Menin inhibition impairs metastatic colonization of Ewing sarcoma

The MEN1 (Multiple Endocrine Neoplasia 1) gene encodes Menin, a unique scaffolding protein that regulates many cellular processes including transcription, DNA damage, cell cycle, and signal transduction through interactions with a wide range of context-dependent partner proteins (1, 2). Menin has broad and fundamental roles in early development. Men1-null mice are early embryonic lethal at E11.5–E13.5 due to widespread defects in organogenesis (3). In addition, studies of human embryonic stem cells implicate Menin in regulating timing of endoderm, cardiomyocyte, and neuronal differentiation (4). Targeted deletion of Menin from embryonic murine neural crest cells and somite precursors disrupts craniofacial skeleton and rib development in vivo (5), and misregulation of Menin expression in mesenchymal progenitor cells in vitro disrupts lineage commitment and terminal differentiation (6–8). Thus, Menin plays diverse roles in regulating fidelity of organogenesis and tissue differentiation during normal development. Importantly, the diverse and highly tissue-specific functions of Menin are determined by its context-dependent interaction partners, including epigenetic modifiers (e.g., MLL1, MLL2), transcription factors (RUNX2, JUND), and signaling pathway components (1, 6, 8–11).

In human disease, MEN1 was first identified as a tumor suppressor gene; germline inactivation of MEN1 is associated with the genetic cancer predisposition syndrome multiple endocrine neoplasia (12–14). Patients with MEN1 syndrome are predisposed to developing parathyroid and pituitary gland tumors, neuroendocrine tumors of the pancreas, and other endocrine and nonendocrine (e.g., meningioma) cancers (15). By contrast in rearranged mixed-lineage leukemia (MLLr), Menin functions as an oncogene (10, 16). Protein-protein interactions between Menin and MLL fusion proteins drive transcriptional activation of leukemogenic genes. Significantly, pharmacologic inhibitors of Menin have been developed that disrupt Menin interactions with MLL1-fusions and with WT MLL1 (KMT2A), and these agents induce terminal differentiation and subsequent death of MLLr and NPM1-mutant (NPM1m) leukemia (17–19). Revumenib (Syndax Pharmaceuticals Inc.) was recently granted FDA approval for treatment of patients with MLLr and NPM1m leukemia (20), and multiple additional Menin inhibitors are currently in development (21). Significantly, recent studies have also identified protumorigenic roles for Menin, and/or Menin-MLL in some solid tumors. These include carcinomas such as breast, lung, endometrial, ovarian, colorectal, and prostate cancer and tumors of mesenchymal origin such as gastrointestinal stromal tumors and Ewing sarcoma (EwS) (reviewed in ref. 22). If and how Menin inhibitors could be leveraged for solid tumor therapies requires that tumor type-specific functions of Menin and its interaction partners be elucidated.

EwS tumors are aggressive cancers of bone and soft tissues that are of presumed mesenchymal and/or neural crest stem cell origin (23). Multiagent systemic chemotherapy combined with local control can induce long-term remission in most patients who present with localized disease (23, 24). However, in up to 25% of patients, tumors recur at distant sites, and patients who relapse with metastatic disease are rarely cured. We previously reported that Menin protein is highly expressed by EwS cells and that partial inhibition of Menin blunts tumorigenicity and pro-oncogenic metabolic programs (25–27). For the current study, we generated Menin-KO EwS cells to definitively characterize how Menin contributes to EwS pathogenesis. Our results show that Menin is required for colonization and metastatic competence of EwS cells in vivo. Mechanistically, transcriptomic profiling revealed reproducible downregulation of nucleic acid biosynthesis programs and MYC target genes in MEN1-KO cells. Conversely, Menin depletion led to upregulated expression of mesenchymal differentiation and development genes. Significantly, tumors that successfully arose from MEN1-KO cells were transcriptionally rewired to reverse these effects; expression of MYC target genes was restored while TGF-β and epithelial-mesenchymal transition (EMT) gene programs were suppressed. Thus, these data suggest that high levels of Menin in EwS cells promote tumorigenicity and metastatic colonization, at least in part, by promoting transcriptional activity of MYC. It was previously reported that, in HT-1080 fibrosarcoma cells, Menin directly interacts with MYC at E-boxes to amplify transcription of MYC-bound target genes (28). We performed coimmunoprecipitation studies and confirmed that Menin and MYC interact in EwS cells. Moreover, our data show that this interaction occurs outside Menin/MLL-containing compass-like complexes. We next exposed EwS cells to VTP50469, the preclinical precursor to revumenib, to test the effect of Menin-MLL interaction inhibitors on Menin-MYC interactions and MYC target gene expression. Notably, VTP50469 partially disrupted the interaction between Menin and MYC and significantly blunted the MYC signature. Given the profound effect of Menin loss on metastatic colonization, we tested whether VTP50469 would inhibit outgrowth of disseminated EwS tumor cells into macroscopic tumors. Feeding mice therapeutic doses of VTP50469 chow significantly reproducibly inhibited colonization of disseminated tumor cells (DTCs). Together, these data confirm the critical importance of Menin to tumorigenic phenotypes in EwS and suggest that Menin inhibitors could be repurposed to inhibit tumor colonization in patients who are at high risk of metastatic relapse.

In vivo tumorigenicity of MEN1-KO EwS cells is impaired. To examine the function of Menin in EwS, we used 2 well established cells lines, A673 and TC32. Both are models of aggressive disease that were originally derived from adolescent female patients with EwS, and both harbor the most common type I EWS-FLI1 fusion (Ex7–Ex6) and a homozygous deletion of the cell cycle inhibitor, CDKN2A. Additionally, A673 is p53 null and has a heterozygous BRAF mutation. TC32 is p53 WT and contains a STAG2 mutation (29). Using these distinct cell lines, we generated MEN1-KO cells using the CRISPR/Cas9 system with 2 guide RNAs, 1 targeting exon 2 (guide 1) and 1 targeting exon 3 (guide 5), to knock out the MEN1 gene (Figure 1A). In parallel, control cell lines were created using a nontargeting gRNA. Single-cell clones were isolated and propagated to generate clonally derived MEN1-KO and control cell lines. MEN1 transcript levels were reduced in all MEN1-KO clones (Figure 1B), and Menin protein was not detected by Western blot, confirming complete loss of the protein (Figure 1C). Interestingly, protein levels of both MLL1 (KMT2A) and MLL2 (KMT2B) were reproducibly reduced in MEN1-KO cells, indicating that both MLL1 and MLL2 proteins are highly expressed by EwS cells and expression is maintained by Menin.

Figure 1

Menin promotes EwS tumorigenesis. (A) CRISPR guides in exons 2 (guide 1) and 3 (guide 5) used to knock out MEN1. (B) MEN1 expression in control and MEN1-KO clones. qPCR determined expression levels were plotted with P values generated using a 1-way ANOVA with a Dunnett’s T3 test comparing MEN1-KO to control. (C) Western blot of Menin, MLL1, MLL2, and GAPDH in control and MEN1-KO clones. (D) Incucyte proliferation assays of control and MEN1-KO clones plotting percent confluence over time. Representative of n = 3–4. (E) S.c. tumor growth of A673 and TC32 control and MEN1-KO cells in NSG mice. Tumor volume is plotted over time for each mouse. The area under the curve for each mouse was used to determine P values calculated using a 1-way ANOVA with a Dunnett’s T3 test comparing MEN1-KO to control or parent. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

To assess whether loss of Menin affected growth of EwS cells in vitro, we used Incucyte assays to measure real-time proliferation. As shown, proliferation of MEN1-KO cells was not significantly changed relative to controls (Figure 1D). Next, we evaluated the in vivo tumorigenic potential of MEN1-KO cells. Notably, despite an absence of observable effects on cell proliferation in vitro, the capacity of A673 and TC32 MEN1-KO cells to generate s.c. tumors in NOD SCID γ (NSG) mice was significantly impaired compared with both parental and clonally derived control gRNA cell lines (Figure 1E). Thus, depleting EwS cells of Menin inhibits their ability to initiate local tumor formation, confirming that Menin plays a critical role in mediating tumorigenicity.

