EZH2-Targeted Therapies in Cancer: Hype or a Reality

The field of cancer epigenetics has recently gained considerable interest due to a greater appreciation of the role of epigenetic genes in the progression of cancer, as well as our increasing ability to pharmacologically target these gene products. The term epigenetics was characterized by Conrad Hal Waddington in 1942, who described epigenetics as the heritable changes in phenotype without genotype alterations (1, 2). Although initially applied only to “heritable” changes, this definition has loosened over time to include the study of all modifications of chromatin or DNA that affect gene transcription independent of mutations in the genetic sequence.

Chromatin, the macro complex of DNA and histone proteins, is generally categorized into hetero- and euchromatin. Heterochromatin (aka restrictive chromatin) is the highly condensed form that prevents active transcription, whereas euchromatin (aka permissive chromatin) is open in configuration and amenable to active transcription. The functional unit is the nucleosome, which is composed of a histone octamer (two copies of each H2A, H2B, H3, and H4 proteins) with 145–147 base pairs of DNA wrapped around it. Chromatin/nucleosomes can be modified by chromatin-remodeling complexes, resulting in changes in gene accessibility at promoters and enhancers, and resultant transcriptional downstream effects (3–9). These complexes include the switch/sucrose nonfermentable (SWI/SNF) complex and the chromodomain helicase DNA-binding (CHD) protein family, and are frequently mutated in human cancers both at the germline and somatic levels. The SWI/SNF complex is composed of a central ATPase (BRG1 or BRM) as well as other proteins termed BRG1/BRM-associated factors (BAF) that are necessary for DNA and protein interactions (ARID1A, ARID1B, ARID2, PBRM1, SMARCD1, and SMARCE1). Many of these genes functionally interact with enhancer of zeste homolog 2 (EZH2; described in detail later).

DNA methylation is another mechanism of epigenetic regulation. Human cancers often exhibit abnormal methylation patterns, with promoter hypermethylation leading to gene suppression (in tumor suppressor genes) and genome-wide hypomethylation resulting in instability and activation of oncogenes (10). DNMT1, DNMT3A, and DNMT3B are examples of DNA methyltransferases, and DNA demethylases include TET1, TET2, and TET3. Mutations in all of these genes have been identified in various human cancers (11).

In addition to chromatin remodeling complexes and DNA methyltransferases/demethylases, chromatin structure and function can be regulated by histone-modifying enzymes (Supplementary Fig. S1). These enzymes catalyze a variety of posttranslational modifications including methylation, acetylation, phosphorylation, ubiquitination, and sumoylation (12). Both loss-of-function and gain-of-function mutations have been described in genes encoding for histone acetyltransferases/deacetylases and histone methyltransferases/demethylases.

Over 30 histone lysine acetyltransferases (HAT) are known. Truncating (inactivating) mutations in p300 and CBP (CREB-binding protein) occur frequently in hematologic malignancies like diffuse large B-cell lymphoma (DLBCL) or acute myeloid leukemia. As histone acetylation generally leads to more open chromatin and active transcription, loss of acetyltransferase activity is associated with general gene repression, which also involves many tumor-suppressor genes. Histone deacetylases (HDAC) on the other hand are frequently overexpressed in cancers and facilitate removal of histone acetyl groups and transcriptional repression of tumor-suppressor genes (13–15). Numerous HDAC inhibitors including belinostat, panobinostat, and vorinostat have been approved as anticancer therapeutics in a variety of hematologic malignancies (16–18). Importantly, histone acetylation and methylation can be functionally competitive and antagonistic as described below.

