibition (Fig. 3b,c), while deletion of EZH2 in MDA-MB-231 cells, which are insensitive to the EZH2 degrader treatment, led to no changes in cell proliferation (Fig. 3a). We also conducted EZH2 KD experiments using both short hairpin RNAs (shRNAs) and small interfering RNAs (siRNAs) in MDA-MB-468 cells and found that KD of EZH2 also significantly inhibited MDA-MB-468 cell growth (Fig. 3d and Supplementary Fig. 12b–d). Collectively, these results support our hypothesis that proliferation of the TNBC cells that are sensitive to the MS1943 treatment (such as MDA-MB-468 and BT549) depends on EZH2, while proliferation of the cells that are insensitive to MS1943 (such as MDA-MB-231) does not depend on EZH2. In addition, deletion of EZH2 (via CRISPR KO) in K562 cells³⁵ did not inhibit cell growth and, consistent with the KO results, MS1943 and C24 had no effect on growth of K562 cells (Supplementary Fig. 9d). The C24 result is in agreement with a previous report that K562 cells were insensitive to EZH2 inhibitors³⁶.

Fig. 3 |. Knockout or knockdown of EZH2 inhibits cell growth in TNBC cells.

a, Growth of EZH2 KO MDA-MB-231 cells for eight days. MFI, mean fluorescence intensity. Values in the plot are shown as mean ± s.d. from three independent replicates. b, Growth of EZH2 KO BT549 cells for eight days. Statistical testing was conducted with paired t-tests between individual time points with Bonferroni correction for multiple comparisons, *P < 0.05, **P < 0.005. Values in the plot are shown as mean ± s.d. from three independent replicates. c, Growth of EZH2 KO MDA-MB-468 cells for eight days. Statistical testing was conducted with paired t-tests between individual time points with Bonferroni correction for multiple comparisons, *P < 0.05, **P < 0.005, ***P < 0.0005. Values in the plot are shown as mean ± s.d. from three independent replicates. d, EZH2 KD inhibits MDA-MB-468 cell growth significantly. Values in the plot are shown as mean ± s.d. from three independent experiments. *P < 0.0001, calculated using two-way analysis of variance (ANOVA).

Taken together, these results suggest that (1) MS1943 has therapeutic potential in treating several types of cancer, including TNBC; (2) the antiproliferative activity of MS1943 is primarily due to its on-target EZH2 degradation effect; (3) MS1943 is not a non-selective cytotoxic agent.

Evaluation of MS1943 activity in vivo.

Next, we evaluated in vivo mouse PK properties of MS1943 and found that MS1943 was bioavailable via both intraperitoneal (i.p.) and per oral (p.o.) administration routes (Fig. 4a). A single i.p. injection of MS1943 at 50 mg per kg body weight achieved a peak plasma concentration (C_max) of 2.9 μM and resulted in plasma concentrations above its cellular IC₅₀ value for ~2 h. A single 150 mg per kg body weight p.o. dose achieved C_max of 1.1 μM, but plasma concentrations were below the cellular IC₅₀ value. Nevertheless, MS1943 was orally bioavailable in mice. Based on the PK results, we decided to evaluate the in vivo antitumor activity of MS1943 by treating mice bearing MDA-MB-468 tumor xenografts with 150 mg per kg body weight once daily i.p. injection of MS1943. Importantly, tumor growth was completely suppressed by MS1943, in comparison to the vehicle group (Fig. 4b). At this dose, MS1943 was well tolerated by the test mice, which did not exhibit any weight loss or other overt toxicities (Fig. 4c). Furthermore, we isolated tumor and plasma samples from mice treated once daily for seven days and determined that sufficient drug concentrations (above its cellular GI₅₀ value) were achieved in both tumor and plasma (Fig. 4d). To further investigate the effects of MS1943 in vivo, tumor samples were analyzed at the endpoint of the experiment using immunohistochemistry. Consistent with our in vitro data, we observed a significant reduction of both EZH2 protein levels and H3K27me3 mark in the tumors from mice treated with MS1943 (Fig. 5a,b). The antitumoral effect of MS1943 was due to increased apoptosis, as measured by cleaved caspase-3 levels, as well as decreased proliferation, as measured by staining with Ki-67 (Fig. 5a,b). Thus, we have established a PK/PD relationship for MS1943 in this tumor xenograft model. Overall, MS1943 was efficacious in vivo and well tolerated in mice at the efficacious dose.

