Main

Tumor masses are mixtures of cancerous and normal cells that collectively form the tumor microenvironment (TME)1. Within the TME, various cell populations communicate through cytokines, chemokines and growth factors, further recruiting additional infiltrating cells, leading to increased tumor heterogeneity2,3,4. Advances in single-cell RNA sequencing (scRNA-seq) technology allow for the characterization of individual components within the TME, offering significant opportunities to enhance our understanding of tumor biology and cancer management5,6. Indeed, in recent years, an increasing number of studies has focused on targeting components of the TME, particularly the immune microenvironment7,8.

The CRISPR screen is a powerful tool to identify vulnerabilities in cancer cells9,10. Although CRISPR screens have been extensively conducted in in vitro cell culture systems11,12,13, including the Cancer Dependency Map (DepMap) project14,15,16, the absence of the TME in these models has limited the ability to identify gene targets that modulate interactions between tumor and the TME. A few recent studies of in vivo CRISPR screens have identified targets that regulate immunotherapy resistance, highlighting the feasibility and significance of functional genomic screens in the context of physiologic TME12,17,18.

In this study, we set out to identify modulators of the tumor–TME interactions. By directly comparing parallel in vitro and in vivo CRISPR screens, we identified MEN1 as the top candidate gene that confers differential effects in cell culture versus xenograft tumors. We found that MEN1 knockout resulted in increased tumor growth in immunodeficient mice but decreased growth in immunocompetent mice, revealing both TME-dependent oncogenic and tumor-suppressive functions. Our findings underscore the effectiveness of in vivo functional genomic screens in identifying clinically relevant drug targets and provide a rationale for the therapeutic targeting of MEN1, either alone or in combination with immunotherapy, in multiple solid and hematologic cancer types.

Results

Parallel in vitro and in vivo CRISPR screens in lung cancer

The number of cells that can be injected in mice to form xenograft tumors is limited. Therefore, we designed a targeted single guide RNA (sgRNA) library to ensure sufficient coverage of each sgRNAs in an in vivo screen. This library, referred to as Epi-Drug19,20, comprises 12,472 sgRNAs targeting 317 epigenetic regulators and 657 DrugBank targets, with an average of 10 sgRNAs per gene (Extended Data Fig. 1a and Supplementary Table 1). A549 lung adenocarcinoma (LUAD) cells with stable Cas9 expression were transduced with the sgRNA library and cultured in Petri dishes for 3 days with puromycin selection (D0). Subsequently, the transduced cells were split into equal aliquots for either Petri dish culture or establishing xenograft tumors via subcutaneous injection into immunodeficient NOD scid gamma (NSG) mice (Fig. 1a). Two-dimensional (2D) cultured cells and xenograft tumors were collected at day 21 (D21) for DNA extraction and next-generation sequencing (NGS).

Fig. 1: Parallel in vivo and in vitro CRISPR screens in A549 and in vivo specific function of MEN1 in lung cancer.
figure 1

a, Schematic representation of the CRISPR screen experiment design. A549 cells with stable Cas9 expression were transduced with Epi-Drug sgRNA library and selected with puromycin for 3 days (D0). Parallel in vitro and in vivo screens were performed for 3 weeks (D21). Samples were collected for PCR amplification and NGS. b, Dropout (genes with sgRNA reduced in D21; blue dots) and enriched (genes with sgRNA increased in D21; red dots) genes in LUAD A549 xenograft compared with A549 2D cultured cells. The P values of positive and negative selections and log2(fold change) were defined and calculated using MAGeCK. MEN1 is the top ranked enriched gene. c, Xenograft tumor growth curve in immunodeficient mice inoculated with control (sgCtrl) or MEN1 knockout (sgMEN1-1, sgMEN1-2) A549 cells. Each data point represents mean ± s.e.m. tumor volumes (n = 5 in sgCtrl, sgMEN1-1 and sgMEN1-2 groups). Two-way ANOVA was used for the growth curves. *P < 0.05, **P < 0.01. d, Colony formation of A549 cells with (sgMEN1) and without (sgCtrl) knockout of MEN1. Cells were seeded in six-well plates in duplicate and allowed to grow for 8 days before staining with crystal violet. Scale bars, 7 mm. e,f, KEGG analysis of differential genes in MEN1 knockout versus control A549 cells in vitro (e) and in vivo (f). The x axis represents the number of genes. Wald tests defined in DEseq2 were used to calculate P values. g, RT–qPCR showing the expression of representative cytokine-related genes in MEN1 knockout A549 cells relative to control. Housekeeping gene TBP was used as a control. Mean ± s.e.m. of three biological replicates is shown (unpaired two-tailed Student’s t-test). h, KEGG analysis of differential genes in MEN1-low versus MEN1-high patients from the TCGA LUAD cohort. Wald tests defined in DEseq2 were used to calculate P values. i, Boxplot showing the abundance of representative cytokine genes in MEN1-high versus MEN1-low patient tumors in the TCGA LUAD cohort. Twenty patients with the highest and the lowest MEN1 expression were assigned to each group. Horizontal lines in the box represent the upper quartile, median and lower quartile from top to bottom. Vertical extending lines mark the 5th to 95th percentile (unpaired two-tailed Student’s t-test). *P < 0.05, **P < 0.01. IgA, immunoglobulin A; IL, interleukin; Padj, adjusted P value; Rep, replicate. a, Created with BioRender.com.

Source data

Full size image

We observed high overall sgRNA coverages, with 99.98% and 97.49% of the 12,472 sgRNAs detected in D21 cultured cells and xenograft tumors, respectively (Extended Data Fig. 1b). Xenograft tumors demonstrated greater variation in sgRNA representation compared with cultured cells (Extended Data Fig. 1c,d), consistent with a previous report21. In both the in vitro and in vivo screens, MAGeCK22 analysis revealed significantly higher dropout rates for the positive control genes (Mann–Whitney U-test; P = 2.10 × 10−21), compared with negative controls (Extended Data Fig. 1e,f), suggesting the high efficacy of the screens.

We then applied MAGeCK to identify genes with sgRNAs that were depleted or enriched in D21 compared with D0 (referred to as dropout and enriched genes hereafter). This led to the identification of 47 and 72 dropout genes in the in vitro and in vivo screens, respectively (Extended Data Fig. 1g,h and Supplementary Table 2). Twenty-two dropout genes were found to be common to both screens (Supplementary Table 2), including DepMap core essential genes POLE and POLE2 (ref. 23). Although no significantly enriched genes were found in vitro, 40 were found in the in vivo screens (Extended Data Fig. 1g,h and Supplementary Table 2). To further identify genes that confer differential essentialities in vitro and in vivo, we used MAGeCK to compare the D21 cell culture and xenograft tumor screens directly. This analysis identified 12 in vivo specific dropout genes and 13 enriched genes (Fig. 1b and Supplementary Table 2). Among these genes, the top hit, MEN1, which encodes menin, is of particular interest. Genetic loss of MEN1 drives tumorigenesis in multiple solid cancer types. However, the dependency score of MEN1 from DepMap showed that it is not an enriched gene in in vitro screens for 953 of 954 cell lines from solid tumor types (Extended Data Fig. 1i). The in vivo specific enrichment in A549 suggests that MEN1 may function through regulating tumor–TME interactions.

MEN1 regulates cytokine genes both in vitro and in vivo

To validate our findings, we designed two sgRNAs for MEN1 deletion (Extended Data Fig. 2a). Tumor growth significantly increased with both MEN1 knockout sgRNAs (Fig. 1c). By contrast, no obvious effect on colony formation or cell proliferation was observed in 2D cultures upon genetic ablation of MEN1 (Fig. 1d and Extended Data Fig. 2b). These data further confirm the in vivo restricted function of MEN1 in this context.