Histone methylation occurs on lysine and arginine residues in mono-, di-, and trimers with trimethylation generally regarded as the most mechanistically effective mark. Lysine methyltransferases (KMT) or protein arginine methyltransferases (PRMT) are the responsible enzyme classes and use S-adenosyl-methionine (SAM) as the methyl donor. Methylation of different amino acid residues on histone 3 is associated with distinct transcriptional effects. H3K4me2/3 is generally linked to transcriptional activation, whereas H3K9me2/3 and H3K27me2/3 are associated with transcriptional repression (19, 20). Importantly, histone methylation may prevent other marks such as acetylation, with resultant additive functional antagonism. Forty-nine histone methyltransferases are currently known in the human genome and include the H3K4 methyltransferase MLL and the H3K27 methyltransferase EZH2 (4, 21). Thus, the global and focal epigenetic transcriptional output depends on the end sum of the various activities of these various epigenetic modifiers (in addition to transcriptional activators and coactivators). It is critical to note, however, that the mechanisms that determine genetic locus specificity to these modifications are an area of intense active investigation and are likely to vary by cellular context.

EZH2 is overexpressed in numerous tumor entities including melanoma, ovarian, breast, endometrial, bladder, renal cell, lung, and liver cancer, and is associated with aggressive disease, leading to its classification as an oncogene (22–31). EZH2 overexpression leads to increases in H3K27me3, with repression of tumor-suppressor genes as well as genes that drive cellular differentiation including p16 and E-cadherin (among numerous others; ref. 32). In prostate cancer, EZH2 is significantly overexpressed in metastatic disease compared with localized cancer and benign prostatic tissue at both the transcript and protein levels. Furthermore, EZH2 plays a role in prostate cancer cell proliferation and depending on its phosphorylation status acts as a coactivator for transcription factors like the androgen receptor. Interestingly, this effect is independent of its function as a transcriptional repressor (33, 34).

The mechanisms by which EZH2 is overexpressed (or activated) in cancer have been carefully evaluated and include transcriptional and posttranscriptional mechanisms. Critical cancer-related transcription factors have been shown to activate EZH2 transcription including E2F, ELK1, HIF1α, and NF-κB (35–38). Furthermore, EZH2 is regulated by different miRNAs at the posttranscriptional level. miR-101 and miR-26a have been shown to decrease EZH2 expression, and therefore, downregulation of these miRNAs leads to EZH2 overexpression in different cancer types (39, 40). Finally, the activity of EZH2 has been shown to be inhibited by AKT-mediated phosphorylation, which can be repressed by MYC-mediated PTEN upregulation and resultant PI3K/AKT pathway inhibition (Fig. 1; ref. 41).

In a subset of hematologic malignancies on the other hand, EZH2 is predominantly altered by somatic genetic mutation, leading to either gain of function (mostly in lymphomas) or loss of function (predominantly in myeloid malignancies). Gain-of-function mutations have been associated with downregulation of tumor-suppressor and differentiation genes (42). Seven percent of follicular lymphomas and 22% of the DLBCLs of germinal-center origin harbor a somatic point mutation in the SET domain of EZH2 (mainly Y641), leading to enhanced H3K27me3 levels and silencing of genes important for B-cell differentiation and cell-cycle control (43, 44).

Loss-of-function mutations in EZH2 occur in 6% of myelodysplastic syndromes, 3% to 13% of myeloproliferative neoplasms, and 25% of T-cell acute lymphatic leukemia, suggesting a unique tumor-suppressor role for EZH2 in this subset of hematologic malignancies (45–47). As NOTCH1 is a main driver of this malignancy, NOTCH1 was found to antagonize the methyltransferase function of EZH2, and inhibition of EZH2 in the setting of NOTCH1 activation resulted in increased tumor progression (47). Studies are ongoing to better understand why EZH2 functions as a tumor suppressor in the setting of T cells. Perhaps T cells harbor a unique H3K27-repressive gene profile due to a unique epigenome, or the downstream signaling effects in T cells are fundamentally different. Regardless, these possible oncogenic effects of EZH2 inhibition in T cells will need to be carefully considered in clinical trials. In myelodysplastic syndromes, RUNX1, a transcription factor critical for hematopoiesis, is frequently mutated (10%–20%), and its mutations are significantly associated with −7/7q− chromosome anomalies. The EZH2 gene is located at 7q36, thus the RUNX1 alterations can lead to EZH2 loss of function (48, 49). In mouse models, EZH2 loss significantly promotes the transformation of hematopoietic stem cells expressing the RUNX1S291fs mutant into myelodysplastic cells (50). These findings suggest that EZH2 may act as an oncogene in most solid tumors and lymphomas, while acting as a tumor suppressor in limited minority of T-cell leukemias and myelodysplastic syndromes.

