MEN1 regulates lipid peroxidation and ferroptosis in lung cancer cells
To determine whether MEN1 regulated ferroptosis, the short-hairpin RNA-mediated MEN1 knockdown (shMEN1) and lentiviral-mediated MEN1 overexpression were examined for potential effects on ferroptosis in lung cancer cells. shMEN1 was found to increase the protein expression of glutathione peroxidase 4 (GPX4), a negative regulatory factor of ferroptosis12, in A549 cells (Extended Data Fig. 1a), whereas overexpression of flag-tagged wild-type (WT) MEN1 (oeMEN1) reduced the GPX4 expression in NCI-H446 cells (Extended Data Fig. 1b). In agreement with this result, the CRISPR/Cas9-mediated genetic knockout of MEN1 (KO) in NCI-H460 cells also yielded similar results and GPX4 upregulation was completely abrogated with MEN1 re-expression (rMEN1) in MEN1-deficient cells (Extended Data Fig. 1c). Notably, the knockdown of MEN1 effectively blocked ferroptosis inducers such as erastin, RSL3, and ROS inducer ROSUP-induced cytosolic ROS generation, as assayed by flow cytometry using the fluorescent probe DCFH-DA (Fig. 1a,b). In contrast, overexpression of MEN1 substantially increased the ROS accumulation induced by erastin or ML-210 (a GPX4 inhibitor) treatment (Extended Data Fig. 1d). Lipid peroxide levels were further compared using lipid peroxidation sensor C11-BODIPY581/591 and MEN1 deficiency was suppressed, whereas MEN1 overexpression promoted lipid peroxide production (Fig. 1c-f and Extended Data Fig. 1e,f). The accumulated effect of erastin and RSL3 on the lipid peroxidation was weaker with shMEN1 or MEN1-KO cells than with WT MEN1 control (Ctrl) cells (Fig. 1c-f), and reduced lipid peroxidation in MEN1-KO NCI-H460 cells was reversed by MEN1 re-expression (Fig. 1e,f). These results demonstrate the important role played by MEN1 plays in executing ferroptosis.
We next investigated the effects of MEN1 on the ferroptosis sensitivity of lung cancer cells and showed that MEN1 deficiency decreased the cell death of A549 and NCI-H460 cells to erastin and RSL3 treatments (Extended Data Fig. 1g,h). MEN1-deficient NCI-H460 cells were relatively insensitive to erastin than MEN1-WT cells, whereas the reduction of erastin-induced ferroptosis was effectively reversed when MEN1 was reintroduced into MEN1-KO cells (Fig. 1g). Subsequently, rMEN1 NCI-H460 cells were treated with erastin or RSL3 alone or in combination with inhibitors of differentially regulated cell death. As expected, erastin and RSL3-induced cell death could be completely rescued by the ferroptosis inhibitor ferrostatin-1 (Fer-1, a lipid peroxidation scavenger) or the iron chelator deferoxamine (DFO), only slightly by the apoptosis inhibitor Z-VAD-FMK (Z-VAD) and hardly by the necroptosis inhibitor necrostatin-1 (Nec-1) (Fig. 1h). These data suggest that MEN1 directly impacted the sensitivity of lung cancer cells to ferroptosis.
Finally, the potential role of MEN1 on the mitochondria was investigated, as this organelle displayed an aberrant morphology when cells undergoing ferroptosis1,13. Transmission electron microscopy revealed that erastin-treated cells did display apparent changes in mitochondrial morphology, such as the loss of structural integrity (Fig. 1i). Compared with MEN1-WT cells, the primary morphological features of MEN1-KO cells involved the mitochondria including dysmorphic large mitochondria, a disarranged structure of crista, a reduced mitochondrial number, and an increased mitochondrial area (Fig. 1i,j), suggesting that MEN1 is at least partially responsible for the maintenance of mitochondrial quality. Quantitative real-time polymerase chain reaction (qPCR) analysis revealed that MEN1 deficiency significantly inhibited erastin-induced mRNA expression of mitochondrial genes, including ribosomal protein L8 (RPL8), tetratricopeptide repeat domain 35 (TTC35), iron response element binding protein 2 (IREB2), and adenosine triphosphate (ATP) synthase F0 complex subunit C3 (ATP5G3) (Extended Data Fig. 1i,j), which play known roles in erastin-induced ferroptosis1. These results indicate that MEN1 is a critical contributor to ferroptosis.
MEN1 -modulated ferroptosis is required for suppressing lung tumor growth and metastasis
To understand the biological consequences of MEN1-mediated ferroptosis, we performed a series of functional experiments. The results in vitro revealed that the proliferative potential of NCI-H460 cells with MEN1 deficiency was faster than those of MEN1-WT cells (Fig. 2a). MEN1-KO significantly enhanced the migratory and invasive capability of NCI-H460 cells compared with MEN1-WT cells (Fig. 2b-d). Conversely, MEN1 overexpression retarded the cell proliferation and attenuated the clonogenic capacity, migration, and invasion of NCI-H446 cells, a small cell lung cancer cell characterized by high invasive and metastatic behaviors, when compared with empty vector-expressing cells (Extended Data Fig. 2a-d).