To investigate the mechanisms underlying the function of MEN1, we conducted RNA sequencing (RNA-seq) in 2D cultured A549 cells with and without MEN1 knockout. Differential gene analysis using DESeq2 (ref. 24) identified 357 upregulated and 252 downregulated genes on MEN1 knockout (Padj ≤ 0.01 and abs(log2(fold change)) > 1; Extended Data Fig. 2c). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed ‘cytokine–cytokine receptor interaction’ as the most significantly enriched pathway in the upregulated genes (Fig. 1e, Extended Data Fig. 2d and Supplementary Table 3). Although a smaller number of differential genes were identified in xenograft tumors (Extended Data Fig. 2e), likely because of the heterogeneity of tumor samples, ‘cytokine–cytokine receptor interaction’ remained the top enriched pathway in the upregulated genes (Fig. 1f and Supplementary Table 3). Quantitative real-time polymerase chain reaction with reverse transcription (RT–qPCR) analysis of representative genes in the ‘cytokine–cytokine receptor interaction’ signature, including CXCL1, IL33, CXCL8 and IL1B, showed marked upregulation upon MEN1 depletion in A549 and NCI-H1792, an additional human LUAD cell line (Fig. 1g and Extended Data Fig. 2f,g). Notably, through reanalysis of five RNA-seq datasets from recent publications on MEN1 functions25,26,27,28,29, we consistently observed cytokine-related signatures as top enriched terms in response to MEN1 perturbation in four of the datasets, utilizing both cancer cell line and patient-derived xenograft (PDX) models (Extended Data Fig. 3a–d).

We extended the gene expression analysis to patient tumors in The Cancer Genome Atlas (TCGA) LUAD RNA-seq dataset, contrasting the top and bottom 10% of patients based on MEN1 mRNA abundance (Extended Data Fig. 3e). In concordance with our observations in the A549 cell line, KEGG analysis identified ‘cytokine–cytokine receptor interaction’ as the top enriched term in the upregulated genes in MEN1-low patients (Fig. 1h,i, Extended Data Fig. 3f,g and Supplementary Table 4). These results suggest that MEN1 regulates the expression of cytokine-related genes in cell lines, PDX models and patient tumors.

MEN1 loss leads to MLL1 chromatin redistribution

Given that menin is a scaffolding protein involved in epigenetic regulation30, we conducted a genome-wide cleavage under targets and release using nuclease (CUT&RUN)31 assay to determine menin chromatin binding in A549 cells. The efficiency and specificity of the menin antibody was confirmed by immunoprecipitation and western blot analysis, as well as by chromatin immunoprecipitation–quantitative PCR (ChIP–qPCR) analysis of representative menin binding regions (Extended Data Fig. 4a,b, left). A total of 26,507 menin chromatin binding sites were identified using MACS32 (Extended Data Fig. 4c,d and Supplementary Table 5). Binding and Expression Target Analysis (BETA)33 revealed that although genes whose expression decreased upon menin depletion were significantly associated with menin binding, no enrichment was observed for upregulated genes, including the cytokine-related genes (Fig. 2a). Similar trends were obtained when applying BETA to differentially expressed genes in MEN1-low compared with MEN1-high subjects from the TCGA LUAD cohort (Extended Data Fig. 4e). These data suggest that the regulation of cytokine-related genes is not directly controlled by cis-regulation of menin.

Fig. 2: MEN1 regulates MLL1 binding at repetitive genomic regions and transcription of dsRNA.
figure 2

a, Correlation between menin binding sites and target gene expression as evaluated by software BETA. b, RT–qPCR showing the abundance of representative cytokine-related genes with and without siRNA silencing of MLL1 in A549 cells. Housekeeping gene TBP was used as a control. Mean ± s.e.m. of two biological replicates is shown (unpaired two-tailed Student’s t-test). *P < 0.05, **P < 0.01, ***P < 0.001. c, RT–qPCR performed in A549 cells with and without MEN1 deletion coupled with siRNA silencing of MLL1. Mean ± s.e.m. of four biological replicates is shown (unpaired two-tailed Student’s t-test). **P < 0.01, ***P < 0.001, ****P < 0.0001. d, Pileup plots showing H3K4me3 ChIP-seq and MLL1 CUT&RUN signal at 1,857 increased peak regions called by MACS2 bdgdiff. e, Heatmap showing repeat loci with differential expression in MEN1 knockout and control A549 cells in 2D culture. f, Donut plot showing the categories of upregulated repeats in MEN1 knockout A549 cells. g, Number of upregulated repeats within a given distance of the 1,857 peaks with increased H3K4me3 and MLL binding. Up-repeats are repeats that are significantly upregulated upon MEN1 knockout. Background are randomly selected repeat regions that did not show differential expression upon MEN1 knockout. The P value was calculated by one-sided paired t-test. h, Northern dot blot showing dsRNA staining in control and MEN1 knockout A549 xenograft tumors. The upper panel demonstrates dsRNA staining using the J2 antibody, and the lower panel illustrates staining of total RNA, serving as a loading control. The experiment was conducted using four biological replicates for each condition. i, Immunofluorescence imaging of control (sgCtrl) or MEN1 knockout (sgMEN1-1, sgMEN1-2) A549 cells or A549 cells treated with poly(I:C) for the detection of dsRNA (red). Red, dsRNA (J2 antibody); blue, DAPI. Scale bars, 20 μm. j, Immunofluorescence imaging of γ-H2AX of control or MEN1 knockout A549 cells or A549 cells. Green, γ-H2AX (γ-H2AX-antibody); blue, DAPI. Scale bars, 20 μm. ERV, endogenous retroviruses; ERVL, endogenous retroviral-like elements; hAT, hobo, Ac and Tam3; LTR, long terminal repeat; MaLR, mammalian apparent LTR-retrotransposons; TcMAR, Tc1/mariner .

Source data

Full size image

Menin is known to interact with MLL1 to regulate target gene expression. Silencing of MLL1 expression with two separate small interfering RNAs in A549 wild-type cells resulted in significant reductions in CXCL1, IL33, CXCL8 and IL1B (Fig. 2b). Unlike MLL1, silencing of other methyltransferases or subunits within the MLL or DOT1L1 complexes29,34, including MLL2, MLL3/4, UTX and DOT1L, did not consistently activate the cytokine genes tested (Extended Data Fig. 4f–i). These data collectively suggest that menin-mediated regulation of these cytokine genes predominantly relies on MLL1. Indeed, induction of these cytokines in MEN1 knockout cells was completely attenuated by silencing of MLL1 (Fig. 2c).

We next expanded the CUT&RUN analysis to MLL1 in A549 cells with and without knockout of MEN1 (Extended Data Fig. 4a,b, right). We identified more than 10,000 MLL1 chromatin binding sites in the control condition, with 68% overlapping with those of menin (Extended Data Fig. 4d (right), Supplementary Fig. 1a and Supplementary Table 5). Because MLL1 is a H3K4-specific methyltransferase, we extended our analysis to H3K4me3 histone modification (Supplementary Fig. 1b). H3K4me3 ChIP-seq in A549 cells identified approximately 30,000 H3K4me3 peak regions, and 55.6% of MLL1 binding overlap with these regions (Supplementary Fig. 1c,d). Differential binding analysis using MACS2 bdgdiff identified 1,857 regions with increased MLL1 binding and H3K4me3 signal, and 1,017 regions with decreased signal, in MEN1 knockout cells (Fig. 2d and Supplementary Fig. 1e–g).