Polycomb-repressive complex 2 (PRC2) is a complex composed of either EZH1 or 2, EED (embryonic ectoderm development), SUZ12 (suppressor of zeste 12), and RbAP46/48 (retinoblastoma binding protein 46/48). The PRC2 complex catalyzes the methylation of H3K27, leading to chromatin compaction and gene repression (51). The mammalian genome encodes for two different EZH orthologs, EZH1 and EZH2. They are the SET domain [Su(var)3–9, enhancer of zeste, trithorax] containing subunits and therefore responsible for the catalytic methyltransferase activity. EZH1/2 are in an inactive state unless associated with EED and SUZ12 (52). The comparative methyltransferase activities of EZH1 and 2 have been intensely studied with the possibility of functional redundancy and compensation in mind. In embryonic stem cells, EZH2 knockdown results in reduced but not complete absence of global H3K27me3. In addition, inhibition of EZH2 via small-molecule inhibitors diminishes H3K27me3 to a great extent, but residual trimethylation can still be detected (53). Knockdown of EZH1 enzyme in EZH2-null cells results in elimination of any residual trimethylation due to EZH1, showing a role for EZH1 as well in H3K27 trimethylation (54). On the other hand, PRC2-EZH1 shows a lower histone methyltransferase activity, and its knockdown alone does not lead to global reduction of H3K27 methylation (55). Finally, EZH1 is present in dividing and differentiated cells, whereas EZH2 is only found in actively proliferating tissues (56). The preponderance of the evidence suggests that EZH2 is the predominant H3K27 methyltransferase in malignant cells, and EZH2-specific inhibitors have been developed with this in mind. However, the residual H3K27me3 after EZH2 inhibition provides a rationale for pursuing dual EZH2 and EZH1 inhibition.

How exactly PRC2 targets certain genes and not others is an area of active investigation. PRC2 seems to be present on CpG island promoters of untranscribed genes (57). Several groups have tried to identify proteins modulating PCR2 function and gene targeting. PHF1, MTF2, PHF19, Polycomb-like 1, -2 and -3, PALI1 (C10ORF12), EPOP (C17ORF96), JARID2, and AEBP2 are some of these proteins. However, further research is needed to illuminate these complex processes (58, 59).

Furthermore, noncatalytic and PRC-independent roles for EZH2 have been described. EZH2 interacts with a variety of transcription factors including GATA4, RORα, and androgen receptor in a PRC2-independent manner (34, 60, 61). EZH2 can directly methylate the natural killer T-cell lineage defining transcription factor PLZF (62). In stem-like glioblastoma multiforme cells, phosphorylation of EZH2 at serine 21 (pS21 EZH2) by AKT signaling facilitates methylation of STAT3 by EZH2, which enhances STAT3 activity (63). In small cell lung cancer (SCLC), EZH2 stabilizes DDB2 and promotes nucleotide excision repair. Consistent with this, EZH2 depletion but not catalytic inhibition sensitizes SCLC cells to cisplatin (64). In breast cancer cells, EZH2 interacts with the RelA/RelB complex coregulating a subset of NF-κB targets increasing the aggressiveness of breast cancer cells (65). In triple-negative breast cancer (TNBC), EZH2 protein binds to p38 MAP kinase, leading to accumulation in the cytoplasm of TNBC cells. In breast cancer tissue samples, cytoplasmic EZH2 phosphorylated at Thr 367 (pT327 EZH2) was associated with low H3K27me3 levels and a triple-negative phenotype. The pT367 EZH2 protein binds to cytoplasmic regulators of cell migration and invasion, and promotes metastasis (66). Taken together, a variety of noncatalytic functions of EZH2 have been described, but many are still not completely characterized. These noncatalytic functions support the rationale of identifying EZH2 small-molecule inhibitors, which function by inhibiting EZH2 protein interactions, without necessarily inhibiting catalytic activity.