Loss of Menin leads to widespread transcriptional changes in cancer-associated gene programs. Menin has well-established roles in regulating gene transcription in development and in cancer. Therefore, to identify Menin-regulated genes in EwS, we performed transcriptomic profiling of control and MEN1-KO A673 and TC32 cells. RNA-seq data showed widespread and highly significant changes in gene expression in both A673 and TC32 MEN1-KO cell lines compared with controls, revealing a critical role for Menin as a master regulator of gene transcription (Supplemental Figure 1, A–D, and Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.202419DS1). To identify changes that were most likely to mediate loss of tumorigenicity, we focused on genes that were reproducibly altered upon Menin loss in both cell lines. Despite evidence of cell context specificity (Supplemental Figure 1, E–H), 1,124 transcripts were commonly downregulated (Figure 2A) and 1,258 genes were upregulated (Figure 2B) in both models. Given the breadth of shared genes, we applied Hallmark and Gene Ontology Biological Processes (GO-BP) gene set enrichment tools to the gene lists to identify key biologic programs that were reproducibly altered in MEN1-KO cells. mTORC1 signaling, E2F targets, MYC targets, and G2-M checkpoint genes were highly enriched among genes that were downregulated in MEN1-KO cells (i.e., Menin-activated genes) (Figure 2C). GO-BP analysis of Menin-activated genes further labeled enrichment of nucleotide metabolism, a key process in cell growth and proliferation. Though no changes in proliferation were detected in vitro, growth of MEN1-KO cells was inhibited in vivo (Figure 1), nominating the tumor microenvironment (TME) as a player in mediating the effects of Menin on cell proliferation. Genes reproducibly upregulated in MEN1-KO cells (i.e., Menin-repressed genes) were enriched for mesenchymal developmental programs including EMT and myogenesis, as well as the apical junction pathway that regulates cell-cell contacts (Figure 2D). Biological processes involved in cell migration and modulation of extracellular matrix (ECM) and the TME were enriched in Menin-repressed genes. Consistent with this, MEN1-KO cells were more invasive than controls when embedded in 3D collagen (Figure 2, E and F). Thus, these findings suggest that Menin maintains the tumorigenicity of EwS cells by sustaining activation of growth-promoting programs and by suppressing differentiation programs. Moreover, the effect of these Menin-regulated genes on tumor behavior is minimal in standard 2D culture but highly evident in the context of an in vivo TME.

Figure 2

MEN1-KO EwS cells have altered expression of proliferation and differentiation programs. (A and B) Overlap of downregulated (A) and upregulated (B) genes in A673 and TC32 MEN1-KO cells. Significant genes in each model were defined as P_adj < 0.05 as described in Methods. The statistical significance of the overlap in the Venn diagrams was calculated using a Fisher’s exact test. (C and D) The top 10 enriched Hallmark pathways and the top 5 enriched GO-BP pathways among overlapping downregulated (C) and upregulated (D) genes as determined by Enrichr. The number of overlapping/total genes is shown for each pathway. (E) Invasion of control and MEN1-KO cells embedded in rat tail collagen for 5 days. Representative phase and phalloidin (red)/DAPI (blue) stained images are shown. Scale bars: 100 μm. (F) Invasion in collagen was quantified and compared between control and MEN1-KO cells using spheroid circularity, as described in Methods. Statistical comparisons were performed using a Welch’s t test with a 2-tailed P value at 95% CI.

Menin is required for metastatic competence of EwS cells. EwS cells are highly heterogeneous and exist along a transcriptional spectrum from highly undifferentiated to more mesenchymal cell states (30–32). EwS cells that upregulate mesenchymal genes show enhanced capacity to metastasize in vivo. Given our discovery that EMT genes and invasion were augmented in MEN1-KO cells in vitro, we next tested their metastatic competence in vivo. We first evaluated metastatic colonization using tail vein models. Despite their more mesenchymal and invasive profiles, MEN1-KO cells were largely incapable of generating macroscopic tumors when intravenously administered to recipient mice. This was evident by serial luciferase imaging (Figure 3A) and at necropsy (Figure 3B). Whereas control cells generated numerous tumors in the livers of NSG mice that nearly completely effaced normal liver, MEN1-KO cells generated few to no visible tumors (Figure 3B).

Figure 3

Menin promotes colonization and metastasis. (A) IVIS images of NSG mice 3 or 5 weeks after tail vein inoculation with 1 × 10⁶ control or MEN1-KO luciferase-tagged A673 and TC32 cells, respectively. Sex of mice are indicated (M, male; F, female). (B) Photos of livers at necropsy show markedly diminished macroscopic tumor burden in mice that received MEN1-KO vs control cells. Scale bars: 1 cm. (C) Diagram of the spontaneous metastasis model. In total, 2 × 10⁵ A673 control or MEN1-KO cells were injected in the subrenal capsule of the left kidney of NSG mice and 26 days after injection the left kidney and the liver were harvested. (D and E) Photographic images of left kidney (D) and liver (E) from each mouse are shown. Scale bars: 1 cm. (F and G) H&E-stained sections of the livers (F) and kidneys (G) from organs marked by asterisks in D and E. (G) Right panel in is 10X amplified image to show tumor-kidney border. Scale bars in F are 2,000 µm for top and 1,000 µm for bottom and in G are 1 mm in top left, 2 mm in bottom left, 100 µm in top right, and 200 µm in bottom right.

Having established that colonization of circulating MEN1-KO cells was inhibited in the tail vein model, we next used a subrenal capsule model to study spontaneous metastasis of EwS cells from kidney to liver. A673 cells were injected by ultrasound guidance into the subrenal capsule of the left kidney, as we have previously described (Figure 3C) (33). Control (N = 4) and MEN1-KO (N = 5) cell recipient mice were euthanized after 26 days, at which point all mice had developed kidney tumors. In keeping with s.c. tumor models, tumors derived from MEN1-KO cells were mainly smaller than control tumors (Figure 3D). Significantly, macroscopic tumors were detected in the livers of all control and none of the MEN1-KO cell recipient mice (Figure 3E). Importantly, this defect in spontaneous metastasis of MEN1-KO cells could not be entirely explained by the smaller size of the primary tumors. As shown in Figure 3, D and E, a control tumor of comparable size to MEN1-KO tumors generated dozens of metastases that completely replaced normal liver. The marked difference in liver tumor burden, in conjunction with comparable primary tumor size, were confirmed by H&E staining (Figure 3F). Interestingly, despite their evident loss of metastatic competence, MEN1-KO cells successfully invaded normal adjacent kidney parenchyma (Figure 3G). Thus, these in vivo studies of MEN1-KO cells collectively demonstrate that metastatic competence of EwS cells is profoundly dependent on Menin. In particular, they suggest that Menin is required for successful colonization of EwS cells at metastatic sites.