To further assess the effects of MEN1 on metastatic colonization of lung cancer cells, an experimental lung metastasis model was established using WT or KO NCI-H460 cells by injecting it into the tail vein of severe immunodeficient M-NSG mice (Fig. 2e). Although the incidence (WT, 4/4; KO, 4/4) of metastatic lung tumors formed by these cells was comparable, the extent and tumor burden of lung metastasis was dramatically increased for MEN1-KO cells compared with that of MEN1-WT cells (Fig. 2e,g). Immunohistochemistry (IHC) demonstrated that positive staining of the proliferation marker Ki67 and metastatic indicator HMGB1 was significantly increased in MEN1-KO tumors (Fig. 2f-h). Accordingly, MEN1 knockout markedly promoted the protein expression of metastasis-associated contributors, such as matrix metalloproteinase 2 (MMP2), MMP9, HMGB1, and epithelial-mesenchymal transition mediators, Snail, ZEB1, and vimentin, as detected by immunoblotting (Extended Data Fig. 2e). Consistent with in vitro results, MEN1-depleted tumors showed a significant suppression of ferroptosis as demonstrated by the augmentation of GPX4 protein expression and the reduction of COX2 expression (Extended Data Fig. 2e). Furthermore, the GPX4 colocalization with MMP9 was observed in these lung tumors, which was further increased by MEN1 deletion (Extended Data Fig. 2e,g), suggesting that MEN1 facilitates ferroptotic cell death and thus suppresses the metastatic progression of lung cancer.
To corroborate this finding in spontaneous tumors, we crossed mice expressing tamoxifen-inducible alveolar type 2 (AT2)-specific Cre recombinase (Sftpc-CreER) with mice harbouring conditional alleles of oncogenic KrasG12D (Kraslox−stop−lox(LSL)−G12D/+) to generate a LUAD mouse model (KrasLSL−G12D/+;Sftpc-CreER (hereafter referred to as KS)) in which KrasG12D expression was induced in AT2 cells, the predominant cell of origin for LUAD14–17. The other two genetically engineered mouse models (GEMMs), Men1flox/flox(f/f);Sftpc-CreER (hereafter referred to as MS) and KrasLSL−G12D/+;Men1f/f;Sftpc-CreER (hereafter referred to as KMS), were established using similar experimental procedures (Extended Data Fig. 2h). Lung tumors were initiated and induced by tamoxifen injection and examined for 8 weeks later, a time point when lung tumors predominate in the KMS mice (Fig. 2i,j). IHC results revealed that KS tumors exhibited strong SFTPC staining with accompanying low expression of menin; however, Men1 deletion in the KMS mice decreased SFTPC protein expression (Extended Data Fig. 2i). The lung weight and lung-to-body weight ratio were significantly elevated in KMS mice relative to KS mice (Extended Data Fig. 2j). Although both KS and KMS mice (but not WT control mice) develop primary lung tumors, the KrasLSL−G12D/+;Men1f/f;Sftpc-CreER mice readily undergo malignant progression to highly malignant neuroendocrine carcinomas following Men1 deletion (Fig. 2k). KrasG12D activation increased tumor burden, with augmented tumor numbers and elevated tumor sizes; these malignant characteristics were further intensified in KMS mice (Fig. 2k,l), suggesting that Men1 is a super LUAD suppressor. Consistently, tumor growth and metastatic potential were significantly enhanced in KMS tumors relative to KS tumors (Fig. 2k,m), demonstrating that Men1 deficiency accelerates Kras mutation-driven lung tumorigenesis and progression. Notably, RNA in situ hybridization showed low basal GPX4 expression in the WT mice, with moderately higher GPX4 levels in KS mice and a dramatically elevated expression in KMS mice (Fig. 2n,o). The ferroptosis was reduced in the lung tumors of KS or KMS mice as compared with that in their adjacent non-tumor tissues, and interestingly, Men1 loss led to a greater reduction in ferroptosis in KMS tumors than that in KS tumors (Fig. 2n,o). These findings in vivo underscore the importance of Men1-modulated ferroptosis in suppressing lung cancer initiation and progression.
MEN1 regulation of ferroptosis in lung cancer cells depends on CD44 variant isoforms
RNA sequencing (RNA-seq) on lung tissues were performed from WT, KS, and KMS mice at 6 weeks after inducting tamoxifen, an early time point in tumor development, to elucidate molecular mechanisms of MEN1 action. Lungs of each genotype displayed distinct transcriptional profiles, with 5,156 Men1-dependent differentially expressed genes (DEGs) (Extended Data Fig. 3a and Supplementary Table 1). To pinpoint pathways pivotal for tumor progression upon Men1 deletion, 1,139 Men1-dependent genes that were more highly activated in KMS tumors than in KS tumors were determined (Fig. 3a and Supplementary Table 2). Functional annotation of Men1-hyperactivated genes identified high enrichment (top 10) of tumorigenesis pathways (small cell lung cancer, hippo signaling pathway, and hedgehog signaling pathway) and metastasis events (extracellular matrix (ECM)-receptor interaction, cell adhesion molecules, and cytoskeleton proteins) (Fig. 3b). Gene set enrichment analysis (GESA) further revealed that two cellular components of gene ontology (GO) terms, ECM and neural cell adhesion molecule 1 (NCAM1), were distinctly enriched in KMS tumors (Fig. 3c).