Several EZH2 inhibitors have been developed within the last decade. 3-Deazaneplanocin A (DZNep) is an S-adenosylhomocysteine (SAH) hydrolase inhibitor leading to accumulation of SAH. SAH is the byproduct after methyl transfer from SAM, and buildup of this byproduct inhibits further SAM-mediated methyl transfer. DZNep induces cell-cycle genes and induces apoptosis in primary acute myeloid leukemia cells (67). It also depletes EZH2 protein levels, which results in a global decrease in H3K27me3. However, because other methyltransferases are dependent on SAM as a methyl donor, it also decreases other histone methylation marks including H3K4me3, H3K9me1/2/3, H3K36me3, H3K79me3, H3R2me2, and H4K20me3, and is thus not specific for EZH2 (68, 69). Other more specific inhibitors including GSK126 and EI1 have been tested in cell culture or xenografts. GSK126, a SAM-competitive, small-molecule inhibitor with a Ki value of approximately 0.5 nmol/L, was shown to be more than 1,000-fold more selective for EZH2 versus 20 other human methyltransferases (70, 71). GSK126 inhibits proliferation of multiple myeloma cells lines, lymphoma cells, and gastric cancer cells (among others; refs. 72–74). GSK126 also suppresses antitumor immunity by increasing myeloid-derived suppressor cells and decreasing CD4⁺ and INFγ⁺CD8⁺ cells (75). In mouse breast cancer models of BRCA1-deficient mice, GSK126 combination with cisplatin shows decreased cell proliferation and improves survival (76). EI1 is also an SAM-competitive inhibitor, which decreases H3K27 methylation without affecting other histone H3 methylation marks. DLBCL cells harboring the Y641-activating EZH2 mutation showed decreased proliferation, cell-cycle arrest, and apoptosis upon treatment with EI1 (77). Tazemetostat (EPZ-6438), another SAM-competitive inhibitor, shows a good oral bioavailability (78). It inhibits H3K27 methylation in EZH2 wild-type and mutant lymphoma cells (79). Other EZH2 SAM-competitive inhibitors including PF-06821497 (80), SHR2554, and CPI-1205 are currently in clinical trials. The latter has displayed potent antitumor effects in preclinical models (81).

In an attempt to identify small molecules that inhibit PRC2 through a different mechanism, protein–protein interaction inhibitors have been developed. MAK683 binds to the domain of EED, which interacts with H3K27me3, thus disrupting the interaction with EZH2 and decreasing methyltransferase activity. This inhibitor is currently under investigation in a phase I/II trial of patients with advanced malignancies including nasopharyngeal carcinoma, gastric cancer, ovarian cancer, prostate cancer, sarcoma, and DLBCL (clinicaltrials.gov NCT02900651).

Furthermore, dual EZH1 and EZH2 inhibitors have been developed. UNC1999 is an orally bioavailable EZH1/2 inhibitor with a 10-fold selective toward EZH2 (compared with >150-fold with GSK126). In MLL-rearranged leukemia, it suppresses tumor growth (82, 83). Valemetostat Tosylate (DS-3201b) is another EZH1/2 dual inhibitor that has demonstrated synthetic lethality in malignancies overexpressing EZH2 or harboring mutations to histone-modifying genes in preclinical models (Supplementary Fig. S2; refs. 53, 84).

Synthetic lethality describes the relationship between two genes or pathways in which loss of either one alone causes no or minimal phenotype, but loss of both leads to cell death. As loss-of-function defects in tumor-suppressor genes are difficult to therapeutically target, investigators have sought to identify biological vulnerabilities conferred by inactive tumor suppressors that could in turn be targeted through a synthetic lethality approach.