BETA indicated that the differential binding sites of H3K4me3 were not enriched near the upregulated genes upon MEN1 knockout (Supplementary Fig. 1h), suggesting an indirect mechanism of regulation. Although the majority of the MLL1 binding sites were shorter than 500 bp and enriched in the promoter regions, we observed regions in the genome with strong binding intensities that exceeded 2 kb (Supplementary Fig. 1g). Because MACS is designed to identify short peaks, we used CREAM software to identify broad peaks35. This led to the identification of 422 broad regions ranging from 2 kb to 8 Mbp, most of which harbor repetitive elements and showed increased MLL1 occupancy upon MEN1 knockout (Extended Data Fig. 5a and Supplementary Table 5).

To determine whether loss of MEN1 led to altered transcription of repetitive elements, we reanalyzed the A549 RNA-seq data by mapping the reads to repetitive genomic regions. Differential expression analysis identified 2,737 repeat loci with upregulated expression, and 865 repeat loci with downregulated expression (Fig. 2e). Most of the upregulated loci contained short interspersed nuclear elements(SINE), long interspersed nuclear elements (LINE) and simple repeats, with 740 of 2,737 located in the 442 CREAM peak regions (Fig. 2f and Extended Data Fig. 5b,c). Furthermore, the genomic regions with elevated MLL1 and H3K4me3 signals following MEN1 depletion were significantly enriched in proximity to upregulated repeats, whereas the decreased ones were not (Fig. 2g and Extended Data Fig. 5d). ChIP–qPCR analysis of representative repeat regions showed that, whereas the binding of menin, UTX and H3K27me3 diminished upon MEN1 depletion, there was a substantial increase in MLL1 binding and H3K4me3 signal at these repeat regions (Extended Data Fig. 5e–i).

Transcripts from repetitive genomic regions tend to form double-stranded RNA, such as inverted Alu repeat (Extended Data Fig. 5j). A significant induction of dsRNA staining was observed in MEN1 knockout compared with control A549 cells (Fig. 2h,i and Extended Data Fig. 5k, left). Transcription from repetitive genomic regions has been reported to induce DNA damage36. Indeed, significantly higher DNA damage, as indicated by stronger γ-H2AX staining, was observed upon MEN1 knockout (Fig. 2j and Extended Data Fig. 5k (right)). dsRNA expression is known to induce interferon signaling. MEN1-low subjects in the TCGA LUAD cohort demonstrated a significantly higher level of interferon signaling genes compared with MEN1-high patients (Extended Data Fig. 5l). Furthermore, MEN1 knockout significantly induced the expression of ISG15 and IRF7 (Extended Data Fig. 5m). Together, these data suggest that MEN1 knockout increases MLL1 occupancy at repetitive genomic regions, leading to the activation of dsRNA transcription.

MEN1 regulates tumor growth depending on mammalian apparent retroviral-like vesicles (MAVS) and cyclic GMP-AMP synthase (cGAS)–stimulator of interferon genes (STING)

dsRNA can induce a viral mimicry response via the RIG-1/MDA5-MAVS RNA sensing pathway or cGAS–cGAMP STING reverse-transcribed DNA sensing pathway37,38,39, with Phosphorylation of TBK1 (p-TBK1) serving as an indicator of MAVS and/or cGAS–STING pathway activation40. Silencing of MAVS or cGAS/STING significantly attenuated the induction of cytokines in MEN1 knockout cells (Fig. 3a and Extended Data Fig. 6a). In addition, MEN1 knockout resulted in induction of p-TBK1 in MEN1 knockout compared with control cells, whereas total TBK1 levels remained unchanged (Extended Data Fig. 6b).

Fig. 3: Antagonizing function of MEN1 and MLL1 in regulating cytokine gene signature and TME infiltration.
figure 3

a, RT–qPCR performed in A549 cells with and without MEN1 deletion coupled with deletion of MAVS and/or cGAS. Mean ± s.e.m. of two biological replicates is shown (unpaired two-tailed Student’s t-test). *P < 0.05, **P < 0.01. b, A549 xenograft tumor growth rate in immunodeficient mice with and without knockout of MEN1 or in combination with MAVS and/or cGAS knockout. Each data point represents mean ± s.e.m. tumor volumes (n = 10 for each arm). Two-way ANOVA was used for statistical analysis. ***P < 0.001, ****P < 0.0001. c, Dot plot showing enriched KEGG terms of mouse differential genes from MEN1 knockout A549 xenografts. d, Quantification of neutrophil infiltration in A549 xenograft tumors with and without MEN1 deletion. Mean ± s.e.m. of neutrophil percentage from 12 tumors is shown (unpaired two-tailed Student’s t-test). e, IHC staining showing percentage of neutrophil or CD8+ T cells in MEN1-high versus MEN1-low tumor samples from a LUAD microarray (TMA). Mean ± s.e.m. of 20 tumors with the highest and the lowest MEN1 TMA scores were assigned to each group. **P < 0.01, ***P < 0.001. f, Representative IHC staining images for MEN1, myeloperoxidase (neutrophil) and CD8 (CD8+ T cell). Scale bars, 200 μm. g, A549 xenograft tumor growth rate in immunodeficient mice with and without knockout of MEN1 or in combination with anti-Ly6G antibody injection. Each data point represents mean ± s.e.m. tumor volumes (n = 5 for each arm). Two-way ANOVA was used for statistical analysis. OR, odds ratio.

Source data

Full size image

In vivo analysis showed that MAVS or cGAS knockout in control cells had no noticeable effect on tumor growth (Extended Data Fig. 6c), but significantly reduced the induction of tumor growth in the MEN1 knockout condition (Fig. 3b). These data suggest that MEN1 regulates cytokine-related genes and tumor growth through a MAVS- and cGAS–STING-dependent viral mimicry mechanism.

MEN1 knockout induces neutrophil infiltration

To determine whether and how MEN1 knockout modulates TME cell infiltration and interaction with tumor cells, we identified reads that specifically mapped to the mouse genome in the A549 xenograft tumor RNA-seq data. Differential gene and KEGG pathway analysis revealed a strong enrichment of terms related to leukocyte function in genes upregulated in MEN1 knockout tumors (Fig. 3c and Extended Data Fig. 6d). Pathway analysis further demonstrated that the function of neutrophils, a subtype of leukocytes, is enriched in MEN1 knockout conditions (Extended Data Fig. 6e).

Immunohistochemistry (IHC) staining of anti-myeloperoxidase, a marker for neutrophil, revealed profound induction of neutrophil infiltration in MEN1 knockout tumors compared with controls (Fig. 3d and Extended Data Fig. 6f). In addition, reanalysis of a publicly available RNA-seq dataset of leukemia xenograft tumors41 revealed that menin inhibition induces cytokine gene expression and neutrophil infiltration in immunodeficient mice (Extended Data Fig. 6g,h).

To validate immune cell infiltrations in patient tumors, we performed IHC staining for neutrophil and CD8+ T cells in a tissue microarray of the LUAD cohort. Tumors with lower MEN1 expression indeed exhibited much stronger neutrophil and CD8+ T cell infiltration (Fig. 3e,f). Neutrophil neutralization with an antibody against myeloperoxidase significantly attenuated tumor growth induced by MEN1 knockout in A549 xenografts (Fig. 3g), indicating that the tumor-promoting effect of MEN1 knockout is dependent on neutrophils. Consistent with this, the percentage neutrophil infiltration in A549 xenograft tumors significantly increased with MEN1 knockout but was reduced to baseline levels with knockout of cGAS or/and MAVS (Extended Data Fig. 6i).

