Mutations in the DNA mismatch repair gene MSH2 are causative of microsatellite instability (MSI) in multiple cancers. Here, we discovered that besides its well-established role in DNA repair, MSH2 exerts a novel epigenomic function in gastric cancer. Unbiased CRISPR-based mass spectrometry combined with genome-wide CRISPR functional screening revealed that in early-stage gastric cancer MSH2 genomic binding is not randomly distributed but rather is associated specifically with tumor-associated super-enhancers controlling the expression of cell adhesion genes. At these loci, MSH2 genomic binding was required for chromatin rewiring, de novo enhancer–promoter interactions, maintenance of histone acetylation levels, and regulation of cell adhesion pathway expression. The chromatin function of MSH2 was independent of its DNA repair catalytic activity but required MSH6, another DNA repair gene, and recruitment to gene loci by the SWI/SNF chromatin remodeler SMARCA4/BRG1. Loss of MSH2 in advanced gastric cancers was accompanied by deficient cell adhesion pathway expression, epithelial–mesenchymal transition, and enhanced tumorigenesis in vitro and in vivo. However, MSH2-deficient gastric cancers also displayed addiction to BAZ1B, a bromodomain-containing family member, and consequent synthetic lethality to bromodomain and extraterminal motif (BET) inhibition. Our results reveal a role for MSH2 in gastric cancer epigenomic regulation and identify BET inhibition as a potential therapy in MSH2-deficient gastric malignancies.

Significance:

DNA repair protein MSH2 binds and regulates cell adhesion genes by enabling enhancer–promoter interactions, and loss of MSH2 causes deficient cell adhesion and bromodomain and extraterminal motif inhibitor synthetic lethality in gastric cancer.

Genetic and epigenetic alterations are universal features of human cancer, regulating cancer-specific gene expression. Super-enhancers (SE) are large clusters (10–60 kb) of putative enhancers in close genomic proximity (1), and cancer-specific activation of SEs are associated with many cancer hallmarks, including oncogene activation (e.g., c-MYC, MYB, EVI1, LMO1, and RAD51B) and aberrant short and long-range enhancer–promoter interactions (2–4). SEs are frequently targeted by noncoding chromosomal alterations to reinforce cancer gene expression (5, 6). Although previous studies have reported that BRD4, MED1, MYB, and GATA factors are important for SE function, our current knowledge of trans-acting factors regulating SEs remains nascent, particularly for factors lacking sequence specific DNA-binding motifs (1, 4, 7).

Gastric cancer is the fifth most common cancer worldwide and a leading cause of global cancer-related mortality (8, 9). Although driver gastric cancer genomic alterations have been identified (10, 11), improving patient outcomes through targeted therapies remain modest due to interpatient and intrapatient genomic heterogeneity (12, 13). Our previous survey of the gastric cancer epigenome identified gastric cancer–specific SEs essential for tumor-specific gene expression (14). Here, we combined CRISPR/dCas9 chromatin immunoprecipitation mass spectrometry (CRISPR/dCas9-ChIP-MS) with CRISPR/Cas9 genome-wide screening (CRISPR/Cas9-GW) to identify key regulators associated with gastric cancer–specific SEs. We found that MSH2, a DNA mismatch repair (MMR) protein, regulates gastric cancer SEs by rewiring chromatin interactions and chromatin accessibility.

Cell lines and treatments

The following gastric cancer cell lines, SNU16 (ATCC; RRID:CVCL_0076), YCC21 (Yonsei Cancer Center, South Korea; RRID:CVCL_9654), OCUM1 (Health Science Research Resource Bank; RRID:CVCL_3084), SNU216 (Korean Cell Line Bank; RRID:CVCL_3946), GES1 (Alfred Cheng, Chinese University of Hong Kong; RRID:CVCL_EQ22), HFE145 (Hassan Asktorab, Howard University, USA), HCC2998 (Chang Young-Tae (SBIC); RRID:CVCL_1266), RKO (ATCC; RRID:CVCL_0504), MOLM13 (Charles CHUAH/SinTiong Ong, CSCB at Duke-NUS; RRID:CVCL_2119), MOLT4 (Charles CHUAH/SinTiong Ong, CSCB at Duke-NUS; RRID:CVCL_0013), NCC59 (Korean Cell Line Bank, RRID:CVCL_8900), KLE (Steve Rozen, Duke-NUS; RRID:CVCL_1329), HEC-1-A (Steve Rozen, Duke-NUS; RRID:CVCL_0293), were cultured in RPMI, MEM, DMEM, RPMI, RPMI, RPMI, RPMI, RPMI, RPMI, RPMI, RPMI, DMEM:F12, DMEM, respectively, with 10% FBS, 1% Penstrep, and 1% NEAA. Cell lines were authenticated by STR and tested for Mycoplasma time to time. For animal studies, experimental procedures were undertaken with female mice ages 6 to 8 weeks in specific pathogen-free facility and performed in accordance to the Institutional Animal Care and Utilization Committee of Singapore.

CRISPR/dCas9-based chromatin immunoprecipitation followed by mass spectrometry

YCC21 cells were sequentially infected with lentiviral constructs pHR'CMV-FB-dCas9-BSD and pHR'CMV-BirA-V5-puro to express FB-dCas9 and BirA (derived from addgene #100547, RRID:Addgene_100547 and 100548, RRID:Addgene_100548 respectively). To target either the CLDN4-promoter or a control region, lentiviral-constructs were either cloned with CLDN4-specific (4 sgRNAs) or random-targeting guides (4 sgRNAs) using Multiplex CRISPR/Cas9 assembly systems (Addgene# 1000000055). The guides were phosphorylated, annealed, and ligated into BbsI digested pX330S-2 (Addgene # 58778; RRID:Addgene_58778), pX330S-3 (Addgene # 58779; RRID:Addgene_58779), pX330S-4 (Addgene # 58780, RRID:Addgene_58780), and pX330S-5 (Addgene # 58781, RRID:Addgene_58781), respectively. Next, CLDN4-specific or random-targeting sgRNAs were assembled into a lentiviral construct pHR’ CMV hygro WSin18 4x using a Golden Gate reaction (single-step digestion and ligation process) as shown in ref. 15. YCC21 Cells expressing FB-dCas9 and BirA were then infected with lentiviral constructs containing either pHR’ CMV hygro WSin18 (CLDN4 sgRNAs) or pHR’ CMV hygro WSin18 (random-targeting sgRNAs) to express specific guides. Proteins bound to the CLDN4 promoter region were isolated using chromatin immunoprecipitation and mass spectrometry as described previously in ref. 16 and detailed in Supplementary Text.

CRISPR/Cas9 genome-wide screening to detect genes essential for proliferation of cells and CLDN4 expression

Genome-wide CRISPR/Cas9 knockout screens were performed on YCC21 (wild-type or MSH2-knockout) cells using the human CRISPR Knockout Pooled Library (Brunello, Pooled Library, Addgene #73179, RRID:Addgene_73179) as previously described (17). Briefly, the CRISPR library was amplified and packaged in HEK293T cells. Next, 30 million cells were infected with CRISPR libraries at a multiplicity of infection of 0.3 O/N followed by puromycin selection for 3 days. On day 5, to ensure complete representation of sgRNA libraries in a cell population, 8 million of transduced cells were aliquoted. The rest of the cells were allowed to propagate for 14 days. Next, to determine genes required for cell proliferation by identifying loss of sgRNAs compared with the original sgRNA representation, approximately 8 million of transduced cells were aliquoted on day 19. For CLDN4 transcriptional regulator identification, 10–20 million of CRISPR library transduced cells aliquoted from day 19 were stained with αCLDN4-FITC antibody (RRID:AB_2726819), followed by sorting of cells with CLDN4-low (bottom 5% FITC-positive cells that contain putative CLDN4 transcription activators) and CLDN4-high (top 5% FITC-positive cells contain putative CLDN4 transcription repressor) on BD AriaII. Genomic DNA was then extracted from 4 samples (days 5 and 19, top 5% FITC-High, and bottom 5% FITC-low) and used for NGS library preparation. The NGS libraries of the CRISPR screening were sequenced and analyzed using MAGeCK-VISPR (18).

ChIP-seq

For each cell line, 2×107 cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature and quenched with 0.2 mol/L glycine. Chromatin was extracted, sonicated to approximately 500bp (Vibra cell, SONICS), and subjected to immunoprecipitation using 5 μg of antibody targeting MSH2 (Abcam 70270) overnight at 4°C. Protein G beads were added to the mixture for 3 hours at 4°C. Beads were washed with high salt, followed by low salt, and wash buffer at room temperature. Next, purified ChIP DNA (10 ng) was used to prepare libraries using the NEBNext ChIP-seq library prep reagent set (NEB) and sequenced to an average depth of 30–50 million reads on a HiSeq4000 sequencer. Sequences were mapped against the human reference genome (hg19) using Burrows-Wheeler Aligner (BWA-mem; version 0.7.10; ref. 19). Only reads with mapQ >10 and with duplicates removed by rmdup were used in the subsequent analysis. MACS2 was used to call significant peaks (q < 0.05; ref. 19). For stringent MSH2 ChIP-seq results, peaks obtained in the wild-type cells were manually filtered to exclude peaks found in MSH2-knockout cells to improve specificity. Differential H3K27ac ChIP-seq peaks for wild-type versus MSH2-knockout cells were detected using the callpeak and bdgdiff function of MACS2 with the default cutoff value of log10 likelihood ratio >3.

