MOZ and Menin–MLL Complexes Are Complementary Regulators of Chromatin Association and Transcriptional Output in Gastrointestinal Stromal Tumor

Gastrointestinal stromal tumor (GIST) is among the most common sarcomas and arises from the interstitial cells of Cajal or their progenitors (1, 2). The majority of GIST cases are characterized by activating mutations in the receptor tyrosine kinase KIT, and inhibition of KIT with tyrosine kinase inhibitors (TKI) is the only established therapeutic strategy for this disease (3, 4). Although TKIs are efficacious for the majority of GIST patients, the evolution of TKI resistance mutations is inexorable in the metastatic setting (5, 6), leading to poor patient outcomes. Novel targeted therapeutic strategies are urgently needed for GIST clinical care, either as monotherapy or in combination with TKIs, although translational efforts to date have not yet met with success.

In contrast to many other malignancies, GIST does not amplify the mutant KIT gene but instead relies upon epigenetic mechanisms of gene regulation for oncogene expression (7). We and others have previously characterized the enhancer and transcriptional landscapes of GIST, which emphasized a conserved core network of transcription factors (TF) that are essential for GIST viability, regulate the expression of GIST-associated genes, including KIT, and are predictive of patient outcomes (8–11). Therapeutic targeting of these key transcriptional dependencies represents an attractive treatment strategy for this disease. In addition, understanding how these TFs establish codependency with essential chromatin regulators would provide significant insight into gene regulation and epigenetic mechanisms of disease (12, 13). Recent advances in the targeting of chromatin-modifying enzymes using inhibitory small molecules have led to new discoveries on the function of these epigenetic regulators and opened opportunities to therapeutically disrupt their activity; several of these drugs have been approved or are under active clinical investigation for cancer treatment.

Here, using genome-scale CRISPR/Cas9 dependency screening, we further determine the paramount importance of epigenetic regulation in GIST oncogenesis, defining essential chromatin regulators that maintain the GIST epigenome and providing a map of experimentally determined therapeutic targets. Among chromatin-modifying enzymes enriched in GIST, KAT6A/MOZ and KMT2A/MLL1 were established as previously unknown codependencies and, more broadly, were found to exhibit similar coregulation across select cancer subtypes. MOZ is a histone acetyltransferase and together with the assembly and chromatin reader proteins BRPF1, MEAF6, and ING5 constitutes the MOZ complex, which mediates activating histone acetylation including locus-specific H3K9ac (14, 15). KMT2A/MLL1 is a member of the Menin–MLL complex and is responsible for H3K4 methylation and transcriptional activation (16, 17). MOZ and Menin–MLL complexes were found to colocalize to GIST-associated genes genome-wide and regulate oncogenic transcription and cell-cycle progression by the regulation of TF gene-expression programs. Catalytic inactivation of MOZ or disruption of Menin–MLL protein interaction using small-molecule inhibitors led to changes in the chromatin association and protein–protein interactions of key mediators of transcriptional regulation, such as disruptor of telomeric silencing 1-like (DOT1L). Inhibition of MOZ and Menin significantly reduced GIST tumor growth in vivo, with the combination of Menin and KIT inhibition showing combinatorial toxicity in cell line and patient-derived xenograft (PDX) models of GIST. Our findings identify novel complementary roles for MOZ and Menin–MLL complexes in maintaining the cancer-associated transcriptional program in GIST and nominate novel therapeutic strategies either alone or in combination with TKIs.

Previously, we and others have characterized the global transcriptional and enhancer landscape of GIST using RNA sequencing (RNA-seq), chromatin immunoprecipitation with sequencing (ChIP-seq), and assay for transposase-accessible chromatin using sequencing (ATAC-seq). These investigations focused on TFs relevant to GIST biology, including core TFs such as ETV1, FOXF1, HIC1, and OSR1 present across GIST samples, and accessory TFs BARX1 and HAND1 expressed in disease state–specific patterns (11). However, how these TFs integrate with other epigenetic regulators to establish the GIST-associated gene expression program is unknown. To establish genes essential to GIST biology, we performed a genome-wide CRISPR/Cas9-based dropout screen in two KIT-mutant GIST cell lines, GIST-T1 and GIST430. A split-library approach utilizing paired human whole-genome single-guide RNA (sgRNA) libraries (denoted H1 and H2) was used, with approximately 5 sgRNAs per gene in each library targeting 18,436 genes with a total of 185,634 sgRNAs in the screen. Significant correlation in dependency (β) scores was observed between the H1 and H2 libraries (Fig. 1A) as well as between the two GIST cell lines (Fig. 1B); data sets were then merged for subsequent analysis to improve statistical power. Genes were stratified as “pan-essential,” having been previously determined to be universally essential for cellular viability (18, 19); “GIST essential” with an FDR <0.05 in the screen but absent from the pan-essential list; or “nonessential” (Fig. 1C; Supplementary Table S1).

As anticipated, KIT was among the strongest detected dependencies, with sgRNA-level data showing near-complete dropout of most (9/10) sgRNAs during the screen (Fig. 1D). Other canonical downstream signaling mediators of the KIT pathway, such as mTOR, showed significant dropout in the screen but less so compared with KIT (Fig. 1E). To identify biological processes that may be specifically enriched in GIST, we focused on the GIST essential subset of genes (Fig. 1F). These unique dependencies in GIST were evaluated by gene ontology enrichment analysis, which revealed that 8 of the top 18 terms were associated with epigenetic regulatory mechanisms, including chromatin and chromosomal organization (Fig. 1G; Supplementary Fig. S1A). Taken together, data from these unbiased dependency screens characterize GIST as having remarkable reliance upon epigenetic mechanisms to maintain its oncogenic program.

To better define which chromatin regulatory complexes may be most relevant and unique to GIST biology, we compared β-scores for all chromatin-modifying enzymes in GIST cell lines to analogous CERES dependency scores averaged across all cell lines in the Dependency Map (DepMap) project (20). Only 7 of the 77 assessed chromatin-modifying enzymes were unique and essential to both GIST cell lines, with β-score <−0.7 and CERES score >−0.25 (Fig. 2A; Supplementary Fig. S1B), with dependency score cutoffs chosen to select for chromatin regulators likely to be unique dependencies. Enriched enzymes included members of the lysine acetyltransferase (KAT), MYST, lysine demethylase (KDM), and lysine methyltransferase (KMT) families. KAT6A was a significant outlier in this comparison, with the highest β-score among all chromatin-modifying enzymes and an average CERES score near 0. To establish which of these seven chromatin-modifying enzymes may function collaboratively to maintain the epigenome, we evaluated gene-level codependency data within DepMap. Comparative analysis of the top 50 codependencies of each chromatin-modifying enzyme showed the highest interaction at gene and ontology term levels between KAT6A, KMT2A, and EZH2 (Fig. 2B), suggesting their genetic codependency. KAT6A, the catalytic member of MOZ, and complex members MEAF6 and BRPF1 were all significant dependencies in GIST (Supplementary Fig. S1C). KMT2A, the catalytic member of the Menin–MLL complex, also had multiple recognized complex members with significant dependency scores including MEN1/Menin and ASH2L (Supplementary Fig. S1D). Plotting the top DepMap codependencies between KAT6A and BRPF1, multiple members of the Menin–MLL complex are highly correlated codependencies, indicating genetic interaction between the MOZ and Menin–MLL complexes (Fig. 2C). As DepMap CRISPR dependency screening efforts have not yet profiled GIST, we leveraged available comparative screening results from Project DRIVE (21), which included GIST-T1 among nearly 400 cell lines profiled by RNAi. Among all cell lines profiled, GIST-T1 had the highest sensitivity score for KAT6A, had the second highest for BRPF1, and was in the top 5% for KMT2A and ASH2L (Fig. 2D–G), further indicating through an independent and comparative screening approach the essential and codependent nature of the MOZ and Menin–MLL complexes in GIST. Several other chromatin regulatory complexes were found to have multiple members with significant dependencies in our screen and also showed enrichment for GIST-T1 in Project DRIVE, including members of the INO80 complex, NuA4 histone acetyltransferase complex, FACT complex, and PAF1 complex (Supplementary Fig. S1E–S1O). EZH2, SUZ12, and EED, the core members of the PRC2 complex (22), were all dependencies in our screen, and GIST-T1 had among the highest sensitivity scores for EZH2 and EED in Project DRIVE (Supplementary Fig. S2A–S2C). Although few cell lines in DepMap had significant dependencies on core PRC2 complex members, the top correlated codependencies of EZH2, including DOT1L, EP300, and MEN1, showed overlap with MOZ and Menin–MLL complex codependencies (Supplementary Fig. S2D and S2E), indicating the complementary function of the transcriptionally repressive PRC2 complex.