Fig. 4 |. MS1943 suppresses tumor growth in vivo.

a, Plasma concentrations of MS1943 following a single i.p. injection (50 mg per kg body weight) or a single oral dose (150 mg per kg) over the 24 h period. Values in the plot are shown as mean ± s.d. from three mice per time point per administration route. b, Tumor growth data from MDA-MB-468 xenografts treated once daily i.p. with MS1943 (150 mg per kg) or vehicle control for 36 days. Values in the plot are shown as mean ± s.d. from eight mice per treatment arm. *P < 0.0001, Sidak’s multiple comparisons test following one-way ANOVA. c, Body weight changes of mice at the end of the in vivo experiment. Sidak’s multiple comparisons test was used following one-way ANOVA for statistics (n = 8). The P value was calculated between day 0 and day 36. d, MS1943 concentrations in the plasma and tumor samples isolated from mice treated once daily with 150 mg per kg via i.p. injection for seven days at 2 h after the last dose. Sidak’s multiple comparisons test was used following one-way ANOVA for statistics (n = 8). The P value was calculated between plasma and tumor.

Fig. 5 |. MS1943 induces apoptosis in the MDA-MB-468 xenograft model.

a, Representative EZH2, H3K27me3, cleaved caspase-3 and Ki-67 immunohistochemistry images of tumor sections from each treatment group. At least three tumor sections per treatment group were analyzed. b, Immunohistochemistry positive cells (in %). Results are shown as individual dots for independent experiments and colors indicate the treatment group (n = 3 independent experiments for each treatment group). Results in black dots represent the mean and error bars indicate ±s.d. of cell numbers from three different fields (two-tailed Student’s t-test, mean ± s.d.). P values of two-tailed Student’s t-tests with unequal variance between vehicle and MS1943 treatment are provided for each protein.

MS1943 induces prolonged activation of the UPR pathway.

To gain mechanistic insights into how MS1943 induces cell death, MDA-MB-468 cells were treated with MS1943 or DMSO control and changes of gene expression were assessed using RNA-seq experiments. Interestingly, MS1943-treated cells were characterized by a unique set of deregulated genes that could readily separate them from control cells (Fig. 6a). We first investigated how the gene expression of PRC2 target genes identified previously⁴ was altered in our MS1943-treated cells. We identified that several PRC2 target gene programs, including Wnt/ß-catenin signaling (c-Myc, cyclin D1 and Axin2), RUNX3, CK5 and CK6, were significantly altered in MS1943-treated cells with a false discovery rate (FDR) at 5%, as when EZH2 is degraded (Supplementary Dataset 1). We identified 8,730 significant differentially expressed genes with an FDR at 5%, of which 2,120 genes had an absolute log fold change above 1. We next performed gene set enrichment analyses (GSEA) to capture pathways perturbed towards both directions simultaneously using the 24,448 ranked genes identified in our dataset and annotated in ENSEMBL (version 94) against the KEGG pathways and the hallmarks gene set collection (MSigDB V6.2). Three KEGG pathways (proteasome, DNA replication and protein processing in the endoplasmic reticulum (ER)) and 10 of the 50 hallmarks gene set collection in MSigDB were statistically significant after correcting for multiple tests with an FDR at 5% (Supplementary Datasets 2 and 3). Induction of the UPR/ER-stress pathway was commonly identified using both types of analysis (Fig. 6b and Supplementary Fig. 13). We therefore pursued this further. It is well established that adaptation to protein-folding stress involves signals that go through distinct stress sensors located at the ER membrane³⁷. The most conserved UPR arm involves IRE1α, which, upon activation, uses its endoribonuclease activity to convert unspliced Xbp1 mRNA into functional XBP1, a multitasking transcription factor that regulates distinct sets of target genes in a cell type-specific manner³⁸. We confirmed through quantitative real-time PCR that Xbp1 and its downstream effectors Chop and Bip were upregulated in response to treatment with MS1943 in MDA-MB-468 cells starting at 4 h of treatment and that induction was sustained for at least two days (Supplementary Fig. 14). Of note, the processed/spliced Xbp1 transcript (Xbp1–207) that can result in active XBP1³⁹ was the only transcript that was significantly upregulated after treatment with the degrader as evidenced by our RNA-seq data (Fig. 6c). Taken together, these data suggest that EZH2 degradation could result in sustained overactivation of the UPR pathway in MS1943-sensitive cells due to prolonged ER stress, which in turn could be deleterious and lead to apoptosis. To test this hypothesis, we treated MDA-MB-468 cells (which are sensitive to MS1943) and MDA-MB-231 cells (which are insensitive to MS1943) with the ER-stress inducer tunicamycin and found that it effectively induced cell death in MDA-MB-468 cells but not in MDA-MB-231 cells (Fig. 6d and Supplementary Fig. 15). These results, together with the previous report that BT549 cells were sensitive to ER stress⁴⁰, support our hypothesis that MS1943-sensitive cells are susceptible to prolonged UPR activation. Furthermore, we examined the effects of both MS1943 and C24 on the UPR-related genes Xbp1, Chop and Bip in three MS1943-sensitive cell lines and one MS1943-insensitive cell line. We found that although C24 had no effect on the expression of these genes, MS1943 resulted in de-repression of these genes in all three MS1943-sensitive lines (MDA-MB-468, BT549 and HCC1187) but not in MDA-MB-231 cells that are insensitive to MS1943 (Fig. 6e,f). Collectively, these results suggest that MS1943 mediates its cytotoxic effects through ER stress and UPR induction in cells that are dependent for their growth on EZH2.