MEN1 regulates tumor growth in an immune cell-dependent manner

Pan-cancer analysis using the TCGA datasets revealed that the ‘cytokine–cytokine receptor interaction’ gene signature is significantly higher in MEN1-low compared with MEN1-high patients in 26 of the 32 cancer (sub-)types analyzed (Extended Data Fig. 7a). The number and types of cytokine genes activated varied across different cancer types (Extended Data Fig. 7b).

Similar to A549 lung cancer cells, knockout of MEN1 in colon cancer cell line HCT116 did not alter cell proliferation in vitro but significantly increased tumor growth in immunodeficient NSG mice (Fig. 4a,b and Extended Data Fig. 7c). Next, we extended our analysis to the CT26 mouse colon cancer cell line (Fig. 4c). Genetic ablation of Men1 did not alter CT26 cell proliferation in vitro but led to significantly faster tumor growth in immunodeficient NSG mice (Fig. 4d and Extended Data Fig. 7d,e). Furthermore, significantly higher neutrophil infiltration was observed in Men1 knockout tumors compared with controls (Extended Data Fig. 7f). By contrast, loss of Men1 resulted in significantly reduced tumor growth in immunocompetent mice (Fig. 4e). A similar effect was also observed in the syngeneic breast cancer model 4T1, with increased and decreased tumor growth upon Men1 knockout in immunodeficient and immunocompetent mice, respectively (Extended Data Fig. 8a,b). We further extended the Men1 knockout study to pancreatic cancer model HKP1 and prostate cancer model DKO, and consistently observed decreased tumor growth following Men1 depletion in immunocompetent mice (Extended Data Fig. 8c–e).

Fig. 4: Immunocompetence-dependent tumor-promoting and -inhibiting function of MEN1 in colon cancer.
figure 4

a, Cell proliferation rate of HCT116 in 2D cell culture with and without knockout of MEN1. NS, not significant. Each data point represents mean ± s.e.m. cell counts (n = 3 for each arm). Two-way ANOVA was used for statistical analysis. b, HCT116 xenograft tumor growth rate in immunodeficient mice with and without knockout of MEN1. Each data point represents mean ± s.e.m. tumor volumes (n = 5 for each arm). Two-way ANOVA was used for statistical analysis. *P < 0.05, **P < 0.01. c, Schematic view of murine CT26 engraftment experiment design. d, Tumor growth in immunodeficient NSG mice inoculated with control (sgEV) or Men1 knockout (sgMen1-1, sgMen1-2) CT26 cells. Each data point represents mean ± s.e.m. tumor volumes (n = 12 for each arm). **P < 0.01. e, Tumor growth in immunocompetent BALB/c mice inoculated with control (sgEV) or Men1 knockout (sgMen1-1, sgMen1-2) CT26 cells. Each data point represents mean ± s.e.m. tumor volumes (n = 5 for each arm). Two-way ANOVA was used for statistical analysis. ***P < 0.001. f, Heatmap showing expression value (z-score based on DESeq normalized RNA-seq counts) of differential genes from Men1 knockout versus control tumors in immunocompetent mice. Two control samples and four Men1 knockout tumors were subjected to RNA-seq analysis. g, GO analysis of differentially expressed genes in Men1 knockout versus control cells was performed and the top five terms are shown. The x axis represents the number of genes. Wald tests defined in DEseq2 were used to calculate P values. h,i, Quantification (h) and representative immunofluorescence images (i) of control (sgCtrl) or Men1 knockout (KO) CT26 cells for the detection of dsRNA. Red signal, dsRNA (J2 antibody); blue signal, DAPI. Scale bars, 20 μm. Each bar in the left-hand panel represents the mean of quantifications from 30 randomly picked fields (unpaired two-tailed Student’s t-test). **P < 0.01, ***P < 0.001. CTCF, corrected total cell fluorescence.

Source data

Full size image

To elucidate how the loss of Men1 restricts CT26 tumor growth in immunocompetent mice, we transcriptionally profiled Men1-proficient and -deficient tumors. Differential gene analysis identified 350 upregulated and 309 downregulated genes, respectively, in the knockout compared with control conditions (Fig. 4f). Pathway analysis revealed activation of antiviral immune response pathways in Men1-deficient tumors (Fig. 4g and Extended Data Fig. 8f,g), indicating increased tumor immunogenicity. Cytokine-related genes with high expression levels in the MEN1-low colon cancer patient tumors were highly upregulated in Men1-deficient tumors of CT26, HKP1 and DKO (Extended Data Fig. 8h,i). The strong enrichment of virus response-related signatures further supports the role of Men1 in suppressing viral mimicry, as observed in A549 cells. Indeed, Men1 knockout exhibited strong upregulation of dsRNA species compared with control CT26 cells (Fig. 4h,i).

Men1 knockout increases CD8+ T cell infiltration

To systematically investigate the TME populations affected by Men1 knockout, we conducted scRNA-seq and mass cytometry by time-of-flight (CyTOF) analysis of Men1-proficient and -deficient CT26 tumors grown in immunocompetent mice (Extended Data Fig. 9a), using standard data processing and quality control procedures42,43. Analysis of transcriptomic profiles of 7,595 cells identified 7 cell types with 16 distinct sub-clusters (Fig. 5a, Extended Data Fig. 9b,c and Supplementary Fig. 2). The proportion of macrophages and T cells significantly increased in Men1 knockout tumors (Fig. 5b and Extended Data Fig. 9d). These data suggest that Men1 knockout impacts immune cell infiltration.

Fig. 5: scRNA-seq, CyTOF and IHC analysis confirmed increased immune cell infiltration in Men1 knockout CT26 tumors.
figure 5

a, UMAP view of 7,595 single cells from scRNA-seq profiling of CT26 tumors with and without depletion of Men1, color coded by assigned clusters. Dotted circles mark cell types as determined by relevant marker genes. b, Percentages of T cells and macrophages in control (sgCtrl) and Men1 knockout (sgMen1-1, sgMen1-2) CT26 tumor samples. Mean ± s.e.m. of two to four biological replicates are shown (unpaired two-tailed Student’s t-test). *P < 0.05 c, UMAP view of 606,301 single cells from CyTOF profiling of 10 CT26 tumors with and without depletion of Men1. d, Percentage of CD45+ cells (left) and CD8+ T cells (right) of live cells captured by CyTOF. Mean ± s.e.m. of four biological replicates (sgCtrl) for control and six biological replicates for Men1 knockout (sgMen1) are shown (unpaired two-tailed Student’s t-test). e, Representative IHC images (left) and quantifications (right) showing the abundance of CD8+ T cells in control (sgCtrl) and Men1 knockout (sgMen1-1, sgMen1-2) CT26 tumors. Mean ± s.e.m. of quantifications from 10 tumor IHC sections are shown (unpaired two-tailed Student’s t-test). **P < 0.01, ***P < 0.001. Scale bars, 200 μm. DC, dendritic cells; NK cell, natural killer cell.

Source data

Full size image

To further characterize the enhanced immune response upon Men1 knockout in immunocompetent mice, we profiled 606,301 CD45+ immune cells from CT26 tumors with CyTOF using an antibody panel targeting 35 immune markers (Fig. 5c, Extended Data Fig. 9e and Supplementary Table 6). We observed an overall increase in the percentage of immune cells in Men1-deficient tumors (Fig. 5d and Extended Data Fig. 9f). In addition to the overall increase in CD45+ cells, the relative composition of the immune cells also changed, with a significant enrichment of CD8+ T cells, dendritic cells and a subset of macrophages (Fig. 5d and Extended Data Fig. 9g). To further validate the impact of MEN1 knockout on CD8+ T cells, we established A549 xenografts in humanized NSG mice, which were reconstituted with a human immune system. MEN1 knockout resulted in significantly reduced tumor growth (Extended Data Fig. 9h), in contrast to the increased tumor growth in nonhumanized NSG mice (Fig. 1c). IHC staining revealed significantly increased CD8+ T infiltration in MEN1 knockout tumors compared with control tumors (Extended Data Fig. 9i,j). Similarly, IHC analysis of CT26 tumors with Men1 depletion also showed increased CD8+ T cell infiltration (Fig. 5e), findings that are consistent with the scRNA-seq and CyTOF analysis.

