IRF4 requires ARID1A to establish plasma cell identity in multiple myeloma

Summary

Multiple myeloma (MM) is an incurable plasma cell malignancy that exploits transcriptional networks driven by IRF4. We employ a multi-omics approach to discover IRF4 vulnerabilities, integrating functional genomics screening, spatial proteomics, and global chromatin mapping. ARID1A, a member of the SWI/SNF chromatin remodeling complex, is both required for IRF4 expression and functionally associates with IRF4 protein on chromatin. Deleting Arid1a in activated murine B cells disrupts IRF4-dependent transcriptional networks and blocks plasma cell differentiation. Targeting SWI/SNF activity leads to rapid loss of IRF4-target gene expression and quenches global amplification of oncogenic gene expression by MYC, resulting in profound toxicity to MM cells. Notably, MM patients with aggressive disease bear the signature of SWI/SNF activity, and SMARCA2/4 inhibitors remain effective in immunomodulatory drug (IMiD)-resistant MM cells. Moreover, combinations of SWI/SNF and MEK inhibitors demonstrate synergistic toxicity to MM cells, providing a promising strategy for relapsed/refractory disease.

Graphical Abstract

eTOC blurb

IRF4 orchestrates oncogenic transcription in multiple myeloma. Here, Bolomsky et al. employ a multi-omic examination of IRF4 in multiple myeloma to reveal that ARID1A and SWI/SNF activity are required for IRF4-dependent transcription. Inhibition of SWI/SNF blocks plasma cell development and is toxic to myeloma cells, including cells resistant to IMiDs.

Introduction

Multiple myeloma (MM) is characterized by extensive clinical and genetic heterogeneity, which are responsible for variable responses to therapy and disparate patient outcomes. Despite these challenges, the past two decades have witnessed considerable improvements in five-year survival rates for MM patients^1–3. This achievement was in large part driven by the introduction of proteosome inhibitors and immunomodulatory drugs (IMiDs) that target molecular vulnerabilities associated with the plasmacytic origins of MM – the unifying feature of MM despite its molecular and phenotypic heterogeneity. Accordingly, recent innovations in immunotherapy approaches targeting plasma cell-specific surface proteins extended these gains. Yet even with these improved therapeutic options, almost all MM patients eventually relapse and succumb to disease⁴. Effectively treating relapsed and refractory patients remains a significant clinical challenge that necessitates new approaches targeting core vulnerabilities of plasma cell biology.

Myeloma cells coopt transcriptional networks that underpin normal plasma cell biology to drive malignant growth and survival. Transcription factors are frequently targeted by chromosomal translocation or mutation in MM, including MAF, MYC and IRF4⁵. In particular, the transcription factor IRF4 serves as the principal architect of gene expression programs supporting normal and malignant plasma cell biology⁶. IRF4 cooperates with MYC in a feed-forward loop to drive gene expression networks responsible for MM growth and survival⁷. Consequently, IRF4 is an important molecular target for treating MM. IMiDs, such as lenalidomide, indirectly decrease IRF4 expression through CRBN-dependent degradation of IKZF1 and IKZF3, which transcriptionally regulate IRF4 expression^8,9. IMiDs are a cornerstone of MM therapy but are eventually circumvented through a variety of mechanisms including CRBN mutation or loss^10,11. Directly inhibiting IRF4 function has proven difficult, and there are currently few effective options to target this critical transcription factor in IMiD-refractory MM patients. Thus, we employed a comprehensive unbiased multi-omics screening approach centered on the discovery of essential regulators of IRF4 expression and function.

Results

ARID1A promotes IRF4 expression in multiple myeloma

We sought to discover vulnerabilities in MM using a clinically refined multi-omics approach combining functional genomic CRISPR screening, unbiased spatial proteomics, and global chromatin mapping in malignant plasma cells (Fig. 1A). We first performed CRISPR-Cas9 fitness screens using the whole genome-targeted Brunello sgRNA library¹² to identify genes essential for growth and survival in 22 human MM cell lines (Fig. 1B; Table S1), expanding upon our previously reported CRISPR screening data¹³. IRF4 was the most essential MM gene after filtering for common essential genes required for basic cellular functions across all tumor types¹⁴ (Fig. 1B; Table S2), in agreement with previous data^7,15. IRF4 is a transcription factor central to imprinting plasma cell identity and MM cells remain addicted to IRF4 expression⁷. Lenalidomide and other IMiDs indirectly target IRF4 expression, and resistance to these drugs represents a major clinical challenge since there are few options to target IRF4 in IMiD-refractory MM.

Figure 1. ARID1A promotes IRF4 expression in multiple myeloma.

A) Multi-omics screening strategy to define MM-specific vulnerabilities. B) Ranked list of average CRISPR screen scores (CSS) of 22 MM cell lines. C) Average IRF4 expression score from KMS12PE, SKMM1, and H1112 IRF4-GFP knock-in cells. Left panel shows top hits. D) IRF4-BioID2 enrichment over empty vector in SKMM1 cells (FDR-corrected 2-way ANOVA, n=3). E) Gene ontology cellular component hypergeometric analysis of IRF4-BioID2 hits with an average Log2FC≥0.8 and P<0.05. F) Venn diagram of hits obtained from CRISPR essentiality screens (CSS≤-1.25), IRF4-GFP knock-in CRISPR screens (CSS≤-1.25), and IRF4-BioID2 (Log2FC≥1.0). G) Heatmap of Log2FC enrichment of SWI/SNF members in IRF4-BioID2 experiments. H) Scatter plot of average Log2FC BioID2-IRF4 enrichment from Fig. 1G (x-axis) vs. the corresponding average CSS values from Fig. 1B (y-axis). I) Western blot analysis of IRF4, ARID1A and GAPDH in MM cell lines 3 days after transduction with control shRNA (NT) or ARID1A targeting shRNAs (n=2). J) Proximity ligation assay (PLA) (red) of IRF4 and SWI/SNF family members in NCI-H929 and SKMM1 cells. DAPI (blue), wheat germ agglutinin (WGA; green). Scale bar is 10 μm, n=3. K) Representative images of ARID1A-IRF4 PLA in bone marrow aspirates from MM patients with PLA (red), CD138 (white) and DAPI (blue). Scale bar is 10 μm. L) Chromatin occupancy (CUT&RUN) of IRF4, ARID1A, SMARCA4, and SMARCB1 among IRF4 peaks in NCI-H929 and SKMM1. M) Venn diagram of IRF4 and ARID1A peaks in NCI-H929 and SKMM1. N) Transcription factor motif enrichment of ARID1A peaks (P<0.01 vs. isotype) in NCI-H929 and SKMM1. O) ARID1A mutation frequency in TCGA and MMRF-CoMMpass. See also Figure S1 and Tables S1–4.

To identify alternative routes to inhibit IRF4, we performed CRISPR-Cas9 screens to discover genes that regulate IRF4 expression. We therefore employed CRISPR-Cas9-mediated knock-in technology to generate MM cell lines (KMS12PE, H1112, and SKMM1) harboring IRF4-GFP fusion proteins under the control of endogenous regulatory elements. IRF4-GFP engineered MM cell lines were then transduced with the Brunello sgRNA library¹² and Cas9 expression was induced for one week to enable targeted gene deletion (Fig. S1A). We next sorted cells into IRF4-GFP-high and IRF4-GFP-low subpopulations (Fig. S1B) and assessed the abundance of sgRNAs in each subpopulation by next-generation sequencing. We determined an “IRF4 expression score”, defined as a Z-score of the log2 fold change (log2fc) in abundance of sgRNAs for each gene between GFP-high and GFP-low subpopulations relative to the input sample. As expected, directly targeting IRF4 yielded the most negative average IRF4 expression score (Fig. 1C; Table S3), demonstrating the validity of this screening approach. Additionally, these screens confirmed reports that EP300¹⁶ and DOT1L¹⁷ maintain IRF4 expression in MM (Fig. 1C). We also identified a host of additional genes that potentially regulated IRF4 expression, including genes associated with mTORC1 (RPS6, RPTOR, MTOR), chromatin remodeling (SMARCB1, SMARCD1, ARID1A, SMARC4A), and NF-κB signaling (REL, NFKB1, NFKBIA) (Fig. 1C).

To prioritize potential regulators of IRF4 expression for further study, we elected to focus on gene products physically associated with IRF4 protein, which we posited would directly regulate IRF4 expression. We determined protein interaction partners for IRF4 using a BioID2-based approach, which can biotinylate neighboring proteins within a 10–30 nm radius¹⁸. We ectopically expressed BioID2-IRF4 fusion constructs in four MM cell lines (RPMI 8226, SKMM1, H1112 and KMS26). Following provision of biotin for 16 hours, cells were lysed, and biotinylated proteins were purified by streptavidin pulldown and enumerated via mass spectrometry (MS). We found that IRF4 was significantly associated with scores of nuclear proteins, including IKZF1 and IKZF3 (Fig. 1D, Fig. S1C; Table S4), as previously reported^8,9. Gene Ontology (GO) cellular component enrichment analysis of protein interaction partners for IRF4 (≥0.8 log2fc) revealed a strong enrichment of the SWI/SNF chromatin remodeling machinery (Fig. 1E), including ARID1A, ARID1B, SMARCA2, SMARCA4, and SMARCB1 (Fig. 1D, Fig. S1C). An integrated analysis of our CRISPR screening and spatial proteomics datasets highlighted genes essential for IRF4 expression in MM. Comparison of genes identified by IRF4 expression CRISPR screens (≤ −1.25 ave. IRF4 expression score; Fig. 1C) with IRF4 protein interaction partners (≥ 1 ave. log2fc, Fig. 1D) and filtered for essential genes in MM (≤ −1.25 ave. CSS; Fig. 1A) resolved on five common genes – IRF4, EP300, ARID1A, SMARCB1, and SMARCD1) (Fig. 1F). Three of these genes (ARID1A, SMARCB1, and SMARCD1) were members of the SWI/SNF family.

SWI/SNF is a family of chromatin remodeling complexes that regulates chromatin accessibility and gene expression. Distinct SWI/SNF complexes have been characterized that respond to various histone modifications, termed canonical BAF (cBAF), polybromo-associated BAF (PBAF) and noncanonical BAF (ncBAF)^19,20. A granular analysis of our BioID2-IRF4 proteomic studies revealed that IRF4 associated with members of cBAF (ARID1A, ARID1B, DPF2) and SWI/SNF components shared between cBAF and other subcomplexes (SS18, SMARCA4, SMARCA2, SMARCB1) (Fig. 1G). Comparison of BioID2-IRF4 enrichments for SWI/SNF members to their respective CSSs highlighted the role of ARID1A as both an essential MM gene and IRF4 interaction partner (Fig. 1H), and knockdown of ARID1A or shared SWI/SNF components decreased IRF4 expression in MM lines (Fig. 1I, S1D–E). Additionally, we confirmed that IRF4 associated with cBAF members in MM cells by proximity ligation assay (PLA). PLA is an antibody-based method to detect protein-protein associations within 40 nm as fluorescent puncta in situ²¹. IRF4-ARID1A PLA yielded strong puncta throughout the nucleus of MM cells (Fig. 1J, S1F). Likewise, we observed robust PLA signal in the nucleus of MM cell lines between IRF4 and SWI/SNF components SMARCA2, SMARCA4, and SMARCB1 (Fig. 1J). PLA also detected IRF4-ARID1A associations in primary MM tumor samples (Fig. 1K).

To determine whether cBAF and IRF4 were functionally associated on chromatin, we performed CUT&RUN analysis²² to globally map the chromatin binding locations of IRF4 and SWI/SNF members ARID1A, SMARCA4, and SMARCB1, along with H3K4me3 and H3K27ac to illuminate sites of active transcription in NCI-H929 and SKMM1 MM lines. We found that IRF4, ARID1A, SMARCA4, and SMARCB1 CUT&RUN peaks were globally aligned, including nearly identical chromosomal positions among well-established IRF4 target genes (CD48 & SLAMF7)⁷ (Fig. 1L, S1G). Global analysis of all significant CUT&RUN peaks for IRF4 and ARID1A revealed that a remarkable 90% of IRF4 peaks overlapped with ARID1A peaks, while over 37% of ARID1A peaks coincided with IRF4 (Fig. 1M). We observed similar intersections between IRF4 and SMARCA4 or SMARCB1 (Fig. S1H). Moreover, ARID1A CUT&RUN peaks were most enriched for IRF4 DNA binding motifs (Fig. 1N), signifying that IRF4 and ARID1A were functionally associated on chromatin within MM.

Our data were consistent with ARID1A promoting IRF4-dependent MM pathogenesis. Yet SWI/SNF is widely considered to act as a tumor suppressor in human cancers, and many SWI/SNF genes, including ARID1A, are targeted by loss-of-function genomic alterations²³. Analysis of sequencing data from MMRF-CoMMpass²⁴ found that ARID1A mutations were relatively rare among MM patients compared to other tumor types profiled in TCGA, such as gynecological (TCGA-UCEC) and stomach (TCGA-STAD) cancers (Fig. 1O), suggesting that ARID1A function is preserved in MM. Examination of CRISPR essentiality data among 39 cancer types profiled in DepMap²⁵ revealed that ARID1A was preferentially essential in MM (Fig. S1I)¹⁵, confirming that ARID1A does not act as a tumor suppressor in this disease. Mutations targeting other SWI/SNF members were also generally uncommon in MM, except for BCL7A (Fig. S1J). BCL7A is targeted by aberrant somatic hypermutation²⁶, leading to loss of expression in lymphoma²⁷. We found similarly low levels of BCL7A RNA in MM (Fig. S1K), and a recent report suggests that BCL7A acts as a tumor suppressor in MM ²⁸.

Arid1a is required for plasma cell differentiation

Since IRF4 coordinates transcriptional programs responsible for normal plasma cell differentiation⁷, we hypothesized that ARID1A could be required for normal plasma cell (PC) development following B cell activation and exit from the germinal center (GC). Previous studies found that knockout of Arid1a in murine hematopoietic stem cells (HSCs) resulted in widespread disruption of hematopoiesis²⁹, while conditional deletion of Brg1 (Smarca4) in B cells interrupted B cell activation, GC formation, and PC differentiation in mice³⁰. To characterize the role of Arid1a in PC development, we examined GC formation and PC differentiation in mice with conditional deletion of Arid1a in GC B cells and their progeny. Aicda^cre/+; Arid1a^flox/flox mice with GC-specific loss of Arid1a and littermate controls were immunized with sheep red blood cells (SRBCs) and euthanized 1 week later to enumerate GC B cells and PCs from the spleen and Peyer’s patches (PPs) by FACS. We observed significant reductions of PCs in SRBC-immunized spleen and PPs following deletion of Arid1a (Fig. 2A–B; Fig. S2A). In contrast, there was only a trend toward reduction in GC B cells in spleen and no change in GC B cell numbers in PPs among Arid1a knockout animals relative to littermate controls (Fig. 2C–D). Because of the relatively small numbers of PCs in the spleen and PP, we assessed PCs in the lamina propria (LP) of the small intestine, which houses large amounts of IgA-expressing PCs derived from PP B cells, many of which are derived from PP GC B cells³¹. Loss of Arid1a in the GC resulted in a significant reduction of IgA⁺ PCs from the LP (Fig. 2E–G). Lastly, we observed a significant reduction in PC numbers in the bone marrow of Arid1a knockout animals 14 days following immunization with SRBCs (Fig. 2H).