Fig. 6 |. MS1943 results in activation of the uPR pathway.

a, Heatmap showing differential gene expression patterns in MDA-MB-468 cells treated with 5 μM of MS1943 or DMSO for three days. Each column represents a different biological replicate. b, Genes induced upon activation of the UPR pathway (108 genes from MSigDB V6.2) are enriched for upregulation on treatment with MS1943 (24,448 ranked genes identified in our dataset and annotated in ENSEMBL (version 94) using GSEA). NES, normalized enrichment score. FDR q, the estimated statistical significance of the enrichment score using FDR. c, Xbp1–207, the transcript that encodes the functionally active isoform transcription factor XBP1 (S), was significantly induced on treatment with MS1943 in MDA-MB-468 cells. The transcript, Xbp1–201, is unspliced mRNA and encodes the isoform XBP1 (U). The other transcripts (Xbp1–202, Xbp1–203, Xbp1–204, Xbp1–205 and Xbp1–206) are less expressed in cells and with unknown function. Results are shown as individual dots for each isoform, treatment and independent experiment (n = 2 independent experiments for MS1943-treated MDA-MB-468 cells and two independent experiments for DMSO-treated MDA-MB-468 cells). d, MDA-MB-468 cells were highly sensitive to treatment with the ER-stress inducer tunicamycin in a live-cell imaging proliferation assay. Results are shown as mean ± s.d. from three independent experiments. e, Xbp1, Chop and Bip were significantly induced on treatment with MS1943 in three MS1943-sensitive TNBC cell lines (MDA-MB-468, BT549 and HCC1187) but not in MDA-MB-231 cells, which are insensitive to MS1943. Results are shown as individual dots for independent experiments and colors indicate the cell types. Results in black dots represent the mean and error bars indicate ±s.d. for four independent cell lines and for each gene (n = 3 independent experiments per cell line). P values of two-tailed Student’s t-tests with unequal variance between DMSO (mean = 1, s.d. = 0) and MS1943 treatment are provided for four cell lines. f, C24 had no effect on the expression of Xbp1, Chop and Bip in all four TNBC cell lines. Results are shown as individual dots for independent experiments and colors indicate cell types. Results in black dots represent the mean and error bars indicate ±s.d. for four independent cell lines and for each gene (n = 3 independent experiments per cell line). P values of two-tailed Student’s t-tests with unequal variance between DMSO (mean = 1, s.d. = 0) and C24 treatment are provided for four cell lines.

Discussion

EZH2 has been pursued by the biomedical community as a therapeutic target for treating sarcoma, lymphoma and MRT, but not TNBC, despite the fact that multiple studies have confirmed overexpression of EZH2 in TNBC and its association with increased tumor size and disease stage, and poor survival^19,41. In addition to contributing to tumorigenesis, EZH2 also appears to increase the risk of distant metastasis in patients with familial early-stage breast cancer⁴². The fact that EZH2 has not been pursued as a therapeutic target for treating TNBC is probably due to the observation that EZH2 inhibitors are ineffective in inhibiting TNBC cell growth. In this study, we confirm that EZH2 inhibitors including EPZ6438, CPI-1205, GSK126 and C24 failed to inhibit growth and induce apoptosis in TNBC cells, even though EZH2 KD was deleterious for the cells. These data suggest that non-canonical functions of EZH2 might be essential for the survival and growth of TNBC cells. Interestingly, several research groups have also alluded to such functions in prostate cancer^13–15.