Based on these analyses, abnormal ECM and cell adhesion activation were hypothesized to contribute to phenotypic alterations in lung cancer cells. Among the members of the cell adhesion family, we were interested in CD44 because its expression has been implicated in the development and progression of multiple cancers18. Here, the CD44 expression levels of human lung cancer samples were also assessed in different tumor/node/metastasis classification system (TNM) stages with paired adjacent non-tumor lung tissues. Although CD44 expression was significantly higher in lung cancer specimens than in adjacent non-tumor tissues, there were no significant differences between its expression and tumor stage, metastasis, tumor size or other clinicopathological characteristics (Extended Data Fig. 3b,c, and Supplementary Table 3). Moreover, PrognoScan analyses revealed that 11 of 30 publicly available RNA-seq data of human LUADs showed a negative correlation between high CD44 mRNA expression levels and overall survival (OS), and 19 of them showed that CD44 expression was not associated with OS of patients with LUAD (Supplementary Table 4). Consistently, among 15 non-small cell lung cancer (NSCLC) databases, only three displayed an anti-correlation between CD44 expression and OS, whereas 12 showed no significant correlation (Supplementary Table 4). Subsequently, we downloaded a transcriptome sequence dataset of 516 patients with LUAD that included detailed clinicopathological annotations with comprehensive mRNA splicing profiling and analyzed the relationship of CD44 variant isoforms with patient survival. These patients were regrouped into high- and low-percent spliced-in (PSI) groups based on the median PSI of the CD44 gene. Survival analysis found that high PSI of the CD44 gene (representing high levels of CD44 variant isoforms), but not its mRNA expression, was dramatically positively correlated with poor OS (Fig. 3d). Immunofluorescence (IF) staining indicated that the expression of CD44 variant exon 6 (CD44v6) in 115 clinical lung cancer samples was significantly increased compared with that in 36 adjacent non-tumor lung tissues (Supplementary information, Fig. S3d. e). Again, qPCR and immunoblotting showed that the protein expressions of the variable exons v3, v5, and v6-containing CD44 splicing isoforms, rather than the total CD44 expression, were positively correlated with lung cancer progression (Extended Data Fig. 3f and Supplementary Table 3). These results were also supported by comparing the CD44v3 and CD44v6 expressions between metastatic- and non-metastatic lung cancers via IF co-staining (Fig. 3e,f). These correlative analyses suggest that CD44 variants could be used as prognostic indicators and are more specific than total CD44 expression to evaluate lung cancer progression.
Given these findings, the MEN1 regulatory role in malignant suppression and ferroptosis of lung cancer cells is likely dependent on the CD44 variant isoforms. In agreement with this hypothesis, metastatic lung tumors from MEN1 KO M-NSG mice exhibited obvious increases in CD44v3 and CD44v6 protein expressions (Fig. 3g,h). A similar inducible effect of MEN1 insufficiency on CD44 variant generation was further confirmed in MEN1-deficient A549 and NCI-H460 cells, as determined by IF staining and flow cytometry assays, showing that MEN1 deficiency enhanced the CD44v6 generation induced by phorbol myristate acetate (PMA) as compared with their control WT counterparts (Fig. 3i, and Extended Data Fig. 3g,h). siRNA-mediated CD44v6 knockdown (siCD44v6, designated siV6) suppressed the invasive capability of A549 cells and, importantly, strongly attenuated the high invasiveness induced by knocking down MEN1 (Extended Data Fig. 3i,j). To corroborate this observation, we performed an anti-CD44v6 antibody-neutralizing assay. As shown in Fig. 3j, the invasive capacity of NCI-H460 cells was significantly diminished by the anti-CD44v6 blockade treatment; the invasion–inhibitory effect of anti-CD44v6 antibody neutralization was stronger in MEN1-KO cells than that observed in MEN1-WT cells. Our data strongly demonstrate that enhanced invasion ability of MEN1-deficient cells is caused at least partly by the increased CD44v6 accumulation mediated by MEN1 inactivation.
Notably, previous studies have shown that repression of CD44 variant expression is critical for lipid ROS generation19. To this end, we examined whether MEN1 influences cellular sensitivity to ferroptosis by inhibiting CD44 variant production. Immunoblotting and qPCR analyses revealed that knockdown of CD44v6 prominently reduced the MEN1 KO-induced GPX4 expression along with rescued the COX2 expression (Fig. 3k and Extended Data Fig. 3k). Importantly, the potentiated cell proliferation by MEN1 deletion was suppressed by siV6 transfection; more importantly, siV6 successfully restored the sensitivity of MEN1-KO NCI-H460 cells to erastin-induced ferroptosis analogous to that of siNC-treated MEN1-WT control cells (Fig. 3l). Similar results were obtained in shNC and shMEN1-A549 cells when ferroptosis was activated by erastin treatment, showing that siV6 transfection abrogated MEN1 deletion-mediated increase in colony formation (Extended Data Fig. 3l,m). We speculated that the suppressive effect of CD44v6 knockdown on lung cancer cells was caused by its intrinsic oxidant potential. A flow cytometry assay was performed using fluorescent probes DCFHDA and C11-BODIPY, respectively, to monitor the oxidant activity of CD44v6. As shown in Fig. 3m,n, cytosolic ROS and lipid peroxide levels were decreased in MEN1-deficient cells to levels approximately 50% lower than those of MEN1-WT cells; this reduction was reversed by siV6 transfection to the same levels as in siNC-treated MEN1-WT control (DMSO) cells. Remarkably, the negative effect of MEN1 deletion on erastin-induced cytosolic and lipid ROS accumulation was effectively counteracted by CD44v6 knockdown (Fig. 3m,n).