To validate dependency upon MOZ and Menin–MLL complex members, we performed a growth over time assay utilizing unique sgRNAs targeting MOZ complex members KAT6A, BRPF1, and MEAF6 and Menin–MLL complex members KMT2A and MEN1. For each gene and with two independent sgRNAs, sgRNA treatment significantly reduced cell proliferation (Fig. 2H). To compare the relative toxicity of these sgRNAs to a control cell line, we utilized GIST48B, which has a similar growth rate as GIST-T1 but through in vitro selection has lost KIT expression and the GIST-associated epigenetic and transcriptional program (10, 23). Although all sgRNAs targeting MOZ and Menin–MLL complexes significantly reduce GIST-T1 cell proliferation, GIST48B showed little or no change in cell proliferation (Fig. 2I). Taken together, these data show a complementary codependency in GIST and across select cell lines in DepMap between MOZ and Menin–MLL complexes, suggesting their collaborative role in maintaining the GIST epigenome.

To define where in the GIST genome MOZ and Menin–MLL complexes bind and modify histones, we performed ChIP-seq for histone marks H3K27ac, H3K9ac (deposited by MOZ and other acetyltransferases), H3K4me3 (deposited by the Menin–MLL complexes), BRPF1, and KAT6A. Although we were unable to successfully perform ChIP-seq for the Menin–MLL complex in GIST cell lines, we were able to identify genomic regions of binding of Menin and MLL1 using the analogous method CUT&Tag (24). Both MOZ and Menin–MLL complex members were found to be localized at the transcription start sites (TSS) of active genes, as determined by their co-occupancy with H3K27ac and H3K9ac (Fig. 3A, top row). In contrast, we observed very little occupancy of these chromatin complex members at H3K27ac-defined enhancers (Fig. 3A, middle row). ATAC peaks, which include accessible DNA sites at both TSSs and enhancers, showed an intermediate level of MOZ and Menin–MLL complex binding (Fig. 3A, bottom row). As suggested by heat maps, there was considerable overlap in peaks within and between MOZ and Menin–MLL complexes (Fig. 3B–D), indicating that the genomic binding of these complexes converges at TSSs and further suggesting their colocalization.

We next evaluated genomic regions that displayed strong enrichment for BRPF1 and Menin in our ChIP-seq and CUT&Tag data sets, reasoning that these factors are representative of MOZ and Menin–MLL complexes. Although both proteins bind to thousands of sites genome-wide, disproportionate enrichment was seen in 3% to 5% of these genomic regions, many of which have clear relevance to GIST (Fig. 3E and F; Supplementary Fig. S3A). TFs, particularly those among the group of core and accessory GIST TFs (10, 11), were included in these enriched regions, as were negative regulators of KIT signaling from DUSP and sprouty families and genes used as biomarkers for GIST (e.g., GPR20, CD34; ref. 25). ChIP-seq and CUT&Tag tracks showed binding of MOZ and Menin–MLL complex members at the TSS and gene body of these enriched genes, analogous to H3K4me3, such as at core TF members DUSP6 and NPR3 (Fig. 3G–I; Supplementary Fig. S3B); by contrast, H3K27ac and H3K9ac were enriched at both enhancer regions and gene bodies, and the GIST accessory TF HAND1 was bound exclusively to enhancers. MOZ and Menin–MLL complex members were notably not enriched at the KIT locus, though both these regulators were bound to the TSS and Menin and MLL1 to a region downstream of the gene body (Supplementary Fig. S3C). Other regions of enrichment showed maximal binding of MOZ and Menin–MLL complex members within and immediately downstream of the TSS, with detectable signal evident at some enhancers of these highly regulated genes (Supplementary Fig. S3D–S3F). These data indicate that MOZ and Menin–MLL complexes are globally present at active genes, with enrichment at a subset of genes relevant to the GIST transcriptional program.

Based on the genetic data and genomic colocalization of MOZ and Menin–MLL complexes, we reasoned that small-molecule inhibitors targeting these complexes would be a viable therapeutic approach. To explore the functional consequences of MOZ and Menin–MLL disruption, we treated GIST-T1 with the selective KAT6A inhibitor WM-1119 (26) alone or in combination with the Menin inhibitor VTP-50469 (27). At submicromolar concentrations, both WM-1119 and VTP-50469 decreased GIST cell proliferation in a growth over time assay (Fig. 4A and B), with a greater effect seen with the combination of the two drugs (Fig. 4C). Unlike imatinib (Supplementary Fig. S4A), antiproliferative effects of MOZ and Menin inhibitors are seen with prolonged drug exposure, in line with other studies utilizing these compounds in leukemia (26, 27). Demonstrating selective toxicity of these inhibitors in KIT-dependent GIST cell lines, the KIT-independent GIST48B cell line showed modest or no alterations in proliferation after a 21-day drug treatment (Fig. 4D), in agreement with genetic data from CRISPR experiments (Fig. 2I). In previous studies of the KIT enhancer, we used an sgRNA directed at the KIT TSS to ablate endogenous KIT expression while rescuing cell viability by expressing a viral promoter-driven KIT construct bearing the same activating mutation (23). This KIT-dependent KIT rescue cell line was similarly susceptible to WM-1119 and VTP-50469 alone or in combination (Fig. 4D), indicating that regulation of the endogenous KIT locus is not the principal mechanism of toxicity of these compounds. To confirm the proliferative effects of these inhibitors in additional GIST cell lines, we treated the slower growing KIT-mutant cell lines GIST430, GIST882, and GIST48 (23) with WM-1119 and VTP-50469 alone or in combination, observing analogous antiproliferative effects arising from drug treatment (Fig. 4E; Supplementary Fig. S4B).

To evaluate the cellular phenotypic consequences of WM-1119 and VTP-50469 treatment, we performed cell-cycle and apoptosis assays utilizing these drugs and the TKI imatinib as a comparator. Although imatinib acutely and potently caused G₀–G₁ cell-cycle arrest within 72 hours, 8 days of treatment with either WM-1119 or VTP-50469 alone led to a modest increase in the fraction of cells in G₀–G₁; the combination of WM-1119 and VTP-50469 led to more marked disruption of the cell cycle after 8-day treatment (Fig. 4F). Although 72-hour imatinib treatment produced a significant increase in early- and late-phase apoptosis and cell death, an 8-day treatment of WM-1119 and VTP-50469 alone or in combination did not significantly increase apoptosis or cell death compared with DMSO control (Fig. 4G). We next evaluated molecular disruptors of the MOZ complex that bind outside the KAT6A active site. BRPF1, considered a scaffolding protein for its multiple DNA and chromatin binding domains, bears a bromodomain that can be selectively and potently targeted by inhibitors PFI-4 and GSK6853, among others (28, 29). Despite the genetic dependency upon BRPF1 in GIST, neither PFI-4 nor GSK6853 disrupted GIST cell proliferation (Supplementary Fig. S4C). A similar absence of toxicity arising from these compounds has been observed in other cell types (28, 29), although concomitant genetic dependency upon BRPF1 has not previously been well established; these findings suggest that BRPF1 bromodomain–targeting compounds may represent ideal candidates for the development of proteolysis-targeting chimera-based molecular degraders. Taken together, these data show that MOZ and Menin–MLL complexes are targetable and unique vulnerabilities in GIST, and in keeping with their distribution across the genome at TSSs of active genes, may play essential roles in gene regulation other than just KIT gene expression. Disruption of these complexes alone or in combination causes alterations in the cell cycle but not in programmed cell death.

