Histone methylation antagonism drives tumor immune evasion in squamous cell carcinomas

Abstract

How cancer-associated chromatin abnormalities shape tumor-immune interaction remains incompletely understood. Recent studies have linked DNA hypomethylation and de-repression of retrotransposons to anti-tumor immunity through the induction of interferon response. Here, we report that inactivation of the histone H3K36 methyltransferase NSD1, which is frequently found in squamous cell carcinomas (SCCs) and induces DNA hypomethylation, unexpectedly results in diminished tumor immune infiltration. In syngeneic and genetically engineered mouse models of head and neck SCCs, NSD1-deficient tumors exhibit immune exclusion and reduced interferon response despite high retrotransposon expression. Mechanistically, NSD1 loss results in silencing of innate immunity genes, including the type III interferon receptor IFNLR1, through depletion of H3K36 di-methylation (H3K36me2) and gain of H3K27 tri-methylation (H3K27me3). Inhibition of EZH2 restores immune infiltration and impairs the growth of Nsd1-mutant tumors. Thus, our work uncovers a druggable chromatin cross talk that regulates the viral mimicry response and enables immune evasion of DNA hypomethylated tumors.

Keywords: DNA methylation; EZH2; NSD1; Tazemetostat; epigenetics; head and neck cancers; histone methylation; immune evasion; squamous cell carcinomas.

Figure 1:. NSD1 loss reduces tumor immune infiltration in syngeneic HNSCC mouse model.

A, Schematic illustrating the implantation of Nsd1 wildtype (WT) or knockout (KO) MOC1 cells into C57BL/6 mice and tumor immune microenvironment analysis. B, Representative Hematoxylin and eosin (H&E) staining and immunofluorescent staining of keratin 5 (KRT5), NSD1 and H3K36me2 in Nsd1 WT or KO MOC1 tumors. C, Percentage of intra-tumoral T cells, CD8+ T cells, natural killer (NK) cells and macrophages determined by multi-channel flow cytometry from subcutaneously injected Nsd1 WT or KO MOC1 tumors (n = 6–10). Scale bar, 50 μm. D-F, Representative immunofluorescent staining of KRT5 and CD8+ T cells (D), CD11b+ macrophages (E) and NK1.1+ natural killer (NK) cells (F) in Nsd1 WT or KO MOC1 tumors. Bar graphs below show the average % immune infiltration into the Krt5+ tumor compartment from at least three regions of interest (ROI) per mouse, quantified for each immune cell type (4–6 mice per group; two-sided Student’s test; data are represented as mean ± SD). Scale bar, 50 μm. G-H, Volumes of Nsd1 wildtype (WT) (G) or knockout (KO) MOC1 (H) tumors in NSG mice or C57BL/6 mice. Data represent mean ± SEM. n=8 per group. I, Comparison of wildtype and Nsd1 knockout (KO) MOC1 tumor volume in C57BL/6 mice and NSG mice in (G) and (H) at day 25 after injection. *, p<0.05, **, p<0.01, ***, p<0.001, ****, p<0.0001. See also Figure S1.

Figure 2:. Ablation of Nsd1 induces immune-cold phenotype in a model of carcinogen-induced oral SCC.

A, Left, schematic illustrating the generation of Nsd1 conditional knockout mice and development of primary oral lesions induced by 4NQO. Right: H&E staining and immunofluorescent images of tongue lesions at the 26-week time point from control (top) and Nsd1-KO mice (bottom). B-D, Representative immunofluorescent staining of keratin 5 (KRT5) and CD8+ T cells (B), CD11b+ macrophages (C) and NK1.1+ natural killer (NK) cells (D) in carcinogen-induced tongue lesions from Nsd1 wildtype (WT) mice (top) and Nsd1 knockout (KO) mice (bottom). Bar graphs below show the average % immune infiltration into the Krt5+ tumor compartment from at least three regions of interest (ROI) per lesion, quantified for each immune cell type (5–6 mice per group; two-sided Student’s test; data are represented as mean ± SD). Scale bar, 100 μm. E, Representative immunofluorescent staining of Krt5, CD8+ T cells and CD161+ NK cells in primary human HNSCC samples highlighting reduced abundance of immune cells (yellow arrowheads) in NSD1 mutant (MT) compared to wildtype (WT) tumors. Scale bar, 50 μm. F, Quantification of at least three regions of interest (ROI) per human sample reveals reduced CD8+ T cell and NK cell infiltration into human NSD1 mutant (Mut) tumors (Krt5+ compartment) (n=10 for NSD1 WT patient samples, n=9 for NSD1 Mut patient samples; two-sided Student’s test; data are represented as mean ± SD). Patient characteristics are listed in Table S1. *, p<0.05; **, p<0.01, ***, p<0.001, ****, p<0.0001. See also Figure S2 and Table S1.