HiChIP and analysis

HiChIP experiments were performed as previously described (20). Briefly, 20 million cells were cross-linked with 1% formaldehyde at room temperature for 10 minutes followed by quenching of formaldehyde using 125 mmol/L glycine. Fixed cells were lysed to isolate nuclei and subjected for contact generation. The nuclear pellet was pretreated with 0.5% SDS before restriction digestion with 25 U/μL MboI restriction enzyme (NEB, R0147) overnight at 37°C. Next, biotin fill-in reactions were performed by incubation with master mix consisting of biotin-ATP (Thermo Fisher Scientific, 19524016), mmol/L dCTP (N0441S, NEB), dTTP (N0443S, NEB), dGTP (N0442S, NEB), and DNA Polymerase I, Large (Klenow) Fragment (NEB, M0210) for 1 hour at 37°C, followed by proximity ligation with T4 DNA Ligase (NEB, M0202) in the presence of ATP (NEB, B0202), at room temperature for 4 hours. The nuclear pellet with in situ generated contacts was sonicated to approximately 400bp (Covaris S220) and subjected to immunoprecipitation using 15 μg of antibody targeting MSH2 (Abcam 70270). Protein G beads were then added to the mixture for 3 hours at 4°C. Beads were washed with high salt, followed by low salt, and wash buffer at room temperature. Next, DNA was eluted, reverse-crosslinked, and purified with AMPure XP beads. Purified DNA samples were then incubated with Streptavidin C-1 beads to capture biotin-labeled HiC contacts. DNA bound to C-1 beads (On beads) was used to prepare NGS library using KAPA HyperPrep Kit (Roche), followed by size selection (300–700 bp). The size-selected library was paired-end sequenced on an Illumina platform (HiSeq3000) for 2×300 read length. HiChIP libraries were processed using HiC-Pro. Significant interactions were calculated using CPU HICCUPS in Juicer Tools. HiChIP interaction reads and interactions were visualized using WashU Epigenome Browser (http://epigenomegateway.wustl.edu/).

Other Materials and Methods, such as 4C-seq, MNNG assay, cell proliferation, and xenograft assays are provided in the Supplementary Text

SEs gained in early-stage primary gastric cancers exhibit enrichment in cell adhesion pathways

To identify biological themes associated with gastric cancer–specific SEs, we identified recurrent tumor-specific SEs from epigenomic profiles of early-stage primary gastric cancers (Fig. 1A; ref. 14). Performing gene ontology (GO) enrichment analysis using the Genomic Regions Enrichment of Annotations Tool, we identified 1214 SEs exhibiting recurrent somatic activation in at least 10% of gastric cancers (21). Early-stage tumor-associated SEs exhibited enrichment in cell adhesion pathways (P < 1.3×10–6), and other cancer-related processes, including transcription factor binding and programmed cell death (Fig. 1B; Supplementary Fig. S1A).

Figure 1.

Gastric cancer super-enhancers exhibit enrichment in cell adhesion pathways. A, Recurrent tumor-associated SEs in primary gastric cancer. B, GO terms associated with gastric cancer SEs. *, cell adhesion related. C, Upregulated cell adhesion genes in gastric cancer (%). D, SEs ranked in gastric cancer lines. E,CLDN4 expression in gastric cancer (red) or normal gastric lines (green). F,CLDN4 SE H3K27ac tracks in gastric cancer and normal lines.

Figure 1.

Gastric cancer super-enhancers exhibit enrichment in cell adhesion pathways. A, Recurrent tumor-associated SEs in primary gastric cancer. B, GO terms associated with gastric cancer SEs. *, cell adhesion related. C, Upregulated cell adhesion genes in gastric cancer (%). D, SEs ranked in gastric cancer lines. E,CLDN4 expression in gastric cancer (red) or normal gastric lines (green). F,CLDN4 SE H3K27ac tracks in gastric cancer and normal lines.

Close modal

Deregulated cell adhesion is a prevalent feature of several cancer types (22). We selected Claudin-4 (CLDN4), highly expressed in approximately 40% of gastric cancers, as a candidate to study SE-mediated cell adhesion (Fig. 1C; Supplementary Table S1). Claudins are components of tight junctions that promote cell adhesion and maintain cell polarity (23). CLDN4 is also upregulated in other epithelial cancer types, including esophageal, biliary, ovarian, and prostate (24).

DNA repair protein MSH2 regulates CLDN4 expression in gastric cancer

We used two gastric cancer cell lines (SNU16 and YCC21) to study CLDN4 SE-mediated gene expression. Compared with normal gastric lines GES1 and HFE145, both gastric cancer lines exhibited gastric cancer–specific CLDN4 SEs and elevated CLDN4 expression (Fig. 1DF). To identify trans-acting factors required for CLDN4 SE-mediated expression, we integrated two complementary approaches. First, using CRISPR/dCas9-based CRISPR/dCas9-ChIP-MS (16), we identified proteins binding to the CLDN4 promoter. As chromatin fixation preserves short-range and long-range enhancer–promoter loops, targeting the CLDN4 promoter effectively captures proteins binding to both gene promoter and distal enhancers. As controls, cells expressing either no-sgRNAs (Supplementary Table S2A) or random-targeting sgRNAs (Supplementary Table S2B) were used (16). Second, we performed CRISPR/Cas9 genome-wide (CRISPR/Cas9-GW) screening by transducing cells with sgRNA libraries, followed by CLDN4 staining to identify regulators of CLDN4 expression (see methods). Notably, CRISPR-mediated deletion of CLDN4 in YCC21 cells does not affect cell growth or proliferation (Supplementary Fig. S1B). The CRISPR/dCas9-ChIP-MS screen identified 173 proteins binding to the CLDN4 locus with known functions in chromatin remodeling and transcriptional regulation [Fig. 2A, left (Approach 1); Supplementary Fig. S2A–S2D; Supplementary Table S2A and S2B]. In parallel, the CRISPR GW screen identified 575 genes required for CLDN4 expression (P < 0.05; Fig. 2A, right (Approach 2); Supplementary Fig. S2E; Supplementary Table S3A).

Figure 2.

DNA repair protein MSH2 regulates CLDN4 expression in gastric cancer. A, Left, proteins identified by CRISPR/dCas9-ChIP-MS. Middle, overlay of proteins obtained from CRISPR/dCas9-ChIP-MS and CRISPR/Cas9-GW. Right, CLDN4-activators identified by CRISPR/Cas9-GW. B,MSH2 or CLDN4 expression fold change upon either control or MSH2 targeting siRNA treatment in YCC21 cells. *, P < 0.05; ***, P < 0.0005. C, Immunoblot showing CLDN4 and GAPDH levels upon treatment with either control or MSH2-targeting siRNAs in YCC21. D, Immunoblot showing MSH2 levels in wild-type or MSH2 knockouts with or without V5-tagged MSH2 reconstitution in YCC21. E,MSH2 or CLDN4 expression fold change in wild-type or MSH2 knockouts with or without V5-tagged MSH2 reconstitution in YCC21. Black *, CLDN4 expression; red *, significant loss of MSH2 or CLDN4 expression in MSH2 knockouts. *, P < 0.05; **, P < 0.005; ***, P < 0.0005. F,MSH2 expression in TCGA-STAD versus normals and across TCGA subtypes. *, P < 0.05; ***, P < 0.001. G,MSH2 and CLDN4 coexpression analysis in the TCGA-STAD dataset. H,MSH2 or CLDN4 expression fold change in wild-type or MSH2-knockout with or without MSH2-WT, -R524P, or -G674S reconstitution in YCC21. **, P < 0.005; ***, P < 0.0005.

Figure 2.

DNA repair protein MSH2 regulates CLDN4 expression in gastric cancer. A, Left, proteins identified by CRISPR/dCas9-ChIP-MS. Middle, overlay of proteins obtained from CRISPR/dCas9-ChIP-MS and CRISPR/Cas9-GW. Right, CLDN4-activators identified by CRISPR/Cas9-GW. B,MSH2 or CLDN4 expression fold change upon either control or MSH2 targeting siRNA treatment in YCC21 cells. *, P < 0.05; ***, P < 0.0005. C, Immunoblot showing CLDN4 and GAPDH levels upon treatment with either control or MSH2-targeting siRNAs in YCC21. D, Immunoblot showing MSH2 levels in wild-type or MSH2 knockouts with or without V5-tagged MSH2 reconstitution in YCC21. E,MSH2 or CLDN4 expression fold change in wild-type or MSH2 knockouts with or without V5-tagged MSH2 reconstitution in YCC21. Black *, CLDN4 expression; red *, significant loss of MSH2 or CLDN4 expression in MSH2 knockouts. *, P < 0.05; **, P < 0.005; ***, P < 0.0005. F,MSH2 expression in TCGA-STAD versus normals and across TCGA subtypes. *, P < 0.05; ***, P < 0.001. G,MSH2 and CLDN4 coexpression analysis in the TCGA-STAD dataset. H,MSH2 or CLDN4 expression fold change in wild-type or MSH2-knockout with or without MSH2-WT, -R524P, or -G674S reconstitution in YCC21. **, P < 0.005; ***, P < 0.0005.