To explore the possibility that pharmacological degradation of EZH2 could overcome the limitations of EZH2 inhibitors, we discovered a first-in-class EZH2 selective degrader, MS1943, which was highly selective for EZH2 over other methyltransferases and common drug targets. MS1943, but not C24, significantly reduced EZH2 protein levels in numerous TNBC and other cancer and non-cancerous cell lines, and effectively blocked proliferation of multiple TNBC and other cancer cell lines. MS1943 displayed little or no cytotoxicity in non-transformed MCF10A and PNT2 cells and several cancer cell lines, indicating that MS1943 is not a non-selective cytotoxic agent. Deletion of EZH2 in BT549 and MDA-MB-468 cells, which are sensitive to MS1943, resulted in significant cell growth inhibition, while EZH2 KO in MDA-MB-231 cells, which are insensitive to MS1943, did not. KD of EZH2 in MDA-MB-468 cells also significantly inhibited cell proliferation. Importantly, MS1943 completely suppressed tumor growth in vivo with on-target activity and PK/PD relationship established and was well tolerated by mice at the efficacious dose. Collectively, these results suggest that MS1943 has therapeutic potential in treating TNBC and several other cancers, and the antiproliferative activity of MS1943 is mainly due to its on-target EZH2 degradation effect. It is also worth noting that MS1943 is an effective, hydrophobic tag-based degrader that utilizes a non-covalent inhibitor to bind the target protein.

MS1943 as a hydrophobic tag-based EZH2 degrader, probably induces EZH2 misfolding or unfolding, similarly to Her3 misfolding or unfolding by the Her3 degrader³². To profile the transcriptome changes induced by MS1943, we performed RNA-seq and found that differentially expressed genes were associated with induction of the UPR, suggesting that the cells were experiencing ER stress due to the treatment. Indeed, we confirmed by rtPCR that MS1943 but not C24 induced prolonged ER stress in MS1943-sensitive but not in MS1943-insensitive TNBC cell lines, as evidenced by induction of Xbp1, Chop and Bip. We found that the ER-stress inducer tunicamycin effectively induced cell death in MS1943-sensitive but not in MS1943-insensitive TNBC cells. Importantly, even though activation of the UPR is meant to provide adaptation to abnormal accumulation of unfolded proteins, if the overload of unfolded or misfolded proteins in the ER is not resolved, the prolonged UPR will induce ER stress-associated programmed cell death⁴³. We show that C24 more effectively reduced H3K27me3 than MS1943, but only MS1943 induced apoptosis in MDA-MB-468 cells. These data suggest that the mechanism of action of MS1943 is not due, at least not solely, to inhibition of the enzymatic activity of EZH2.

Finally, it is worth noting that although the stapled peptide SAH-EZH2 was reported to disrupt the interaction between EZH2 and EED leading to EZH2 degradation⁴⁴, SAH-EZH2 also degrades EZH1, has poor cell permeability and is not suitable for in vivo studies. DZNep, a potent inhibitor of S-adenosyl-l-homocysteine hydrolase, was also reported to degrade EZH2 and PRC2⁴⁵. However, DZNep is a non-selective inhibitor of all S-adenosyl-l-methionine (SAM)-dependent methyltransferases^46,47. Gambogenic acid derivatives were reported to be EZH2 degraders⁴⁸. However, they are not selective for EZH2⁴⁹ and share several common features with pan-assay interference compounds (PAINS)⁵⁰. In contrast, MS1943 is selective for EZH2 over >100 methyltransferases (including EZH1) and other drug targets. It has robust on-target activity in multiple TNBC cell lines and is orally bioavailable and efficacious in vivo, with an established PK/PD relationship. MS1943 treatment phenocopies the effect of EZH2 KO or KD in sensitive and insensitive cell lines. Our results suggest that pharmacological degradation of EZH2 may offer a potential therapeutic strategy for TNBC. Furthermore, based on the oncogenic role of EZH2 in other cancers including prostate cancer⁴, we hypothesize that EZH2 degraders could be effective in treating these tumors. MS1943 is a valuable tool to test this therapeutic hypothesis.

Methods

Common reagents, antibodies and software.

Detailed information regarding the reagents, antibodies and software used in this study is provided in Supplementrary Table 3.

EZH2 methyltransferase inhibition assay.

In this biochemical assay, which monitors the transfer of a³H-labeled methyl group from the cofactor SAM to the substrate, the five-component PRC2 complex (EZH2, EED, SUZ12, RBAP48 and AEBP2) was used as the enzyme (5 nM). The concentration of the substrate core histone was 0.05 μM and the concentration of the cofactor SAM was 1 μM. The assay was performed in three technical replicates from two independent experiments by Reaction Biology Corp.

Selectivity assays.

