Developing EZH2-Targeted Therapy for Lung Cancer

Cancer cell growth depends upon selective gene expression. Although mutations in “driver” oncogenes and tumor-suppressor genes can alter gene expression, other important mechanisms involve changes to the cancer epigenome. If these changes can be identified and thwarted, such approaches would represent important new tools for the early detection, prevention, and rational treatment of cancers, such as lung cancer, the biggest cancer killer of men and women in the United States. Zhang and colleagues address one important such mechanism (and target) involving EZH2 of the polycomb repressive complex 2 (PRC2; ref. 1). As part of their study, they developed genetically engineered mouse models of lung cancer driven by overexpression of EZH2 in target tissues and examined human lung cancer lines and databases to study lung adenocarcinomas that greatly overexpress EZH2. EZH2 was then targeted genetically and chemically with the new open-source EZH2 inhibitor probe JQEZ5 that they developed, which demonstrated anticancer effects. Although the compound (or its derivatives) will need additional refinement for optimal clinical applications, it serves as an excellent probe for research and reinforces EZH2 as an important target in lung adenocarcinoma.

Chromatin-associated complexes can alter chromosomal DNA or its associated histones, leading to altered transcription. PRC2 is an assembly of polycomb group (PcG) proteins including embryonic ectoderm development protein (EED), suppressor of zeste protein (SUZ12), retinoblastoma associated protein (RbAp46/48) and the enhancer of zeste homolog polypeptide (EZH1 or EZH2; ref. 2). EZH2, but not EZH1, is expressed in proliferating cells. The 746–amino acid EZH2 has domains providing scaffolding for PRC2 proteins and a C-terminal Su(var)3-9 enhancer of zeste trithorax (SET) enzymatic domain (3–5). Chemical biology is at the heart of Zhang and colleagues' work, requiring understanding of the enzymology involved. The EZH2 SET domain has S-adenosyl-methionine (SAM) and H3 histone N-terminal tail lysine binding sites connected by a narrow channel for methyl transfer. After the binding of positively charged SAM and positively charged histone tail lysine to SET, water-mediated deprotonation of the H3K27 ϵ-amino group occurs at the active site (6). The H3K27 ϵ-amino group then performs a nucleophilic attack on the methyl donor SAM. The methyl group from SAM is transferred to the ϵ-amino group of the substrate lysine by an S_N2 substitution reaction sequentially producing S-adenosyl-homocysteine (SAH) and H3K27me3 (Fig. 1A). Notably, PRC2 contains an unusual split catalytic domain consisting of the SET activation loop (SAL) from the N-terminal portion of EZH2 and the SET domain at its C-terminus (4).

H3K27me3-marked chromatin recruits PRC1 complex with H2AK119 monoubiquitination activity, DNA methyltransferases that methylate CpG islands, and histone deacetylase–containing complexes. The result is steric hindrance for transcription machinery binding, chromatin compaction, and gene silencing, which leads to changes in gene expression in cancer. Although nonenzymatic transcriptional activator functions of EZH2 have been observed in some breast and prostate cancers, most research has analyzed the impact of altered H3K27me3. Global H3K27me3 increase leads to blocks in expression of differentiation genes and tumor-suppressor genes and malignant transformation. Zhang and colleagues demonstrated shifts in superenhancer signatures from H3K27ac to H3K27me3 by chromatin immunoprecipitation sequencing with associated decreases in gene expression in these regions by RNA sequencing (1).

Prior definitive evidence for the role of EZH2 in carcinogenesis leading up to the current study emerged with the discovery of activating mutations in the SET domain of EZH2 in B-cell lymphomas, overexpression of EZH2 in lung adenocarcinomas, and mutational inactivation of UTX histone demethylase and of the SWI/SNF complex in over 20% of cancers (7, 8). UTX mutations occur in the JmjC catalytic domain, yielding increased H3K27me3. SWI/SNF is an ATP-dependent antagonist of PRC2 gene repression such that mutational loss of one of its components—ARID1A, PBRM1, SMARCB1, or SMARCA4—also leads to increased H3K27me3 marks. These classes of mutations mimic EZH2 overactivity. As shown by Zhang and colleagues, EZH2 shRNAs inhibit growth of preclinical mouse transgenic and human lung cancer xenograft models and thus confirm the importance of EZH2 in the establishment and preservation of EZH2-positive (but pAKT-negative and pERK-negative) lung adenocarcinomas.

