EZH2 loss promotes gastric squamous cell carcinoma

Abstract

Gastric Squamous Cell Carcinoma (GSCC) is a rare but aggressive subtype of gastric cancer with unique histopathology, whose etiology remains poorly understood. Here, we perform genomics analyses of twenty GSCC samples and find that epigenetic regulation genes are among the most frequently mutated genes, including Enhancer of zeste homolog 2 (EZH2). Ezh2 loss induces squamous feature both in gastric organoids in vitro and in vivo mouse model. Ezh2 deficiency, together with Trp53 and Pten loss, both of which are also frequently mutated in GSCC, give rise to full-blown GSCC in mice. Mechanistically, we find that Ezh2 could repress the expression of Transcription factor AP-2 gamma (Tfap2c), a transcription factor with the ability to initiate epidermal squamous differentiation, through H3K27 methylation. Disruption of Tfap2c reduces the squamous characteristics of the Ezh2 loss-driven GSCC and reverses its resistance to chemo treatment. Our findings elucidate key molecular mechanisms underlying GSCC pathogenesis and identify potential therapeutic targets for this aggressive malignancy.

Fig. 1. EZH2 deficiency is associated with GSCC.

a Curated list of frequently altered genes in 20 GSCC patients. b Protein domain structure of EZH2 with mutations identified in GSCC patients. c Summary of mutations, deletions, and methylation levels of EZH2 in 20 GSCC patients. d Boxplot showing the expression levels of EZH2 in squamous carcinoma gene signature high (n = 58) and low (n = 313) samples in the TCGA-STAD cohort. Two-tailed Wilcoxon rank-sum test. The boxes indicate the median value, interquartile range, with whiskers extending from the box boundaries to upper/lower quartile ± 1.5 interquartile range. The p-value was determined by the two-tailed unpaired Wilcoxon rank-sum test. e GSEA showed that the GO_KERATINIZATION gene sets were significant enriched in GC patients with high expression of squamous carcinoma gene signatures (n = 58) compared to low expression ones (n = 313) in the TCGA-STAD cohort. Gene set enrichment significance was assessed by the unpaired two-tailed permutation test. f Survival curves of GC patients stratified by high and low expression of squamous carcinoma gene signature in the TCGA-STAD cohort. The cut-off values were determined by maximally selected rank statistics. The p-value was determined by the log-rank test. g The scatter plot showing the negative correlation between expression levels of squamous signature genes and EZH2 in the TCGA-STAD cohort. The p-value was determined by the two-tailed ttest. r, Pearson’s correlation coefficient. The shaded band around the regression line indicates the 95% confidence interval (CI) of the linear fit. h Representative IF staining of EZH2 (red) and P40 (green) in GAC tumor tissues (left) and GSCC tumor tissues (right). Scale bar, 100 µm. Source data are provided as a Source Data file.

Fig. 2. Ezh2 disruption promotes squamous features in vivo and in vitro.

a Western blot analysis of EZH2 in sgScr (Cas9-sgScramble), sgEzh2-1(Cas9-sgEzh2-1), and sgEzh2-2(Cas9-sgEzh2-2) gastric organoids (−1 and −2 represent two independent sgRNAs targeting Ezh2). Representative blot (a) of n  =  2 technical replicates. b Western blot analysis of H3K27me3 in sgScr, sgEzh2-1, and sgEzh2-2 gastric organoids. Representative blot (b) of n  =  2 technical replicates. c Representative bright-field(top), H&E(middle), CK14 and P40 IF staining(bottom) images of sgScr, sgEzh2-1, and sgEzh2-2 mouse gastric organoids. Scale bars, 50 µm (top) and 20 µm (middle and bottom). d The heatmap showing the differential expression genes (p < 0.05 and absolute log₂ fold-change > 0.5, p-value was determined by unpaired two-tailed Wald test) between sgScr, sgEzh2-1 and sgEzh2-2 gastric organoids, measured by RNA-seq analyses. e Statistical graphs showing the diameter of TP (Trp53^-/-; sgPten) and TPE (Trp53^-/-; sgPten; sgEzh2) gastric organoids. f Statistical graphs showing the organoids formation rate of TP and TPE gastric organoids (n = 3 biological replicates). g The bright-field image of subcutaneous tumors of TP (n = 3 mice) and TPE (n = 3 mice), Scale bar, 1 cm. h Tumor weight of subcutaneously transplanted TP (n = 3 mice) and TPE (n = 3 mice). i H&E staining of subcutaneous tumor tissues of TP and TPE mice. Scale bar, 20 µm. j Representative images of CK14(left), CK5(middle) and P40(right) IF staining of subcutaneous tumor tissues of TP and TPE mice. Scale bars, 20 µm and 50 µm. Data are shown as means ± SD, p-value was determined by unpaired two-tailed t test(e, f, h). Source data are provided as a Source Data file.