Tumors derived from MEN1-KO cells are transcriptionally rewired to restore Menin-regulated gene programs. Though there was a marked delay in engraftment, some A673 MEN1-KO cells did eventually generate tumors in vivo despite persistent loss of Menin protein (Figure 1E, Figure 4A, and Supplemental Figure 2). We hypothesized that defining the mechanisms by which these MEN1-KO cells had bypassed the effects of Menin loss would illuminate the most critical transcriptional targets of Menin that mediate its effects on tumor colonization. To address this, we performed RNA-seq on 12 tumors: 3 generated from the control gRNA cell line and 3 from each of the 3 MEN1-KO cell lines (Figure 4A). Comparison of tumor transcriptomes revealed thousands of differentially expressed transcripts that were up- or downregulated in MEN1-KO tumors compared with controls (Figure 4, B and C, and Supplemental Table 2). Focusing on the transcripts that were reproducibly altered in tumors from all 3 MEN1-KO cell lines identified 936 commonly upregulated and 1,059 downregulated genes. Commonly upregulated genes were enriched for G2-M checkpoint, E2F, and MYC targets (Figure 4D). Remarkably, these same pathways were significantly downregulated in MEN1-KO cells (Figure 2C). Comparing expression patterns of specific genes in these overrepresented pathways confirms that expression of core oncogenic genes was reestablished in MEN1-KO cells coincident with tumor outgrowth (Figure 4E). Similarly, transcriptional rewiring of Menin-repressed genes was also evident in MEN1-KO cell–derived tumors (Figure 4F). Restoration of tumorigenicity was associated with renewed suppression of genes that were relatively upregulated in MEN1-KO cells (Figure 2D), most predominantly genes that regulate NF-κB, hypoxia, and EMT pathways (Figure 4, F and G). Thus, these studies of in vivo tumors confirm that Menin serves as a master regulator of oncogenic gene programs in EwS. Moreover, they reinforce the hypothesis that Menin-dependent activation of MYC target gene expression is critical for maintenance of tumorigenicity.

Figure 4

Tumors derived from MEN1-KO cells are transcriptionally rewired. (A) Photos of tumors removed from NSG mice 28 days after s.c. injection of A673 control and MEN1-KO cells (from Figure 1E). Three tumors from each cell line were processed for RNA-seq (denoted by asterisk). Sex of mice is indicated (M, male; F, female). Scale bars: 1 cm. (B and C) Overlap of genes that were significantly upregulated (B) or downregulated (C) (P_adj < 0.05) in MEN1-KO cell-derived tumors compared with controls. (D) Enriched Hallmark pathways (P_adj < 0.05) as determined by Enrichr among shared upregulated genes in MEN1-KO tumors. Overlap/total genes shown for each pathway. Green arrows indicate pathways that were identified as being downregulated in A673 MEN1-KO cells in vitro (Figure 2C). (E) Heatmap of log₂FC of genes from Hallmark pathways that were downregulated in MEN1-KO cells in vitro and upregulated in MEN1-KO cell-derived tumors in vivo. (F) Enriched Hallmark pathways (P_adj < 0.05) as determined by Enrichr among shared downregulated genes in MEN1-KO tumors. Overlap/total genes shown for each pathway. Green arrows indicate pathways that were identified as being upregulated in A673 MEN1-KO cells in vitro (Figure 2D). (G) Heatmap of log₂FC of genes from Hallmark pathways that were upregulated in MEN1-KO cells in vitro and downregulated in MEN1-KO cell-derived tumors in vivo.

Menin and MYC proteins interact outside MLL-complexes in EwS cells. Our transcriptomic analyses identified MYC target genes as being among the most highly and reproducibly altered by Menin depletion. We therefore sought to address whether this is mediated via direct regulation of MYC levels. qPCR and RNA-seq analyses revealed no change in MYC mRNA expression in either MEN1-KO cells or tumors (Figure 5, A and B). Likewise, MYC protein levels were largely unchanged, although expression was modestly increased in a few tumors derived from MEN1-KO cells (Figure 5, C and D). Thus, these data suggest that Menin does not activate the MYC signature by directly inducing expression of MYC itself. Importantly, it was previously reported that Menin can directly interact with MYC and that protein-protein interactions between Menin and MYC at E-boxes leads to RNA polymerase II and P-TEFb recruitment and transcriptional amplification (28). To assess whether Menin and MYC interact in EwS cells, we performed coimmunoprecipitation experiments (co-IP). As shown (Figure 5E and Supplemental Figure 3), MYC was pulled down by the Menin antibody in A673, CHLA10, and TC32 EwS cells but not in U2OS osteosarcoma cells. By contrast, the H3K4 methyltransferase complex proteins MLL2 and WDR5 coimmunoprecipitated with Menin in all cell lines. These findings confirm that, unlike its canonical trithorax complex partners, Menin interactions with MYC are context specific. Next, we reversed the IP antibodies to validate that MYC can also pull-down Menin. These studies confirmed interactions between MYC and Menin, along with its canonical binding partner MAX (Figure 5F). Conversely, MYC did not pull down either MLL2 or WDR5, revealing that Menin interactions with MYC occur outside of MLL-containing methyltransferase complexes. These data suggest that Menin-dependent activation of the MYC signature in EwS is mediated by Menin-MYC interactions. Whether this is orchestrated by selective recruitment of RNA polymerase II to MYC bound E-boxes, or other EwS-specific mechanisms, requires additional investigation.

Figure 5

Menin and MYC form a complex in EwS cells. (A) MYC transcript levels in control and MEN1-KO cells as determined by qPCR. (B) MYC raw counts from RNA-seq of A673 control and MEN1-KO tumors. (C) Western blot of MYC and Vinculin levels in A673 and TC32 control and MEN1-KO cells. (D) Western blot of MYC and Vinculin in tumors derived from A673 control and MEN1-KO cells. (E) Coimmunoprecipitations with an anti-Menin antibody were performed on A673, CHLA10, and U2OS nuclear extracts and immunoblotted for Menin, MYC, MLL2, and WDR5. (F) Coimmunoprecipitations with an anti-MYC antibody were performed on A673 and CHLA10 nuclear extracts and immunoblotted for MYC, Menin, MAX, MLL2, and WDR5. Statistical comparisons of MEN-KO to control cell lines in A and B were performed using a 1-way ANOVA with a Dunnett’s T3 test.

Pharmacologic inhibition of Menin downregulates expression of MYC targets. Studies of MEN1-KO cells identified Menin as a master regulator of MYC targets and other pro-oncogenic gene programs in EwS. We next sought to test whether pharmacologic inhibitors of Menin could in whole or in part reverse this activity. For these investigations, we utilized VTP50469, the preclinical precursor of revumenib that was recently FDA approved for treatment of MLLr and NPM1m leukemia. Importantly, exposing EwS cells to VTP50469 had no effect on Menin transcript (Supplemental Figure 4A) or protein expression (Supplemental Figure 4B), allowing us to compare the effect of pharmacologic loss of function to the effects of protein loss. Consistent with genetic KO, pharmacologic inhibition of Menin had no effect on EwS cell proliferation in vitro (Figure 6A). However, unlike MEN1-KO cells, VTP50469-treated cells did not exhibit augmented invasion in 3D collagen (Figure 6, B and C), indicating that Menin-inhibition using a Menin-MLL interaction inhibitor does not completely phenocopy the effects of Menin depletion.