Close modal

Integration of the two GW lists identified two common proteins (MSH2 and FKBP1A) after excluding factors either required for general viability and/or CLDN4 regulation in an indirect fashion. We focused on MSH2, a component of the DNA MMR complex that functions to correct DNA mismatches during DNA replication (25). To validate the role of MSH2 in CLDN4 transcriptional regulation, we treated YCC21 cells with independent siRNAs targeting MSH2 and confirmed that MSH2 loss (Fig. 2B; Supplementary Fig. S2F) reduced CLDN4 expression, at both the RNA (P < 0.005; Fig. 2B) and protein level (Fig. 2C). Testing two other gastric cancer lines (OCUM1 and SNU216) yielded similar results (Supplementary Fig. S2G and S2H). In addition, we generated three MSH2 knockouts in two gastric cancer lines (SNU16 and YCC21), and confirmed that all MSH2 knockouts displayed reduced CLDN4 expression compared with isogenic wild-type cells (P < 0.05; Fig. 2D and E, Supplementary Fig. S2I and S2J). Genetic rescue experiments by re-expressing wild-type MSH2 in MSH2-knockout cells restored CLDN4 expression (P < 0.05; Fig. 2D and E).

To evaluate relationships between expression of MSH2 and CLDN4 in primary gastric cancers, we analyzed The Cancer Genome Atlas (TCGA) stomach adenocarcinoma (STAD) samples (gastric cancer = 408; normal = 36). MSH2 was upregulated in all four TCGA gastric cancer molecular subtypes compared with normal samples, with highest expression in EBV-positive and microsatellite instability (MSI)–positive gastric cancers, followed by chromosomal instability (CIN) and genome-stable gastric cancers (P < 0.05; Fig. 2F). We observed significant positive correlations between MSH2 and CLDN4 expressions (Spearman correlation coefficient = 0.28; P = 2.6×10–4), supporting the observation that MSH2 is a regulator of CLDN4 expression in primary gastric cancers (Fig. 2G). Analysis of three additional gastric cancer patient cohorts confirmed these positive correlations (Spearman coefficient = 0.41; P < 0.001; Supplementary Fig. S2K–S2M; ref. 26).

MSH2 mutations are associated with defective MMR (dMMR) and Lynch syndrome imparting genetic predisposition to cancer (hereditary nonpolyposis colorectal cancer; ref. 27). To assess whether the DNA repair functions of MSH2 are required for CLDN4 transcriptional regulation, we tested two MSH2-mutants (R524P and G674S) mimicking mutations found in patients with Lynch syndrome. R524P is partially MMR deficient due to a point mutation in the clamp domain, whereas G674S is altered in the MSH2 ATPase domain and defective in subsequent ATP processing (27, 28). To validate the loss of MMR activity by both mutations, we assessed their ability to resist N-Methyl-N′-nitro-N-nitrosoguanidine (MNNG)-mediated cell death (27). Both mutant constructs (R524P and G674S), when re-expressed in MSH2-deficient cells, were significantly resistant to MNNG treatment compared with wild-type MSH2 (P < 0.01; Supplementary Fig. S2N). Importantly, when expressed in MSH2-deficient cells, both MSH2(R524P) and MSH2(G674S) mutants still activated CLDN4 expression (Fig. 2H), suggesting that the transcriptional regulatory function of MSH2 does not require mismatch-binding or ATPase activity.

MSH2 binds and activates expression of cell–cell adhesion genes

To characterize genomic loci occupied by endogenous MSH2, we performed MSH2 ChIP-sequencing (ChIP-seq) in wild-type and MSH2-knockout YCC21 cells. We found that MSH2 binds to 422 loci, whereas MSH2-knockout cells did not reveal a significant enrichment of genomic binding (Supplementary Table S4). Visual inspection of MSH2 ChIP-seq tracks at individual genomic loci such as CLDN4, ID1, and FOSB confirmed robust MSH2 occupancy (Fig. 3A). Despite general positive correlations between gene expression levels and MSH2 occupancy (Fig. 3B), MSH2 occupancy is unlikely completely due to transcription-coupled DNA repair (29) because not all highly expressed genes showed MSH2 enrichment (Supplementary Fig. S3A and S3B). A global analysis revealed that MSH2 binds predominantly to promoters (∼60%; Supplementary Fig. S3C). MSH2 enrichment at promoter regions correlated with promoter marks (H3K27ac and H3K4me3), and at distal regions correlated with enhancer marks (H3K27ac and H3K4me1; Fig. 3C). We also used ChromHMM and found that MSH2 binding is significantly enriched in open chromatin promoter regions (Fig. 3D).

Figure 3.

MSH2 binds to and activates expression of cell adhesion genes. A, MSH2 ChIP-seq tracks in YCC21. B, Correlation of global gene expression with MSH2 ChIP enrichment in YCC21. C, Heatmap of histone modification signals around MSH2 peaks and correlation with occupancy of H3K4me1, H3K4me3, or H3K27ac for YCC21. Top, promoters (<± 2 kb from TSSs); bottom, distal elements (>± 2 kb from TSS). D, MSH2 binding across chromatin states in YCC21. E, GO terms associated with MSH2 ChIP-seq peaks. *, cell adhesion related. F, Permutation tests assessing association between the union of HNF4α peaks in two gastric cancer lines (left) or GATA4/6-binding sites in three gastric cancer lines (right) and MSH2 binding sites. G, Top GSEA plots of differentially expressed genes in MSH2 knockouts versus wild-type YCC21 (P < 0.005). H, Venn diagram of genes associated with MSH2 ChIP-seq peaks and genes differentially expressed in MSH2-knockout versus wild-type; pathway enrichment of genes in the overlap. I, Absolute mRNA quantifications of MSH2-target genes in either wild-type or MSH2 knockouts.

Figure 3.

MSH2 binds to and activates expression of cell adhesion genes. A, MSH2 ChIP-seq tracks in YCC21. B, Correlation of global gene expression with MSH2 ChIP enrichment in YCC21. C, Heatmap of histone modification signals around MSH2 peaks and correlation with occupancy of H3K4me1, H3K4me3, or H3K27ac for YCC21. Top, promoters (<± 2 kb from TSSs); bottom, distal elements (>± 2 kb from TSS). D, MSH2 binding across chromatin states in YCC21. E, GO terms associated with MSH2 ChIP-seq peaks. *, cell adhesion related. F, Permutation tests assessing association between the union of HNF4α peaks in two gastric cancer lines (left) or GATA4/6-binding sites in three gastric cancer lines (right) and MSH2 binding sites. G, Top GSEA plots of differentially expressed genes in MSH2 knockouts versus wild-type YCC21 (P < 0.005). H, Venn diagram of genes associated with MSH2 ChIP-seq peaks and genes differentially expressed in MSH2-knockout versus wild-type; pathway enrichment of genes in the overlap. I, Absolute mRNA quantifications of MSH2-target genes in either wild-type or MSH2 knockouts.

Close modal

Supporting the functional specificity of MSH2 binding, GO analysis revealed that MSH2 occupancy is not randomly distributed but significantly enriched in cell adhesion–related pathways (Fig. 3E). We confirmed enrichment of MSH2-binding and H3K27ac signals at multiple highly expressed cell adhesion genes (e.g., CLDN4, CDH1, and FLNA; Supplementary Fig. S3A and S3D). Sequence motif analysis of MSH2-binding sites highlighted significant enrichment of GATA and HNF4α-binding motifs—known master regulators of gastric cancer (Supplementary Fig. S3E). Using gastric cancer–specific GATA and HNF4α ChIP-seq data, we confirmed significant overlaps of both GATA and HNF4α binding with MSH2 genomic occupancy profiles, suggesting that GATA and HNF4α activity may contribute to the specificity of MSH2 genomic occupancy (Fig. 3F). To assess the tumor-specificity of MSH2 binding, we performed MSH2 ChIP-seq in GES1 normal gastric cells. Compared with gastric cancer lines (YCC21 and SNU16), very few peaks were found in common between normal and gastric cancer lines (1.7% overlap, Supplementary Fig. S3F, left), whereas in contrast there was a 65.57% overlap between the two gastric cancer lines (Supplementary Fig. S3F, right).

To evaluate the impact of MSH2 on the global gastric cancer transcriptome, we performed RNA-seq on three independent MSH2 knockouts (#15, #62, #69; Fig. 3G). In parallel, we also performed RNA-seq comparing transcripts of YCC21 cells treated with either control or independent siRNAs targeting MSH2 (Supplementary Fig. S4A). MSH2 loss induced a broad swath of transcriptional reprogramming involving 1,335 genes in MSH2-knockout (Supplementary Table S5A) and S11405 genes (Supplementary Table S6) in MSH2-knockdown cells. Importantly, mirroring the epigenomic findings, genes downregulated in MSH2-deficient cells exhibited a significant enrichment of cell adhesion gene signatures (Fig. 3G; Supplementary Fig. S4B, Supplementary Tables S7 and S8). Similarly, GO analyses of genes differentially expressed in RNA-seq and also associated with MSH2 ChIP-seq peaks revealed enrichment of cell-adhesion pathways (cell–cell junction organization; Fig. 3H). Using digital PCR, we validated the altered expression of cell adhesion genes identified by intersecting RNA-seq and ChIP-seq (Fig. 3I). These data suggest that MSH2 directly binds and activates cell adhesion genes. Supplementary Tables S7 and S8 present additional GO terms enriched in MSH2 target genes.