Fig. 3. The primary and orthotopic TPE mouse model recapitulates the molecular and clinical characteristic of human diseases.

a Survival curve of mice orthotopically transplanted with TP, TPE-1, and TPE-2 organoids (n = 6 mice). All curves were analyzed by log-rank (Mentel-Cox) test (− 1 and − 2 represent two independent sgRNAs targeting Ezh2). b Representative bright-field (top) and red fluorescence(bottom) images of gastric cancer transduced with TP, TPE-1 and TPE-2 organoids. Scale bar, 1 mm. c Statistical graphs showing the tumor weight of mice orthotopically transplanted with TP, TPE-1 and TPE-2 organoids (n = 6 mice). d Representative IHC staining of EZH2 in gastric tumor sections of TP (left), TPE-1 (middle), and TPE-2(right) mice. Scale bar, 20 µm. e Western blot analysis of EZH2 in TP, TPE-1 and TPE-2 mouse gastric tumor organoids. Representative blot (e) of n  =  2 biological replicates. f Representative IHC staining of H3K27me3 in gastric cancer sections of TP (left), TPE-1 (middle), and TPE-2(right) mice. Scale bar, 20 µm. g Western blot analysis of H3K27me3 in TP, TPE-1, and TPE-2 gastric tumor organoids. Representative blot (g) of n  =  2 biological replicates. h Representative H&E, P40, CK14, CK5/6, and CK5 IHC staining of TP (top), TPE-1 (middle), and TPE-2(bottom) tumor section from mice. Scale bar, 50 µm. i Statistical graphs of keratin pearl area percentages in tumor tissues from TP, TPE-1, and TPE-2 mice (n = 4 mice). j Representative bright-field (left) and red-fluorescence(right) images of liver from TP (top) and TPE (bottom) tumor-bearing mice. Scale bars, 2 mm(top) and 1 mm(bottom). k H&E staining of liver sections from TP (top) and TPE (bottom) tumor-bearing mice. Scale bar, 50 µm. l Representative IHC staining of CAS9, CK14, CK5/6 and P40 in liver sections from TPE tumor-bearing mice. Scale bar, 50 µm. Data are shown as means ± SD, p-value was determined by unpaired two-tailed t test(c, i). The samples derive from the same experiment, and that gels were processed in parallel for quantitative comparisons (e.g.,). Source data are provided as a Source Data file.

Fig. 4. Ezh2 represses the expressions of squamous signature genes through H3K27me3 activation.

a The heatmap showing the differential expression genes (p < 0.05 and absolute log₂ fold-change > 0.5, p-value was determined by the unpaired two-tailed Wald test) between TP (n = 2 mice) and TPE (n = 2 mice) tumor organoids. b GSEA showing that the squamous signature genes were significantly positively enriched in the upregulated genes of TPE organoids compared to TP ones. Gene set enrichment significance was assessed by the unpaired two-tailed permutation test. c Box plots illustrating the gastric squamous signature scores in TP (n = 2 mice) and TPE (n = 2 mice) tumor organoids, measured by RNA-seq analyses. The boxes indicate the median value, interquartile range, with whiskers extending from the box boundaries to upper/lower quartile ± 1.5 interquartile range. p-value was determined by the two-sided Likelihood ratio test. d Pie chart depicting the distribution of EZH2 binding peaks across annotated genomic regions in TP gastric organoids. e Average intensity curves (top) and tornado plots (bottom) showing the binding intensity of EZH2 (left) and H3K27me3modification levels in these corresponding regions between TP and TPE organoids (right). Scale bars denote BPM (bins per million mapped reads) for CUT&Tag signal. f Pie chart (left) showing the proportion of genes with reduced H3K27me3 modification levels in TPE gastric organoids compared to TP ones, categorized by the presence or absence of EZH2 binding. Pie chart (right) displaying the proportion of transcriptional changes in genes with both EZH2 binding and reduced H3K27me3 levels in TPE gastric organoids compared to TP ones. g Gene ontology enrichment results of overlapping genes associated with EZH2 binding, significantly reduced H3K27me3 modification levels, and up-regulated expression in TPE gastric organoids compared to the TP group. p-values were determined by the unpaired one-tailed hypergeometric test. h The network showing protein-protein interactions among EZH2-bound squamous signature genes with both reduced H3K27me3 and up-regulated expression levels in TPE vs. TP organoids, analyzed using the STRING database. Source data are provided as a Source Data file.