Figure 3:. NSD1 loss decreases the expression of interferon-stimulated genes.

A, Gene ontology (GO) analysis of differentially expressed genes between NSD1 wildtype and knockout Cal27 cells. The top statistically significantly enriched GO groups among downregulated genes in NSD1 knockout cells are highlighted in blue. B, Gene set enrichment analysis (GSEA) showing interferon-stimulated genes (ISGs) are significantly enriched among downregulated genes in NSD1 knockout (KO) Cal27 cells (top) or NSD1 mutant (Mut) SKN-3 and BICR78 cells (bottom) versus wildtype (WT) Cal27 cells. C, qRT-PCR analysis of STAT1 and representative ISGs in NSD1 wildtype (Cal27 and PE/CA-PJ15) or mutant (BICR78 and SKN-3) HNSCC cells. Data represent mean ± SD, n=3. D-F, qRT-PCR analysis of STAT1 and representative ISGs comparing NSD1 knockout (KO) to wildtype (WT) Cal27 (D), Fadu (E) or Detroit 562 (F) HNSCC cells. Data represent mean ± SD, n=3. G, qRT-PCR analysis of representative ISGs in NSD1 knockout (KO) or wildtype (WT) Cal27 cells with or without 10 μg/mL Lipopolysaccharide (LPS) treatment for 24 hrs. Data represent mean ± SD, n=3. H, qRT-PCR analysis of representative ISGs in NSD1 knockout (KO) or wildtype (WT) Cal27 cells with or without transfecting with 1 μg/mL herring testis DNA (htDNA) for 5 hrs. Data represent mean ± SD, n=3. I, Representative immunofluorescent staining of KRT5 and IRF7 in carcinogen-induced tongue lesions from Nsd1 wildtype (WT) mice (top) and Nsd1 knockout (KO) mice (bottom). Scale bar, 50 μm. J, Representative immunofluorescent staining of KRT5 and IRF7 or MX1 in NSD1 wildtype (WT) or NSD1 mutant (MT) human HNSCC patient samples. *, p<0.05; **, p<0.01, ***, p<0.001, ****, p<0.0001. Scale bar, 50 μm. See also Figure S3.

Figure 4:. NSD1 loss impairs interferon signaling response.

A, Schematic view of dsRNA production, sensing and the downstream type I & III interferon response pathways. B, NSD1 wildtype (WT) or knockout (KO) Cal27 cells were stained with J2 antibody (anti-dsRNA) and subjected to flow cytometry quantification (left) or immunofluorescence staining (right). Data represent mean ± SD of fold increase of mean fluorescence intensity (MFI) in NSD1 knockout (KO) Cal27 cells versus in NSD1 wildtype (WT) Cal27 cells, n=4. C, Western blots showing expression of interferon pathway proteins in NSD1 wildtype (Cal27, PE/CA-PJ15) or mutant (BICR78, SKN-3) HNSCC cells. D, Western blots showing expression of interferon pathway proteins in NSD1 wildtype (WT) or knockout (KO) Cal27 cells. E, Western blots showing STAT1 expression and phosphorylation in NSD1 wildtype (WT), NSD1 knockout (KO) Cal27 cells and KO cells with ectopic expression of truncated wildtype or catalytic-dead R2017Q mutant NSD1 (NSD1RQ). F, qRT-PCR analysis of STAT1 and representative ISGs in cell lines listed in (E). Data represent mean ± SD, n=3. G, Western blots showing increased STAT1 expression and phosphorylation by CRISPR activation in Cal27 NSD1 knockout (KO) cells. H, qRT-PCR analysis of representative ISGs in Cal27 wildtype (WT) and NSD1 knockout (KO) cells with or without STAT1 activation. Data represent mean ± SD, n=3. *, p<0.05; **, p<0.01, ***, p<0.001, ns, not significant. See also Figure S4.

Figure 5:. Epigenetic silencing of IFNLR1 contributes to NSD1 loss-induced immune evasion.