MSH2 genomic occupancy rewires chromatin interactions and chromatin accessibility

We hypothesized that MSH2 genomic binding might facilitate promoter–enhancer interactions thereby regulating transcription. To test whether MSH2 genomic occupancy regulates enhancer–promoter looping, we measured MSH2-associated 3D chromatin structures using MSH2 HiChIP. Using HICCUPS, we identified 10,190 significant long-range chromatin interactions associated with MSH2 (FDR<0.05). Once again, many MSH2-associated interactions involved cell adhesion genes (CLDN4, CDH1; Fig. 4A; Supplementary Fig. S5A and S5B; Supplementary Table S9). There was a significant overlap in MSH2 peaks identified by ChIP-seq (13,672 peaks) and Hi-ChIP (18,582 peaks, see Materials and Methods) with 5544 high-confidence overlapping peaks (chi-sq test, P < 2.2e–16, Supplementary Fig. S5C). Supporting the validity of the enhancer–promoter interactions, we detected MSH2-associated interactions at the CLDN4 promoter region with two enhancers (enhancer1 and enhancer2) previously shown to interact with the CLDN4 promoter (Fig. 4B; ref. 14).

Figure 4.

MSH2 genomic occupancy rewiring chromatin interactions and chromatin accessibility. A, MSH2 HiChIP at CLDN4 and CDH1 loci. B, MSH2 HiChIP showing interactions of the CLDN4 promoter with enhancers e1 and e2. C, 4C-seq interactions of CLDN4 promoter and enhancers in wild-type and MSH2-knockout YCC21. Red dotted circles, differential interactions. D, H3K27ac ChIP-seq tracks of cell adhesion genes in wild-type or MSH2-knockout YCC21. E, Histogram with distribution of H3K27ac peaks in wild-type versus MSH2 knockout. Heatmap of H3K27ac ChIP-seq signals >±1kb or <±1kb around the center of the peak in wild-type (red) and MSH2-knockout (black) YCC21.

Figure 4.

MSH2 genomic occupancy rewiring chromatin interactions and chromatin accessibility. A, MSH2 HiChIP at CLDN4 and CDH1 loci. B, MSH2 HiChIP showing interactions of the CLDN4 promoter with enhancers e1 and e2. C, 4C-seq interactions of CLDN4 promoter and enhancers in wild-type and MSH2-knockout YCC21. Red dotted circles, differential interactions. D, H3K27ac ChIP-seq tracks of cell adhesion genes in wild-type or MSH2-knockout YCC21. E, Histogram with distribution of H3K27ac peaks in wild-type versus MSH2 knockout. Heatmap of H3K27ac ChIP-seq signals >±1kb or <±1kb around the center of the peak in wild-type (red) and MSH2-knockout (black) YCC21.

Close modal

To confirm that MSH2 deficiency compromises enhancer–promoter interactions, we performed 4C-seq targeting the CLDN4 promoter region in wild-type and MSH2-knockout cells in gastric cancer cell lines, YCC21 and SNU16 (Fig. 4C, Supplementary Fig. S6A). In MSH2-deficient YCC21 cells, we observed significantly lower CLDN4 enhancer–promoter interactions for enhancer1 and enhancer2 (P value for differential interaction ranged between 0.02 and 0.1; calculated by DESeq2; Supplementary Table S10, Fig. 4C). To further validate a requirement of MSH2 binding at one of the enhancers, e2, in establishing promoter–enhancer interactions and transcriptional activation, we CRISPR-deleted the e2 MSH2-binding site. As predicted, e2 deletion resulted in decreased interactions between promoter and e2, and CLDN4 expression (Supplementary Fig. S6B and S6C). e2 deletion did not completely abolish CLDN4 expression, consistent with findings that eukaryotic genes are often regulated by multiple enhancers and upstream regulators where the loss of a single activator often causes a modest yet significant decrease (5, 30). To extend these findings beyond CLDN4, we also performed 4C-seq targeting the CDH1 promoter, another cell-adhesion gene displaying MSH2 enrichment by HiChIP and SE gain (14). Like CLDN4, in wild-type cells CDH1 enhancer elements interacted with the CDH1 promoter (14), whereas MSH2-deficient cells showed reduced CDH1 promoter/enhancer 1 interactions (Supplementary Fig. S6D). These data establish a role for MSH2 in the regulation of enhancer–promoter interactions.

Disruption of enhancer–promoter chromatin interactions can also cause loss of local H3K27ac acetylation (31). We observed that MSH2-knockout cells exhibited H3K27ac loss at cell adhesion genes (CLDN4 and CDH1) at regions showing decreased 4C-seq interactions (Fig. 4D). Other cell adhesion genes such as FOS and FLRT3 also showed H3K27ac loss at their gene loci (Fig. 4D). To understand the global effects of MSH2-loss on histone acetylation, we compared H3K27ac peaks between MSH2-knockout and wild-type cells. MSH2 deficiency caused a loss of 4701 acetylation peaks (Fig. 4E; Supplementary Table S11), and enrichment of GO-MSigDB terms, including cell adhesion (Supplementary Fig. S7A). These data suggest that MSH2 is required to maintain chromatin interactions and histone acetylation at cell adhesion genes. Interestingly, loss of MSH2 also resulted in a gain of 1740 acetylation peaks, for which GO analysis showed enrichment of Wnt signaling pathways (Fig. 4E, Supplementary Fig. S7B, Supplementary Table S11).

SMARCA4 (BRG1) is required to dock MSH2–MSH6 dimer at genomic loci to activate transcription of cell adhesion genes

We sought to examine relationships between MSH2 and its known heterodimerization partners—MSH6 (MutSa) or MSH3 (MutSb) in MSH2-mediated transcriptional regulation (Fig. 5A). During MMR, MSH2 heterodimerizes with either MSH6 (MutSa) or MSH3 (MutSb), and the MSH2–MutSa/b complex recognizes mismatched bases (25, 32). Analyzing both in-house primary gastric cancer RNA-seq and TCGA-STAD datasets, MSH6, but not MSH3, exhibited correlations in expression with MSH2 and CLDN4 (Fig. 5B and C; Spearman coefficient = 0.75; P = 5.5e–49), suggesting a potential involvement of MSH2–MSH6 heterodimer (MutSa) in transcriptional activation of cell adhesion genes (Fig. 5A). To test this, we compared MSH6-knockout cells with wild-type YCC21 cells. Like MSH2, expressions of cell adhesion genes, including CLDN4, CDH1, FLRT3, and FOS, were reduced in MSH6-knockout cells (Fig. 5D; Supplementary Fig. S8A). Notably, although MSH6 loss did not affect MSH2 transcript levels, loss of MSH6 decreased MSH2 protein levels, rendering it possible that the MSH6-knockout transcriptional effects may be explained (at least partially) by secondary MSH2 protein loss (Fig. 5D, right). To determine whether MSH2 and MSH6 are epistatic or collaborate to activate cell adhesion genes, we treated MSH6-knockout cells with siRNAs to target the residual MSH2. We found that MSH6 loss combined with MSH2-knockdown did not further decrease CLDN4 expression and other cell adhesion genes compared with MSH6 single knockouts (Fig. 5E; Supplementary Fig. S8B). In addition, MSH6-knockout cells failed to recruit residual MSH2 to the CLDN4 locus (Fig. 5F). We also ruled out the involvement of MutLα (MLH1 and PMS2; a part of the MSH2 complex in MMR) in MSH2-mediated transcription regulation by treating YCC21 cells with siRNAs targeting MLH1, demonstrating that MLH1 loss does not affect CLDN4 expression (Supplementary Fig. S8C).

Figure 5.

SMARCA4 is required for MSH2-MSH6 occupancy at genomic loci. A, Candidate MSH2-interacting partners at the CLDN4 promoter. B, mRNA expression of either CLDN4, MSH2, MSH3, or MSH6 in primary gastric cancers. C,MSH2 and MSH6 coexpression analysis in TCGA-STAD dataset. D, Left, CLDN4 expression fold change in either wild-type or MSH6 knockouts in YCC21. Right, immunoblot showing protein levels of MSH6, MSH2, or p84 in either wild-type or MSH6 knockouts. E, MSH2 or CLDN4 expression fold change in MSH6-knockout or wild-type cells upon control or MSH2 targeting siRNA treatment in YCC21. F, MSH2 ChIP-qPCR at the CLDN4 promoter in wild-type and MSH6-knockout. G, SMARCA4 and MSH2 coexpression analysis in the TCGA-STAD dataset. H,SMARCA4 and CLDN4 coexpression analysis in the TCGA-STAD dataset. I, Immunoblot showing SMARCA4 protein levels in wild-type or SMARCA4 knockouts. J,CLDN4 expression fold change in either wild-type or SMARCA4 knockouts in YCC21. K,MSH2 or CLDN4 expression fold change in wild-type or SMARCA4 knockout upon either control or MSH2-targeting siRNA treatment in YCC21. L, MSH2 ChIP-qPCR at the CLDN4 promoter in wild-type, SMARCA4-knockout, or JUNB-knockout. *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P < 0.00005.

Figure 5.