Figure 6

Menin inhibition regulates key tumorigenic pathways. (A) Incucyte proliferation assay plotting percent confluence for A673 and TC32 cells treated with 0.1% DMSO or 10 μM VTP50469 starting at time 0. Representative of n = 3–4. (B and C) Invasion of cells from a sphere of A673 cells embedded in rat tail collagen and treated with 0.1% DMSO or 10 μM VTP50469 for 5 days. (B) Representative phase and phalloidin (red)/DAPI (blue) stained images are shown. Scale bars: 100 μm. (C) Plot of circularity quantified from replicate spheroids treated as in B. (D) A673 and TC32 cells were treated with 0.1% DMSO or 10 μM VTP50469 for 72 hours, and genes that were significantly downregulated in VTP50469-treated cells (P_adj < 0.05) were overlapped with genes downregulated in MEN1-KO cells. The statistical significance of the overlap in the Venn diagrams was calculated using a Fisher’s exact test. The top 5 enriched Hallmark pathways for the overlapping downregulated genes in each cell line are shown along with the overlap/total genes for each pathway. The green arrows indicate pathways that were reactivated in A673 MEN1-KO tumors (Figure 4D). (E) MYC transcript levels in A673, CHLA10, and TC32 cells treated with 0.1% DMSO or 10 μM VTP50469 for 72 hours as determined by qPCR. (F) Western of MYC and Vinculin in cells treated with 0.1% DMSO or 10 μM VTP50469 for 72 hours. (G) Coimmunoprecipitations with an anti-Menin antibody were performed on nuclear extracts from A673 and CHLA10 cells treated with 0.1% DMSO or 10 μM VTP50469 for 72 hours and immunoblotted for Menin, MYC, MLL2, and WDR5. Statistical comparisons in C and E were performed using a Welch’s t test with a 2-tailed P value at 95% CI.

We next evaluated the effects of VTP50469 on gene transcription. Despite having no obvious effects on in vitro cell phenotypes, Menin inhibition led to up- or downregulation of thousands of genes (Supplemental Figure 4, C and D, and Supplemental Table 3). Though highly significant differences were observed between 2 different EwS cell lines, genes involved in cell signaling and cell cycle regulation were commonly overrepresented among downregulated genes (Supplemental Figure 4E), while metabolic and developmental regulators were enriched in upregulated genes (Supplemental Figure 4F). Among cell line–specific changes (Supplemental Figure 4, G–J), we noted profound downregulation of ribosome biogenesis genes in A673 cells and upregulated cell migration pathways in TC32 cells. To identify overlapping transcriptional effects of genetic and pharmacologic Menin inhibition strategies, we compared VTP50469-treated to MEN1-KO cells. Intriguingly, like Menin depletion, VTP50469 significantly suppressed expression of MYC target genes (Figure 6D). Moreover, this effect of VTP50469 occurred without any significant changes to MYC transcript or protein (Figure 6, E and F). Thus, we used co-IP studies to test if the drug inhibited interactions between Menin and MYC. As expected, and consistent with its mechanism of action as a Menin-MLL interaction inhibitor, VTP50469 dramatically and reproducibly reduced Menin-dependent pull down of both MLL2 and WDR5 proteins (Figure 6G). Significantly, VTP50469-treated cells also showed diminished interactions between Menin and MYC (Figure 6G).

Thus, pharmacologic disruption of Menin-MLL interactions in EwS cells leads to downregulated expression of pro-oncogenic MYC target genes, and this is associated with partial disruption of protein-protein interactions between Menin and MYC.

Menin inhibition reduces metastatic colonization in vivo. Our studies of MEN1-KO cells demonstrated that Menin is required for successful metastatic colonization of EwS tumor cells in vivo. In addition, transcriptomic profiling of MEN1-KO cells and tumors identified amplification of MYC targets as a likely promoter of Menin-dependent oncogenic states. Given our findings that VTP50469 disrupts Menin-MYC interactions and blunts expression of MYC target genes in vitro, we sought to test if the drug could be leveraged to block tumor growth or colonization in vivo. A673 and TC32 tumors were established in the paratibial regions of NSG mice, and the effect of VTP50469 on primary tumor growth was determined after 3–4 weeks. As shown in Figure 7, A and B, Menin inhibition did not alter growth of A673 xenografts. A statistically significant, albeit modest, decrease in the size of TC32 tumors was detected (Figure 7B). These studies of orthotopic EwS xenografts are consistent with published data from s.c. models, which showed heterogeneity of response between models (34). To test tumor colonization, we designed a model of disseminated micrometastatic disease that would allow comparison of metastatic outgrowth over time in control and VTP50469-treated NSG mice (Figure 7C). GFP-luciferase–labeled A673 cells were injected by tail vein, and IVIS imaging 1 hour after cell injection confirmed that tumor cells had successfully entered the systemic circulation. One week after cell injections, no tumor was detectable by IVIS in any mice, and animals were randomized to control or treatment groups. Inhibitor-treated mice were fed 0.1% VTP50469 chow (Syndax Pharmaceuticals Inc.), a dose that effectively inhibits Menin in mouse models of MLLr leukemia (18). IVIS imaging studies were performed weekly, and these documented near-uniform engraftment and outgrowth of tumors in control mice. By day 33 all control mice had evidence of disseminated disease (Figure 7, D–F). By contrast, tumor formation was both delayed and highly variable in VTP50469-treated mice (Figure 7D). Half of the drug-exposed mice showed no evidence of tumor by IVIS imaging 40 days after cell injection (Figure 7E). This difference in tumor colonization between the groups led to significantly improved survival of drug-exposed mice (Figure 7G). On day 40, all but 4 treated mice were euthanized for necropsy. In the 4 surviving mice, the drug was stopped and mice were followed for an additional 40 days — or humane endpoint, if it came first. Then, necropsies were performed.

Figure 7

Menin inhibition reduces colonization in EwS xenografts. (A) Photos of primary tumors in leg. In total, 5 × 10⁵ A673 or TC32 GFP-Luciferase cells were injected paratibially into 10 female NSG mice and 3 days (A673) or 5 days (TC32) after injection, the mice were randomized to either regular (Control) or 0.1% VTP50469 (~120–180 mg/kg/day) mouse chow. (B) The volume of each tumor was measured and plotted for both cell lines. (C) Diagram of mouse experiment. In total, 5 × 10⁴ A673 GFP-Luciferase cells were injected into the tail veins of 30 NSG mice (male and female), and 7 days later, mice were randomized to regular (Control) or 0.1% VTP50469 (~120–180 mg/kg/day) mouse chow. Four VTP50469-treated mice were taken off drug at study end and followed for up to 41 days as described in text. (D) Spider plots of IVIS signal for each control and VTP50469-treated mouse. (E) Plot of time to engraftment with engraftment set at an IVIS signal of 1 × 10⁸. (F) Luminescence of all mice at 33 days after injection. Sex of each mouse is indicated (M, male; F, female). (G) Survival plot comparing the control and VTP50469-treated mice. (H) Images of representative livers from the control and VTP50469-treated mice are shown. Scale bar: 1 cm. The number of macroscopic tumors in the livers was counted and plotted for each mouse. (I) Spider plot of IVIS signals from 4 mice taken off VTP50469 and followed. The liver of one of the released mice (asterisk in J) is shown. Statistical comparisons were performed using a Welch’s t test with a 2-tailed P value at 95% CI in B and H and a Logrank (Mantel-Cox) test to calculate the P value in E and G.