The SWI/SNF chromatin-remodeling complex modulates transcription and is essential for differentiation, proliferation, and DNA repair. Loss-of-function mutations in nearly every component of this 12–15 protein complex have been identified across numerous cancers (85). In fact, mutations in subunits of the SWI/SNF complex occur in 20% of all human cancers, rivaling the prevalence of other tumor suppressors such as p53 (86). The functionally antagonistic roles of the SWI/SNF complex and PRC2 are well known and evolutionarily conserved. Thus, targeting PRC2 complexes in SWI/SNF mutant cancer cells could result in synthetic lethality (Supplementary Fig. S3). This concept was first tested in SMARCB1 (aka SNF5 or INI1) deficient tumors, including pediatric rhabdoid tumors. Genetic loss of SMARCB1 leads to highly penetrant rhabdoid tumors and lymphomas in mice, and this phenotype can be rescued by genetic or pharmacologic inhibition of EZH2 (78, 87). These studies have led to the current phase I trial investigating the EZH2 inhibitor tazemetostat in pediatric rhabdoid tumors with SMARCB1 mutations (clinicaltrials.gov NCT02601937).

Mutations in other subunits of the SWI/SNF complex have been found to confer sensitivity to EZH2 inhibitors as well. Notably, ARID1A is the DNA-binding component of the SWI/SNF complex and is the most frequently mutated protein in the complex. In fact, it is the most commonly mutated epigenetic gene in all of human cancer with nearly 50% of ovarian clear cell carcinoma (OCCC), 20% of gastric cancer, 20% of bladder cancer, 14% of hepatocellular, 12% of melanoma, and 10% of colorectal cancers harboring inactivating ARID1A mutations (88). Notably, OCCC cells in culture that are normally resistant to EZH2 inhibition with GSK-126 become sensitive upon ARID1A knockdown (89). These cells' dependence on EZH2 was found to be related to EZH2-mediated transcriptional repression of PIK3IP1 (which decreases proliferation and promotes apoptosis by inhibiting PI3K-AKT signaling; ref. 89). In addition, a potential resistance mechanism was found in cells harboring concomitant activating mutations in the Ras pathway (90).

Furthermore, mutations in BRG1 (SMARCA4), an ATPase subunit of SWI/SNF, have been shown to confer sensitivity to EZH2 and Topo II inhibition in lung cancer cell lines (90). All of these preclinical data have led to the phase I and II investigations of EZH2 inhibitors (Tazemetostat) in solid tumors harboring somatic mutations in SWI/SNF components (91). Kadoch and colleagues have offered convincing molecular data describing the mechanisms of this process and showed that SWI/SNF machinery actively evicts PRC2 and EZH2 via its ATPase activity, thus decreasing the H3K27Me3 mark (presumably through histone turnover and competing demethylase activity) to allow active transcription (92). Thus, mutations in SWI/SNF members allow uninhibited PRC2-mediated gene silencing, which in cancer cells may predominantly consist of tumor suppressors (but may also orchestrate tumor immunity as described below). Inhibition of EZH2 then allows accumulation of tumor suppressors and cell death. Of course, the downstream transcriptional effects of this interplay are likely highly context dependent and can be affected by the presence and activity of demethylases, nucleosome turnover, and functionally redundant histone methyltransferases.

The histone lysine demethylase KDM6A (aka UTX) directly antagonizes EZH2/PRC2 by removing methyl marks on H3K27. Thus, it would be hypothesized to act as a tumor suppressor, to be mutated/inactivated in tumors cells, and to confer sensitivity to EZH2 inhibitors in its absence. KDM6A is found to harbor mutations in multiple tumor types including bladder cancer, multiple myeloma, pancreatic, and esophageal cancers (93, 94). In multiple myeloma, cell lines harboring mutations in KDM6A are exquisitely sensitive to EZH2 inhibition, whereas wild-type KDM6A cells are not. Importantly, KDM6A knockdown in wild-type cells or KDM6A reconstitution in KDM6A-mutant cells can alter EZH2 inhibitor sensitivity, both in vitro and in mouse xenografts (95). Similar findings were identified in bladder cancer cells (96).

Finally, mesothelioma cells lacking the H2aK119Ub deubiquitinase BAP1 are sensitive to EZH2 pharmacologic inhibition (97). The authors suggest that this is due to upregulation of EZH2 transcription via BAP1-mediated deubiquitination of a chromatin modulator that regulates the levels of H4K20me1 at the EZH2 locus.