Observing the genome-wide occupancy of MOZ and Menin–MLL complexes, with limited regions of enrichment, we next probed for selective changes in gene expression arising from the pharmacologic or genetic disruption of these complexes. We performed RNA-seq on GIST-T1 cells treated for 5 days with WM-1119, VTP-50469, or the combination of these two drugs. Compared with DMSO control, all treatment groups showed changes in global gene expression, with VTP-50469 and the combination groups showing the greatest deviation from control (Fig. 5A). We similarly evaluated gene expression changes following genetic disruption of MOZ and Menin–MLL complexes using sgRNAs targeting two complex members each. Using this genetic system, sgRNAs targeting MEN1 led to global changes in gene expression, with disruption of other MOZ and Menin–MLL1 complex members showing less dramatic changes (Fig. 5B). To integrate and compare transcriptional changes arising from either pharmacologic or genetic disruption, we evaluated the correlation of gene expression changes from control across the transcriptome. For comparison, we also evaluated transcriptional changes from sgRNA-mediated disruption of the GIST TFs HAND1 and ETV1 (11). sgRNAs targeting MOZ complex members KAT6A and BRPF1 showed the highest degree of correlation and, moreover, induced similar global transcriptional changes, as did the disruption of HAND1, ETV1, or KMT2A (Fig. 5C–E). Pharmacologic inhibition of MOZ and/or Menin–MLL complexes had a comparatively weak but positive correlation with sgRNAs targeting GIST TFs, MOZ complex members, and KMT2A, whereas genetic disruption of MEN1 showed the least correlation with other conditions (Fig. 5C and F).

As these results disclosed disparate global transcriptional consequences arising from genetic or chemical disruption of these epigenetic regulators, we sought to identify core GIST transcriptional programs exhibiting common alterations that could account for central functions of MOZ and Menin–MLL complexes. Using gene set enrichment analysis (GSEA), we evaluated for changes in the expression of genes enriched for MOZ and Menin–MLL complexes, H3K27ac-defined superenhancer (SE)–associated genes, and genes regulated by TFs HAND1 and ETV1. Among these GIST-associated gene lists, genes regulated by TFs were the most markedly affected, with drug or sgRNA treatment causing reduced expression of genes upregulated by HAND1 and increased expression of genes normally downregulated by HAND1 or ETV1 function (Fig. 5G). Genetic or pharmacologic MOZ disruption exhibited the greatest effect on genes bound by Menin or BRPF1 (Fig. 5H), while only targeting HAND1 or the combination of WM-1119 and VTP-50469 decreased the expression of SE-associated genes (Fig. 5I). However, common to all conditions was the disruption of HAND1-associated gene expression, with expression values of MOZ and Menin–MLL complex disruption using either inhibitors or sgRNAs phenocopying direct HAND1 knockout (Fig. 5J–L).

KIT gene expression was most markedly affected by pharmacologic or genetic disruption of Menin, whereas there was a common loss of DUSP6, a negative regulator of KIT signaling, and the GIST biomarker CD34 (Fig. 5M). Expression of several core GIST TFs was altered by genetic or pharmacologic MOZ or Menin–MLL complex disruption—most notably FOXF1, HAND2, and PITX1—with WM-1119 exerting the greatest global reduction in TF expression (Fig. 5N; Supplementary Fig. S5A and S5B). Several other genes highly regulated by HAND1 expression, including NPR3, ITGA4, and RASL11A, showed similar loss of expression with disruption of MOZ and Menin–MLL complexes (Fig. 5O). Comparable reductions in gene expression of DUSP6 and NPR3 were seen in the KIT-dependent GIST cell lines GIST430, GIST882, and GIST48 by qRT-PCR (Supplementary Fig. S5C–S5F). Among all Reactome gene sets, processes related to protein translation were the most recurrently altered gene sets among treatment conditions, with most drug and sgRNA data sets showing a decrease in gene expression (Supplementary Fig. S5G and S5H). Taken together, these results show that both genetic and pharmacologic means of MOZ and Menin–MLL complex disruption lead to selective alteration of transcriptional programs associated with GIST TFs and most remarkably HAND1. Further, dual inhibition of Menin and MOZ with small molecules induces complementary effects on global gene expression and reduces expression of GIST SE-associated genes.

As MOZ and Menin–MLL complexes function in a coordinated fashion at highly regulated genomic regions together with other chromatin regulators, we sought to evaluate effects on local protein interactions in the presence or absence of WM-1119 and/or VTP-50469. We utilized the BioID system (30) to append the biotin ligase BirA* to the N-terminus of the MOZ complex member MEAF6, which enabled the covalent labeling of proteins localized within 10 nm with a biotin moiety. To ensure appropriate incorporation of the BirA*-tagged MEAF6 into the MOZ complex, we used CRISPR/Cas9 and a MEAF6-targeted sgRNA to disrupt the endogenous MEAF6, which would otherwise be lethal if not functionally replaced by the stably expressed MEAF6-BirA construct (Fig. 2H). Stable expression of the N-terminally tagged MEAF6-BirA in GIST-T1 led to high levels of protein production, with evidence of N-terminal degradation products observed on Western blot (Fig. 6A). As a BioID control, we used a construct that fused BirA* to the DNA binding domain of IKZF1, which retained its nuclear localization signal and served as a nuclear background control. Following labeling of cells for 24 hours with biotin, we performed a streptavidin pulldown followed by mass spectrometry, identifying 243 proteins labeled by MEAF6-BirA and enriched above control (Fig. 6B; Supplementary Table S2). Labeled interactors included chromatin regulatory proteins such as KMT2A/MLL1, KMT2B/MLL2, JADE3, and RUVBL1, in addition to MOZ complex members, enhancer-associated proteins such as BRD4, and the core GIST TF HIC1. Ordering these MOZ-proximal proteins by gene ontology, there was enrichment for cellular processes including DNA repair, mRNA processing, and chromatin complex regulation (Fig. 6C). These data demonstrate the integrated cellular function of these transcriptional regulatory proteins and their complex interactions between splicing factors, enhancers, and chromatin complexes.

To evaluate for changes in the MEAF6 proximal proteome as a consequence of MOZ and Menin inhibition, we pretreated MEAF6-BirA–expressing GIST-T1 cells for 3 days with the drug prior to labeling with biotin and subsequent label-free quantification using mass spectrometry. Although the majority of MEAF6-proximal proteins remained the same, a subset of proteins showed significant alterations in abundance with drug treatment, with a significant correlation in changes seen with WM-1119 and/or VTP-50469 treatment (Fig. 6D and E). A subset of proximal proteins increased only with WM-1119 or combination treatment, including MOZ complex members KAT6A and BRPF1 and splicing factor XAB2 (Fig. 6F); that BRPF1 association increased only with the combination treatment may suggest differences in MOZ complex interactions following dual inhibitor treatment, as compared with initial complex formation (31). In contrast, MLL family member KMT2A/MLL1, MLL family member KMT2B/MLL2, and DNA and RNA binding antiapoptotic protein GPATCH4 (32) interactions were decreased only with VTP-50469 treatment (Fig. 6G). The proximity of select chromatin regulators, splicing factors, and polymerase regulatory proteins was altered in a similar fashion with either drug, most notably DOT1L, which was significantly decreased in all treatment conditions (Fig. 6H). Leveraging DepMap data, DOT1L dependency was found to be highly correlated with codependency upon Menin–MLL complex, MOZ complex, and PRC2 complex members (Supplementary Fig. S6A), suggesting the importance of this protein–protein interaction. To determine how MOZ or Menin inhibition altered the chromatin association of DOT1L, we performed spike-in normalized ChIP-seq in GIST-T1 cells treated with WM-1119 or VTP-50469. Both drugs significantly decreased DOT1L association with chromatin at all DOT1L binding sites without changing DOT1L expression (Fig. 6I and J; Supplementary Fig. S6B–S6E).