Necropsies of the mice euthanized on day 40 showed that tumors had mainly engrafted in the liver, consistent with prior studies of EwS in NSG models. Significantly, whereas multiple macroscopic tumors were clearly visible in the livers of all control mice, tumor burden was markedly reduced in VTP50469-treated mice, and in 6 of 16 mice, no tumors could be identified (Figure 7H). Of the 4 mice that were observed after stopping drug, 2 failed to develop tumors after 40 days of drug-free follow-up (Figure 7I). Conversely, 2 mice developed liver masses, and, in both mice, an increase in IVIS signal was evident at the time of drug cessation, presaging formation of macroscopic tumors (Figure 7I). Intriguingly, in one of the mice that developed tumors off drug, necropsy revealed that the liver contained a single very large tumor in addition to multiple much smaller tumors (Figure 7J). We speculate that these smaller tumors emerged only after VTP50469 was stopped, supporting a role for the Menin inhibitor in preventing outgrowth of micrometastatic foci.

To test if the inhibitory effects of VTP50469 on metastatic colonization would extend to an additional model, we repeated the experiment with TC32 cells (Supplemental Figure 5A). In this experiment, we followed half of the mice (N = 10) for 25 days after drug removal to monitor tumor outgrowth after treatment cessation. As with A673 cells, metastatic colonization of disseminated TC32 cells was inhibited by VTP50469, and macroscopic tumor burden in livers after 6 weeks of drug treatment was significantly reduced (Supplemental Figure 5, B and C). Intriguingly, the effect of the drug was much more significant in male compared with female mice, largely because colonization of TC32 cells in control mice was more rapid and efficient in males than females (Supplemental Figure 5C). This was not observed with A673 cells (Supplemental Figure 5D), and the reason for this sex discrepancy in the TC32 model is unknown. Both A673 and TC32 cell lines were originally derived from female patients with localized tumors. Regardless of mouse sex, removal of VTP50469 led to rapid outgrowth of macroscopic tumors in all mice within 25 days and livers of both male and female mice were nearly completely replaced by tumor (Supplemental Figure 5E).

Thus, pharmacologic inhibition of Menin using a Menin-MLL inhibitor reproducibly inhibited EwS tumor colonization. In most instances, withdrawing the drug resulted in rapid tumor outgrowth, supporting the hypothesis that inhibiting Menin activity reversibly suppresses the capacity of disseminated EwS cells to form tumors in metastatic niches.

Finally, we profiled transcriptomes of liver tumors that were harvested from control and VTP50469-treated mice to define significant differences in gene expression that were induced by exposure to the drug in vivo and to compare this with effects of the drug in vitro. Very limited changes to gene expression were observed in VTP50469-treated A673 tumors (71 upregulated and 139 downregulated transcripts, P_adj < 0.05), and these genes were largely nonoverlapping with in vitro targets (Supplemental Table 4). Only 32 of 71 upregulated and 37 of 139 downregulated genes in VTP50469-treated tumors were similarly altered by the drug in vitro. Effects of VTP50469 on TC32 tumors were much more profound. Nearly 1,800 genes were reproducibly upregulated, and over 1,300 genes were downregulated (P_adj < 0.05) in VTP50469-treated tumors (Supplemental Table 4). However, there was again virtually no overlap between VTP50469-modulated genes in vitro and in vivo (0 of 1,790 upregulated, 114 of 1,317 downregulated). We speculate that the marked differences between in vitro and in vivo effects of VTP50469 on gene expression are most likely due to (a) the prolonged period of in vivo exposure compared with 72 hours in vitro and/or (b) the probability that, at the time of necropsy, drug treated tumors had successfully established programs to escape the effect of Menin inhibition, which are most evident early in colonization. Nevertheless, the p53 pathway was the most highly enriched Hallmark among upregulated genes in TC32 tumors in vivo (Supplemental Table 4). This raises the possibility that, in the context of p53 WT EwS cells and an in vivo TME, Menin inhibition might activate the p53 tumor suppressor pathway. Additional studies will be needed to address this hypothesis.

Despite a robust and most often complete clinical response to primary therapy, patients with both localized and metastatic EwS often experience tumor relapse months or years after their original diagnosis (24). Metastatic relapse remains the primary cause of EwS patient death, yet very little is understood about the mechanisms that underlie this enduring clinical problem, and investigational agents have thus far offered no improvement in outcomes (35). Metastasis is a complex, multistage process that begins with invasion of cells beyond the confines of a primary tumor, continues with transit and survival of cells through the circulation, and culminates with engraftment and outgrowth of tumors at metastatic sites (36). Outgrowth of microscopic tumor foci into macroscopic metastases, or colonization, can take years and is the final and most inefficient step in the metastatic cascade (37). In the current work, we have identified the scaffolding protein Menin as a key driver of EwS metastasis. Both genetic and pharmacologic loss-of-function approaches revealed that, despite having little effect on cell phenotypes in vitro, inhibiting Menin activity in EwS cells significantly impairs their capacity to metastasize in vivo. Moreover, this was found to be primarily mediated by a defect in tumor colonization.

To robustly study the biologic programs that drive cancer metastasis, and to study inhibition strategies, models of tumor metastasis should ideally study the entire process from growth of a local tumor at an orthotopic site to successful outgrowth of the tumor in physiologically relevant distant organs. Studies of metastasis in EwS are challenged by the absence of syngeneic mouse models and a reliance on xenografts in immune deficient mice (38). In addition, metastasis of orthotopic EwS xenografts from bones to lungs of recipient mice is neither rapid nor efficient enough to be feasibly studied without amputation of the tumor-bearing limb (39, 40). Thus, we elected to use renal subcapsule models to study the first stages of spontaneous metastasis, as previously described (41). Tail vein models were added to isolate and study the final stages of the metastatic cascade, engraftment, and colonization. When analyzed together, the results of these 2 models identified that defects in colonization are likely to be the primary mediators of reduced metastatic competence in Menin-loss of function cells. Though this conclusion is supported by our data, we recognize that a major disadvantage of these EwS xenograft models is that cells preferentially metastasize to livers, an uncommon metastatic site in humans. In addition, the immune system is emerging as a key determinant of the response of both leukemias and solid tumors to genetic and pharmacologic inhibition of Menin (42, 43). Thus, it will be important that the potential contribution of the immune system be considered in future in vivo studies of Menin inhibitors. Importantly, Daley et al. recently showed that EwS tumor xenografts readily engraft in the paratibial space and reliably metastasize to the lungs of humanized NSG mice. In addition, lung metastases can be histologically detected within 4 weeks of cell inoculation (44). Future studies of Menin inhibitors in this model of spontaneous metastasis in an immune reconstituted TME will provide additional clarity regarding the mechanisms by which Menin promotes EwS progression.