Figure 2. ARID1A is required for plasma cell development.

A) FACS gating of PCs from spleen and Peyer’s patches (PP). B) % live PCs in spleen and PP in Aicda^cre/+; Arid1a^flox/flox mice (red, n=8 and n=7, respectively) versus littermates (gray, n=10). Line represents the mean and comparisons between groups were performed using a two-tailed unpaired t-test. C) FACS gating of GC B cells from spleen and PP. D) % live GC B cells in spleen and PP in Aicda^cre/+; Arid1a^flox/flox mice (red, n=8 and n=7, respectively) versus littermates (gray, n=10). Line represents the mean and comparisons between groups were performed using a two-tailed unpaired t-test. E) FACS gating of PCs from LP. F) % live PCs in LP in Aicda^cre/+; Arid1a^flox/flox mice (red, n=4) versus littermates (gray, n=4). Line represents the mean and comparisons between groups were performed using a two-tailed unpaired t-test. G) Representative images of IgA⁺ PCs (green) in LP. DAPI (blue) and EPCAM (white), scale bar =100 μm. H) % live PCs in BM in Aicda^cre/+; Arid1a^flox/flox mice (red, n=7), versus littermates (gray, n=6). Line represents the mean and comparisons between groups were performed using a two-tailed unpaired t-test. I) CUT&RUN tracks of ARID1A in naïve B cells, GC B cells, and PCs. J) ARID1A CUT&RUN in FACS-purified B cell subsets (columns) among enriched (P<0.01 vs isotype antibody) ARID1A peaks in the same subsets (rows). K) Top 3 motifs from transcription factor motif analysis among the B cell subset specific ARID1A peaks shown in Fig. 2J. L) LOLA of ARID1A CUT&RUN peaks in naïve B cells, GC B cells, and PCs versus public chromatin profiling datasets in murine cells (n=786). A lower max TF rank indicates stronger overlap. See also Figure S2.

To understand the role of Arid1a in PC differentiation, we mapped chromosomal locations of ARID1A during B cell differentiation. Splenic naïve B cells, GC B cells, and PCs were FACS-sorted from SRBC-immunized C57BL/6 mice and subjected to CUT&RUN analysis to identify ARID1A binding peaks on chromatin versus isotype control. As expected, we found that ARID1A peaks were enriched in PCs among PC lineage genes Xbp1, Sdc1, and Prdm1 (Fig. 2I), consistent with our findings in human MM cell lines (Fig. 1L). In contrast, ARID1A peaks were found among subtype specific genes for naïve B cells (Cxcr5, Ccr7, Ssbp2, Cd55) and GC B cells (Aicda, Bcl6, Fas), but had minimal signal in PCs (Fig. 2I, S2B). For a global perspective, we identified significantly enriched peaks for ARID1A relative to isotype control in naïve B cells, GC B cells, and PCs (Fig. 2J). Notably, we observed a stepwise increase in ARID1A binding to PC-specific sites from naïve to GCBs to functional PCs (Fig 2J). Transcription factor motif enrichment analysis among ARID1A peaks highlighted the different transcriptional factors targeted by ARID1A during B cell differentiation, with IRF-binding domains only found in PCs (Fig. 2K).

Locus overlap analysis (LOLA)³², which queries reference ChIP-seq datasets to create a ranked list of significant binding overlap, of our CUT&RUN data in naïve B cells, GC B cells, and PCs determined that Arid1a preferentially associated with IRF4, EP300, and BATF binding sites in PCs (Fig. 2L). In contrast, GC B cells preferentially associated with areas bound by SPI1 (PU.1), which is known to work with BCL6 to organize gene expression in GC B cells and suppress PC development³³. Interestingly, ARID1A bound to the same regions as POL2 and POLR2A in naïve B cells. POL2 may be paused on genes in naïve B cells to induce rapid gene expression following activation³⁴. These data suggest a broader role for ARID1A in regulating transcription through a variety of transcription factors specific to the B cell differentiation state, consistent with results from the Arid1a knockout in HSCs²⁹.

SMARCA2/4 inhibition targets the IRF4 transcriptional network

Regulation of IRF4 by ARID1A could be exploited to target transcriptional networks in MM. To explore this notion, we utilized the SMARCA2/4 degrader AU-15330 as a tool compound to block SWI/SNF-mediated chromatin remodeling³⁵. NCI-H929 and SKMM1 MM lines, and a primary MM patient sample (Fig. S3A–B) were treated with 1μM AU-15330 or DMSO control before performing ATAC-seq to identify regions of open chromatin. AU-15330 treatment led to a global loss of chromatin accessibility in MM cells among enriched ATAC peaks (Fig. 3A). LOLA found the most significantly downregulated ATAC-seq peaks in MM lines (P<0.0001, n=9980) intersected with binding sites of transcription factors known to regulate lymphocyte differentiation (IRF4, IKZF1, EP300, MEF2A, BATF) (Fig 3B). Gene ontology analysis (GREAT) of genes proximal to significantly reduced ATAC-seq peaks (P<0.0001) confirmed loss of pathways responsible for lymphocyte differentiation (Fig. 3C), consistent with a role for SWI/SNF in regulating IRF4 expression. Integrating IRF4 CUT&RUN data (Fig. 1L) corroborated this finding. ATAC-seq peaks associated with IRF4 target genes defined by CUT&RUN were significantly more downregulated by AU-15330 treatment compared to non-IRF4 target genes (Fig. 3D), and accordingly, IRF4 CUT&RUN peaks were strongly enriched among the significantly AU-15330-downregulated ATAC-seq peak regions (Fig. 3E). Moreover, CUT&RUN analysis of ARID1A and IRF4 found that three hours of AU-15330 treatment abrogated chromatin binding within IRF4-dependent genes CD48 and SLAMF7 (Fig. 3F, Fig. S3C), in agreement with the ATAC-seq results (Fig. 3A–E).

Figure 3. SMARCA2/4 inhibition targets IRF4 and its underlying transcriptional network.

A) Heatmap of identified ATAC-seq peaks in SKMM1, NCI-H929, and a MM patient sample treated with DMSO or 1 μM AU-15330. B) LOLA of downregulated ATAC-seq peaks (P<0.0001) from Fig 3A. Peak locations queried against public human chromatin profiling datasets (n=689) and inverse max ranks are plotted against −Log10 P-value. C) GREAT analysis of the top downregulated ATAC-seq peaks (P<0.0001) from Fig. 3A (ranking based on FDR-corrected binomial P-value). D) Differential ATAC-seq peak downregulation among gene locations co-occupied by IRF4 CUT&RUN peaks (IRF4 genes) or not (non-IRF4 genes). Box plots represent median and 25–75% of data, whiskers incorporate 5–95% of data, and outliers are displayed as dots. P-values were determined by Mann Whitney U test. E) IRF4 chromatin occupancy among significantly (P<0.05) and non-significantly downregulated ATAC-seq peaks from Fig. 3A. F) Representative ARID1A, IRF4 (CUT&RUN), and ATAC-seq tracks in SKMM1 cells following treatment with DMSO or 1 μM AU-15330. G) Volcano plot of differentially expressed genes by RNA-seq in SKMM1 cells 24 hours following DMSO or 1 μM AU-15330 (FDR-corrected Wald test, n=2). H) Overlap of genes significantly downregulated by RNA-seq and ATAC-seq in SKMM1 and NCI-H929. I-J) Heatmap (I) and J) transcription factor motif analysis of the 264 overlapping gene regions. K) Top 25 genes in the AU-15330-down signature. L) Kaplan-Meier plot of disease-specific survival in MMRF CoMMpass trial divided into quartiles based on the expression of the AU-15330 down signature. Significance determined by a two-sided likelihood-ratio test based on a Cox proportional hazard model with the AU-15330 down signature treated as a continuous variable. (M-Q) Average AU-15330 down signature among CoMMpass samples for M) indicated MM molecular subsets (P-values from one way ANOVA versus all samples), N) 1q21 tetraploidy, O) TP53 copy number, P) disease type (GSE13591 ⁶¹), or Q) in new diagnoses versus first relapse. Lines indicate mean expression values, comparisons between two groups (N-Q) were performed by Mann-Whitney U test. See also Figure S3 and Table S5.

We performed RNA sequencing to directly assess how AU-15330 altered gene expression in NCI-H929 and SKMM1 relative to DMSO control after 6 and 24 hours of treatment. AU-15330 substantially reduced expression of key MM and PC genes, including IRF4 and MYC, at both 6 and 24 hours (Fig. 3G; Fig. S3D), confirming that IRF4 expression was rapidly lost in the face of SWI/SNF inhibition. Among the downregulated genes were established IRF4-dependent genes CD48, SLAMF7, and CCL3 (Fig. 3G; Fig. S3D–E)^7,36. An integrated analysis of the AU-15330 ATAC-seq and RNA-seq datasets identified 264 overlapping genes (Fig. 3H). Pathway enrichment analysis (GREAT) of these 264 genes confirmed they were responsible for lymphocyte activation (Fig. S3F). We ranked these 264 genes by average decrease in gene expression between MM lines at both time points and found that IRF4 and MYC were among the most downregulated genes (Fig. 3I). Moreover, transcription factor motifs enriched within the ATAC-seq peaks from the 264 genes revealed that IRF4 motifs were overrepresented (Fig. 3J). Taken together, these data demonstrate that inhibition of SWI/SNF activity with AU-15330 leads to rapid loss of IRF4 and disruption of the IRF4-dependent transcriptional networks.

We next created a gene signature of SWI/SNF inhibition composed of 93 genes we identified that were significantly reduced following 24 hours of AU-15330 treatment in NCI-H929 and SKMM1 cells (Log2FC>2.0, P<0.01) and filtered to ensure expression in primary MM patient samples (ave. DGE ≥6.0 from CoMMpass RNA-seq data) (Fig. 3K; Table S5). The resulting “AU-15330 down” gene signature was highly correlated to patient survival in the CoMMpass study by Cox proportional hazard model (P=5.52E-8), and we found that patients in the highest quartile for expression of these genes had significantly worse disease-specific survival (Fig. 3L). Accordingly, the AU-15330 down signature was significantly enriched in MM patient samples with known clinical markers of aggressive disease, including the proliferation (PR) molecular subtype (Fig. 3M)³⁷, 1q21 tetraploidy (Fig. 3N)³⁸, and loss of TP53 (Fig. 3N,S3G)³⁹. Of note, patients with CD20⁺ MM had significantly lower expression of the AU-15330 down signature (Fig. S3H), suggesting a correlation between SWI/SNF activity and PC identity. Moreover, the AU-15330 down signature correlated with disease stage (Fig. 3P), ISS staging (Fig. S3I), and was increased at relapse (Fig. 3Q). We validated the AU-15330 down signature in an independent cohort of 559 patients with Affymetrix-based gene expression data³⁷, in which 72 of the 93 signature genes were profiled, where it again predicted poor survival (P=0.0065) (Fig. S3J) and significantly associated with markers of aggressive disease (Fig. S3K–L). Notably, the PR and AU-15330 down signature only shared one gene (Fig. S3M), demonstrating that high SWI/SNF activity is a distinct marker of aggressive MM regardless of the molecular drivers.

SMARCA2/4 inhibition is toxic to MM

We next evaluated the dose-dependent toxicity of AU-15330 and FHD-286, a SMARCA2/4 ATPase inhibitor in phase 1 clinical trials (NCT04879017, NCT04891757), in a panel of MM cell lines (Fig. 4A). We found that the SMARCA2/4 degrader AU-15330 and inhibitor FHD-286 were toxic to MM cell lines at low nM concentrations (Fig. 4A, S4A). AU-15330 and FHD-286 treatment increased p27 and BIM expression, both known to be repressed by IRF4³⁶ (Fig. S4B); hindered the cell cycle in G1 (Fig. S4C); and induced apoptosis by 72 hours (Fig. S4D). Primary MM patient samples treated with AU-15330 or FHD-286 had a dose-dependent increase in apoptosis measured by Annexin V (Fig. 4B). Moreover, AU-15330 and FHD-286 treatment decreased IRF4 expression in primary MM patient samples treated ex vivo for 48 hours, as measured by intracellular FACS, whereas lenalidomide treatment had no consistent effect on IRF4 expression in this group of samples (Fig. 4C).

Figure 4. SMARCA2/4 inhibition is toxic to MM by targeting IRF4 and MYC.

A) Dose-response of AU-15330 and FHD-286 in MM lines 7 days after treatment (mean±SD, n=3). B) Apoptosis in primary MM cells 72h after treatment. C) Intracellular IRF4 staining (MFI relative to DMSO control) 48h after treatment. D) CSS values of SWI/SNF members in 22 MM cell lines. Lines indicate median CSS. E) Ranked SWI/SNF dependency among disease lineages (DepMap). Heatmap indicates average CHRONOS scores for SWI/SNF sub-complexes in each lineage; Ranking is based on average dependency rank of all three sub-complexes. F) FHD-286 IC₅₀ values obtained 7 days after treatment in MM (n=13), GCB-DLBCL (n=15), ABC-DLBCL (n=3), and adenocarcinoma cell lines (n=5). Red indicates ARID1A mutation/loss. Lines indicate mean IC₅₀ (n=3). G) Essential MM genes associated with transcription factor motifs under SMARCA4 CUT&RUN peaks. H) Western Blot analysis in NCI-H929 and SKMM1 24h after treatment (n=2). I) Venn diagram of SMARCA4, IRF4, and MYC peaks (CUT&RUN) in NCI-H929 and SKMM1. J) GREAT analysis of shared IRF4-MYC-SMARCA4 and MYC-SMARCA4 CUT&RUN peaks (ranking based on FDR-corrected binomial P-value). K) LOLA of IRF4-MYC-SMARCA4 versus MYC-SMARCA4 CUT&RUN peaks. L) Distribution of IRF4-MYC-SMARCA4 (pink) and MYC-SMARCA4 (blue) peaks relative to transcription start sites. M) Chromatin occupancy of MYC and H3K4me3 (CUT&RUN) among significantly (P<0.05) and non-significantly downregulated ATAC-seq peaks in NCI-H929 and SKMM1 4h after AU-15330 treatment. N) Tumor volume for MM.1S cells treated with vehicle (grey), 1.5mg/kg FHD-286 (orange), or 3 mg/kg FHD-286 (red) (n=5 per group). Error bars represent SEM of tumor size. See also Figure S4.