Pharmacological inhibition of menin reduces tumor growth

Pharmacologic inhibition of the menin–MLL interaction is an effective treatment in preclinical models and clinical trials of MLL1-rearranged and NPM1-mutant leukemias44,45,46. To assess the efficacy of inhibiting menin in solid cancer types, we treated CT26 cancer cells with the clinical stage menin inhibitor ziftomenib (KO-539)46. In line with the Men1 knockout phenotype, we observed a dose-dependent induction of representative cytokine genes in CT26 following ziftomenib treatment (Fig. 6a), validating the on-target effect of ziftomenib in these cells. In addition, cGAS or MAVS deletion attenuated ziftomenib treatment-induced upregulation of cytokine genes (Extended Data Fig. 10a). Consistently, ziftomenib treatment also results in an increase in phosphorylated Tbk1 (Extended Data Fig. 10b). Enhanced cytokine expression was consistently detected in lung and pancreatic cancer explant PDX tissues and organoid models following treatment with ziftomenib (Extended Data Fig. 10c–f). To further confirm the effect of ziftomenib on chromatin dynamics, we conducted MLL1 and H3K4me3 ChIP assays in wild-type A549 cells treated with either dimethyl sulfoxide or ziftomenib. Ziftomenib treatment significantly reduced menin abundance (Extended Data Fig. 10g). ChIP–qPCR analysis validated significantly increased MLL1 and H3K4me3 signals at selected repeat regions tested, consistent with that observed with CRISPR mediated MEN1 depletion (Extended Data Fig. 10h).

Fig. 6: Pharmacological inhibition of MEN1 reduces tumor growth in a CD8+ T cell-dependent manner and demonstrates additive effect with anti-PD-1 treatment.
figure 6

a, RT–qPCR showing the relative expression of cytokine-related genes in CT26 cells treated with different dosages of menin inhibitor ziftomenib. Mean ± s.e.m. of two biological replicates are shown (unpaired two-tailed Student’s t-test). b, CT26 tumor growth with vehicle control or ziftomenib treatment at dosages of 100 and 150 mg kg−1. Each data point represents mean ± s.e.m. tumor volumes (n = 5 for each arm). Two-way ANOVA was used for statistical analysis. *P < 0.05, **P < 0.01, ***P < 0.001. c, Representative IHC images (left) and quantifications (right) showing the infiltration of CD8+ T cells in CT26 tumors with vehicle control and KO5-39 treatment. Scale bars, 100 μm. Mean ± s.e.m. of quantifications from 10 tumor IHC sections are shown (unpaired two-tailed Student’s t-test). *P < 0.05, ***P < 0.001. d, Tumor growth rate in immunocompetent mice inoculated with CT26 cells following IgG or CD8 antibody treatment in combination with vehicle control or ziftomenib (150 mg kg−1). Each data point represents mean ± s.e.m. tumor volumes (n = 10 in each group). Two-way ANOVA was used for statistical analysis. ***P < 0.001. e, Relative abundance of Cd274/PD-L1 in CT26 tumors with and without deletion of Men1 from bulk RNA-seq data (left). Mean ± s.e.m. of two biological replicates are shown (unpaired two-tailed Student’s t-test). Relative abundance of CD274/PD-L1 in MEN1-high and MEN1-low patient tumors in the TCGA COAD cohort (right). Mean ± s.e.m. of 50 patient tumors with the highest and the lowest MEN1 expression were shown (unpaired two-tailed Student’s t-test). *P < 0.05, **P < 0.01, ***P < 0.001. f, Tumor growth rate in immunocompetent mice inoculated with CT26 cells following IgG or PD-1 antibody treatment in combination with vehicle control or ziftomenib (150 mg kg−1). Each data point represents mean ± s.e.m. tumor volumes (n = 10 in each group). **P < 0.01, ****P < 0.0001. g, Schematic view of the dual function of MEN1 in modulating tumor–microenvironment interactions. MEN1 suppression reshapes MLL1 chromatin binding, triggering cytokine gene expression via MAVS- and cGAS-dependent viral mimicry response. Consequently, this leads to increased infiltration of tumor-promoting neutrophils and tumor-inhibiting CD8+ T cells in immunodeficient and immunocompetent conditions, respectively. DMSO, dimethyl sulfoxide. g, Created with BioRender.com.

Source data

Full size image

In concordance with Men1 knockout, drug treatment significantly reduced tumor growth compared with the control treatment in a dosage-dependent manner (Fig. 6b). Ziftomenib treatment had a similar growth inhibitory effect in the syngeneic 4T1 breast cancer allograft model in immunocompetent mice (Extended Data Fig. 10i,j). A significantly increased tumor growth rate was observed in the pancreatic cancer PDX model OCIP200 in immunodeficient mice following ziftomenib treatment (Extended Data Fig. 10k), consistent with the TME-dependent effects observed with MEN1 genetic depletion. Furthermore, treatment with ziftomenib resulted in a marked increase in CD8+ T cell infiltration, with stronger induction observed at a higher dosage (Fig. 6c). Although CD8+ T cell neutralization alone had no obvious effect on tumor growth in the control condition, it completely abolished the tumor-suppressive effect of ziftomenib (Fig. 6d), suggesting that MEN1 regulation of tumor growth depends on CD8+ T cells and that menin blockade might constitute a new immunotherapeutic approach.

Bulk RNA-seq showed significantly higher levels of Cd274/PD-L1 and a trend of increased Pdcd1/PD-1 expression in Men1-depleted CT26 xenograft tumors and MEN1-low patient tumors in the TCGA LUAD cohort (Fig. 6e), likely due to increased immune cell infiltrations. scRNA-seq and CyTOF showed that a subset of tumor and macrophage cells express PD-L1, and the CD8+ T cells express high levels of PD-1, with a modest increase in the knockout tumors (Supplementary Fig. 3a–e). Ziftomenib or anti-PD-1 treatment alone substantially reduced the CT26 tumor growth and the combination demonstrated a significantly stronger effect (Fig. 6f). These data suggest that pharmacological inhibition of menin alone or in combination with immunotherapy may benefit patients with functional menin protein (Fig. 6g).

Discussion

Tumors develop within a complex microenvironment and identifying clinically relevant therapeutic targets amid such complexity is a daunting task. In this study, we used parallel in vitro and in vivo CRISPR knockout screens to identify genes that modulate tumor and TME interactions. We found that the epigenetic regulator MEN1 markedly influences tumor growth under certain conditions in vivo, but does not affect cell growth in vitro.

MEN1 plays a complex multifaceted role in cancer. It is a well-characterized tumor suppressor, loss of which causes multiple endocrine neoplasia type 1 and development of lung, prostate, skin and central nerve system tumors47,48,49,50. Paradoxically, however, MEN1 also functions as an oncogene, notably in acute myeloid leukemia with NPM1 mutations of MLL1-rearrangements46,51. As a chromatin adapter, menin is known to directly interact with the N-terminal domain of H3K4-specific methyltransferase MLL1/MLL2. Our study reveals the existence of an antagonist role for menin, distinct from its previously described cofactor functions. Menin accomplishes this role potentially by squelching MLL1 in the COMPASS complex, resulting in reduced MLL1–chromatin interactions at repetitive genomic regions and subsequent suppression of cytokine-related gene expression. In addition, although menin–MLL1 coactivated target genes vary in different tissues, the suppression of cytokine signaling is highly consistent across different tissues and species, suggesting a highly conserved and common mechanism.