Taking all the data above together, it is evident that many different cancer cell types are dependent on a dysregulation of one or more components of the epigenetic machinery, which in turn leads to dependence on PRC2/EZH2 activity that can be targeted pharmacologically via a synthetic lethality approach.

In addition to functioning as an oncogene in tumor cells, EZH2 is also critical in coordinating the immune response to the tumor via transcriptional modulation both in tumor cells and immune cells.

As EZH2 is a master epigenetic regulator of transcription, its effects are dependent upon the transcriptional program that it modulates, which can be highly context- and cell type–dependent. In cancer cells, not only does EZH2 work to repress the expression of tumor-suppressor genes as described above, but can also modulate the expression of genes involved in orchestrating the tumor immune microenvironment (98). Peng and colleagues determined that EZH2-mediated transcriptional silencing of the cytokines CXCL9 and CXCL10 prevents efficient T-cell trafficking in ovarian cancer cell xenograft models, and can be reversed by EZH2 inhibition in combination with anti–PD-L1 checkpoint blockade (99). Furthermore, outcomes of patients with ovarian cancer negatively correlate with EZH2 expression and tumor-infiltrating CD8⁺ T cells. Similar findings have been described for colon cancer (100). The same group recently discovered that ARID1A mutations result in higher EZH2-dependent immunosilencing via CXCL9/10, which is resistant to PD-L1 blockade in colon and ovarian cancer xenograft models (101). However, the effect of combination therapy with EZH2 inhibitors and immunotherapy was not reported. Zingg and colleagues uncovered an interesting interplay between infiltrating tumor immune cells and melanoma cancer cell EZH2 expression (102). They showed that CD8⁺ T cells in the immune infiltrate secrete TNFα, leading to upregulation of EZH2 in tumor cells. This EZH2 upregulation results in transcriptional silencing of genes involved in immunogenicity and antigen presentation, which could be reversed by EZH2 inhibition in conjunction with CTLA4 and IL2 immunotherapy. Importantly, because PD-L1 is downregulated in these cells, combination therapy with anti–PD-L1 produced no additive response.

In addition to its role in cancer cells, EZH2 has been shown to play a critical role in immune cell differentiation and function, and specifically in antitumor immunity (98). Although EZH2 has been investigated in all immune cell lineages including natural killer cells, dendritic cells, B lymphocytes, and macrophages, its role in T-cell biology has been the most well-characterized (103). Notably, although EZH2 appears to prevent naïve CD4 cells from differentiating into Th1/Th2 effector cells, it also appears to promote CD8⁺ effector cell survival and regulatory T cell (Treg) development. Because naïve CD4 cells and Tregs generally function to repress cancer immunity, whereas CD8⁺ effector T cells promote it, the implications of systemic EZH2 inhibition for cancer immunotherapy could be highly context dependent.

Recently, the role of EZH2 in maintaining the cancer immune-inhibitory phenotype of Tregs has been intensively studied. Goswami and colleagues showed that EZH2 inhibition with CPI-1205 slowed bladder tumor xenograft growth in immunocompetent mice, and this was therapeutically additive with anti–CTLA4 therapy. Specifically, mice with EZH2-deficient Tregs were more sensitive to anti–CTLA4 therapy than wild-type mice, and Rag1^−/− mice (lymphocyte deficient) did not respond to EZH2 inhibition (104). Similar results were found by Wang and colleagues, who showed that Treg-specific EZH2 deletion was sufficient to inhibit tumor growth in immunocompetent mice with colon, prostate, and melanoma tumors (105).

However, Zhao and colleagues focused on EZH2 expression in CD8⁺ effector T cells and found EZH2 to be necessary for their survival and antitumor immunity (106). In both melanoma and ovarian cancer cell models, mice with T cells with pharmacologically or genetically inhibited EZH2 showed higher tumor burden. Furthermore, correlative studies in humans showed that patients with increased EZH2⁺CD8 T cells in their tumors had favorable clinical outcomes, presumably due to a more active tumor immune microenvironment. These divergent functions of EZH2 in different T-cell subsets will need to be considered in design of any human clinical trials. Notably, there are currently at least 3 phase I clinical trials investigating the synergy of EZH2 inhibition with immunotherapies.