Collectively, MEN1 suppresses lung tumorigenesis and progression by maintaining CD44 variant-dependent physiological ferroptosis levels.
MEN1 prevents CD44 variant generation by repressing RNA alternative splicing
The CD44 gene comprises 10 constant exons (C1–C5 and C16–C19) and 9 clustered variable exons (v1–v10) and generates various variant isoforms, some being involved in tumor progression and metastasis (Extended Data Fig. 4a)20–24. Exon-specific reverse-transcription PCR (RT-PCR) analysis showed that the expression of nearly all CD44 variants (v3–v10) in the lung tissue of whole-body Men1-KO (Men1Δ/Δ) mice, which was established as our previously described10, was increased compared with that of WT Men1 (Men1f/f) mice (Extended Data Fig. 4b). A similar variant isoform profiling was observed in the M-NSG mice, in which MEN1-KO cell-derived metastatic lung tumors showed a marked increase in the expression abundance of CD44v3–v10 isoforms (Extended Data Fig. 4c). Immunoblotting analysis confirmed that protein expressions of the variant exon v3, v5, or v6-containing CD44 isoforms were elevated in the MEN1-KO NCI-H460 cells, both with and without PMA induction (Extended Data Fig. 4d).
We assumed that the overwhelming, MEN1 deficiency-dependent accumulation of CD44 variants is what initiates lung tumorigenesis and drives malignant progression. To test this hypothesis, the expression pattern of CD44 splicing isoforms in the lungs was first determined from WT, KS, and KMS mice. Although a small fraction of CD44 variant expression was also detected in WT mouse lungs, differential expressions on WT control mice versus KS or KMS mice were very apparent at 8 weeks after tamoxifen induction. Notably, the mRNA abundance of CD44 variant isoforms, specifically v3, v4, v5, v6, and v9, in the lung tissues of KMS mice was significantly higher than that in KS mice, in which lung tumors have lower menin expressions (Fig. 4a and Extended Data Fig. 4e). IF co-staining further validated that KMS mice had significantly more CD44v3 and CD44v6 expressions in lung tissues (Fig. 4b-d and Extended Data Fig. 4f). Moreover, in both KS and KMS mice, CD44v3 and CD44v6 isoforms appear to be expressed at higher levels in tumor regions than those in their non-tumor counterparts (Fig. 4b-d and Extended Data Fig. 4f). To confirm this observation, we isolated tumors (T) and adjacent non-tumor (N) tissues from lung tissue sections of KS and KMS mice using the laser capture microdissection (Fig. 4e). qPCR analysis showed that KS and KMS tumors more than doubled in CD44 variant expression compared with their adjacent non-tumor tissues. In Kras-mutant tumors, the abundance of mRNAs containing variant exons v3, v6, and v9 increased by > 10-fold (Fig. 4f). The strongest mRNA abundance was observed at the expression of v3, v5, v6, v8, and v9 isoforms (> 5-fold increase) in Men1-deficient Kras-mutant tumors (Fig. 4g). Unexpectedly, the abundance of CD44 variants was also higher in the adjacent non-tumor lung tissues of KS or KMS mice compared with that of normal lung tissues of WT control mice; these variant expressions in non-tumor lung tissue of KMS mice were dramatically higher than those of KS mice (Extended Data Fig. 4g). Expression changes in the normal lung of WT mice and non-tumor lung tissue of KS and KMS mice were associated with the severity degree of pathological progression (Fig. 4h, Extended Data Fig. 4h and Supplementary Table 5); because at 8 weeks of age, Men1-deficient Kras-mutant mice displayed numerous macroscopically visible tumors in a whole lung with premalignant lesions, whereas Kras-mutant mice revealed several sporadic tumors in a partial damaged lung (Fig. 4c,d, H&E staining). This strict correlation of increased splicing isoforms in lung tumors highlights that high levels of intratumoral CD44 variants are critical for tumorigenicity.
AS of the precursor messenger RNA (pre-mRNA) is a foremost mechanism that generates multiple RNA splicing isoforms derived from one same gene. We therefore focused on testing the possibility that augmentation of CD44 variant isoforms after MEN1 deficiency might be due to enhanced pre-mRNA splicing efficiency. The compound 5,6-dichlorobenzimidazole 1-β-D-ribofuranoside (DRB) arrests gene transcriptional progress from the initiation to the elongation phase, but it does not block the elongation of previously initiated transcripts already within the gene25. NCI-H460 cells were treated with DRB for different times and determined for unspliced pre-mRNA levels with primers specific for CD44 variant exons. RT-PCR analysis revealed that unspliced pre-mRNAs were progressively diminished due to being processed to mature mRNA variants by the splicing machinery during the DRB treatment (Fig. 4i); MEN1 deletion expedited the AS procession and thus generated abundant CD44 mRNAs containing variant exons v4–v5 and v6–v7 (Fig. 4i). To confirm that altered splicing efficiency upon MEN1 deletion correlated with an accumulation of CD44 variant isoforms, qPCR primer pairs were designed to dynamically monitor the intron-containing pre-mRNA (unspliced primers), and the mature mRNA (spliced primers) that can be stimulated by PMA treatment (Fig. 4j and Extended Data Fig. 4j). During the treatment of NCI-H460 cells with PMA, knocking out MEN1 resulted in a time-dependent increase in mRNAs containing variant exons v3–v4 and v6–v7 (Fig. 4k and Extended Data Fig. 4j). To determine pre-mRNA splicing, the ratio of mature mRNA to unspliced was defined as the splicing efficiency; the ratio of mature mRNA to unspliced was lowest at the earliest time points after PMA induction, suggesting that splicing has just initiated and is a slow process relative to transcription and increases over time during the mRNA production (Fig. 4l and Extended Data Fig. 4k). Strikingly, an obvious increase was observed in the splicing efficiency in MEN1-KO cells as compared to those in WT cells 6 h after PMA induction; the enhanced effect on splicing was attenuated by 12 h after PMA induction (Fig. 4l and Extended Data Fig. 4k). These experiments strongly demonstrate that MEN1 prevents the generation and accumulation of splicing isoforms harbored pro-tumor potency by inhibiting the AS of CD44 pre-mRNA.