SMARCA4 is required for MSH2-MSH6 occupancy at genomic loci. A, Candidate MSH2-interacting partners at the CLDN4 promoter. B, mRNA expression of either CLDN4, MSH2, MSH3, or MSH6 in primary gastric cancers. C,MSH2 and MSH6 coexpression analysis in TCGA-STAD dataset. D, Left, CLDN4 expression fold change in either wild-type or MSH6 knockouts in YCC21. Right, immunoblot showing protein levels of MSH6, MSH2, or p84 in either wild-type or MSH6 knockouts. E, MSH2 or CLDN4 expression fold change in MSH6-knockout or wild-type cells upon control or MSH2 targeting siRNA treatment in YCC21. F, MSH2 ChIP-qPCR at the CLDN4 promoter in wild-type and MSH6-knockout. G, SMARCA4 and MSH2 coexpression analysis in the TCGA-STAD dataset. H,SMARCA4 and CLDN4 coexpression analysis in the TCGA-STAD dataset. I, Immunoblot showing SMARCA4 protein levels in wild-type or SMARCA4 knockouts. J,CLDN4 expression fold change in either wild-type or SMARCA4 knockouts in YCC21. K,MSH2 or CLDN4 expression fold change in wild-type or SMARCA4 knockout upon either control or MSH2-targeting siRNA treatment in YCC21. L, MSH2 ChIP-qPCR at the CLDN4 promoter in wild-type, SMARCA4-knockout, or JUNB-knockout. *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P < 0.00005.

Close modal

To identify mechanisms used by MSH2–MSH6 to dock at loci such as CLDN4 (Fig. 5A), we noted a previous report that ARID1A, a subunit of the SWI/SNF (BAF) complex, can recruit MSH2 to chromatin during MMR (33). We thus scanned the CRISPR/dCas9-ChIP-MS dataset for ARID1A-like SWI/SNF complex proteins binding to the CLDN4 locus and identified SMARCA4 (BRG1; Fig. 2A, Approach 1). Supporting this finding, SMARCA4 is also predicted by ENCODE to bind to the CLDN4 locus (5kb upstream of TSS; ref. 34). In primary gastric cancers (TCGA-STAD dataset), SMARCA4 is overexpressed in tumors compared with normal, and exhibited positive correlations with MSH2 (Spearman correlation coefficient = 0.33; P = 4.4e–8) and CLDN4 (Spearman correlation coefficient = 0.28; P = 3.5e–6) expression (Fig. 5G and H; Supplementary Fig. S8D). We generated SMARCA4-knockout cells and confirmed that SMARCA4 loss reduces expression of CLDN4 and other cell adhesion genes (Fig. 5I and J; Supplementary Fig. S8E). To evaluate whether MSH2 requires SMARCA4 to activate cell adhesion genes, we treated SMARCA4-knockout cells with siRNAs targeting MSH2 or control, and measured expression levels of CLDN4 and other cell adhesion genes. We found that SMARCA4 and MSH2 likely act in the same pathway to activate CLDN4 and other cell adhesion genes (Fig. 5K; Supplementary Fig. S8F). We also performed ChIP-qPCR using MSH2 antibodies at the CLDN4 locus on wild-type and SMARCA4 knockout cells, and found that SMARCA4-deficient cells were unable to recruit MSH2 (Fig. 5L). As a control, JUNB-deficient cells (another transcription factor that binds to the CLDN4 promoter region as predicted by ENCODE) maintained enrichment of MSH2 occupancy at the CLDN4 locus (Fig. 5L; Supplementary Fig. S8G). Similarly, SETD2, a histone methyl transferase responsible for H3K36 trimethylation, is required for recruitment of MutSa during MMR (35). We tested whether SETD2 is also involved in MSH2 recruitment by performing MSH2 ChIP-qPCR at the CLDN4 locus on SETD2 knockout cells, and found that SETD2 loss compromised MSH2 enrichment at the CLDN4 promoter (Supplementary Fig. S8H and S8I). Together, these data suggest that MSH6, SMARCA4, and SETD2 are likely involved in MutSa recruitment to cell adhesion genes for transcription regulation.

Late-stage MSH2-deficient gastric cancers display enhanced tumorigenesis

In contrast with early-stage tumors, loss of cell adhesion is often observed in advanced cancers and associated with tumor progression processes like epithelial–mesenchymal transition (EMT). We hypothesized that MSH2 deficiency, particularly in advanced gastric cancer, might contribute to tumor aggressiveness through the loss of cell adhesion gene expression. Supporting this model, gastric cancer lines with CLDN4-low expression showed a classic EMT gene expression signature compared with CLDN4-high expressing lines (Fig. 6A; ref. 36). In the TCGA dataset, a comparison of combined stages I, II, III to stage IV (metastatic) in CIN/MSI/EBV gastric cancer subtypes revealed a loss of MSH2 expression in stage IV samples (Supplementary Fig. S9A, P = 0.022, Wilcoxon test; genome-stable subtype tumors were excluded as they already show low MSH2 expression; see Fig. 2F). In a subset analysis, TCGA-STAD samples exhibiting both low CLDN4 expression and low MSH2 expression were enriched in stage IV or advanced tumors (Supplementary Fig. S9B), and MSH2 null MSI tumors exhibited significantly lower CLDN4 expressions compared with MSH2 wild-type MSI gastric cancers (Fig. 6B). In an independent cohort of paired patient-matched primary and metastatic gastric cancer samples, we also observed loss of CLDN4 and CDH1 expression in metastatic lesions compared with primary lesions (Fig. 6C). To functionally test the impact of MSH2 expression on gastric cancer cell proliferation, we performed cell proliferation assays on YCC21 cells treated with either control or siRNAs targeting MSH2, and found that MSH2 knockdown cells indeed grew faster than control cells (P = 0.0022, Fig. 6D). Loss of cell–cell adhesion and anchorage-independent growth are hallmarks of cancer, and are repressed by cell adhesion molecules, including E-cadherin (37–39). Consistently, we found that in soft agar assays MSH2 knockout cells grew 2-fold more colonies than wild-type YCC21 cells (Fig. 6E). To validate these data in vivo, we performed xenograft experiments by transplanting either wild-type or MSH2-knockout cells in mice (6 mice/group). We found that tumors grew faster in the absence of MSH2 (P = 0.0022, Fig. 6F). These data suggest that in late-stage gastric cancer, MSH2 deficiency may reduce CLDN4 expression, in turn contributing to aggressive tumor growth. Of note, MNNG-treated MSH2-deleted cells also exhibited altered commitment to mitosis as previously reported (Supplementary Fig. S9C; ref. 27). However, we did not observe MSH2 binding to cell-cycle genes or an enrichment of cell-cycle programs after MSH-silencing, suggesting that MSH2 role in cell adhesion is likely independent of cell-cycle progression (Fig. 3E and G; Supplementary Tables S7 and S8).

Figure 6.

MSH2 deficiency drives tumor aggressiveness and addiction to bromodomain protein BAZ1B. A, Heatmap representing expression of EMT gene signatures in gastric cancer lines stratified for CLDN4 expression. B,CLDN4 expression in MSH2-null or wild-type cancers from TCGA-STAD cohort. C, Fold change in expression of cell adhesion genes in primary versus metastatic gastric cancer samples using Nanostring. D, Cell proliferation assays of YCC21 transfected with either control or MSH2-targeting siRNAs (P = 0.0022). E, Number of colonies formed on a soft agar plate by wild-type and MSH2-knockout in YCC21. **, P < 0.005. F, Tumor volumes of either wild-type or MSH2-deficient YCC21 in NSG mice (P = 0.0022). G, Four-way Venn diagram of (i) genes synthetically lethal with MSH2 loss, (ii) genes required for the proliferation of wild-type cells, (iii) genes downregulated with MSH2 loss, and (iv) genes in GO_KINASE_ACTIVITY (MSIGDB geneset). H,BAZ1B or CLDN4 expression fold change in MSH2 knockout or wild-type YCC21 upon control or BAZ1B-targeting siRNA treatment. **, P < 0.005; ***, P < 0.0005; ****, P < 0.00005. I, IC50 values of either wild-type or MSH2-knockout YCC21 treated with JQ1. J, Tumor volumes of JQ1 or vehicle-treated MSH2-knockout YCC21 in NSG mice (P = 0.0002).

Figure 6.

MSH2 deficiency drives tumor aggressiveness and addiction to bromodomain protein BAZ1B. A, Heatmap representing expression of EMT gene signatures in gastric cancer lines stratified for CLDN4 expression. B,CLDN4 expression in MSH2-null or wild-type cancers from TCGA-STAD cohort. C, Fold change in expression of cell adhesion genes in primary versus metastatic gastric cancer samples using Nanostring. D, Cell proliferation assays of YCC21 transfected with either control or MSH2-targeting siRNAs (P = 0.0022). E, Number of colonies formed on a soft agar plate by wild-type and MSH2-knockout in YCC21. **, P < 0.005. F, Tumor volumes of either wild-type or MSH2-deficient YCC21 in NSG mice (P = 0.0022). G, Four-way Venn diagram of (i) genes synthetically lethal with MSH2 loss, (ii) genes required for the proliferation of wild-type cells, (iii) genes downregulated with MSH2 loss, and (iv) genes in GO_KINASE_ACTIVITY (MSIGDB geneset). H,BAZ1B or CLDN4 expression fold change in MSH2 knockout or wild-type YCC21 upon control or BAZ1B-targeting siRNA treatment. **, P < 0.005; ***, P < 0.0005; ****, P < 0.00005. I, IC50 values of either wild-type or MSH2-knockout YCC21 treated with JQ1. J, Tumor volumes of JQ1 or vehicle-treated MSH2-knockout YCC21 in NSG mice (P = 0.0002).