A, Violin plots showing the global levels of H3K36me2 and H3K27me3 (counts per million of CUT&Tag reads) in NSD1 wildtype (WT) or knockout (KO) Cal27 cells. Bin size = 10 Mb. The center line in the embedded boxplots represents the median, the box limits are the 25th and 75th percentiles, and the whiskers are the minimum to maximum values. P-values were determined by two-tailed Student’s t-test. B, Gene ontology (GO) analysis of genes that gained H3K27me3 in NSD1 knockout (KO) Cal27 cells. The top statistically significantly enriched GO groups are highlighted in red. C, IGV snapshot showing the chromatin landscape of H3K36me2, H3K27me3, and gene transcription at the IFNLR1 locus in wildtype or NSD1 knockout (KO) Cal27 cells with or without the treatment of 3 μM EPZ-6438 for 10 days. D, Flow cytometry analysis of cell surface expression of IFNLR1 in NSD1 wildtype (WT) or knockout (KO) Cal27 cells. The percentage of IFNLR1-positive cells is shown as bar graph. Data represent mean ± SD, n=3. E, Representative immunofluorescent staining of KRT5 and IFNLR1 in carcinogen-induced tongue lesions from Nsd1 wildtype (WT) mice (top) and Nsd1 knockout (KO) mice (bottom). Scale bar, 50 μm. F, Western blots showing STAT1 expression and phosphorylation in NSD1 wildtype (WT) or knockout (KO) Cal27 cells, or Cal27 cells transfected with IFNLR1 siRNA. G, qRT-PCR analysis of representative ISGs in Cal27 cells transfected with control or IFNLR1 siRNA. Data represent mean ± SD, n=3. H, Representative immunofluorescent staining of KRT5 and IFNLR1 in Nsd1 wildtype (WT), Nsd1 knockout (KO) or Ifnlr1 knockout (KO) MOC1 cells. Scale bar, 50 μm. I, Representative immunofluorescent staining of KRT5 and CD8+ T cells in Ifnlr1 wildtype (WT) or knockout (KO) MOC1 tumors. Bar graph shows the average % immune infiltration into the Krt5+ tumor compartment from at least three regions of interest (ROI) per mouse. Data represent mean ± SD, n=5. Scale bar, 50 μm. *, p<0.05; **, p<0.01, ***, p<0.001, ****, p<0.0001. See also Figure S5 and Table S2 and S3.

Figure 6:. Pharmacologic inhibition of EZH2 but not DNMT1 restores interferon response in NSD1-deficient cells.

A, Gene ontology (GO) analysis of differentially expressed genes between DMSO and EPZ-6438-treated (3 μM for 10 days) NSD1 knockout (KO) Cal27 cells. The top statistically significantly enriched GO groups among upregulated genes after EPZ-6438 treatment are highlighted in red. B, Violin plots showing the normalized ISG expression (Z score) of NSD1 wildtype (WT) and knockout (KO) Cal27 cells with or without the treatment of 3 μM EPZ-6438 for 10 days. The center line in the embedded boxplots represents the median, the box limits are the 25th and 75th percentiles, and the whiskers are the minimum to maximum values. P-values were determined by two-tailed Student’s t-test. C, Western blots showing expression of interferon pathway proteins in NSD1 wildtype (WT) and knockout (KO) Cal27 cells with or without the treatment of 3 μM EPZ-6438 for 10 days. D, qRT-PCR analysis of representative ISGs in NSD1 wildtype (WT) or knockout (KO) Cal27 cells with or without the treatment of 3 μM EPZ-6438 for 10 days. Data represent mean ± SD, n=3. E, qRT-PCR analysis of STAT1 and representative ISGs in NSD1 knockout (KO) Cal27 cells transfected with control or IFNLR1 siRNA and with or without treatment of EPZ-6438. Data represent mean ± SD, n=3. F, Representative immunofluorescent staining of KRT5 and IRF7 in Nsd1 wildtype (WT), left, or Nsd1 knockout (KO), right, MOC1 tumors with or without EPZ-6438 treatment. Scale bar, 50 μm. *, p<0.05; **, p<0.01, ***, p<0.001, ns, not significant. See also Figure S6.

Figure 7:. EZH2 inhibitor elicits immune infiltration and inhibits the growth of NSD1-deficient SCC tumors.

A, Volumes of Nsd1 wildtype (WT) or knockout (KO) MOC1 tumors in C57BL/6 mice treated with vehicle or EPZ-6438 at indicated dosage. Data represent mean ± SEM. n=6–8 per group. B, Percentage survival of C57BL/6 mice engrafted with Nsd1 knockout (KO) MOC1 cells treated with vehicle or 200 mg/kg EPZ-6438. Log-rank (Mantel-Cox) test was used to determine significance, n=9–10 per group. C, Volumes of Nsd1 knockout (KO) MOC1 tumors in Foxn1^nu immunodeficient mice treated with vehicle or EPZ-6438 at indicated dosage. Data represent mean ± SEM. n=6–8 per group. D-F, Representative immunofluorescent staining of Krt5 and CD8+ T cells (D), CD11b+ macrophages (E) and NK1.1+ natural killer (NK) cells (F) in Nsd1 knockout (KO) MOC1 tumors treated with vehicle or EPZ-6438. Bar graphs below show the average % immune infiltration into the Krt5+ tumor compartment from at least three regions of interest (ROI) per mouse, quantified for each immune cell type (3–7 mice per group; two-sided Student’s test; data are represented as mean ± SD). Scale bar, 50 μm. *, p<0.05; **, p<0.01, ***, p<0.001, ****, p<0.0001, ns, not significant. See also Figure S7.