In both normal development and in cancer, Menin functions to activate and repress gene transcription via its highly context-specific interactions with diverse partner proteins, including transcription factors and chromatin modifying enzymes (2, 22). The most well-studied and ubiquitous interaction partners of Menin are the MLL1 and MLL2 histone methyltransferases, and our studies show that MLL2 robustly interacts with Menin in EwS cells. Disruption of Menin-MLL interactions can be achieved by pharmacologic inhibitors of Menin that were developed for treatment of MLLr leukemias (21). Here, we have shown that VTP50469 broadly alters gene expression profiles of EwS cells, and this is associated with disruption of Menin-MLL2 interactions. Intriguingly, however, genes that were most reproducibly downregulated by genetic or pharmacologic inhibition of Menin in EwS cells were MYC target genes. In addition, we found that Menin forms a complex with MYC in EwS and that Menin-MYC interactions are partially disrupted by VTP50469. These data, combined with the more extensive changes to cell phenotype and gene expression that we identified in MEN1-KO compared with VTP50469-treated cells, lead us to conclude that Menin contributes to the pathogenesis of EwS via mechanisms that extend beyond MLL-dependent gene regulation. Furthermore, RNA-seq studies of VTP50469-treated A673 tumors revealed only subtle changes to gene expression, which contrasted with the major transcriptional rewiring that was apparent in A673 MEN1-KO tumors. We conclude that interactions with proteins beyond MLL contributed to the severe colonization defect that was reproducibly observed with all MEN1-KO cell line models. Significantly, Menin inhibitors have shown activity in preclinical models of other solid tumors (45–49), and clinical trials of these drugs in cancers beyond leukemia are now underway (https://clinicaltrials.gov/; NCT06655246 [gastrointestinal stromal tumors], NCT05731947 [colorectal cancer and solid tumors]). It remains to be elucidated whether the antitumor effects of Menin inhibitors in solid tumors are mediated in whole or in part by disrupting Menin interactions with MLL proteins or if direct or allosteric effects on other interaction partners, including MYC, are important. Given its broad and context-dependent partners and functions in normal organogenesis and development, we propose that the mechanisms by which Menin inhibitors impede oncogenic states in cancer will be, at least in part, tumor-type dependent.

In addition to downregulation of MYC signature genes, our studies revealed that genetic loss or pharmacologic inhibition of Menin led to upregulation of developmental, hypoxia, and stress-response programs. Notably, the specific genes and pathways that were suppressed by Menin expression and activity differed between A673 and TC32 cells, again demonstrating the highly cell context–dependent nature of Menin-regulated programs. Nevertheless, genes that control EMT were overrepresented among Menin-repressed genes in both cell lines and in both genetic and pharmacologic loss-of-function studies. These findings in a sarcoma are consistent with the critical role for Menin in regulating development of mesoderm and neural crest tissues during organogenesis and mesenchymal lineage differentiation (3, 5–8). Importantly, overexpression of Menin suppresses terminal differentiation of osteoblasts, and this is mediated by its antagonism of the JUND transcription factor (6). JUND interacts with Menin in the same binding pocket as MLL, and this interaction blocks JNK/ERK2-mediated phosphorylation of JUND (50, 51). Given that Menin-MLL inhibitors also disrupt Menin-JUND interactions, we hypothesize that the biologic and transcriptomic effects of Menin loss of function on mesenchymal gene programs in EwS cells may be, in part, secondary to effects on JUND activity. In addition, EWS-ETS fusion proteins are the oncogenic drivers of EwS and are required for tumor initiation and maintenance (23). EWS-ETS proteins directly and indirectly repress expression of mesenchymal differentiation genes, and loss of this repressive activity promotes migratory and metastatic cell phenotypes (30, 52–55). Thus, it is possible that Menin also promotes metastatic colonization of EwS cells by modifying the transcriptional activity of the EWS-ETS driver. It is also possible that differences in the EWS-ETS driver oncogene or other secondary mutations may be important for Menin interactions and Menin function in different EwS tumors. Both cell lines used in this study express the most common type I EWS-FLI1 fusion. However, differences in their p53, STAG2, and/or other mutational profiles may have contributed to the marked intertumoral differences that we observed in their transcriptional responses to Menin inhibition. It will be important that future preclinical studies of Menin inhibitors test the response of a broader diversity of EwS models to identify potential predictive biomarkers of response that could inform future clinical trial designs.

In summary, these studies identify Menin as a master regulator of gene transcription in EwS and a critical mediator of tumor metastasis. Our preclinical studies with the Menin inhibitor VTP50469 (revumenib) suggest that Menin-MLL inhibitors warrant investigation as potentially novel agents for patients with high-risk EwS. We propose that Menin inhibition could be leveraged to delay or prevent metastatic relapse in patients who are suspected to harbor subclinical residual disease at the end of primary therapy. Additional studies are now needed to determine if and how other agents could be combined with Menin inhibitors to eliminate rather than suppress growth of residual disease.

Sex as a biological variable

A mix of male and female mice were used in the in vivo experiments described below to control for sex differences in tumor engraftment and metastasis. In VTP50469 studies, mice were randomized to treatment groups, stratifying to include both male and female mice. The sex of each mouse is indicated in the figures.

Cell culture

A673, TC32, CHLA10, and U2OS cell lines were obtained from ATCC or COG. A673 and TC32 EwS cell lines were grown in RPMI (Gibco 11875093) with 10% FBS (Atlas Biologicals EF-0500-A) and 1% L-glutamine (Gibco 25030081), CHLA10 EwS cell line was grown in IMDM (Gibco 12440053) with 20% FBS, 1% L-glutamine and 1X Insulin-Transferrin-Selenium-Ethanolamine (ITS-X, Gibco 51500056), and U2OS (osteosarcoma) cells were grown in McCoy’s 5A (Gibco 16600082) with 10% FBS, 1% L-glutamine at 37°C with 5% CO₂. Cells were washed with Dulbecco’s phosphate-buffered saline (DPBS, Gibco 14040117) and passaged using 0.05% Trypsin-EDTA (Gibco 25300120). The CRISPR/Cas9 system was used to knock out Menin using guide 1 (gMEN1 1) to exon 2 (GCTGCGCTCCATCGACGACG) or guide 5 (gMEN1 5) to exon 3 (CAAATTGGACAGCTCCGGTG) and a nontargeting guide (gControl, GGTGGACATCCCGGAGACCC) for the control using the LentiCRISPRv2 vector (Addgene plasmid 52961). A673 and TC32 cells were transduced with each CRISPR lentivirus, transduced cells selected with puromycin, single-cell clones isolated and verified for loss of Menin protein by Western blot. The following CRISPR clones were generated with the last number in the name indicating the clone number (A673 gControl 1, A673 gMEN1 1-3, A673 gMEN1 5-1, gMEN1 5-2, TC32 gControl 3, TC32 gMEN1 1-5, TC32 gMEN1 5-5), and all were grown in the same conditions as the parental cell line. For animal experiments, cell lines were transduced with a GFP-Luciferase vector (pj01668-16revised-egfp-ffluc_ephiv7). To inhibit Menin, cells were treated with 0.1% DMSO (Vehicle) or 10 μM VTP50469 (MedChemExpress HY-114162) for 72 hours. All cell lines used in this study were verified by STR profiling and confirmed to be Mycoplasma negative using the IDEXX CellCheck 9 Plus service.

qPCR assay

Cells were trypsinized, pelleted, washed with ice-cold DPBS (Gibco 14040117) and total RNA (DNA-free) purified using the QIAshredder (Qiagen 79656), RNeasy Mini Kit (Qiagen 74106), and RNase-Free DNase Set (Qiagen 79254). cDNA was generated with iScript cDNA Synthesis Kit (Bio-Rad 1708891). qPCR reactions were done using iTaq Universal SYBR Green Supermix (Bio-Rad 1725122) with primers for MYC (TACAACACCCGAGCAAGGAC and GAGGCTGCTGGTTTTCCACT), EEF1A1 (CGTTACAACGGAAGTAAAATC and CAGGATAATCACCTGAGC), and 18S (GCAATTATTCCCCATGAACGA and GGCCTCACTAAACCATCCAAT) or TaqMan Fast Universal PCR Master Mix (Applied Biosystems 4352042) with primers for MEN1 (Hs00365720_m1), B2M (Hs_00984230_m1), and 18S (Hs_03003631_s1) and run on the Roche LightCycler 480 II. MYC crossing point (Cp) values were normalized to the geomean of the control primers Cp values (EEF1A1 and 18S), and MEN1 Cp values were normalized to the geomean of the control primers Cp values (B2M and 18S) and plotted as 2^ΔCp for Expression or relative to Cp value of the control cell sample normalized to the geomean of the control primers Cp values for Relative Expression (2^ΔΔCp).