Close modal

MSH2-deficient gastric cancers display addiction to the bromodomain-containing family member BAZ1B

To explore druggable genetic vulnerabilities of MSH2-deficient cells, we performed CRISPR/Cas9-GW screening in MSH2-knockout YCC21 cells and found 952 synthetic lethal partners (Supplementary Table S12). To enrich potential therapeutic targets, we plotted Venn diagrams intersecting: (i) genes necessary for MSH2-knockout cell proliferation, (ii) genes necessary for proliferation of wild-type cells (Supplementary Table S13), (iii) genes downregulated by MSH2-loss (Supplementary Table S5B), and (iv) genes (proteins) that have kinase activity (targetable by kinase inhibitors; MSigDB-GO_KINASE_ACTIVITY). Using this approach, we identified BAZ1B as a gene required for proliferation in MSH2-knockout cells but not wild-type cells (Fig. 6G). BAZ1B is an atypical tyrosine–protein kinase that is a chromatin remodeler and bromodomain-containing protein (40). We validated that BAZ1B is the only bromodomain protein (out of 43 bromodomain proteins analyzed) exclusively required for proliferation of MSH2-knockout cells (Supplementary Fig. S9D and S9E; ref. 41). Like MSH2, BAZ1B is required for CLDN4 expression (Fig. 6H; Supplementary Fig. S9F). To evaluate whether BAZ1B mediates these transcriptional effects through MSH2, we performed ChIP-qPCR using MSH2 antibodies in BAZ1B knockout YCC21 cells and found that BAZ1B loss does not prevent MSH2 binding (Supplementary Fig. S9G and S9H). A possible explanation is that BAZ1B addiction arises upon MSH2 loss because both proteins serve a similar function in regulating the transcription circuitry of cell adhesion genes. To assess the therapeutic implications of this finding, we tested BAZ1B inhibition via the bromodomain inhibitor JQ1 in MSH2-knockout and wild-type cells, and found that MSH2 deficiency indeed renders MSH2-knockout cells susceptible to JQ1 treatment (Fig. 6I). To determine the efficacy of JQ1 in vivo, we treated mouse xenograft models of MSH2-deficient tumors with either JQ1 or vehicle (10 mice/group) and observed that JQ1 treatment significantly inhibited tumor growth (P = 0.0002, Fig. 6J).

Finally, genetic mutations in MSH2 and other MMR genes (MSH6, MLH1, and PMS2) are associated with MSI (42–44). We explored whether MSI gastric cancer cell lines also display sensitivity to bromodomain and extraterminal motif (BET) inhibition similar to MSH2 deficiency. We treated the MSI gastric cancer cell line NCC59 and non-MSI gastric cancer cell line YCC21 with JQ1. We found that NCC59 is highly susceptible (low IC50) to JQ1 compared with YCC21 (high IC50 value; Supplementary Fig. S9I). Notably, NCC59 cells display reduced CLDN4 and MSH2 expression compared with YCC21 (Supplementary Fig. S9I, inset). Previous studies have shown that MSH2-deficient or MLH1-deficient MSI cells may require activity of DNA polymerases POLB and POLG (45); however, we found that BET inhibition does not repress but rather increases the expression of POLB and POLG (Supplementary Fig. S9J). Together, these data suggest that MSH2 deficiency promotes the proliferation of tumor cells addicted to BAZ1B.

Here, we report a noncanonical role for the DNA repair protein MSH2 in regulating the expression of gastric cancer–associated cell-adhesion genes, by binding to cell adhesion gene loci and altering chromatin architecture. Loss of MSH2 in advanced and metastatic gastric cancers deregulates cell adhesion pathway, resulting in aggressive tumor cells that are addicted to BAZ1B, sensitizing such tumor cells to bromodomain/BET inhibition (Fig. 7). Although this report focuses on gastric cancer, preliminary analysis suggests that this sensitivity may apply to other MSI tumor types (Supplementary Fig. S10). Other non MMR-functions of MSH2 include estrogen receptor activation (46) or suppression of homologous recombination (47). Previous studies have also demonstrated vulnerabilities of MSH2/MMR-deficient cells against chemotherapy, methotrexate, or POLB, POLG, or PINK silencing (45, 48). The mechanisms underlying BAZ1B-addiction displayed in MSH2-deficient cells are unlikely mediated through inhibiting POLB/POLG.

Figure 7.

A schematic model for MSH2 loss–dependent reprogramming of cell adhesion and addiction to BAZ1B.

Figure 7.

A schematic model for MSH2 loss–dependent reprogramming of cell adhesion and addiction to BAZ1B.

Close modal

The present work extends our earlier study where analysis of sequence motifs at gastric cancer SEs identified master regulatory transcription factors such as CDX2 and HNF4α  (14). Similar approaches were used to identify SE-associated factors in other tumor types (49). This approach is limited to sequence-specific transcription factors, whereas the CRISPR/dCas9-ChIP-MS approach represents an unbiased approach to identifying novel factors associated with cis-regulatory elements, and has been used to identify regulators of β−globulin genes, erythroid enhancers (16), and Sox9 enhancers (50). However, a limitation of CRISPR/dCas9-ChIP-MS is its moderate specificity, thereby requiring data to be integrated against other functional genomic approaches, which in our case was a genome-wide screen to identify genes required for CLDN4 expression.

We found that MSH2 genomic occupancy is specifically associated with SEs linked to cell adhesion pathways. MSH2 occupied sites were enriched in both GATA and HNF4α-binding sites, defined either by sequence motif or experimentally determined ChIP sequencing. Our data also suggest that MSH2 binding exhibits cancer specificity as multiple gastric cancer lines showed overlapping binding peaks absent in normal epithelial gastric cells. At these cancer-specific cell adhesion gene loci, we found that MSH2 binding is sufficient to affect chromatin rewiring, altering enhancer–promoter interactions and gene expression. Chromatin rewiring is emerging as an important mechanism to drive and maintain cancer gene expression without overt genomic changes (51). Our data suggest a role for MSH2 in chromatin organization involving MSH2 enhancer and promoter binding to regulate transcription (52). Notably, two MSH2 mutants (R524P and G674S) deficient in MMR activity maintained their transcriptome regulatory function, suggesting that the two functions are separable. To confirm this, future studies are needed to map the specific MSH2 domains involved in transcription regulatory function, and to identify MSH2 mutants competent in MMR but unable to regulate transcription.

Our study found that BAZ1B is a synthetic lethal partner of MSH2 and an activator of CLDN4. BAZ1B, an atypical tyrosine-protein kinase, is known to play a role in chromatin remodeling, transcription, and topoisomerase I activity during DNA replication (40). Future studies are required to resolve the crystal structure of BAZ1B to identify inhibitors of BAZ1B, and to dissect the mechanism of BAZ1B addiction in MSH2-deficient cells. Our current data on BAZ1B-knockout cells suggest that BAZ1B and MSH2 act in parallel pathways, and could be further reinforced by additional studies such as JQ1 treatments in vitro and in vivo on MSH2-deficient gastric cancer cells reconstituted with DNA repair–defective MSH2 mutants capable of regulating transcription.

Our findings that MSH2 loss in late- and advanced-stage gastric cancers is associated with EMT and aggressive tumor behavior is reconcilable with the general perception that MSI cancers are associated with good prognosis. Because of tumor heterogeneity, not all MSI-cancers are MSH2-deficient, and not all MSH2-deficient gastric cancers are MSI. For example, in the TCGA dataset, only 9%–10% of MSH2-low tumors are MSI (“MSH2-low” being defined as the 10%–20% of tumors showing the lowest MSH2 expression), whereas 20%–25% of MSH6-low tumors are MSI, with an overlap of 15%–20%. Moreover, although MSI-positive cancers are generally associated with good clinical responses to immune checkpoint inhibitors (e.g., anti-PD1 antibodies; ref. 53), these same cancers are also typically non-responsive to chemotherapy thus representing a negative predictive factor for cytotoxic agents (44). Even for checkpoint inhibition, heterogeneity exists within the MSI subtype, as only 40% of MSI patients show long-term robust responses with the majority of MMR-deficient tumors being refractory to anti–PD-1 therapy (44, 54). In this respect, the potential for repurposing BET-inhibitors to treat MSH2-deficient gastric cancers via targeting BAZ1B may warrant further investigation, perhaps in combination with immunotherapy (26).

Taken collectively, our results highlight an unexpected role for the MSH2 in regulating gastric cancer enhancer biology. Although our data suggest a potential therapy for MSH2-deficient gastric cancers, these results are still preclinical and should be carefully interpreted given previous suggested alternatives for treating dMMR tumors, such as methotrexate or PARP inhibitors that induce oxidative DNA damage in the absence of MSH2, and were not validated in human trials ultimately (NCT00952016; ref. 55). It will be interesting to further investigate the role of DNA repair complexes in integrating genomic and epigenomic alterations to establish hallmarks of cancer.