On January 23, 2020, the EZH2 inhibitor tazemetostat (Tazverik; Epizyme) was approved by the FDA for treatment of patients aged ≥ 16 years with metastatic or locally advanced epithelioid sarcoma not eligible for complete resection. In preclinical studies, tazemetostat showed antitumor activity in cell culture and xenograft models (78, 79). The accelerated FDA approval of tazemetostat was based on a single-arm clinical trial (NCT02601950). The study included 62 patients with 9 patients (15%) showing a response with 1 (1.6%) complete response and 8 (13%) partial responses. Of these 9 patients, 6 had responses that lasted 6 months or longer (107, 108).

The first clinical trial of tazemetostat (NCT01897571) in patients with B-cell non-Hodgkin lymphoma (21 patients) and advanced solid tumors (43 patients) conducted between June 13, 2013, and September 21, 2016, revealed tolerable side effects including asthenia [21 (33%)], anemia [9 (14%)], anorexia [4 (6%)], muscle spasms [9 (14%)], nausea [13 (20%)], and vomiting [6 (9%)]. Durable objective responses, including complete responses, were observed in 8 (38%) of 21 patients with B-cell non-Hodgkin lymphoma and 2 (5%) of 43 patients with solid tumors. None of the patients died of treatment-related causes, and 7 (11%) patients experienced non–treatment-related deaths (91).

An open‐label, multicenter, phase II study evaluated tazemetostat in patients with relapsed or refractory DLBCL or follicular lymphoma (grades 1–3b) harboring either wild-type EZH2 or mutationally hyperactive EZH2. Interim data as of February 1, 2019, on the 95 patients with follicular lymphoma were as follows. Forty-one patients harbored EZH2 active-mutant tumors and 54 were EZH2 wild-type. The objective response rate in the EZH2-mutant groups with 39 patients being evaluable was 74% [95% confidence interval (CI), 57.9–87.0] with a complete response rate of 10%, partial response of 64%, and stable disease of 26%. None of the patients showed progressive disease. The median progression-free survival was 60 weeks (95% CI, 46.7–83.9). In patients with EZH2 wild-type tumors, 53 were evaluable. The objective response rate was 34% (95% CI, 21.5–48.3) with a complete response rate of 6%, partial response of 28%, and stable disease of 30%. Twenty-eight percent of patients showed progressive disease. The median progression-free survival was 24.6 weeks (95% CI, 15.1–47.9). Response duration was greater than 24 weeks in 83% of patients, and 50% of patients had a response duration greater than 1 year. Treatment‐emergent adverse events (TEAE) leading to dose reductions occurred in 8% of patients, and study drug discontinuation due to TEAEs occurred in 9% (109). Interim data as of May 2018 on 226 patients with DLBCL were as follows. Thirty-six patients with EZH2-mutant tumors were treated with tazemetostat monotherapy. The objective response rate was 17% (95% CI, 6–33) with a complete response rate of 3% and a partial response rate of 14%. Progressive disease was seen in 39% of patients. The median progression-free survival was 15.9 weeks (range, 0.1–84.1 weeks). In the 121 patients containing EZH2 wild-type cancers who received tazemetostat monotherapy, the objective response rate was 17% (95% CI, 10–25) with a complete response rate of 9% and a partial response rate of 7%. Progressive disease was seen in 60% of patients. The median progression-free survival was 8 weeks (range, 0.1–103.9 weeks; ref. 110). Taken together, tazemetostat demonstrated enhanced activity in lymphomas harboring EZH2-activating mutations, with acceptable tolerability.