While our proteogenomic screens identified ARID1A and cBAF as regulators of IRF4 expression, AU-15330 and FHD-286 inhibit SMARCA2/4, which power all SWI/SNF sub-complexes. In addition to ARID1A, our CRISPR fitness screens in MM identified that components of PBAF and ncBAF scored as essential genes in numerous MM cell lines (Fig. 4D). This realization prompted us to scrutinize SWI/SNF requirements more globally by examining CRISPR screening data within DepMap²⁵. We summed the average CHRONOS gene essentiality scores for members of cBAF (ARID1A, ARID1B, DPF1, DPF2, DPF3), PBAF (ARID2, PBRM1, PHF10, BRD7), and ncBAF (BRD9, GLTSCR1, GLTSCR1L) for each cell line (n=1014) and grouped them into 39 distinct tumor types, ranked by their overall dependency on SWI/SNF genes. Among all cancer types, MM had the highest dependency on SWI/SNF activity stemming from all BAF complexes, followed next by DLBCL (Fig. 4E). DLBCL is a common aggressive lymphoma that can be broadly subdivided into germinal center-like B cell (GCB) and activated B cell-like (ABC) DLBCL⁴⁰. We determined the IC₅₀ of FHD-286 on a panel of MM, GCB DLBCL, ABC DLBCL, and solid tumor adenocarcinoma (ADC) cell lines (Fig. 4F). In accordance with our DepMap analysis, little response was seen in the adenocarcinoma lines. In contrast, MM cells were most sensitive to FHD-286, even in the MM line EJM, which harbors an ARID1A^R635K mutation²⁵ (Fig. 4F; red). GCB DLBCL was highly sensitive to FHD-286, while ABC DLBCL was comparatively insensitive, even though IRF4 is an essential gene in ABC DLBCL⁴⁰. This result may be due to IRF4 expression being differentially regulated in MM and ABC DLBCL. IRF4 expression is chiefly driven by NF-κB signaling in ABC DLBCL⁴¹, whereas MYC and IRF4 reciprocally control expression of each other in MM⁷. Accordingly, ARID1A knockdown led to a loss of both IRF4 and MYC protein in MM (Fig. S4E). Of note, both GCB DLBCL and MM are characterized by frequent deregulation of MYC expression^42,43.

The above data suggested that mechanisms of SMARCA2/4 inhibitor toxicity in MM extended beyond IRF4 repression. To investigate additional genes targeted by SMARCA2/4 inhibitors in MM, we examined SMARCA4 CUT&RUN peaks from Figure 1L. Global motif enrichment of all SMARCA4 bound regions within a broad window of +/−500 bp identified over 300 associated genes. Among these genes, we found that only 18 were essential in our MM CRISPR screen data, with IRF4 and MYC ranked highest (Fig. 4G). Interestingly, MYC was more essential in MM than all other cancer types profiled in DepMap (Fig. S4F), reflecting the frequent deregulation of MYC expression by translocation, amplification, or mutation in this disease^44–46.

AU-15330 was initially reported to quench MYC expression in prostate cancer³⁵, and we observed a substantial loss of MYC RNA expression in MM lines following AU-15330 treatment (Fig. 3G, 3I). We confirmed loss of both IRF4 and MYC protein by western blot analysis in MM lines following 24 hours of treatment with either AU-15330 or FHD-286, while lenalidomide only modestly decreased IRF4 expression in the IMiD-sensitive NCI-H929 line (Fig. 4H, S4G). These findings established that inhibiting SMARCA2/4 rapidly disrupted both IRF4 and MYC protein expression in MM.

We sought to define the relative contributions of IRF4 and MYC downstream of SWI/SNF. We performed MYC CUT&RUN in NCI-H929 and SKMM1 MM lines to map locations of MYC protein on chromatin, which yielded over 35,000 significantly enriched peaks. Comparison of CUT&RUN peaks for MYC, IRF4, and SMARCA4 revealed that MYC had substantial overlap with the majority of IRF4 and SMARCA4 peaks (Fig. 4I), consistent with its role as global amplifier of oncogenic transcription⁴⁷. We subdivided SMARCA4 peaks into those bound to MYC alone or to both MYC and IRF4. GREAT gene ontology analysis determined that IRF4/MYC/SMARCA4 shared peaks were enriched for genes that regulated lymphocyte differentiation (Fig. 4J), whereas MYC/SMARCA4 shared peaks were enriched for genes that controlled cellular growth and metabolism (Fig. 4J). Similar results were seen when focused on cBAF peaks (ARID1A, SMARCA4, SMARCB1) (Fig. 4H–I). Accordingly, LOLA identified distinct binding patterns between IRF4/MYC/SMARCA4 (IRF4) and MYC/SMARCA4 (MYC) peaks (Fig. 4K). IRF4 peaks intersected with regions bound by EP300, IKZF1, NF-κB, and RUNX1, whereas MYC peaks predominantly overlapped with regions indicative of active transcription bound by POL2 and H3K4me3⁴⁸. Notably, while IRF4/MYC/SMARCA4 (IRF4) peaks had a broad distribution over their target genes, MYC/SMARCA4 (MYC) peaks were distinctly centered over transcriptional start sites (TSS), in agreement with reports that MYC preferentially binds core promoter regions⁴⁷ (Fig. 4L).

From the above data, we hypothesized that SWI/SNF regulated IRF4 in a targeted fashion while simultaneously fostering general MYC-dependent amplification of oncogenic transcriptional programs. To test this notion, we analyzed ATAC-seq data from Figure 3A to ascertain whether short-term AU-15330 treatment specifically disrupted open chromatin over MYC binding peaks. Notably, we did not observe a selective enrichment of MYC peaks among AU-15330 targeted regions (Fig. 4M), as we did for IRF4 or ARID1A peaks (Fig. 3D–E; Fig. S4J). We found a similar pattern for H3K27ac CUT&RUN peaks, while H3K4me3 marking active promoters was strongly enriched in the non-significantly downregulated regions after short-term SMARCA2/4 inhibition. (Fig. 4M; Fig. S4J). Correspondingly, we observed the genomic distribution of AU-15330-targeted regions of open chromatin was strongly enriched within intergenic locations, whereas regions bound to core promoters were exclusively found in ATAC-seq peaks unaffected by AU-15330 (Fig S4K), reminiscent of IRF4 and MYC data in Figure 4L.

Our data suggested a role for ARID1A-independent regulation of MYC promoter binding. Recent studies provided evidence that ncBAF or PBAF controlled MYC in MM^49,50. Analysis of published BRD9 ChIP-seq data from OPM2 MM cells⁴⁹ overlayed on AU-15330 sensitive and insensitive ATAC-seq peaks revealed a pattern similar to H3K4me3 (Fig. S4J). Taken as a whole, these data support a model in which MYC amplifies global transcription of IRF4/ARID1A bound PC lineage-specific genes at intergenic (enhancer) locations, while also acting to promote general oncogenic transcription of genes implicated in cell division, survival, and cellular metabolism by binding to active promoters which are preferentially targeted by PBAF and ncBAF SWI/SNF subcomplexes²⁰.

Given the potential that broad SWI/SNF inhibition could disrupt transcription in normal cells, we tested whether these drugs could be used to treat MM safely and effectively in vivo. We employed mouse xenograft models of MM to test whether FHD-286 could suppress MM growth without overt toxicity. MM.1S or SKMM1 cells were transplanted subcutaneously in NSG mice and established tumors were treated with FHD-286 at either 1.5 mg/kg/day or 3.0 mg/kg/day by oral gavage for 2 weeks. FHD-286 effectively inhibited tumor growth in MM.1S (Fig. 4N) and reduced tumor volume in SKMM1 (Fig. S4L). Moreover, FHD-286 was well tolerated at both doses with no overt toxicity (Fig. S4M), although MM.1S xenograft mice treated with 3 mg/kg/day had a reduction in body weight that approached the acceptable limits of our study. These data are in general agreement with initial reports from two recent Phase 1 trials of FHD-286 in patients with uveal melanoma and acute myeloid leukemia (AML) that reported a manageable safety profile and clinical responses in heavily pretreated patients^51,52.

SMARCA2/4 inhibitors retain efficacy in lenalidomide-resistant cells

A principal challenge in treating MM is finding therapies for patients refractory to existing treatments, particularly IMiD-refractory disease⁵³. We found that lenalidomide did not reduce IRF4 association with cBAF (Fig. S5A), whereas AU-15330 led to an expected loss of SMARCA2 and SMARCA4, as well as a reduction in IRF4 associations with components of the SWI/SNF ATPase module (SS18, ACTL6A, BCL7C) (Fig. S5B–E). These data suggest that SMARCA2/4 inhibition should still target IRF4 expression and kill MM cells resistant to lenalidomide. To test this, we generated MM.1S and NCI-H929 MM lines made resistant to lenalidomide through serial passage with increasing concentrations of lenalidomide (Fig. 5A, left). We observed no difference in viability between lenalidomide sensitive and resistant lines challenged with AU-15330 and FHD-286 (Fig. 5A, right), and IRF4 expression was quenched in IMiD-resistant cells, whereas lenalidomide was unable to disrupt IKZF3 and IRF4 protein expression in resistant cells (Fig. 5B).

Figure 5. SMARCA2/4 inhibitors effectively target IRF4 in lenalidomide-resistant cells.

A) Dose-response curves in lenalidomide sensitive and resistant variants of MM.1S and NCI-H929 72h after treatment (mean±SD, n=3). B) Western Blot analysis in NCI-H929 lenalidomide sensitive and resistant variants 24h after treatment (n=2). C) PLA between IRF4 and IKZF1 in lenalidomide sensitive and resistant NCI-H929 cells 24h after treatment with DMSO (0.1%), AU-15330 (1 μM), or lenalidomide (10 μM). Scale bar is 10 μm, (n=3). D) Corresponding median PLA scores between IRF4 and IKZF1. Box plots represent median and 25–75% of data, whiskers incorporate 5–95% of data, and outliers are displayed as dots. One-way ANOVA with Dunnet’s post test. E) PLA between IRF4 and ARID1A in lenalidomide sensitive and resistant NCI-H929 cells 24h after treatment with DMSO (0.1%), AU-15330 (1 μM), or lenalidomide (10 μM). Scale bar is 10 μm, (n=3). F) Corresponding median PLA scores between IRF4 and ARID1A. Box plots represent median and 25–75% of data, whiskers incorporate 5–95% of data, and outliers are displayed as dots. One-way ANOVA with Dunnet’s post test. H) Synergy heatmaps from XG7, INA6, and KMS12PE 72h after treatment with FHD-286 and lenalidomide at the indicated concentrations. See also Figure S5.

We next used PLA to measure interactions between IRF4 and either IKZF1 or ARID1A in lenalidomide sensitive or resistant NCI-H929 cells. As expected, lenalidomide treatment abrogated associations between IRF4 and IKZF1 in lenalidomide-sensitive, but not in lenalidomide-resistant, NCI-H929 cells (Fig. 5C–D). However, AU-15330 reduced the number of IRF4-IKZF1 puncta regardless of their lenalidomide resistance status (Fig. 5C–D), consistent with the toxicity data in Figure 5A. We also found that AU-15330 treatment significantly disrupted associations between endogenous IRF4 and ARID1A, presumably since IRF4 protein expression was lost. Lenalidomide treatment led to a significant increase of IRF4-ARID1A PLA signal exclusively in the lenalidomide-resistant cells (Fig. 5E–F).

These data suggested engagement of IRF4 by cBAF may be involved in the transcriptional plasticity reported to mediate resistance to IMiD therapies^54,55, indicating that combination therapies of SWI/SNF inhibitors and IMiDs may synergize to kill MM cells. We tested combinations of increasing doses of FHD-286 and lenalidomide in XG7, INA6, and KMS12PE MM lines. Although these cells display innate IMiD resistance, we found that FHD-286 was able to sensitize these cells to lenalidomide, resulting in synergistic toxicity (Fig. 5G).

FHD-286 synergizes with RAS pathway inhibitors

We have demonstrated that inhibiting SMARCA2/4 can effectively quench IRF4 expression and induce toxicity in MM cells, even in models of IMiD-resistance. Moreover, we found that combinations of FHD-286 and lenalidomide can enhance toxicity in MM cell lines. We next sought to identify additional combination therapies that would synergize with FHD-286 to kill MM cell lines. We implemented high-throughput combinatorial drug screens using the MIPE v6.0 library of 2804 mechanistically annotated compounds⁵⁶. SKMM1 and XG7 MM lines were screened using a series of 6x6 matrix blocks of FHD-286 versus all library compounds (Fig. S6A). We found that FHD-286 had exceptional synergy with drugs targeting effectors of the classical MAPK signaling pathway, including ERK, MEK, RAF, as well as KRAS itself in XG7, which harbors a therapeutically actionable KRAS^G12C mutation (Fig. 6A). Several small molecule inhibitors targeting MEK, including trametinib, were highly synergistically toxic with FHD-286 (Fig. 6A) and displayed a synthetic lethal phenotype (Fig. 6B–D).

Figure 6. FHD-286 synergizes with RAS pathway inhibitors.

A) Heatmap of the FHD-286 drug interaction landscape in SKMM1 and XG7. Drug synergy is ranked by average Excess HSA. B) MEK inhibitor enrichment plot from the Drug Set Enrichment Analysis (DSEA) of the FHD-286 vs MIPE6.0 screen. C) Response matrix for the FHD-286 trametinib combination in SKMM1. D) Excess HSA matrix for the FHD-286 trametinib combination in SKMM1. E) Dose-response curves of RAS-dependent MM cell lines treated with FHD-286 alone or in combination with trametinib at the indicated concentrations for 72h (mean±SD, n=3). F) MM cell apoptosis 48h after treatment (mean±SD, n=3). G) Tumor volume for MM.1S cells treated with vehicle (grey), 1 mg/kg trametinib (blue), 1.5mg/kg FHD-286 (red), or the combination (pink) (n=5 per group). Error bars represent SEM of tumor size. See also Figure S6.