A.M. Nargund reports a patent for the use of BET inhibitors in MSH2-deficient cancers pending. J.M.N. Teo reports a patent for the use of BET inhibitors in MSH2-deficient cancers pending to his institution. R. Sundar reports other support from Bristol Myers Squibb, Merck, Eisai, Bayer, Taiho, Novartis, MSD, Eli Lilly, Roche, Taiho, AstraZeneca, and DKSH, and grants from Paxman Coolers, MSD outside the submitted work. H. Ashktorab reports grants and other support from NCI during the conduct of the study and outside the submitted work. H.I. Grabsch reports personal fees from AstraZeneca and Bristol Myers Squibb outside the submitted work. S. Li reports grants from Duke-NUS Medical School during the conduct of the study, as well as grants from NMRC and MOE outside the submitted work and reports a patent for PCT/SG2021/050789 issued. P. Tan reports a patent for the use of BET inhibitors in MSH2-deficient cancers filed by his institution pending. No disclosures were reported by the other authors.

A.M. Nargund: Conceptualization, resources, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Xu: Software, formal analysis, validation, investigation. A. Mandoli: Software, formal analysis, validation, investigation. A. Okabe: Software, investigation. G.B. Chen: Software, investigation. K.K. Huang: Formal analysis. T. Sheng: Formal analysis, validation, investigation. X. Yao: Formal analysis, investigation. J.M.N. Teo: Validation, investigation. R. Sundar: Software, investigation. Y.J. Kok: Investigation. Y.X. See: Validation, investigation. M. Xing: Formal analysis, investigation. Z. Li: Formal analysis, investigation. C.H. Yong: Formal analysis, validation, investigation. A. Anand: Formal analysis, investigation. Z.F. Bin Adam Isa: Investigation. L.F. Poon: Investigation. M.S.W. Ng: Resources, investigation. J.Y.P. Koh: Investigation. W.F. Ooi: Resources, investigation. S.T. Tay: Resources, investigation. X. Ong: Investigation, methodology, project administration. A.L.K. Tan: Resources, methodology, project administration. D.T. Smoot: Resources. H. Ashktorab: Resources. H.I. Grabsch: Methodology, project administration. M.J. Fullwood: Methodology, project administration. B.T. Teh: Formal analysis, project administration. X. Bi: Conceptualization, methodology, project administration, writing–review and editing. A. Kaneda: Resources, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing. S. Li: Conceptualization, software, formal analysis, validation, methodology, project administration, writing–review and editing. P. Tan: Resources, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