Insertion of transposable elements resulted in a complex distribution of interspersed repeats comprising almost half of the human genome52,53. The transcribed interspersed repeats form dsRNA, which is sensed by the RIG-1/MDA5-MAVS signaling pathway or reverse-transcribed and detected via the cGAS–cGAMP STING pathway37, leading to increased innate immune responsiveness through a viral mimicry mechanism38,39,54,55,56,57. Our data show that upon MEN1 depletion, MLL1 activates cytokine signals at least partially via the MAVS and cGAS–STING pathways. The potential cooperative effects of these epigenetic regulators warrant further exploration.

The immunocompetence-dependent tumor-inhibiting and -promoting function provides a mechanism underlying the paradoxical function of MEN1. Several inhibitors that block menin–MLL interactions have entered clinical development in acute myeloid leukemia25,26. Our data suggest that menin inhibition may also offer therapeutic value related to activation of the immune response in solid tumors. Because infiltrating immune cells may have both tumor-promoting and -inhibiting effects, the efficacy of menin inhibition will likely be context-specific. Additional studies identifying biomarkers predicting the efficacy are warranted.

In summary, our study presents a mechanism contributing to the paradoxical tumor-suppressive and oncogenic function of MEN1 and provides a strong rationale for targeting menin alone or in combination with immunotherapy for both hematologic and solid cancer.

Methods

All samples obtained in this study complied with the relevant ethical regulations approved by the institutional ethics committee and Research Ethics Board at the University Health Network (UHN).

Western blotting

Cells were lysed in RIPA buffer containing phosphatase and protease inhibitor cocktails (Roche, cat. no. 11697498001). Protein was quantified using a bicinchoninic acid assay. Secondary antibodies were used at a 1:5,000 dilution. Primary antibodies were diluted according to the manufacturers’ recommended ratios. Used antibodies included anti-menin (Bethyl Laboratories, cat. no. A300-105A), anti-MLL1 (Abcam, cat. no. ab272023), anti-H4K4me3 (Abcam, cat. no. ab8580), anti-UTX (Bethyl Laboratories, cat. no. A302-374A), anti-cGAS (Cell Signaling Technology, cat. no. 151025), anti-MAVS (Abcam, cat. no. ab89825), anti-TBK1 (Cell Signaling Technology, cat. no. 3504T), anti-pTBK1 (Cell Signaling Technology, cat. no. 5483T), GAPDH (Santa Cruz Biotechnology, cat. no. sc-47724) and Vinculin (Cell Signaling Technology, cat. no. 13901S), H3 (Cell Signaling Technology, cat. no. 4499L) and β-tubulin (Cell Signaling Technology, cat. no. 2128S).

CUT&RUN assays

In total, 250,000 cells per condition were washed twice with 1 ml of wash buffer by pelleting for 3 min at 600g. As previously described58, the cells were bound to Concanavalin A magnetic beads (Cell Signaling Technology, cat. no. 93569S) using nuclear binding buffer and incubated at room temperature on a rotator at 15 rpm for 10 min. Antibodies (anti-menin (Bethyl Laboratories, cat. no. A300-105A) and anti-MLL1 (Abcam, cat. no. ab272023)), were diluted to a ratio of 1:100 using antibody buffer. After the 10-min incubation, the samples were placed on a magnetic separator. The nuclear binding buffer was removed and the antibody mixture was added to the bead-bound samples. Samples were incubated overnight at 4 °C on a rotator. Following overnight incubation, pAG-MNAse enzyme (Cell Signaling Technology, cat. no. 40366S) was diluted 1:1,000 in digitonin buffer. The samples were placed on a magnetic separator, supernatant was removed and pAG-MNAse buffer was added. The samples were mixed on a rotator (15 rpm) and incubated at 4 °C for 1 h. To activate the pAG-MNAse, 3 μl of cold 100 mM CaCl2 was added and the samples were incubated on ice for 30 min. After 30 min, stop buffer was immediately added to quench the pAG-MNAse reaction. DNA fragments were released from the cells by incubating at 37 °C for 10 min with 700 rpm shaking, followed by a 5-min centrifugation at 16,000g at 4 °C. Supernatant containing DNA fragments was moved to a new tube containing 3 μl of 10% SDS and 2.5 μl of Proteinase K (Thermo Fisher Scientific, cat. no. EO0492) then incubated for 10 min at 70 °C. The DNA fragments were purified using phenol–chloroform extraction. Purified DNA was quantified via Qubit and 7.5 ng was used for input into the library preparation. Library preparation was completed following the Takara SMARTer ThruPLEX DNA-Seq protocol (Takara Bio, cat. no. R400676). Samples were amplified using 11 PCR cycles as determined via qPCR. Final library traces were generated using the Agilent 2100 Bioanalyzer and quantified via qPCR before sequencing. Libraries were sequenced to a depth of 50–60 million reads per sample using paired-end 50 sequencing configuration on Illumina’s NovaSeq 6000.

Chromatin immunoprecipitation assay

ChIP assays on A549 cells were conducted as previously described59, both with and without MEN1 knockdown. In brief, 5 million cells underwent crosslinking with 1% formaldehyde for 10 min at room temperature, followed by quenching with 125 mM glycine. Subsequently, the cell pellets were twice washed with cold PBS buffer and resuspended in 300 μl of modified RIPA buffer (10 mM Tris–HCl pH 8, 1 mM EDTA, 140 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate) supplemented with protease inhibitor. The resuspended pellet underwent sonication for 18 cycles (30 s on, 30 s off) at 4 °C using Bioruptor (Diagenode).

Following sonication, the cell lysate was precleared with 40 μl of protein A/G beads before incubation with antibody-coated beads at 4 °C overnight. To prepare the antibody-coated beads, 5 μg of target antibody of interest were incubated with 11 μl of protein A and 11 μl of protein G beads (Thermo Fisher Scientific, cat. no. 10002D and 10004D) with rotation for at least 6 h at 4 °C. Antibodies used for ChIP assays included anti-menin (Bethyl Laboratories, cat. no. A300-105A), anti-MLL1 (Abcam, cat. no. ab272023), anti-H3K27me3 (Abcam, cat. no. ab6002), anti-H4K4me3 (Abcam, cat. no. ab8580) and anti-UTX (Bethyl Laboratories, cat. no. A302-374A).