We similarly performed spike-in normalized ChIP-seq for MEAF6 using the HA-tagged MEAF6-BirA construct, which showed no loss of signal with VTP-50469 treatment but modest reduction with WM-1119 treatment (Fig. 6K; Supplementary Fig. S6F). Together with BioID data, which showed an increased association of other MOZ complex members with MEAF6-BirA following WM-1119 treatment, these data suggest that inhibition of KAT6A leads to loss of the intact MOZ complex and associated factors from chromatin; this accounts for both the reduction in ChIP-seq signal and increased MOZ complex association in BioID data following WM-1119 treatment. In an analogous fashion, VTP-50469 treatment was found to decrease labeling of MLL proteins by chromatin-bound MEAF6-BirA, suggesting loss of Menin–MLL complexes from chromatin analogous to what has been shown in MLL-rearranged leukemia (27).

Like other MOZ and Menin–MLL complex members, DOT1L, H3K79me2 (the histone mark deposited by DOT1L), and MEAF6 all showed enrichment genome-wide at the TSS and gene body of active genes, with enrichment at loci relevant to GIST biology and reduction in DOT1L signal with WM-1119 or VTP-50469 treatment (Fig. 6L; Supplementary Fig. S6B and S6G–S6J). To further explore whether loss of DOT1L function could constitute a mechanism of cellular toxicity downstream of MOZ and Menin inhibition, we treated GIST-T1 cells, or GIST48B as control, in a growth over time assay at various doses of the selective DOT1L inhibitor EPZ-5676 (33). At all doses tested, GIST-T1 exhibited significantly reduced cellular proliferation compared with DMSO control or GIST48B, indicating selective toxicity of DOT1L inhibition in a manner similar to MOZ and Menin inhibition (Fig. 6M). DOT1L-targeting sgRNAs led to significant reductions in GIST-T1 cell proliferation, although more modest than that seen with Menin–MLL- and MOZ complex–targeting sgRNAs, consistent with findings of the genome-scale CRISPR screen (Supplementary Fig. S6K; Fig. 2A). To better characterize the transcriptional consequences of DOT1L inhibition in GIST, we treated GIST-T1 cells with EPZ-5676 for 5 days followed by RNA-seq. Globally, transcriptional changes associated with EPZ-5676 treatment were highly correlated with those arising from Menin inhibition with VTP-50469 (Fig. 6N and O), a phenomenon previously observed in MLL-rearranged leukemia (27). Like Menin–MLL and MOZ inhibition, DOT1L inhibition led to significant disruption in HAND1-regulated transcriptional programs (Fig. 6P), with alterations in the expression of KIT, CD34, NPR3, and GIST TFs (Fig. 6Q and R). Together, these data demonstrate the complexity of proximal protein interactions between these epigenetic complexes, alterations in the protein and chromatin associations of multiple transcriptional regulators with MOZ or Menin inhibition, and that DOT1L function is a dependency in KIT-dependent GIST, with loss of DOT1L chromatin association serving as a downstream consequence of MOZ or Menin inhibition.

To evaluate the effects of genetic loss of KAT6A, Menin, or DOT1L on tumor growth in vivo, we expressed and selected for cells expressing an sgRNA directed against each of these targets or luciferase control in GIST-T1 cells coexpressing Cas9. Following implantation of an equal number of modified cells, mice were monitored for tumor formation and growth. Although all implants generated tumors, those derived from cells treated with sgKAT6A or sgMEN1 had significantly reduced growth compared with sgLuc control, whereas expression of sgDOT1L led to a nonsignificant trend toward reduced growth (Supplementary Fig. S7A). Although growth restriction was more modest than analogous in vitro experiments (Fig. 2H; Supplementary Fig. S6K), sgKAT6A and sgMEN1 conditions required selection and propagation for 2 weeks in vitro to produce enough cells for implantation, likely selecting for cells with less deleterious gene alterations. As these genetic results supported a more prominent role for MOZ and Menin–MLL complexes in tumor growth, we next evaluated the effects of MOZ and Menin pharmacologic inhibition in vivo. We treated GIST-T1–engrafted mice with either WM-1119, dosed 3 times daily 7 days/week, or VTP-50469, given continuously in chow. WM-1119 has been shown to be rapidly cleared after administration (26), necessitating frequent daily administration to achieve therapeutic effects. Whereas 28-day administration of WM-1119 or VTP-50469 alone led to nonsignificant reductions in tumor size (72% and 65% of vehicle control, respectively), the combination treatment completely halted tumor growth over the treatment period (Fig. 7A). We continued monitoring tumor recovery without further drug treatment, and all conditions showed comparable tumor growth rates similar to vehicle control. Mice tolerated treatment with WM-1119, VTP-50469, or the combination without evidence of overt toxicity or weight loss (Supplementary Fig. S7B).

Multiple Menin inhibitors have advanced to early-phase clinical trials for the treatment of leukemia (NCT04067336, NCT04811560, and NCT04065399) and may represent an optimal strategy for targeting GIST's chromatin regulatory dependencies in combination with TKIs. To assess the effects of Menin inhibition alone or in combination with imatinib in vivo, we treated mice engrafted with GIST-T1 cells with imatinib, VTP-50469, the combination of imatinib and VTP-50469, or vehicle as control. At the end of a 28-day treatment period, both monotherapy treatment groups showed similar significant reductions in tumor growth compared with vehicle, whereas the combination group showed complete cessation of tumor growth. In a subsequent posttreatment observation period, although the tumors from imatinib and VTP-50469 monotherapy groups regained a tumor growth trajectory similar to the vehicle group, the combination of imatinib and VTP-50469 sustained a 3- to 4-fold reduced slope of tumor recovery (Fig. 7B). All treatment arms tolerated drug administration without overt toxicity or weight loss (Supplementary Fig. S7C).

To evaluate for changes in the GIST transcriptional program arising from Menin and/or KIT inhibition in vivo, we performed RNA-seq on GIST-T1 xenografts after 5 and 10 days of imatinib and/or VTP-50469 treatment. Although all treatment conditions led to global transcriptional changes compared with vehicle control, greater changes were seen following treatment with VTP-50469 and the combination of imatinib and VTP-50469 at both time points, with the gene expression profile of imatinib treatment more closely correlating with vehicle-treated tumors (Fig. 7C). Imatinib, VTP-50469, and combination treatments all led to a decrease in the expression of genes regulated by HAND1 (Supplementary Fig. S7D). Although GIST-associated transcripts including KIT, CD34, and NPR3 were preferentially reduced by VTP-50469 treatment, other KIT signaling-dependent transcripts including TMEM100 and SPRY2 (34) were preferentially reduced with imatinib treatment; PCNA, a marker of cellular proliferation, was reduced only with the combination of imatinib and VTP-50469 at both 5- and 10-day time points (Fig. 7D), consistent with the greater effect of the combination treatment on tumor growth.

We next evaluated the effects of imatinib and VTP-50469 treatment on PG27, a KIT-mutant PDX model of GIST (23). Although imatinib administration below the maximum tolerated dose had a significant but modest growth-inhibitory effect compared with the GIST-T1 cell line xenograft, treatment with VTP-50469 alone or in combination with imatinib resulted in a significant reduction in tumor growth (Fig. 7E). At the end of the treatment period, PG27 tumors were harvested, fixed, and sectioned to evaluate tumor histology. Although vehicle- and imatinib-treated tumors had monomorphic sheets of tumor cells, xenografts treated with VTP-50469 or the drug combination exhibited areas of tumor necrosis (Fig. 7F). Whereas mice treated with VTP-50469 in chow showed no weight loss in the GIST-T1 xenograft experiments, PG27-engrafted mice treated with VTP-50469 at the same concentration exhibited modest weight loss (Supplementary Fig. S7E), possibly related to systemic effects of the observed tumor necrosis. Despite the restriction of tumor growth with VTP-50469 treatment, viable areas of the tumor showed similar levels of Ki-67 and cleaved caspase-3 across conditions (Supplementary Fig. S7F). These data demonstrate the therapeutic activity of VTP-50469 alone or in combination with imatinib, with Menin inhibition decreasing GIST xenograft growth, producing tumor necrosis, and, when combined with imatinib, generating durable tumor responses after cessation of treatment.