To rapidly and quantitatively measure alterations of CD44 splicing efficiency, a luciferase splicing minigene reporter, pETCatEBLucv6, was constructed in which the inclusion of variant exon 6 was measured by luciferase activity (Fig. 4m). As shown in Fig. 4n, knockout of MEN1 increased the basal luciferase reporter activity and markedly enhanced PMA-stimulated reporter activity. After the PMA stimulation, the luciferase activity of the splice backbone vector pETLuc was only marginally higher than in DMSO-treated cells (Fig. 4o), ruling out the possibility that increased transcription from the splice reporter vector induced by PMA increased the luciferase activity. Moreover, RT-PCR analysis of RNA from NCI-H460 or A549 cells transfected with pETCatEBLucv6 revealed that MEN1 deficiency promoted the basal level of v6-exon inclusion. Again, it significantly augmented the PMA-induced inclusion of the variant exon 6 (Extended Data Fig. 4l-o). These splicing reporter assays confirmed the inhibitory effect of MEN1 on variant exon inclusion in the CD44 pre-mRNA.
Menin binds to the target RNA but not to CD44 and modulates RNA alternative splicing
Next, we sought to understand the molecular mechanism underlying the involvement of MEN1 in CD44 AS regulation. We evaluated whether MEN1 directly regulates RNA splicing through its interaction with specific RNAs and performed RNA immunoprecipitation and sequencing (RIP-seq) of menin-bound RNA complexes from DMSO (Ctrl) and aphidicolin (APH)-treated NCI-H460 cells (Extended Data Fig. 5a). By normalizing the IP signal with size-matched Input RIP libraries, we identified quantitative estimates of menin-binding intensity, resulting in 2,134 and 3,623 menin high-confidence peaks from 899 and 1,556 menin-bound transcripts on a genome-wide scale in the Ctrl and APH-treated cells, respectively (log2FC(IP/Input) ≥ 1.5, adjusted pvalue < 0.05) (Fig. 5a,b, and Supplementary Table 6). We further analyzed the menin distribution patterns within total peaks. A similar pattern of menin distribution in the Ctrl and APH-treated cells was observed when the RNA species were divided into exon, intron, promoter, intergenic, 3′UTR, and 5′UTR regions (Fig. 5c). However, menin does not bind to the CD44 RNA in NCI-H460 cells regardless of whether APH induction is present or not (Supplementary Table 6), suggesting that menin modulates CD44 AS in an RNA-binding protein-independent manner.
In view of RIP-seq analysis, we reasoned that MEN1 controls the AS through both RNA-binding protein-dependent and protein-independent mechanisms. Hence, we decided to investigate the mechanisms by which menin regulates AS through its RNA-binding protein. In the APH-treated cells, 2,502 novel menin peaks appeared with the disappearance of 923 peaks, and the other 1,121 peaks were found in both the Ctrl and APH-treated NCI-H460 cells (Fig. 5a,b). Since DNA damage induced AS pattern changes by interfering with the binding of RNA splicing factors with pre-mRNA26, 1,121 peaks are expected to contain genuine targets of menin. To illustrate the biological functions of loci related to menin peaks, we conducted GO analysis of the biological process and molecular function of common peaks, revealing genes for RNA metabolism such as mRNA processing, RNA splicing, and RNA-binding (Extended Data Fig. 5b). In addition, KEGG pathway enrichment analysis showed that these peaks are predominantly enriched in pathways including necroptosis, tight junction, focal adhesion, and Wnt signaling pathway (Fig. 5d).
Mapping the location of common menin-binding sites across the gene body displayed that the majority of menin-binding sites are located in exons and proximal introns of > 100 nucleotides from splice sites (Fig. 5e). Although menin exhibited less binding to exons in APH-treated cells compared with control cells, it binds to 3′UTR regions more significantly in APH-treated cells. Differences in binding to 5′UTR or distal and proximal intron regions demonstrated no statistical significance (Fig. 5e). Metagene analysis to evaluate menin-binding intensity across all human introns, exons and coding sequences (CDSs) revealed dramatically more frequent binding of menin to exons or CDSs compared with introns (Fig. 5f). Intriguingly, menin intronic binding in control cells appears skewed toward the 5′ splice sites (5′SSs) and 3′SSs whereas exonic and CDSs binding favored 5′SSs. After PAH induction, the binding of menin to intronic regions is more inclined to 3′SSs. Unexpectedly, the binding of menin to exons or CDSs was drastically diminished or even disappeared at 5′SSs, while sharply increasing at 3′SSs (Fig. 5f). Quantification of the menin-binding ability in the exons, 3′UTR, 5′UTR, and distal and proximal introns showed prominently higher binding in APH-treated cells (Fig. 5g). This further supports our observations that menin binds to RNAs in lung cancer cells more frequently and strongly in response to external stimuli.