MEK is typically activated downstream of oncogenic RAS, and around 50% of newly diagnosed MM tumors express oncogenic mutations in KRAS, NRAS, BRAF, or FGFR3⁵⁷. However, we have previously reported that RAS does not exclusively signal through MAPK in MM, but rather mTORC1 is the chief target of oncogenic RAS in this disease¹³. Accordingly, we observed muted activity of trametinib as a single agent in most MM cell lines relative to KRAS-dependent solid tumor cell lines¹³. Regardless, the combination of FHD-286 with trametinib resulted in substantial toxicity in RAS-dependent cell lines (SKMM1, MM.1S, XG7, NCI H929) (Fig. 6E). We observed no drug synergy in RAS-independent cell lines (KMS12PE and LP1) (Fig. S6B). The combination of FHD-286 with trametinib induced potent apoptosis in MM cells (Fig. 6F), although there was no additional reduction of IRF4 expression or changes in chromatin accessibility in IRF4-dependent genes with trametinib or combination treatments beyond what we observed for FHD-286 alone (Fig. S6C–D). Moreover, the combinatorial drug screens identified additional drugs that synergized with FHD-286 regardless of RAS activity, including mTORC1 inhibitors, such as everolimus, as well as inhibitors of CDK8. We confirmed that these combinations were synergistically toxic to MM cells regardless of RAS pathway activity (Fig. S6E–F).

Since the frequency of RAS pathway mutations is particularly high (up to 70%) in relapsed/refractory MM ⁵, combination therapies of FHD-286 and MAPK pathway inhibitors are clinically relevant. To test whether such combinations could be safe and effective in vivo, we used xenograft models of MM.1S and SKMM1, which express mutant RAS isoforms, in NSG mice. Mice were treated with either 1 mg/kg/day trametinib, 1.5 mg/kg/day FHD-286, both drugs in combination, or vehicle controls for 5 days a week for 3 weeks. We found that the combination of FHD-286 with trametinib was significantly more effective than single agents at suppressing tumor growth in both MM.1S and SKMM1 (Fig. 6G; Fig. S6G). This drug combination was not overtly toxic to the animals, and we observed no significant weight loss in treated mice (Fig. S6H). These data underscore that combination therapies centered on SMARCA2/4 inhibitors enable effective precision medicine therapies for MM patients with relapsed/refractory disease by targeting core plasma cell vulnerabilities.

Discussion

Our multi-omics approach revealed that ARID1A and SWI/SNF chromatin remodeling complexes regulate IRF4 and MYC, two essential oncogenic drivers of MM. Conditional knockout of Arid1a in GC B cells inhibited subsequent plasma cell development, highlighting the essential role of ARID1A in controlling IRF4 expression. Accordingly, SWI/SNF inhibitors repressed IRF4 expression and IRF4-dependent transcriptional networks in malignant plasma cells. The effects of these drugs were compounded by their ability to concomitantly block MYC expression, and this co-repression yielded a ‘one-two punch’ that effectively quenched oncogenic transcription and was profoundly toxic to MM cells. Importantly, these drugs effectively targeted IRF4 in IMiD-resistant MM cells, and combination therapies dramatically enhanced the efficacy of SWI/SNF inhibitors against MM by promoting rapid apoptosis of malignant cells.

Mutations targeting ARID1A and members of the SWI/SNF complex are common in cancer, contributing to the notion that SWI/SNF acts as a tumor suppressor²³. However, our data demonstrate that ARID1A and SWI/SNF activity power oncogenic transcription in MM. SWI/SNF dependency is not exclusive to MM, as SMARCA2/4 inhibitors have shown efficacy in prostate adenocarcinomas, uveal melanoma, AML, and DLBCL^35,51,52,58. However, MM stands out for its exceptional reliance on SWI/SNF subcomplexes, followed closely by other hematological malignancies. Interestingly, recent reports demonstrated that haploinsufficiency of ARID1A or SMARCA4 can drive oncogenic transcription in lymphomas via discrete lineage-specific transcription factors^58,59. Likewise, we observed distinct chromatin binding patterns for Arid1a with lineage-specific transcription factors among B cells at different differentiation stages in our murine CUT&RUN experiments. We speculate that ARID1A may cooperate with a variety of pioneer factors to enable rapid engagement of lineage-specific transcriptional programs that determine cell fate during hematopoiesis. Accordingly, SMARCA2/4 inhibitors are more effective in lineage-enhancer addicted cells, including MM³⁵.

IRF4 and MYC are fundamental to MM pathogenesis, making them prime candidates for targeted therapies. IMiDs indirectly target these factors^8,9, but these drugs are typically employed in combination with other agents, suggesting that their inhibition of the IRF4-MYC axis is insufficient. Alternative approaches targeting IRF4 have largely focused on EP300/CREBBP inhibitors¹⁶. EP300 was the only gene aside from SWI/SNF that scored in all three screening approaches implemented in our study, and a recent clinical trial suggested that this strategy can be effective in relapsed hematological malignancies, including MM⁶⁰. However, EP300 inhibitors might face the same resistance mechanisms as IMiDs, which are largely driven by rapid compensatory recruitment of alternative transcription factors to PC-specific enhancers to sustain IRF4 and MYC transcription^54,55. SWI/SNF inhibition might represent a superior strategy since the resulting shutdown of PC-specific enhancer regions in tandem with direct interference of IRF4 may prevent the recruitment of alternative transcription factors.

Our study suggests prioritizing MM for SWI/SNF inhibitor clinical trials. The SMARCA2/4 inhibitor FHD-286 is already in Phase I clinical trials for other cancers (NCT04879017, NCT04891757) and initial reports indicate that this drug is safe in humans^51,52. Not only was MM acutely sensitive to FHD-286, but we identified numerous combination therapies based on FHD-286 that can be pursued in the clinic. Notably, combination therapy can improve outcomes while lowering the effective doses needed to achieve remission. For example, treatment with FHD-286 and trametinib abolished tumor growth in vivo without obvious toxicity. Given the conserved reliance of MM on SWI/SNF-dependent IRF4 and MYC expression, we propose that FHD-286 and trametinib could serve as a backbone for multi-agent combination regimens optimized to safely treat de novo and treatment-refractory MM.

STAR Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Ryan Young (youngrm@nih.gov).

Materials Availability

Unique reagents generated in this study are available from the lead contact after completion of a Materials Transfer Agreement (MTA).