Selectivity assays against other methyltransferases were performed by Reaction Biology Corp. using the same³H-labeled SAM assay format as described above. The EZH1 assay was performed in duplicates. Selectivity assays against 22 other methyltransferases were performed once with a range of 10 concentrations.

In the selectivity assays against kinases, MS1943 was tested at 10 μM. The compound enzyme inhibition effect was calculated as percent of inhibition of control enzyme activity. The assays were performed in duplicate by Eurofins Cerep.

Selectivity of MS1943 against the 45 GPCRs, ion channels and transporters was performed by NIMH-PDSP (http://pdsp.med.unc.edu/) in radioligand binding assays⁵¹. MS1943 was tested at 10 μM in duplicates.

Western blotting analysis.

Cells were collected and lysed in radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitor cocktail. After lysis, the concentration of protein was quantified by bicinchoninic acid (BCA) protein assay kit. Equal amounts of total protein were loaded onto 4–15% SDS–PAGE and separated. The gel was transferred to polyvinylidene difluoride membranes. Then, membranes were blocked for1 h in PBS blocking solution (LI-COR) and probed at 4 °C for 12 h with the following primary antibodies: EZH2 (Cell Signaling Technology), SUZ12 (Cell Signaling Technology), EED (Millipore), H3K27me3 (Cell Signaling Technology), H3K27me2 (Cell Signaling Technology), H3K27me (Cell Signaling Technology), H3K4me3 (Cell Signaling Technology), H3 (Cell Signaling Technology), EZH1 (Cell Signaling Technology) and actin (Proteintech). Membranes were then incubated with the specific rabbit IRDye 800CW secondary antibody (LI-COR) for 1 h at room temperature. After washing, the blot was detected using an Odyssey CLx Imaging system (LI-COR). Data analysis was performed using Image Studio (LI-COR). Source blot images are provided in Supplementary Figs. 16–22.

Mechanism of action studies.

For the study using MDA-MB-468 cells, cells were cultured for 24 h in the presence of MS1943 (5 μM), alone or with different concentrations of the proteasome inhibitor MG132, and EZH2 protein levels were detected via western blot as described already. Vinculin was used as a loading control to normalize EZH2 protein levels. For the study using HCC70 cells, cells were cultured for 6 h in the presence of MS1943 (4 μM), alone or with different concentrations of MG132, and EZH2 protein levels were detected via western blot as described before. Actin was used as a loading control to normalize EZH2 protein levels.

CRISPR/Cas9 knockout of EZH2 in MDA-MB-231, MDA-MB-468 and BT549 cells.

EZH2 KO cells were generated as described previously³⁵. Briefly, 3 × 10⁵ cells were seeded in six-well plates one day before infection. Cells were infected with lentiviruses (multiplicity of infection of 5 and 8 μg ml⁻¹ polybrene) harboring a lentiCRISPR V2 plasmid targeting either EZH2 or no target (empty vector control). Two days post infection, cells were selected with 5 μg ml⁻¹ puromycin for two days before beginning experiments.

• Target sequences:

• gRNA: TTATGATGGGAAAGTACACG

• Source of gRNA sequence: 10.1172/JCI90793

Characterization of EZH2 knockout cell lines.

In vitro cell viability assay.

Cells were seeded at 1 × 10³ cells per well in black 96-well plates (Corning) in triplicate. Viability was quantified using Cell Titer Glo (Promega), according to the manufacturer’s instructions, from one to eight days post seeding.

Western blotting.

Cells were collected (2 × 10⁶) four days post infection and lysed in 2× Laemmli buffer with 4% β-mercaptoethanol. Western blots were performed as described previously³⁵. Briefly, proteins were run on bis-tris gels and separated with SDS–PAGE. After incubation for 1 h at room temperature in 5% milk PBST (PBS + 1% Tween-20), gels were probed with antibodies against EZH2 at 1:1,000 (D2C9, Cell Signaling Technology) and β-actin at 1:10,000 (8H10D10, Cell Signaling Technology) overnight in 5% milk PBST. The next day, membranes were washed three times in PBST and incubated for 1 h at room temperature with anti-rabbit or anti-mouse horseradish peroxidase secondary antibody (GE Healthcare) in 5% milk PBST. After washing, membranes were exposed to detection reagent (Amersham) and visualized on film.

siRNA-mediated EZH2 knockdown.