To identify the binding site consensus sequences for menin, we plotted menin binding peaks and identified the top-ranking binding motifs. Fractions of targeted sequences were comparable in the Ctrl and APH-treated cells (Fig. 5h), suggesting that menin cooperates with other splicing regulators to recognize different types of sequences. The menin high-fidelity motifs at least contain two consecutive three-nucleotide stretches consisting of one cytosine and two uridine nucleotides (Fig. 5h). Further, the menin-bound peaks of RIP-seq data were compared from control NCI-H460 cells with the AS profiling of RNA-seq data from WT and KO NCI-H460 cells. We identified 133 differentially spliced RNAs that are both regulated by MEN1 and bound by menin (Extended Data Fig. 5c). Similarly, 373 AS events affected by both APH induction and bound by menin were identified (Extended Data Fig. 5d). Integrated genomics view showed that these representative RNAs encompass one or more menin consensus binding motifs (Fig. 5I and Extended Data Fig. 5e). The menin-bound to targeted RNAs was validated by RIP-qPCR assays in vector and MEN1 overexpressed NCI-H460 cells (Fig. 5j). Moreover, we performed RNA pull-down assays using RNA oligos targeting the SLC4A7 and SLTM transcripts, which contain one and three menin-binding motifs, respectively (Fig. 5i and Extended Data Fig. 5e). As shown in Fig. 5k, menin binds to the WT sequence but not to their mutant (Mut) counterparts. We designed PCR primer pairs targeting two neighboring constitutive exons in MEN1-regulated genes with exon skipping. RT-PCR results indicated that knocking out MEN1 in NCI-H460 cells gave rise to a marked increase in SLC4A7, BAG1, and VCP exon skipping, whereas it inhibited the exon skipping of SLTM, CLSTN1, and THOC1 genes (Fig. 5l and Extended Data Fig. 5f). A similar regulatory effect of MEN1 on AS was further confirmed in MEN1 knockdown NCI-H446 cells; we also observed that APH treatment moderately affected MEN1 deletion-induced exon skipping of these genes (Extended Data Fig. 5g,h).
These analyses suggest that menin is an RNA-binding protein and directly regulates the AS of a specific subset of RNAs in lung cancer cells.
Menin regulates the release of Pol II from CD44 gene body through the PAF1 recruitment
Since menin does not directly bind to the CD44 RNA, we excluded the possibility that MEN1-affected CD44 variant production is mediated by a trans-acting mechanism. Releasing and pausing RNA polymerase II (Pol II) during early transcriptional elongation is a critical step in pre-mRNA processing regulation27,28. Because pausing Pol II can lead to AS changes through cotranscriptional coupling29, we decided to explore whether the MEN1 effect on CD44 AS was caused by alteration of Pol II progression. Immunoblotting analysis from chromatin-associated fractions revealed that PMA induction resulted in a time-dependent decrease in association abundance of the total Pol II (8WG16), Ser-phosphorylated Pol II [p-Pol II (Ser 5)], p-Pol II (Ser 2), and Rbp1 CTD (phosphorylated and unphsophorylated Pol II forms) (Fig. 6a). Importantly, MEN1-KO dramatically diminished the enrichment of chromatin with Pol II (8WG16), p-Pol II (Ser 5) (a hallmark of paused Pol II30), and Rbp1 CTD but increased the chromatinic enrichment of p-Pol II (Ser 2) (a mark for elongating Pol II30) (Fig. 6a). More importantly, decreased recruitment of Rbp1 CTD to the 5′SSs or 3′SSs of the CD44 gene following MEN1 depletion is accompanied by augmented p-Pol II (Ser 2) occupancy, effects that were completely reversed by re-expression of menin in KO NCI-H460 cells, as determined by chromatin immunoprecipitation (ChIP) assays (Fig. 6b). These results suggest that MEN1 deletion facilitates the Pol II release from the splice sites, thereby expediting the transcriptional elongation of the CD44 gene.