Data and Code Availability

Gene expression (RNA-seq), ATAC-seq, and CUT&RUN datasets generated in this study have been deposited in Gene Expression Omnibus (GEO) under accession number GSE253716. Mass spectrometry data was uploaded to Mass Spectrometry Interactive Virtual Environment (MassIVE) under accession number MSV000093842. Analyzed data from all CRISPR screens are provided in Supplemental Tables 1–3 and sequencing reads for all CRISPR screens are made available through our website at https://lymphochip.nih.gov/local/IRF4_MM/. This paper also analyzes existing, publicly available data: CHRONOS scores were obtained from the depmap portal (https://depmap.org/portal/) using the 23Q4 data release. The results of The Cancer Genome Atlas Progam (TCGA) analysis shown in this manuscript are in whole or part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. Public MM patient RNA-seq and whole exome sequencing data were obtained from the MMRF-CoMMpass study (IA17 release) and are available at https://research.themmrf.org. Individual patient subgroup information within the MMRF CoMMpass cohort was retrieved from a recent in-depth analysis of the dataset ⁶². These data were generated as part of the Multiple Myeloma Research Foundation Personalized Medicine Initiatives (https://research.themmrf.org and www.themmrf.org). In addition, we accessed publicly available gene expression and CHIP-seq datasets from the Gene Expression Omnibus (Accession numbers GSE2658, GSE13591, GSE197492). This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov). The accession numbers for all analyzed datasets including sources are displayed in the Key Resource Table and are publicly available as of the date of publication. All original code has been deposited at github (https://github.com/janwisn/EIB_PLA_Analysis) and is publicly available as of the date of publication.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-human IRF4	Santa Cruz	Cat.# sc-6059; RRID: AB_2127145
anti-human ARID1A	Cell Signaling Technology	Cat.# 12354; RRID: AB_2637010
anti-human SMARCA2	Cell Signaling Technology	Cat.# 11966; RRID: AB_2797783
anti-human SMARCA4	Cell Signaling Technology	Cat.# 49360; RRID: AB_2728743
anti-human SMARCB1	Cell Signaling Technology	Cat.# 91735; RRID: AB_2800172
anti-human H3K27Ac	Cell Signaling Technology	Cat.# 8173; RRID: AB_10949503
anti-human SMARCD1	Santa Cruz	Cat.# 135843; RRID: AB_2192137
anti-human AIOLOS	Cell Signaling Technology	Cat.# 15103; RRID: AB_2744524
anti-human IKAROS	Cell Signaling Technology	Cat.# 14859; RRID: AB_2744523
anti-human c-MYC	Cell Signaling Technology	Cat.# 18583; RRID: AB_2895543
anti-human GAPDH	Santa Cruz	Cat.# sc-47724; RRID: AB_627678
anti-human P27	Cell Signaling Technology	Cat.# 3686; RRID: AB_2077850+
anti-human BIM	Cell Signaling Technology	Cat.# 2933; RRID: AB_1030947
anti-human CD138 Alexa Fluor 647	Biolegend	Cat.# 356524; RRID: AB_2564251
anti-human CD54-APC	Biolegend	Cat.# 322712; RRID: AB_535984
anti-human CD98 (SLC3A2) AlexaFluor 647	Santa Cruz	Cat.# sc-59145; RRID: AB_1120783
anti-human CD38-FITC	BD	Cat.# 555459; RRID: AB_395852
anti-human CD138-BV421	BD	Cat.# 562935; RRID: AB_2737904
anti-human CD45-PerCP/C5.5	Biolegend	Cat.# 304027; RRID: AB_1236444
anti-mouse B220-FITC	Biolegend	Cat.# 103205; RRID: AB_312990
anti-mouse B220-BV785	Biolegend	Cat.# 103245; RRID: AB_11218795
anti-mouse CD4-PerCP Cy5.5	BD	Cat.# 569847; RRID: AB_3095980
anti-mouse CD4-BUV395	BD	Cat.# 568375; RRID: AB_3095981
anti-mouse TCRb-PerCP Cy5.5	Biolegend	Cat.# 109227; RRID: AB_1575173
anti-mouse GL7 Alexa Fluor 647	Biolegend	Cat.# 144605; RRID: AB_2562184
anti-mouse IgD FITC	Biolegend	Cat.# 405703; RRID: AB_315025
anti-mouse IgD BV650	Biolegend	Cat.# 405721; RRID: AB_2562731
anti-mouse CD38 PerCP Cy5.5	Biolegend	Cat.# 102721; RRID: AB_2563332
anti-mouse CD38 PE-Cy7	Biolegend	Cat.# 102717; RRID: AB_2072892
anti-mouse Fas PE-Cy7	BD	Cat.# 557653; RRID: AB_396768
anti-mouse Fas BV421	BD	Cat.# 562633; RRID: AB_2737690
anti-mouse CD138 BV421	Biolegend	Cat.# 142507; RRID: AB_11204257
anti-mouse TACI PE	Biolegend	Cat.# 133403; RRID: AB_2203542
anti-mouse IgA FITC	Southern	Cat.# 1040-02; RRID: AB_2794370
anti-mouse IgA Alexa Fluor 647	Southern	Cat.# 1040-31; RRID: AB_2794377
anti-mouse Epcam Biotin	Biolegend	Cat.# 118204; RRID: AB_1134178
anti-mouse CD4 Biotin	Biolegend	Cat.# 100508; RRID: AB_312711
anti-mouse CD8 Biotin	Biolegend	Cat.# 100704; RRID: AB_312743
anti-mouse Ter119 Biotin	Biolegend	Cat.# 116204; RRID: AB_313705
anti-mouse IgD Biotin	ebioscience	Cat.# 13-5993-82; RRID: AB_466860
anti-human IRF4-AF647	Biolegend	Cat.# 646408; RRID: AB_2564048
anti-human IRF4	eBioscience	Cat.# 14985882; RRID: AB_10804654
H3K4me3 Antibody	Epicypher	Cat.# 13-0041; RRID: AB_3076423
CUTANA™ Rabbit IgG CUT&RUN Negative Control Antibody	Epicypher	Cat.# 13-0042; RRID: AB_2923178
anti-human c-MYC	Cell Signaling Technology	Cat.# 13987; RRID: AB_2631168
anti-rabbit HRP	Cell Signaling Technology	Cat.# 7074; RRID: AB_2099233
anti-mouse HRP	Cell Signaling Technology	Cat.# 7076; RRID: AB_330924
anti-goat HRP	Santa Cruz	Cat.# sc-2354; RRID: AB_628490
Bacterial and virus strains
One Shot™ Stbl3™ E.coli	Thermo Fisher Scientific	Cat.# C737303
Chemicals, peptides, and recombinant proteins
AU-15330	MedChemExpress	Cat.# HY-145388
FHD-286	SelleckChem	Cat.# E1178
Trametinib	SelleckChem	Cat.# S2673
Everolimus	SelleckChem	Cat.# S1120
JHX-VI-178	MedChemExpress	Cat.# HY-139875
Lenalidomide	SelleckChem	Cat.#S1029
Dimethylsulfoxid (DMSO)	SigmaAldrich	Cat.#D2650
Biotin	SigmaAldrich	Cat.#B4639
5M NaCl	KD Medical	Cat.#RGF-3270
Na3VO4	SigmaAldrich	Cat.#567540
NaF	SigmaAldrich	Cat.#S7920
Tris-HCl (1M), pH7.5	Quality Biological	Cat.#351-006-101
NP-40 (10% in H₂O)	Biovision	Cat.#2111-100
Sodium deoxycholate	SigmaAldrich	D6750
cOmplete™, Mini, EDTA-free Protease Inhibitor Cocktail	SigmaAldrich	Cat.#11836170001
Ficoll-Paque PLUS density gradient media	Cytiva	Cat.#17144002
BfuA1	New England Biolabs	Cat.#R0701L
SnaB1	New England Biolabs	Cat.#R0130L
Puromycin Dihydrochloride	Thermo Fisher Scientific	Cat.#A1113802
Geneticin™ Selective Antibiotic (G418 Sulfate) (50 mg/mL)	Thermo Fisher Scientific	Cat.#10131027
Doxycyclin-Hyclat	SigmaAldrich	Cat.#D5207
Phosphate buffered saline (PBS)	Thermo Fisher Scientific	Cat.# 10010023
Advanced RPMI 1640 Medium	Thermo Fisher Scientific	Cat.# 12633012
DMEM, high glucose	Thermo Fisher Scientific	Cat.# 11965092
Fetal bovine serum	R&D Systems	Cat.# S10350
Penicillin-Streptomycin-Glutamine (100X)	Thermo Fisher Scientific	Cat.# 10378016
Opti-MEM™ Reduced Serum Medium	Thermo Fisher Scientific	Cat.#31985070
Pierce™ Universal Nuclease	Thermo Fisher	Cat.#88700
Paraformaldehyde, 16% solution	Electron Microscopy Sciences	Cat.#15710
Digitonin	Promega	Cat.#G9441
SPRIselect Beads	Beckman	Cat.#B23318
Polybrene	Santa Cruz	Cat.# sc-134220
Critical commercial assays
MycoAlert Mycoplasma Detection Kit	Lonza	Cat.# LT07-318
QIAmp DNA Blood Maxi kits	Qiagen	Cat.#13362
TMTpro 18-plex Label Reagent Set	Thermo Fisher	Cat.#A52045
EasyPep Mini MS Sample PrepKit	Thermo Fisher	Cat.#A40006
EasyPep MS Sample PrepKit Lysis Buffer	Thermo Fisher	Cat.#A45735
Acclaim™ PepMap™ 100 C18 HPLC Columns	Thermo Fisher	Cat.#164535
EASY-Spray™ HPLC Columns	Thermo Fisher	Cat.#ES902
ReproSil-Pur C18 AQ 1.9 μm reversed phase resin	Dr. Maisch GmbH	Cat.#r119.aq.
QIAamp DNA Blood Mini Kit	Qiagen	Cat.#51106
QIAGEN DNA/RNA AllPrep Kit	Qiagen	Cat.#80204
CD34 MicroBeads Kit	Miltenyi Biotec	Cat.#130-046-702
TruSeq stranded mRNA library kit	Illumina	Cat.# 20020594
4-15% gradient polyacrylamide gel	BioRad	Cat.#4561083EDU
4-12% NuPAGE BisTris Gels	Invitrogen	Cat.# NP0322BOX
Immobilon-p PVDF membrane	Millipore	Cat.#IPVH00010
2x NEBuilder HiFi DNA Assembly Master Mix	New England Biolabs	Cat.#E2621
DNA Clean and Concentrator-5 kit	ZymoResearch	Cat.#D4013
CUTANA CUT&RUN kit	Epicypher	Cat.#14-1048
CUTANA CUT&RUN Library Prep Kit	Epicypher	Cat.#14-1001
Qubit dsDNA HS Assay	ThermoFisher	Cat.#Q32851
Agilent High Sensitivity DNA Kit	Agilent	Cat.#5067-4626
E-Gel SizeSelect II Agarose 2% gel	Invitrogen	Cat.#G661012
Duolink In Situ PLA Probe Anti-Rabbit PLUS, Affinity purified Donkey anti-Rabbit IgG (H+L)	Sigma Millipore	Cat.#DUO92002-100RXN
Duolink In Situ PLA Probe Anti-Mouse MINUS, Affinity purified Donkey anti-Mouse IgG (H+L)	Sigma Millipore	Cat.#DUO92004-100RXN
Duolink In Situ Detection Reagents Far Red kit	Sigma Millipore	Cat.#DUO92013-100RXN
Duolink In Situ Wash Buffers, Fluorescence	Sigma Millipore	Cat.#DUO82049-20L
Duolink In Situ PLA Probe Anti-Goat MINUS	Sigma Millipore	Cat.#DUO92006-100RXN
Ex Taq DNA Polymerase	Takara	Cat.# RR001C
Cell Counting Kit 8 (CCK8)	SelleckChem	Cat.# B34304
MethoCult™ H4434 Classic	Stem Cell Technologies	Cat.# 04434
Mounting Medium With DAPI - Aqueous, Fluoroshield	abcam	Cat.#ab104139
Fluoromount-G	Southern Biotech	Cat.# 0100-01
Wheat Germ Agglutinin, Alexa Fluor™ 488 Conjugate	Invitrogen	Cat.#W11261
Fixable Viability Dye eFluor 780	eBioscience	Cat.# 65-0865-14
APC Annexin V	Biolegend	Cat.# 640920
Tissue-Tek O.C.T Compound	Sakura	Cat.# 4583
Pierce™ High Capacity Streptavidin Agarose	Thermo Fisher	Cat.#20359
Dynabeads™ Mouse CD8 (Lyt2)	Thermo Fisher Scientific	Cat.#11447D
Pierce BCA protein assay kit	Thermo Fisher	Cat.#23225
4x Laemmli sample buffer	BioRad	Cat #1610747
Lenti-X concentrator	CloneTech	Cat.#631231
Trans-IT 293T	Mirus	Cat.#6603
Ex Taq Polymerase	TaKaRa	Cat.#RR006
BD Transcription Factor Buffer Set	BD	Cat.# 562574
BD Cytofix/Cytoperm kit	BD	Cat.# 554714
FxCycle PI/RNase Staining Solution	Thermo Fisher	Cat.# F10797
Deposited data
Raw ATAC-seq data	This manuscript	GSE253716
Raw CUT&RUN data	This manuscript	GSE253716
Raw RNA sequencing data	This manuscript	GSE253716
Raw proteomic data	This manuscript	MSV000093842
CRISPR screen read count data	This manuscript	https://lymphochip.nih.gov/local/IRF4_MM/
CoMMpass study IA17 release	MMRF	https://research.themmrf.org
BRD9 CHIP-seq data	Kurata et al. 2023 49	GSE197492
Gene expression data of TT2 and TT3 pre-treatment samples	Zhan et al. 2006 37	GSE2658
Gene expression of different MM stages	Agnelli et al. 2009 61	GSE13591
TCGA data	TCGA Research Network	https://www.cancer.gov/tcga
CHRONOS scores 23Q4 release	Depmap portal	https://depmap.org/portal/
Experimental models: Cell lines
Human: ARP1	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_D523
Human: EJM	DSMZ	Cellosaurus; RRID: CVCL_2030
Human: H1112	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_A485
Human: INA-6	DSMZ	Cellosaurus; RRID: CVCL_5209
Human: JJN-3	DSMZ	Cellosaurus; RRID: CVCL_2078
Human: JK-6L	DSMZ	Cellosaurus; RRID: CVCL_0C41
Human: KMS-12-PE	JCRB	Cellosaurus; RRID: CVCL_1333
Human: KMS-26	JCRB	Cellosaurus; RRID: CVCL_2992
Human: KMS34	JCRB	Cellosaurus; RRID: CVCL_2996
Human: L-363	DSMZ	Cellosaurus; RRID: CVCL_1357
Human: LP-1	DSMZ	Cellosaurus; RRID: CVCL_0012
Human: MM.1-144	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_EI97
Human: OCI-My5	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_E332
Human: RPMI-8226	ATCC	Cellosaurus; RRID: CVCL_0014
Human: SK-MM-1	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_A478
Human: XG-7	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_A772
Human: XG-6	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_A771
Human: MM1.S	ATCC	Cellosaurus; RRID: CVCL_8792
Human: NCI-H929	ATCC	Cellosaurus; RRID: CVCL_1600
Human: OPM-2	DSMZ	Cellosaurus; RRID: CVCL_1625
Human:NCI-H929 LEN resistant	This manuscript	N/A
Human: MM.1S LEN resistant	This manuscript	N/A
Human: SU-DHL-6	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_2206
Human: ULA	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1854
Human: MC116	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1399
Human: TK	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_3216
Human: WILL-2	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1901
Human: Mieu	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_H208
Human: OCI-Ly4	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_8801
Human: SU-DHL-7	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_4380
Human: OZ	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_M710
Human: RC	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_9U45
Human: WSU-DLCL2	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1902
Human: SU-DHL-5	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1735
Human: HKBML	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_8161
Human: LIB	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_H209
Human: ShK	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_B3QN
Human: RIVA	Lab of Louis Staudt	Cellosaurus; RRID: CVCL_1885
Human: TMD8	Lab of Louis Staudt	CelloSaurus, RRID: CVCL_A442
Human: HBL1	Lab of Louis Staudt	CelloSaurus, RRID: CVCL_4213
Human: NCI-H1299	ATCC	Cellosaurus; RRID: CVCL_0060
Human: A-549	ATCC	Cellosaurus; RRID: CVCL_0023
Human: LS180	ATCC	Cellosaurus; RRID: CVCL_0397
Human: GP2d	Sigma	Cellosaurus; RRID: CVCL_2450
Human: LoVo	ATCC	Cellosaurus; RRID: CVCL_0399
Human: HCT-8	ATCC	Cellosaurus; RRID: CVCL_2478
Human: PANC-1	ATCC	Cellosaurus; RRID: CVCL_0480
Human: HEK293-FT	Thermo Fisher	Cellosaurus; RRID: CVCL_6911
Experimental models: Organisms/Strains
Mouse: B6.129P2-Aicdatm1(cre)Mnz/J	The Jackson Laboratory	Stock# 007770; RRID: IMSR_JAX:007770
Mouse: Arid1atm1.1Zhwa/J	The Jackson Laboratory	Stock# 027717; RRID: IMSR_JAX:027717
Mouse: Female NSG mice	NCI	N/A
Oligonucleotides
shControl non-targeting GCCAAGATTCAGAATCCCAAA	Yang et al. 2022 13	N/A
shARID1A.1 CGGCTCACAATGAAAGACATT	Mission shRNA Library (Sigma)	N/A
shARID1A.1 CCTCTCTTATACACAGCAGAT	Mission shRNA Library (Sigma)	N/A
shSMARCA4.1 CCATATTTATACAGCAGAGAA	Mission shRNA Library (Sigma)	N/A
shSMARCA4.2 CCGAGGTCTGATAGTGAAGAA	Mission shRNA Library (Sigma)	N/A
shSMARCB1.1 CCACCAGTGTGACCCTGTTAA	Mission shRNA Library (Sigma)	N/A
shSMARCB1.2 ACGGAGCATCTCAGAAGATTG	Mission shRNA Library (Sigma)	N/A
shSMARCD1.1 CTTGGTGATTGAACTGGACAA	Mission shRNA Library (Sigma)	N/A
shSMARCD1.2 CCATGAGACAATAGAAACCAT	Mission shRNA Library (Sigma)	N/A
Recombinant DNA
Brunello sgRNA library	Addgene	Cat.#73178
pMD2.G	Addgene	Cat.#12259
psPAX2	Addgene	Cat.#12260
pRetroCMV/TO-Cas9-Hygro	Phelan et al. 2018 40	N/A
pHIT60	Phelan et al. 2018 40	N/A
pHIT/EA6x3*	Phelan et al. 2018 40	N/A
pLKO.1 Puro U6	Addgene	Cat.#52628
pLKO.1 Puro-GFP	Phelan et al. 2018 40	N/A
pBMN-BioID2-IRF4-IRES-LYT2	This manuscript	N/A
GFP-P2A-CD34 donor plasmid for IRF4 tagging	Scheich et al. 2023 64	N/A
sgIRF4-pLKO.1-puro-U6 for IRF4 tagging	Scheich et al. 2023 64	N/A
pBMN-MCS-IRES-LYT2	Lab of Louis Staudt	N/A
Software and algorithms
Basespace	Illumina	https://www.illumina.com/products/by-type/informatics-products/basespace-sequence-hub/apps.html
Bowtie 2 version 2.2.9	N/A	https://bioweb.pasteur.fr/packages/pack@bowtie2@2.2.9
Graphpad Prism v.9	https://www.graphpad.com/scientific-software/prism/	RRID:SCR_002798
BD FACSDiva software version 8.0.1	BD	RRID:SCR_001456
Image Lab Touch Software (v2.3.0.07)	Biorad	https://www.bio-rad.com/en-us/product/image-lab-touch-software?ID=PJW3UUTU86LJ
GREAT	McLean et al. 2010 72	http://great.stanford.edu/great/public/html/
LOLA	Sheffield et al. 2016 32	http://lolaweb.databio.org/
STAR	Dobin et al. 2012 74	https://github.com/alexdobin/STAR
GSEA (v4.0.3)	Subramanian et al. 2005 78	RRID:SCR_003199
FlowJo (v.10)	BD	RRID:SCR_008520
FIJI 1.53f51	ImageJ	RRID:SCR_002285
Proteome Discoverer 2.4	Thermo Fisher	https://www.thermofisher.com/order/catalog/product/OPTON-31014?SID=srch-srp-OPTON-31014
R for statistical computing (version 4.2.2)	R Core Team	RRID: SCR_001905
CytExpert software	Beckman Coulter	RRID:SCR_017217
Leica Application Suite X version 4.5.0.2531	Leica	RRID:SCR_013673
CellQuest Pro version 6.0	BD	RRID:SCR_014489
Other
Sony MA900 FACS cell sorter	Sony Biotechnology	N/A
Neon Transfection System	Invitrogen	N/A
Illumina NextSeq2000 sequencer	Illumina	N/A
Leica CM1950 cryostat	Leica	N/A
Leica Stellaris 5 Confocal microscope	Leica	N/A
Beckman Coulter Cytoflex LX flow cytometer	Beckman Coulter	N/A
Beckman Coulter Cytoflex S flow cytometer	Beckman Coulter	N/A
UltiMate 3000 RSLCnano HPLC system	Thermo Fisher Scientific	N/A
Q Exactive HF orbitrap mass spectrometer	Thermo Fisher Scientific	N/A
BD FACS Calibur flow cytometer	BD Biosciences	N/A

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

Female NSG (non-obese diabetic/severe combined immunodeficient/common gamma chain deficient) mice were obtained from NCI Fredrick Biological Testing Branch and used for the xenograft experiments between 6–8 weeks of age. Mice were housed in specific pathogen-free facility in ventilated microisolator cages with 12 hour light and 12 hour dark cycles at 72F and 40–60% relative humidity. Approved protocols allowed tumor growth below 20 mm in any dimension; no animals had tumors which exceeded these limits.

Aicda^cre/+ Arid1a^f/f and littermate controls were generated by intercrossing Aicda^cre (B6.129P2-Aicda^tm1(cre)Mnz/J; Stock number 007770) mice on a B6 background and Arid1a^f/f (Arid1a^tm1.1Zhwa/J; Stock number 027717) mice that were on a mixed background 129/B6 background from The Jackson Laboratory. Animals of both sexes were analyzed between 7 and 12 weeks of age. Mice were injected sheep red blood cells (SRBC; Colorado Serum Company) i.p. on days −2 and 0 and analyzed at day 7 or 14. All mouse experiments were approved by the National Cancer Institute Animal Care and Use Committee (NCI-ACUC) and were performed in accordance with NCI-ACUC guidelines and under approved protocols.