For EZH2 knockdown experiments, siGENOME SMART pool against human EZH2 (cat. no. M-004218-03-0005) and non-targeting control pool (cat. no. D-001206-13-05) were purchased from Dharmacon and used according to the manufacturer’s instructions. For protein analysis, cells were incubated at 37 °C and 5% CO₂ for two days. For proliferation analysis, cells were incubated at 37 °C and 5% CO₂ overnight and the next day, 3 × 10³ cells were seeded into 96-well plates and growth was calculated as described below using the IncuCyte live-cell imaging system.

shRNA-mediated EZH2 knockdown.

PMD (VSVG)/pCMVΔ8.2/pLKO.1 plasmids were transfected into HEK293T cells for lentivirus packaging. Virus was harvested 48 h after transfection, then 2 ml of virus was used per 6 cm dish to infect MDA-MB-468 cells. At 24 h after infection, medium was changed to fresh medium with 2 μg ml⁻¹ of puromycin. Cells were selected in medium with puromycin for 48 h, and then the same number of shControl or shEZH2 cells, in nonuplicate, were seeded into 96-well plates for five days of growth. Cell viability was measured by MTT assay every day. Knockdown efficiency was examined by western blot.

The reagents were as follows: plasmids PMD (VSVG) and pCMVΔ8.2 (from C. M. Lee, Icahn School of Medicine at Mount Sinai); pLKO.1-shRNA Control (scramble) (from W. Ma, Memorial Sloan Kettering Cancer Center); pLKO.1-shRNA EZH2#1 (TRCN0000040074, Sigma); pLKO.1-shRNA EZH2#2 (TRCN0000040077, Sigma); pLKO.1-shRNA EZH2#3 (TRCN0000286227, Sigma). Puromycin (P8833) was from Sigma. Lipofectamine 3000 transfection reagent (L3000008) was purchased from Thermo Fisher Scientific.

Proliferation and apoptosis assays.

A total of 1 to 3 × 10³ cells were seeded in 96-well plates in duplicates and dosed at the indicated concentrations. Cells were then monitored using the IncuCyte live-cell imaging system (Essen BioScience), which was placed in a cell culture incubator operated at 37 °C and 5% CO₂. Cell confluence was determined using calculations derived from phase-contrast images. For KARPAS-422 and SUDHL8, 5 × 10³ cells were seeded in 96-well plates coated with 0.01% poly-l-ornithine solution (Sigma, P4957) and cell confluence was measured as described already. Tunicamycin was bought from Sigma-Aldrich (SML1287–1ml). For measurement of cell death, DRAQ7 (Cell Signaling, cat.no. 7406) at 1.5 μM was included in the medium and apoptotic red counts were measured using the IncuCyte FLR automated incubator microscope.

Cell viability assays.

Cells (1,000–5,000 cells per well) were seeded into 96-well microplates in triplicate. After 20 h, cells were treated with DMSO or indicated serial dilutions of compounds for three days. Cell viability was assessed using MTT or CCK-8 (Cell Counting Kit-8, WST-8). Briefly, 12 mM MTT (Thermo Fisher, M6494) was prepared in DPBS (Dulbecco’s PBS; Thermo Fisher, 14190250) or 1× solution of CCK-8 (Dojindo, CK04) was warmed up at room temperature. MTT or CCK-8 was then added to each well (20 μl per 150 μl medium) and the plates incubated at 37 °C for 3 h in the absence of light. The medium was then replaced with 200 μl of DMSO, and the plates incubated at 37 °C for 30 min for MTT assay. For CCK-8, no further process was needed after incubation. The absorbance of plates was read at 540 nm for MTT and 450 nm for CCK-8, while the wells were measured at 690 nm as reference using an Infinite F PLEX plate reader (TECAN). GI₅₀ values were analyzed using GraphPad Prism 6. Error bars represent ±s.d. for three or four independent experiments. For MDA-MB-468 cells treated with EZH2 inhibitors or MS1943, the statistical significance for each column in the plot was obtained by comparing the cell viability signal of every concentration with that of DMSO in each group using the two-tailed Student’s t-test.

Crystal violet assay.

Cells (5,000–20,000 per well) were seeded in 12-well plates. After 20 h, cells were treated with DMSO or indicated compounds for six days, and media containing compound were changed every three days. Cells were then fixed with glutaraldehyde (Sigma-Aldrich, G5882) (6.0% vol/vol) and stained with crystal violet (0.5% wt/vol). The assay was performed in at least two biological replicates.

RNA-seq studies.

We sequenced duplicated RNA samples (DMSO-treated versus MS1943-treated (5 μM for three days)), which were purified using Qiagen RNeasy Plus Mini Kit according to the protocols in the RNeasy Plus Mini Handbook published by Qiagen.