Fig. 5. TFAP2C is upregulated in GSCC and disruption of TFAP2C inhibits the squamous features of TPE GSCC tumors.

a The scatter plot showing the integrated ranking of the EZH2 regulated genes, The integrated score combines EZH2 binding intensity, H3K27me3 modification levels, and RNA expression abundance. b Integrative Genomics Viewer showing the EZH2 binding and H3K27me3 modification levels at the Tfap2c locus in TP and TPE gastric organoids. c Box plot showing the relative expression levels of Tfap2c in TP (n = 3 mice) and TPE (n = 3 mice). d Western blot analysis of TFAP2C in TP, TPE-1, and TPE-2 mice gastric tumor organoids. Representative blot (d) of n  =  2 biological replicates. The samples derive from the same experiment and that gels were processed in parallel for quantitative comparisons. e IF staining of P40 (red), TFAP2C (green), and EZH2(white) in TPE tumors. Scale bar, 50 µm. f mIHC staining of P40 (red), TFAP2C (green), and EZH2 (white) in GSCC16 tumor tissue and normal tissue. Scale bars, 200 µm and 20 µm. g Statistical graphs showing the P40 and TFAP2C fluorescence intensity of TPE tumor organoids with sgScr, sgTfap2c-1 and sgTfap2c-2 (n = 4 biological replicates) (− 1 and − 2 represent two sgRNAs targeting Tfap2c). h Representative images of H&E and TFAP2C, MUC1 and MUC5AC IHC staining of TPE tumors with sgScr, sgTfap2c-1 and sgTfap2c-2. Scale bar, 20 µm. i Statistical graphs showing the percentages of keratin pearl area of tumors in TPE tumor organoids with sgScr, sgTfap2c-1 and sgTfap2c-2 (n = 4 mice). j Representative IF staining of TFAP2C (green) and P40 (red) in TPE-sgScr and TPE-sgTfap2c tumor tissues. Scale bar, 50 µm. Data are shown as means ± SD, p-values were determined by unpaired two-tailed t test(c, g, i). Source data are provided as a Source Data file.

Fig. 6. Tfap2c mediates the functions of Ezh2 in GSCC and suggests a potential therapeutic target.

a Pie chart showing the distribution of TFAP2C binding peaks in annotated regions of the genome in TPE gastric organoids. b Integrative Genomics Viewer showing the TFAP2C binding levels at ΔNp63, Krt5, and Krt14 gene loci in TPE gastric organoids. c The heatmap showing the differential expression genes (p < 0.05 and absolute log₂ fold-change > 0.5, p-value was determined by the unpaired two-tailed Wald test) between TPE-sgScr, TPE-sgTfap2c−1, and TPE-sgTfap2c−2 gastric tumor organoids. d GSEA showing the negative enrichment of the DARWICHE_SQUAMOUS_CELL_CARCINOMA_UP gene set in TPE-sgTfap2c−1, TPE-sgTfap2c−2 compared to TPE- sgScr gastric tumor organoids. Gene set enrichment significance was assessed by the unpaired two-tailed permutation test. e Pie chart showing the proportion of genes with reduced expression levels in TPE-sgTfap2c gastric tumor organoids compared to TPE-sgScr ones, categorized by the presence or absence of TFAP2C binding. f GO analysis of overlapping genes in Fig. 6e. P-values were determined by the unpaired one-tailed hypergeometric test. g The box plots showing the gastric squamous signature scores in TPE-sgScr, TPE-sgTfap2c-1 and TPE-sgTfap2c-2 tumor organoids (n = 3 biological replicates). The boxes indicate the median value, interquartile range, with whiskers extending from the box boundaries to upper/lower quartile ± 1.5 interquartile range. P-values were determined by the two-sided Likelihood ratio test. h The Venn diagram showing the overlapping genes between the top 1000 upregulated genes in TPE tumor organoids, and the 1000 downregulated genes in TPE-sgTfap2c, relative to TP and TPE-sgScr groups, respectively. P-value was determined by a hypergeometric distribution test. i Gene ontology enrichment analysis of overlapping genes in Fig. 6h. P-values were determined by the unpaired one-tailed hypergeometric test. Source data are provided as a Source Data file.