Cell lines

Human multiple myeloma (ARP1, female donor; EJM, female donor; H1112, unspecified sex; INA-6, male donor; JJN-3, female donor; JK-6L, male donor; KMS-12-PE, female donor; KMS-26, male donor; KMS34, female donor; L-363, female donor; LP-1, female donor; MM.1-144, female donor; OCI-My5, unspecified sex; RPMI-8226, male donor; SK-MM-1, male donor; XG-7, female donor; XG-6, female donor; MM1.S, female donor; NCI-H929, female donor; OPM-2, female donor; NCI-H929 LEN resistant, female donor; MM.1S LEN resistant, female donor), diffuse large B cell lymphoma (SU-DHL-6, male donor; ULA, male donor; MC116, male donor; TK, female donor; WILL-2, female donor; MIEU, unspecified sex; OCI-Ly4, male donor; SU-DHL-7, female donor; OZ, male donor; RC, female donor; WSU-DLCL2, male donor; SU-DHL-5, female donor; HKBML, male donor; LIB, unspecified sex; ShK, unspecified sex; RIVA, female donor; TMD8, male donor; HBL1, male donor) and adenocarcinoma cell lines (NCI-H1299, male donor; A-549, male donor; LS180, female donor; GP2d, female donor; LoVo, male donor; HCT-8, male donor; PANC-1, male donor; HEK293-FT, female donor) were obtained from the sources indicated in the Key Resources Table. Lenalidomide resistant variants of NCI-H929 and MM.1S were generated by continuous exposure to gradually increasing doses of lenalidomide for >6 months. Routine growth of cell lines was performed in advanced RPMI medium supplemented with 4% fetal bovine serum (Tet tested, R&D Systems) and 1x penicillin/streptomycin/L-glutamine. HEK293FT cells were cultured in Dulbecco’s modified Eagle medium (DMEM) containing 10% FBS and 1% penicillin/streptomycin/L-glutamine. All cell lines were grown at 37°C with 5% CO₂, regularly tested for mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza), and DNA fingerprinted by examining 16 regions of copy number variants⁶³.

Primary cell cultures

Primary bone marrow mononuclear cells (BM MNCs) were obtained from myeloma patients undergoing routine BM aspiration at the NIH Clinical Center. The study was approved by the institutional review board (protocol #01-C-0129), and written informed consent for the use of material for scientific studies was obtained from these patients in accordance with the Declaration of Helsinki. BM MNCs were collected from routine BM aspirates by Ficoll-Paque (Cytiva) density gradient centrifugation for 40 min at 400 g at room temperature (breaks off), washed twice with PBS and resuspended in advanced RPMI medium supplemented with 4% fetal bovine serum (Tet tested, R&D Systems) and 1x penicillin/streptomycin/glutamine. Cells were incubated at 37 °C in humidified incubators with 5% CO₂ in the presence of DMSO (Sigma), AU-15330 (MedChemExpress) or FHD-286 (SelleckChem) for 2.5 hours or up to 72 hours before performing ATAC-seq or flow cytometry-based assays (see method details). All primary patient samples were de-identified before analysis.

METHOD DETAILS

CRISPR essentiality screens

For the generation of Cas9 MM clones and CRISPR essentiality screens, MM cells were transduced with pRetroCMV/TO-Cas9-Hygro and selected for 2 weeks. Single cell clones were subsequently derived from Cas9-transduced cell pools and tested for Cas9 activity using sgRNAs for CD54 or CD98. Only clones with exceptional Cas9 activity (> 90% negative cells for CD54 (1:200, Biolegend, clone HDCD54) or CD98 (1:200, Santa Cruz Biotechnologies, clone E-5) surface expression by FACS) were used for further experiments. CRISPR screens were performed using the Brunello sgRNA library ¹² (Addgene #73178) which was packaged in 293FT cells (Invitrogen) with helper plasmids pPAX2 (Addgene #12260) and pMD2.g (Addgene #12259) in a 4:3:1 ratio. 293FT supernatants were harvested at 24, 48, and 72 hours, pre-cleared by centrifugation at 1000g for 5 min and concentrated (40X) using Lenti-X concentrator (Takara) following the manufacturer’s instructions. Concentrated Brunello lentiviral library was added to Cas9 MM clones (2 biological replicates per screen) to yield ~ 30% infection efficiency and maintain ~1 sgRNA per cell with an average of 500 copies per sgRNA in total. Puromycin selection was started 3 days after transduction. At day 0 (72h post puromycin), 50x10⁶ cells were harvested and saved as input sample at −80C. At least 50x10⁶ cells were cultured in the presence of 200 ng/ml doxycycline and 0.5 μg/ml puromycin to induce Cas9 expression and passed every other day for 21 days. 50x10⁶ cells were harvested for the day 21 timepoint. DNA was extracted from Day 0 and 21 cell pellets with QIAmp DNA Blood Maxi kits (Qiagen).

CRISPR IRF4-GFP knock-in screens

IRF4-GFP knock-in cells were generated as recently described ⁶⁴. In brief, 2x 10⁵ Cas9 expressing MM cells were electroporated (Neon Transfection System, Thermo Fisher) with the pLKO.1 vector carrying a sgRNA to target IRF4 at the C-terminus and a donor vector carrying a GFP-P2A-truncCD34 sequence for knock-in via non-homology directed repair mechanisms. After recovery from electroporation, knock-in cells were enriched using the CD34 MicroBeads Kit (Miltenyi Biotec) and subsequently sorted for GFP-positivity on a MA900 cell sorter (Sony) to obtain pure knock-in cell pools. Successfully generated SKMM1, H1112, and KMS12PE IRF4-GFP knock-in cells were then transduced in duplicate with the Brunello sgRNA library as described above. Cells were selected with puromycin for 3 days, followed by Cas9 induction with doxycycline and cell expansion for 7 days. At this point, 50x10⁶ cells per biological replicate were collected and reserved as an input sample. The remaining ~80x10⁶ cells were washed in PBS with 0.5% FBS and resuspended at 20x10⁶ cells/ml in PBS with 0.5% FBS. Live cells were sorted for the top 8% highest and lowest GFP signal on a Sony MA900 to obtain ~1.5x10⁶ GFP-low and GFP-high cells, respectively. DNA from sorted samples was extracted using the QIAamp Blood Mini Kit (QIAGEN) following the manufacturer’s instructions.

CRISPR Library amplification

CRISPR screens were amplified according to established protocols ¹³. In brief, a nested PCR strategy was used to first isolate sgRNA sequences from genomic DNA followed by the addition of a nextgen sequencing adapter compatible with the Illumina NextSeq2000. All products were amplified for 17 cycles per each round using ExTaq polymerase (Takara). PCR products were isolated using SizeSelect II 2% agarose gels (Invitrogen) and quantified using the Qubit dsDNA high-sensitivity assay (Thermo). The resulting libraries were sequenced with a NextSeq2000 (Illumina) running NextSeq 1000/2000 Control Software (v. 1.2.036376) (Illumina) and demultiplexed using DRAGEN (v.3.7.4) (Illumina) and aligned using Bowtie2 (version 2.2.9).

Protein interactomes (BioID2)

Synthetic gene fragments of IRF4 (Twist) were cloned into the BioID2-8Xlinker-pBMN-LYT2 vector ¹³ via Gibson cloning (New England Biolabs). Gene fragments (150 ng) were mixed with 1 μl SnaBI cut BioID-2 vector and 4 μl of the 2x NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs), and incubated for 1 hour at 50°C. The assembled vector was then transformed into Stbl3 cells for amplification. Sequence-verified BioID2 constructs were packaged into retrovirus using 293T cells (ATCC) with helper plasmids pHIT60 and pHIT/EA6x3* in a 2:1:1 ratio. Transduced MM cells were purified with anti-LYT2 (mouse CD8) magnetic beads (ThermoFisher). Once >95% lyt2-positive cell pools were obtained, cells were expanded for approximately 2 weeks to obtain 40x10⁶ cells. On the day prior to lysis, biotin (Sigma) was added to a final concentration of 50 μM to BioID2-IRF4 carrying cells. Cells were incubated for 16 hours in the presence of biotin and then lysed at 1 x 10⁷ cells per ml in RIPA buffer modified for MS analysis (1% NP-40, 0.5% deoxycholate, 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM Na₃VO₄, 5mM NaF, 1 mM AEBSF) for 10 min on ice. Lysates were cleared by centrifugation at 14,000xg for 20 min at 4°C. Subsequently, 30 μl of pre-washed streptavidin agarose beads (ThermoFisher) was added and samples were rotated at 4°C for 2 hours. At the end of the incubation period, samples were washed two times in 1X RIPA buffer, 1x in PBS, and 1x in 50 mM HEPES and stored at −80°C for MS analysis. Three independent biological replicates were collected for all BioID2 experiments.

For MS analysis, proteins were separated by one-dimensional gel electrophoresis (4–12% NuPAGE BisTris Gel; Invitrogen), and the entire lane of a Coomassie blue-stained gel was cut into 23 slices. All slices were processed as described previously ⁶⁵. After tryptic digestion of the proteins the resulting peptides were resuspended in sample loading buffer (2% acetonitrile and 0.05% trifluoroacetic acid) and were separated by an UltiMate 3000 RSLCnano HPLC system (Thermo Fisher Scientific) coupled online to a Q Exactive HF mass spectrometer (Thermo Fisher Scientific). First, peptides were desalted on a reverse phase C18 precolumn (Dionex 5 mm length, 0.3 mm inner diameter) for 3 minutes. After 3 minutes the precolumn was switched online to the analytical column (30cm length, 75 mm inner diameter) prepared in-house using ReproSil-Pur C18 AQ 1.9 μm reversed phase resin (Dr. Maisch GmbH). Buffer A consisted of 0.1 % formic acid in H₂O, and buffer B consisted of 80% acetonitrile and 0.1% formic acid in H₂O. The peptides eluted from buffer B (5 to 42 % gradient) at a flow rate of 300 nl/min over 76 min. The temperature of the precolumn and the analytical column was set to 50°C during the chromatography. The mass spectrometer was operated in a TopN data-dependent mode, where the 30 most intense precursors from survey MS1 scans were selected with an isolation window of 1.6 Th for MS2 fragmentation under a normalized collision energy of 28. Only precursor ions with a charge state between 2 and 5 were selected. MS1 scans were acquired with a mass range from 350 to 1600 m/z at a resolution of 60,000 at 200 m/z. MS2 scans were acquired with a starting mass of 110 Th at a resolution of 15,000 at 200 m/z with maximum IT of 54ms. AGC targets for MS1 and MS2 scans were set to 1E6 and 1E5, respectively. Dynamic exclusion was set to 20 seconds.

Mass spectrometry

For the digestion of IRF4 BioID2 proteins and TMTpro labeling, each sample bound to capture beads containing 50μL 50mM HEPES pH 8 was treated with 150μL of digestion buffer containing the following composition: 10mM TCEP, 40mM chloroacetamide, and 15ng/μL trypsin/LysC in lysis buffer provided with the EasyPep kit (Thermo Fisher). Samples were incubated at 37°C overnight in the dark at 1000rpm. After digestion, 175μL of each solution was transferred to a new tube and treated with 10μL of 10μg/μL TMTpro (ThermoFisher) reagent and incubated for 1hr at 25 °C with shaking. Excess TMTpro was quenched with 50μL of 5% hydroxylamine, 20% Formic acid for 10min and samples within each plex were then combined. Combined plexes were cleaned using EasyPep mini columns (ThermoFisher) as described in the manual. Eluted peptides were dried in speed-vac.

For the digestion of global cellular proteins and TMTpro labeling, each cell pellet was lysed in 500μL EasyPep Lysis buffer (Thermo Fisher) and treated with 1μL universal nuclease (Thermo Fisher). Protein concentration was determined by the BCA method and 10μg was taken from each sample for digestion. Samples were adjusted to 100μL total with lysis buffer and treated with 100 μL of digestion buffer containing the following composition: 10mM TCEP, 40mM chloroacetamide, and 15ng/μL trypsin/LysC in 100mM HEPES pH 8. Samples were incubated at 37°C overnight in the dark. Then 10μL of 10μg/μL TMTpro 18-plex label (Thermo Fisher) was added to the samples followed by incubation for 1 hr at 25°C. Excess TMTpro was quenched with 50μL of 5% hydroxylamine, 20% Formic acid for 10min, and samples within each plex were then combined. Samples were cleaned using EasyPep mini columns provided with the EasyPep kit as described in the manual. Eluted peptides were dried in speed-vac. For LC/MS analysis of BioID2 experiments, peptides were resuspended in 50μL of 0.1% FA and 5μL was analyzed in duplicate using a Dionex U3000 RSLC in front of a Orbitrap Eclipse (Thermo) equipped with an EasySpray ion source. Solvent A consisted of 0.1%FA in water and Solvent B consisted of 0.1%FA in 80%ACN. Loading pump consisted of Solvent A and was operated at 7 μL/min for the first 6 minutes of the run and then dropped to 2 μL/min when the valve was switched to bring the trap column (Acclaim^™ PepMap^™ 100 C18 HPLC Column, 3μm, 75μm I.D., 2cm) in-line with the analytical column (EasySpray C18 HPLC Column, 2μm, 75μm I.D., 25cm). The gradient pump was operated at a flow rate of 300nL/min and each run used a linear LC gradient of 5-7%B for 1min, 7-30%B for 83min, 30-50%B for 25min, 50-95%B for 4min, holding at 95%B for 7min, then re-equilibration of analytical column at 5%B for 17min. All MS injections employed the TopSpeed method with three FAIMS compensation voltages (CVs) and a 1 second cycle time for each CV (3 second cycle time total) that consisted of the following: Spray voltage was 2200V and ion transfer temperature of 300 °C. MS1 scans were acquired in the Orbitrap with resolution of 120,000, AGC of 4e5 ions, and max injection time of 50ms, mass range of 350–1600 m/z; MS2 scans were acquired in the Orbitrap using TurboTMT method with resolution of 15,000, AGC of 1.25e5, max injection time of 22ms, HCD energy of 38%, isolation width of 0.4Da, intensity threshold of 2.5e4 and charges 2–6 for MS2 selection. Advanced Peak Determination, Monoisotopic Precursor selection (MIPS), and EASY-IC for internal calibration were enabled and dynamic exclusion was set to a count of 1 for 15sec.