The organization and remodeling of chromatin are essential to cellular lineage, identity, and function. Posttranslational modifications of histones serve as a nexus of epigenetic regulation that controls the binding of TFs and chromatin regulators, ultimately administrating gene expression and chromosomal structure. As chromatin modifications are dynamic and reversible, they require active maintenance by cell type– and state-specific chromatin-modifying enzymes. Cancer exploits or appropriates the chromatin state of its precursor cells to sustain a malignant phenotype through maintenance of an environment permissive of oncogene activation or by gain-of-function alterations in chromatin regulators such as MLL gene fusions (17). Here, we show that specific chromatin regulators are essential to sustain the GIST epigenome, with the MOZ and Menin–MLL complexes binding to actively expressed genes genome-wide, regulating GIST-associated gene expression programs, coordinating protein–protein interactions between multiple regulators of gene expression, and ultimately regulating cellular proliferation and tumor growth.

MOZ is essential for embryonic development and locus-specific gene regulation (14, 35). Through the development of selective KAT6A/B inhibitors, this complex has been found to induce cell-cycle arrest and senescence by maintaining H3K9ac at cell type–specific genes (26). Menin, encoded by the multiple endocrine neoplasia 1 (MEN1) gene, has classically been described as a tumor suppressor, with mutations in MEN1 promoting endocrine tumor formation. However, in other tissues, multiple functions have been ascribed to Menin arising from the protein's ability to positively or negatively regulate gene expression, associate with different chromatin complexes, integrate inputs from upstream signaling pathways, and modulate DNA replication and repair (36). Menin has been best studied as an oncogenic dependency in the context of MLL-rearranged leukemia, in which it binds to the MLL fusion protein and, together with the recruitment of DOT1L, executes the leukemogenic gene expression program (27, 37, 38). In GIST, both MOZ and Menin–MLL complex members are essential for global chromatin regulation and, ultimately, tumor cell proliferation. Compared with hundreds of other cell types profiled in Project DRIVE and DepMap, GIST has exceptional sensitivity to the targeted disruption of MOZ and Menin–MLL complex members. In agreement with the conservation of TFs and transcriptional and chromatin landscapes in KIT-dependent GIST (10, 37), sensitivity to genetic or pharmacologic MOZ and Menin–MLL complex disruption was lost in a KIT-independent GIST cell line. These data indicate that, unlike the oncogenic hijacking seen in MLL-rearranged leukemias, GIST depends upon the native function of the MOZ and Menin–MLL complexes and their associated dependencies to maintain a chromatin landscape that provides a foundation for a malignant gene expression program.

Multiple lines of evidence suggest collaboration between MOZ and Menin–MLL complexes in transcriptional regulation. Here, we show the genome-wide colocalization of MOZ and Menin–MLL complex members at the TSS of actively expressed genes, changes in gene expression arising from inhibition of either complex, proximal protein interactions between these two complexes, coordinated regulation of DOT1L and other transcription-associated proteins, and that effects on cell-cycle and proliferation were more marked when inhibiting both complexes rather than one at a time. In agreement with our findings, interactions between MOZ and MLL complexes promoting gene expression have been previously described at the HOXA locus in hematopoietic progenitor cells (39) as well as their coordinated colocalization genome-wide (40). Leveraging DepMap data, we highlight a previously underappreciated and complementary genetic codependency of these chromatin regulatory complexes in a minority of cancer cell lines. Dependency upon the PRC2 complex was also seen in GIST, with a similar codependency observed across DepMap data, suggesting the contrasting but complementary role of PRC2 in chromatin silencing balancing the activating functions of MOZ and Menin–MLL complexes. These data suggest that MOZ, Menin–MLL, and PRC2 complexes cooperatively function genome-wide to control chromatin state and transcriptional output. We found the superior activity of simultaneous inhibition of MOZ and Menin–MLL complexes with WM-1119 and VTP-50469 on cell cycle, cellular proliferation assays, transcriptional output, and tumor growth. Whereas the MOZ complex was enriched at core GIST TF genes and WM-1119 treatment decreased TF expression/function, the Menin–MLL complex showed greater binding at the KIT locus and VTP-50469 decreased KIT expression, indicating at least some nonoverlapping functions of these complexes. Although combination drug treatment led to greater alterations in global and SE-associated transcription, the expression of some GIST-associated genes and disruption in protein–protein interactions showed many similarities between monotherapy and combination treatments. Although the mechanism of combinatorial toxicity requires further investigation, these results suggest that disruption of one complex may maximally deregulate both at specific target loci and that nonoverlapping functions of MOZ and Menin–MLL complexes may also exist.

In keeping with their association with the TSSs of active genes genome-wide, genetic or pharmacologic disruption of MOZ and Menin–MLL complexes led to alterations in transcription that were enriched in specific pathways. We found these transcriptional changes to closely phenocopy those observed with acute loss of GIST TFs and most notably HAND1. Selective disruption of HAND1 and other GIST TF-mediated gene expression may represent a primary antineoplastic mechanism of MOZ and Menin–MLL complex inhibition in GIST. This toxic mechanism may explain the delayed and progressive effects on GIST cell proliferation, as has been observed in leukemias (27, 38, 41), and is contrasted by the acutely toxic effects of direct KIT inhibition. In agreement with our findings, KAT6A was recently identified to support oncogenic MYC-associated gene expression and block differentiation in acute myeloid leukemia (AML) through H3K9 acetylation and ENL recruitment (42). ENL (MLLT1) was also found to be a dependency in our GIST genome-scale CRISPR screen (Supplementary Table S1) and was identified in MOZ complex proximity proteomics experiments (Supplementary Table S2), suggesting similar chromatin regulation strategies between GIST and AML. These results suggest a common utilization of epigenetic regulators supporting oncogenic transcriptional programs in subsets of these mesoderm-derived cancers.

Using BioID to assess the proximal proteome of the MOZ complex member MEAF6, we identified multiple potential regulatory interactions between proteins involved in chromatin modification, transcriptional elongation, mRNA processing, and DNA repair. Several interactions with enhancer-associated proteins were also observed, including BRD4, MED1, and the core GIST TF HIC1, which, in light of the principal localization of MOZ and Menin–MLL complexes at the TSS of target genes, suggests possible interaction with looping enhancers. In agreement with these findings, Menin has previously been shown to interact with looping enhancers and cancer-specific TFs to promote oncogenic gene expression (43). That HAND1 was not detected in proximity proteomics experiments, yet HAND1's gene expression program was highly sensitive to the disruption of MOZ and Menin–MLL complexes, suggests that these complexes may control TF programs by both direct association and indirectly through the modulation of downstream transcriptional regulatory processes. Our findings suggest that MOZ and Menin–MLL complexes serve as central integrators of the GIST gene expression program, regulating multiple steps in transcription through interactions with enhancer elements and chromatin and transcriptional regulators, and depositing histone marks supportive of productive gene transcription. Downstream consequences of MOZ or Menin inhibition include disruption of proximal interactions between transcriptional regulators, including loss of DOT1L from chromatin. DOT1L methylates H3K79 to support an active transcriptional state and has been investigated in leukemia where its recruitment by the MLL fusion protein is essential for leukemogenesis (44). In solid tumors, DOT1L has been found to cooperate with oncogenic TFs (45, 46), though DOT1L inhibitors have thus far not been evaluated in clinical trials for solid tumors. With prior studies demonstrating TF dependencies in GIST, and current work showing the vulnerability of GIST cells to both DOT1L genetic and pharmacologic disruption, DOT1L may function as a downstream integrator of TF, MOZ, and Menin–MLL complex activity in establishing a transcriptionally active state of select cancer-associated genes. Unexpectedly, using BioID, we identified proximal protein interactions with regulators of ribonucleoprotein and ribosomal biogenesis and associated loss of expression of genes involved in protein translation upon disruption of MOZ and/or Menin–MLL complexes. Although the relative contribution of changes in translation to the observed antineoplastic activity of these inhibitors is uncertain, this may represent an understudied downstream consequence of such chromatin complex alterations. Taken together, these data demonstrate the essential function of MOZ and Menin–MLL complexes in GIST, which serve as integral components of chromatin regulation and the oncogenic GIST gene expression program.