An indirect approach of evaluating the effects of MEN1 on Pol II elongation is monitoring at the total Pol II distribution across a given gene via ChIP. We normalize the ChIP enrichment of WT and KO cells to 100% at the intron 1 region (i1, promoter-proximal), the Pol II distribution on the CD44 gene in MEN1-KO cells displays an accumulation toward the gene distal regions (i18) (Extended Data Fig. 6a). Similarly, also in this assay, the loss of menin expression decreased by 30% in the ratio of promoter-proximal (position + 1) to distal (position + 4,828) on the c-Myc gene (Extended Data Fig. 6b), whose transcription being negatively regulated by the menin/MLL complexes31, which is consistent with a fasted Pol II progression. This highly processive elongation of Pol II induced by MEN1 inactivation was further validated by using an adenovirus E1a reporter minigene (pME1A), whose presents a single 3′SS and three alternative 5′SSs, giving rise to four splicing isoforms designated 13S, 12S, 11S, and 9S (Fig. 6c)32. The pME1A replication provokes a preferred use of the upstream 5′SS generating higher proportions of shorter isoforms. This has been attributed to alterations in the conformation of template DNA that in turn slows down Pol II elongation efficiency33. Using this reporter, we observed that overexpression of MEN1 in NCI-H446 cells increased by 2.3-fold the proportion of shorter isoforms (9S/13S ratio) (Extended Data Fig. 6c). Conversely, a 38% reduction in the 9S/13S ratio was observed when the pME1A was transcribed by the fast Pol II in MEN1 knockout NCI-H460 cells (Fig. 6d). Finally, we employed DRB release approach to calculate the elongation rate of Pol II across the CD44 gene. Pre-mRNA expression was detected by qPCR with primers specific for exon-intron boundary sequences. NCI-H460 cells were incubated with DRB for 3 h, followed by qPCR analysis reporting that the expression of the proximal region of the CD44 gene, located at 25 bp downstream to the transcriptional start site (TSS), which was recovered already within 8 min of DRB removal in both MEN1-WT and MEN1-KO cells (Fig. 6e, left). Reciprocally, the transcriptional recovery of the distal region, located 37 kb downstream to the TSS, occurred approximately 18 min after a drug release. Then, the elongation rate of Pol II was estimated to be 3.7 kb/min in MEN1-WT cells (Fig. 6e, left). In MEN1-KO cells, however, the recovery of transcription of the distal region was advanced to about 10 min, and the estimated Pol II elongation rate (7.4 kb/min) was at least twice that of the MEN1-WT cells (Fig. 6e, right). Similarly, the elongation rate of DRB-sensitive OPA1 gene was calculated in the NCI-H460 cells, showing that Pol II transcribed the OPA1 gene at a rate of 3.3 kb/min in MEN1-WT cells, whereas without MEN1, Pol II transcribed the same region of the OPA1 gene at a rate of > 3.3 kb/min (Extended Data Fig. 6d). The fast Pol II processivity induced by MEN1 deficiency significantly promoted the production of nascent transcripts of CD44 variant exons and c-Myc gene, as detected by nuclear run-on (NRO)-qPCR assays (Fig. 6f and Extended Data Fig. 6e); however, reintroduction of MEN1 into MEN1-KO cells effectively suppressed the generation of nascent transcripts of these two genes (Fig. 6f and Extended Data Fig. 6e). Our findings suggest that MEN1 slows transcriptional elongation by decreasing the paused Pol II release into the CD44 gene body.
How do the observed changes in menin expression affect Pol II release from the CD44 gene body? The controlled release of Pol II from promoter-proximal pausing (PPP) sites requires the recruitment of PAF134,35. Immunoblotting analysis of chromatin-associated proteins from NCI-H460 and A549 cells revealed that the association abundance of PAF1 was markedly augmented upon MEN1 deficiency (Fig. 6g and Extended Data Fig. 6f). Notably, ChIP analysis of PAF1 after MEN1 deletion indicates increased recruitment at the PPP site, i1, C16, and variant exons of the CD44 gene and the transcription units of the c-Myc gene in comparison with MEN1-WT cells (Fig. 6h and Extended Data Fig. 6g). Knockdown by siRNA directed against PAF1 (Supplementary information, Fig. S6h) partially recovered MEN1 deletion-induced enrichment reduction of p-Pol II (Ser 5) within variant exons (v3-v9) but further increased the occupancy of p-Pol II (Ser 2) (Fig. 6i,j), suggesting that MEN1 regulates the Pol II release from the CD44 gene body in a PAF1-dependent manner.
A relatively slow elongation would allow more time for the assembly and recruitment of splicing complexes at the splice sites, thereby controlling the generation of splicing isoforms36. As expected, RIP-qPCR analysis showed that MEN1 knockdown reduced the recruitment of epithelial splicing regulatory protein 1 (ESRP1) at 5′SSs and 3′SSs of several variant exons including v4, v6, v7, and v9 while increasing the recruitment of hnRNPLL, but did not affect the ESRP1 recruitment at the C5 3′SS or v10 5′SS (Extended Data Fig. 6i,j). This experiment was repeated in the WT and KO NCI-H460 cells, a similar pattern of reduced ESRP1 recruitment also occurs in MEN1 knockout cells (Fig. 6k). Again, we failed to detect a significant change in ESRP1 recruitment at C5 3′SS or v10 5′SS in MEN1-deficient cells (Fig. 6k). Importantly, PAF1 knockdown obviously restored the recruitment of ESRP1 in MEN1-deficient cells (Fig. 6k), further supporting that menin regulates the recruitment of splicing factors in the context of the PAF1 complex. In agreement with this observation, immunoblotting analysis showed that the protein level of chromatin-associated ESRP1 was increased while the hnRNPLL protein level was decreased upon MEN1 deficiency (Fig. 6g and Extended Data Fig. 6f).
Collectively, these findings indicate that the increased recruitment of PAF1 by MEN1 depletion is responsible for the accelerated Pol II release from the CD44 gene body. The rapid Pol II release ratio speeds up the transcriptional elongation and generates a large number of unspliced pre-mRNAs, leaving insufficient time for splicing factors to assemble and recruit, thus disturbing the splice site selection and RNA splicing of splicing factors (Fig. 6l).
