EZH2 inhibition: it’s all about the context

Enhancer of zeste homolog 2 (EZH2) is a histone methyltransferase and subunit of the polycomb repressive complex 2 (PRC2), which functions as a chromatin regulator (1). Within the PRC2, EZH2 facilitates the trimethylation of histone H3K27 and is associated with transcriptional repression. EZH2 has other independent functions outside of its activity within PRC2, including methylating nonhistone protein substrates and acting as a transcriptional activator (2). In addition, EZH2 can cooperate with other epigenetic silencing mechanisms, including DNA methyltransferases and histone deacetylases, resulting in a complex interplay between these enzymes to regulate a large array of gene targets. By serving as this master regulator, EZH2 plays a role in cell fate determination and cellular differentiation, and high levels of EZH2 are associated with suppression of apoptosis, cellular proliferation, and survival (3).

Increased activity of EZH2 is also known to be pathogenic in several cancer types; high expression correlates with tumor aggressiveness and is essential for proliferation of cancer cell lines (4). In addition, EZH2 activity is increased where activity of the SWI/SNF complex is lost. The SWI/SNF complex is a multisubunit chromosome remodeler that exists in all normal cells (5). One such subunit, SMARCB1 (INI1) acts as a tumor suppressor gene, and germline loss-of-function alterations in SMARCB1 predispose patients to rhabdoid tumors (6). Mouse models with biallelic SMARCB1 loss have embryonic lethality, but heterozygous or conditional knockouts develop rhabdoid tumors as well (7). INI acts to oppose EZH2, so the loss of INI1 and subsequent deregulation of EZH2 has an oncogenic effect. This observation has now been extended to other members of the SWI/SNF complex, including ARID1A and SMARCA4 (8). Moreover, alterations in specific SWI/SNF complex genes are enriched in particular cancers, suggesting unique or cancer type–specific roles for individual components of this complex. This observation is also recapitulated in mouse models harboring alterations in individual SWI/SNF genes, where alterations in individual SWI/SNF complex components can confer vulnerability to different cancer types.

EZH2 inhibitors were developed as cancer therapeutics to target the epigenetic regulation of transcription in these various malignancies. An open question remains, however, as to whether targeting EZH2 can be an effective therapeutic strategy agnostic of tumor type. There are rare cases of targeted therapies, such as the TRK inhibitor larotrectinib, which demonstrates universal activity across tumor histologies, likely because TRK fusion–driven malignancies have otherwise simple genomes that lack concurrent drivers (9). In contrast, BRAF V600E–driven cancers often possess concurrent molecular alterations and lineage-specific genes that modulate the activity of dual RAF/MEK inhibition in different cancer types (10).

Tazemetostat was developed in adults to treat EZH2-dysregulated malignancies in multiple contexts. For example, tazemetostat resulted in clinically meaningful responses in patients with follicular lymphoma and remarkably showed a 69% overall response rate in EZH2-mutant patients (11). In adult epithelioid sarcoma, where all patients have INI loss, 15% of patients showed an objective response, and a median progression-free survival of 5.5 months was observed (12). Molecular analysis of these patients’ tumors showed multiple genetic mechanisms leading to INI1 loss, and further analysis yielded a tumor mutation burden of 25.8 variations/megabase, showing a greater genetic complexity than other INI-negative tumors. Conversely, where tazemetostat was trialed for rhabdoid tumors in adults (13), only 2 of 31 objective responses were seen. These rhabdoid tumors have few oncogenic drivers and no genomic instability, yet they do not respond as well as epithelioid sarcoma to EZH2 inhibition.

In this issue of the Journal, Chi and colleagues (14) report on arm C of the National Cancer Institute (NCI)-Children’s Oncology Group (COG) Pediatric Molecular Analysis for Therapy Choice (MATCH) trial, in which tazemetostat was used to treat pediatric patients with tumors harboring EZH2 hotspot mutations or alterations in the SWI/SNF pathway (SMARCB1 or SMARCA4 loss by immunohistochemistry), adding a chapter to this emerging story. The NCI-COG Pediatric MATCH, analogous to the adult NCI-MATCH trial, is a precision medicine trial designed to identify signals of efficacy to a targeted agent in a biomarker-selected but tumor type–agnostic population. Twenty patients were enrolled; the most common histologies were atypical teratoid rhabdoid tumor (ATRT; n = 8) and malignant rhabdoid tumor (MRT; n = 4) but also included an array of other diagnoses (Ewing sarcoma, epithelioid sarcoma, renal medullary sarcoma, hepatocellular carcinoma, non-Langerhans cell histiocytosis, and ependymoma). The most common enrolling genetic alterations were SMARCB1 loss (n = 16), EZH2 mutations (n = 3), and SMARCA4 loss (n = 1). Pediatric tumors are generally characterized as having low genomic complexity, and this trial was somewhat unusual in that 11 of the 20 treated patients were younger than 10 years of age, because ATRT and MRT are chiefly diseases that occur in very young children. The only objective response was seen in a 10-year-old patient with non-Langerhans cell histiocytosis with SMARCA4 loss. Five additional patients with SMARCB1 loss had a best response of stable disease for 6 or more months, including 2 of 8 patients with ATRT, 2 of 2 patients with epithelioid sarcoma, and the only patient with renal medullary carcinoma; none of the 4 patients with MRT derived clinical benefit. No new adverse events were observed compared with previous data on tazemetostat, and the drug was quite well tolerated.