Through in vivo studies, we found that the genetic or pharmacologic disruption of MOZ and Menin–MLL complexes led to reductions in GIST tumor burden, with the combination of WM-1119 and VTP-50469 fully halting tumor growth. Multiple Menin inhibitors have been developed that disrupt the association between Menin and MLL and are now under clinical investigation (27, 47, 48), and these represent ideal strategies to combine with TKIs in GIST. To assess the in vivo effect of Menin inhibition on xenograft models of GIST, we treated cell line xenografts and PDXs with imatinib, VTP-50469, or the combination treatment, which demonstrated activity of VTP-50469 as a monotherapy and superior activity with the combination of KIT and Menin inhibition. After the treatment period, tumors in both monotherapy arms regained their growth trajectory, whereas tumors treated with the combination of imatinib and VTP-50469 sustained prolonged tumor-suppressive effects observed weeks after treatment withdrawal. Both imatinib and VTP-50469 treatment led to distinctive transcriptional changes in GIST xenografts, with the combination of KIT and Menin inhibition resulting in a greater deviation from the baseline tumor transcriptional state. In a PDX model of GIST, we saw the potent antitumor activity of VTP-50469, with histology showing areas of necrosis interspersed with viable tumors. These results support the clinical development of Menin inhibitors for GIST patients, either alone or ideally in combination with TKIs.

As TKIs are the only clinically available therapeutic strategy in GIST, which has native resistance to cytotoxic chemotherapy (49), targeting Menin and other essential components of the GIST epigenome may prove therapeutically advantageous. The conserved transcriptional and enhancer landscape seen in GIST tumors and cell lines, together with oncogenic KIT gene expression being regulated by disease-specific TFs and enhancer elements, harbingers this disease's dependency upon epigenetic mechanisms. We describe here the collaborating chromatin regulators responsible for maintaining the GIST epigenome and how their disruption at multiple disparate nodes with small-molecule inhibitors (WM-1119, VTP-50469, and EPZ-5676) displays promising and selective anticancer activity; members of each of these inhibitor classes have reached clinical trial (e.g., NCT04606446 and NCT02141828). As compared with leukemias bearing oncogenic alterations in chromatin regulators, GIST may be an outlier among solid tumors in its dependency upon these pathways and susceptibility to their disruption.

All cell lines tested negative for Mycoplasma infection on routine surveillance (MilliporeSigma; cat. #MP0025-1KT). Human embryonic kidney 293FT (Thermo Fisher Scientific; cat. #R70007, RRID: CVCL_6911) and GIST cell lines GIST-T1 (Cosmo Bio; cat. #PMC-GIST01-COS, RRID:CVCL_4976; KIT mutation in exon 11 Δ560–578), GIST430 (RRID: CVCL_7040; KIT mutation in exon 11 Δ560–576), GIST48 (RRID: CVCL_7041; KIT mutations in exon 11 V560D and exon 17 D820A), GIST48B (RRID: CVCL_M441; KIT-independent), and GIST882 (RRID: CVCL_7044; KIT mutation in exon 13 K642E) were cultured in Dulbecco's modified Eagle medium containing 10% FBS, 2 mmol/L L-glutamine, 100 mg/mL penicillin, and 100 mg/mL streptomycin. KIT rescue cell lines, which are independent of the KIT enhancer, were generated as previously described (23). Noncommercial cell lines were obtained from the laboratory of Jonathan Fletcher (Brigham and Women's Hospital, Boston, MA) between 2014 and 2016. KIT exons were sequenced to confirm the expected coding mutations and cell identity of GIST cell lines, and cells were thawed from original or derived stocks and used in the described experiments within approximately 3 months. Transfections were performed with X-tremeGene (Roche; cat. #6365809001). Lentiviral production was performed as previously described (50). Briefly, 293FT cells were cotransfected with pMD2.G (Addgene; #12259), psPAX2 (Addgene; #12260), and the lentiviral expression plasmid. Viral supernatant was collected at approximately 72 hours, and debris was removed by centrifugation at 1,000 × g for 5 minutes. Cells were transduced with viral supernatant and polybrene at 8 μg/mL by spinoculation at 680 × g for 60 minutes. Drugs were used at the indicated concentrations and included imatinib (LC Laboratories; cat. #I-5508), WM-1119 (Selleck Chemicals; cat. #S8776 or MedChemExpress; cat. #HY-102058), VTP-50469 (gift of Syndax Pharmaceuticals), and EPZ-5676 (Selleck Chemicals; cat. #S7062). For growth over time assays, 15 × 10³ cells were dispensed per well in a 96-well plate, transduced with virus, or treated with the drug, and cell count was performed approximately twice per week on a Guava easyCyte Flow Cytometer (Luminex Corporation), with normalization of cell count to the control condition.

The Liu Human CRISPR Knockout Library (Addgene; #1000000132; ref. 51) targeting 18,436 genes with 185,634 sgRNAs is divided into two pooled libraries, H1 and H2, containing approximately 5 sgRNAs per gene in each library. Each virion contained an sgRNA, Cas9, and a puromycin-resistance gene derived from lentiCRISPRv2. The cell lines GIST-T1 and GIST430 were transduced in duplicate with each library (n = 8 total). For each library transduction, 44.64 × 10⁶ cells were transduced at a target multiplicity of infection (MOI) of 0.3, with an estimated library coverage of 134×. Puromycin was applied at 72 hours for selection. Cells were passaged at confluency to maintain a library coverage >134× for approximately 30 days. At the termination of the experiment, genomic DNA was extracted from 30 × 10⁶ cells per library. The region of the sgRNA between U6 and EF-1α was amplified from 200 μg of genomic DNA from each experimental replicate in 32 separate 100 μL reactions. The product was pooled, and a second PCR reaction was performed to incorporate Illumina adapters and a 6-bp barcode. A third PCR reaction was performed to enrich for full-length amplicons (primers detailed in Supplementary Table S3). Final amplicon libraries were purified by agarose gel electrophoresis and extraction with a QIAquick Gel Extraction Kit (Qiagen; cat. #28704). Next-generation sequencing was performed on a NovaSeq 6000 (Illumina). MAGeCK software (version 0.5.8) was used to analyze screen data (52). The “count” command was used to generate read counts of all libraries (n = 8), with the initial plasmid library (n = 2) used as baseline control. Total counts were normalized between samples to minimize the effects of sequencing depth. The maximum likelihood estimate command was used to generate β-scores for each screen, with data normalized to control AAVS1 sgRNAs contained within H1 and H2 libraries. Metascape was used for gene ontology enrichment analysis (53).

Cell lines stably expressing a human codon-optimized Streptococcus pyogenes Cas9 (Addgene; #73310) were generated by viral transduction. CRISPR sgRNAs were designed using CHOPCHOP (ref. 54; https://chopchop.cbu.uib.no/) and cloned into Lenti-sgRNA-EFS-GFP (LRG, Addgene; #65656) modified with GFP replaced by copGFP linked to a puromycin resistance gene by a 2A peptide, and are detailed in Supplementary Table S3. The BioID expression vectors were synthesized with codon optimization to alter sgRNA-binding sequences (Twist Bioscience). DepMap portal data were accessed through https://depmap.org/portal/ (20), utilizing the CRISPR (Avana) Public 20Q3 through 20Q4 releases.

