Epigenetic Rewiring Underlies SMARCA4-Dependent Maintenance of Progenitor State in Pediatric H3K27M Diffuse Midline Glioma

Pediatric diffuse midline gliomas (DMG) driven by lysine 27 to methionine mutations in histone H3.1 and histone H3.3 (collectively noted here as H3K27M DMG) arise most commonly in the pons, precluding surgical resection. Chemotherapy is ineffective. All patients ultimately succumb to their disease after tumors become refractory to standard-of-care radiotherapy. Epigenetic dysregulation in H3K27M DMG leads to a stalled developmental state resembling that of highly proliferative, stem-like oligodendrocyte precursor cells (OPC; ref. 1).

The H3K27M mutation rewires the epigenome by inhibiting polycomb repressive complex 2 activity, globally reducing repressive H3K27 trimethylation (H3K27me3), and increasing activating H3K27 acetylation (H3K27ac; refs. 2, 3). However, H3K27me3 at promoters of key PRC2 target genes is required for tumor proliferation and maintenance (2, 3). This argues for a model in which the H3K27M mutation alters the epigenetic landscape and creates novel dependencies that may be targeted therapeutically.

In this issue of Cancer Discovery, Mo and colleagues (4) and Panditharatna and colleagues (5) independently identified BRG1, the catalytic subunit of the mammalian SWI/SNF (BAF) complex, as a dependency in H3K27M DMG through focused CRISPR/Cas9 screening of epigenetic regulators. To elucidate how BRG1 loss sensitizes H3K27M DMG, they per­formed cell viability and colony formation assays in H3K27M DMG cell lines, observing that ablation of BRG1 reduced cell viability and the ability to form colonies. BRG1 loss also led to reduced proliferation and increased apoptosis. Exogenous expression of BRG1 rescued the proliferation defects, providing evidence that BRG1 is essential for the proliferation of H3K27M DMG cells in vitro. In vivo, single-guide RNA–mediated depletion of BRG1 in orthotopically implanted DMG cells reduced tumor growth and improved survival.

To delineate molecular mechanisms underlying BRG1-mediated tumorigenesis and maintenance of the OPC state, the authors analyzed SMARCA4-driven gene regulatory networks using single-cell regulatory network inference and clustering (SCENIC) analysis. SCENIC combines correlative expression analysis of transcriptional regulators and putative target genes with regulatory sequence analysis to define gene regulons or modules (6). SCENIC identified SMARCA4 regulons in the OPC, oligodendrocyte cell, and cycling cell states.

Hypothesizing that BRG1 remodels chromatin to maintain the OPC state, Panditharatna and colleagues performed chromatin immunoprecipitation followed by DNA sequencing for BRG1, H3K27ac, and H3K27me3 in patient-derived DMG neurospheres. BRG1 binding sites partially overlapped with H3K27ac peaks and did not overlap with H3K27me3 peaks, suggesting that BRG1 localizes to active enhancer regions. Assessing chromatin binding of BRG1 in DMG cell lines by cleavage under targets and release using nuclease (CUT&RUN; ref. 7) followed by next-generation sequencing, Mo and colleagues confirmed that BRG1 peaks localized to enhancer (high H3K27ac, low H3K4me3) and promoter (high H3K4me3) regions. Because BRG1 is part of a chromatin remodeling complex, the authors investigated the effect of BRG1 loss on chromatin accessibility. After the genetic depletion of SMARCA4, transposase-accessible chromatin using sequencing demonstrated decreased accessibility at BRG1 binding sites, specifically at OPC-like H3K27M DMG markers. These data indicated that BRG1 regulates chromatin accessibility and binds regulatory elements to maintain an OPC-like program.

Mo and colleagues then sought to identify the BRG1-dependent genetic pathways that could drive tumorigenesis. Combining BRG1 CUT&RUN data, RNA sequencing of genes downregulated upon BRG1 depletion, and gene ontology analysis, the authors found that downregulated genes with BRG1 peaks in their promoters and enhancers enriched for extracellular matrix (ECM) and cell proliferation programs. These data argued for a model in which BRG1 binds to gene regulatory elements to direct expression of genes involved in cell proliferation and ECM pathways.