These clinical data raise 2 clear questions: 1) why are response rates so low in these cancer types for this targeted therapy, and 2) what differentiates the few patients who do respond to treatment from the many patients who do not. Our first inclination to address the question is whether histology is the determining factor for sensitivity of EZH2-deficient tumors to tazemetostat. Even though SMARCB1-deficient tumors are uniformly genomically quiet tumors, the initial phase 1 pediatric study indicates important differences based on histologic subtype (15). Per tumor category in the dose expansion, overall response rates were 24% (5/21) in those with ATRT, 33% (2/6) in those with poorly differentiated chordoma, 22% (2/9) in those with epithelioid sarcoma, and 0% in those with non–central nervous system rhabdoid tumors (n = 21) and in tumors of other histology (n = 6). The 0% in non–central nervous system MRT is especially striking and was reproduced in both the NCI-COG Pediatric MATCH trial (although only 4 patients with rhabdoid tumors were enrolled) and in the adult experience. However, in the cohort of patients on this trial, only one objective response was seen in a rare proliferative entity with SMARCA4 loss, and as a whole, the patients gaining the most clinical benefit harbored SMARCB1-deficient tumors across disparate tumor types.

Of additional concern are the mechanisms of resistance to tazemetostat in this cohort of patients both at the time of progression on therapy and perhaps more importantly, intrinsically at the time of enrollment. Previously reported data show a variety of mechanisms of on- and off-target resistance to EZH2 inhibition. An analysis of patient-level data in lymphomas treated under the selective pressure of EZH2 inhibition revealed the development of EZH2 resistance mutations (16). Modeling data have also shown, however, that resistance to EZH2 inhibition can occur through bypass mechanisms, either through activation of the insulin-like growth factor receptor 1 (IGF-1R), MEK, or phosphatidylinositol 3-kinase pathways (17). Furthermore, newly emerging data within SMARCB1-deficient tumors have identified a diverse set of acquired mutations that converge on RB1/E2F, which is sufficient to confer resistance in these tumors (18). Finally, it has recently been shown that loss of the H3K36 methyltransferase NSD1 is yet another resistance mechanism to EZH2 inhibition in SMARCB1-deficient tumors (19). Taken together, these data expand our understanding of the landscape of de novo and acquired resistance to EZH2 inhibition and underscore the need to incorporate iterative genetic and epigenetic profiling into future studies to help identify hallmarks of susceptibility and resistance as these data have the potential to generate testable hypotheses for future combination therapies.

These clinical data from the NCI-COG Pediatric MATCH trial unfortunately show a relatively low objective response rate to tazemetostat as a single agent but serve as an inspiration for future research and trial design. Even when tumors appear to be genomically quiet, the “1 gene, 1 drug” hypothesis is increasingly the exception, not the rule in precision medicine. In the setting of our increasing knowledge of the pleiomorphic biology of EZH2 and the complexity of chromatin remodeling in cancer biology, it has now become imperative to use genetic and epigenetic profiling to identify constellations of susceptibility to EZH2 inhibition in individual patients with certain histologies, such as ATRT and epithelioid sarcoma, both at baseline and post-therapy. Moreover, it is a charge to better understand the biology of a disease such as non–CNS MRT, which may uncover new roles for SMARCB1 and the SWI/SNF pathway. As suggested by the authors, clinical response to alteration of genomic programs through chromatin remodeling may also be a slow process requiring more time than is typically afforded by rapidly proliferating tumors. The data presented here support future efforts to identify combinatorial strategies using EZH2 inhibition in biomarker-selected patients using both targeted and empiric cytotoxic therapies in the context of histology, which is no less precise.

Data availability

No new data were presented in this manuscript.

Author contributions

Ezra Y. Rosen, MD, PhD (Writing – original draft; Writing — review & editing), Neerav N. Shukla, MD (Writing — original draft; Writing — review & editing), Julia L. Glade Bender, MD (Writing — original draft; Writing — review & editing).

Funding

No funding was used for this editorial.

Conflicts of interest

E.Y.R. and N.N.S. have no disclosures to report. J.L.G.B. reports the following disclosures: she received grant support from NCI P30 CA008748 and NCI P50 CA217694. She also receives support from the COG, the consortium that sponsored this trial through NIH National Clinical Trials Network grants U10CA180886 and UM1CA228823. She was part of the leadership for the NCI MATCH trial and the principal investigator of a different subarm (subarm H; APEC1621H), directed at a different molecular target, but had no role in the planning of subarm C (APEC1621C). She has served as a paid consultant for Jazz Pharmaceuticals (limited to 1 pediatric advisory board), an uncompensated consultant on the data and safety monitoring board for Springworks, Merck, and Pfizer and an uncompensated consultant on pediatric advisory boards for BMS and Eisai. She receives institutional research support for clinical trials from Jazz Pharmaceuticals, Lilly, Eisai, Loxo-Oncology, Cellectar, and Bayer.