Cell-cycle analysis was performed following drug treatment for 72 hours (imatinib) or 8 days (VTP-50469, WM-1119). Cells were trypsinized, washed in PBS, and fixed in 70% ethanol. Propidium iodide at 25 μg/mL (Life Technologies; cat. #P1304MP) and RNAse A at 0.2 mg/mL (Thermo Fisher Scientific; cat. #EN0531) were used to stain nuclear DNA. Analysis was performed on a Guava easyCyte Flow Cytometer (Luminex Corporation), and single cells were assessed for nuclear content using Guava InCyte software. Apoptosis and cell death were measured following 72 hours or 8 days of drug treatment using Guava Nexin Reagent (Luminex Corporation; cat. #4500-0450) per the manufacturer's recommendations. Nonapoptotic cells stain negative for Annexin V and 7-AAD, early apoptotic cells stain positive for Annexin V but negative for 7-AAD, and late apoptotic and dead cells stain positive for both Annexin V and 7-AAD. Staining was assayed on a Guava easyCyte Flow Cytometer, and data were analyzed using Guava InCyte software.

Cells were trypsinized and washed in PBS for RNA extraction using the RNeasy Mini Kit (Qiagen; cat. #74106). Libraries of cDNA were made using the SuperScript IV VILO cDNA Synthesis Kit (Invitrogen; cat. #11766050). RT-PCR was performed using Power SYBR Green PCR Master Mix (Life Technologies; cat. #4367659) on a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher Scientific). Relative mRNA levels were calculated by the ΔΔCt method using GAPDH expression as a reference. Primers are listed in Supplementary Table S3.

Total RNA was isolated using an RNeasy Plus Kit (Qiagen; cat. #74136), and concentration was measured by NanoDrop (Thermo Fisher Scientific) and quality by TapeStation 4200 (Agilent). Library preparation was performed using the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs; cat. #E7645S). Paired-end 150-bp sequencing was performed on a NovaSeq 6000 (Illumina). RNA-seq data were aligned to hg19 using STAR (55) with expression quantification using Cufflinks (56) to generate gene expression values in fragments per kilobase of transcript per million mapped reads (FPKM) units. GSEA (RRID:SCR_003199; ref. 57) was performed using custom gene lists or those in the Molecular Signatures Database (software.broadinstitute.org/gsea/).

For ChIP-seq, approximately 20 × 10⁶ cells were incubated in 1% formaldehyde for 10 minutes. Following fixation, excess formaldehyde was quenched with glycine at 0.125 M for 5 minutes. Samples were washed with PBS, and intact nuclei were suspended in SDS Buffer [0.5% SDS, 50 mmol/L Tris, 100 mmol/L NaCl, 5 mmol/L EDTA with protease inhibitor cocktail (Roche; cat. #11873580001)] and sonicated in a E220 Focused-Ultrasonicator (Covaris, Inc.). Sonicated samples were spun 20,000 × g for clarification, and supernatant was diluted to <0.1% SDS and then incubated with Dynabeads Protein A (Life Technologies; cat. #10002D) prebound with antibody (H3K9ac, Active Motif; cat. #39137, RRID:AB_2561017; H3K4me3, Abcam; cat. #ab8580, RRID:AB_306649; BRPF1, Thermo Fisher Scientific; cat. #PA5-27783, RRID:AB_2545259; KAT6A, Cell Signaling Technology; cat. #78462; HA, Cell Signaling Technology; cat. #3724, RRID:AB_1549585; DOT1L, Cell Signaling Technology; cat. #77087, RRID:AB_2799889; H3K79me2, Cell Signaling Technology; cat. #5427, RRID:AB_10693787) overnight. Samples were washed serially with Buffer A (150 mmol/L NaCl, 5 mmol/L EDTA, 5% sucrose, 1% Triton X-100, 0.2% SDS, 20 mmol/L Tris), Buffer B (5 mmol/L EDTA, 1% Triton X-100, 0.1% Deoxycholate, 20 mmol/L Tris), Buffer C (250 mmol/L LiCl, 1 mmol/L EDTA, 0.5% NP40, 0.5% Deoxycholate, 10 mmol/L Tris), and TE following resuspension of beads in Elution Buffer (200 mmol/L NaCl, 100 mmol/L NaHCO₃, 1% SDS) and incubation at 65°C to reverse cross-links for 12 to 15 hours. DNA was purified using AMPureXP beads (Beckman Coulter; cat. #A63881) per manufacturer's recommendation, and quality was assessed by the Qubit dsDNA HS Assay Kit (Life Technologies; cat. #Q32854) and TapeStation 4200 (Agilent). Sequencing libraries were prepared using a ThruPLEX DNA-seq Kit (Takara Bio; cat. #R400675) and sequenced on a NextSeq 500 or 550 System (Illumina). ChIP-seq spike-in normalization was performed by prebinding spike-in antibody (Active Motif; cat. #61686) together with the IP antibody of interest to Dynabeads. Equal amounts of Drosophila melanogaster chromatin (Active Motif; cat. #53083) were added to prepared GIST cell chromatin per the manufacturer's recommendations. Resultant sequenced samples were aligned to the Drosophila genome, with total Drosophila read counts used to normalize Homo sapiens read counts across samples.

Cut&Tag was performed as previously described (24) using a protein A and Tn5 transposase fusion protein (Addgene; #124601). In brief, 100,000 GIST-T1 cells were washed in Wash Buffer (20 mmol/L HEPES pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L Spermidine, protease inhibitor cocktail) and bound to Concanavalin A beads (Bangs Laboratories; cat. #BP531) for 15 minutes at room temperature. Bound cells were resuspended in 50 μL Dig-Wash Buffer (20 mmol/L HEPES pH 7.5, 150 mmol/L NaCl, 0.5 mmol/L Spermidine, protease inhibitor cocktail, 2 mmol/L EDTA, 0.05% Digitonin) and incubated with antibody diluted 1:100 overnight at 4°C (Menin, Bethyl; cat. #A300-105A, RRID:AB_2143306; MLL1n, Bethyl; cat. #A300-086A, RRID:AB_242510). A magnet was used to collect beads, and cells were resuspended in 100 μL Dig-Wash Buffer with a secondary antibody diluted 1:100 and incubated at room temperature for 30 minutes. Cells were washed three times in Dig-Wash Buffer and resuspended in Dig-Med Buffer (0.05% Digitonin, 20 mmol/L HEPES, pH 7.5, 300 mmol/L NaCl, 0.5 mmol/L Spermidine, protease inhibitor cocktail) containing 1:250 pA-Tn5 transposase and incubated at room temperature for 1 hour. Cells were washed three times in Dig-Med Buffer and resuspended in 300 μL Tagmentation Buffer (10 mmol/L MgCl2 in Dig-Med Buffer) and incubated at 37°C for 1 hour. Tagmentation was stopped by adding 10 μL of 0.5 M EDTA, 3 μL of 10% SDS, and 2.5 μL of 20 mg/mL Proteinase K (Invitrogen; cat. #25530049), and samples were incubated 50°C for 1 hour. Tagmented DNA was purified by phenol:cholorphorm:isoamyl alcohol extraction, the aqueous layer was subjected to ethanol precipitation, and DNA was resuspended in 30 μL TE. For each sample, 21 μL DNA was mixed with a universal i5 and uniquely barcoded i7 primer and amplified using NEBNext High Fidelity 2x PCR Master Mix (New England Biolabs; cat. #M0541S) in a thermocycler using the following conditions: 98°C for 30 seconds; 14 cycles of 98°C for 10 seconds, 63°C for 10 seconds; and 72°C for 2 minutes. DNA was purified with AMPureXP beads per the manufacturer's recommendation, and quality was assessed by the Qubit dsDNA HS Assay Kit and TapeStation 4200. Samples were sequenced on a NextSeq 550 System (Illumina).

All sequencing data were aligned to the human reference genome assembly hg19 using Bowtie2 (58). Normalized read density was calculated using the Bamliquidator (version 1.0) read density calculator. Aligned reads were extended by 200 bp, and the density of reads per base pair was calculated. In each region, the density of reads was normalized to the total number of million mapped reads, generating read density in units of reads per million mapped reads per bp (rpm/bp). Peak finding was performed by model-based analysis for ChIP-seq (MACS, version 1.4.2; ref. 59), and ROSE2 (60) was used to identify regions of signal enrichment. Individual ChIP-seq track displays were generated using bamplot (github.com/linlabbcm). Heat map visualizations of ChIP-seq data were generated using ChAsE (61).