Western blot analysis

Cells were trypsinized, pelleted, and washed with ice-cold DPBS (Gibco 14040117), and whole cell extracts were generated with a 10-minute incubation on ice in RIPA buffer (Thermo Fisher 89900) with protease inhibitors (Roche 04 693 116 001) and phosphatase inhibitors (Roche 04 906 837 001), and sonicated for 10 × 30 second cycles in the Bioruptor Pico (Diagenode), and debris were removed by centrifuging for 5 minutes at 16,000g. Protein extracts were quantitated using the Bio-Rad Protein Assay Kit II (catalog 5000002) and 30 μg of total protein was loaded on 4–15% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad 4561084 or 4561086) with a Tris-Glycine-SDS buffer system (Bio-Rad 1610732) using Precision Plus Protein Dual Color Standards (Bio-Rad 161037) for molecular weight estimation. Proteins were transferred to a nitrocellulose membrane (Bio-Rad 1620115) using the Bio-Rad Mini Trans-Blot Electrophoretic Transfer Cell with Tris-Glycine buffer system (Bio-Rad 1610734). Membranes were blocked with Intercept (TBS) Blocking Buffer (LICORbio 927-60001) and probed with the primary antibodies GAPDH Rabbit mAb (Cell Signaling 2118), MAX Rabbit pAb (Cell Signaling 4739), Menin Goat pAb (Bethyl A300-106A), MLL1-C (C-terminal) Rabbit pAb (Bethyl A300-374A), MLL2-C (C-terminal), Rabbit mAb (Cell Signaling 63735), MYC Rabbit mAb (Cell Signaling 18583), Vinculin Rabbit mAb (Cell Signaling 13901), or WDR5 Rabbit pAb (Bethyl A302-430A) followed by the secondary antibody Goat anti-Rabbit 800CW (LICORbio 926-32211), Donkey anti-Rabbit 800CW (LICORbio 926-32213), or Donkey anti-Goat 680RD (LICORbio 926-68074). All antibodies were diluted according to the manufacturer recommendations in Intercept (TBS) Blocking Buffer with 0.1% Tween-20. Membranes were scanned with the Odyssey (LICORbio) in the 700 and 800 nm channel.

Mouse models of tumor growth and metastasis

S.c. mouse model. Cells were prepared by trypsinizing, washing with DPBS, resuspending in DPBS at 1 × 10⁶ cells/50 μL and diluting 2-fold with 50 μL of Matrigel matrix (Corning 354234). Cells were injected into the flank region on the right or left side just below the skin layer in a 7-week-old NSG mouse (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) from Jackson Laboratory (strain no. 005557). Three to 5 mice were used for each model. Tumors were measured 2–5 times per week with calipers in 2 dimensions, the volume calculated (V = [L × W²]/2) was plotted over time for each mouse. Once the first mouse reached endpoint, all mice were euthanized at the same time.

Tail vein mouse model of tumor colonization. Cells transduced with the GFP-Luciferase vector were prepared by trypsinizing, washing with DPBS, and resuspending in DPBS at 1 × 10⁶ cells/100 μL. Cells were injected into the tail vein of a 7-week-old NSG mouse. Five mice were used for each cell line. Imaging with the In Vivo Imaging System (IVIS) was performed at 1 hour to confirm successful injection and then every 3–4 days until the first mouse reached endpoint and then all mice were euthanized at the same time.

Subrenal capsule spontaneous metastasis mouse model. Cells transduced with the GFP-Luciferase vector were prepared by trypsinizing, washing with DPBS, resuspending in DPBS at 2 × 10⁵ cells/5 μL and diluting 2-fold with 5 μL of Matrigel matrix. Live ultrasound imaging was used to insert the needle through the skin into the back muscle and into the subrenal capsule space to inject cells into a 7-week-old NSG mouse (33). Five mice were used for each cell line. Tumor size in kidney was monitored by ultrasound, and IVIS imaging was done 2× per week. Once the tumor in the kidney reached about half the size of the kidney in 1 mouse, all the mice were euthanized at once.

In vivo studies of VTP50469

Local tumor growth. To assess the effect of VTP50469 treatment on local tumor growth, we used an orthotopic model, as previously described (44, 56). Cells transduced with the GFP-Luciferase vector were prepared by trypsinizing, washing with DPBS, resuspending in DPBS at 5 × 10⁵ cells/25 μL, diluting 2-fold with growth factor reduced Geltrex (Gibco A1413201). Cells were injected into the paratibial regions of ten 7-week-old female NSG mice. Tumor engraftment was monitored by IVIS and when signal reached greater than 1 × 10⁸, mice were randomized to control or 0.1% VTP50469 mouse chow (~120–180 mg/kg/day, provided by Syndax Pharmaceuticals Inc. and prepared by Inotiv, catalog TD.170873). All mice were euthanized when the first mouse in each model reached humane endpoint (25 days, A673; 32 days, TC32) and tumor volumes measured.

DTC model. Tail vein mouse models as above were used to assess the impact of VTP50469 on engraftment and colonization of DTCs. Cells transduced with the GFP-Luciferase vector were prepared by trypsinizing, washing with DPBS, and resuspending in DPBS at 5 × 10⁴ cells/100 μL. Cells were injected into the tail veins of 30 NSG mice that were 7 weeks old, and IVIS imaging was performed after 1 hour to ensure presence of circulating tumor cells in the lungs of all mice. After 7 days, mice were randomized to control or 0.1% VTP50469 chow, as above. Tumor engraftment and outgrowth was monitored by serial IVIS imaging. Treatment continued until the last control mouse reached humane endpoint (A673), or all control mice had an IVIS signal greater than 1 × 10¹⁰ (TC32), at which point most mice were euthanized and livers harvested for evaluation of macroscopic tumor burden. To assess the effect of drug withdrawal, a subset of VTP50469-treated mice (n = 4 A673; n = 10 TC32) were switched to control chow and monitored by IVIS imaging for an additional 3–6 weeks, at which point remaining mice were euthanized.

Tissue histology

Formalin-fixed, paraffin-embedded (FFPE) organs (liver and kidney) from the xenograft mouse experiments were sectioned, mounted on slides, deparaffinized, and stained with H&E. For immunofluorescence, following antigen retrieval, slides were blocked in blocking buffer (0.2% bovine serum albumin in PBS) for 1 hour, washed, incubated with 1:100 primary antibody, CD99 rabbit mAb (Biocare Medical CME392A), in blocking buffer for 1 hour and then washed, incubated with 1:200 secondary antibody, Goat anti-Rabbit Alexa Fluor 647 (Invitrogen A-21245), in blocking buffer for 1 hour with 1:2,000 DAPI (2.5 μg/mL, Invitrogen D1306); coverslips were mounted with ProLong Gold Antifade Mountant (Invitrogen P36934). Whole slides were scanned with ZEISS Axioscan 7 or imaged with Leica DMi8 Thunder Imager at 20X magnification.

mRNA-seq

RNA was purified from in vitro cultured cells, and from liver tumors harvested at necropsy in the tail vein models as above. Biological triplicate RNA samples from independent experiments or mice were sent to Novogene for polyA enrichment library preparation using NEBNext Ultra II RNA Library Prep kit and sequencing with the Illumina NovaSeq 6000 or NovaSeq X Plus Series sequencer utilizing a paired-end 150 bp (PE150) sequencing strategy. Sequencing quality was assessed using FastQC. Adapter sequences, low-quality bases, and short reads were removed with Trim Galore. Trimmed reads were aligned to the human reference genome (GRCh38) using STAR, and gene-level counts were generated with STAR (batches 1, 2, and 4) or Salmon (batch 3), depending on the dataset. Additional quality control was performed with MultiQC, and differential gene expression analysis was conducted using DESeq2. Genes with a P_adj < 0.05, without applying a fold-change threshold, were used to generate Venn diagrams with Vennerable. Hallmark gene set enrichment analysis of ranked gene sets was performed using Gene Set Enrichment Analysis (GSEA) (57, 58). Hallmark and GO-BP gene set enrichment of commonly regulated genes was assessed using MSigDB (59, 60) or Enrichr (61). Heatmaps of differentially expressed genes were generated using pheatmap. All data have been submitted to GEO (accession no. GSE338238).