After incubation, beads underwent a series of washes: once with 0.5 ml of modified RIPA buffer, once with high salt modified RIPA buffer (NaCl increased to 500 mM), once with LiCl buffer (10 mM Tris–HCl pH 8, 1 mM EDTA, 250 mM LiCl, 0.5% NP-40, 0.1% sodium deoxycholate) on a rotator for 5 min each at 4 °C, and finally twice with 0.5 ml of TE buffer (pH 8). The beads were then resuspended in 100 μl of de-crosslinking buffer (1% SDS, 0.1 M NaHCO3). Subsequently, 1 μl of RNase A (Thermo Fisher Scientific, cat. no. EN0531) was added, and the mixture was incubated at 37 °C for 30 min with shaking, followed by incubation at 55 °C for 30 min with shaking in the presence of 2 μl of Proteinase K (Thermo Fisher Scientific, cat. no. AM2546). The crosslinking was reversed by incubating at 65 °C overnight with shaking. DNA purification was carried out using the PCR purification kit (Qiagen, cat. no. 28004), and the purified DNA was then subjected to qPCR or Illumina ChIP-Seq library construction using the ThruPLEX DNA-seq kit (Takara, cat. no. R400676).

sgRNA pooled library design and synthesis

The Epi-Drug library consists of ~12,500 sgRNAs targeting 317 epigenetic regulators, 657 US Food and Drug Administration-approved drug targets based on DrugBank 4.3 (ref. 60), and control genes, with 10 sgRNAs per gene on average (Supplementary Table 1). The sgRNAs were designed using the CRISPR-DO tool that accounted for sgRNA specificity and cutting efficiency61. sgRNAs were synthesized as 73-base polymer/oligonucleotides (CustomArray). After being amplified by PCR, the PCR product was purified and cloned in the lentiGuide-Puro vector using BsmBI (NEB). Ligation was performed using the NEBuilder HiFi DNA Assembly Cloning Kit and transformed into an electrocompetent stain (Stbl4; Thermo Fisher Scientific) to achieve ~300× coverage. Colonies were scraped off agar plates with LB medium. Plasmid DNA was extracted using a NA0310 Sigma GenElute HP Plasmid Maxiprep Kit and adequate library representation of each sgRNA was confirmed by NGS.

CRISPR–Cas9 screening, sequencing and analysis

Library viruses were produced in HEK293FT cells and multiplicity of infection was determined for the A549 cell line as previously described9. The Cas9-expressing A549 cell line was infected with the library at an multiplicity of infection of ~0.3 and coverage of >200×. At 24 h post-infection, cells were selected with puromycin for 72 h (1 μg ml−1) and then cultured for ~21 days in Petri dishes or subcutaneously inoculated into mice for ~21 days. Day 0 genomic DNA was harvested after 3 days of puromycin selection. Genomic DNA was extracted and sgRNA inserts were amplified by PCR as previously described62. Screens were performed in duplicate, and the libraries were sequenced on an Illumina HiSeq2500.

The NGS data from CRISPR screens were first aligned to the library sgRNAs using bowtie. The resulting count matrix was the input to the tool MAGeCK22, which estimates the enrichment/depletion of individual sgRNAs using a negative binomial model and estimates the enrichment of genes using the robust ranking aggression model.

Mouse tumor growth experiments

Four- to six-week-old male NSG, C57BL6 and BALB/c mice were maintained at the UHN Animal Resources Centre. NSG, C57BL6 or BALB/c mice were injected subcutaneously with 50 μl of a single-cell suspension containing 5 × 105 MEN1/Men1 knockout and control A549, HCT116, CT26, 4T1, HKP1 or DKO cells. Syngeneic CT26, 4T1 models were injected subcutaneously in 50 μl of a single-cell suspension into BALB/c mice. Cryo-preserved OCIP200 PDX tissues were implanted subcutaneously in NSG mice. The tumors were expanded in mouse replicates (n = 10) to evaluate the agent ziftomenib. Once tumors had initiated growth, the vehicle or drug was administered daily by oral gavage. To generate humanized mice, human peripheral blood mononuclear cells were engrafted into NSG pups after the pups had been irradiated at 1 Gy. Ten to twelve weeks after the injection, blood samples were collected from the mice, and human immune cell engraftment was determined by flow cytometry. Humanized mice with engraftment efficiency (percent human CD45+/(percent human CD45+ + percent mouse CD45+)) values between 20% and 50% were used in subsequent A549 xenograft experiments. Mouse weight and tumor volume were measured twice weekly. Xenograft (tum) or sizes were measured twice per week. Tumor volumes were calculated using the formula: tumor volume (mm3) = L × W2/2, where L is the major axis (largest cross-sectional diameter) of the tumor and W is the minor axis. The data are presented as the mean ± s.e.m. Mice were sacrificed when the diameter of the tumor reaches 1.5 cm. All procedures were performed in accordance with the International Guidelines for the Use of Animals and approved by the Animal Care Committee at UHN.

Gene knockout and transfection

Guide RNA plasmids targeting MEN1 was constructed into a lentiCRISPR V2-Blast vector (Addgene, cat. no. 83480). Guide RNA plasmids targeting MAVS and cGAS were constructed in a lentiCRISPR V2 vector (Addgene, cat. nos. 83480, 98291, 98290). siRNAs were transfected into the cancer cells by RNAiMAX reagent (Thermo Fisher Scientific, cat. no. 13778075), following the manufacturer’s protocol. siRNAs used in this study were ordered from Thermo Fisher Scientific and sequences are listed in Supplementary Table 7.

Statistical analyses and reproducibility

Statistical analyses for in vitro and in vivo assays were performed using GraphPad Prism v.6 (GraphPad Software). Data were expressed as mean ± s.e.m. Two-tailed unpaired Student’s t-test and one-way analysis of variance (ANOVA) were performed to identify significant differences between groups in our experiments. R v.3.3.0 was used to perform statistical analysis. Details of the test method used for statistical analysis is specified in the relevant figure captions and Methods. All values were considered significantly different at P < 0.05. Western blotting for MEN1 knockout samples was performed in biological replicates.

RNA-seq mapping and data processing

RNA was extracted with TRIzol reagent (Thermo Fisher Scientific, cat. no. 15596026) and processed using a RNA-seq library preparation kit (Illumina, cat. no. RS-122-2101) to produce libraries for deep sequencing on NextSeq. Library preparation and sequencing were performed according to the manufacturer’s protocol. RNA-seq raw reads were first filtered by trim_galore (v.0.5.0) then mapped to the human genome (hg38) using STAR (v.2.4.2a)63 software with default parameters. The hg38 GENCODE gene list was used for all transcription level analysis. RNA-seq reads strands were determined by RSeQC (v.2.6.1)64,65. HTSeq (v.0.11.0)66 was used to obtain gene-level read counts from STAR-mapped bam files. The resultant gene read count table was subjected to DESeq2 (v.1.22.2)24 for differential gene analysis and a false discovery rate cutoff of 0.01 was chosen to identify significant differential genes. A log2(fold change) value less than −1 and a log2(fold change) value above 1 were chosen as upregulated genes and downregulated genes, respectively.

For visualization of RNA-seq data, we generated bedGraph files using genomeCoverageBed function in the BEDTools suite with signal scaled to reads per 10 million for both plus and minus strands. The bedGraph files were converted to bigwig format using the bedGraphToBigWig function in the BLAT suite.

CUT&RUN and ChIP-seq mapping and data processing

CUT&RUN and ChIP-seq reads were mapped to the human genome (hg38) using bowtie. MACS2 callpeak was used for peak calling with the parameter ‘–SPMR’ on. MACS232 bdgdiff was used to call differential peaks. Resultant bedgraph files were converted to big wiggle files with the bedGraphToBigWig function. CREAM67 was applied to call broad peak regions followed by the removal of ENCODE blacklist regions.

Gene Ontology analysis

Gene Ontology (GO) and KEGG analyses were performed using R package clusterProfiler v.3.10.1 (ref. 68). Terms with a false discovery rate <0.01 were considered significantly enriched. A complete list of terms enriched can be found in the supplementary datasets.

BETA analysis

BETA (v.1.0.7)33 software was used to predict menin directly regulated genes by combining menin binding peaks and differentially expressed genes.

Repeats analysis

The human repeat masker file was downloaded from the website (http://www.repeatmasker.org/genomes/hg38/RepeatMasker-rm405-db20140131/hg38.fa.out.gz). The reference file was converted to bed format using bedops/2.4.37. RNA-seq reads were remapped to genome hg38 by STAR (v.2.4.2a)63 with parameter –outFilterMultimapNMax set to 100 and –winAnchorMultimapNmax set to 100. Exon reads were calculated by HTSeq (v.0.11.0)66. Exons with more than five reads were subtracted from the repeats masker file by running bedtools (v.2.27.1)69 subtract. Read counts on repeat regions were calculated by bedtools (v.2.27.1)69 multicov. DESeq2 was used to call differential repeats.