This study was supported by National Medical Research Council grants NMRC/STaR/0026/2015, MOH-000967–00 and OFLCG18May-0003, and A*ccelerate GAP fund ETPL/15-R15 GAP-0021 (to P. Tan) and MOE tier 2 grant (MOE2017-T2–1-105) and NMRC CS-IRG grant (NMRC/CIRG/1481/2017 to S. Li) and SCISSOR (A*STAR IAF-PP) H18/01/a0/020. Funding was also provided by Cancer Science Institute of Singapore, NUS, under the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centers of Excellence initiative, and block funding from Duke-NUS Medical School. A.M. Nargund is also supported by St. Baldrick's Foundation Research Award. R. Sundar is supported by a National Medical Research Council (NMRC) Fellowship, Singapore. M.J. Fullwood is supported by the RNA Biology Center at the Cancer Science Institute of Singapore, NUS, as part of funding under the Singapore Ministry of Education Academic Research Fund Tier 3 awarded to Daniel Tenen (MOE2014-T3–1-006) and the National Research Foundation Singapore and the Singapore Ministry of Education under its Research Centers of Excellence initiative. The authors thank the CRISPR Core and Flow Cytometry Core at Duke-NUS, and the Genome Biology Facility at Duke-NUS.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Jia
Y
,
Chng
WJ
,
Zhou
J
.
Super-enhancers: critical roles and therapeutic targets in hematologic malignancies
.
J Hematol Oncol
2019
;
12
:
77
.
2.
Huang
J
,
Li
K
,
Cai
W
,
Liu
X
,
Zhang
Y
,
Orkin
SH
, et al
.
Dissecting super-enhancer hierarchy based on chromatin interactions
.
Nat Commun
2018
;
9
:
943
.
3.
Schoenfelder
S
,
Fraser
P
.
Long-range enhancer–promoter contacts in gene expression control
.
Nat Rev Genet
2019
;
20
:
437
55
.
4.
Tang
F
,
Yang
Z
,
Tan
Y
,
Li
Y
.
Super-enhancer function and its application in cancer targeted therapy
.
NPJ Precis Oncol
2020
;
4
:
2
.
5.
Ooi
WF
,
Nargund
AM
,
Lim
KJ
,
Zhang
S
,
Xing
M
,
Mandoli
A
, et al
.
Integrated paired-end enhancer profiling and whole-genome sequencing reveals recurrent CCNE1 and IGF2 enhancer hijacking in primary gastric adenocarcinoma
.
Gut
2020
;
69
:
1039
52
.
6.
Northcott
PA
,
Lee
C
,
Zichner
T
,
Stutz
AM
,
Erkek
S
,
Kawauchi
D
, et al
.
Enhancer hijacking activates GFI1 family oncogenes in medulloblastoma
.
Nature
2014
;
511
:
428
34
.
7.
Wang
X
,
Cairns
MJ
,
Yan
J
.
Super-enhancers in transcriptional regulation and genome organization
.
Nucleic Acids Res
2019
;
47
:
11481
96
.
8.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
.
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
9.
Collaborators
GBDSC
.
The global, regional, and national burden of stomach cancer in 195 countries, 1990–2017: a systematic analysis for the Global Burden of Disease study 2017
.
Lancet Gastroenterol Hepatol
2020
;
5
:
42
54
.
10.
Wang
K
,
Yuen
ST
,
Xu
J
,
Lee
SP
,
Yan
HH
,
Shi
ST
, et al
.
Whole-genome sequencing and comprehensive molecular profiling identify new driver mutations in gastric cancer
.
Nat Genet
2014
;
46
:
573
82
.
11.
Chia
NY
,
Deng
N
,
Das
K
,
Huang
D
,
Hu
L
,
Zhu
Y
, et al
.
Regulatory crosstalk between lineage-survival oncogenes KLF5, GATA4 and GATA6 cooperatively promotes gastric cancer development
.
Gut
2015
;
64
:
707
19
.
12.
Janjigian
YY
,
Sanchez-Vega
F
,
Jonsson
P
,
Chatila
WK
,
Hechtman
JF
,
Ku
GY
, et al
.
Genetic predictors of response to systemic therapy in esophagogastric cancer
.
Cancer Discov
2018
;
8
:
49
58
.
13.
Pectasides
E
,
Stachler
MD
,
Derks
S
,
Liu
Y
,
Maron
S
,
Islam
M
, et al
.
Genomic heterogeneity as a barrier to precision medicine in gastroesophageal adenocarcinoma
.
Cancer Discov
2018
;
8
:
37
48
.
14.
Ooi
WF
,
Xing
M
,
Xu
C
,
Yao
X
,
Ramlee
MK
,
Lim
MC
, et al
.
Epigenomic profiling of primary gastric adenocarcinoma reveals super-enhancer heterogeneity
.
Nat Commun
2016
;
7
:
12983
.
15.
Sakuma
T
,
Nishikawa
A
,
Kume
S
,
Chayama
K
,
Yamamoto
T
.
Multiplex genome engineering in human cells using all-in-one CRISPR/Cas9 vector system
.
Sci Rep
2014
;
4
:
5400
.
16.
Liu
X
,
Zhang
Y
,
Chen
Y
,
Li
M
,
Zhou
F
,
Li
K
, et al
.
In situ capture of chromatin interactions by biotinylated dCas9
.
Cell.
2017
;
170
:
1028
43
.
17.
Joung
J
,
Konermann
S
,
Gootenberg
JS
,
Abudayyeh
OO
,
Platt
RJ
,
Brigham
MD
, et al
.
Genome-scale CRISPR-Cas9 knockout and transcriptional activation screening
.
Nat Protoc
2017
;
12
:
828
63
.
18.
Li
W
,
Koster
J
,
Xu
H
,
Chen
CH
,
Xiao
T
,
Liu
JS
, et al
.
Quality control, modeling, and visualization of CRISPR screens with MAGeCK-VISPR
.
Genome Biol
2015
;
16
:
281
.
19.
Giannoulatou
E
,
Park
SH
,
Humphreys
DT
,
Ho
JW
.
Verification and validation of bioinformatics software without a gold standard: a case study of BWA and Bowtie
.
BMC Bioinf
2014
;
15
:
S15
.
20.
Mumbach
MR
,
Rubin
AJ
,
Flynn
RA
,
Dai
C
,
Khavari
PA
,
Greenleaf
WJ
, et al
.
HiChIP: efficient and sensitive analysis of protein-directed genome architecture
.
Nat Methods
2016
;
13
:
919
22
.
21.
McLean
CY
,
Bristor
D
,
Hiller
M
,
Clarke
SL
,
Schaar
BT
,
Lowe
CB
, et al
.
GREAT improves functional interpretation of cis-regulatory regions
.
Nat Biotechnol
2010
;
28
:
495
501
.
22.
Schnell
U
,
Cirulli
V
,
Giepmans
BN
.
EpCAM: structure and function in health and disease
.
Biochim Biophys Acta
2013
;
1828
:
1989
2001
.
23.
Tsukita
S
,
Tanaka
H
,
Tamura
A
.
The Claudins: from tight junctions to biological systems
.
Trends Biochem Sci
2019
;
44
:
141
52
.
24.
Neesse
A
,
Griesmann
H
,
Gress
TM
,
Michl
P
.
Claudin-4 as therapeutic target in cancer
.
Arch Biochem Biophys
2012
;
524
:
64
70
.
25.
Hingorani
MM
.
TIRF(ing) reveals Msh2-Msh6 surfing on DNA
.
Nat Struct Mol Biol
2007
;
14
:
1124
5
.
26.
Yong
WP
,
Rha
SY
,
Tan
IB
,
Choo
SP
,
Syn
NL
,
Koh
V
, et al
.
Real-time tumor gene expression profiling to direct gastric cancer chemotherapy: proof-of-concept "3G" trial
.
Clin Cancer Res.
2018
;
24
:
5272
81
.
27.
Mastrocola
AS
,
Heinen
CD
.
Lynch syndrome-associated mutations in MSH2 alter DNA repair and checkpoint response functions in vivo
.
Hum Mutat
2010
;
31
:
E1699
708
.
28.
Heinen
CD
,
Wilson
T
,
Mazurek
A
,
Berardini
M
,
Butz
C
,
Fishel
R
.
HNPCC mutations in hMSH2 result in reduced hMSH2-hMSH6 molecular switch functions
.
Cancer Cell
2002
;
1
:
469
78
.
29.
Kobayashi
K
,
Karran
P
,
Oda
S
,
Yanaga
K
.
Involvement of mismatch repair in transcription-coupled nucleotide excision repair
.
Hum Cell
2005
;
18
:
103
15
.
30.
Gasperini
M
,
Hill
AJ
,
McFaline-Figueroa
JL
,
Martin
B
,
Kim
S
,
Zhang
MD
, et al
.
A Genome-wide framework for mapping gene regulation via cellular genetic screens
.
Cell
2019
;
176
:
377
90
.
31.
Splinter
E
,
Heath
H
,
Kooren
J
,
Palstra
RJ
,
Klous
P
,
Grosveld
F
, et al
.
CTCF mediates long-range chromatin looping and local histone modification in the beta-globin locus
.
Genes Dev
2006
;
20
:
2349
54
.
32.
Hura
GL
,
Tsai
CL
,
Claridge
SA
,
Mendillo
ML
,
Smith
JM
,
Williams
GJ
, et al
.
DNA conformations in mismatch repair probed in solution by X-ray scattering from gold nanocrystals
.
Proc Natl Acad Sci U S A
2013
;
110
:
17308
13
.
33.
Shen
J
,
Ju
Z
,
Zhao
W
,
Wang
L
,
Peng
Y
,
Ge
Z
, et al
.
ARID1A deficiency promotes mutability and potentiates therapeutic antitumor immunity unleashed by immune checkpoint blockade
.
Nat Med
2018
;
24
:
556
62
.
34.
Davis
CA
,
Hitz
BC
,
Sloan
CA
,
Chan
ET
,
Davidson
JM
,
Gabdank
I
, et al
.
The Encyclopedia of DNA elements (ENCODE): data portal update
.
Nucleic Acids Res
2018
;
46
:
D794
801
.
35.
Li
F
,
Mao
G
,
Tong
D
,
Huang
J
,
Gu
L
,
Yang
W
, et al
.
The histone mark H3K36me3 regulates human DNA mismatch repair through its interaction with MutSalpha
.
Cell
2013
;
153
:
590
600
.
36.
Lee
J
,
Kim
H
,
Lee
JE
,
Shin
SJ
,
Oh
S
,
Kwon
G
, et al
.
Selective cytotoxicity of the NAMPT inhibitor FK866 toward gastric cancer cells with markers of the Epithelial-mesenchymal transition, due to loss of NAPRT
.
Gastroenterology
2018
;
155
:
799
814
.
37.
Jolly MK,
Ware KE,
Xu S,
Gilja S,
Shetler S,
Yang
Y
, et al
.
E-cadherin represses anchorage-independent growth in sarcomas through both signaling and mechanical mechanisms
. Mol Cancer Res 2019;17:1391–1402.
38.
Gava
F
,
Rigal
L
,
Mondesert
O
,
Pesce
E
,
Ducommun
B
,
Lobjois
V
.
Gap junctions contribute to anchorage-independent clustering of breast cancer cells
.
BMC Cancer
2018
;
18
:
221
.
39.
Janiszewska
M
,
Primi
MC
,
Izard
T
.
Cell adhesion in cancer: beyond the migration of single cells
.
J Biol Chem
2020
;
295
:
2495
505
.
40.
Barnett
C
,
Krebs
JE
.
WSTF does it all: a multifunctional protein in transcription, repair, and replication
.
Biochem Cell Biol
2011
;
89
:
12
23
.
41.
Muller
S
,
Filippakopoulos
P
,
Knapp
S
.
Bromodomains as therapeutic targets
.
Expert Rev Mol Med
2011
;
13
:
e29
.
42.
de la Chapelle
A
,
Hampel
H
.
Clinical relevance of microsatellite instability in colorectal cancer
.
J Clin Oncol
2010
;
28
:
3380
7
.
43.
Bonneville
R
,
Krook
MA
,
Kautto
EA
,
Miya
J
,
Wing
MR
,
Chen
HZ
, et al
.
Landscape of microsatellite instability across 39 cancer types
.
JCO Precis Oncol
2017 Oct 3
. [
Epub ahead of print
].
44.
McGrail
DJ
,
Garnett
J
,
Yin
J
,
Dai
H
,
Shih
DJH
,
Lam
TNA
, et al
.
Proteome instability is a therapeutic vulnerability in mismatch repair-deficient cancer
.
Cancer Cell
2020
;
37
:
371
86
.
45.
Martin
SA
,
McCabe
N
,
Mullarkey
M
,
Cummins
R
,
Burgess
DJ
,
Nakabeppu
Y
, et al
.
DNA polymerases as potential therapeutic targets for cancers deficient in the DNA mismatch repair proteins MSH2 or MLH1
.
Cancer Cell
2010
;
17
:
235
48
.
46.
Wada-Hiraike
O
,
Yano
T
,
Nei
T
,
Matsumoto
Y
,
Nagasaka
K
,
Takizawa
S
, et al
.
The DNA mismatch repair gene hMSH2 is a potent coactivator of oestrogen receptor alpha
.
Br J Cancer
2005
;
92
:
2286
91
.
47.
Smith
JA
,
Bannister
LA
,
Bhattacharjee
V
,
Wang
Y
,
Waldman
BC
,
Waldman
AS
.
Accurate homologous recombination is a prominent double-strand break repair pathway in mammalian chromosomes and is modulated by mismatch repair protein Msh2
.
Mol Cell Biol
2007
;
27
:
7816
27
.
48.
Martin
SA
,
Hewish
M
,
Sims
D
,
Lord
CJ
,
Ashworth
A
.
Parallel high-throughput RNA interference screens identify PINK1 as a potential therapeutic target for the treatment of DNA mismatch repair-deficient cancers
.
Cancer Res
2011
;
71
:
1836
48
.
49.
Suva
ML
,
Rheinbay
E
,
Gillespie
SM
,
Patel
AP
,
Wakimoto
H
,
Rabkin
SD
, et al
.
Reconstructing and reprogramming the tumor-propagating potential of glioblastoma stem-like cells
.
Cell
2014
;
157
:
580
94
.
50.
Mochizuki
Y
,
Chiba
T
,
Kataoka
K
,
Yamashita
S
,
Sato
T
,
Kato
T
, et al
.
Combinatorial CRISPR/Cas9 approach to elucidate a far-upstream enhancer complex for tissue-specific Sox9 expression
.
Dev Cell
2018
;
46
:
794
806
.
51.
Okabe
A
,
Huang
KK
,
Matsusaka
K
,
Fukuyo
M
,
Xing
M
,
Ong
X
, et al
.
Cross-species chromatin interactions drive transcriptional rewiring in Epstein-Barr virus-positive gastric adenocarcinoma
.
Nat Genet
2020
;
52
:
919
30
.
52.
Dixon
JR
,
Selvaraj
S
,
Yue
F
,
Kim
A
,
Li
Y
,
Shen
Y
, et al
.
Topological domains in mammalian genomes identified by analysis of chromatin interactions
.
Nature
2012
;
485
:
376
80
.
53.
Le
DT
,
Durham
JN
,
Smith
KN
,
Wang
H
,
Bartlett
BR
,
Aulakh
LK
, et al
.
Mismatch repair deficiency predicts response of solid tumors to PD-1 blockade
.
Science
2017
;
357
:
409
13
.
54.
Mandal
R
,
Samstein
RM
,
Lee
KW
,
Havel
JJ
,
Wang
H
,
Krishna
C
, et al
.
Genetic diversity of tumors with mismatch repair deficiency influences anti–PD-1 immunotherapy response
.
Science
2019
;
364
:
485
91
.
55.
Goyal
G
,
Fan
T
,
Silberstein
PT
.
Hereditary cancer syndromes: utilizing DNA repair deficiency as therapeutic target
.
Fam Cancer
2016
;
15
:
359
66
.