Real-time cell proliferation assays

Cells were plated in a 96-well plate with 5,000 or 15,000 cells in 50 μL media per well and incubated overnight to adhere to plate. Media were added to 200 μL final, and DMSO or VTP50469 were added to a final concentration of 0.1% and 10 μM, respectively, for the Menin inhibition assays. DPBS was added to the empty wells and interspaces in the plate to keep plate humidified. The plate was loaded into the Incucyte SX5 Live-Cell Analysis System (Sartorius), and images were captured every 2 hours. Confluence was measured using the Incucyte software and plotted as a percentage over time.

3D-collagen invasion assays

Small spheroids (10–1,000 cells) were formed by trypsinizing cells, resuspending at 50,000 cells/mL, plating 4 mL/well in a nonadherent 6-well plate (Corning 3471) and incubating overnight. Collagen gels were formed by combining 1.8 mg/mL rat tail collagen I (Gibco A1048301) in 1X DMEM (Sigma-Aldrich D2429) and 8 mM acetic acid, neutralizing the mixture with 0.5 M NaOH and incubating on ice for 45–60 minutes to allow polymerization. A small flat circular underlay was formed by pipetting 25 μL of polymerized collagen mixture into each well of a 24-well plate warmed to 37°C. Spheroids embedded in collagen were formed by resuspending 1 mL of spheroids collected by centrifuging at 300g for 10 seconds in 100 μL of polymerized collagen mixture, pipetting 60 μL of spheroid-collagen mixture on top of overlay, incubating at 37°C for ~30 minutes to set gel, and adding appropriate culture media. For VTP50469-treated cells, 0.1% DMSO or 10 μM VTP50469 were added with the media after the gels were formed. Plates were incubated for 5 days; then, cells were fixed with 4% paraformaldehyde in PBS for 15 minutes, washed with PBS, permeabilized, and stained with 0.1% Triton-X100, 1:2,000 DAPI (2.5 μg/mL) and 1:800 Alexa Fluor 568 Phalloidin (400x, Invitrogen A12380) for 2 hours, before being washed with PBS and imaged in PBS on Leica DMi8 Thunder Imager at 10X magnification. Images were analyzed by ImageJ (NIH) to determine circularity. F-actin images were converted to 8-bit, the threshold was set using the Otsu method, and the perimeter, area, and shape descriptors (includes circularity) were selected in Set Measurements. Analyze Particles was used to generate outlines of spheres (set size range from 6,000 to infinity pixel2, circularity from 0 to 1, and show outlines), and then the circularity measurements for each sphere outline in the image was outputted. Circularity is calculated using the following equation, 4π(area/perimeter²). A perfect circle has a value of 1.0 and decreases as the object becomes less circular. Invasion increases the perimeter with minimal change in the area resulting in a decrease in circularity.

Coimmunoprecipitations

Menin (rabbit pAb, Bethyl Labs A300-105A); MYC (mouse mAb, Santa Cruz SC-42); Rabbit IgG, polyclonal - Isotype Control (Abcam ab37415); or Mouse IgG1, kappa monoclonal - Isotype Control (Abcam ab18443) was cross-linked to Protein A Dynabeads (Invitrogen 10001D) with 20 mM dimethyl pimelimidate dihydrochloride (Millipore Sigma D8388) at 8 μg human IgG/mg beads. Antibody-beads were resuspended at 30 mg/mL.

The nuclear extract was prepared using the Nuclear Complex Co-IP Kit (Active Motif 54001). Briefly, cells were trypsinized, washed in PBS with Phosphatase Inhibitors, resuspended in Hypotonic Buffer, and lysed with Detergent; nuclei were pelleted by centrifuging at 10,000g and were lysed in Digestion Buffer with Protease Inhibitor Cocktail and 0.5 mM PMSF, before shearing chromatin with the Enzymatic Shearing Cocktail. Enzymes were inactivated with 10 mM EDTA, debris was removed by centrifuging at 12,000g, and the supernatant was collected for the nuclear extract. Nuclear extracts were quantitated using the Bio-Rad Protein Assay Kit II.

For each immunoprecipitation, 1–2 mg of nuclear extract were incubated with 0.25 mg beads (2 μg Ab) per mg of nuclear extract in IP buffer (20 mM HEPES•KOH [pH 7.5], 150 mM NaCl, 0.5% Igepal, protease inhibitors [Roche 04 693 116 001] and phosphatase inhibitors [Roche 04 906 837 001]). Immunoprecipitations were incubated on rotator at 4°C for 2 hours and washed 3X with IP Buffer (20 mM HEPES•KOH [pH 7.5], 150 mM NaCl, 0.5% Igepal, protease inhibitors [Roche 04 693 116 001] and phosphatase inhibitors [Roche 04 906 837 001]), and the protein was eluted with Laemmli Sample Buffer (Bio-Rad 1610747) at room temperature for 10 minutes. Eluted samples and 5% and 1% of input (nuclear extract) prepared in Laemmli Sample Buffer were heated at 95°C for 5 minutes. Input and 50% of the eluted sample were analyzed by Western blots as described above.

Statistics

For cell culture experiments, graphical data are plotted as the mean with a SD of at least 3 biological replicates, and a Welch’s t test with a 2-tailed P value at 95% CI was used to compare cell lines or treatment. For comparison of 3 or more groups, a 1-way ANOVA with a Dunnett’s T3 multiple comparison test for unequal variances was used to generate pairwise P values. For Venn diagrams, the GeneOverlap R package was used to calculate the P value of the overlap using a Fisher’s exact test. For s.c. mouse experiments, the AUC of tumor volume measurements over time was calculated for each mouse, and the P values were calculated using a 1-way ANOVA with a Dunnett’s T3 multiple comparison test for unequal variances to compare area under the curve values of MEN1-KO compared with control xenografts. Time to engraftment and survival plots in DTC mouse experiments used a Logrank (Mantel-Cox) test to calculate the P value. The number of tumors in the DTC mouse experiments were plotted as the mean ± SD and a Welch’s t test with a 2-tailed P value at 95% CI to compare treated to untreated mice. P ≤ 0.05 was considered significant.

Study approval

All experimental procedures adhere to the Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011) and were approved by Seattle Children’s Research Institute IACUC protocol no. ACUC00635. The institute is fully AAALAC accredited and has a Public Health Service approved Animal Welfare Assurance.

Data availability

Sequencing data generated by this study have been deposited to NCBI Gene Expression Omnibus database (accession no. GSE338238). Values for all data points in graphs are reported in the Supporting Data Values file.