CD44v6 peptides suppress the growth and metastasis of pre-existing lung cancers by reactivating ferroptosis
Since MEN1 inactivation was to promote a CD44 variant generation essential for cancer cell viability, then reducing the levels of this variant would sensitize lung cancer cells to ferroptosis. The KrasLSL−G12D/+;Men1f/f;Sftpc-CreER and KrasLSL−G12D/+;Sftpc-CreER mice were induced by an intraperitoneal (i.p.) injection of 100 mg/kg tamoxifen, and let them feed for 6 and 8 weeks, respectively, when all animals had developed primary lung tumors. Then, these animals were treated with murine CD44v6 peptide (designated mv6pep), control peptide (Ctrlpep), or erastin 3 times per week for another 3 weeks (Extended Data Fig. 7a). Mice were killed at the end of treatment, and body weights of mv6pep- or erastin-treated animals did not significantly change during the observation period (66 days for KMS mice and 80 days for KS mice) (data not shown). Autopsy examinations revealed that numerous tumor nodules on the lung surface could be detected in all animals treated with Ctrlpep. Strikingly, most mice treated with mv6pep or erastin were tumor-free, and small sporadic tumor nodules appeared on the lung surface in two mv6pep-treated KS mice and one erastin-treated KMS mouse (Fig. 7a,f). Importantly, a significant therapeutic effect of both drugs on the outgrowth of lung tumors was observed in KS and KMS mice; the lung weight, tumor number, tumor burden, and tumor size were reduced in the mv6pep- or erastin-treated mice when compared to Ctrlpep-treated mice (Fig. 7b-e). IHC results showed that the treatment with both drugs notably decreased the Ki67 expression in the tumor tissues of KS and KMS mice (Fig. 7f and Extended Data Fig. 7g). Intriguingly, the treatment efficacy of mv6pep in Men1-deficient Kras-mutant tumors was stronger than that observed in Kras-mutant tumors. In contrast, the inhibitory effect of erastin on Kras-mutant tumors was stronger than that on Men1-deficient Kras-mutant tumors (Supplementary Table 7).
We next examined whether mv6pep induced ferroptosis in two established lung tumors. Indeed, as a classical inducer of ferroptosis, erastin treatment did give rise to a significant induction in ferroptosis in KS and KMS tumors, as demonstrated by elevating the iron content and lipid peroxide level and 4-HNE expression and by reducing the GPX4 expression (Fig. 7h-m). Similar to erastin, mv6pep treatment prominently increased the iron content and lipid peroxidation accumulation and significantly promoted the 4-HNE expression in lung tumors from KS and KMS mice (Fig. 7h-k), suggesting activated ferroptosis signaling. An augmentation of ferroptosis was evidenced by a decreased focus number of GPX4 probes in lung tissues, particularly in tumor regions, from both KS and KMS mice treated with mv6pep, as analyzed by RNA in situ hybridization (Fig. 7l,m). Of note, although both drugs triggered ferroptotic cell death in the two tumor models, the efficacy of mv6pep was much more effective in Men1-deficient Kras-mutant tumors than the treatment with erastin, which was most effective in Kras-mutant tumors (Supplementary Table 7).
To investigate whether CD44v6 peptides can be used in a therapeutic setting on established metastatic tumors, WT and KO NCI-H460 cells were injected into the tail vein of M-NSG mice and 1 week later (when all animals from WT groups already developed lung carcinomas with metastases in the liver and kidney). Then, the human CD44v6 peptide (hv6pep) or Ctrlpep was injected intravenously three times per week for 3 weeks (Extended Data Fig. 7b). In animals treated with Ctrlpep, the three animals (two for KO mice and one for WT mice) were moribund. At that time (24 days), the body weights of hv6pep-treated animals did not significantly change compared to before treatment, but all animals treated with Ctrlpep were drastically reduced (Extended Data Fig. 7c). The lung weight, tumor burden, and tumor size in KO mice treated with Ctrlpep were significantly higher compared with that in WT mice (Extended Data Fig. 7d-g). Treatment with hv6pep effectively attenuated malignant phenotypes of the lung tumor from both the experimental mice (Extended Data Fig. 7d-g). Likewise, we also noticed that hv6pep notably reduced the positive rate of Ki67 and HMGB1 staining in the lung tumors as demonstrated by IHC analysis (Extended Data Fig. 7d,h). These results indicate that the already metastatic lung tumors can also be eliminated by human CD44v6-interfering peptide treatment. In agreement with previous observations, in the Ctrlpep-treated mice, the expression and colocalization of GPX4 and MMP9 were obviously increased in MEN1-KO lung tumors compared with that in MEN1-WT tumors (Extended Data Fig. 7i,j). Here, these protein expressions in lung tumors were markedly inhibited by hv6pep treatment (Extended Data Fig. 7i,j). Consistently, hv6pep-treated tumors showed a significant induction of ferroptosis as indicated by the accumulation of iron content and lipid peroxidation (Extended Data Fig. 7k,l). Finally, we also observed that the role of hv6pep on ferroptosis was stronger in MEN1-deficient tumors than in WT tumors (Supplementary Table 7).
Altogether, our findings suggest that CD44v6-interfering peptide is a potent ferroptosis inducer and that predominantly represses the growth and metastasis of lung tumors with MEN1 inactivation.