Because of the identification of EZH2 pathway mutations, previously high-throughput screens had been used to identify SAM-competitive inhibitors of EZH2 (9). Analogues were optimized for oral bioavailability and improved pharmacokinetics, and three—tazemetostat, GSK2816126, and CPI-1205—have entered clinical trials (NCT01897571, NCT02395601, NCT02601937, NCT02601950, and NCT02082977; Fig. 1B). The first two have a pyridine-amide core, and the last has a tetramethylpiperidinyl benzamide framework. Three other reports describe pyridine-amide–based EZH2 inhibitor chemical probes for preclinical experiments—inhibitor 1, EI1, and JQEZ5 (1, 3, 10). The structure of the latter differs from G2816126 by four methyl groups (Fig. 1B). Both cocrystallization X-ray structures of inhibitor with a human EED, human SUZ12(VEFS) domain and engineered American chameleon EZH2 subunits and in silico docking with EZH2 homology models built on the X-ray crystal structure of the H3K9 lysine methyltransferases G9a-like protein were used. A step-by-step description of “hit” optimization was reported for tazemetostat (11). The pyridine-amide type of inhibitor binds to the unique split catalytic domain of PRC2 on the interface between the EZH2 SAL and SET domains, which accounts for the remarkable selectivity displayed by these inhibitors (Fig. 1C; ref. 3).

Clinical results with three EZH2 inhibitors (tazemetostat, CPI-1205, and GSK2816126) are preliminary but encouraging. Tazemetostat was associated with cytopenias, hypertension, anorexia, and transaminasemia with a recommended phase II dose of 800 mg orally twice daily. Thirteen of 47 evaluable patients with non-Hodgkin lymphoma (NHL) achieved objective responses, including four complete remissions and nine partial remissions (12). Complete and partial responses have also been achieved in patients with malignant rhabdoid tumors. CPI-1205 has been administered orally in a dose-escalation schedule twice daily, with low-grade diarrhea as a side effect, and has led to disease stabilization in patients with B-cell NHL. Finally, GSK2816126 has been given on a dose-escalation twice-weekly i.v. schedule to patients with NHL, multiple myeloma, or solid tumors and has been well tolerated to date (13).

Patient selection based on likely tumor EZH2 dependency will be critical for the first phase of lung cancer clinical trials, and analyses for non–small cell lung cancer tumors using Clinical Laboratory Improvement Amendments–certified molecular tests that are becoming routine will facilitate this. Similarly, lymphomas with EZH2^Y641 mutations, UTX histone demethylase mutations, and tumors with SWI/SNF core subunit inactivation should be more likely to respond—but time will tell. Resistance mechanisms need to be studied in preclinical models and should be anticipated, such as secondary mutations in EZH2 or the development of non-PRC2 MAPK pathway or AKT pathway activation (8, 14). Solutions to overcome such resistance may include combinations of EZH2 inhibitors or combinations with cytotoxic chemotherapy or other epigenetics or tyrosine kinase–targeted agents. The study by Zhang and colleagues indicates that EZH2 inhibitors will likely be at center stage of epigenetic cancer therapy development, and point the way to new opportunities and provide important reagents for further preclinical studies.

No potential conflicts of interest were disclosed.

X. Liu was supported by Welch Foundation research grant I-1790, CPRIT research grant R1119, a Rita Allen Foundation research grant, and NIH grant GM114576. J.D. Minna was supported by NCI SPORE P50CA70907, CTD2N CA176284, and CPRIT RP110708.