Immunofluorescence confocal microscopy

A549 cells constitutively expressing Cas9 protein were infected with lentivirus containing sgRNA targeting LacZ or MEN1. A round coverslip was loaded to a 24-well plate and incubated with 1 ml of 100% ethanol for 5 min at room temperature. After one wash with PBS buffer, 75,000 cells in growth media were seeded to each well. The plates were incubated at 37 °C overnight to allow attachment of the cells to the coverslip. The next day, the medium was aspirated, and the cells were washed once with PBS and fixed with ice cold methanol for 15 min at −20 °C. The cells were washed three times with PBS and blocked with 1% bovine serum albumin in PBS at room temperature for 1 h. Primary antibody targeting dsRNA (J2; Scicons, cat. no. 10010500) and γ-H2AX (EMD Millipore, cat. no. JBW30) diluted with blocking solution at a ratio of 1:500 was added to each well, and the plates were incubated at 4 °C overnight with shaking. The next day, the diluted antibodies were aspirated, and the cells were washed three times with PBS for 10 min at room temperature with shaking. The cells were then incubated with the secondary antibody (anti-mouse immunoglobulin G (IgG) (H+L), F(ab′)2 fragment (Alexa Fluor 647 conjugate; Cell Signaling Technology, cat. no. 4410s) diluted with blocking solution at a ratio of 1:1,000 in a dark environment for 1 h at room temperature with shaking. The cells were washed three times with PBS in a dark environment at room temperature for 10 min with shaking. The coverslips were mounted to glass slides with ProLong Gold Antifade Mountant with 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific, cat. no. P36935). Confocal microscopy was performed with Zeiss LSM700 (oil ×40 magnification) and ImageJ software was used to analyze the images acquired. Corrected total cell fluorescence was calculated as: corrected total cell fluorescence = integrated density − (area of selected cell × mean fluorescence of background readings).

Tumor dissociation to single cells

Fresh tumor samples were harvested from mice and minced into small, approximately 1-mm, cubic pieces on ice. The samples were washed five times with PBS and enzymatically digested in DMEM/F12 media containing 1× collagenase/hyaluronidase (STEMCELL Technologies, cat. no. 07912) and 5% FBS buffer for 2 h at 37 °C with rotation. After dissociation, samples were centrifuged at 350g for 5 min. After the supernatant was removed, 5 ml of 0.25% Trypsin-EDTA (STEMCELL Technologies, cat. no. 07901) was added to the pelleted cells followed by incubate for 5 min at 37 °C. Then 10 ml of cold modified Hanks’ balanced salt solution buffer (HBSS modified; STEMCELL Technologies, cat. no. 37150) supplemented with 2% FBS was added to the solution followed by centrifugation for 5 min. After the supernatant was removed, 1 ml of dispase (STEMCELL Technologies, cat. no. 07913) in 5 mg ml−1 and 100 μl of DNase I (STEMCELL Technologies, cat. no. 07900) in 1 mg ml−1 were added to the cell pellets and were pipetted up and down with a 27 G syringe five times. The cell pellet was resuspended in 10 ml of HBSS modified buffer supplemented with 2% FBS and filtered through a 40μm cell strainer (STEMCELL Technologies, cat. no. 27305). The product can be washed with the appropriate buffer and is ready for CyTOF or scRNA-seq.

Mass cytometry by time-of-flight

Details of each antibody are listed Supplementary Table 7. Purified unconjugated antibodies were labeled with metal tags at the SickKids-UHN Flow and Mass Cytometry Facility using a MaxPar Antibody Labeling kit (Fluidigm). Directly conjugated antibodies were purchased from Fluidigm. All working antibody concentrations were determined by titration.

Single-cell suspensions from individual samples were washed with PBS and pulsed with 12.5 μM cisplatin (Fluidigm) in PBS for 1 min before quenching with CyTOF staining media (Mg+/Ca+ HBSS containing 2% FBS (Multicell), 10 mM HEPES (Corning) and FBS underlay). Cells were then fixed for 12 min at room temperature with transcription factor fixative (Bioscience, cat. no. 00-5523-00), permeabilized and individual samples were barcoded according to manufacturer’s instructions (Fluidigm 20-Plex Pd Barcoding Kit, cat. no. 201060), before being combined. Combined samples were resuspended in staining media containing metal-tagged surface antibodies and Fc block (Bioxcell, cat. no. CD16/32) for 30 min at 4 °C. Cells were washed, then permeabilized and stained with metal-tagged intracellular antibodies using a transcription factor staining buffer set. Cells were incubated in PBS containing 0.3% (w/v) saponin (Sigma, cat,. no. S7900-25G), 1.6% (v/v) paraformaldehyde (Polysciences) and 50 nM iridium (Fluidigm) for 1 h at room temperature. Cells were analyzed on a Helios or Helios2 mass cytometer (Fluidigm) at the SickKids-UHN Flow and Mass Cytometry Facility. EQ Four Element Calibration Beads (Fluidigm) were used to normalize signal intensity over time using CyTOF software v.6.7. Flow cytometry standard files were manually debarcoded and analyzed using Cytobank6.2 (Cytobank).

Data were pre-transformed by ArcSinh transformation (cofactor = 5) in R. Phenogragh analysis was performed with R package ‘Cytofkit’70,71 (seeds = 42, k = 30). Uniform manifold approximation and projection (UMAP) analysis was performed using the R package ‘UMAP’72 with default settings. Differential states and differential abundance were calculated using the limma and edgeR tests, respectively, with the R package ‘diffcyt’73. An adjusted P value <0.05 was considered significant for the comparisons made. Median signal intensity values were used to generate heatmaps. Heatmaps were generated in R using the blue–red, plasma or viridis color packages and the R package ‘gplots’.

scRNA sequencing (10X Genomics)

Cell capture and complementary DNA synthesis was carried out using the Single Cell 3′ Library and Gel Bead Kit V2 (10X Genomics, cat. no. 120237) and a Chromium Single Cell A Chip Kit (10X Genomics, cat. no. 120236). The cell suspension (300–600 living cells per microliter) was loaded onto the chromium single-cell controller (10X Genomics) to generate single-cell gel beads in the emulsion, according to the manufacturer’s protocol. In short, single cells were suspended in PBS containing 0.04% bovine serum albumin. Approximately 7,000 cells were added to each channel, and the target cell recovery rate was estimated to be 3,000 cells. Captured cells were lysed and the released RNA was barcoded through reverse transcription in individual gel beads in emulsion.

Using an S1000TM Touch Thermal Cycler (Bio-Rad) to reverse transcribe, the gel beads in emulsion were programmed at 53 °C for 45 min, followed by 85 °C for 5 min, then held at 4 °C. Complementary DNA was generated and then amplified, and the quality was assessed using an Agilent 4200 system.

For scRNA-seq library preparation, according to the manufacturer’s protocol, scRNA-seq libraries were constructed using a Single Cell 3′ Library Gel Bead Kit V2. Sequencing was performed on an Illumina NovaSeq 6000 sequencer with a sequencing depth of at least 100,000 reads per cell and 150-bp (PE150) paired-end reads.

Data analysis and visualization

Data analysis was conducted using R (v.3.4.1). Data visualization was conducted using ggplot2 (v.3.1.0). Figures 1a and 6g and Extended Data Fig. 10a were created using BioRender.com.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.