RNA-seq analysis.

RNA-seq libraries were constructed from the PolyA selected mRNA using the TruSeq RNA sample preparation guide (Illumina) and submitted for paired-end 100 base pairs sequencing on a HiSeq2500 system (Illumina) by the Genomics Core Facility at the Department of Genetics and Genomic Sciences at Icahn School of Medicine at Mount Sinai, following the manufacturer’s instructions.

After quality control of FASTQ files using the FASTQC tool (version 0.11.7) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), we trimmed low-quality bases (Phred < 10) and adapter sequences and then discarded short reads (length < 60 nt) using the bbduk tool (version 37.53) (https://jgi.doe.gov/data-and-tools/bbtools/bb-tools-user-guide/). When either forward or reverse of a paired-end read were discarded, we discarded the complete paired-end read. We quantified the expression of transcripts using cleaned paired-end reads with the Salmon tool (version 0.9.1)⁵² using the human reference transcriptome from The Cancer Genome Atlas (TCGA; GDC.h38 GENCODE v22)⁵³. We performed the differential expression between the two MDA-MB-468 cell lines treated with DMSO and the two MDA-MB-468 cell lines treated with EZH2 degrader at gene level using the R tximport library (version 1.6.0) on R (version 3.4.3)⁵⁴. We pre-filtered genes to keep only genes that had at least 10 reads in total (Ngene = 24,924) and then performed differential gene expression using the R DESeq2 library (version 1.81.1)⁵⁵. We identified as differential expressed between two conditions when the P value adjusted was at 5% and log₂(fold change) was more than 1. The heatmap was drawn using the R pheatmap library (version 1.0.10). We next performed GSEA to capture pathways perturbed towards both directions simultaneously using the 24,448 ranked genes identified in our dataset and annotated in ENSEMBL (version 94) against the KEGG pathways using R GAGE (version 2.28.2) and Pathview (version 1.18.2) libraries⁵⁶ and the hallmarks gene set collection (MSigDB V6.2) using GSEA^57,58. We also performed a differential transcript usage using Salmon quantification (trimmed mean of M-values (TMM) values) at transcript level for Xbp1 using two-tailed Student’s t-tests.

Statistical analysis.

All signaling analysis experiments were performed in at least two biological replicates. Proliferation assays were performed in three independent experiments with three technical replicates for each experiment, and the results are shown as the mean ± s.d. for three independent experiments. Apoptosis assays were performed in triplicates and the results are shown as the mean ± s.d. for three replicates. Cell viability assays were performed in three or four independent experiments with three technical replicates for each experiment, and the results are shown as the mean ± s.d. for three or four independent experiments. Two-tailed Student’s t-tests were used to test means between groups. Statistical comparison among groups in the xenograft study was carried out with Sidak’s multiple comparisons test (GraphPad Prism). Statistical comparison between paired samples was performed with a Wilcoxon test for paired samples (R version 3.5.1).

Quantitative real-time PCR.

MDA-MB-468 cells were treated with DMSO or 4 μM MS1943 for 4, 24 and 48 h. RNA was prepared from cells plated in six-well plates using the Qiagen RNeasy kit. Approximately 0.5 μg of RNA was used to generate cDNA with Superscript Reverse Transcriptase II (Invitrogen). Samples were analyzed using SYBR Green reagents and the AB7500 real-time system from Applied Biosystems. Experimental results are the average of three independent experiments. Expression levels were normalized to actin. To determine the relative abundance of RNA molecules, the ΔΔCT method was utilized. For amplification, we used the following primer pairs:

• Actin: forward, TCACCCACACTGTGCCCATCTACGA—reverse, AGCGGAACCGCTCATTGCCAATG

• Xbp1: forward, TGCTGAGTCCGCAGCAGGTG—reverse, GCTGGCAGGCTCTGGGGAAG

• Chop: forward, GCACCTCCCAGAGCCCTCACTCTCC—reverse, GTCTACTCCAAGCCTTCCCCCTGCG

• BiP: forward, CGAGGAGGAGGACAAGAAGG—reverse, CACCTTGAACGGCAAGAACT

Formulation for mouse pharmacokinetic studies.