For LC/MS analysis of cellular proteomes, global cellular peptides were resuspended in 50μL of 0.1% FA and loaded 5μL twice onto a Dionex U3000 RSLC in front of a Orbitrap Eclipse (Thermo) equipped with an EasySpray ion source. Solvent composition and column configuration was the same as described in the previous section. The gradient pump was operated at a flow rate of 300nL/min and each run used a linear LC gradient of 5-7%B for 1min, 7-30%B for 133min, 30-50%B for 35min, 50-95%B for 4min, holding at 95%B for 7min, then re-equilibration of analytical column at 5%B for 17min. All MS injections employed the TopSpeed method with four FAIMS compensation voltages (CVs) and a 0.75 second cycle time for each CV (3 second cycle time total) that consisted of the following: Spray voltage was 2200V and ion transfer temperature of 300 °C. MS1 scans were acquired in the Orbitrap with resolution of 120,000, AGC of 4e5 ions, and max injection time of 50ms, mass range of 350–1600 m/z; MS2 scans were acquired in the Orbitrap using TurboTMT method with resolution of 15,000, AGC of 1.25e5, max injection time of 22ms, HCD energy of 38%, isolation width of 0.4Da, intensity threshold of 2.5e4 and charges 2–6 for MS2 selection. Advanced Peak Determination, Monoisotopic Precursor selection (MIPS), and EASY-IC for internal calibration were enabled and dynamic exclusion was set to a count of 1 for 15sec.

shRNA and sgRNA mediated gene silencing

Lentiviral packaging and transduction of shRNAs and sgRNAs was performed as described above. In brief, shRNA and sgRNA vectors were packaged in 293FT cells (Invitrogen) with helper plasmids pPAX2 (Addgene 12260) and pMD2.g (Addgene 12259) in a 4:3:1 ratio. 293FT supernatants were harvested at 24 and/or 48 hours post transfection. MM cells were transduced for 90 min at 2500 rpm at 30°C, supplemented with fresh complete media and incubated overnight before further experiments. For shRNA experiments, transduced MM cells were selected with 1 μg/ml puromycin (Gibco) for 2 days and then probed for Western Blot analysis. Individual shRNAs were obtained from the MISSION shRNA Library from the RNAi Consortium in the pLKO.1 vector (Sigma). For sgRNA experiments, transduced MM cells were left untreated or selected with 1μg/ml puromycin. sgRNAs were cloned into the pLKO.1-puro-U6 or pLKO.1-puro-GFP vector.

Proximity Ligation Assay

MM cells were plated onto a 15 well m-Slide Angiogenesis ibiTreat chamber slide (Ibidi) and allowed to adhere to the surface for 30 min at 37°C. Cells were next fixed with 4% paraformaldehyde (Electron Microscopy Sciences) for 20 min at room temperature and then washed in PBS (Gibco). Cells were permeabilized in cold methanol for 20 min at −20°C, washed in PBS, and then blocked in Duolink Blocking buffer (Sigma) for 30 min at room temperature. Primary antibodies were diluted in Duolink Antibody Diluent (Sigma) and incubated overnight at 4°C. Cells were next washed for 2x 4 min in TBST with 0.05% tween-20, followed by addition of the appropriate Duolink secondary antibodies (Sigma), diluted and mixed according to the manufacturer’s instructions. Cellular membranes were labeled by the addition of 5 mg/ml wheat germ agglutinin (WGA) conjugated to Alexa Fluor 488 (Thermo Fisher), and cells were incubated for 1 hour at 37°C, after which the plates were washed in TBST with 0.05% tween-20 for 2x5 min. Ligation and amplification steps of the PLA were performed using the Duolink in situ Detection Reagents Far Red kit (Sigma) according to the manufacturer’s instructions. Following the PLA, cells were mounted in Fluoroshield Mounting Medium with DAPI (Abcam). Images were acquired on a Leica Stellaris 5 Confocal microscope using Leica Application Suite X version 4.5.0.2531. Images for display were prepared with NIH ImageJ/FIJI software version 2.9.0/1.53t ⁶⁶.

PLA on formalin-fixed, paraffin-embedded (FFPE) biopsy samples from MM patients was performed in a similar manner. Samples were deparaffinized in xylene and rehydrated in graded alcohol and distilled water. Heat induced antigen retrieval was performed at pH 6.0 for 30 minutes. Slides were then placed in tris-buffered solution and prepared for proximity ligation assay, as described above. MM samples were co-stained with mouse anti-human CD138-Alexa647 (Biolegend, clone MI15). All primary patient samples were de-identified before PLA analysis.

Western blot analysis

For Western Blot analysis, cells were lysed on ice at 1 x 10⁷ cells per ml in modified RIPA buffer (1% NP-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM Na₃VO₄, 5mM NaF, 1 mM AEBSF) for 15 min. Lysates were cleared by centrifugation at 14,000xg for 20 min at 4°C, and stored at −80°C. Protein concentrations were determined using the Pierce BCA protein assay kit (Thermo) following the manufacturer’s instructions. Total protein lysates were combined with 4X Laemmli sample buffer (Bio-Rad) supplemented with 1% b-mercaptoethanol (BioRad) and boiled for 5 min. 15 μg of each lysate was run on a 4–15% gradient gel (Bio-Rad) and transferred to a PVDF membrane (Millipore) using the Trans-Blot Turbo Transfer System (Bio-Rad). PVDF membranes were blocked with 5% milk (Cell Signaling Technologies) in TBST and then probed with primary antibodies overnight at 4°C. Primary antibodies against IRF4, SMARCD1, GAPDH (all Santa Cruz), ARID1A, SMARCA2, SMARCA4, SMARCB1, AIOLOS, IKAROS, c-MYC, p27, BIM (all from Cell Signaling Technologies) are listed in the Key Resources Table and were all used at a concentration of 1:1000. After incubation with listed HRP-conjugated secondary antibodies (1:2000) for 1 hour, blots were imaged with a ChemiDoc Imaging System (Bio-Rad) and analyzed using Image Lab Touch Software (v2.3.0.07) (Bio-Rad).

Viability Assays

MM, DLBCL, and solid tumor cell lines were seeded at 2000–10000 cells/well in triplicate in 96-well plates. Three independent biological replicates were run for each cell line. FHD-286, lenalidomide, trametinib, everolimus, (all from SelleckChem), AU-15330, and JHX-VI-178 (all from MedChemExpress) dissolved in DMSO were diluted in equal volumes at the indicated concentrations. Cells were cultured with drugs for 3–7 days. Drugs were replenished after 3 days, and metabolic activity was measured at the end of the incubation period using CellCountingKit-8 (SelleckChem) according to the manual. Absorbance was measured at 450nm using a 96-well Tecan Infinite 200 Pro plate reader. Synergy scores for drug combination experiments were determined with synergyfinder ⁶⁷.

Colony Formation Assay

For colony formation assays, 2x10³ MM cells either treated or untreated with FHD-286 at 100 nM were plated into 6-wells in duplicate in 1.1 ml methylcellulose-based medium (MethoCult Classic, StemCell Technologies). Cells were incubated for 14 days (37°C, 5% CO₂) before the number of colonies (>40 cells) per well was scored using a Zeiss AXIO Vert.A1 inverted microscope. Three independent biological replicates were run for each cell line.

Flow cytometry and cell sorting of murine cells

Spleen and PP cell suspensions were generated by mashing the organs through 70-μm cell strainers and BM cell suspensions were generated by flushing femurs with RPMI containing 2% (v/v) FBS, antibiotics (penicillin (50 IU/ml) and streptomycin (50 mg/ml); Cellgro) and 10 mM HEPES, pH 7.2 (Cellgro). Small intestinal (SI) lamina propria (LP) cells were isolated as previously described ⁶⁸. Briefly, the SI was opened longitudinally and vortexed in HBSS containing 5% FBS and 10 mM HEPES, ph 7.2. Epithelial cells were removed by gently agitating the SI tissue in RPMI containing 5% FBS, 10 mM Hepes, pH 7.2 and 10 mM EDTA for 30 min at 37 °C. The SI was then washed with RPMI containing 10% FBS, 10 mM Hepes, pH 7.2 and antibiotics. SI tissue was chopped with scissors and digested at 37 °C for 30 min in RPMI medium containing 5% FBS, 10 mM Hepes, pH 7.2, 1 mg/ml Collagenase IV (Worthington Biochemical) and 25 mg/ml DNase I (Sigma). Digested tissue was passed through a 70-μm cell strainer and resuspended in 40% Percoll (Cytiva)-RPMI, layered with 80% Percoll-RPMI and subsequently centrifuged for 20 min at 650g. The cell ring at the interface was collected and washed twice with RPMI containing 10% FBS, 10 mM Hepes, pH 7.2 and antibiotics. Cells were stained with the following antibodies and dyes: Fixable Viability Dye eFluor 780 (ebiosciences), FITC or BV786–conjugated anti-B220 (RA3-6B2; Biolegend), BUV395 or PerCP Cy5.5-conjugated anti-CD4 (RM4-5; BD), PerCP Cy5.5-conjugated anti-TCRb (H57-597; Biolegend), Alexa Fluor 647–conjugated GL7 (GL-7; BioLegend), FITC or BV650-conjugated anti-IgD (11-26c.2a; BioLegend), PerCP Cy5.5 or PE-Cy7– conjugated anti-CD38 (90; BioLegend), PE-Cy7 or BV421-conjugated anti-Fas (Jo2; BD), BV421-conjugated anti-CD138 (281–2; Biolegend), PE-conjugated anti-TACI (8F10; Biolegend) or FITC-conjugated anti-IgA (polyclonal: Southern). To stain for intracellular IgA, cells were first stained for surface markers and then fixed and permeabilized using the BD cytofix/cytoperm kit per manufacturer’s instructions. Flow cytometry was performed on a Cytoflex LX (Beckman Coulter). Cells were sorted on a Sony MA900 sorter.

FACS analysis and cell sorting of human cells

For apoptosis evaluation via Annexin V staining, MM cell lines were seeded in a 96-well plate at 50,000–75,000 cells per 200 μl and treated for 24, 48, or 72 hours with DMSO, AU-15330, FHD-286, lenalidomide, and/or trametinib at the indicated concentrations. At the end of the incubation period cells were 1x washed with cold PBS and stained with APC-Annexin V (BioLegend) at a dilution of 1:40 in Annexin V Binding Buffer (BioLegend). Cells were incubated for 15 min at RT in the dark, resuspended in Annexin V Binding Buffer and subsequently acquired on a CytoFLEX (Beckman Coulter). Data was processed with FlowJo version 10.8.1.

For cell cycle analysis, MM cell lines were seeded at 200,000–500,000 cells per mL and treated for 24 or 48 hr with DMSO, 100 nM or 1000 nM AU-15330, 100 nM or 1000 nM FHD-286, or 10 μM lenalidomide. At the end of the incubation period, cells were washed with PBS once and resuspended in in 0.5 mL PBS. Cells were fixed by adding 4.5 mL ice-cold 70% EtOH dropwise and stored at −20°C overnight. For staining, cells were washed in cold PBS and stained with the FxCycle PI/Rnase Staining Solution (Invitrogen) following the manufacturer’s protocol. Stained cells were analyzed on a CytoFLEX (Beckman Coulter) and data was analyzed using FlowJo version 10.8.1.

For the intracellular staining of IRF4 in primary patient samples, we used the BD Transcription Factor Buffer set (BD Biosciences) according to the manufacturer’s instructions. Samples were first stained for surface expression of CD38-FITC, CD45-PerCP/Cy5.5, and CD138-BV421 to define MM cells, then fixed and permeabilized for staining with anti-IRF4-AF647 antibody (Biolegend). For the evaluation of apoptotic MM cells in patient samples we performed the identical surface stain followed by staining with AnnexinV-APC as described above. Patient cells were acquired on a CytoFLEX LX (Beckman Coulter) and either apoptosis or IRF4 expression in CD45-/CD38++/CD138+ MM cells was analyzed with FlowJo version 10.8.1. For ATAC-seq in primary MM cells, we stained whole bone marrow mononuclear cells of a MM patient with anti CD38-FITC (BD), CD45-PerCP/Cy5.5 (Biolegend), CD138-BV421 (BD) antibodies and the viability Dye eFluor 780 (eBiosciences). Subsequently, 60,000 single live CD45-/CD38++/CD138+ MM cells were sorted using a Sony MA900 sorter and immediately processed for ATAC-seq as described below. Post-sort analysis of all sorted samples indicated >95% enrichment for MM cells. All primary patient samples were de-identified before analysis.

Immunofluorescence

Pieces of small intestine were fixed in 4% paraformaldehyde (PFA; Electron Microscopy Sciences) and 10% sucrose in PBS for 1 hour at 4 °C then moved to 30% sucrose in PBS overnight. Tissues were flash frozen in OCT compound (Sakura) the following day. 20-μM sections were cut on a Leica CM1950 cryostat and were adhered to Super Frost Plus slides (Fisher Scientific). Sections were fixed in ice cold acetone for 10 minutes and then air dried for 1 h. Sections were blocked for 1 h in PBS containing 0.3% Triton X-100, 1% BSA, 2% normal rat serum, 2% goat serum and 2% anti-CD16/32 (BioXcell). Sections were stained with Alexa Fluor 647–conjugated anti-IgA (Southern; 1:1000), biotinylated anti-Epcam (G8.8; Biolegend; 1:200). Secondary antibodies were Alexa Flour 488-conjugated streptavidin (Invitrogen) and were incubated with slides for 3h at 27 °C. Cell nuclei were stained with DAPI (Invitrogen). Stained slides were mounted with Fluoromount G (Southern Biotech) and sealed with a glass coverslip. Tile scans of small intestine were acquired using a Stellaris 5 confocal microscope (Leica Microsystems, Exton, PA) with a 63X oil immersion objective NA1.4 at a voxel density of 2056 x 2056.

ATAC-seq

Analysis of open chromatin sites was performed using established protocols ⁶⁹. Briefly, 50 000 SKMM1 or NCIH929 cells were harvested at the indicated time points, lysed in 50 μl ATAC-seq Lysis Buffer (10 mM Tris-HCl pH7.5, 10 mM NaCl, 3 mM MgCl₂, 0.1% NP-40, 0.1% Tween-20, 0.01% Digitonin) for 3 min on ice and washed with 1 ml ATAC-seq Wash Buffer (10 mM Tris-HCl pH7.5, 10 mM NaCl, 3 mM MgCl₂, 0.1% Tween-20). Isolated nuclei were pelleted (10 min, 600g, 4°C), resuspended in 50 μl Transposition Mix and incubated at 37°C for 30 min in a thermomixer set to 1000 RPM. At the end of the incubation period the transposition reaction was stopped by adding 5 volumes of DNA Binding Buffer (ZymoResearch) and DNA was cleaned up using the DNA Clean and Concentrator-5 Kit (ZymoResearch) according to the manufacturer’s instructions.