Another phase 2, multicenter study evaluated tazemetostat monotherapy in adults with relapsed or refractory malignant mesothelioma with BAP1 inactivation (NCT02860286). Seventy-four patients were enrolled. All had at least one prior systemic therapy. Thirty-one patients (51%) achieved disease control at 12 weeks, and 15 patients (25%) sustained disease control (CR or PR+SD) at 24 weeks. Two of 61 patients had a confirmed partial response. None of the patients discontinued due to adverse events; however, 5 patients underwent dose reductions due to adverse events. The most frequently observed adverse events include fatigue (32%), decreased appetite (28%), dyspnea (28%), and nausea (27%; completed clinical trials are summarized in Supplementary Table S1; ref. 111).

In addition to monotherapy, tazemetostat is being evaluated in combination with other drugs including PD-L1 and PD-1 inhibitors, antiandrogens, or antibodies against CD20. One example includes a phase Ib clinical trial on patients with relapsed/refractory DLBCL treated with atezolizumab (anti–PD-L1 antibody) and tazemetostat (NCT02220842). As of August 28, 2018, 43 patients were enrolled. Median progression‐free survival was 1.9 months (95% CI, 1.8–2.8). Best overall response rate was 16% [complete response, n = 2 (5%); partial response, n = 5 (12%)]. Eight (19%) patients had stable disease; 19 (44%) had progressive disease. In 5 patients (17%), EZH2 mutations were identified. At least one adverse event occurred in 95% of patients, with anemia (26%) and fatigue (23%) being the most frequent ones. Six patients (14%) had adverse events, leading to discontinuation of either study drug (atezolizumab, n = 3; tazemetostat, n = 3). Two grade 5 adverse events occurred: sepsis, deemed unrelated to treatment; and hyponatremia, related to both study treatments (112).

Further ongoing clinical trials are summarized in Table 1. They include the evaluation of tazemetostat as a single agent in B-cell lymphomas, advanced solid tumors, malignant mesotheliomas, rhabdoid tumors, and sarcomas. Further, tazemetostat is being investigated in combination with antiandrogens, checkpoint inhibitors, chemotherapy, and targeted therapeutics like rituximab. Other EZH2 inhibitors including CPI-1205, PF-06821497, SHR2554, and Valemetostat Tosylate are likewise currently being investigated in clinical trials. The preliminary data above suggest that patient selection using molecular biomarkers such as EZH2 activating mutations and/or synthetically lethal mutations like those described above will be crucial to optimize clinical outcomes.

EZH2 overexpression and mutation play a critical role in the development, progression, and metastasis of many types of cancer. Extensive research in the last 20 years on the histone methyltransferase EZH2 in cancer has led to the development of different inhibitors that target both wild-type and mutant EZH2, as well as some that target the PRC2 complex member EED. EZH2 now serves as a target for precision medicine, and many studies have shown that EZH2 is a target for synthetic lethality and hence serves as a valuable drug in cancers with specific mutations in trithorax/SWI/SNF family members like BAP1 and ARID1A (88–91, 111).

EZH2 inhibition alone may not be highly effective in certain tumors, however, and combination with other drugs and immunotherapies may prove beneficial in cancers that are not sensitive to EZH2 inhibition alone. For example, activation of IGF-1R, PI3K, and MEK pathways can lead to resistance to EZH2 inhibitors. Furthermore, some mutations in the EZH2 gene prevent small molecular inhibitors from binding to EZH2. Other acquired resistance mechanisms include inhibiting proapoptotic proteins TNFSF10 and BAD via a FOXO3A-dependent mechanism (113).

Therefore, identification of additional mutations that confer EZH2 dependence to tumors could increase patient and tumor eligibility. Although our ability to predict EZH2 sensitivity is improving, more work is required to develop highly predictive biomarkers for EZH2 therapeutic response. In addition, further combination trials will be crucial to maximize therapeutic benefit, while keeping toxicity tolerable. Overall, EZH2 is one molecular target that has elicited tremendous interest both in tumor biologists with a focus on epigenetics, and in clinical oncologists. With increased testing and trials, the future of targeting EZH2 to treat specific type of cancers either as a single agent or in combination is highly promising.

No potential conflicts of interest were disclosed.

Dr. S. Varambally is supported by Department of Defense funding (W81XWH1910588). Part of this work is supported by supplemental funding from U54CA118948 and startup funding from UAB to S. Varambally.