Cells were lysed in RIPA Buffer containing protease inhibitor cocktail (Roche; cat. #11873580001) and centrifuged at 14,000 × g for 10 minutes to remove genomic DNA and debris. Protein concentrations were determined using a bicinchoninic acid-based assay (Pierce Biotechnology; cat. #23225). Protein samples were subjected to SDS-PAGE and Western blotting with the following antibodies: HA (1:1,000, Cell Signaling Technology; cat. #2367, RRID:AB_10691311), MEAF6 (1:500, Proteintech; cat. #26465-1-AP, RRID:AB_2880524), Actin (1:1,000, Cell Signaling Technology; cat. #4967, RRID:AB_330288), Menin (1:10,000, Bethyl; cat. #A300-105A, RRID:AB_2143306), DOT1L (1:1,000; Cell Signaling Technology; cat. #77087, RRID:AB_2799889), or streptavidin–HRP (1:40,000; Abcam; cat. #ab7403). Western blots were probed with anti-mouse or anti-rabbit secondary antibodies and detected using the Odyssey CLx infrared imaging system (LI-COR Biosciences) or streptavidin–HRP by chemiluminescence (MilliporeSigma; cat. #WBKLS0500). Immunoblots shown are representative of at least three inde­pendent experiments.

GIST-T1 cell lines were generated that stably expressed control or experimental mutant biotin ligase (BirA* R118G)–tagged fusion proteins. Twenty-four-hour biotin-labeled whole-cell lysate was subject to affinity pulldown overnight at 4°C using streptavidin–sepharose beads (GE Healthcare; cat. #17-5113-01). Beads were washed three times in 2% SDS in 50 mmol/L Tris, twice in BioID Buffer (50 mmol/L Tris, 500 mmol/L NaCl, 0.4% SDS), and six times in 50 mmol/L Tris and resuspended in 100 μL of ammonium bicarbonate. Samples were subject to tryptic digestion, and beads and salts were removed in a reverse-phase cleanup step. Extracts were dried on a speed-vac and later reconstituted in 5 to 10 μL of 2.5% acetonitrile and 0.1% formic acid. A nanoscale reverse-phase HPLC capillary column was created by packing 2.6-μm C18 spherical silica beads into a fused silica capillary (100 μm inner diameter × ∼30 cm length) with a flame-drawn tip. After equilibrating the column, each sample was loaded via a Famos Auto Sampler (LC Packings). A gradient was formed, and peptides were eluted with increasing concentrations of 97.5% acetonitrile and 0.1% formic acid. As peptides eluted, they were subjected to electrospray ionization and then entered into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein databases with the acquired fragmentation pattern by Sequest (Thermo Fisher Scientific). All databases include a reversed version of all the sequences, and the data were filtered to between 1% and 2% peptide FDR. Label-free quantification of signal intensity was used in replicate samples for quantitative comparisons. Heat maps of log₂ fold change in signal compared with DMSO were generated using Morpheus (software.broadinstitute.org/morpheus).

The PG27 PDX was obtained from a patient undergoing clinically indicated surgery and following written informed consent to a Dana-Farber Cancer Institute (DFCI) Institutional Review Board–approved research protocol. Cryopreserved tumor or the GIST-T1 cell line mixed 1:1 with matrigel was implanted subcutaneously into 6-week-old, female nude mice (NU/NU; Charles River Laboratories). GIST-T1 tested negative for Mycoplasma and rodent pathogens (Charles River Laboratories). For in vivo assessment of CRISPR/Cas9-modified cell line growth, GIST-T1/Cas9 cells were treated with the indicated sgRNAs and selected with puromycin in vitro for 14 days prior to bilateral flank implantation. For drug treatment studies, singly engrafted mice were enrolled into treatment groups when tumors reached approximately 100 to 200 mm³ in size, as measured by calipers and determined by the tumor volume equation: volume = long diameter² × short diameter × 0.5. Mice were randomly assigned to treatment groups administered imatinib (50 mg/kg gavage daily, 5 days per week), WM-1119 (50 mg/kg gavage 3 times daily, 7 days per week), VTP-50469 (0.1% in chow), or combination treatments. Imatinib was administered below the maximum tolerated dose to facilitate testing of combination therapy. No statistical methods were used to predetermine the sample size, and no animals died from drug treatment. Two GIST-T1 cell line xenograft mice in the control groups were excluded from analysis, as the subcutaneous implants failed to grow. One outlier tumor-bearing mouse in the VTP-50469 arm in Fig. 7A was excluded due to early termination from rapid tumor growth (full data are available in Supplementary Table S4). Tumors were dissected and fixed in 10% formalin for corollary studies, including hematoxylin and eosin staining and IHC of sectioned tumors. Sections (4 μm) were cut from fixed and embedded tumors and stained with Ki-67 (1:400, Cell Signaling Technology; cat. #9027, RRID:AB_2636984) and cleaved caspase-3 (1:250, Cell Signaling Technology; cat. #9579, RRID:AB_10897512). Reactions were developed using DAB (Cell Signaling Technology; cat. #8059) or NovaRed (Vector Laboratories; cat. #SK-4800) substrate kits per the manufacturers’ recommendations. All procedures were conducted under protocols approved by the Institutional Animal Care and Use Committee at the DFCI.

Center values, error bars, P-value cutoffs, number of replicates, and statistical tests are identified in the corresponding figure legends. For box plots, the box extends from the 25th to 75th percentiles, with the center line indicating the median and whiskers drawn to the 10th and 90th percentiles. Sample sizes were not predetermined.

Novel sequencing data are available through the Gene Expression Omnibus (GEO) and are available under GEO Publication Reference ID GSE172154. Additional RNA-seq, ATAC-seq, and ChIP-seq data sets analyzed in this study include GSE95864 (10), GSE113207, and GSE113217 (23).

M.L. Hemming reports a patent for the use of Menin and MOZ inhibitors in GIST pending. B.J. Aubrey is a former employee of the Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia, and has received proceeds from royalties and milestone payments related to the BCL2 inhibitor ABT-199/venetoclax and aryl sulfonohydrazide MOZ inhibitors. S.A. Armstrong reports personal fees from Neomorph, C4 Therapeutics, Imago Biosciences, Accent Therapeutics, Mana Therapeutics, and Cyteir Therapeutics and grants from Syndax, Janssen, and Novartis outside the submitted work, as well as a patent for MENIN inhibition in leukemia (WO/2017/132398A1) issued and a patent for the use of Menin and MOZ inhibitors in GIST pending. No disclosures were reported by the other authors.

M.L. Hemming: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M.R. Benson: Data curation, investigation, writing–review and editing. M.A. Loycano: Data curation, investigation, writing–review and editing. J.A. Anderson: Data curation, investigation, writing–review and editing. J.L. Andersen: Data curation, investigation, writing–review and editing. M.L. Taddei: Data curation, investigation, writing–review and editing. A.V. Krivtsov: Data curation, investigation, writing–review and editing. B.J. Aubrey: Resources, data curation, investigation, writing–review and editing. J.A. Cutler: Resources, data curation, investigation, writing–review and editing. C. Hatton: Resources, data curation, investigation, writing–review and editing. E. Sicinska: Conceptualization, resources, data curation, supervision, investigation, methodology, writing–review and editing. S.A. Armstrong: Resources, data curation, supervision, funding acquisition, investigation, project administration, writing–review and editing.

We thank the members of the Armstrong lab, the Experimental Therapeutics Core at DFCI (Prafulla Ghokhale and Benjamin Springer), and the Taplin Mass Spectrometry Facility (Ross Tomaino) for experimental guidance and support. Funding for this study was provided by NIH award R01 CA176745 (S.A. Armstrong), the Harvard Catalyst Medical Research Investigator Training Program, NIH Award UL 1TR002541 (M.L. Hemming), NIH Award K08 CA245235 (M.L. Hemming), and the DFCI Medical Oncology Grant Program (M.L. Hemming).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.