For ATAC-seq in primary MM cells, total BM MNCs were treated with DMSO or 1000 nM AU-15330 for 2.5 h and FACS sorted as described above. Cells were then immediately processed for nuclei isolation in 50 μl ATAC-seq Lysis Buffer, similar to the cell lines. Purified DNA was stored at −20°C for further processing. Amplification of tagmented DNA samples was performed for 7 PCR cycles with a previously described PCR primer set ⁷⁰. Resultant libraries were cleaned up using SPRIselect beads (Beckman), quantified with the Bioanalyzer DNA High-Sensitivity Kit (Agilent), and sequenced on a NextSeq 2000 using a P2 flow cell (Illumina, 100 cycles, 2x 50bp paired-end). Data processing and peak calling were performed as previously described ⁷¹. Track visualizations were performed with the Integrative Genomics Viewer (https://igv.org/). Significantly deregulated peaks were determined with DeSeq2 and then used to run downstream analyses. In brief, we used GREAT ⁷² to associate the ATAC-seq peaks of interest to neighboring genes (two closest genes within ±100 kb of the ATAC-seq peak). Gene ontology (GO) analyses of the peak-associated genes was performed with the webtool version of GREAT (https://great.stanford.edu/great/public/html/). Locus overlap analysis (LOLA) ³² was used to compare the generated peak locations in this study with numerous public CHIP-seq datasets. Identified overlaps were ranked as outlined in the results section and plotted as either scatter plot or heatmaps using Graphpad Prism v9. In addition, LOLA was used to determine the distribution of our peak locations across genomic partitions and relative to transcription start sites, respectively. Peak occupancies were derived from merged bigwig files (using bigwig average in galaxy) and displayed as heatmaps using deeptools capturematrix and plotheatmap functions in galaxy (www.usegalaxy.org).

CUT&RUN

CUT&RUN analysis in SKMM1 and NCIH929 cells was performed using a commercial kit (CUTANA CUT&RUN kit, Epicypher) according to the instructions of the supplier. In short, 1.5x10⁶ cells were washed in 1 ml cold PBS, resuspended in 150 μl cold Nuclear Extraction Buffer (20 mM HEPES pH7.9, 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 1 mM MnCl₂, 0.5 mM Spermidine, 1x protease inhibitor (Roche)) and incubated for 10 min on ice. Nuclei were harvested via centrifugation (5 min, 600g, 4°C), resuspended in 110 μl cold Nuclear Extraction Buffer per sample and mixed with 10 μl activated Concanavalin A (ConA) conjugated paramagnetic beads. Nuclei were incubated with ConA beads for 10 min at RT and the mix was subsequently purified on a magnet. Beads were resuspended in 50 μl Antibody Buffer and incubated with approximately 1 μg primary antibodies (anti-IRF4, ebiosciences; anti-ARID1A, anti-SMARCA4, anti-SMARCB1, anti-MYC, anti-H3K27ac, all Cell Signaling Technologies; anti-H3K4me3 and isotype control, Epicypher) on a nutator overnight at 4°C. Beads were then washed with cold Digitonin Buffer, incubated with 1x pAG-MNase for 10 min at RT, washed and resuspended in 50 μl cold Digitonin Buffer and incubated on a nutator for 2 hours at 4°C in the presence of 2 mM CaCl₂. The reaction was stopped by adding 33 μl StopBuffer followed by a 10 min incubation at 37°C to release cleaved chromatin. DNA was cleaned up and eluted in 12 μl Elution Buffer with supplied materials before quantification via the Qubit ds DNA High-Sensitivity assay (ThermoFisher). For CUT&RUN experiments in primary mouse cells, we harvested spleens from up to 10 wt mice per biological replicate and performed negative enrichment using the Easysep system (StemCellTechnologies). We therefore resuspended 10⁸ cells/ml in EasySep Buffer and added 50 μl/ml rat serum (StemCellTechnologies) as well as biotin anti-mouse CD4 (Biolegend), biotin anti-mouse CD8 (Biolegend), biotin anti-mouse Ter119 (Biolegend), and biotin anti-mouse IgD (ebioscience) antibodies at a concentration of 1:200. Cells were incubated for 25 min on ice before 75 μl/ml rapidshperes (StemCellTechnologies) were added. After 2.5 min incubation at RT, EasySepBuffer was added to a final volume of 8 ml and samples were incubated for 2.5 min in an EasySep magnet (StemCellTechnologies) before the negative fraction containing GCB and PC subsets was transferred into a fresh tube. This fraction was used for subsequent FACS sorting of single live B220⁺/CD38^low/GL7⁺ GC B cells and B220⁻/TACI⁺/CD138⁺ PCs. In addition, single live CD4⁻/CD19⁺/IgD⁺/CD38⁺ naïve B cells were FACS sorted from spleens undergoing no prior selection. Isolated single cell suspensions (0.5–1.0x10⁶ sorted cells per sample) were used for CUT&RUN experiments as described above. Library preparation was performed with the CUTANA CUT&RUN Library Prep Kit (Epicypher) according to the manual, except for the following minor modifications: Instead of 5 ng, we used approximately 10 ng DNA input per sample and performed two additional DNA cleanup steps after PCR amplification with SPRIselect beads (0.9x reaction volume). Libraries were quantified using the Bioanalyzer DNA High-Sensitivity Kit (Agilent) and sequenced on a NextSeq2000 (Illumina, paired-end 2x 50 bp). Resultant fastq files were mapped to the hg19 or mm10 reference genomes using Bowtie. Reads were aligned to either mm10 (mouse experiments) or hg19 (human experiments) via Bowtie2 using default parameters. Next, we divided the genome into bins of 20 (mouse experiments) or 100 bp (human experiments) and performed peak calling as previously described ⁷¹. Overlaps between different CUT&RUN datasets were determined with the bedtools Multiple Intersect tool in galaxy. Motif enrichment analysis was performed with RSAT ⁷³. For this purpose, peak locations were transformed into sequences using the MEME bed2fasta tool and analyzed with the RSAT-motif webtool (http://rsat.sb-roscoff.fr/peak-motifs_form.cgi). Peak sequences (peak center ±80bp unless otherwise stated) were queried against human Jaspar, Hocomoco, and Homer or mouse Jaspar and Hocomoco transcription factor databases to define the top enriched transcription factors. Track visualizations, heatmaps, gene-associated (GREAT), and locus associated (LOLA) downstream analyses were performed as described above for ATAC-seq.

RNA-seq and Signature Enrichment

SKMM1 and NCI-H929 MM cells were treated with 1000 nM AU-15330 for 6 or 24 hours and harvested for RNA-seq. RNA was extracted using the AllPrep kit (QIAGEN) and RNA libraries were prepared using the TruSeq V3 chemistry (Illumina) according to the manufacturer’s protocol. Sequencing of libraries was done on a NovaSeq S1 with a read length of 2x100 bp. Alignment to the human genome (hg19) was done using STAR-aligner ⁷⁴. Normalized reads were Log2 transformed to calculate Digital Gene Expression values, as previously described ⁷⁵. Changes in gene expression between AU-15330 treated and DMSO control cells were determined using DeSeq2. Significantly downregulated genes with an average log2 fold change of less than −2.0 in both cell lines at 24 hours with average DGE expression values of 6.0 or greater from RNA-seq in CoMMpass were included in the AU-15330 down signature.

Analysis of publicly available datasets

Publicly available data was downloaded from the sources specified in the Key Resource Table. Data from the MMRF CoMMpass dataset and TCGA was downloaded and processed using the GDC standard pipelines (https://docs.gdc.cancer.gov/Data/Bioinformatics_Pipelines/DNA_Seq_Variant_Calling_Pipeline/). Then, processed WES result was further annotated as previously described ⁷⁵ to call gene mutations. For mutation analyses, for a given gene, the somatic mutation frequency in each TCGA and the MMRF CoMMpass cohort (the unique mutated cases/ total unique cases in the cohort) was calculated using GDC Mutation Frequency Tool provided on GDC portal (https://portal.gdc.cancer.gov/). RNA-Seq data was further processed and normalized as previously described ⁷⁵. Public CHIP-seq data (GSE197492) was processed as previously described⁷⁰.Normalized microarray data matrix was downloaded from GEO (Accession GSE2658), which was originally processed with the Affymetrix Microarray Suite GCOS1.1 software as reported³⁷. Affymetrix signals were then transformed by the log-base 2 for each sample. The standard two-tail T-test was used for differential gene expression analysis.

High-throughput drug combination screening

Drug combination screening was performed as previously described ⁵⁶. Briefly, 10 nL of compounds were acoustically dispensed into 1536-well white polystyrene tissue culture-treated plates with an Echo 550 acoustic liquid handler (Labcyte). Cells were then added to compound-containing plates at a density of 500-cells/well in 5 μL of medium. A 5-point custom concentration range, with constant 1:4 dilution, was used for all the MIPE 6.0 drugs ⁷⁶ in the primary 6×6 matrix screening against FHD-286 (1:3 dilution).

Plates were incubated for 48 hours at standard incubator conditions covered by a stainless steel gasketed lid to prevent evaporation. 48h post compound addition, 3 μL of Cell Titer Glo (Promega) was added to each well, and plates were incubated at room temperature for 15 minutes with the stainless-steel lid in place. Luminescence readings were taken using a Viewlux imager (PerkinElmer) with a 2 second exposure time per plate.

Xenograft experiments

All mouse experiments were approved by the National Cancer Institute Animal Care and Use Committee (NCI-ACUC) and were performed in accordance with NCI-ACUC guidelines and under approved protocols. Female NSG (non-obese diabetic/severe combined immunodeficient/common gamma chain deficient) mice were obtained from NCI Fredrick Biological Testing Branch and used for the xenograft experiments between 6–8 weeks of age. Mice were housed in specific pathogen-free facility in ventilated microisolator cages with 12 hour light and 12 hour dark cycles at 72F and 40–60% relative humidity. Approved protocols allowed tumor growth below 20 mm in any dimension; no animals had tumors which exceeded these limits. MM.1S or SKMM1 multiple myeloma tumors were established by subcutaneous injection of 10⁶ cells in a 1:1 Matrigel/PBS suspension. Treatments were initiated when tumor volume reached a mean of 200mm³. Trametinib and FHD-286 (SelleckChem) were prepared in sterile-filtered 20% (2-Hydroxypropyl)-β-cyclodextrin (Sigma) in PBS and administered p.o. once per day (1mg/kg/day for trametinib, 1.5mg/kg/day or 3 mg/kg/day for FHD-286). For the combination arm, drugs were given at the same concentration and schedule as single agents. Each treatment group contained 5 mice. Tumor growth was monitored every other day by measuring tumor size in two orthogonal dimensions, and tumor volume was calculated by the following equation: tumor volume = (length × widtĥ2)/2.

Artwork

Biorender.com was used for the generation of the graphical abstract.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed with Prism (GraphPad). P-values ≤ 0.05 were generally considered significant, with asterisks indicating significance: ***, P ≤ 0.001; **, P ≤ 0.01; *, P ≤ 0.05. n.s. indicates not significant. Where applicable, FDR-corrected q-values ≤ 0.05 were considered as statistically significant. Preclinical data were analyzed by one-way ANOVA or a two-tailed Student t test. In the case of patient subgroup comparisons, a Kruskal-Wallis test followed by a Mann-Whitney U test for individual group comparisons was applied. P-values for the association between the AU-15330 down signature and survival were obtained from a two-sided likelihood-ratio test based on a Cox proportional hazard model with the AU-15330 down signature treated as a continuous variable. Bar graphs show the mean +/− SD of three independent experiments performed in triplicates. Patient subgroup data is displayed as individual AU-down signature expression scores and lines indicate the mean. Box and whisker plots display the median with whiskers incorporating 10–90% of all data (outliers are displayed as dots) unless otherwise specified.

CRISPR analysis

The DESeq2 algorithm ⁷⁷ was used to estimate the log-fold change of the read count between Day 21 and Day 0 samples, or GFP-low and GFP-high samples, of the sgRNA guides in each cell line. Of the 77,441 guides targeting genes, 9,919 (13%) were removed for having poor performance across a large number of essential gene experiments ^40,75. For each gene, the log-ratios of the remaining guides associated with that gene were averaged to estimate a gene-level, log-fold change. For each cell, these gene-level, log-fold changes were normalized by subtracting the mode of their distribution (estimated with the R-function “density”) and then divided by the root-mean-square deviation (RMSD) from that mode. Further analysis was performed using Microsoft Excel v16.63.1 as previously described ¹³.

PLA image analysis

PLA spots were counted as follows: Original images were analyzed in ImageJ (fiji.sc) using custom macro available at https://github.com/janwisn/EIB_PLA_Analysis. In short, nuclei and cells were identified in the corresponding color channels. Then PLA spots were detected, and their number counted separately over each nucleus and cell. Results were compiled into a spreadsheet for further analysis. Color-coded maps of selected areas and detected points were combined with the image to facilitate quick assessment of analysis accuracy. PLA Score was determined by normalizing the number of PLA spots counted in each sample to the average number of PLA spots counted in the control sample, which was set to 100. Box plots represent median and 25–75% of data, whiskers incorporate 5–95% of data, and outliers are displayed as dots. Statistical comparisons were made by one-way ANOVA with Dunnett’s post test using Prism.

Drug-Target Set Enrichment Analysis (DTSEA)

To enable the unbiased identification of over-represented drug targets that synergized with FHD-286 in MM cell lines, we used the Excess over the Highest Single Agent (ExcessHSA) metric to quantitatively assess synergism and antagonism throughout the FHD-286 vs MIPE 6.0 combination screenings. We then ranked the entire MIPE 6.0 drug-universe based on the average ExcessHSA score in SKMM1 and XG7 cells. We used this ranked list to run a pre-ranked Drug-Target Set Enrichment Analysis (DTSEA), against a custom collection of drug-target sets representing any MIPE 6.0 drug-target that is covered by at least 3 small-molecule drugs (n=278). The pre-ranked enrichment analysis was performed using the GSEA software (v4.0.3) ⁷⁸ with a weighted enrichment statistic.

MS Database search and post-processing analysis

Both injections for each sample were batched together as fractions and all MS files were searched with Proteome Discoverer 2.4 using the Sequest node. Data was searched against the Uniprot Human database from Feb 2020 using a full tryptic digest, 2 max missed cleavages, minimum peptide length of 6 amino acids and maximum peptide length of 40 amino acids, an MS1 mass tolerance of 10 ppm, MS2 mass tolerance of 0.02 Da, variable oxidation on methionine (+15.995 Da) and fixed modifications of carbamidomethyl on cysteine (+57.021), TMTpro (+304.207) on lysine and peptide N-terminus Percolator was used for FDR correction of 2-way ANOVA based P-values obtained in Proteome Discoverer 2.4 and TMTpro reporter ions were quantified using the Reporter Ion Quantifier node and normalized on total peptide intensity of each channel. TMTpro channel assignment for conditions can be found in Table S6.

Highlights.

• ARID1A promotes IRF4 expression through the cBAF complex

• ARID1A is required for plasma cell development and myeloma survival

• SMARCA2/4 inhibition disrupts oncogenic transcription in multiple myeloma

• SMARCA2/4 inhibition overcomes IMiD resistance and synergizes with MEK inhibitors