For the i.p. formulation, the weighed quantity (20.52 mg) of MS1943 for i.p. dosing was added in a bottle. To this, a 0.185 ml volume of NMP (N-methyl-2-pyrrolidone), 0.741 ml of captisol (20% wt/vol), 0.741 ml of PEG-400 and 2.039 ml of normal saline were added, with continuous vortexing after each addition. The final formulation was vortexed for another 2 min to obtain a clear solution. For the p.o. formulation, a weighed quantity (53.22 mg) of MS1943 was added to a bottle. To this, a volume of 0.321 ml NMP, 0.160 ml of solutol HS-15, 0.641 ml of captisol (20% wt/vol), 0.641 ml of PEG-400 and 1.442 ml of normal saline were added, with continuous vortexing after each addition. The final formulation was vortexed for another 2 min to obtain a clear solution.

Mouse pharmacokinetic studies.

Male Swiss albino mice were dosed at 50 mg per kg body weight via i.p. injection or at 150 mg per kg via oral gavage. Blood samples (~60 μl) were collected from three mice at each time point at pre-dose, 0.08, 0.25, 0.5, 1, 2, 4, 8 and 24 h (i.p.) or at pre-dose, 0.25, 0.5, 1, 2, 4, 8 and 12 h (p.o.). Samples were collected into labeled microtubes containing K2EDTA solution (20% K2EDTA solution) as an anticoagulant. Plasma was immediately collected from the blood by centrifugation at 4,000 r.p.m. for 10 min at 4 ± 2 °C and stored below −70 °C until bioanalysis. Concentrations of MS1943 in mouse plasma were determined by a fit-for-purpose liquid chromatography (LC)-MS/MS method. The non-compartmental-analysis tool of Phoenix WinNonlin (version 6.3) was used to assess the PK parameters. Peak plasma concentration (C_max) and time to peak plasma concentration (T_max) were the observed values.

In vivo treatment study.

Approximately eight-week-old female BALB/c nude mice were injected subcutaneously with 1 × 10⁶ MDA-MB-468 cells in the right flank. Females bearing tumor grafts were randomized in groups of eight mice per group when tumor volumes reached ~100 mm³. Animals were treated daily with MS1943 at 150 mg per kg body weight (dissolved in 10% NMP, 10% captisol, 20% PEG-400, 60% normal saline) or vehicle by i.p. injection for 36 days. Tumor size was measured using calipers and tumor volume was calculated as follows: (long measurement × short measurement²) × 0.5. Tumor sizes were recorded every three days over the course of the studies. Animal body weights were measured before and after the end of the treatment. Experiments involving mice were performed according to Mount Sinai School of Medicine Institutional Animal Care and Use Committee-approved protocols. We have complied with all relevant ethical regulations for the experiments involving mice.

Tumor and plasma samples preparation and intratumor/plasma drug concentration analysis.

Mice were treated via i.p. with 150 mg per kg body weight of MS1943 once daily for seven days. On day 7, mice were euthanized ~2 h after the last treatment. For plasma preparation, ~200 μl of blood was collected in an Eppendorf tube pre-treated with EDTA. Samples were centrifuged at 2,000g for 15 min using a refrigerated centrifuge, and the resulting supernatant was transferred in a clean Eppendorf tube and stored at −20 °C. Tumors were harvested immediately after animals were euthanized and then cut into smaller specimens, snap-frozen in liquid nitrogen and stored at −20 °C. All tumor samples, except no. 1, were homogenized in 80:20 (vol/vol) water:acetonitrile at a 1:9 (wt/vol) ratio. Total homogenization dilution was 10×. The no. 1 tumor sample was homogenized at a 1:19 (wt/vol) ratio due to the small sample size; this sample had a 20× total homogenization dilution.

All tumor samples were diluted 5× in plasma and analyzed against plasma calibration curves. MS1943 concentrations in plasma and tumor samples were analyzed using LC-MS. A Mac Mod Ace C18 column (2.1 × 50 mm, 3 μm) was used for LC. Mobile phase A: 95:5:0.1 (vol/vol/vol) water:acetonitrile:formic acid. Mobile phase B: 50:50:0.1 (vol/vol/vol) methanol:acetonitrile:formic acid. API 5500 was used for MS/MS analysis. Electrospray was used for the ionization method (positive ion).

Immunohistochemistry.

Tumors were harvested 1 h after the last treatment, fixed overnight in 10% buffered formalin and then dehydrated and embedded in paraffin. Paraffin blocks were sectioned at 5 μm. Immunophenotyping was performed with primary antibodies against EZH2 (Cell Signaling, cat no. 5246), H3K27me3 (Millipore, cat no. 07–449), Ki-67 (Abcam, cat no. ab15580) and cleaved caspase-3 (Cell Signaling, cat no. 9661). Percentages of positive staining cells were estimated after counting positive and negative cells in three representative fields from at least two tumors treated with either vehicle or MS1943.

Reporting Summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.