To gain insight into how BRG1 is recruited to chromatin to drive changes in gene expression in H3K27M DMG, the authors analyzed DNA sequence motifs of BRG1 binding sites in DMG cell lines. They found that BRG1 peaks are most highly enriched for a set of transcription factors, including SOX10, which is expressed in neural crest cells during early development and is required for oligodendrocyte specification and differentiation (8). By coimmunoprecipitation, they found that BRG1 interacted with SOX10 in DMG cells and CUT&RUN data showed that BRG1 and SOX10 peaks colocalized. Depletion of SOX10 from DMG cells downregulated the expression of ECM genes and reduced the ability of these cells to proliferate, migrate, and invade. Mo and colleagues additionally depleted H3.3K27M in DMG cells. These experiments demonstrated that the dependence of DMG cells on BRG1 was due to the H3.3K27M mutation. They did not perform depletion experiments in H3.1K27M DMG cells. Future work will be required to determine a mechanistic link between H3.1K27M mutation and BRG1 in DMG cells. Collectively, these data suggest that in H3K27M-mutant DMG, SOX10 recruits BRG1 to chromatin to direct expression of ECM and cell proliferation genes, paving the way for the therapeutic targeting of BRG1 as a rational treatment strategy.

To test the efficacy of pharmacologic inhibition of BRG1, the authors treated DMG cells with dual ATPase inhibitors of BRG1 (encoded by SMARCA4) and BRM (a second SWI/SNF ATPase encoded by SMARCA2). Treatment of human H3K27M DMG cells reduced cell viability, ability to form colonies, and cell proliferation. Importantly, combined treatment with radiation and an ATPase inhibitor significantly reduced the viability of DMG cells more than either treatment alone. As an additional therapeutic strategy, Panditharatna and colleagues utilized heterobifunctional proteolysis-targeting chimera (PROTAC) molecules to target BRG1 for proteosomal degradation. PROTAC treatment resulted in reduced proliferation and increased apoptosis in vitro.

Building upon their in vitro studies, they assessed the utility of pharmacologic inhibition or degradation of BRG1 in vivo. ATPase inhibitor treatment of a subcutaneous xenograft mouse model significantly inhibited tumor growth and improved survival. However, PROTAC treatment only marginally reduced tumor volume and did not improve survival, likely due to poor absorption and tissue penetrance. Although subcutaneous models fail to accurately recapitulate some aspects of the human disease, limiting our ability to generalize the results of these drug studies, targeting BRG1 is a rational therapeutic approach, especially in a disease with such a dismal prognosis.

In summary, the authors identified BRG1 as a dependency in pediatric H3K27M DMG. They outlined a molecular mechanism by which, in H3K27M-mutant cells, SOX10 recruits BRG1 to gene regulatory elements to drive expression of ECM and cell proliferation pathway genes (Fig. 1A and B). Pharmacologically targeting BRG1 with either ATPase inhibitors or PROTACs resulted in antiproliferative effects and induction of apoptosis in vitro and reduced tumor volume and increased survival in vivo. Inhibiting BRG1 ATPase activity represents a potential therapeutic strategy for pediatric H3K27M DMG.

However, it is important to take a step back and ask why preclinical studies of inhibitors targeting epigenetic regulators in pediatric H3K27M DMG have thus far failed to provide meaningful benefits to patients. One technical reason may be the difficulty in generating blood–brain barrier–penetrant versions of epigenetic inhibitors, a significant hurdle in advancing compounds clinically in patients with brain tumors. Although a degrader of BRG1 showed efficacy preclinically in a prostate cancer model (9), PROTAC degraders of BRG1 generated by Panditharatna and colleagues failed to penetrate the blood–brain barrier. Degraders that pass efficiently into the brain are required for the treatment of brain tumors.

Another reason for therapeutic failure to date may be the plasticity of the epigenome and the importance of context dependence in defining the function of epigenetic regulators. SMARCA4 was first described as a tumor suppressor gene (10), and how the H3K27M mutation rewires the epigenome to become dependent on BRG1 is not clear. Theoretically, targeting BRG1 may lead to secondary malignancies, especially with long-term treatment, and future studies should assess this risk. However, in pediatric cancers like H3K27M DMG, where epigenetic dysregulation causes differentiation block, long-term treatment likely may not be necessary. Short-term treatment with BRG1 inhibitors (perhaps in conjunction with radiation) may relieve the differentiation block, shunting cells from a proliferative, less differentiated OPC-like state to a differentiated nonproliferative or apoptotic state. BRG1 inhibition potentially offers much-needed differentiation therapy for pediatric H3K27M DMG, with promise to improve survival for patients.

W.A. Weiss reports other support from StemSynergy Therapeutics outside the submitted work. No disclosures were reported by the other author.